diff options
Diffstat (limited to 'servers/rendering/renderer_rd')
154 files changed, 73312 insertions, 0 deletions
diff --git a/servers/rendering/renderer_rd/SCsub b/servers/rendering/renderer_rd/SCsub new file mode 100644 index 0000000000..10b83dca11 --- /dev/null +++ b/servers/rendering/renderer_rd/SCsub @@ -0,0 +1,12 @@ +#!/usr/bin/env python + +Import("env") + +env.add_source_files(env.servers_sources, "*.cpp") + +SConscript("effects/SCsub") +SConscript("environment/SCsub") +SConscript("forward_clustered/SCsub") +SConscript("forward_mobile/SCsub") +SConscript("shaders/SCsub") +SConscript("storage_rd/SCsub") diff --git a/servers/rendering/renderer_rd/cluster_builder_rd.cpp b/servers/rendering/renderer_rd/cluster_builder_rd.cpp new file mode 100644 index 0000000000..1bb45cbcc1 --- /dev/null +++ b/servers/rendering/renderer_rd/cluster_builder_rd.cpp @@ -0,0 +1,547 @@ +/*************************************************************************/ +/* cluster_builder_rd.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "cluster_builder_rd.h" +#include "servers/rendering/rendering_device.h" +#include "servers/rendering/rendering_server_globals.h" + +ClusterBuilderSharedDataRD::ClusterBuilderSharedDataRD() { + RD::VertexFormatID vertex_format; + + { + Vector<RD::VertexAttribute> attributes; + { + RD::VertexAttribute va; + va.format = RD::DATA_FORMAT_R32G32B32_SFLOAT; + va.stride = sizeof(float) * 3; + attributes.push_back(va); + } + vertex_format = RD::get_singleton()->vertex_format_create(attributes); + } + + { + Vector<String> versions; + versions.push_back(""); + cluster_render.cluster_render_shader.initialize(versions); + cluster_render.shader_version = cluster_render.cluster_render_shader.version_create(); + cluster_render.shader = cluster_render.cluster_render_shader.version_get_shader(cluster_render.shader_version, 0); + cluster_render.shader_pipelines[ClusterRender::PIPELINE_NORMAL] = RD::get_singleton()->render_pipeline_create(cluster_render.shader, RD::get_singleton()->framebuffer_format_create_empty(), vertex_format, RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RD::PipelineColorBlendState(), 0); + RD::PipelineMultisampleState ms; + ms.sample_count = RD::TEXTURE_SAMPLES_4; + cluster_render.shader_pipelines[ClusterRender::PIPELINE_MSAA] = RD::get_singleton()->render_pipeline_create(cluster_render.shader, RD::get_singleton()->framebuffer_format_create_empty(), vertex_format, RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), ms, RD::PipelineDepthStencilState(), RD::PipelineColorBlendState(), 0); + } + { + Vector<String> versions; + versions.push_back(""); + cluster_store.cluster_store_shader.initialize(versions); + cluster_store.shader_version = cluster_store.cluster_store_shader.version_create(); + cluster_store.shader = cluster_store.cluster_store_shader.version_get_shader(cluster_store.shader_version, 0); + cluster_store.shader_pipeline = RD::get_singleton()->compute_pipeline_create(cluster_store.shader); + } + { + Vector<String> versions; + versions.push_back(""); + cluster_debug.cluster_debug_shader.initialize(versions); + cluster_debug.shader_version = cluster_debug.cluster_debug_shader.version_create(); + cluster_debug.shader = cluster_debug.cluster_debug_shader.version_get_shader(cluster_debug.shader_version, 0); + cluster_debug.shader_pipeline = RD::get_singleton()->compute_pipeline_create(cluster_debug.shader); + } + + { // SPHERE + static const uint32_t icosphere_vertex_count = 42; + static const float icosphere_vertices[icosphere_vertex_count * 3] = { + 0, 0, -1, 0.7236073, -0.5257253, -0.4472195, -0.276388, -0.8506492, -0.4472199, -0.8944262, 0, -0.4472156, -0.276388, 0.8506492, -0.4472199, 0.7236073, 0.5257253, -0.4472195, 0.276388, -0.8506492, 0.4472199, -0.7236073, -0.5257253, 0.4472195, -0.7236073, 0.5257253, 0.4472195, 0.276388, 0.8506492, 0.4472199, 0.8944262, 0, 0.4472156, 0, 0, 1, -0.1624555, -0.4999952, -0.8506544, 0.4253227, -0.3090114, -0.8506542, 0.2628688, -0.8090116, -0.5257377, 0.8506479, 0, -0.5257359, 0.4253227, 0.3090114, -0.8506542, -0.5257298, 0, -0.8506517, -0.6881894, -0.4999969, -0.5257362, -0.1624555, 0.4999952, -0.8506544, -0.6881894, 0.4999969, -0.5257362, 0.2628688, 0.8090116, -0.5257377, 0.9510579, -0.3090126, 0, 0.9510579, 0.3090126, 0, 0, -1, 0, 0.5877856, -0.8090167, 0, -0.9510579, -0.3090126, 0, -0.5877856, -0.8090167, 0, -0.5877856, 0.8090167, 0, -0.9510579, 0.3090126, 0, 0.5877856, 0.8090167, 0, 0, 1, 0, 0.6881894, -0.4999969, 0.5257362, -0.2628688, -0.8090116, 0.5257377, -0.8506479, 0, 0.5257359, -0.2628688, 0.8090116, 0.5257377, 0.6881894, 0.4999969, 0.5257362, 0.1624555, -0.4999952, 0.8506544, 0.5257298, 0, 0.8506517, -0.4253227, -0.3090114, 0.8506542, -0.4253227, 0.3090114, 0.8506542, 0.1624555, 0.4999952, 0.8506544 + }; + static const uint32_t icosphere_triangle_count = 80; + static const uint32_t icosphere_triangle_indices[icosphere_triangle_count * 3] = { + 0, 13, 12, 1, 13, 15, 0, 12, 17, 0, 17, 19, 0, 19, 16, 1, 15, 22, 2, 14, 24, 3, 18, 26, 4, 20, 28, 5, 21, 30, 1, 22, 25, 2, 24, 27, 3, 26, 29, 4, 28, 31, 5, 30, 23, 6, 32, 37, 7, 33, 39, 8, 34, 40, 9, 35, 41, 10, 36, 38, 38, 41, 11, 38, 36, 41, 36, 9, 41, 41, 40, 11, 41, 35, 40, 35, 8, 40, 40, 39, 11, 40, 34, 39, 34, 7, 39, 39, 37, 11, 39, 33, 37, 33, 6, 37, 37, 38, 11, 37, 32, 38, 32, 10, 38, 23, 36, 10, 23, 30, 36, 30, 9, 36, 31, 35, 9, 31, 28, 35, 28, 8, 35, 29, 34, 8, 29, 26, 34, 26, 7, 34, 27, 33, 7, 27, 24, 33, 24, 6, 33, 25, 32, 6, 25, 22, 32, 22, 10, 32, 30, 31, 9, 30, 21, 31, 21, 4, 31, 28, 29, 8, 28, 20, 29, 20, 3, 29, 26, 27, 7, 26, 18, 27, 18, 2, 27, 24, 25, 6, 24, 14, 25, 14, 1, 25, 22, 23, 10, 22, 15, 23, 15, 5, 23, 16, 21, 5, 16, 19, 21, 19, 4, 21, 19, 20, 4, 19, 17, 20, 17, 3, 20, 17, 18, 3, 17, 12, 18, 12, 2, 18, 15, 16, 5, 15, 13, 16, 13, 0, 16, 12, 14, 2, 12, 13, 14, 13, 1, 14 + }; + + Vector<uint8_t> vertex_data; + vertex_data.resize(sizeof(float) * icosphere_vertex_count * 3); + memcpy(vertex_data.ptrw(), icosphere_vertices, vertex_data.size()); + + sphere_vertex_buffer = RD::get_singleton()->vertex_buffer_create(vertex_data.size(), vertex_data); + + Vector<uint8_t> index_data; + index_data.resize(sizeof(uint32_t) * icosphere_triangle_count * 3); + memcpy(index_data.ptrw(), icosphere_triangle_indices, index_data.size()); + + sphere_index_buffer = RD::get_singleton()->index_buffer_create(icosphere_triangle_count * 3, RD::INDEX_BUFFER_FORMAT_UINT32, index_data); + + Vector<RID> buffers; + buffers.push_back(sphere_vertex_buffer); + + sphere_vertex_array = RD::get_singleton()->vertex_array_create(icosphere_vertex_count, vertex_format, buffers); + + sphere_index_array = RD::get_singleton()->index_array_create(sphere_index_buffer, 0, icosphere_triangle_count * 3); + + float min_d = 1e20; + for (uint32_t i = 0; i < icosphere_triangle_count; i++) { + Vector3 vertices[3]; + for (uint32_t j = 0; j < 3; j++) { + uint32_t index = icosphere_triangle_indices[i * 3 + j]; + for (uint32_t k = 0; k < 3; k++) { + vertices[j][k] = icosphere_vertices[index * 3 + k]; + } + } + Plane p(vertices[0], vertices[1], vertices[2]); + min_d = MIN(Math::abs(p.d), min_d); + } + sphere_overfit = 1.0 / min_d; + } + + { // CONE + static const uint32_t cone_vertex_count = 99; + static const float cone_vertices[cone_vertex_count * 3] = { + 0, 1, -1, 0.1950903, 0.9807853, -1, 0.3826835, 0.9238795, -1, 0.5555703, 0.8314696, -1, 0.7071068, 0.7071068, -1, 0.8314697, 0.5555702, -1, 0.9238795, 0.3826834, -1, 0.9807853, 0.1950903, -1, 1, 0, -1, 0.9807853, -0.1950902, -1, 0.9238796, -0.3826833, -1, 0.8314697, -0.5555702, -1, 0.7071068, -0.7071068, -1, 0.5555702, -0.8314697, -1, 0.3826833, -0.9238796, -1, 0.1950901, -0.9807853, -1, -3.25841e-7, -1, -1, -0.1950907, -0.9807852, -1, -0.3826839, -0.9238793, -1, -0.5555707, -0.8314693, -1, -0.7071073, -0.7071063, -1, -0.83147, -0.5555697, -1, -0.9238799, -0.3826827, -1, 0, 0, 0, -0.9807854, -0.1950894, -1, -1, 9.65599e-7, -1, -0.9807851, 0.1950913, -1, -0.9238791, 0.3826845, -1, -0.8314689, 0.5555713, -1, -0.7071059, 0.7071077, -1, -0.5555691, 0.8314704, -1, -0.3826821, 0.9238801, -1, -0.1950888, 0.9807856, -1 + }; + static const uint32_t cone_triangle_count = 62; + static const uint32_t cone_triangle_indices[cone_triangle_count * 3] = { + 0, 23, 1, 1, 23, 2, 2, 23, 3, 3, 23, 4, 4, 23, 5, 5, 23, 6, 6, 23, 7, 7, 23, 8, 8, 23, 9, 9, 23, 10, 10, 23, 11, 11, 23, 12, 12, 23, 13, 13, 23, 14, 14, 23, 15, 15, 23, 16, 16, 23, 17, 17, 23, 18, 18, 23, 19, 19, 23, 20, 20, 23, 21, 21, 23, 22, 22, 23, 24, 24, 23, 25, 25, 23, 26, 26, 23, 27, 27, 23, 28, 28, 23, 29, 29, 23, 30, 30, 23, 31, 31, 23, 32, 32, 23, 0, 7, 15, 24, 32, 0, 1, 1, 2, 3, 3, 4, 5, 5, 6, 3, 6, 7, 3, 7, 8, 9, 9, 10, 7, 10, 11, 7, 11, 12, 15, 12, 13, 15, 13, 14, 15, 15, 16, 17, 17, 18, 19, 19, 20, 24, 20, 21, 24, 21, 22, 24, 24, 25, 26, 26, 27, 28, 28, 29, 30, 30, 31, 32, 32, 1, 3, 15, 17, 24, 17, 19, 24, 24, 26, 32, 26, 28, 32, 28, 30, 32, 32, 3, 7, 7, 11, 15, 32, 7, 24 + }; + + Vector<uint8_t> vertex_data; + vertex_data.resize(sizeof(float) * cone_vertex_count * 3); + memcpy(vertex_data.ptrw(), cone_vertices, vertex_data.size()); + + cone_vertex_buffer = RD::get_singleton()->vertex_buffer_create(vertex_data.size(), vertex_data); + + Vector<uint8_t> index_data; + index_data.resize(sizeof(uint32_t) * cone_triangle_count * 3); + memcpy(index_data.ptrw(), cone_triangle_indices, index_data.size()); + + cone_index_buffer = RD::get_singleton()->index_buffer_create(cone_triangle_count * 3, RD::INDEX_BUFFER_FORMAT_UINT32, index_data); + + Vector<RID> buffers; + buffers.push_back(cone_vertex_buffer); + + cone_vertex_array = RD::get_singleton()->vertex_array_create(cone_vertex_count, vertex_format, buffers); + + cone_index_array = RD::get_singleton()->index_array_create(cone_index_buffer, 0, cone_triangle_count * 3); + + float min_d = 1e20; + for (uint32_t i = 0; i < cone_triangle_count; i++) { + Vector3 vertices[3]; + int32_t zero_index = -1; + for (uint32_t j = 0; j < 3; j++) { + uint32_t index = cone_triangle_indices[i * 3 + j]; + for (uint32_t k = 0; k < 3; k++) { + vertices[j][k] = cone_vertices[index * 3 + k]; + } + if (vertices[j] == Vector3()) { + zero_index = j; + } + } + + if (zero_index != -1) { + Vector3 a = vertices[(zero_index + 1) % 3]; + Vector3 b = vertices[(zero_index + 2) % 3]; + Vector3 c = a + Vector3(0, 0, 1); + Plane p(a, b, c); + min_d = MIN(Math::abs(p.d), min_d); + } + } + cone_overfit = 1.0 / min_d; + } + + { // BOX + static const uint32_t box_vertex_count = 8; + static const float box_vertices[box_vertex_count * 3] = { + -1, -1, -1, -1, -1, 1, -1, 1, -1, -1, 1, 1, 1, -1, -1, 1, -1, 1, 1, 1, -1, 1, 1, 1 + }; + static const uint32_t box_triangle_count = 12; + static const uint32_t box_triangle_indices[box_triangle_count * 3] = { + 1, 2, 0, 3, 6, 2, 7, 4, 6, 5, 0, 4, 6, 0, 2, 3, 5, 7, 1, 3, 2, 3, 7, 6, 7, 5, 4, 5, 1, 0, 6, 4, 0, 3, 1, 5 + }; + + Vector<uint8_t> vertex_data; + vertex_data.resize(sizeof(float) * box_vertex_count * 3); + memcpy(vertex_data.ptrw(), box_vertices, vertex_data.size()); + + box_vertex_buffer = RD::get_singleton()->vertex_buffer_create(vertex_data.size(), vertex_data); + + Vector<uint8_t> index_data; + index_data.resize(sizeof(uint32_t) * box_triangle_count * 3); + memcpy(index_data.ptrw(), box_triangle_indices, index_data.size()); + + box_index_buffer = RD::get_singleton()->index_buffer_create(box_triangle_count * 3, RD::INDEX_BUFFER_FORMAT_UINT32, index_data); + + Vector<RID> buffers; + buffers.push_back(box_vertex_buffer); + + box_vertex_array = RD::get_singleton()->vertex_array_create(box_vertex_count, vertex_format, buffers); + + box_index_array = RD::get_singleton()->index_array_create(box_index_buffer, 0, box_triangle_count * 3); + } +} +ClusterBuilderSharedDataRD::~ClusterBuilderSharedDataRD() { + RD::get_singleton()->free(sphere_vertex_buffer); + RD::get_singleton()->free(sphere_index_buffer); + RD::get_singleton()->free(cone_vertex_buffer); + RD::get_singleton()->free(cone_index_buffer); + RD::get_singleton()->free(box_vertex_buffer); + RD::get_singleton()->free(box_index_buffer); + + cluster_render.cluster_render_shader.version_free(cluster_render.shader_version); + cluster_store.cluster_store_shader.version_free(cluster_store.shader_version); + cluster_debug.cluster_debug_shader.version_free(cluster_debug.shader_version); +} + +///////////////////////////// + +void ClusterBuilderRD::_clear() { + if (cluster_buffer.is_null()) { + return; //nothing to clear + } + RD::get_singleton()->free(cluster_buffer); + RD::get_singleton()->free(cluster_render_buffer); + RD::get_singleton()->free(element_buffer); + cluster_buffer = RID(); + cluster_render_buffer = RID(); + element_buffer = RID(); + + memfree(render_elements); + + render_elements = nullptr; + render_element_max = 0; + render_element_count = 0; + + RD::get_singleton()->free(framebuffer); + framebuffer = RID(); + + cluster_render_uniform_set = RID(); + cluster_store_uniform_set = RID(); +} + +void ClusterBuilderRD::setup(Size2i p_screen_size, uint32_t p_max_elements, RID p_depth_buffer, RID p_depth_buffer_sampler, RID p_color_buffer) { + ERR_FAIL_COND(p_max_elements == 0); + ERR_FAIL_COND(p_screen_size.x < 1); + ERR_FAIL_COND(p_screen_size.y < 1); + + _clear(); + + screen_size = p_screen_size; + + cluster_screen_size.width = (p_screen_size.width - 1) / cluster_size + 1; + cluster_screen_size.height = (p_screen_size.height - 1) / cluster_size + 1; + + max_elements_by_type = p_max_elements; + if (max_elements_by_type % 32) { //need to be 32 aligned + max_elements_by_type += 32 - (max_elements_by_type % 32); + } + + cluster_buffer_size = cluster_screen_size.x * cluster_screen_size.y * (max_elements_by_type / 32 + 32) * ELEMENT_TYPE_MAX * 4; + + render_element_max = max_elements_by_type * ELEMENT_TYPE_MAX; + + uint32_t element_tag_bits_size = render_element_max / 32; + uint32_t element_tag_depth_bits_size = render_element_max; + cluster_render_buffer_size = cluster_screen_size.x * cluster_screen_size.y * (element_tag_bits_size + element_tag_depth_bits_size) * 4; // tag bits (element was used) and tag depth (depth range in which it was used) + + cluster_render_buffer = RD::get_singleton()->storage_buffer_create(cluster_render_buffer_size); + cluster_buffer = RD::get_singleton()->storage_buffer_create(cluster_buffer_size); + + render_elements = static_cast<RenderElementData *>(memalloc(sizeof(RenderElementData *) * render_element_max)); + render_element_count = 0; + + element_buffer = RD::get_singleton()->storage_buffer_create(sizeof(RenderElementData) * render_element_max); + + uint32_t div_value = 1 << divisor; + if (use_msaa) { + framebuffer = RD::get_singleton()->framebuffer_create_empty(p_screen_size / div_value, RD::TEXTURE_SAMPLES_4); + } else { + framebuffer = RD::get_singleton()->framebuffer_create_empty(p_screen_size / div_value); + } + + { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 1; + u.append_id(state_uniform); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 2; + u.append_id(element_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 3; + u.append_id(cluster_render_buffer); + uniforms.push_back(u); + } + + cluster_render_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, shared->cluster_render.shader, 0); + } + + { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 1; + u.append_id(cluster_render_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 2; + u.append_id(cluster_buffer); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 3; + u.append_id(element_buffer); + uniforms.push_back(u); + } + + cluster_store_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, shared->cluster_store.shader, 0); + } + + if (p_color_buffer.is_valid()) { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 1; + u.append_id(cluster_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 2; + u.append_id(p_color_buffer); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 3; + u.append_id(p_depth_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 4; + u.append_id(p_depth_buffer_sampler); + uniforms.push_back(u); + } + + debug_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, shared->cluster_debug.shader, 0); + } else { + debug_uniform_set = RID(); + } +} + +void ClusterBuilderRD::begin(const Transform3D &p_view_transform, const Projection &p_cam_projection, bool p_flip_y) { + view_xform = p_view_transform.affine_inverse(); + projection = p_cam_projection; + z_near = projection.get_z_near(); + z_far = projection.get_z_far(); + orthogonal = p_cam_projection.is_orthogonal(); + adjusted_projection = projection; + if (!orthogonal) { + adjusted_projection.adjust_perspective_znear(0.0001); + } + + Projection correction; + correction.set_depth_correction(p_flip_y); + projection = correction * projection; + adjusted_projection = correction * adjusted_projection; + + //reset counts + render_element_count = 0; + for (uint32_t i = 0; i < ELEMENT_TYPE_MAX; i++) { + cluster_count_by_type[i] = 0; + } +} + +void ClusterBuilderRD::bake_cluster() { + RENDER_TIMESTAMP("> Bake 3D Cluster"); + + RD::get_singleton()->draw_command_begin_label("Bake Light Cluster"); + + //clear cluster buffer + RD::get_singleton()->buffer_clear(cluster_buffer, 0, cluster_buffer_size, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); + + if (render_element_count > 0) { + //clear render buffer + RD::get_singleton()->buffer_clear(cluster_render_buffer, 0, cluster_render_buffer_size, RD::BARRIER_MASK_RASTER); + + { //fill state uniform + + StateUniform state; + + RendererRD::MaterialStorage::store_camera(adjusted_projection, state.projection); + state.inv_z_far = 1.0 / z_far; + state.screen_to_clusters_shift = get_shift_from_power_of_2(cluster_size); + state.screen_to_clusters_shift -= divisor; //screen is smaller, shift one less + + state.cluster_screen_width = cluster_screen_size.x; + state.cluster_depth_offset = (render_element_max / 32); + state.cluster_data_size = state.cluster_depth_offset + render_element_max; + + RD::get_singleton()->buffer_update(state_uniform, 0, sizeof(StateUniform), &state, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); + } + + //update instances + + RD::get_singleton()->buffer_update(element_buffer, 0, sizeof(RenderElementData) * render_element_count, render_elements, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); + + RENDER_TIMESTAMP("Render 3D Cluster Elements"); + + //render elements + { + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(framebuffer, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD); + ClusterBuilderSharedDataRD::ClusterRender::PushConstant push_constant = {}; + + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, shared->cluster_render.shader_pipelines[use_msaa ? ClusterBuilderSharedDataRD::ClusterRender::PIPELINE_MSAA : ClusterBuilderSharedDataRD::ClusterRender::PIPELINE_NORMAL]); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, cluster_render_uniform_set, 0); + + for (uint32_t i = 0; i < render_element_count;) { + push_constant.base_index = i; + switch (render_elements[i].type) { + case ELEMENT_TYPE_OMNI_LIGHT: { + RD::get_singleton()->draw_list_bind_vertex_array(draw_list, shared->sphere_vertex_array); + RD::get_singleton()->draw_list_bind_index_array(draw_list, shared->sphere_index_array); + } break; + case ELEMENT_TYPE_SPOT_LIGHT: { + RD::get_singleton()->draw_list_bind_vertex_array(draw_list, shared->cone_vertex_array); + RD::get_singleton()->draw_list_bind_index_array(draw_list, shared->cone_index_array); + } break; + case ELEMENT_TYPE_DECAL: + case ELEMENT_TYPE_REFLECTION_PROBE: { + RD::get_singleton()->draw_list_bind_vertex_array(draw_list, shared->box_vertex_array); + RD::get_singleton()->draw_list_bind_index_array(draw_list, shared->box_index_array); + } break; + } + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &push_constant, sizeof(ClusterBuilderSharedDataRD::ClusterRender::PushConstant)); + + uint32_t instances = 1; + RD::get_singleton()->draw_list_draw(draw_list, true, instances); + i += instances; + } + RD::get_singleton()->draw_list_end(RD::BARRIER_MASK_COMPUTE); + } + //store elements + RENDER_TIMESTAMP("Pack 3D Cluster Elements"); + + { + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, shared->cluster_store.shader_pipeline); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, cluster_store_uniform_set, 0); + + ClusterBuilderSharedDataRD::ClusterStore::PushConstant push_constant; + push_constant.cluster_render_data_size = render_element_max / 32 + render_element_max; + push_constant.max_render_element_count_div_32 = render_element_max / 32; + push_constant.cluster_screen_size[0] = cluster_screen_size.x; + push_constant.cluster_screen_size[1] = cluster_screen_size.y; + push_constant.render_element_count_div_32 = render_element_count > 0 ? (render_element_count - 1) / 32 + 1 : 0; + push_constant.max_cluster_element_count_div_32 = max_elements_by_type / 32; + push_constant.pad1 = 0; + push_constant.pad2 = 0; + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ClusterBuilderSharedDataRD::ClusterStore::PushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, cluster_screen_size.x, cluster_screen_size.y, 1); + + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); + } + } else { + RD::get_singleton()->barrier(RD::BARRIER_MASK_TRANSFER, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); + } + RENDER_TIMESTAMP("< Bake 3D Cluster"); + RD::get_singleton()->draw_command_end_label(); +} + +void ClusterBuilderRD::debug(ElementType p_element) { + ERR_FAIL_COND(debug_uniform_set.is_null()); + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, shared->cluster_debug.shader_pipeline); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, debug_uniform_set, 0); + + ClusterBuilderSharedDataRD::ClusterDebug::PushConstant push_constant; + push_constant.screen_size[0] = screen_size.x; + push_constant.screen_size[1] = screen_size.y; + push_constant.cluster_screen_size[0] = cluster_screen_size.x; + push_constant.cluster_screen_size[1] = cluster_screen_size.y; + push_constant.cluster_shift = get_shift_from_power_of_2(cluster_size); + push_constant.cluster_type = p_element; + push_constant.orthogonal = orthogonal; + push_constant.z_far = z_far; + push_constant.z_near = z_near; + push_constant.max_cluster_element_count_div_32 = max_elements_by_type / 32; + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ClusterBuilderSharedDataRD::ClusterDebug::PushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, screen_size.x, screen_size.y, 1); + + RD::get_singleton()->compute_list_end(); +} + +RID ClusterBuilderRD::get_cluster_buffer() const { + return cluster_buffer; +} + +uint32_t ClusterBuilderRD::get_cluster_size() const { + return cluster_size; +} + +uint32_t ClusterBuilderRD::get_max_cluster_elements() const { + return max_elements_by_type; +} + +void ClusterBuilderRD::set_shared(ClusterBuilderSharedDataRD *p_shared) { + shared = p_shared; +} + +ClusterBuilderRD::ClusterBuilderRD() { + state_uniform = RD::get_singleton()->uniform_buffer_create(sizeof(StateUniform)); +} + +ClusterBuilderRD::~ClusterBuilderRD() { + _clear(); + RD::get_singleton()->free(state_uniform); +} diff --git a/servers/rendering/renderer_rd/cluster_builder_rd.h b/servers/rendering/renderer_rd/cluster_builder_rd.h new file mode 100644 index 0000000000..ef17ceb98c --- /dev/null +++ b/servers/rendering/renderer_rd/cluster_builder_rd.h @@ -0,0 +1,378 @@ +/*************************************************************************/ +/* cluster_builder_rd.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef CLUSTER_BUILDER_RD_H +#define CLUSTER_BUILDER_RD_H + +#include "servers/rendering/renderer_rd/shaders/cluster_debug.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/cluster_render.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/cluster_store.glsl.gen.h" +#include "servers/rendering/renderer_rd/storage_rd/material_storage.h" + +class ClusterBuilderSharedDataRD { + friend class ClusterBuilderRD; + + RID sphere_vertex_buffer; + RID sphere_vertex_array; + RID sphere_index_buffer; + RID sphere_index_array; + float sphere_overfit = 0.0; //because an icosphere is not a perfect sphere, we need to enlarge it to cover the sphere area + + RID cone_vertex_buffer; + RID cone_vertex_array; + RID cone_index_buffer; + RID cone_index_array; + float cone_overfit = 0.0; //because an cone mesh is not a perfect sphere, we need to enlarge it to cover the actual cone area + + RID box_vertex_buffer; + RID box_vertex_array; + RID box_index_buffer; + RID box_index_array; + + enum Divisor { + DIVISOR_1, + DIVISOR_2, + DIVISOR_4, + }; + + struct ClusterRender { + struct PushConstant { + uint32_t base_index; + uint32_t pad0; + uint32_t pad1; + uint32_t pad2; + }; + + ClusterRenderShaderRD cluster_render_shader; + RID shader_version; + RID shader; + enum PipelineVersion { + PIPELINE_NORMAL, + PIPELINE_MSAA, + PIPELINE_MAX + }; + + RID shader_pipelines[PIPELINE_MAX]; + } cluster_render; + + struct ClusterStore { + struct PushConstant { + uint32_t cluster_render_data_size; // how much data for a single cluster takes + uint32_t max_render_element_count_div_32; //divided by 32 + uint32_t cluster_screen_size[2]; + uint32_t render_element_count_div_32; //divided by 32 + uint32_t max_cluster_element_count_div_32; //divided by 32 + uint32_t pad1; + uint32_t pad2; + }; + + ClusterStoreShaderRD cluster_store_shader; + RID shader_version; + RID shader; + RID shader_pipeline; + } cluster_store; + + struct ClusterDebug { + struct PushConstant { + uint32_t screen_size[2]; + uint32_t cluster_screen_size[2]; + + uint32_t cluster_shift; + uint32_t cluster_type; + float z_near; + float z_far; + + uint32_t orthogonal; + uint32_t max_cluster_element_count_div_32; + uint32_t pad1; + uint32_t pad2; + }; + + ClusterDebugShaderRD cluster_debug_shader; + RID shader_version; + RID shader; + RID shader_pipeline; + } cluster_debug; + +public: + ClusterBuilderSharedDataRD(); + ~ClusterBuilderSharedDataRD(); +}; + +class ClusterBuilderRD { +public: + enum LightType { + LIGHT_TYPE_OMNI, + LIGHT_TYPE_SPOT + }; + + enum BoxType { + BOX_TYPE_REFLECTION_PROBE, + BOX_TYPE_DECAL, + }; + + enum ElementType { + ELEMENT_TYPE_OMNI_LIGHT, + ELEMENT_TYPE_SPOT_LIGHT, + ELEMENT_TYPE_DECAL, + ELEMENT_TYPE_REFLECTION_PROBE, + ELEMENT_TYPE_MAX, + + }; + +private: + ClusterBuilderSharedDataRD *shared = nullptr; + + struct RenderElementData { + uint32_t type; //0-4 + uint32_t touches_near; + uint32_t touches_far; + uint32_t original_index; + float transform_inv[12]; //transposed transform for less space + float scale[3]; + uint32_t pad; + }; + + uint32_t cluster_count_by_type[ELEMENT_TYPE_MAX] = {}; + uint32_t max_elements_by_type = 0; + + RenderElementData *render_elements = nullptr; + uint32_t render_element_count = 0; + uint32_t render_element_max = 0; + + Transform3D view_xform; + Projection adjusted_projection; + Projection projection; + float z_far = 0; + float z_near = 0; + bool orthogonal = false; + + enum Divisor { + DIVISOR_1, + DIVISOR_2, + DIVISOR_4, + }; + + uint32_t cluster_size = 32; + bool use_msaa = true; + Divisor divisor = DIVISOR_4; + + Size2i screen_size; + Size2i cluster_screen_size; + + RID framebuffer; + RID cluster_render_buffer; //used for creating + RID cluster_buffer; //used for rendering + RID element_buffer; //used for storing, to hint element touches far plane or near plane + uint32_t cluster_render_buffer_size = 0; + uint32_t cluster_buffer_size = 0; + + RID cluster_render_uniform_set; + RID cluster_store_uniform_set; + + //persistent data + + void _clear(); + + struct StateUniform { + float projection[16]; + float inv_z_far; + uint32_t screen_to_clusters_shift; // shift to obtain coordinates in block indices + uint32_t cluster_screen_width; // + uint32_t cluster_data_size; // how much data for a single cluster takes + uint32_t cluster_depth_offset; + uint32_t pad0; + uint32_t pad1; + uint32_t pad2; + }; + + RID state_uniform; + + RID debug_uniform_set; + +public: + void setup(Size2i p_screen_size, uint32_t p_max_elements, RID p_depth_buffer, RID p_depth_buffer_sampler, RID p_color_buffer); + + void begin(const Transform3D &p_view_transform, const Projection &p_cam_projection, bool p_flip_y); + + _FORCE_INLINE_ void add_light(LightType p_type, const Transform3D &p_transform, float p_radius, float p_spot_aperture) { + if (p_type == LIGHT_TYPE_OMNI && cluster_count_by_type[ELEMENT_TYPE_OMNI_LIGHT] == max_elements_by_type) { + return; //max number elements reached + } + if (p_type == LIGHT_TYPE_SPOT && cluster_count_by_type[ELEMENT_TYPE_SPOT_LIGHT] == max_elements_by_type) { + return; //max number elements reached + } + + RenderElementData &e = render_elements[render_element_count]; + + Transform3D xform = view_xform * p_transform; + + float radius = xform.basis.get_uniform_scale(); + if (radius < 0.98 || radius > 1.02) { + xform.basis.orthonormalize(); + } + + radius *= p_radius; + + if (p_type == LIGHT_TYPE_OMNI) { + radius *= shared->sphere_overfit; // overfit icosphere + + //omni + float depth = -xform.origin.z; + if (orthogonal) { + e.touches_near = (depth - radius) < z_near; + } else { + //contains camera inside light + float radius2 = radius * shared->sphere_overfit; // overfit again for outer size (camera may be outside actual sphere but behind an icosphere vertex) + e.touches_near = xform.origin.length_squared() < radius2 * radius2; + } + + e.touches_far = (depth + radius) > z_far; + e.scale[0] = radius; + e.scale[1] = radius; + e.scale[2] = radius; + e.type = ELEMENT_TYPE_OMNI_LIGHT; + e.original_index = cluster_count_by_type[ELEMENT_TYPE_OMNI_LIGHT]; + + RendererRD::MaterialStorage::store_transform_transposed_3x4(xform, e.transform_inv); + + cluster_count_by_type[ELEMENT_TYPE_OMNI_LIGHT]++; + + } else { + //spot + radius *= shared->cone_overfit; // overfit icosphere + + real_t len = Math::tan(Math::deg_to_rad(p_spot_aperture)) * radius; + //approximate, probably better to use a cone support function + float max_d = -1e20; + float min_d = 1e20; +#define CONE_MINMAX(m_x, m_y) \ + { \ + float d = -xform.xform(Vector3(len * m_x, len * m_y, -radius)).z; \ + min_d = MIN(d, min_d); \ + max_d = MAX(d, max_d); \ + } + + CONE_MINMAX(1, 1); + CONE_MINMAX(-1, 1); + CONE_MINMAX(-1, -1); + CONE_MINMAX(1, -1); + + if (orthogonal) { + e.touches_near = min_d < z_near; + } else { + //contains camera inside light + Plane base_plane(-xform.basis.get_column(Vector3::AXIS_Z), xform.origin); + float dist = base_plane.distance_to(Vector3()); + if (dist >= 0 && dist < radius) { + //inside, check angle + float angle = Math::rad_to_deg(Math::acos((-xform.origin.normalized()).dot(-xform.basis.get_column(Vector3::AXIS_Z)))); + e.touches_near = angle < p_spot_aperture * 1.05; //overfit aperture a little due to cone overfit + } else { + e.touches_near = false; + } + } + + e.touches_far = max_d > z_far; + + e.scale[0] = len * shared->cone_overfit; + e.scale[1] = len * shared->cone_overfit; + e.scale[2] = radius; + + e.type = ELEMENT_TYPE_SPOT_LIGHT; + e.original_index = cluster_count_by_type[ELEMENT_TYPE_SPOT_LIGHT]; //use omni since they share index + + RendererRD::MaterialStorage::store_transform_transposed_3x4(xform, e.transform_inv); + + cluster_count_by_type[ELEMENT_TYPE_SPOT_LIGHT]++; + } + + render_element_count++; + } + + _FORCE_INLINE_ void add_box(BoxType p_box_type, const Transform3D &p_transform, const Vector3 &p_half_extents) { + if (p_box_type == BOX_TYPE_DECAL && cluster_count_by_type[ELEMENT_TYPE_DECAL] == max_elements_by_type) { + return; //max number elements reached + } + if (p_box_type == BOX_TYPE_REFLECTION_PROBE && cluster_count_by_type[ELEMENT_TYPE_REFLECTION_PROBE] == max_elements_by_type) { + return; //max number elements reached + } + + RenderElementData &e = render_elements[render_element_count]; + Transform3D xform = view_xform * p_transform; + + //extract scale and scale the matrix by it, makes things simpler + Vector3 scale = p_half_extents; + for (uint32_t i = 0; i < 3; i++) { + float s = xform.basis.rows[i].length(); + scale[i] *= s; + xform.basis.rows[i] /= s; + }; + + float box_depth = Math::abs(xform.basis.xform_inv(Vector3(0, 0, -1)).dot(scale)); + float depth = -xform.origin.z; + + if (orthogonal) { + e.touches_near = depth - box_depth < z_near; + } else { + //contains camera inside box + Vector3 inside = xform.xform_inv(Vector3(0, 0, 0)).abs(); + e.touches_near = inside.x < scale.x && inside.y < scale.y && inside.z < scale.z; + } + + e.touches_far = depth + box_depth > z_far; + + e.scale[0] = scale.x; + e.scale[1] = scale.y; + e.scale[2] = scale.z; + + e.type = (p_box_type == BOX_TYPE_DECAL) ? ELEMENT_TYPE_DECAL : ELEMENT_TYPE_REFLECTION_PROBE; + e.original_index = cluster_count_by_type[e.type]; + + RendererRD::MaterialStorage::store_transform_transposed_3x4(xform, e.transform_inv); + + cluster_count_by_type[e.type]++; + render_element_count++; + } + + void bake_cluster(); + void debug(ElementType p_element); + + RID get_cluster_buffer() const; + uint32_t get_cluster_size() const; + uint32_t get_max_cluster_elements() const; + + void set_shared(ClusterBuilderSharedDataRD *p_shared); + + ClusterBuilderRD(); + ~ClusterBuilderRD(); +}; + +#endif // CLUSTER_BUILDER_RD_H diff --git a/servers/rendering/renderer_rd/effects/SCsub b/servers/rendering/renderer_rd/effects/SCsub new file mode 100644 index 0000000000..86681f9c74 --- /dev/null +++ b/servers/rendering/renderer_rd/effects/SCsub @@ -0,0 +1,5 @@ +#!/usr/bin/env python + +Import("env") + +env.add_source_files(env.servers_sources, "*.cpp") diff --git a/servers/rendering/renderer_rd/effects/bokeh_dof.cpp b/servers/rendering/renderer_rd/effects/bokeh_dof.cpp new file mode 100644 index 0000000000..27850695b0 --- /dev/null +++ b/servers/rendering/renderer_rd/effects/bokeh_dof.cpp @@ -0,0 +1,505 @@ +/*************************************************************************/ +/* bokeh_dof.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "bokeh_dof.h" +#include "copy_effects.h" +#include "servers/rendering/renderer_rd/renderer_compositor_rd.h" +#include "servers/rendering/renderer_rd/storage_rd/material_storage.h" +#include "servers/rendering/renderer_rd/uniform_set_cache_rd.h" +#include "servers/rendering/rendering_server_default.h" +#include "servers/rendering/storage/camera_attributes_storage.h" + +using namespace RendererRD; + +BokehDOF::BokehDOF(bool p_prefer_raster_effects) { + prefer_raster_effects = p_prefer_raster_effects; + + // Initialize bokeh + Vector<String> bokeh_modes; + bokeh_modes.push_back("\n#define MODE_GEN_BLUR_SIZE\n"); + bokeh_modes.push_back("\n#define MODE_BOKEH_BOX\n#define OUTPUT_WEIGHT\n"); + bokeh_modes.push_back("\n#define MODE_BOKEH_BOX\n"); + bokeh_modes.push_back("\n#define MODE_BOKEH_HEXAGONAL\n#define OUTPUT_WEIGHT\n"); + bokeh_modes.push_back("\n#define MODE_BOKEH_HEXAGONAL\n"); + bokeh_modes.push_back("\n#define MODE_BOKEH_CIRCULAR\n#define OUTPUT_WEIGHT\n"); + bokeh_modes.push_back("\n#define MODE_COMPOSITE_BOKEH\n"); + if (prefer_raster_effects) { + bokeh.raster_shader.initialize(bokeh_modes); + + bokeh.shader_version = bokeh.raster_shader.version_create(); + + const int att_count[BOKEH_MAX] = { 1, 2, 1, 2, 1, 2, 1 }; + for (int i = 0; i < BOKEH_MAX; i++) { + RD::PipelineColorBlendState blend_state = (i == BOKEH_COMPOSITE) ? RD::PipelineColorBlendState::create_blend(att_count[i]) : RD::PipelineColorBlendState::create_disabled(att_count[i]); + bokeh.raster_pipelines[i].setup(bokeh.raster_shader.version_get_shader(bokeh.shader_version, i), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), blend_state, 0); + } + } else { + bokeh.compute_shader.initialize(bokeh_modes); + bokeh.compute_shader.set_variant_enabled(BOKEH_GEN_BOKEH_BOX_NOWEIGHT, false); + bokeh.compute_shader.set_variant_enabled(BOKEH_GEN_BOKEH_HEXAGONAL_NOWEIGHT, false); + bokeh.shader_version = bokeh.compute_shader.version_create(); + + for (int i = 0; i < BOKEH_MAX; i++) { + if (bokeh.compute_shader.is_variant_enabled(i)) { + bokeh.compute_pipelines[i] = RD::get_singleton()->compute_pipeline_create(bokeh.compute_shader.version_get_shader(bokeh.shader_version, i)); + } + } + + for (int i = 0; i < BOKEH_MAX; i++) { + bokeh.raster_pipelines[i].clear(); + } + } +} + +BokehDOF::~BokehDOF() { + if (prefer_raster_effects) { + bokeh.raster_shader.version_free(bokeh.shader_version); + } else { + bokeh.compute_shader.version_free(bokeh.shader_version); + } +} + +void BokehDOF::bokeh_dof_compute(const BokehBuffers &p_buffers, RID p_camera_attributes, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal) { + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use compute version of bokeh depth of field with the mobile renderer."); + + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + bool dof_far = RSG::camera_attributes->camera_attributes_get_dof_far_enabled(p_camera_attributes); + float dof_far_begin = RSG::camera_attributes->camera_attributes_get_dof_far_distance(p_camera_attributes); + float dof_far_size = RSG::camera_attributes->camera_attributes_get_dof_far_transition(p_camera_attributes); + bool dof_near = RSG::camera_attributes->camera_attributes_get_dof_near_enabled(p_camera_attributes); + float dof_near_begin = RSG::camera_attributes->camera_attributes_get_dof_near_distance(p_camera_attributes); + float dof_near_size = RSG::camera_attributes->camera_attributes_get_dof_near_transition(p_camera_attributes); + float bokeh_size = RSG::camera_attributes->camera_attributes_get_dof_blur_amount(p_camera_attributes) * 64; // Base 64 pixel radius. + + bool use_jitter = RSG::camera_attributes->camera_attributes_get_dof_blur_use_jitter(); + RS::DOFBokehShape bokeh_shape = RSG::camera_attributes->camera_attributes_get_dof_blur_bokeh_shape(); + RS::DOFBlurQuality blur_quality = RSG::camera_attributes->camera_attributes_get_dof_blur_quality(); + + // setup our push constant + memset(&bokeh.push_constant, 0, sizeof(BokehPushConstant)); + bokeh.push_constant.blur_far_active = dof_far; + bokeh.push_constant.blur_far_begin = dof_far_begin; + bokeh.push_constant.blur_far_end = dof_far_begin + dof_far_size; // Only used with non-physically-based. + bokeh.push_constant.use_physical_far = dof_far_size < 0.0; + bokeh.push_constant.blur_size_far = bokeh_size; // Only used with physically-based. + + bokeh.push_constant.blur_near_active = dof_near; + bokeh.push_constant.blur_near_begin = dof_near_begin; + bokeh.push_constant.blur_near_end = dof_near_begin - dof_near_size; // Only used with non-physically-based. + bokeh.push_constant.use_physical_near = dof_near_size < 0.0; + bokeh.push_constant.blur_size_near = bokeh_size; // Only used with physically-based. + + bokeh.push_constant.use_jitter = use_jitter; + bokeh.push_constant.jitter_seed = Math::randf() * 1000.0; + + bokeh.push_constant.z_near = p_cam_znear; + bokeh.push_constant.z_far = p_cam_zfar; + bokeh.push_constant.orthogonal = p_cam_orthogonal; + bokeh.push_constant.blur_size = (dof_near_size < 0.0 && dof_far_size < 0.0) ? 32 : bokeh_size; // Cap with physically-based to keep performance reasonable. + + bokeh.push_constant.second_pass = false; + bokeh.push_constant.half_size = false; + + bokeh.push_constant.blur_scale = 0.5; + + // setup our uniforms + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_base_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_buffers.base_texture })); + RD::Uniform u_depth_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_buffers.depth_texture })); + RD::Uniform u_secondary_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_buffers.secondary_texture })); + RD::Uniform u_half_texture0(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_buffers.half_texture[0] })); + RD::Uniform u_half_texture1(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_buffers.half_texture[1] })); + + RD::Uniform u_base_image(RD::UNIFORM_TYPE_IMAGE, 0, p_buffers.base_texture); + RD::Uniform u_secondary_image(RD::UNIFORM_TYPE_IMAGE, 0, p_buffers.secondary_texture); + RD::Uniform u_half_image0(RD::UNIFORM_TYPE_IMAGE, 0, p_buffers.half_texture[0]); + RD::Uniform u_half_image1(RD::UNIFORM_TYPE_IMAGE, 0, p_buffers.half_texture[1]); + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + + /* FIRST PASS */ + // The alpha channel of the source color texture is filled with the expected circle size + // If used for DOF far, the size is positive, if used for near, its negative. + + RID shader = bokeh.compute_shader.version_get_shader(bokeh.shader_version, BOKEH_GEN_BLUR_SIZE); + ERR_FAIL_COND(shader.is_null()); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.compute_pipelines[BOKEH_GEN_BLUR_SIZE]); + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_base_image), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_depth_texture), 1); + + bokeh.push_constant.size[0] = p_buffers.base_texture_size.x; + bokeh.push_constant.size[1] = p_buffers.base_texture_size.y; + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_buffers.base_texture_size.x, p_buffers.base_texture_size.y, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + if (bokeh_shape == RS::DOF_BOKEH_BOX || bokeh_shape == RS::DOF_BOKEH_HEXAGON) { + //second pass + BokehMode mode = bokeh_shape == RS::DOF_BOKEH_BOX ? BOKEH_GEN_BOKEH_BOX : BOKEH_GEN_BOKEH_HEXAGONAL; + shader = bokeh.compute_shader.version_get_shader(bokeh.shader_version, mode); + ERR_FAIL_COND(shader.is_null()); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.compute_pipelines[mode]); + + static const int quality_samples[4] = { 6, 12, 12, 24 }; + + bokeh.push_constant.steps = quality_samples[blur_quality]; + + if (blur_quality == RS::DOF_BLUR_QUALITY_VERY_LOW || blur_quality == RS::DOF_BLUR_QUALITY_LOW) { + //box and hexagon are more or less the same, and they can work in either half (very low and low quality) or full (medium and high quality_ sizes) + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_half_image0), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_base_texture), 1); + + bokeh.push_constant.size[0] = p_buffers.base_texture_size.x >> 1; + bokeh.push_constant.size[1] = p_buffers.base_texture_size.y >> 1; + bokeh.push_constant.half_size = true; + bokeh.push_constant.blur_size *= 0.5; + + } else { + //medium and high quality use full size + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_secondary_image), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_base_texture), 1); + } + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, bokeh.push_constant.size[0], bokeh.push_constant.size[1], 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + //third pass + bokeh.push_constant.second_pass = true; + + if (blur_quality == RS::DOF_BLUR_QUALITY_VERY_LOW || blur_quality == RS::DOF_BLUR_QUALITY_LOW) { + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_half_image1), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_half_texture0), 1); + } else { + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_base_image), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_secondary_texture), 1); + } + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, bokeh.push_constant.size[0], bokeh.push_constant.size[1], 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + if (blur_quality == RS::DOF_BLUR_QUALITY_VERY_LOW || blur_quality == RS::DOF_BLUR_QUALITY_LOW) { + //forth pass, upscale for low quality + + shader = bokeh.compute_shader.version_get_shader(bokeh.shader_version, BOKEH_COMPOSITE); + ERR_FAIL_COND(shader.is_null()); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.compute_pipelines[BOKEH_COMPOSITE]); + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_base_image), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_half_texture1), 1); + + bokeh.push_constant.size[0] = p_buffers.base_texture_size.x; + bokeh.push_constant.size[1] = p_buffers.base_texture_size.y; + bokeh.push_constant.half_size = false; + bokeh.push_constant.second_pass = false; + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_buffers.base_texture_size.x, p_buffers.base_texture_size.y, 1); + } + } else { + //circle + + shader = bokeh.compute_shader.version_get_shader(bokeh.shader_version, BOKEH_GEN_BOKEH_CIRCULAR); + ERR_FAIL_COND(shader.is_null()); + + //second pass + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.compute_pipelines[BOKEH_GEN_BOKEH_CIRCULAR]); + + static const float quality_scale[4] = { 8.0, 4.0, 1.0, 0.5 }; + + bokeh.push_constant.steps = 0; + bokeh.push_constant.blur_scale = quality_scale[blur_quality]; + + //circle always runs in half size, otherwise too expensive + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_half_image0), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_base_texture), 1); + + bokeh.push_constant.size[0] = p_buffers.base_texture_size.x >> 1; + bokeh.push_constant.size[1] = p_buffers.base_texture_size.y >> 1; + bokeh.push_constant.half_size = true; + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, bokeh.push_constant.size[0], bokeh.push_constant.size[1], 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + //circle is just one pass, then upscale + + // upscale + + shader = bokeh.compute_shader.version_get_shader(bokeh.shader_version, BOKEH_COMPOSITE); + ERR_FAIL_COND(shader.is_null()); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.compute_pipelines[BOKEH_COMPOSITE]); + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_base_image), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_half_texture0), 1); + + bokeh.push_constant.size[0] = p_buffers.base_texture_size.x; + bokeh.push_constant.size[1] = p_buffers.base_texture_size.y; + bokeh.push_constant.half_size = false; + bokeh.push_constant.second_pass = false; + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_buffers.base_texture_size.x, p_buffers.base_texture_size.y, 1); + } + + RD::get_singleton()->compute_list_end(); +} + +void BokehDOF::bokeh_dof_raster(const BokehBuffers &p_buffers, RID p_camera_attributes, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal) { + ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't blur-based depth of field with the clustered renderer."); + + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + bool dof_far = RSG::camera_attributes->camera_attributes_get_dof_far_enabled(p_camera_attributes); + float dof_far_begin = RSG::camera_attributes->camera_attributes_get_dof_far_distance(p_camera_attributes); + float dof_far_size = RSG::camera_attributes->camera_attributes_get_dof_far_transition(p_camera_attributes); + bool dof_near = RSG::camera_attributes->camera_attributes_get_dof_near_enabled(p_camera_attributes); + float dof_near_begin = RSG::camera_attributes->camera_attributes_get_dof_near_distance(p_camera_attributes); + float dof_near_size = RSG::camera_attributes->camera_attributes_get_dof_near_transition(p_camera_attributes); + float bokeh_size = RSG::camera_attributes->camera_attributes_get_dof_blur_amount(p_camera_attributes) * 64; // Base 64 pixel radius. + + RS::DOFBokehShape bokeh_shape = RSG::camera_attributes->camera_attributes_get_dof_blur_bokeh_shape(); + RS::DOFBlurQuality blur_quality = RSG::camera_attributes->camera_attributes_get_dof_blur_quality(); + + // setup our base push constant + memset(&bokeh.push_constant, 0, sizeof(BokehPushConstant)); + + bokeh.push_constant.orthogonal = p_cam_orthogonal; + bokeh.push_constant.size[0] = p_buffers.base_texture_size.width; + bokeh.push_constant.size[1] = p_buffers.base_texture_size.height; + bokeh.push_constant.z_far = p_cam_zfar; + bokeh.push_constant.z_near = p_cam_znear; + + bokeh.push_constant.second_pass = false; + bokeh.push_constant.half_size = false; + bokeh.push_constant.blur_size = bokeh_size; + + // setup our uniforms + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_base_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_buffers.base_texture })); + RD::Uniform u_depth_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_buffers.depth_texture })); + RD::Uniform u_secondary_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_buffers.secondary_texture })); + RD::Uniform u_half_texture0(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_buffers.half_texture[0] })); + RD::Uniform u_half_texture1(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_buffers.half_texture[1] })); + RD::Uniform u_weight_texture0(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_buffers.weight_texture[0] })); + RD::Uniform u_weight_texture1(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_buffers.weight_texture[1] })); + RD::Uniform u_weight_texture2(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_buffers.weight_texture[2] })); + RD::Uniform u_weight_texture3(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_buffers.weight_texture[3] })); + + if (dof_far || dof_near) { + if (dof_far) { + bokeh.push_constant.blur_far_active = true; + bokeh.push_constant.blur_far_begin = dof_far_begin; + bokeh.push_constant.blur_far_end = dof_far_begin + dof_far_size; + } + + if (dof_near) { + bokeh.push_constant.blur_near_active = true; + bokeh.push_constant.blur_near_begin = dof_near_begin; + bokeh.push_constant.blur_near_end = dof_near_begin - dof_near_size; + } + + { + // generate our depth data + RID shader = bokeh.raster_shader.version_get_shader(bokeh.shader_version, BOKEH_GEN_BLUR_SIZE); + ERR_FAIL_COND(shader.is_null()); + + RID framebuffer = p_buffers.base_weight_fb; + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, bokeh.raster_pipelines[BOKEH_GEN_BLUR_SIZE].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 0, u_depth_texture), 0); + RD::get_singleton()->draw_list_bind_index_array(draw_list, material_storage->get_quad_index_array()); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &bokeh.push_constant, sizeof(BokehPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + } + + if (bokeh_shape == RS::DOF_BOKEH_BOX || bokeh_shape == RS::DOF_BOKEH_HEXAGON) { + // double pass approach + BokehMode mode = bokeh_shape == RS::DOF_BOKEH_BOX ? BOKEH_GEN_BOKEH_BOX : BOKEH_GEN_BOKEH_HEXAGONAL; + + RID shader = bokeh.raster_shader.version_get_shader(bokeh.shader_version, mode); + ERR_FAIL_COND(shader.is_null()); + + if (blur_quality == RS::DOF_BLUR_QUALITY_VERY_LOW || blur_quality == RS::DOF_BLUR_QUALITY_LOW) { + //box and hexagon are more or less the same, and they can work in either half (very low and low quality) or full (medium and high quality_ sizes) + bokeh.push_constant.size[0] = p_buffers.base_texture_size.x >> 1; + bokeh.push_constant.size[1] = p_buffers.base_texture_size.y >> 1; + bokeh.push_constant.half_size = true; + bokeh.push_constant.blur_size *= 0.5; + } + + static const int quality_samples[4] = { 6, 12, 12, 24 }; + bokeh.push_constant.blur_scale = 0.5; + bokeh.push_constant.steps = quality_samples[blur_quality]; + + RID framebuffer = bokeh.push_constant.half_size ? p_buffers.half_fb[0] : p_buffers.secondary_fb; + + // Pass 1 + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, bokeh.raster_pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 0, u_base_texture), 0); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 1, u_weight_texture0), 1); + RD::get_singleton()->draw_list_bind_index_array(draw_list, material_storage->get_quad_index_array()); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &bokeh.push_constant, sizeof(BokehPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + + // Pass 2 + if (!bokeh.push_constant.half_size) { + // do not output weight, we're writing back into our base buffer + mode = bokeh_shape == RS::DOF_BOKEH_BOX ? BOKEH_GEN_BOKEH_BOX_NOWEIGHT : BOKEH_GEN_BOKEH_HEXAGONAL_NOWEIGHT; + + shader = bokeh.raster_shader.version_get_shader(bokeh.shader_version, mode); + ERR_FAIL_COND(shader.is_null()); + } + bokeh.push_constant.second_pass = true; + + framebuffer = bokeh.push_constant.half_size ? p_buffers.half_fb[1] : p_buffers.base_fb; + RD::Uniform texture = bokeh.push_constant.half_size ? u_half_texture0 : u_secondary_texture; + RD::Uniform weight = bokeh.push_constant.half_size ? u_weight_texture2 : u_weight_texture1; + + draw_list = RD::get_singleton()->draw_list_begin(framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, bokeh.raster_pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 0, texture), 0); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 1, weight), 1); + RD::get_singleton()->draw_list_bind_index_array(draw_list, material_storage->get_quad_index_array()); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &bokeh.push_constant, sizeof(BokehPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + + if (bokeh.push_constant.half_size) { + // Compose pass + mode = BOKEH_COMPOSITE; + shader = bokeh.raster_shader.version_get_shader(bokeh.shader_version, mode); + ERR_FAIL_COND(shader.is_null()); + + framebuffer = p_buffers.base_fb; + + draw_list = RD::get_singleton()->draw_list_begin(framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, bokeh.raster_pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 0, u_half_texture1), 0); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 1, u_weight_texture3), 1); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 2, u_weight_texture0), 2); + RD::get_singleton()->draw_list_bind_index_array(draw_list, material_storage->get_quad_index_array()); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &bokeh.push_constant, sizeof(BokehPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + } + + } else { + // circular is a single pass approach + BokehMode mode = BOKEH_GEN_BOKEH_CIRCULAR; + + RID shader = bokeh.raster_shader.version_get_shader(bokeh.shader_version, mode); + ERR_FAIL_COND(shader.is_null()); + + { + // circle always runs in half size, otherwise too expensive (though the code below does support making this optional) + bokeh.push_constant.size[0] = p_buffers.base_texture_size.x >> 1; + bokeh.push_constant.size[1] = p_buffers.base_texture_size.y >> 1; + bokeh.push_constant.half_size = true; + // bokeh.push_constant.blur_size *= 0.5; + } + + static const float quality_scale[4] = { 8.0, 4.0, 1.0, 0.5 }; + bokeh.push_constant.blur_scale = quality_scale[blur_quality]; + bokeh.push_constant.steps = 0.0; + + RID framebuffer = bokeh.push_constant.half_size ? p_buffers.half_fb[0] : p_buffers.secondary_fb; + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, bokeh.raster_pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 0, u_base_texture), 0); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 1, u_weight_texture0), 1); + RD::get_singleton()->draw_list_bind_index_array(draw_list, material_storage->get_quad_index_array()); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &bokeh.push_constant, sizeof(BokehPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + + if (bokeh.push_constant.half_size) { + // Compose + mode = BOKEH_COMPOSITE; + shader = bokeh.raster_shader.version_get_shader(bokeh.shader_version, mode); + ERR_FAIL_COND(shader.is_null()); + + framebuffer = p_buffers.base_fb; + + draw_list = RD::get_singleton()->draw_list_begin(framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, bokeh.raster_pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 0, u_half_texture0), 0); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 1, u_weight_texture2), 1); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 2, u_weight_texture0), 2); + RD::get_singleton()->draw_list_bind_index_array(draw_list, material_storage->get_quad_index_array()); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &bokeh.push_constant, sizeof(BokehPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + } else { + CopyEffects::get_singleton()->copy_raster(p_buffers.secondary_texture, p_buffers.base_fb); + } + } + } +} diff --git a/servers/rendering/renderer_rd/effects/bokeh_dof.h b/servers/rendering/renderer_rd/effects/bokeh_dof.h new file mode 100644 index 0000000000..33dbdfcdc1 --- /dev/null +++ b/servers/rendering/renderer_rd/effects/bokeh_dof.h @@ -0,0 +1,125 @@ +/*************************************************************************/ +/* bokeh_dof.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef BOKEH_DOF_RD_H +#define BOKEH_DOF_RD_H + +#include "servers/rendering/renderer_rd/pipeline_cache_rd.h" +#include "servers/rendering/renderer_rd/shaders/effects/bokeh_dof.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/bokeh_dof_raster.glsl.gen.h" +#include "servers/rendering/renderer_scene_render.h" + +#include "servers/rendering_server.h" + +namespace RendererRD { + +class BokehDOF { +private: + bool prefer_raster_effects; + + struct BokehPushConstant { + uint32_t size[2]; + float z_far; + float z_near; + + uint32_t orthogonal; + float blur_size; + float blur_scale; + uint32_t steps; + + uint32_t blur_near_active; + float blur_near_begin; + float blur_near_end; + uint32_t blur_far_active; + + float blur_far_begin; + float blur_far_end; + uint32_t second_pass; + uint32_t half_size; + + uint32_t use_jitter; + float jitter_seed; + uint32_t use_physical_near; + uint32_t use_physical_far; + + float blur_size_near; + float blur_size_far; + uint32_t pad[2]; + }; + + enum BokehMode { + BOKEH_GEN_BLUR_SIZE, + BOKEH_GEN_BOKEH_BOX, + BOKEH_GEN_BOKEH_BOX_NOWEIGHT, + BOKEH_GEN_BOKEH_HEXAGONAL, + BOKEH_GEN_BOKEH_HEXAGONAL_NOWEIGHT, + BOKEH_GEN_BOKEH_CIRCULAR, + BOKEH_COMPOSITE, + BOKEH_MAX + }; + + struct Bokeh { + BokehPushConstant push_constant; + BokehDofShaderRD compute_shader; + BokehDofRasterShaderRD raster_shader; + RID shader_version; + RID compute_pipelines[BOKEH_MAX]; + PipelineCacheRD raster_pipelines[BOKEH_MAX]; + } bokeh; + +public: + struct BokehBuffers { + // bokeh buffers + + // textures + Size2i base_texture_size; + RID base_texture; + RID depth_texture; + RID secondary_texture; + RID half_texture[2]; + + // raster only + RID base_fb; + RID secondary_fb; // with weights + RID half_fb[2]; // with weights + RID base_weight_fb; + RID weight_texture[4]; + }; + + BokehDOF(bool p_prefer_raster_effects); + ~BokehDOF(); + + void bokeh_dof_compute(const BokehBuffers &p_buffers, RID p_camera_attributes, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal); + void bokeh_dof_raster(const BokehBuffers &p_buffers, RID p_camera_attributes, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal); +}; + +} // namespace RendererRD + +#endif // BOKEH_DOF_RD_H diff --git a/servers/rendering/renderer_rd/effects/copy_effects.cpp b/servers/rendering/renderer_rd/effects/copy_effects.cpp new file mode 100644 index 0000000000..a14228eb3d --- /dev/null +++ b/servers/rendering/renderer_rd/effects/copy_effects.cpp @@ -0,0 +1,1196 @@ +/*************************************************************************/ +/* copy_effects.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "copy_effects.h" +#include "core/config/project_settings.h" +#include "servers/rendering/renderer_rd/renderer_compositor_rd.h" +#include "servers/rendering/renderer_rd/storage_rd/material_storage.h" +#include "servers/rendering/renderer_rd/uniform_set_cache_rd.h" +#include "thirdparty/misc/cubemap_coeffs.h" + +using namespace RendererRD; + +CopyEffects *CopyEffects::singleton = nullptr; + +CopyEffects *CopyEffects::get_singleton() { + return singleton; +} + +CopyEffects::CopyEffects(bool p_prefer_raster_effects) { + singleton = this; + prefer_raster_effects = p_prefer_raster_effects; + + if (prefer_raster_effects) { + // init blur shader (on compute use copy shader) + + Vector<String> blur_modes; + blur_modes.push_back("\n#define MODE_MIPMAP\n"); // BLUR_MIPMAP + blur_modes.push_back("\n#define MODE_GAUSSIAN_BLUR\n"); // BLUR_MODE_GAUSSIAN_BLUR + blur_modes.push_back("\n#define MODE_GAUSSIAN_GLOW\n"); // BLUR_MODE_GAUSSIAN_GLOW + blur_modes.push_back("\n#define MODE_GAUSSIAN_GLOW\n#define GLOW_USE_AUTO_EXPOSURE\n"); // BLUR_MODE_GAUSSIAN_GLOW_AUTO_EXPOSURE + blur_modes.push_back("\n#define MODE_COPY\n"); // BLUR_MODE_COPY + + blur_raster.shader.initialize(blur_modes); + memset(&blur_raster.push_constant, 0, sizeof(BlurRasterPushConstant)); + blur_raster.shader_version = blur_raster.shader.version_create(); + + for (int i = 0; i < BLUR_MODE_MAX; i++) { + blur_raster.pipelines[i].setup(blur_raster.shader.version_get_shader(blur_raster.shader_version, i), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RD::PipelineColorBlendState::create_disabled(), 0); + } + + } else { + // not used in clustered + for (int i = 0; i < BLUR_MODE_MAX; i++) { + blur_raster.pipelines[i].clear(); + } + + Vector<String> copy_modes; + copy_modes.push_back("\n#define MODE_GAUSSIAN_BLUR\n"); + copy_modes.push_back("\n#define MODE_GAUSSIAN_BLUR\n#define DST_IMAGE_8BIT\n"); + copy_modes.push_back("\n#define MODE_GAUSSIAN_BLUR\n#define MODE_GLOW\n"); + copy_modes.push_back("\n#define MODE_GAUSSIAN_BLUR\n#define MODE_GLOW\n#define GLOW_USE_AUTO_EXPOSURE\n"); + copy_modes.push_back("\n#define MODE_SIMPLE_COPY\n"); + copy_modes.push_back("\n#define MODE_SIMPLE_COPY\n#define DST_IMAGE_8BIT\n"); + copy_modes.push_back("\n#define MODE_SIMPLE_COPY_DEPTH\n"); + copy_modes.push_back("\n#define MODE_SET_COLOR\n"); + copy_modes.push_back("\n#define MODE_SET_COLOR\n#define DST_IMAGE_8BIT\n"); + copy_modes.push_back("\n#define MODE_MIPMAP\n"); + copy_modes.push_back("\n#define MODE_LINEARIZE_DEPTH_COPY\n"); + copy_modes.push_back("\n#define MODE_CUBEMAP_TO_PANORAMA\n"); + copy_modes.push_back("\n#define MODE_CUBEMAP_ARRAY_TO_PANORAMA\n"); + + copy.shader.initialize(copy_modes); + memset(©.push_constant, 0, sizeof(CopyPushConstant)); + + copy.shader_version = copy.shader.version_create(); + + for (int i = 0; i < COPY_MODE_MAX; i++) { + if (copy.shader.is_variant_enabled(i)) { + copy.pipelines[i] = RD::get_singleton()->compute_pipeline_create(copy.shader.version_get_shader(copy.shader_version, i)); + } + } + } + + { + Vector<String> copy_modes; + copy_modes.push_back("\n"); // COPY_TO_FB_COPY + copy_modes.push_back("\n#define MODE_PANORAMA_TO_DP\n"); // COPY_TO_FB_COPY_PANORAMA_TO_DP + copy_modes.push_back("\n#define MODE_TWO_SOURCES\n"); // COPY_TO_FB_COPY2 + copy_modes.push_back("\n#define MULTIVIEW\n"); // COPY_TO_FB_MULTIVIEW + copy_modes.push_back("\n#define MULTIVIEW\n#define MODE_TWO_SOURCES\n"); // COPY_TO_FB_MULTIVIEW_WITH_DEPTH + + copy_to_fb.shader.initialize(copy_modes); + + if (!RendererCompositorRD::singleton->is_xr_enabled()) { + copy_to_fb.shader.set_variant_enabled(COPY_TO_FB_MULTIVIEW, false); + copy_to_fb.shader.set_variant_enabled(COPY_TO_FB_MULTIVIEW_WITH_DEPTH, false); + } + + copy_to_fb.shader_version = copy_to_fb.shader.version_create(); + + //use additive + + for (int i = 0; i < COPY_TO_FB_MAX; i++) { + if (copy_to_fb.shader.is_variant_enabled(i)) { + copy_to_fb.pipelines[i].setup(copy_to_fb.shader.version_get_shader(copy_to_fb.shader_version, i), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RD::PipelineColorBlendState::create_disabled(), 0); + } else { + copy_to_fb.pipelines[i].clear(); + } + } + } + + { + // Initialize copier + Vector<String> copy_modes; + copy_modes.push_back("\n"); + + cube_to_dp.shader.initialize(copy_modes); + + cube_to_dp.shader_version = cube_to_dp.shader.version_create(); + RID shader = cube_to_dp.shader.version_get_shader(cube_to_dp.shader_version, 0); + RD::PipelineDepthStencilState dss; + dss.enable_depth_test = true; + dss.depth_compare_operator = RD::COMPARE_OP_ALWAYS; + dss.enable_depth_write = true; + cube_to_dp.pipeline.setup(shader, RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), dss, RD::PipelineColorBlendState(), 0); + } + + { + //Initialize cubemap downsampler + Vector<String> cubemap_downsampler_modes; + cubemap_downsampler_modes.push_back(""); + + if (prefer_raster_effects) { + cubemap_downsampler.raster_shader.initialize(cubemap_downsampler_modes); + + cubemap_downsampler.shader_version = cubemap_downsampler.raster_shader.version_create(); + + cubemap_downsampler.raster_pipeline.setup(cubemap_downsampler.raster_shader.version_get_shader(cubemap_downsampler.shader_version, 0), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RD::PipelineColorBlendState::create_disabled(), 0); + } else { + cubemap_downsampler.compute_shader.initialize(cubemap_downsampler_modes); + + cubemap_downsampler.shader_version = cubemap_downsampler.compute_shader.version_create(); + + cubemap_downsampler.compute_pipeline = RD::get_singleton()->compute_pipeline_create(cubemap_downsampler.compute_shader.version_get_shader(cubemap_downsampler.shader_version, 0)); + cubemap_downsampler.raster_pipeline.clear(); + } + } + + { + // Initialize cubemap filter + filter.use_high_quality = GLOBAL_GET("rendering/reflections/sky_reflections/fast_filter_high_quality"); + + Vector<String> cubemap_filter_modes; + cubemap_filter_modes.push_back("\n#define USE_HIGH_QUALITY\n"); + cubemap_filter_modes.push_back("\n#define USE_LOW_QUALITY\n"); + cubemap_filter_modes.push_back("\n#define USE_HIGH_QUALITY\n#define USE_TEXTURE_ARRAY\n"); + cubemap_filter_modes.push_back("\n#define USE_LOW_QUALITY\n#define USE_TEXTURE_ARRAY\n"); + + if (filter.use_high_quality) { + filter.coefficient_buffer = RD::get_singleton()->storage_buffer_create(sizeof(high_quality_coeffs)); + RD::get_singleton()->buffer_update(filter.coefficient_buffer, 0, sizeof(high_quality_coeffs), &high_quality_coeffs[0]); + } else { + filter.coefficient_buffer = RD::get_singleton()->storage_buffer_create(sizeof(low_quality_coeffs)); + RD::get_singleton()->buffer_update(filter.coefficient_buffer, 0, sizeof(low_quality_coeffs), &low_quality_coeffs[0]); + } + + if (prefer_raster_effects) { + filter.raster_shader.initialize(cubemap_filter_modes); + + // array variants are not supported in raster + filter.raster_shader.set_variant_enabled(FILTER_MODE_HIGH_QUALITY_ARRAY, false); + filter.raster_shader.set_variant_enabled(FILTER_MODE_LOW_QUALITY_ARRAY, false); + + filter.shader_version = filter.raster_shader.version_create(); + + for (int i = 0; i < FILTER_MODE_MAX; i++) { + if (filter.raster_shader.is_variant_enabled(i)) { + filter.raster_pipelines[i].setup(filter.raster_shader.version_get_shader(filter.shader_version, i), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RD::PipelineColorBlendState::create_disabled(), 0); + } else { + filter.raster_pipelines[i].clear(); + } + } + + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 0; + u.append_id(filter.coefficient_buffer); + uniforms.push_back(u); + } + filter.uniform_set = RD::get_singleton()->uniform_set_create(uniforms, filter.raster_shader.version_get_shader(filter.shader_version, filter.use_high_quality ? 0 : 1), 1); + } else { + filter.compute_shader.initialize(cubemap_filter_modes); + filter.shader_version = filter.compute_shader.version_create(); + + for (int i = 0; i < FILTER_MODE_MAX; i++) { + filter.compute_pipelines[i] = RD::get_singleton()->compute_pipeline_create(filter.compute_shader.version_get_shader(filter.shader_version, i)); + filter.raster_pipelines[i].clear(); + } + + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 0; + u.append_id(filter.coefficient_buffer); + uniforms.push_back(u); + } + filter.uniform_set = RD::get_singleton()->uniform_set_create(uniforms, filter.compute_shader.version_get_shader(filter.shader_version, filter.use_high_quality ? 0 : 1), 1); + } + } + + { + // Initialize roughness + Vector<String> cubemap_roughness_modes; + cubemap_roughness_modes.push_back(""); + + if (prefer_raster_effects) { + roughness.raster_shader.initialize(cubemap_roughness_modes); + + roughness.shader_version = roughness.raster_shader.version_create(); + + roughness.raster_pipeline.setup(roughness.raster_shader.version_get_shader(roughness.shader_version, 0), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RD::PipelineColorBlendState::create_disabled(), 0); + + } else { + roughness.compute_shader.initialize(cubemap_roughness_modes); + + roughness.shader_version = roughness.compute_shader.version_create(); + + roughness.compute_pipeline = RD::get_singleton()->compute_pipeline_create(roughness.compute_shader.version_get_shader(roughness.shader_version, 0)); + roughness.raster_pipeline.clear(); + } + } + + { + Vector<String> specular_modes; + specular_modes.push_back("\n#define MODE_MERGE\n"); // SPECULAR_MERGE_ADD + specular_modes.push_back("\n#define MODE_MERGE\n#define MODE_SSR\n"); // SPECULAR_MERGE_SSR + specular_modes.push_back("\n"); // SPECULAR_MERGE_ADDITIVE_ADD + specular_modes.push_back("\n#define MODE_SSR\n"); // SPECULAR_MERGE_ADDITIVE_SSR + + specular_modes.push_back("\n#define USE_MULTIVIEW\n#define MODE_MERGE\n"); // SPECULAR_MERGE_ADD_MULTIVIEW + specular_modes.push_back("\n#define USE_MULTIVIEW\n#define MODE_MERGE\n#define MODE_SSR\n"); // SPECULAR_MERGE_SSR_MULTIVIEW + specular_modes.push_back("\n#define USE_MULTIVIEW\n"); // SPECULAR_MERGE_ADDITIVE_ADD_MULTIVIEW + specular_modes.push_back("\n#define USE_MULTIVIEW\n#define MODE_SSR\n"); // SPECULAR_MERGE_ADDITIVE_SSR_MULTIVIEW + + specular_merge.shader.initialize(specular_modes); + + if (!RendererCompositorRD::singleton->is_xr_enabled()) { + specular_merge.shader.set_variant_enabled(SPECULAR_MERGE_ADD_MULTIVIEW, false); + specular_merge.shader.set_variant_enabled(SPECULAR_MERGE_SSR_MULTIVIEW, false); + specular_merge.shader.set_variant_enabled(SPECULAR_MERGE_ADDITIVE_ADD_MULTIVIEW, false); + specular_merge.shader.set_variant_enabled(SPECULAR_MERGE_ADDITIVE_SSR_MULTIVIEW, false); + } + + specular_merge.shader_version = specular_merge.shader.version_create(); + + //use additive + + RD::PipelineColorBlendState::Attachment ba; + ba.enable_blend = true; + ba.src_color_blend_factor = RD::BLEND_FACTOR_ONE; + ba.dst_color_blend_factor = RD::BLEND_FACTOR_ONE; + ba.src_alpha_blend_factor = RD::BLEND_FACTOR_ONE; + ba.dst_alpha_blend_factor = RD::BLEND_FACTOR_ONE; + ba.color_blend_op = RD::BLEND_OP_ADD; + ba.alpha_blend_op = RD::BLEND_OP_ADD; + + RD::PipelineColorBlendState blend_additive; + blend_additive.attachments.push_back(ba); + + for (int i = 0; i < SPECULAR_MERGE_MAX; i++) { + if (specular_merge.shader.is_variant_enabled(i)) { + RD::PipelineColorBlendState blend_state; + if (i == SPECULAR_MERGE_ADDITIVE_ADD || i == SPECULAR_MERGE_ADDITIVE_SSR || i == SPECULAR_MERGE_ADDITIVE_ADD_MULTIVIEW || i == SPECULAR_MERGE_ADDITIVE_SSR_MULTIVIEW) { + blend_state = blend_additive; + } else { + blend_state = RD::PipelineColorBlendState::create_disabled(); + } + specular_merge.pipelines[i].setup(specular_merge.shader.version_get_shader(specular_merge.shader_version, i), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), blend_state, 0); + } + } + } +} + +CopyEffects::~CopyEffects() { + if (prefer_raster_effects) { + blur_raster.shader.version_free(blur_raster.shader_version); + cubemap_downsampler.raster_shader.version_free(cubemap_downsampler.shader_version); + filter.raster_shader.version_free(filter.shader_version); + roughness.raster_shader.version_free(roughness.shader_version); + } else { + copy.shader.version_free(copy.shader_version); + cubemap_downsampler.compute_shader.version_free(cubemap_downsampler.shader_version); + filter.compute_shader.version_free(filter.shader_version); + roughness.compute_shader.version_free(roughness.shader_version); + } + + specular_merge.shader.version_free(specular_merge.shader_version); + + RD::get_singleton()->free(filter.coefficient_buffer); + + if (RD::get_singleton()->uniform_set_is_valid(filter.image_uniform_set)) { + RD::get_singleton()->free(filter.image_uniform_set); + } + + if (RD::get_singleton()->uniform_set_is_valid(filter.uniform_set)) { + RD::get_singleton()->free(filter.uniform_set); + } + + copy_to_fb.shader.version_free(copy_to_fb.shader_version); + cube_to_dp.shader.version_free(cube_to_dp.shader_version); + + singleton = nullptr; +} + +void CopyEffects::copy_to_rect(RID p_source_rd_texture, RID p_dest_texture, const Rect2i &p_rect, bool p_flip_y, bool p_force_luminance, bool p_all_source, bool p_8_bit_dst, bool p_alpha_to_one) { + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use the compute version of the copy_to_rect shader with the mobile renderer."); + + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + memset(©.push_constant, 0, sizeof(CopyPushConstant)); + if (p_flip_y) { + copy.push_constant.flags |= COPY_FLAG_FLIP_Y; + } + + if (p_force_luminance) { + copy.push_constant.flags |= COPY_FLAG_FORCE_LUMINANCE; + } + + if (p_all_source) { + copy.push_constant.flags |= COPY_FLAG_ALL_SOURCE; + } + + if (p_alpha_to_one) { + copy.push_constant.flags |= COPY_FLAG_ALPHA_TO_ONE; + } + + copy.push_constant.section[0] = p_rect.position.x; + copy.push_constant.section[1] = p_rect.position.y; + copy.push_constant.section[2] = p_rect.size.width; + copy.push_constant.section[3] = p_rect.size.height; + copy.push_constant.target[0] = p_rect.position.x; + copy.push_constant.target[1] = p_rect.position.y; + + // setup our uniforms + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_source_rd_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_source_rd_texture })); + RD::Uniform u_dest_texture(RD::UNIFORM_TYPE_IMAGE, 0, p_dest_texture); + + CopyMode mode = p_8_bit_dst ? COPY_MODE_SIMPLY_COPY_8BIT : COPY_MODE_SIMPLY_COPY; + RID shader = copy.shader.version_get_shader(copy.shader_version, mode); + ERR_FAIL_COND(shader.is_null()); + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[mode]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_source_rd_texture), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 3, u_dest_texture), 3); + RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_rect.size.width, p_rect.size.height, 1); + RD::get_singleton()->compute_list_end(); +} + +void CopyEffects::copy_cubemap_to_panorama(RID p_source_cube, RID p_dest_panorama, const Size2i &p_panorama_size, float p_lod, bool p_is_array) { + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use the compute version of the copy_cubemap_to_panorama shader with the mobile renderer."); + + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + memset(©.push_constant, 0, sizeof(CopyPushConstant)); + + copy.push_constant.section[0] = 0; + copy.push_constant.section[1] = 0; + copy.push_constant.section[2] = p_panorama_size.width; + copy.push_constant.section[3] = p_panorama_size.height; + copy.push_constant.target[0] = 0; + copy.push_constant.target[1] = 0; + copy.push_constant.camera_z_far = p_lod; + + // setup our uniforms + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_source_cube(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_source_cube })); + RD::Uniform u_dest_panorama(RD::UNIFORM_TYPE_IMAGE, 0, p_dest_panorama); + + CopyMode mode = p_is_array ? COPY_MODE_CUBE_ARRAY_TO_PANORAMA : COPY_MODE_CUBE_TO_PANORAMA; + RID shader = copy.shader.version_get_shader(copy.shader_version, mode); + ERR_FAIL_COND(shader.is_null()); + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[mode]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_source_cube), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 3, u_dest_panorama), 3); + RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_panorama_size.width, p_panorama_size.height, 1); + RD::get_singleton()->compute_list_end(); +} + +void CopyEffects::copy_depth_to_rect(RID p_source_rd_texture, RID p_dest_texture, const Rect2i &p_rect, bool p_flip_y) { + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use the compute version of the copy_depth_to_rect shader with the mobile renderer."); + + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + memset(©.push_constant, 0, sizeof(CopyPushConstant)); + if (p_flip_y) { + copy.push_constant.flags |= COPY_FLAG_FLIP_Y; + } + + copy.push_constant.section[0] = 0; + copy.push_constant.section[1] = 0; + copy.push_constant.section[2] = p_rect.size.width; + copy.push_constant.section[3] = p_rect.size.height; + copy.push_constant.target[0] = p_rect.position.x; + copy.push_constant.target[1] = p_rect.position.y; + + // setup our uniforms + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_source_rd_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_source_rd_texture })); + RD::Uniform u_dest_texture(RD::UNIFORM_TYPE_IMAGE, 0, p_dest_texture); + + CopyMode mode = COPY_MODE_SIMPLY_COPY_DEPTH; + RID shader = copy.shader.version_get_shader(copy.shader_version, mode); + ERR_FAIL_COND(shader.is_null()); + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[mode]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_source_rd_texture), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 3, u_dest_texture), 3); + RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_rect.size.width, p_rect.size.height, 1); + RD::get_singleton()->compute_list_end(); +} + +void CopyEffects::copy_depth_to_rect_and_linearize(RID p_source_rd_texture, RID p_dest_texture, const Rect2i &p_rect, bool p_flip_y, float p_z_near, float p_z_far) { + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use the compute version of the copy_depth_to_rect_and_linearize shader with the mobile renderer."); + + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + memset(©.push_constant, 0, sizeof(CopyPushConstant)); + if (p_flip_y) { + copy.push_constant.flags |= COPY_FLAG_FLIP_Y; + } + + copy.push_constant.section[0] = 0; + copy.push_constant.section[1] = 0; + copy.push_constant.section[2] = p_rect.size.width; + copy.push_constant.section[3] = p_rect.size.height; + copy.push_constant.target[0] = p_rect.position.x; + copy.push_constant.target[1] = p_rect.position.y; + copy.push_constant.camera_z_far = p_z_far; + copy.push_constant.camera_z_near = p_z_near; + + // setup our uniforms + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_source_rd_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_source_rd_texture })); + RD::Uniform u_dest_texture(RD::UNIFORM_TYPE_IMAGE, 0, p_dest_texture); + + CopyMode mode = COPY_MODE_LINEARIZE_DEPTH; + RID shader = copy.shader.version_get_shader(copy.shader_version, mode); + ERR_FAIL_COND(shader.is_null()); + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[mode]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_source_rd_texture), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 3, u_dest_texture), 3); + RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_rect.size.width, p_rect.size.height, 1); + RD::get_singleton()->compute_list_end(); +} + +void CopyEffects::copy_to_atlas_fb(RID p_source_rd_texture, RID p_dest_framebuffer, const Rect2 &p_uv_rect, RD::DrawListID p_draw_list, bool p_flip_y, bool p_panorama) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + memset(©_to_fb.push_constant, 0, sizeof(CopyToFbPushConstant)); + + copy_to_fb.push_constant.use_section = true; + copy_to_fb.push_constant.section[0] = p_uv_rect.position.x; + copy_to_fb.push_constant.section[1] = p_uv_rect.position.y; + copy_to_fb.push_constant.section[2] = p_uv_rect.size.x; + copy_to_fb.push_constant.section[3] = p_uv_rect.size.y; + + if (p_flip_y) { + copy_to_fb.push_constant.flip_y = true; + } + + // setup our uniforms + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_source_rd_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_source_rd_texture })); + + CopyToFBMode mode = p_panorama ? COPY_TO_FB_COPY_PANORAMA_TO_DP : COPY_TO_FB_COPY; + RID shader = copy_to_fb.shader.version_get_shader(copy_to_fb.shader_version, mode); + ERR_FAIL_COND(shader.is_null()); + + RD::DrawListID draw_list = p_draw_list; + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, copy_to_fb.pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dest_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 0, u_source_rd_texture), 0); + RD::get_singleton()->draw_list_bind_index_array(draw_list, material_storage->get_quad_index_array()); + RD::get_singleton()->draw_list_set_push_constant(draw_list, ©_to_fb.push_constant, sizeof(CopyToFbPushConstant)); + RD::get_singleton()->draw_list_draw(draw_list, true); +} + +void CopyEffects::copy_to_fb_rect(RID p_source_rd_texture, RID p_dest_framebuffer, const Rect2i &p_rect, bool p_flip_y, bool p_force_luminance, bool p_alpha_to_zero, bool p_srgb, RID p_secondary, bool p_multiview) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + memset(©_to_fb.push_constant, 0, sizeof(CopyToFbPushConstant)); + + if (p_flip_y) { + copy_to_fb.push_constant.flip_y = true; + } + if (p_force_luminance) { + copy_to_fb.push_constant.force_luminance = true; + } + if (p_alpha_to_zero) { + copy_to_fb.push_constant.alpha_to_zero = true; + } + if (p_srgb) { + copy_to_fb.push_constant.srgb = true; + } + + // setup our uniforms + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_source_rd_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_source_rd_texture })); + + CopyToFBMode mode; + if (p_multiview) { + mode = p_secondary.is_valid() ? COPY_TO_FB_MULTIVIEW_WITH_DEPTH : COPY_TO_FB_MULTIVIEW; + } else { + mode = p_secondary.is_valid() ? COPY_TO_FB_COPY2 : COPY_TO_FB_COPY; + } + + RID shader = copy_to_fb.shader.version_get_shader(copy_to_fb.shader_version, mode); + ERR_FAIL_COND(shader.is_null()); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD, Vector<Color>(), 1.0, 0, p_rect); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, copy_to_fb.pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dest_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 0, u_source_rd_texture), 0); + if (p_secondary.is_valid()) { + // TODO may need to do this differently when reading from depth buffer for multiview + RD::Uniform u_secondary(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_secondary })); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 1, u_secondary), 1); + } + RD::get_singleton()->draw_list_bind_index_array(draw_list, material_storage->get_quad_index_array()); + RD::get_singleton()->draw_list_set_push_constant(draw_list, ©_to_fb.push_constant, sizeof(CopyToFbPushConstant)); + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); +} + +void CopyEffects::copy_raster(RID p_source_texture, RID p_dest_framebuffer) { + ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use the raster version of the copy with the clustered renderer."); + + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + memset(&blur_raster.push_constant, 0, sizeof(BlurRasterPushConstant)); + + // setup our uniforms + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_source_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_source_texture })); + + RID shader = blur_raster.shader.version_get_shader(blur_raster.shader_version, BLUR_MODE_COPY); + ERR_FAIL_COND(shader.is_null()); + + // Just copy it back (we use our blur raster shader here).. + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, blur_raster.pipelines[BLUR_MODE_COPY].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dest_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 0, u_source_texture), 0); + RD::get_singleton()->draw_list_bind_index_array(draw_list, material_storage->get_quad_index_array()); + + memset(&blur_raster.push_constant, 0, sizeof(BlurRasterPushConstant)); + RD::get_singleton()->draw_list_set_push_constant(draw_list, &blur_raster.push_constant, sizeof(BlurRasterPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); +} + +void CopyEffects::gaussian_blur(RID p_source_rd_texture, RID p_texture, const Rect2i &p_region, bool p_8bit_dst) { + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use the compute version of the gaussian blur with the mobile renderer."); + + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + memset(©.push_constant, 0, sizeof(CopyPushConstant)); + + copy.push_constant.section[0] = p_region.position.x; + copy.push_constant.section[1] = p_region.position.y; + copy.push_constant.section[2] = p_region.size.width; + copy.push_constant.section[3] = p_region.size.height; + + // setup our uniforms + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_source_rd_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_source_rd_texture })); + RD::Uniform u_texture(RD::UNIFORM_TYPE_IMAGE, 0, p_texture); + + CopyMode mode = p_8bit_dst ? COPY_MODE_GAUSSIAN_COPY_8BIT : COPY_MODE_GAUSSIAN_COPY; + RID shader = copy.shader.version_get_shader(copy.shader_version, mode); + ERR_FAIL_COND(shader.is_null()); + + //HORIZONTAL + RD::DrawListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[mode]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_source_rd_texture), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 3, u_texture), 3); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_region.size.width, p_region.size.height, 1); + + RD::get_singleton()->compute_list_end(); +} + +void CopyEffects::gaussian_glow(RID p_source_rd_texture, RID p_back_texture, const Size2i &p_size, float p_strength, bool p_high_quality, bool p_first_pass, float p_luminance_cap, float p_exposure, float p_bloom, float p_hdr_bleed_threshold, float p_hdr_bleed_scale, RID p_auto_exposure, float p_auto_exposure_scale) { + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use the compute version of the gaussian glow with the mobile renderer."); + + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + memset(©.push_constant, 0, sizeof(CopyPushConstant)); + + CopyMode copy_mode = p_first_pass && p_auto_exposure.is_valid() ? COPY_MODE_GAUSSIAN_GLOW_AUTO_EXPOSURE : COPY_MODE_GAUSSIAN_GLOW; + uint32_t base_flags = 0; + + copy.push_constant.section[2] = p_size.x; + copy.push_constant.section[3] = p_size.y; + + copy.push_constant.glow_strength = p_strength; + copy.push_constant.glow_bloom = p_bloom; + copy.push_constant.glow_hdr_threshold = p_hdr_bleed_threshold; + copy.push_constant.glow_hdr_scale = p_hdr_bleed_scale; + copy.push_constant.glow_exposure = p_exposure; + copy.push_constant.glow_white = 0; //actually unused + copy.push_constant.glow_luminance_cap = p_luminance_cap; + + copy.push_constant.glow_auto_exposure_scale = p_auto_exposure_scale; //unused also + + // setup our uniforms + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_source_rd_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_source_rd_texture })); + RD::Uniform u_back_texture(RD::UNIFORM_TYPE_IMAGE, 0, p_back_texture); + + RID shader = copy.shader.version_get_shader(copy.shader_version, copy_mode); + ERR_FAIL_COND(shader.is_null()); + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[copy_mode]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_source_rd_texture), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 3, u_back_texture), 3); + if (p_auto_exposure.is_valid() && p_first_pass) { + RD::Uniform u_auto_exposure(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_auto_exposure })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_auto_exposure), 1); + } + + copy.push_constant.flags = base_flags | (p_first_pass ? COPY_FLAG_GLOW_FIRST_PASS : 0) | (p_high_quality ? COPY_FLAG_HIGH_QUALITY_GLOW : 0); + RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_size.width, p_size.height, 1); + RD::get_singleton()->compute_list_end(); +} + +void CopyEffects::gaussian_glow_raster(RID p_source_rd_texture, RID p_half_texture, RID p_dest_texture, float p_luminance_multiplier, const Size2i &p_size, float p_strength, bool p_high_quality, bool p_first_pass, float p_luminance_cap, float p_exposure, float p_bloom, float p_hdr_bleed_threshold, float p_hdr_bleed_scale, RID p_auto_exposure, float p_auto_exposure_scale) { + ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use the raster version of the gaussian glow with the clustered renderer."); + + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + RID half_framebuffer = FramebufferCacheRD::get_singleton()->get_cache(p_half_texture); + RID dest_framebuffer = FramebufferCacheRD::get_singleton()->get_cache(p_dest_texture); + + memset(&blur_raster.push_constant, 0, sizeof(BlurRasterPushConstant)); + + BlurRasterMode blur_mode = p_first_pass && p_auto_exposure.is_valid() ? BLUR_MODE_GAUSSIAN_GLOW_AUTO_EXPOSURE : BLUR_MODE_GAUSSIAN_GLOW; + uint32_t base_flags = 0; + + blur_raster.push_constant.pixel_size[0] = 1.0 / float(p_size.x); + blur_raster.push_constant.pixel_size[1] = 1.0 / float(p_size.y); + + blur_raster.push_constant.glow_strength = p_strength; + blur_raster.push_constant.glow_bloom = p_bloom; + blur_raster.push_constant.glow_hdr_threshold = p_hdr_bleed_threshold; + blur_raster.push_constant.glow_hdr_scale = p_hdr_bleed_scale; + blur_raster.push_constant.glow_exposure = p_exposure; + blur_raster.push_constant.glow_white = 0; //actually unused + blur_raster.push_constant.glow_luminance_cap = p_luminance_cap; + + blur_raster.push_constant.glow_auto_exposure_scale = p_auto_exposure_scale; //unused also + + blur_raster.push_constant.luminance_multiplier = p_luminance_multiplier; + + // setup our uniforms + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_source_rd_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_source_rd_texture })); + RD::Uniform u_half_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_half_texture })); + + RID shader = blur_raster.shader.version_get_shader(blur_raster.shader_version, blur_mode); + ERR_FAIL_COND(shader.is_null()); + + //HORIZONTAL + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(half_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, blur_raster.pipelines[blur_mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(half_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 0, u_source_rd_texture), 0); + if (p_auto_exposure.is_valid() && p_first_pass) { + RD::Uniform u_auto_exposure(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_auto_exposure })); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 1, u_auto_exposure), 1); + } + RD::get_singleton()->draw_list_bind_index_array(draw_list, material_storage->get_quad_index_array()); + + blur_raster.push_constant.flags = base_flags | BLUR_FLAG_HORIZONTAL | (p_first_pass ? BLUR_FLAG_GLOW_FIRST_PASS : 0); + RD::get_singleton()->draw_list_set_push_constant(draw_list, &blur_raster.push_constant, sizeof(BlurRasterPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + + blur_mode = BLUR_MODE_GAUSSIAN_GLOW; + + shader = blur_raster.shader.version_get_shader(blur_raster.shader_version, blur_mode); + ERR_FAIL_COND(shader.is_null()); + + //VERTICAL + draw_list = RD::get_singleton()->draw_list_begin(dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, blur_raster.pipelines[blur_mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(dest_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 0, u_half_texture), 0); + RD::get_singleton()->draw_list_bind_index_array(draw_list, material_storage->get_quad_index_array()); + + blur_raster.push_constant.flags = base_flags; + RD::get_singleton()->draw_list_set_push_constant(draw_list, &blur_raster.push_constant, sizeof(BlurRasterPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); +} + +void CopyEffects::make_mipmap(RID p_source_rd_texture, RID p_dest_texture, const Size2i &p_size) { + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use the compute version of the make_mipmap shader with the mobile renderer."); + + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + memset(©.push_constant, 0, sizeof(CopyPushConstant)); + + copy.push_constant.section[0] = 0; + copy.push_constant.section[1] = 0; + copy.push_constant.section[2] = p_size.width; + copy.push_constant.section[3] = p_size.height; + + // setup our uniforms + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_source_rd_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_source_rd_texture })); + RD::Uniform u_dest_texture(RD::UNIFORM_TYPE_IMAGE, 0, p_dest_texture); + + CopyMode mode = COPY_MODE_MIPMAP; + RID shader = copy.shader.version_get_shader(copy.shader_version, mode); + ERR_FAIL_COND(shader.is_null()); + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[mode]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_source_rd_texture), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 3, u_dest_texture), 3); + RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_size.width, p_size.height, 1); + RD::get_singleton()->compute_list_end(); +} + +void CopyEffects::make_mipmap_raster(RID p_source_rd_texture, RID p_dest_texture, const Size2i &p_size) { + ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use the raster version of mipmap with the clustered renderer."); + + RID dest_framebuffer = FramebufferCacheRD::get_singleton()->get_cache(p_dest_texture); + + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + memset(&blur_raster.push_constant, 0, sizeof(BlurRasterPushConstant)); + + BlurRasterMode mode = BLUR_MIPMAP; + + blur_raster.push_constant.pixel_size[0] = 1.0 / float(p_size.x); + blur_raster.push_constant.pixel_size[1] = 1.0 / float(p_size.y); + + // setup our uniforms + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_source_rd_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_source_rd_texture })); + + RID shader = blur_raster.shader.version_get_shader(blur_raster.shader_version, mode); + ERR_FAIL_COND(shader.is_null()); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, blur_raster.pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(dest_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 0, u_source_rd_texture), 0); + RD::get_singleton()->draw_list_bind_index_array(draw_list, material_storage->get_quad_index_array()); + RD::get_singleton()->draw_list_set_push_constant(draw_list, &blur_raster.push_constant, sizeof(BlurRasterPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); +} + +void CopyEffects::set_color(RID p_dest_texture, const Color &p_color, const Rect2i &p_region, bool p_8bit_dst) { + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use the compute version of the set_color shader with the mobile renderer."); + + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + + memset(©.push_constant, 0, sizeof(CopyPushConstant)); + + copy.push_constant.section[0] = 0; + copy.push_constant.section[1] = 0; + copy.push_constant.section[2] = p_region.size.width; + copy.push_constant.section[3] = p_region.size.height; + copy.push_constant.target[0] = p_region.position.x; + copy.push_constant.target[1] = p_region.position.y; + copy.push_constant.set_color[0] = p_color.r; + copy.push_constant.set_color[1] = p_color.g; + copy.push_constant.set_color[2] = p_color.b; + copy.push_constant.set_color[3] = p_color.a; + + // setup our uniforms + RD::Uniform u_dest_texture(RD::UNIFORM_TYPE_IMAGE, 0, p_dest_texture); + + CopyMode mode = p_8bit_dst ? COPY_MODE_SET_COLOR_8BIT : COPY_MODE_SET_COLOR; + RID shader = copy.shader.version_get_shader(copy.shader_version, mode); + ERR_FAIL_COND(shader.is_null()); + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[mode]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 3, u_dest_texture), 3); + RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_region.size.width, p_region.size.height, 1); + RD::get_singleton()->compute_list_end(); +} + +void CopyEffects::copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dst_framebuffer, const Rect2 &p_rect, const Vector2 &p_dst_size, float p_z_near, float p_z_far, bool p_dp_flip) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + CopyToDPPushConstant push_constant; + push_constant.screen_rect[0] = p_rect.position.x; + push_constant.screen_rect[1] = p_rect.position.y; + push_constant.screen_rect[2] = p_rect.size.width; + push_constant.screen_rect[3] = p_rect.size.height; + push_constant.z_far = p_z_far; + push_constant.z_near = p_z_near; + push_constant.texel_size[0] = 1.0f / p_dst_size.x; + push_constant.texel_size[1] = 1.0f / p_dst_size.y; + push_constant.texel_size[0] *= p_dp_flip ? -1.0f : 1.0f; // Encode dp flip as x size sign + + // setup our uniforms + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_source_rd_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_source_rd_texture })); + + RID shader = cube_to_dp.shader.version_get_shader(cube_to_dp.shader_version, 0); + ERR_FAIL_COND(shader.is_null()); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dst_framebuffer, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, cube_to_dp.pipeline.get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dst_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 0, u_source_rd_texture), 0); + RD::get_singleton()->draw_list_bind_index_array(draw_list, material_storage->get_quad_index_array()); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &push_constant, sizeof(CopyToDPPushConstant)); + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_TRANSFER); +} + +void CopyEffects::cubemap_downsample(RID p_source_cubemap, RID p_dest_cubemap, const Size2i &p_size) { + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use compute based cubemap downsample with the mobile renderer."); + + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + cubemap_downsampler.push_constant.face_size = p_size.x; + cubemap_downsampler.push_constant.face_id = 0; // we render all 6 sides to each layer in one call + + // setup our uniforms + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_source_cubemap(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_source_cubemap })); + RD::Uniform u_dest_cubemap(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_dest_cubemap })); + + RID shader = cubemap_downsampler.compute_shader.version_get_shader(cubemap_downsampler.shader_version, 0); + ERR_FAIL_COND(shader.is_null()); + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, cubemap_downsampler.compute_pipeline); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_source_cubemap), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_dest_cubemap), 1); + + int x_groups = (p_size.x - 1) / 8 + 1; + int y_groups = (p_size.y - 1) / 8 + 1; + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &cubemap_downsampler.push_constant, sizeof(CubemapDownsamplerPushConstant)); + + RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 6); // one z_group for each face + + RD::get_singleton()->compute_list_end(); +} + +void CopyEffects::cubemap_downsample_raster(RID p_source_cubemap, RID p_dest_framebuffer, uint32_t p_face_id, const Size2i &p_size) { + ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use raster based cubemap downsample with the clustered renderer."); + ERR_FAIL_COND_MSG(p_face_id >= 6, "Raster implementation of cubemap downsample must process one side at a time."); + + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + cubemap_downsampler.push_constant.face_size = p_size.x; + cubemap_downsampler.push_constant.face_id = p_face_id; + + // setup our uniforms + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_source_cubemap(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_source_cubemap })); + + RID shader = cubemap_downsampler.raster_shader.version_get_shader(cubemap_downsampler.shader_version, 0); + ERR_FAIL_COND(shader.is_null()); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, cubemap_downsampler.raster_pipeline.get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dest_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 0, u_source_cubemap), 0); + RD::get_singleton()->draw_list_bind_index_array(draw_list, material_storage->get_quad_index_array()); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &cubemap_downsampler.push_constant, sizeof(CubemapDownsamplerPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); +} + +void CopyEffects::cubemap_filter(RID p_source_cubemap, Vector<RID> p_dest_cubemap, bool p_use_array) { + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use compute based cubemap filter with the mobile renderer."); + + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + Vector<RD::Uniform> uniforms; + for (int i = 0; i < p_dest_cubemap.size(); i++) { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = i; + u.append_id(p_dest_cubemap[i]); + uniforms.push_back(u); + } + if (RD::get_singleton()->uniform_set_is_valid(filter.image_uniform_set)) { + RD::get_singleton()->free(filter.image_uniform_set); + } + filter.image_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, filter.compute_shader.version_get_shader(filter.shader_version, 0), 2); + + // setup our uniforms + RID default_mipmap_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_source_cubemap(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_mipmap_sampler, p_source_cubemap })); + + int mode = p_use_array ? FILTER_MODE_HIGH_QUALITY_ARRAY : FILTER_MODE_HIGH_QUALITY; + mode = filter.use_high_quality ? mode : mode + 1; + + RID shader = filter.compute_shader.version_get_shader(filter.shader_version, mode); + ERR_FAIL_COND(shader.is_null()); + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, filter.compute_pipelines[mode]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_source_cubemap), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, filter.uniform_set, 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, filter.image_uniform_set, 2); + + int x_groups = p_use_array ? 1792 : 342; // (128 * 128 * 7) / 64 : (128*128 + 64*64 + 32*32 + 16*16 + 8*8 + 4*4 + 2*2) / 64 + + RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, 6, 1); // one y_group for each face + + RD::get_singleton()->compute_list_end(); +} + +void CopyEffects::cubemap_filter_raster(RID p_source_cubemap, RID p_dest_framebuffer, uint32_t p_face_id, uint32_t p_mip_level) { + ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use raster based cubemap filter with the clustered renderer."); + ERR_FAIL_COND_MSG(p_face_id >= 6, "Raster implementation of cubemap filter must process one side at a time."); + + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + // TODO implement! + CubemapFilterRasterPushConstant push_constant; + push_constant.mip_level = p_mip_level; + push_constant.face_id = p_face_id; + + // setup our uniforms + RID default_mipmap_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_source_cubemap(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_mipmap_sampler, p_source_cubemap })); + + CubemapFilterMode mode = filter.use_high_quality ? FILTER_MODE_HIGH_QUALITY : FILTER_MODE_LOW_QUALITY; + + RID shader = filter.raster_shader.version_get_shader(filter.shader_version, mode); + ERR_FAIL_COND(shader.is_null()); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, filter.raster_pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dest_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 0, u_source_cubemap), 0); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, filter.uniform_set, 1); + RD::get_singleton()->draw_list_bind_index_array(draw_list, material_storage->get_quad_index_array()); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &push_constant, sizeof(CubemapFilterRasterPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); +} + +void CopyEffects::cubemap_roughness(RID p_source_rd_texture, RID p_dest_texture, uint32_t p_face_id, uint32_t p_sample_count, float p_roughness, float p_size) { + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use compute based cubemap roughness with the mobile renderer."); + + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + memset(&roughness.push_constant, 0, sizeof(CubemapRoughnessPushConstant)); + + roughness.push_constant.face_id = p_face_id > 9 ? 0 : p_face_id; + roughness.push_constant.roughness = p_roughness * p_roughness; // Shader expects roughness, not perceptual roughness, so multiply before passing in. + roughness.push_constant.sample_count = p_sample_count; + roughness.push_constant.use_direct_write = p_roughness == 0.0; + roughness.push_constant.face_size = p_size; + + // setup our uniforms + RID default_mipmap_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_source_rd_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_mipmap_sampler, p_source_rd_texture })); + RD::Uniform u_dest_texture(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_dest_texture })); + + RID shader = roughness.compute_shader.version_get_shader(roughness.shader_version, 0); + ERR_FAIL_COND(shader.is_null()); + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, roughness.compute_pipeline); + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_source_rd_texture), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_dest_texture), 1); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &roughness.push_constant, sizeof(CubemapRoughnessPushConstant)); + + int x_groups = (p_size - 1) / 8 + 1; + int y_groups = (p_size - 1) / 8 + 1; + + RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, p_face_id > 9 ? 6 : 1); + + RD::get_singleton()->compute_list_end(); +} + +void CopyEffects::cubemap_roughness_raster(RID p_source_rd_texture, RID p_dest_framebuffer, uint32_t p_face_id, uint32_t p_sample_count, float p_roughness, float p_size) { + ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use raster based cubemap roughness with the clustered renderer."); + ERR_FAIL_COND_MSG(p_face_id >= 6, "Raster implementation of cubemap roughness must process one side at a time."); + + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + memset(&roughness.push_constant, 0, sizeof(CubemapRoughnessPushConstant)); + + roughness.push_constant.face_id = p_face_id; + roughness.push_constant.roughness = p_roughness * p_roughness; // Shader expects roughness, not perceptual roughness, so multiply before passing in. + roughness.push_constant.sample_count = p_sample_count; + roughness.push_constant.use_direct_write = p_roughness == 0.0; + roughness.push_constant.face_size = p_size; + + // setup our uniforms + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_source_rd_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_source_rd_texture })); + + RID shader = roughness.raster_shader.version_get_shader(roughness.shader_version, 0); + ERR_FAIL_COND(shader.is_null()); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, roughness.raster_pipeline.get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dest_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 0, u_source_rd_texture), 0); + RD::get_singleton()->draw_list_bind_index_array(draw_list, material_storage->get_quad_index_array()); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &roughness.push_constant, sizeof(CubemapRoughnessPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); +} + +void CopyEffects::merge_specular(RID p_dest_framebuffer, RID p_specular, RID p_base, RID p_reflection, uint32_t p_view_count) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::get_singleton()->draw_command_begin_label("Merge specular"); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, Vector<Color>()); + + int mode; + if (p_reflection.is_valid()) { + if (p_base.is_valid()) { + mode = SPECULAR_MERGE_SSR; + } else { + mode = SPECULAR_MERGE_ADDITIVE_SSR; + } + } else { + if (p_base.is_valid()) { + mode = SPECULAR_MERGE_ADD; + } else { + mode = SPECULAR_MERGE_ADDITIVE_ADD; + } + } + + if (p_view_count > 1) { + mode += SPECULAR_MERGE_ADD_MULTIVIEW; + } + + RID shader = specular_merge.shader.version_get_shader(specular_merge.shader_version, mode); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, specular_merge.pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dest_framebuffer))); + + if (p_base.is_valid()) { + RD::Uniform u_base(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_base })); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 2, u_base), 2); + } + + RD::Uniform u_specular(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_specular })); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 0, u_specular), 0); + + if (p_reflection.is_valid()) { + RD::Uniform u_reflection(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_reflection })); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 1, u_reflection), 1); + } + + RD::get_singleton()->draw_list_bind_index_array(draw_list, material_storage->get_quad_index_array()); + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + + RD::get_singleton()->draw_command_end_label(); +} diff --git a/servers/rendering/renderer_rd/effects/copy_effects.h b/servers/rendering/renderer_rd/effects/copy_effects.h new file mode 100644 index 0000000000..0ddb60ebef --- /dev/null +++ b/servers/rendering/renderer_rd/effects/copy_effects.h @@ -0,0 +1,346 @@ +/*************************************************************************/ +/* copy_effects.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef COPY_EFFECTS_RD_H +#define COPY_EFFECTS_RD_H + +#include "servers/rendering/renderer_rd/pipeline_cache_rd.h" +#include "servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/copy.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/copy_to_fb.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/cube_to_dp.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/cubemap_downsampler.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/cubemap_downsampler_raster.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/cubemap_filter.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/cubemap_filter_raster.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/cubemap_roughness.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/cubemap_roughness_raster.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/specular_merge.glsl.gen.h" +#include "servers/rendering/renderer_scene_render.h" + +#include "servers/rendering_server.h" + +namespace RendererRD { + +class CopyEffects { +private: + bool prefer_raster_effects; + + // Blur raster shader + + enum BlurRasterMode { + BLUR_MIPMAP, + + BLUR_MODE_GAUSSIAN_BLUR, + BLUR_MODE_GAUSSIAN_GLOW, + BLUR_MODE_GAUSSIAN_GLOW_AUTO_EXPOSURE, + BLUR_MODE_COPY, + + BLUR_MODE_MAX + }; + + enum { + BLUR_FLAG_HORIZONTAL = (1 << 0), + BLUR_FLAG_USE_ORTHOGONAL_PROJECTION = (1 << 1), + BLUR_FLAG_GLOW_FIRST_PASS = (1 << 2), + }; + + struct BlurRasterPushConstant { + float pixel_size[2]; + uint32_t flags; + uint32_t pad; + + //glow + float glow_strength; + float glow_bloom; + float glow_hdr_threshold; + float glow_hdr_scale; + + float glow_exposure; + float glow_white; + float glow_luminance_cap; + float glow_auto_exposure_scale; + + float luminance_multiplier; + float res1; + float res2; + float res3; + }; + + struct BlurRaster { + BlurRasterPushConstant push_constant; + BlurRasterShaderRD shader; + RID shader_version; + PipelineCacheRD pipelines[BLUR_MODE_MAX]; + } blur_raster; + + // Copy shader + + enum CopyMode { + COPY_MODE_GAUSSIAN_COPY, + COPY_MODE_GAUSSIAN_COPY_8BIT, + COPY_MODE_GAUSSIAN_GLOW, + COPY_MODE_GAUSSIAN_GLOW_AUTO_EXPOSURE, + COPY_MODE_SIMPLY_COPY, + COPY_MODE_SIMPLY_COPY_8BIT, + COPY_MODE_SIMPLY_COPY_DEPTH, + COPY_MODE_SET_COLOR, + COPY_MODE_SET_COLOR_8BIT, + COPY_MODE_MIPMAP, + COPY_MODE_LINEARIZE_DEPTH, + COPY_MODE_CUBE_TO_PANORAMA, + COPY_MODE_CUBE_ARRAY_TO_PANORAMA, + COPY_MODE_MAX, + + }; + + enum { + COPY_FLAG_HORIZONTAL = (1 << 0), + COPY_FLAG_USE_COPY_SECTION = (1 << 1), + COPY_FLAG_USE_ORTHOGONAL_PROJECTION = (1 << 2), + COPY_FLAG_DOF_NEAR_FIRST_TAP = (1 << 3), + COPY_FLAG_GLOW_FIRST_PASS = (1 << 4), + COPY_FLAG_FLIP_Y = (1 << 5), + COPY_FLAG_FORCE_LUMINANCE = (1 << 6), + COPY_FLAG_ALL_SOURCE = (1 << 7), + COPY_FLAG_HIGH_QUALITY_GLOW = (1 << 8), + COPY_FLAG_ALPHA_TO_ONE = (1 << 9), + }; + + struct CopyPushConstant { + int32_t section[4]; + int32_t target[2]; + uint32_t flags; + uint32_t pad; + // Glow. + float glow_strength; + float glow_bloom; + float glow_hdr_threshold; + float glow_hdr_scale; + + float glow_exposure; + float glow_white; + float glow_luminance_cap; + float glow_auto_exposure_scale; + // DOF. + float camera_z_far; + float camera_z_near; + uint32_t pad2[2]; + //SET color + float set_color[4]; + }; + + struct Copy { + CopyPushConstant push_constant; + CopyShaderRD shader; + RID shader_version; + RID pipelines[COPY_MODE_MAX]; + + } copy; + + // Copy to FB shader + + enum CopyToFBMode { + COPY_TO_FB_COPY, + COPY_TO_FB_COPY_PANORAMA_TO_DP, + COPY_TO_FB_COPY2, + + COPY_TO_FB_MULTIVIEW, + COPY_TO_FB_MULTIVIEW_WITH_DEPTH, + COPY_TO_FB_MAX, + }; + + struct CopyToFbPushConstant { + float section[4]; + float pixel_size[2]; + uint32_t flip_y; + uint32_t use_section; + + uint32_t force_luminance; + uint32_t alpha_to_zero; + uint32_t srgb; + uint32_t pad; + }; + + struct CopyToFb { + CopyToFbPushConstant push_constant; + CopyToFbShaderRD shader; + RID shader_version; + PipelineCacheRD pipelines[COPY_TO_FB_MAX]; + + } copy_to_fb; + + // Copy to DP + + struct CopyToDPPushConstant { + float z_far; + float z_near; + float texel_size[2]; + float screen_rect[4]; + }; + + struct CopyToDP { + CubeToDpShaderRD shader; + RID shader_version; + PipelineCacheRD pipeline; + } cube_to_dp; + + // Cubemap effects + + struct CubemapDownsamplerPushConstant { + uint32_t face_size; + uint32_t face_id; + float pad[2]; + }; + + struct CubemapDownsampler { + CubemapDownsamplerPushConstant push_constant; + CubemapDownsamplerShaderRD compute_shader; + CubemapDownsamplerRasterShaderRD raster_shader; + RID shader_version; + RID compute_pipeline; + PipelineCacheRD raster_pipeline; + } cubemap_downsampler; + + enum CubemapFilterMode { + FILTER_MODE_HIGH_QUALITY, + FILTER_MODE_LOW_QUALITY, + FILTER_MODE_HIGH_QUALITY_ARRAY, + FILTER_MODE_LOW_QUALITY_ARRAY, + FILTER_MODE_MAX, + }; + + struct CubemapFilterRasterPushConstant { + uint32_t mip_level; + uint32_t face_id; + float pad[2]; + }; + + struct CubemapFilter { + CubemapFilterShaderRD compute_shader; + CubemapFilterRasterShaderRD raster_shader; + RID shader_version; + RID compute_pipelines[FILTER_MODE_MAX]; + PipelineCacheRD raster_pipelines[FILTER_MODE_MAX]; + + RID uniform_set; + RID image_uniform_set; + RID coefficient_buffer; + bool use_high_quality; + + } filter; + + struct CubemapRoughnessPushConstant { + uint32_t face_id; + uint32_t sample_count; + float roughness; + uint32_t use_direct_write; + float face_size; + float pad[3]; + }; + + struct CubemapRoughness { + CubemapRoughnessPushConstant push_constant; + CubemapRoughnessShaderRD compute_shader; + CubemapRoughnessRasterShaderRD raster_shader; + RID shader_version; + RID compute_pipeline; + PipelineCacheRD raster_pipeline; + } roughness; + + // Merge specular + + enum SpecularMergeMode { + SPECULAR_MERGE_ADD, + SPECULAR_MERGE_SSR, + SPECULAR_MERGE_ADDITIVE_ADD, + SPECULAR_MERGE_ADDITIVE_SSR, + + SPECULAR_MERGE_ADD_MULTIVIEW, + SPECULAR_MERGE_SSR_MULTIVIEW, + SPECULAR_MERGE_ADDITIVE_ADD_MULTIVIEW, + SPECULAR_MERGE_ADDITIVE_SSR_MULTIVIEW, + + SPECULAR_MERGE_MAX + }; + + /* Specular merge must be done using raster, rather than compute + * because it must continue the existing color buffer + */ + + struct SpecularMerge { + SpecularMergeShaderRD shader; + RID shader_version; + PipelineCacheRD pipelines[SPECULAR_MERGE_MAX]; + + } specular_merge; + + static CopyEffects *singleton; + +public: + static CopyEffects *get_singleton(); + + CopyEffects(bool p_prefer_raster_effects); + ~CopyEffects(); + + bool get_prefer_raster_effects() { return prefer_raster_effects; } + + void copy_to_rect(RID p_source_rd_texture, RID p_dest_texture, const Rect2i &p_rect, bool p_flip_y = false, bool p_force_luminance = false, bool p_all_source = false, bool p_8_bit_dst = false, bool p_alpha_to_one = false); + void copy_cubemap_to_panorama(RID p_source_cube, RID p_dest_panorama, const Size2i &p_panorama_size, float p_lod, bool p_is_array); + void copy_depth_to_rect(RID p_source_rd_texture, RID p_dest_framebuffer, const Rect2i &p_rect, bool p_flip_y = false); + void copy_depth_to_rect_and_linearize(RID p_source_rd_texture, RID p_dest_texture, const Rect2i &p_rect, bool p_flip_y, float p_z_near, float p_z_far); + void copy_to_fb_rect(RID p_source_rd_texture, RID p_dest_framebuffer, const Rect2i &p_rect, bool p_flip_y = false, bool p_force_luminance = false, bool p_alpha_to_zero = false, bool p_srgb = false, RID p_secondary = RID(), bool p_multiview = false); + void copy_to_atlas_fb(RID p_source_rd_texture, RID p_dest_framebuffer, const Rect2 &p_uv_rect, RD::DrawListID p_draw_list, bool p_flip_y = false, bool p_panorama = false); + void copy_raster(RID p_source_texture, RID p_dest_framebuffer); + + void gaussian_blur(RID p_source_rd_texture, RID p_texture, const Rect2i &p_region, bool p_8bit_dst = false); + void gaussian_glow(RID p_source_rd_texture, RID p_back_texture, const Size2i &p_size, float p_strength = 1.0, bool p_high_quality = false, bool p_first_pass = false, float p_luminance_cap = 16.0, float p_exposure = 1.0, float p_bloom = 0.0, float p_hdr_bleed_threshold = 1.0, float p_hdr_bleed_scale = 1.0, RID p_auto_exposure = RID(), float p_auto_exposure_scale = 1.0); + void gaussian_glow_raster(RID p_source_rd_texture, RID p_half_texture, RID p_dest_texture, float p_luminance_multiplier, const Size2i &p_size, float p_strength = 1.0, bool p_high_quality = false, bool p_first_pass = false, float p_luminance_cap = 16.0, float p_exposure = 1.0, float p_bloom = 0.0, float p_hdr_bleed_threshold = 1.0, float p_hdr_bleed_scale = 1.0, RID p_auto_exposure = RID(), float p_auto_exposure_scale = 1.0); + + void make_mipmap(RID p_source_rd_texture, RID p_dest_texture, const Size2i &p_size); + void make_mipmap_raster(RID p_source_rd_texture, RID p_dest_texture, const Size2i &p_size); + + void set_color(RID p_dest_texture, const Color &p_color, const Rect2i &p_region, bool p_8bit_dst = false); + + void copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dst_framebuffer, const Rect2 &p_rect, const Vector2 &p_dst_size, float p_z_near, float p_z_far, bool p_dp_flip); + void cubemap_downsample(RID p_source_cubemap, RID p_dest_cubemap, const Size2i &p_size); + void cubemap_downsample_raster(RID p_source_cubemap, RID p_dest_framebuffer, uint32_t p_face_id, const Size2i &p_size); + void cubemap_filter(RID p_source_cubemap, Vector<RID> p_dest_cubemap, bool p_use_array); + void cubemap_filter_raster(RID p_source_cubemap, RID p_dest_framebuffer, uint32_t p_face_id, uint32_t p_mip_level); + + void cubemap_roughness(RID p_source_rd_texture, RID p_dest_texture, uint32_t p_face_id, uint32_t p_sample_count, float p_roughness, float p_size); + void cubemap_roughness_raster(RID p_source_rd_texture, RID p_dest_framebuffer, uint32_t p_face_id, uint32_t p_sample_count, float p_roughness, float p_size); + + void merge_specular(RID p_dest_framebuffer, RID p_specular, RID p_base, RID p_reflection, uint32_t p_view_count); +}; + +} // namespace RendererRD + +#endif // COPY_EFFECTS_RD_H diff --git a/servers/rendering/renderer_rd/effects/fsr.cpp b/servers/rendering/renderer_rd/effects/fsr.cpp new file mode 100644 index 0000000000..92b34ede0e --- /dev/null +++ b/servers/rendering/renderer_rd/effects/fsr.cpp @@ -0,0 +1,128 @@ +/*************************************************************************/ +/* fsr.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "fsr.h" +#include "../storage_rd/material_storage.h" +#include "../uniform_set_cache_rd.h" + +using namespace RendererRD; + +FSR::FSR() { + Vector<String> FSR_upscale_modes; + +#if defined(MACOS_ENABLED) || defined(IOS_ENABLED) + // MoltenVK does not support some of the operations used by the normal mode of FSR. Fallback works just fine though. + FSR_upscale_modes.push_back("\n#define MODE_FSR_UPSCALE_FALLBACK\n"); +#else + // Everyone else can use normal mode when available. + if (RD::get_singleton()->has_feature(RD::SUPPORTS_FSR_HALF_FLOAT)) { + FSR_upscale_modes.push_back("\n#define MODE_FSR_UPSCALE_NORMAL\n"); + } else { + FSR_upscale_modes.push_back("\n#define MODE_FSR_UPSCALE_FALLBACK\n"); + } +#endif + + fsr_shader.initialize(FSR_upscale_modes); + + shader_version = fsr_shader.version_create(); + pipeline = RD::get_singleton()->compute_pipeline_create(fsr_shader.version_get_shader(shader_version, 0)); +} + +FSR::~FSR() { + fsr_shader.version_free(shader_version); +} + +void FSR::fsr_upscale(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_source_rd_texture, RID p_destination_texture) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + Size2i internal_size = p_render_buffers->get_internal_size(); + Size2i target_size = p_render_buffers->get_target_size(); + float fsr_upscale_sharpness = p_render_buffers->get_fsr_sharpness(); + + if (!p_render_buffers->has_texture(SNAME("FSR"), SNAME("upscale_texture"))) { + RD::DataFormat format = p_render_buffers->get_base_data_format(); + uint32_t usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + uint32_t layers = 1; // we only need one layer, in multiview we're processing one layer at a time. + + p_render_buffers->create_texture(SNAME("FSR"), SNAME("upscale_texture"), format, usage_bits, RD::TEXTURE_SAMPLES_1, target_size, layers); + } + + RID upscale_texture = p_render_buffers->get_texture(SNAME("FSR"), SNAME("upscale_texture")); + + FSRUpscalePushConstant push_constant; + memset(&push_constant, 0, sizeof(FSRUpscalePushConstant)); + + int dispatch_x = (target_size.x + 15) / 16; + int dispatch_y = (target_size.y + 15) / 16; + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, pipeline); + + push_constant.resolution_width = internal_size.width; + push_constant.resolution_height = internal_size.height; + push_constant.upscaled_width = target_size.width; + push_constant.upscaled_height = target_size.height; + push_constant.sharpness = fsr_upscale_sharpness; + + RID shader = fsr_shader.version_get_shader(shader_version, 0); + ERR_FAIL_COND(shader.is_null()); + + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + //FSR Easc + RD::Uniform u_source_rd_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, { default_sampler, p_source_rd_texture }); + RD::Uniform u_upscale_texture(RD::UNIFORM_TYPE_IMAGE, 0, { upscale_texture }); + + push_constant.pass = FSR_UPSCALE_PASS_EASU; + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_source_rd_texture), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_upscale_texture), 1); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(FSRUpscalePushConstant)); + + RD::get_singleton()->compute_list_dispatch(compute_list, dispatch_x, dispatch_y, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + //FSR Rcas + RD::Uniform u_upscale_texture_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, { default_sampler, upscale_texture }); + RD::Uniform u_destination_texture(RD::UNIFORM_TYPE_IMAGE, 0, { p_destination_texture }); + + push_constant.pass = FSR_UPSCALE_PASS_RCAS; + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_upscale_texture_with_sampler), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_destination_texture), 1); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(FSRUpscalePushConstant)); + + RD::get_singleton()->compute_list_dispatch(compute_list, dispatch_x, dispatch_y, 1); + + RD::get_singleton()->compute_list_end(compute_list); +} diff --git a/servers/rendering/renderer_rd/effects/fsr.h b/servers/rendering/renderer_rd/effects/fsr.h new file mode 100644 index 0000000000..69088e526a --- /dev/null +++ b/servers/rendering/renderer_rd/effects/fsr.h @@ -0,0 +1,72 @@ +/*************************************************************************/ +/* fsr.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef FSR_RD_H +#define FSR_RD_H + +#include "../pipeline_cache_rd.h" +#include "../shaders/effects/fsr_upscale.glsl.gen.h" +#include "../storage_rd/render_scene_buffers_rd.h" +#include "servers/rendering/renderer_scene_render.h" +#include "servers/rendering_server.h" + +namespace RendererRD { + +class FSR { +public: + FSR(); + ~FSR(); + + void fsr_upscale(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_source_rd_texture, RID p_destination_texture); + +private: + enum FSRUpscalePass { + FSR_UPSCALE_PASS_EASU = 0, + FSR_UPSCALE_PASS_RCAS = 1 + }; + + struct FSRUpscalePushConstant { + float resolution_width; + float resolution_height; + float upscaled_width; + float upscaled_height; + float sharpness; + int pass; + int _unused0, _unused1; + }; + + FsrUpscaleShaderRD fsr_shader; + RID shader_version; + RID pipeline; +}; + +} // namespace RendererRD + +#endif // FSR_RD_H diff --git a/servers/rendering/renderer_rd/effects/resolve.cpp b/servers/rendering/renderer_rd/effects/resolve.cpp new file mode 100644 index 0000000000..6c49a2ebce --- /dev/null +++ b/servers/rendering/renderer_rd/effects/resolve.cpp @@ -0,0 +1,130 @@ +/*************************************************************************/ +/* resolve.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "resolve.h" +#include "servers/rendering/renderer_rd/renderer_compositor_rd.h" +#include "servers/rendering/renderer_rd/storage_rd/material_storage.h" +#include "servers/rendering/renderer_rd/uniform_set_cache_rd.h" + +using namespace RendererRD; + +Resolve::Resolve() { + Vector<String> resolve_modes; + resolve_modes.push_back("\n#define MODE_RESOLVE_GI\n"); + resolve_modes.push_back("\n#define MODE_RESOLVE_GI\n#define VOXEL_GI_RESOLVE\n"); + resolve_modes.push_back("\n#define MODE_RESOLVE_DEPTH\n"); + + resolve.shader.initialize(resolve_modes); + + resolve.shader_version = resolve.shader.version_create(); + + for (int i = 0; i < RESOLVE_MODE_MAX; i++) { + resolve.pipelines[i] = RD::get_singleton()->compute_pipeline_create(resolve.shader.version_get_shader(resolve.shader_version, i)); + } +} + +Resolve::~Resolve() { + resolve.shader.version_free(resolve.shader_version); +} + +void Resolve::resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RID p_source_voxel_gi, RID p_dest_depth, RID p_dest_normal_roughness, RID p_dest_voxel_gi, Vector2i p_screen_size, int p_samples, uint32_t p_barrier) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + ResolvePushConstant push_constant; + push_constant.screen_size[0] = p_screen_size.x; + push_constant.screen_size[1] = p_screen_size.y; + push_constant.samples = p_samples; + + // setup our uniforms + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_source_depth(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_source_depth })); + RD::Uniform u_source_normal_roughness(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 1, Vector<RID>({ default_sampler, p_source_normal_roughness })); + RD::Uniform u_dest_depth(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_dest_depth })); + RD::Uniform u_dest_normal_roughness(RD::UNIFORM_TYPE_IMAGE, 1, Vector<RID>({ p_dest_normal_roughness })); + + ResolveMode mode = p_source_voxel_gi.is_valid() ? RESOLVE_MODE_GI_VOXEL_GI : RESOLVE_MODE_GI; + RID shader = resolve.shader.version_get_shader(resolve.shader_version, mode); + ERR_FAIL_COND(shader.is_null()); + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, resolve.pipelines[mode]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_source_depth, u_source_normal_roughness), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_dest_depth, u_dest_normal_roughness), 1); + if (p_source_voxel_gi.is_valid()) { + RD::Uniform u_source_voxel_gi(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_source_voxel_gi })); + RD::Uniform u_dest_voxel_gi(RD::UNIFORM_TYPE_IMAGE, 0, p_dest_voxel_gi); + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 2, u_source_voxel_gi), 2); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 3, u_dest_voxel_gi), 3); + } + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ResolvePushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.x, p_screen_size.y, 1); + + RD::get_singleton()->compute_list_end(p_barrier); +} + +void Resolve::resolve_depth(RID p_source_depth, RID p_dest_depth, Vector2i p_screen_size, int p_samples, uint32_t p_barrier) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + ResolvePushConstant push_constant; + push_constant.screen_size[0] = p_screen_size.x; + push_constant.screen_size[1] = p_screen_size.y; + push_constant.samples = p_samples; + + // setup our uniforms + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_source_depth(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_source_depth })); + RD::Uniform u_dest_depth(RD::UNIFORM_TYPE_IMAGE, 0, p_dest_depth); + + ResolveMode mode = RESOLVE_MODE_DEPTH; + RID shader = resolve.shader.version_get_shader(resolve.shader_version, mode); + ERR_FAIL_COND(shader.is_null()); + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, resolve.pipelines[mode]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_source_depth), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_dest_depth), 1); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ResolvePushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.x, p_screen_size.y, 1); + + RD::get_singleton()->compute_list_end(p_barrier); +} diff --git a/servers/rendering/renderer_rd/effects/resolve.h b/servers/rendering/renderer_rd/effects/resolve.h new file mode 100644 index 0000000000..2a4cd06827 --- /dev/null +++ b/servers/rendering/renderer_rd/effects/resolve.h @@ -0,0 +1,74 @@ +/*************************************************************************/ +/* resolve.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef RESOLVE_RD_H +#define RESOLVE_RD_H + +#include "servers/rendering/renderer_rd/pipeline_cache_rd.h" +#include "servers/rendering/renderer_rd/shaders/effects/resolve.glsl.gen.h" +#include "servers/rendering/renderer_scene_render.h" + +#include "servers/rendering_server.h" + +namespace RendererRD { + +class Resolve { +private: + struct ResolvePushConstant { + int32_t screen_size[2]; + int32_t samples; + uint32_t pad; + }; + + enum ResolveMode { + RESOLVE_MODE_GI, + RESOLVE_MODE_GI_VOXEL_GI, + RESOLVE_MODE_DEPTH, + RESOLVE_MODE_MAX + }; + + struct ResolveShader { + ResolvePushConstant push_constant; + ResolveShaderRD shader; + RID shader_version; + RID pipelines[RESOLVE_MODE_MAX]; //3 quality levels + } resolve; + +public: + Resolve(); + ~Resolve(); + + void resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RID p_source_voxel_gi, RID p_dest_depth, RID p_dest_normal_roughness, RID p_dest_voxel_gi, Vector2i p_screen_size, int p_samples, uint32_t p_barrier = RD::BARRIER_MASK_ALL); + void resolve_depth(RID p_source_depth, RID p_dest_depth, Vector2i p_screen_size, int p_samples, uint32_t p_barrier = RD::BARRIER_MASK_ALL); +}; + +} // namespace RendererRD + +#endif // RESOLVE_RD_H diff --git a/servers/rendering/renderer_rd/effects/ss_effects.cpp b/servers/rendering/renderer_rd/effects/ss_effects.cpp new file mode 100644 index 0000000000..9653382e96 --- /dev/null +++ b/servers/rendering/renderer_rd/effects/ss_effects.cpp @@ -0,0 +1,1876 @@ +/*************************************************************************/ +/* ss_effects.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "ss_effects.h" + +#include "core/config/project_settings.h" +#include "servers/rendering/renderer_rd/renderer_compositor_rd.h" +#include "servers/rendering/renderer_rd/storage_rd/material_storage.h" +#include "servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.h" +#include "servers/rendering/renderer_rd/uniform_set_cache_rd.h" + +using namespace RendererRD; + +SSEffects *SSEffects::singleton = nullptr; + +static _FORCE_INLINE_ void store_camera(const Projection &p_mtx, float *p_array) { + for (int i = 0; i < 4; i++) { + for (int j = 0; j < 4; j++) { + p_array[i * 4 + j] = p_mtx.columns[i][j]; + } + } +} + +SSEffects::SSEffects() { + singleton = this; + + // Initialize depth buffer for screen space effects + { + Vector<String> downsampler_modes; + downsampler_modes.push_back("\n"); + downsampler_modes.push_back("\n#define USE_HALF_SIZE\n"); + downsampler_modes.push_back("\n#define GENERATE_MIPS\n"); + downsampler_modes.push_back("\n#define GENERATE_MIPS\n#define USE_HALF_SIZE\n"); + downsampler_modes.push_back("\n#define USE_HALF_BUFFERS\n"); + downsampler_modes.push_back("\n#define USE_HALF_BUFFERS\n#define USE_HALF_SIZE\n"); + downsampler_modes.push_back("\n#define GENERATE_MIPS\n#define GENERATE_FULL_MIPS"); + + ss_effects.downsample_shader.initialize(downsampler_modes); + + ss_effects.downsample_shader_version = ss_effects.downsample_shader.version_create(); + + for (int i = 0; i < SS_EFFECTS_MAX; i++) { + ss_effects.pipelines[i] = RD::get_singleton()->compute_pipeline_create(ss_effects.downsample_shader.version_get_shader(ss_effects.downsample_shader_version, i)); + } + + ss_effects.gather_constants_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(SSEffectsGatherConstants)); + SSEffectsGatherConstants gather_constants; + + const int sub_pass_count = 5; + for (int pass = 0; pass < 4; pass++) { + for (int subPass = 0; subPass < sub_pass_count; subPass++) { + int a = pass; + int b = subPass; + + int spmap[5]{ 0, 1, 4, 3, 2 }; + b = spmap[subPass]; + + float ca, sa; + float angle0 = (float(a) + float(b) / float(sub_pass_count)) * Math_PI * 0.5f; + + ca = Math::cos(angle0); + sa = Math::sin(angle0); + + float scale = 1.0f + (a - 1.5f + (b - (sub_pass_count - 1.0f) * 0.5f) / float(sub_pass_count)) * 0.07f; + + gather_constants.rotation_matrices[pass * 20 + subPass * 4 + 0] = scale * ca; + gather_constants.rotation_matrices[pass * 20 + subPass * 4 + 1] = scale * -sa; + gather_constants.rotation_matrices[pass * 20 + subPass * 4 + 2] = -scale * sa; + gather_constants.rotation_matrices[pass * 20 + subPass * 4 + 3] = -scale * ca; + } + } + + RD::get_singleton()->buffer_update(ss_effects.gather_constants_buffer, 0, sizeof(SSEffectsGatherConstants), &gather_constants); + } + + // Initialize Screen Space Indirect Lighting (SSIL) + ssil_set_quality(RS::EnvironmentSSILQuality(int(GLOBAL_GET("rendering/environment/ssil/quality"))), GLOBAL_GET("rendering/environment/ssil/half_size"), GLOBAL_GET("rendering/environment/ssil/adaptive_target"), GLOBAL_GET("rendering/environment/ssil/blur_passes"), GLOBAL_GET("rendering/environment/ssil/fadeout_from"), GLOBAL_GET("rendering/environment/ssil/fadeout_to")); + + { + Vector<String> ssil_modes; + ssil_modes.push_back("\n"); + ssil_modes.push_back("\n#define SSIL_BASE\n"); + ssil_modes.push_back("\n#define ADAPTIVE\n"); + + ssil.gather_shader.initialize(ssil_modes); + + ssil.gather_shader_version = ssil.gather_shader.version_create(); + + for (int i = SSIL_GATHER; i <= SSIL_GATHER_ADAPTIVE; i++) { + ssil.pipelines[i] = RD::get_singleton()->compute_pipeline_create(ssil.gather_shader.version_get_shader(ssil.gather_shader_version, i)); + } + ssil.projection_uniform_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(SSILProjectionUniforms)); + } + + { + Vector<String> ssil_modes; + ssil_modes.push_back("\n#define GENERATE_MAP\n"); + ssil_modes.push_back("\n#define PROCESS_MAPA\n"); + ssil_modes.push_back("\n#define PROCESS_MAPB\n"); + + ssil.importance_map_shader.initialize(ssil_modes); + + ssil.importance_map_shader_version = ssil.importance_map_shader.version_create(); + + for (int i = SSIL_GENERATE_IMPORTANCE_MAP; i <= SSIL_PROCESS_IMPORTANCE_MAPB; i++) { + ssil.pipelines[i] = RD::get_singleton()->compute_pipeline_create(ssil.importance_map_shader.version_get_shader(ssil.importance_map_shader_version, i - SSIL_GENERATE_IMPORTANCE_MAP)); + } + ssil.importance_map_load_counter = RD::get_singleton()->storage_buffer_create(sizeof(uint32_t)); + int zero[1] = { 0 }; + RD::get_singleton()->buffer_update(ssil.importance_map_load_counter, 0, sizeof(uint32_t), &zero); + RD::get_singleton()->set_resource_name(ssil.importance_map_load_counter, "Importance Map Load Counter"); + + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 0; + u.append_id(ssil.importance_map_load_counter); + uniforms.push_back(u); + } + ssil.counter_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssil.importance_map_shader.version_get_shader(ssil.importance_map_shader_version, 2), 2); + RD::get_singleton()->set_resource_name(ssil.counter_uniform_set, "Load Counter Uniform Set"); + } + + { + Vector<String> ssil_modes; + ssil_modes.push_back("\n#define MODE_NON_SMART\n"); + ssil_modes.push_back("\n#define MODE_SMART\n"); + ssil_modes.push_back("\n#define MODE_WIDE\n"); + + ssil.blur_shader.initialize(ssil_modes); + + ssil.blur_shader_version = ssil.blur_shader.version_create(); + for (int i = SSIL_BLUR_PASS; i <= SSIL_BLUR_PASS_WIDE; i++) { + ssil.pipelines[i] = RD::get_singleton()->compute_pipeline_create(ssil.blur_shader.version_get_shader(ssil.blur_shader_version, i - SSIL_BLUR_PASS)); + } + } + + { + Vector<String> ssil_modes; + ssil_modes.push_back("\n#define MODE_NON_SMART\n"); + ssil_modes.push_back("\n#define MODE_SMART\n"); + ssil_modes.push_back("\n#define MODE_HALF\n"); + + ssil.interleave_shader.initialize(ssil_modes); + + ssil.interleave_shader_version = ssil.interleave_shader.version_create(); + for (int i = SSIL_INTERLEAVE; i <= SSIL_INTERLEAVE_HALF; i++) { + ssil.pipelines[i] = RD::get_singleton()->compute_pipeline_create(ssil.interleave_shader.version_get_shader(ssil.interleave_shader_version, i - SSIL_INTERLEAVE)); + } + } + + // Initialize Screen Space Ambient Occlusion (SSAO) + ssao_set_quality(RS::EnvironmentSSAOQuality(int(GLOBAL_GET("rendering/environment/ssao/quality"))), GLOBAL_GET("rendering/environment/ssao/half_size"), GLOBAL_GET("rendering/environment/ssao/adaptive_target"), GLOBAL_GET("rendering/environment/ssao/blur_passes"), GLOBAL_GET("rendering/environment/ssao/fadeout_from"), GLOBAL_GET("rendering/environment/ssao/fadeout_to")); + + { + RD::SamplerState sampler; + sampler.mag_filter = RD::SAMPLER_FILTER_NEAREST; + sampler.min_filter = RD::SAMPLER_FILTER_NEAREST; + sampler.mip_filter = RD::SAMPLER_FILTER_NEAREST; + sampler.repeat_u = RD::SAMPLER_REPEAT_MODE_MIRRORED_REPEAT; + sampler.repeat_v = RD::SAMPLER_REPEAT_MODE_MIRRORED_REPEAT; + sampler.repeat_w = RD::SAMPLER_REPEAT_MODE_MIRRORED_REPEAT; + sampler.max_lod = 4; + + uint32_t pipeline = 0; + { + Vector<String> ssao_modes; + + ssao_modes.push_back("\n"); + ssao_modes.push_back("\n#define SSAO_BASE\n"); + ssao_modes.push_back("\n#define ADAPTIVE\n"); + + ssao.gather_shader.initialize(ssao_modes); + + ssao.gather_shader_version = ssao.gather_shader.version_create(); + + for (int i = 0; i <= SSAO_GATHER_ADAPTIVE; i++) { + ssao.pipelines[pipeline] = RD::get_singleton()->compute_pipeline_create(ssao.gather_shader.version_get_shader(ssao.gather_shader_version, i)); + pipeline++; + } + } + + { + Vector<String> ssao_modes; + ssao_modes.push_back("\n#define GENERATE_MAP\n"); + ssao_modes.push_back("\n#define PROCESS_MAPA\n"); + ssao_modes.push_back("\n#define PROCESS_MAPB\n"); + + ssao.importance_map_shader.initialize(ssao_modes); + + ssao.importance_map_shader_version = ssao.importance_map_shader.version_create(); + + for (int i = SSAO_GENERATE_IMPORTANCE_MAP; i <= SSAO_PROCESS_IMPORTANCE_MAPB; i++) { + ssao.pipelines[pipeline] = RD::get_singleton()->compute_pipeline_create(ssao.importance_map_shader.version_get_shader(ssao.importance_map_shader_version, i - SSAO_GENERATE_IMPORTANCE_MAP)); + + pipeline++; + } + + ssao.importance_map_load_counter = RD::get_singleton()->storage_buffer_create(sizeof(uint32_t)); + int zero[1] = { 0 }; + RD::get_singleton()->buffer_update(ssao.importance_map_load_counter, 0, sizeof(uint32_t), &zero); + RD::get_singleton()->set_resource_name(ssao.importance_map_load_counter, "Importance Map Load Counter"); + + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 0; + u.append_id(ssao.importance_map_load_counter); + uniforms.push_back(u); + } + ssao.counter_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssao.importance_map_shader.version_get_shader(ssao.importance_map_shader_version, 2), 2); + RD::get_singleton()->set_resource_name(ssao.counter_uniform_set, "Load Counter Uniform Set"); + } + + { + Vector<String> ssao_modes; + ssao_modes.push_back("\n#define MODE_NON_SMART\n"); + ssao_modes.push_back("\n#define MODE_SMART\n"); + ssao_modes.push_back("\n#define MODE_WIDE\n"); + + ssao.blur_shader.initialize(ssao_modes); + + ssao.blur_shader_version = ssao.blur_shader.version_create(); + + for (int i = SSAO_BLUR_PASS; i <= SSAO_BLUR_PASS_WIDE; i++) { + ssao.pipelines[pipeline] = RD::get_singleton()->compute_pipeline_create(ssao.blur_shader.version_get_shader(ssao.blur_shader_version, i - SSAO_BLUR_PASS)); + + pipeline++; + } + } + + { + Vector<String> ssao_modes; + ssao_modes.push_back("\n#define MODE_NON_SMART\n"); + ssao_modes.push_back("\n#define MODE_SMART\n"); + ssao_modes.push_back("\n#define MODE_HALF\n"); + + ssao.interleave_shader.initialize(ssao_modes); + + ssao.interleave_shader_version = ssao.interleave_shader.version_create(); + for (int i = SSAO_INTERLEAVE; i <= SSAO_INTERLEAVE_HALF; i++) { + ssao.pipelines[pipeline] = RD::get_singleton()->compute_pipeline_create(ssao.interleave_shader.version_get_shader(ssao.interleave_shader_version, i - SSAO_INTERLEAVE)); + RD::get_singleton()->set_resource_name(ssao.pipelines[pipeline], "Interleave Pipeline " + itos(i)); + pipeline++; + } + } + + ERR_FAIL_COND(pipeline != SSAO_MAX); + + ss_effects.mirror_sampler = RD::get_singleton()->sampler_create(sampler); + } + + // Screen Space Reflections + ssr_roughness_quality = RS::EnvironmentSSRRoughnessQuality(int(GLOBAL_GET("rendering/environment/screen_space_reflection/roughness_quality"))); + + { + Vector<RD::PipelineSpecializationConstant> specialization_constants; + + { + RD::PipelineSpecializationConstant sc; + sc.type = RD::PIPELINE_SPECIALIZATION_CONSTANT_TYPE_BOOL; + sc.constant_id = 0; // SSR_USE_FULL_PROJECTION_MATRIX + sc.bool_value = false; + specialization_constants.push_back(sc); + } + + { + Vector<String> ssr_scale_modes; + ssr_scale_modes.push_back("\n"); + + ssr_scale.shader.initialize(ssr_scale_modes); + ssr_scale.shader_version = ssr_scale.shader.version_create(); + + for (int v = 0; v < SSR_VARIATIONS; v++) { + specialization_constants.ptrw()[0].bool_value = (v & SSR_MULTIVIEW) ? true : false; + ssr_scale.pipelines[v] = RD::get_singleton()->compute_pipeline_create(ssr_scale.shader.version_get_shader(ssr_scale.shader_version, 0), specialization_constants); + } + } + + { + Vector<String> ssr_modes; + ssr_modes.push_back("\n"); // SCREEN_SPACE_REFLECTION_NORMAL + ssr_modes.push_back("\n#define MODE_ROUGH\n"); // SCREEN_SPACE_REFLECTION_ROUGH + + ssr.shader.initialize(ssr_modes); + ssr.shader_version = ssr.shader.version_create(); + + for (int v = 0; v < SSR_VARIATIONS; v++) { + specialization_constants.ptrw()[0].bool_value = (v & SSR_MULTIVIEW) ? true : false; + for (int i = 0; i < SCREEN_SPACE_REFLECTION_MAX; i++) { + ssr.pipelines[v][i] = RD::get_singleton()->compute_pipeline_create(ssr.shader.version_get_shader(ssr.shader_version, i), specialization_constants); + } + } + } + + { + Vector<String> ssr_filter_modes; + ssr_filter_modes.push_back("\n"); // SCREEN_SPACE_REFLECTION_FILTER_HORIZONTAL + ssr_filter_modes.push_back("\n#define VERTICAL_PASS\n"); // SCREEN_SPACE_REFLECTION_FILTER_VERTICAL + + ssr_filter.shader.initialize(ssr_filter_modes); + ssr_filter.shader_version = ssr_filter.shader.version_create(); + + for (int v = 0; v < SSR_VARIATIONS; v++) { + specialization_constants.ptrw()[0].bool_value = (v & SSR_MULTIVIEW) ? true : false; + for (int i = 0; i < SCREEN_SPACE_REFLECTION_FILTER_MAX; i++) { + ssr_filter.pipelines[v][i] = RD::get_singleton()->compute_pipeline_create(ssr_filter.shader.version_get_shader(ssr_filter.shader_version, i), specialization_constants); + } + } + } + } + + // Subsurface scattering + sss_quality = RS::SubSurfaceScatteringQuality(int(GLOBAL_GET("rendering/environment/subsurface_scattering/subsurface_scattering_quality"))); + sss_scale = GLOBAL_GET("rendering/environment/subsurface_scattering/subsurface_scattering_scale"); + sss_depth_scale = GLOBAL_GET("rendering/environment/subsurface_scattering/subsurface_scattering_depth_scale"); + + { + Vector<String> sss_modes; + sss_modes.push_back("\n#define USE_11_SAMPLES\n"); + sss_modes.push_back("\n#define USE_17_SAMPLES\n"); + sss_modes.push_back("\n#define USE_25_SAMPLES\n"); + + sss.shader.initialize(sss_modes); + + sss.shader_version = sss.shader.version_create(); + + for (int i = 0; i < sss_modes.size(); i++) { + sss.pipelines[i] = RD::get_singleton()->compute_pipeline_create(sss.shader.version_get_shader(sss.shader_version, i)); + } + } +} + +SSEffects::~SSEffects() { + { + // Cleanup SS Reflections + ssr.shader.version_free(ssr.shader_version); + ssr_filter.shader.version_free(ssr_filter.shader_version); + ssr_scale.shader.version_free(ssr_scale.shader_version); + + if (ssr.ubo.is_valid()) { + RD::get_singleton()->free(ssr.ubo); + } + } + + { + // Cleanup SS downsampler + ss_effects.downsample_shader.version_free(ss_effects.downsample_shader_version); + + RD::get_singleton()->free(ss_effects.mirror_sampler); + RD::get_singleton()->free(ss_effects.gather_constants_buffer); + } + + { + // Cleanup SSIL + ssil.blur_shader.version_free(ssil.blur_shader_version); + ssil.gather_shader.version_free(ssil.gather_shader_version); + ssil.interleave_shader.version_free(ssil.interleave_shader_version); + ssil.importance_map_shader.version_free(ssil.importance_map_shader_version); + + RD::get_singleton()->free(ssil.importance_map_load_counter); + RD::get_singleton()->free(ssil.projection_uniform_buffer); + } + + { + // Cleanup SSAO + ssao.blur_shader.version_free(ssao.blur_shader_version); + ssao.gather_shader.version_free(ssao.gather_shader_version); + ssao.interleave_shader.version_free(ssao.interleave_shader_version); + ssao.importance_map_shader.version_free(ssao.importance_map_shader_version); + + RD::get_singleton()->free(ssao.importance_map_load_counter); + } + + { + // Cleanup Subsurface scattering + sss.shader.version_free(sss.shader_version); + } + + singleton = nullptr; +} + +/* SS Downsampler */ + +void SSEffects::downsample_depth(RID p_depth_buffer, const Vector<RID> &p_depth_mipmaps, bool p_invalidate_uniform_set, Size2i p_full_screen_size, const Projection &p_projection) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + // Downsample and deinterleave the depth buffer for SSAO and SSIL + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + + int downsample_mode = SS_EFFECTS_DOWNSAMPLE; + bool use_mips = ssao_quality > RS::ENV_SSAO_QUALITY_MEDIUM || ssil_quality > RS::ENV_SSIL_QUALITY_MEDIUM; + + if (ssao_quality == RS::ENV_SSAO_QUALITY_VERY_LOW && ssil_quality == RS::ENV_SSIL_QUALITY_VERY_LOW) { + downsample_mode = SS_EFFECTS_DOWNSAMPLE_HALF; + } else if (use_mips) { + downsample_mode = SS_EFFECTS_DOWNSAMPLE_MIPMAP; + } + + bool use_half_size = false; + bool use_full_mips = false; + + if (ssao_half_size && ssil_half_size) { + downsample_mode++; + use_half_size = true; + } else if (ssao_half_size != ssil_half_size) { + if (use_mips) { + downsample_mode = SS_EFFECTS_DOWNSAMPLE_FULL_MIPS; + use_full_mips = true; + } else { + // Only need the first two mipmaps, but the cost to generate the next two is trivial + // TODO investigate the benefit of a shader version to generate only 2 mips + downsample_mode = SS_EFFECTS_DOWNSAMPLE_MIPMAP; + use_mips = true; + } + } + + int depth_index = use_half_size ? 1 : 0; + + RD::get_singleton()->draw_command_begin_label("Downsample Depth"); + if (p_invalidate_uniform_set || use_full_mips != ss_effects.used_full_mips_last_frame || use_half_size != ss_effects.used_half_size_last_frame || use_mips != ss_effects.used_mips_last_frame) { + if (ss_effects.downsample_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(ss_effects.downsample_uniform_set)) { + RD::get_singleton()->free(ss_effects.downsample_uniform_set); + ss_effects.downsample_uniform_set = RID(); + } + + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 0; + u.append_id(p_depth_mipmaps[depth_index + 1]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 1; + u.append_id(p_depth_mipmaps[depth_index + 2]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 2; + u.append_id(p_depth_mipmaps[depth_index + 3]); + uniforms.push_back(u); + } + if (use_full_mips) { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 3; + u.append_id(p_depth_mipmaps[4]); + uniforms.push_back(u); + } + ss_effects.downsample_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ss_effects.downsample_shader.version_get_shader(ss_effects.downsample_shader_version, use_full_mips ? 6 : 2), 2); + } + + float depth_linearize_mul = -p_projection.columns[3][2]; + float depth_linearize_add = p_projection.columns[2][2]; + if (depth_linearize_mul * depth_linearize_add < 0) { + depth_linearize_add = -depth_linearize_add; + } + + ss_effects.downsample_push_constant.orthogonal = p_projection.is_orthogonal(); + ss_effects.downsample_push_constant.z_near = depth_linearize_mul; + ss_effects.downsample_push_constant.z_far = depth_linearize_add; + if (ss_effects.downsample_push_constant.orthogonal) { + ss_effects.downsample_push_constant.z_near = p_projection.get_z_near(); + ss_effects.downsample_push_constant.z_far = p_projection.get_z_far(); + } + ss_effects.downsample_push_constant.pixel_size[0] = 1.0 / p_full_screen_size.x; + ss_effects.downsample_push_constant.pixel_size[1] = 1.0 / p_full_screen_size.y; + ss_effects.downsample_push_constant.radius_sq = 1.0; + + RID shader = ss_effects.downsample_shader.version_get_shader(ss_effects.downsample_shader_version, downsample_mode); + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_depth_buffer(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_depth_buffer })); + RD::Uniform u_depth_mipmaps(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_depth_mipmaps[depth_index + 0] })); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ss_effects.pipelines[downsample_mode]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_depth_buffer), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_depth_mipmaps), 1); + if (use_mips) { + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, ss_effects.downsample_uniform_set, 2); + } + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ss_effects.downsample_push_constant, sizeof(SSEffectsDownsamplePushConstant)); + + Size2i size(MAX(1, p_full_screen_size.x >> (use_half_size ? 2 : 1)), MAX(1, p_full_screen_size.y >> (use_half_size ? 2 : 1))); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, size.x, size.y, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + RD::get_singleton()->draw_command_end_label(); + + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_COMPUTE); + + ss_effects.used_full_mips_last_frame = use_full_mips; + ss_effects.used_half_size_last_frame = use_half_size; + ss_effects.used_mips_last_frame = use_mips; +} + +/* SSIL */ + +void SSEffects::ssil_set_quality(RS::EnvironmentSSILQuality p_quality, bool p_half_size, float p_adaptive_target, int p_blur_passes, float p_fadeout_from, float p_fadeout_to) { + ssil_quality = p_quality; + ssil_half_size = p_half_size; + ssil_adaptive_target = p_adaptive_target; + ssil_blur_passes = p_blur_passes; + ssil_fadeout_from = p_fadeout_from; + ssil_fadeout_to = p_fadeout_to; +} + +void SSEffects::gather_ssil(RD::ComputeListID p_compute_list, const Vector<RID> p_ssil_slices, const Vector<RID> p_edges_slices, const SSILSettings &p_settings, bool p_adaptive_base_pass, RID p_gather_uniform_set, RID p_importance_map_uniform_set, RID p_projection_uniform_set) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + + RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, p_gather_uniform_set, 0); + if ((ssil_quality == RS::ENV_SSIL_QUALITY_ULTRA) && !p_adaptive_base_pass) { + RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, p_importance_map_uniform_set, 1); + } + RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, p_projection_uniform_set, 3); + + RID shader = ssil.gather_shader.version_get_shader(ssil.gather_shader_version, 0); + + for (int i = 0; i < 4; i++) { + if ((ssil_quality == RS::ENV_SSIL_QUALITY_VERY_LOW) && ((i == 1) || (i == 2))) { + continue; + } + + RD::Uniform u_ssil_slice(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssil_slices[i] })); + RD::Uniform u_edges_slice(RD::UNIFORM_TYPE_IMAGE, 1, Vector<RID>({ p_edges_slices[i] })); + + ssil.gather_push_constant.pass_coord_offset[0] = i % 2; + ssil.gather_push_constant.pass_coord_offset[1] = i / 2; + ssil.gather_push_constant.pass_uv_offset[0] = ((i % 2) - 0.0) / p_settings.full_screen_size.x; + ssil.gather_push_constant.pass_uv_offset[1] = ((i / 2) - 0.0) / p_settings.full_screen_size.y; + ssil.gather_push_constant.pass = i; + RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, uniform_set_cache->get_cache(shader, 2, u_ssil_slice, u_edges_slice), 2); + RD::get_singleton()->compute_list_set_push_constant(p_compute_list, &ssil.gather_push_constant, sizeof(SSILGatherPushConstant)); + + Size2i size = Size2i(p_settings.full_screen_size.x >> (ssil_half_size ? 2 : 1), p_settings.full_screen_size.y >> (ssil_half_size ? 2 : 1)); + + RD::get_singleton()->compute_list_dispatch_threads(p_compute_list, size.x, size.y, 1); + } + RD::get_singleton()->compute_list_add_barrier(p_compute_list); +} + +void SSEffects::ssil_allocate_buffers(SSILRenderBuffers &p_ssil_buffers, const SSILSettings &p_settings, RID p_linear_depth) { + if (p_ssil_buffers.half_size != ssil_half_size) { + ssil_free(p_ssil_buffers); + } + + if (ssil_half_size) { + p_ssil_buffers.buffer_width = (p_settings.full_screen_size.x + 3) / 4; + p_ssil_buffers.buffer_height = (p_settings.full_screen_size.y + 3) / 4; + p_ssil_buffers.half_buffer_width = (p_settings.full_screen_size.x + 7) / 8; + p_ssil_buffers.half_buffer_height = (p_settings.full_screen_size.y + 7) / 8; + } else { + p_ssil_buffers.buffer_width = (p_settings.full_screen_size.x + 1) / 2; + p_ssil_buffers.buffer_height = (p_settings.full_screen_size.y + 1) / 2; + p_ssil_buffers.half_buffer_width = (p_settings.full_screen_size.x + 3) / 4; + p_ssil_buffers.half_buffer_height = (p_settings.full_screen_size.y + 3) / 4; + } + + if (p_ssil_buffers.ssil_final.is_null()) { + { + p_ssil_buffers.depth_texture_view = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_linear_depth, 0, ssil_half_size ? 1 : 0, 4, RD::TEXTURE_SLICE_2D_ARRAY); + } + { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.width = p_settings.full_screen_size.x; + tf.height = p_settings.full_screen_size.y; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; + p_ssil_buffers.ssil_final = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssil_buffers.ssil_final, "SSIL texture"); + RD::get_singleton()->texture_clear(p_ssil_buffers.ssil_final, Color(0, 0, 0, 0), 0, 1, 0, 1); + if (p_ssil_buffers.last_frame.is_null()) { + tf.mipmaps = 6; + p_ssil_buffers.last_frame = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssil_buffers.last_frame, "Last Frame Radiance"); + RD::get_singleton()->texture_clear(p_ssil_buffers.last_frame, Color(0, 0, 0, 0), 0, tf.mipmaps, 0, 1); + for (uint32_t i = 0; i < 6; i++) { + RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_ssil_buffers.last_frame, 0, i); + p_ssil_buffers.last_frame_slices.push_back(slice); + RD::get_singleton()->set_resource_name(slice, "Last Frame Radiance Mip " + itos(i) + " "); + } + } + } + { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; + tf.width = p_ssil_buffers.buffer_width; + tf.height = p_ssil_buffers.buffer_height; + tf.array_layers = 4; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + p_ssil_buffers.deinterleaved = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssil_buffers.deinterleaved, "SSIL deinterleaved buffer"); + for (uint32_t i = 0; i < 4; i++) { + RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_ssil_buffers.deinterleaved, i, 0); + p_ssil_buffers.deinterleaved_slices.push_back(slice); + RD::get_singleton()->set_resource_name(slice, "SSIL deinterleaved buffer array " + itos(i) + " "); + } + } + + { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; + tf.width = p_ssil_buffers.buffer_width; + tf.height = p_ssil_buffers.buffer_height; + tf.array_layers = 4; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + p_ssil_buffers.pong = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssil_buffers.pong, "SSIL deinterleaved pong buffer"); + for (uint32_t i = 0; i < 4; i++) { + RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_ssil_buffers.pong, i, 0); + p_ssil_buffers.pong_slices.push_back(slice); + RD::get_singleton()->set_resource_name(slice, "SSIL deinterleaved buffer pong array " + itos(i) + " "); + } + } + + { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R8_UNORM; + tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; + tf.width = p_ssil_buffers.buffer_width; + tf.height = p_ssil_buffers.buffer_height; + tf.array_layers = 4; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + p_ssil_buffers.edges = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssil_buffers.edges, "SSIL edges buffer"); + for (uint32_t i = 0; i < 4; i++) { + RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_ssil_buffers.edges, i, 0); + p_ssil_buffers.edges_slices.push_back(slice); + RD::get_singleton()->set_resource_name(slice, "SSIL edges buffer slice " + itos(i) + " "); + } + } + + { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R8_UNORM; + tf.width = p_ssil_buffers.half_buffer_width; + tf.height = p_ssil_buffers.half_buffer_height; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + p_ssil_buffers.importance_map[0] = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssil_buffers.importance_map[0], "SSIL Importance Map"); + p_ssil_buffers.importance_map[1] = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssil_buffers.importance_map[1], "SSIL Importance Map Pong"); + } + p_ssil_buffers.half_size = ssil_half_size; + } +} + +void SSEffects::screen_space_indirect_lighting(SSILRenderBuffers &p_ssil_buffers, RID p_normal_buffer, const Projection &p_projection, const Projection &p_last_projection, const SSILSettings &p_settings) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + RD::get_singleton()->draw_command_begin_label("Process Screen Space Indirect Lighting"); + //Store projection info before starting the compute list + SSILProjectionUniforms projection_uniforms; + store_camera(p_last_projection, projection_uniforms.inv_last_frame_projection_matrix); + + RD::get_singleton()->buffer_update(ssil.projection_uniform_buffer, 0, sizeof(SSILProjectionUniforms), &projection_uniforms); + + memset(&ssil.gather_push_constant, 0, sizeof(SSILGatherPushConstant)); + + RID shader = ssil.gather_shader.version_get_shader(ssil.gather_shader_version, 0); + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + RID default_mipmap_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + { + RD::get_singleton()->draw_command_begin_label("Gather Samples"); + ssil.gather_push_constant.screen_size[0] = p_settings.full_screen_size.x; + ssil.gather_push_constant.screen_size[1] = p_settings.full_screen_size.y; + + ssil.gather_push_constant.half_screen_pixel_size[0] = 1.0 / p_ssil_buffers.buffer_width; + ssil.gather_push_constant.half_screen_pixel_size[1] = 1.0 / p_ssil_buffers.buffer_height; + float tan_half_fov_x = 1.0 / p_projection.columns[0][0]; + float tan_half_fov_y = 1.0 / p_projection.columns[1][1]; + ssil.gather_push_constant.NDC_to_view_mul[0] = tan_half_fov_x * 2.0; + ssil.gather_push_constant.NDC_to_view_mul[1] = tan_half_fov_y * -2.0; + ssil.gather_push_constant.NDC_to_view_add[0] = tan_half_fov_x * -1.0; + ssil.gather_push_constant.NDC_to_view_add[1] = tan_half_fov_y; + ssil.gather_push_constant.z_near = p_projection.get_z_near(); + ssil.gather_push_constant.z_far = p_projection.get_z_far(); + ssil.gather_push_constant.is_orthogonal = p_projection.is_orthogonal(); + + ssil.gather_push_constant.half_screen_pixel_size_x025[0] = ssil.gather_push_constant.half_screen_pixel_size[0] * 0.25; + ssil.gather_push_constant.half_screen_pixel_size_x025[1] = ssil.gather_push_constant.half_screen_pixel_size[1] * 0.25; + + ssil.gather_push_constant.radius = p_settings.radius; + float radius_near_limit = (p_settings.radius * 1.2f); + if (ssil_quality <= RS::ENV_SSIL_QUALITY_LOW) { + radius_near_limit *= 1.50f; + + if (ssil_quality == RS::ENV_SSIL_QUALITY_VERY_LOW) { + ssil.gather_push_constant.radius *= 0.8f; + } + } + radius_near_limit /= tan_half_fov_y; + ssil.gather_push_constant.intensity = p_settings.intensity * Math_PI; + ssil.gather_push_constant.fade_out_mul = -1.0 / (ssil_fadeout_to - ssil_fadeout_from); + ssil.gather_push_constant.fade_out_add = ssil_fadeout_from / (ssil_fadeout_to - ssil_fadeout_from) + 1.0; + ssil.gather_push_constant.inv_radius_near_limit = 1.0f / radius_near_limit; + ssil.gather_push_constant.neg_inv_radius = -1.0 / ssil.gather_push_constant.radius; + ssil.gather_push_constant.normal_rejection_amount = p_settings.normal_rejection; + + ssil.gather_push_constant.load_counter_avg_div = 9.0 / float((p_ssil_buffers.half_buffer_width) * (p_ssil_buffers.half_buffer_height) * 255); + ssil.gather_push_constant.adaptive_sample_limit = ssil_adaptive_target; + + ssil.gather_push_constant.quality = MAX(0, ssil_quality - 1); + ssil.gather_push_constant.size_multiplier = ssil_half_size ? 2 : 1; + + if (p_ssil_buffers.projection_uniform_set.is_null()) { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; + u.binding = 0; + u.append_id(default_mipmap_sampler); + u.append_id(p_ssil_buffers.last_frame); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 1; + u.append_id(ssil.projection_uniform_buffer); + uniforms.push_back(u); + } + p_ssil_buffers.projection_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssil.gather_shader.version_get_shader(ssil.gather_shader_version, 0), 3); + } + + if (p_ssil_buffers.gather_uniform_set.is_null()) { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; + u.binding = 0; + u.append_id(default_sampler); + u.append_id(p_ssil_buffers.depth_texture_view); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 1; + u.append_id(p_normal_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 2; + u.append_id(ss_effects.gather_constants_buffer); + uniforms.push_back(u); + } + p_ssil_buffers.gather_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssil.gather_shader.version_get_shader(ssil.gather_shader_version, 0), 0); + } + + if (p_ssil_buffers.importance_map_uniform_set.is_null()) { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 0; + u.append_id(p_ssil_buffers.pong); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; + u.binding = 1; + u.append_id(default_sampler); + u.append_id(p_ssil_buffers.importance_map[0]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 2; + u.append_id(ssil.importance_map_load_counter); + uniforms.push_back(u); + } + p_ssil_buffers.importance_map_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssil.gather_shader.version_get_shader(ssil.gather_shader_version, 2), 1); + } + + if (ssil_quality == RS::ENV_SSIL_QUALITY_ULTRA) { + RD::get_singleton()->draw_command_begin_label("Generate Importance Map"); + ssil.importance_map_push_constant.half_screen_pixel_size[0] = 1.0 / p_ssil_buffers.buffer_width; + ssil.importance_map_push_constant.half_screen_pixel_size[1] = 1.0 / p_ssil_buffers.buffer_height; + ssil.importance_map_push_constant.intensity = p_settings.intensity * Math_PI; + //base pass + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssil.pipelines[SSIL_GATHER_BASE]); + gather_ssil(compute_list, p_ssil_buffers.pong_slices, p_ssil_buffers.edges_slices, p_settings, true, p_ssil_buffers.gather_uniform_set, p_ssil_buffers.importance_map_uniform_set, p_ssil_buffers.projection_uniform_set); + + //generate importance map + RD::Uniform u_ssil_pong_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssil_buffers.pong })); + RD::Uniform u_importance_map(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssil_buffers.importance_map[0] })); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssil.pipelines[SSIL_GENERATE_IMPORTANCE_MAP]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_ssil_pong_with_sampler), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_importance_map), 1); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssil.importance_map_push_constant, sizeof(SSILImportanceMapPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_ssil_buffers.half_buffer_width, p_ssil_buffers.half_buffer_height, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + // process Importance Map A + RD::Uniform u_importance_map_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssil_buffers.importance_map[0] })); + RD::Uniform u_importance_map_pong(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssil_buffers.importance_map[1] })); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssil.pipelines[SSIL_PROCESS_IMPORTANCE_MAPA]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_importance_map_with_sampler), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_importance_map_pong), 1); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssil.importance_map_push_constant, sizeof(SSILImportanceMapPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_ssil_buffers.half_buffer_width, p_ssil_buffers.half_buffer_height, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + // process Importance Map B + RD::Uniform u_importance_map_pong_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssil_buffers.importance_map[1] })); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssil.pipelines[SSIL_PROCESS_IMPORTANCE_MAPB]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_importance_map_pong_with_sampler), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_importance_map), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, ssil.counter_uniform_set, 2); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssil.importance_map_push_constant, sizeof(SSILImportanceMapPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_ssil_buffers.half_buffer_width, p_ssil_buffers.half_buffer_height, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + RD::get_singleton()->draw_command_end_label(); // Importance Map + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssil.pipelines[SSIL_GATHER_ADAPTIVE]); + } else { + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssil.pipelines[SSIL_GATHER]); + } + + gather_ssil(compute_list, p_ssil_buffers.deinterleaved_slices, p_ssil_buffers.edges_slices, p_settings, false, p_ssil_buffers.gather_uniform_set, p_ssil_buffers.importance_map_uniform_set, p_ssil_buffers.projection_uniform_set); + RD::get_singleton()->draw_command_end_label(); //Gather + } + + { + RD::get_singleton()->draw_command_begin_label("Edge Aware Blur"); + ssil.blur_push_constant.edge_sharpness = 1.0 - p_settings.sharpness; + ssil.blur_push_constant.half_screen_pixel_size[0] = 1.0 / p_ssil_buffers.buffer_width; + ssil.blur_push_constant.half_screen_pixel_size[1] = 1.0 / p_ssil_buffers.buffer_height; + + int blur_passes = ssil_quality > RS::ENV_SSIL_QUALITY_VERY_LOW ? ssil_blur_passes : 1; + + shader = ssil.blur_shader.version_get_shader(ssil.blur_shader_version, 0); + + for (int pass = 0; pass < blur_passes; pass++) { + int blur_pipeline = SSIL_BLUR_PASS; + if (ssil_quality > RS::ENV_SSIL_QUALITY_VERY_LOW) { + blur_pipeline = SSIL_BLUR_PASS_SMART; + if (pass < blur_passes - 2) { + blur_pipeline = SSIL_BLUR_PASS_WIDE; + } + } + + for (int i = 0; i < 4; i++) { + if ((ssil_quality == RS::ENV_SSIL_QUALITY_VERY_LOW) && ((i == 1) || (i == 2))) { + continue; + } + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssil.pipelines[blur_pipeline]); + if (pass % 2 == 0) { + if (ssil_quality == RS::ENV_SSIL_QUALITY_VERY_LOW) { + RD::Uniform u_ssil_slice(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssil_buffers.deinterleaved_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_ssil_slice), 0); + } else { + RD::Uniform u_ssil_slice(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ ss_effects.mirror_sampler, p_ssil_buffers.deinterleaved_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_ssil_slice), 0); + } + + RD::Uniform u_ssil_pong_slice(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssil_buffers.pong_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_ssil_pong_slice), 1); + } else { + if (ssil_quality == RS::ENV_SSIL_QUALITY_VERY_LOW) { + RD::Uniform u_ssil_pong_slice(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssil_buffers.pong_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_ssil_pong_slice), 0); + } else { + RD::Uniform u_ssil_pong_slice(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ ss_effects.mirror_sampler, p_ssil_buffers.pong_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_ssil_pong_slice), 0); + } + + RD::Uniform u_ssil_slice(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssil_buffers.deinterleaved_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_ssil_slice), 1); + } + + RD::Uniform u_edges_slice(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssil_buffers.edges_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 2, u_edges_slice), 2); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssil.blur_push_constant, sizeof(SSILBlurPushConstant)); + + int x_groups = (p_settings.full_screen_size.x >> (ssil_half_size ? 2 : 1)); + int y_groups = (p_settings.full_screen_size.y >> (ssil_half_size ? 2 : 1)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, x_groups, y_groups, 1); + if (ssil_quality > RS::ENV_SSIL_QUALITY_VERY_LOW) { + RD::get_singleton()->compute_list_add_barrier(compute_list); + } + } + } + + RD::get_singleton()->draw_command_end_label(); // Blur + } + + { + RD::get_singleton()->draw_command_begin_label("Interleave Buffers"); + ssil.interleave_push_constant.inv_sharpness = 1.0 - p_settings.sharpness; + ssil.interleave_push_constant.pixel_size[0] = 1.0 / p_settings.full_screen_size.x; + ssil.interleave_push_constant.pixel_size[1] = 1.0 / p_settings.full_screen_size.y; + ssil.interleave_push_constant.size_modifier = uint32_t(ssil_half_size ? 4 : 2); + + int interleave_pipeline = SSIL_INTERLEAVE_HALF; + if (ssil_quality == RS::ENV_SSIL_QUALITY_LOW) { + interleave_pipeline = SSIL_INTERLEAVE; + } else if (ssil_quality >= RS::ENV_SSIL_QUALITY_MEDIUM) { + interleave_pipeline = SSIL_INTERLEAVE_SMART; + } + + shader = ssil.interleave_shader.version_get_shader(ssil.interleave_shader_version, 0); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssil.pipelines[interleave_pipeline]); + + RD::Uniform u_destination(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssil_buffers.ssil_final })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_destination), 0); + + if (ssil_quality > RS::ENV_SSIL_QUALITY_VERY_LOW && ssil_blur_passes % 2 == 0) { + RD::Uniform u_ssil(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssil_buffers.deinterleaved })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_ssil), 1); + } else { + RD::Uniform u_ssil_pong(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssil_buffers.pong })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_ssil_pong), 1); + } + + RD::Uniform u_edges(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssil_buffers.edges })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 2, u_edges), 2); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssil.interleave_push_constant, sizeof(SSILInterleavePushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_settings.full_screen_size.x, p_settings.full_screen_size.y, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + RD::get_singleton()->draw_command_end_label(); // Interleave + } + + RD::get_singleton()->draw_command_end_label(); // SSIL + + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_NO_BARRIER); + + int zero[1] = { 0 }; + RD::get_singleton()->buffer_update(ssil.importance_map_load_counter, 0, sizeof(uint32_t), &zero, 0); //no barrier +} + +void SSEffects::ssil_free(SSILRenderBuffers &p_ssil_buffers) { + if (p_ssil_buffers.ssil_final.is_valid()) { + RD::get_singleton()->free(p_ssil_buffers.ssil_final); + RD::get_singleton()->free(p_ssil_buffers.deinterleaved); + RD::get_singleton()->free(p_ssil_buffers.pong); + RD::get_singleton()->free(p_ssil_buffers.edges); + RD::get_singleton()->free(p_ssil_buffers.importance_map[0]); + RD::get_singleton()->free(p_ssil_buffers.importance_map[1]); + RD::get_singleton()->free(p_ssil_buffers.last_frame); + + p_ssil_buffers.ssil_final = RID(); + p_ssil_buffers.deinterleaved = RID(); + p_ssil_buffers.pong = RID(); + p_ssil_buffers.edges = RID(); + p_ssil_buffers.deinterleaved_slices.clear(); + p_ssil_buffers.pong_slices.clear(); + p_ssil_buffers.edges_slices.clear(); + p_ssil_buffers.importance_map[0] = RID(); + p_ssil_buffers.importance_map[1] = RID(); + p_ssil_buffers.last_frame = RID(); + p_ssil_buffers.last_frame_slices.clear(); + + p_ssil_buffers.gather_uniform_set = RID(); + p_ssil_buffers.importance_map_uniform_set = RID(); + p_ssil_buffers.projection_uniform_set = RID(); + } +} + +/* SSAO */ + +void SSEffects::ssao_set_quality(RS::EnvironmentSSAOQuality p_quality, bool p_half_size, float p_adaptive_target, int p_blur_passes, float p_fadeout_from, float p_fadeout_to) { + ssao_quality = p_quality; + ssao_half_size = p_half_size; + ssao_adaptive_target = p_adaptive_target; + ssao_blur_passes = p_blur_passes; + ssao_fadeout_from = p_fadeout_from; + ssao_fadeout_to = p_fadeout_to; +} + +void SSEffects::gather_ssao(RD::ComputeListID p_compute_list, const Vector<RID> p_ao_slices, const SSAOSettings &p_settings, bool p_adaptive_base_pass, RID p_gather_uniform_set, RID p_importance_map_uniform_set) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + + RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, p_gather_uniform_set, 0); + if ((ssao_quality == RS::ENV_SSAO_QUALITY_ULTRA) && !p_adaptive_base_pass) { + RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, p_importance_map_uniform_set, 0); + } + + RID shader = ssao.gather_shader.version_get_shader(ssao.gather_shader_version, 1); // + + for (int i = 0; i < 4; i++) { + if ((ssao_quality == RS::ENV_SSAO_QUALITY_VERY_LOW) && ((i == 1) || (i == 2))) { + continue; + } + + RD::Uniform u_ao_slice(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ao_slices[i] })); + + ssao.gather_push_constant.pass_coord_offset[0] = i % 2; + ssao.gather_push_constant.pass_coord_offset[1] = i / 2; + ssao.gather_push_constant.pass_uv_offset[0] = ((i % 2) - 0.0) / p_settings.full_screen_size.x; + ssao.gather_push_constant.pass_uv_offset[1] = ((i / 2) - 0.0) / p_settings.full_screen_size.y; + ssao.gather_push_constant.pass = i; + RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, uniform_set_cache->get_cache(shader, 2, u_ao_slice), 2); + RD::get_singleton()->compute_list_set_push_constant(p_compute_list, &ssao.gather_push_constant, sizeof(SSAOGatherPushConstant)); + + Size2i size = Size2i(p_settings.full_screen_size.x >> (ssao_half_size ? 2 : 1), p_settings.full_screen_size.y >> (ssao_half_size ? 2 : 1)); + + RD::get_singleton()->compute_list_dispatch_threads(p_compute_list, size.x, size.y, 1); + } + RD::get_singleton()->compute_list_add_barrier(p_compute_list); +} + +void SSEffects::ssao_allocate_buffers(SSAORenderBuffers &p_ssao_buffers, const SSAOSettings &p_settings, RID p_linear_depth) { + if (p_ssao_buffers.half_size != ssao_half_size) { + ssao_free(p_ssao_buffers); + } + + if (ssao_half_size) { + p_ssao_buffers.buffer_width = (p_settings.full_screen_size.x + 3) / 4; + p_ssao_buffers.buffer_height = (p_settings.full_screen_size.y + 3) / 4; + p_ssao_buffers.half_buffer_width = (p_settings.full_screen_size.x + 7) / 8; + p_ssao_buffers.half_buffer_height = (p_settings.full_screen_size.y + 7) / 8; + } else { + p_ssao_buffers.buffer_width = (p_settings.full_screen_size.x + 1) / 2; + p_ssao_buffers.buffer_height = (p_settings.full_screen_size.y + 1) / 2; + p_ssao_buffers.half_buffer_width = (p_settings.full_screen_size.x + 3) / 4; + p_ssao_buffers.half_buffer_height = (p_settings.full_screen_size.y + 3) / 4; + } + + if (p_ssao_buffers.ao_deinterleaved.is_null()) { + { + p_ssao_buffers.depth_texture_view = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_linear_depth, 0, ssao_half_size ? 1 : 0, 4, RD::TEXTURE_SLICE_2D_ARRAY); + } + { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R8G8_UNORM; + tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; + tf.width = p_ssao_buffers.buffer_width; + tf.height = p_ssao_buffers.buffer_height; + tf.array_layers = 4; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + p_ssao_buffers.ao_deinterleaved = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssao_buffers.ao_deinterleaved, "SSAO De-interleaved Array"); + for (uint32_t i = 0; i < 4; i++) { + RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_ssao_buffers.ao_deinterleaved, i, 0); + p_ssao_buffers.ao_deinterleaved_slices.push_back(slice); + RD::get_singleton()->set_resource_name(slice, "SSAO De-interleaved Array Layer " + itos(i) + " "); + } + } + + { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R8G8_UNORM; + tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; + tf.width = p_ssao_buffers.buffer_width; + tf.height = p_ssao_buffers.buffer_height; + tf.array_layers = 4; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + p_ssao_buffers.ao_pong = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssao_buffers.ao_pong, "SSAO De-interleaved Array Pong"); + for (uint32_t i = 0; i < 4; i++) { + RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_ssao_buffers.ao_pong, i, 0); + p_ssao_buffers.ao_pong_slices.push_back(slice); + RD::get_singleton()->set_resource_name(slice, "SSAO De-interleaved Array Layer " + itos(i) + " Pong"); + } + } + + { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R8_UNORM; + tf.width = p_ssao_buffers.buffer_width; + tf.height = p_ssao_buffers.buffer_height; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + p_ssao_buffers.importance_map[0] = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssao_buffers.importance_map[0], "SSAO Importance Map"); + p_ssao_buffers.importance_map[1] = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssao_buffers.importance_map[1], "SSAO Importance Map Pong"); + } + { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R8_UNORM; + tf.width = p_settings.full_screen_size.x; + tf.height = p_settings.full_screen_size.y; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + p_ssao_buffers.ao_final = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssao_buffers.ao_final, "SSAO Final"); + } + p_ssao_buffers.half_size = ssao_half_size; + } +} + +void SSEffects::generate_ssao(SSAORenderBuffers &p_ssao_buffers, RID p_normal_buffer, const Projection &p_projection, const SSAOSettings &p_settings) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + memset(&ssao.gather_push_constant, 0, sizeof(SSAOGatherPushConstant)); + /* FIRST PASS */ + + RID shader = ssao.gather_shader.version_get_shader(ssao.gather_shader_version, 0); + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::get_singleton()->draw_command_begin_label("Process Screen Space Ambient Occlusion"); + /* SECOND PASS */ + // Sample SSAO + { + RD::get_singleton()->draw_command_begin_label("Gather Samples"); + ssao.gather_push_constant.screen_size[0] = p_settings.full_screen_size.x; + ssao.gather_push_constant.screen_size[1] = p_settings.full_screen_size.y; + + ssao.gather_push_constant.half_screen_pixel_size[0] = 1.0 / p_ssao_buffers.buffer_width; + ssao.gather_push_constant.half_screen_pixel_size[1] = 1.0 / p_ssao_buffers.buffer_height; + float tan_half_fov_x = 1.0 / p_projection.columns[0][0]; + float tan_half_fov_y = 1.0 / p_projection.columns[1][1]; + ssao.gather_push_constant.NDC_to_view_mul[0] = tan_half_fov_x * 2.0; + ssao.gather_push_constant.NDC_to_view_mul[1] = tan_half_fov_y * -2.0; + ssao.gather_push_constant.NDC_to_view_add[0] = tan_half_fov_x * -1.0; + ssao.gather_push_constant.NDC_to_view_add[1] = tan_half_fov_y; + ssao.gather_push_constant.is_orthogonal = p_projection.is_orthogonal(); + + ssao.gather_push_constant.half_screen_pixel_size_x025[0] = ssao.gather_push_constant.half_screen_pixel_size[0] * 0.25; + ssao.gather_push_constant.half_screen_pixel_size_x025[1] = ssao.gather_push_constant.half_screen_pixel_size[1] * 0.25; + + ssao.gather_push_constant.radius = p_settings.radius; + float radius_near_limit = (p_settings.radius * 1.2f); + if (ssao_quality <= RS::ENV_SSAO_QUALITY_LOW) { + radius_near_limit *= 1.50f; + + if (ssao_quality == RS::ENV_SSAO_QUALITY_VERY_LOW) { + ssao.gather_push_constant.radius *= 0.8f; + } + } + radius_near_limit /= tan_half_fov_y; + ssao.gather_push_constant.intensity = p_settings.intensity; + ssao.gather_push_constant.shadow_power = p_settings.power; + ssao.gather_push_constant.shadow_clamp = 0.98; + ssao.gather_push_constant.fade_out_mul = -1.0 / (ssao_fadeout_to - ssao_fadeout_from); + ssao.gather_push_constant.fade_out_add = ssao_fadeout_from / (ssao_fadeout_to - ssao_fadeout_from) + 1.0; + ssao.gather_push_constant.horizon_angle_threshold = p_settings.horizon; + ssao.gather_push_constant.inv_radius_near_limit = 1.0f / radius_near_limit; + ssao.gather_push_constant.neg_inv_radius = -1.0 / ssao.gather_push_constant.radius; + + ssao.gather_push_constant.load_counter_avg_div = 9.0 / float((p_ssao_buffers.half_buffer_width) * (p_ssao_buffers.half_buffer_height) * 255); + ssao.gather_push_constant.adaptive_sample_limit = ssao_adaptive_target; + + ssao.gather_push_constant.detail_intensity = p_settings.detail; + ssao.gather_push_constant.quality = MAX(0, ssao_quality - 1); + ssao.gather_push_constant.size_multiplier = ssao_half_size ? 2 : 1; + + if (p_ssao_buffers.gather_uniform_set.is_null()) { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; + u.binding = 0; + u.append_id(default_sampler); + u.append_id(p_ssao_buffers.depth_texture_view); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 1; + u.append_id(p_normal_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 2; + u.append_id(ss_effects.gather_constants_buffer); + uniforms.push_back(u); + } + p_ssao_buffers.gather_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, shader, 0); + RD::get_singleton()->set_resource_name(p_ssao_buffers.gather_uniform_set, "SSAO Gather Uniform Set"); + } + + if (p_ssao_buffers.importance_map_uniform_set.is_null()) { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 0; + u.append_id(p_ssao_buffers.ao_pong); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; + u.binding = 1; + u.append_id(default_sampler); + u.append_id(p_ssao_buffers.importance_map[0]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 2; + u.append_id(ssao.importance_map_load_counter); + uniforms.push_back(u); + } + p_ssao_buffers.importance_map_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssao.gather_shader.version_get_shader(ssao.gather_shader_version, 2), 1); + RD::get_singleton()->set_resource_name(p_ssao_buffers.importance_map_uniform_set, "SSAO Importance Map Uniform Set"); + } + + if (ssao_quality == RS::ENV_SSAO_QUALITY_ULTRA) { + RD::get_singleton()->draw_command_begin_label("Generate Importance Map"); + ssao.importance_map_push_constant.half_screen_pixel_size[0] = 1.0 / p_ssao_buffers.buffer_width; + ssao.importance_map_push_constant.half_screen_pixel_size[1] = 1.0 / p_ssao_buffers.buffer_height; + ssao.importance_map_push_constant.intensity = p_settings.intensity; + ssao.importance_map_push_constant.power = p_settings.power; + + //base pass + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GATHER_BASE]); + gather_ssao(compute_list, p_ssao_buffers.ao_pong_slices, p_settings, true, p_ssao_buffers.gather_uniform_set, RID()); + + //generate importance map + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GENERATE_IMPORTANCE_MAP]); + + RD::Uniform u_ao_pong_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssao_buffers.ao_pong })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_ao_pong_with_sampler), 0); + + RD::Uniform u_importance_map(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssao_buffers.importance_map[0] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_importance_map), 1); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.importance_map_push_constant, sizeof(SSAOImportanceMapPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_ssao_buffers.half_buffer_width, p_ssao_buffers.half_buffer_height, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + //process importance map A + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_PROCESS_IMPORTANCE_MAPA]); + + RD::Uniform u_importance_map_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssao_buffers.importance_map[0] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_importance_map_with_sampler), 0); + + RD::Uniform u_importance_map_pong(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssao_buffers.importance_map[1] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_importance_map_pong), 1); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.importance_map_push_constant, sizeof(SSAOImportanceMapPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_ssao_buffers.half_buffer_width, p_ssao_buffers.half_buffer_height, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + //process Importance Map B + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_PROCESS_IMPORTANCE_MAPB]); + + RD::Uniform u_importance_map_pong_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssao_buffers.importance_map[1] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_importance_map_pong_with_sampler), 0); + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_importance_map), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, ssao.counter_uniform_set, 2); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.importance_map_push_constant, sizeof(SSAOImportanceMapPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_ssao_buffers.half_buffer_width, p_ssao_buffers.half_buffer_height, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GATHER_ADAPTIVE]); + RD::get_singleton()->draw_command_end_label(); // Importance Map + } else { + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GATHER]); + } + + gather_ssao(compute_list, p_ssao_buffers.ao_deinterleaved_slices, p_settings, false, p_ssao_buffers.gather_uniform_set, p_ssao_buffers.importance_map_uniform_set); + RD::get_singleton()->draw_command_end_label(); // Gather SSAO + } + + // /* THIRD PASS */ + // // Blur + // + { + RD::get_singleton()->draw_command_begin_label("Edge Aware Blur"); + ssao.blur_push_constant.edge_sharpness = 1.0 - p_settings.sharpness; + ssao.blur_push_constant.half_screen_pixel_size[0] = 1.0 / p_ssao_buffers.buffer_width; + ssao.blur_push_constant.half_screen_pixel_size[1] = 1.0 / p_ssao_buffers.buffer_height; + + int blur_passes = ssao_quality > RS::ENV_SSAO_QUALITY_VERY_LOW ? ssao_blur_passes : 1; + + shader = ssao.blur_shader.version_get_shader(ssao.blur_shader_version, 0); + + for (int pass = 0; pass < blur_passes; pass++) { + int blur_pipeline = SSAO_BLUR_PASS; + if (ssao_quality > RS::ENV_SSAO_QUALITY_VERY_LOW) { + blur_pipeline = SSAO_BLUR_PASS_SMART; + if (pass < blur_passes - 2) { + blur_pipeline = SSAO_BLUR_PASS_WIDE; + } else { + blur_pipeline = SSAO_BLUR_PASS_SMART; + } + } + + for (int i = 0; i < 4; i++) { + if ((ssao_quality == RS::ENV_SSAO_QUALITY_VERY_LOW) && ((i == 1) || (i == 2))) { + continue; + } + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[blur_pipeline]); + if (pass % 2 == 0) { + if (ssao_quality == RS::ENV_SSAO_QUALITY_VERY_LOW) { + RD::Uniform u_ao_slices_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssao_buffers.ao_deinterleaved_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_ao_slices_with_sampler), 0); + } else { + RD::Uniform u_ao_slices_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ ss_effects.mirror_sampler, p_ssao_buffers.ao_deinterleaved_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_ao_slices_with_sampler), 0); + } + + RD::Uniform u_ao_pong_slices(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssao_buffers.ao_pong_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_ao_pong_slices), 1); + } else { + if (ssao_quality == RS::ENV_SSAO_QUALITY_VERY_LOW) { + RD::Uniform u_ao_pong_slices_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssao_buffers.ao_pong_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_ao_pong_slices_with_sampler), 0); + } else { + RD::Uniform u_ao_pong_slices_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ ss_effects.mirror_sampler, p_ssao_buffers.ao_pong_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_ao_pong_slices_with_sampler), 0); + } + + RD::Uniform u_ao_slices(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssao_buffers.ao_deinterleaved_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_ao_slices), 1); + } + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.blur_push_constant, sizeof(SSAOBlurPushConstant)); + + Size2i size(p_settings.full_screen_size.x >> (ssao_half_size ? 2 : 1), p_settings.full_screen_size.y >> (ssao_half_size ? 2 : 1)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, size.x, size.y, 1); + } + + if (ssao_quality > RS::ENV_SSAO_QUALITY_VERY_LOW) { + RD::get_singleton()->compute_list_add_barrier(compute_list); + } + } + RD::get_singleton()->draw_command_end_label(); // Blur + } + + /* FOURTH PASS */ + // Interleave buffers + // back to full size + { + RD::get_singleton()->draw_command_begin_label("Interleave Buffers"); + ssao.interleave_push_constant.inv_sharpness = 1.0 - p_settings.sharpness; + ssao.interleave_push_constant.pixel_size[0] = 1.0 / p_settings.full_screen_size.x; + ssao.interleave_push_constant.pixel_size[1] = 1.0 / p_settings.full_screen_size.y; + ssao.interleave_push_constant.size_modifier = uint32_t(ssao_half_size ? 4 : 2); + + shader = ssao.interleave_shader.version_get_shader(ssao.interleave_shader_version, 0); + + int interleave_pipeline = SSAO_INTERLEAVE_HALF; + if (ssao_quality == RS::ENV_SSAO_QUALITY_LOW) { + interleave_pipeline = SSAO_INTERLEAVE; + } else if (ssao_quality >= RS::ENV_SSAO_QUALITY_MEDIUM) { + interleave_pipeline = SSAO_INTERLEAVE_SMART; + } + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[interleave_pipeline]); + + RD::Uniform u_upscale_buffer(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssao_buffers.ao_final })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_upscale_buffer), 0); + + if (ssao_quality > RS::ENV_SSAO_QUALITY_VERY_LOW && ssao_blur_passes % 2 == 0) { + RD::Uniform u_ao(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssao_buffers.ao_deinterleaved })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_ao), 1); + } else { + RD::Uniform u_ao(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssao_buffers.ao_pong })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_ao), 1); + } + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.interleave_push_constant, sizeof(SSAOInterleavePushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_settings.full_screen_size.x, p_settings.full_screen_size.y, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + RD::get_singleton()->draw_command_end_label(); // Interleave + } + RD::get_singleton()->draw_command_end_label(); //SSAO + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_NO_BARRIER); //wait for upcoming transfer + + int zero[1] = { 0 }; + RD::get_singleton()->buffer_update(ssao.importance_map_load_counter, 0, sizeof(uint32_t), &zero, 0); //no barrier +} + +void SSEffects::ssao_free(SSAORenderBuffers &p_ssao_buffers) { + if (p_ssao_buffers.ao_final.is_valid()) { + RD::get_singleton()->free(p_ssao_buffers.ao_deinterleaved); + RD::get_singleton()->free(p_ssao_buffers.ao_pong); + RD::get_singleton()->free(p_ssao_buffers.ao_final); + + RD::get_singleton()->free(p_ssao_buffers.importance_map[0]); + RD::get_singleton()->free(p_ssao_buffers.importance_map[1]); + + p_ssao_buffers.ao_deinterleaved = RID(); + p_ssao_buffers.ao_pong = RID(); + p_ssao_buffers.ao_final = RID(); + p_ssao_buffers.importance_map[0] = RID(); + p_ssao_buffers.importance_map[1] = RID(); + p_ssao_buffers.ao_deinterleaved_slices.clear(); + p_ssao_buffers.ao_pong_slices.clear(); + + p_ssao_buffers.gather_uniform_set = RID(); + p_ssao_buffers.importance_map_uniform_set = RID(); + } +} + +/* Screen Space Reflection */ + +void SSEffects::ssr_set_roughness_quality(RS::EnvironmentSSRRoughnessQuality p_quality) { + ssr_roughness_quality = p_quality; +} + +void SSEffects::ssr_allocate_buffers(SSRRenderBuffers &p_ssr_buffers, const RenderingDevice::DataFormat p_color_format, const Size2i &p_screen_size, const uint32_t p_view_count) { + // As we are processing one view at a time, we can reuse buffers, only our output needs to have layers for each view. + if (p_ssr_buffers.size != p_screen_size || p_ssr_buffers.roughness_quality != ssr_roughness_quality) { + ssr_free(p_ssr_buffers); + } + + if (p_ssr_buffers.output.is_valid()) { + // already allocated + return; + } + + p_ssr_buffers.size = p_screen_size; + p_ssr_buffers.roughness_quality = ssr_roughness_quality; + + if (p_ssr_buffers.depth_scaled.is_null()) { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R32_SFLOAT; + tf.width = p_screen_size.x; + tf.height = p_screen_size.y; + tf.texture_type = RD::TEXTURE_TYPE_2D; + tf.array_layers = 1; + tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT; + + p_ssr_buffers.depth_scaled = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssr_buffers.depth_scaled, "SSR Depth Scaled"); + + tf.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + + p_ssr_buffers.normal_scaled = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssr_buffers.normal_scaled, "SSR Normal Scaled"); + } + + if (ssr_roughness_quality != RS::ENV_SSR_ROUGHNESS_QUALITY_DISABLED && !p_ssr_buffers.blur_radius[0].is_valid()) { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R8_UNORM; + tf.width = p_screen_size.x; + tf.height = p_screen_size.y; + tf.texture_type = RD::TEXTURE_TYPE_2D; + tf.array_layers = 1; + tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; + + p_ssr_buffers.blur_radius[0] = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssr_buffers.blur_radius[0], "SSR Blur Radius 0"); + p_ssr_buffers.blur_radius[1] = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssr_buffers.blur_radius[1], "SSR Blur Radius 1"); + } + + if (p_ssr_buffers.intermediate.is_null()) { + RD::TextureFormat tf; + tf.format = p_color_format; + tf.width = p_screen_size.x; + tf.height = p_screen_size.y; + tf.texture_type = RD::TEXTURE_TYPE_2D; + tf.array_layers = 1; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + + p_ssr_buffers.intermediate = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssr_buffers.intermediate, "SSR Intermediate"); + + if (p_view_count > 1) { + tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; + tf.array_layers = p_view_count; + } else { + tf.texture_type = RD::TEXTURE_TYPE_2D; + tf.array_layers = 1; + } + + p_ssr_buffers.output = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssr_buffers.output, "SSR Output"); + + for (uint32_t v = 0; v < p_view_count; v++) { + p_ssr_buffers.output_slices[v] = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_ssr_buffers.output, v, 0); + } + } +} + +void SSEffects::screen_space_reflection(SSRRenderBuffers &p_ssr_buffers, const RID *p_diffuse_slices, const RID *p_normal_roughness_slices, const RID *p_metallic_slices, const RID *p_depth_slices, const Size2i &p_screen_size, int p_max_steps, float p_fade_in, float p_fade_out, float p_tolerance, const uint32_t p_view_count, const Projection *p_projections, const Vector3 *p_eye_offsets) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + { + // Store some scene data in a UBO, in the near future we will use a UBO shared with other shaders + ScreenSpaceReflectionSceneData scene_data; + + if (ssr.ubo.is_null()) { + ssr.ubo = RD::get_singleton()->uniform_buffer_create(sizeof(ScreenSpaceReflectionSceneData)); + } + + for (uint32_t v = 0; v < p_view_count; v++) { + store_camera(p_projections[v], scene_data.projection[v]); + store_camera(p_projections[v].inverse(), scene_data.inv_projection[v]); + scene_data.eye_offset[v][0] = p_eye_offsets[v].x; + scene_data.eye_offset[v][1] = p_eye_offsets[v].y; + scene_data.eye_offset[v][2] = p_eye_offsets[v].z; + scene_data.eye_offset[v][3] = 0.0; + } + + RD::get_singleton()->buffer_update(ssr.ubo, 0, sizeof(ScreenSpaceReflectionSceneData), &scene_data, RD::BARRIER_MASK_COMPUTE); + } + + uint32_t pipeline_specialization = 0; + if (p_view_count > 1) { + pipeline_specialization |= SSR_MULTIVIEW; + } + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + + for (uint32_t v = 0; v < p_view_count; v++) { + RD::get_singleton()->draw_command_begin_label(String("SSR View ") + itos(v)); + + { //scale color and depth to half + RD::get_singleton()->draw_command_begin_label("SSR Scale"); + + ScreenSpaceReflectionScalePushConstant push_constant; + push_constant.view_index = v; + push_constant.camera_z_far = p_projections[v].get_z_far(); + push_constant.camera_z_near = p_projections[v].get_z_near(); + push_constant.orthogonal = p_projections[v].is_orthogonal(); + push_constant.filter = false; //enabling causes arctifacts + push_constant.screen_size[0] = p_screen_size.x; + push_constant.screen_size[1] = p_screen_size.y; + + RID shader = ssr_scale.shader.version_get_shader(ssr_scale.shader_version, 0); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssr_scale.pipelines[pipeline_specialization]); + + RD::Uniform u_diffuse(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_diffuse_slices[v] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_diffuse), 0); + + RD::Uniform u_depth(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_depth_slices[v] })); + RD::Uniform u_normal_roughness(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 1, Vector<RID>({ default_sampler, p_normal_roughness_slices[v] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_depth, u_normal_roughness), 1); + + if (ssr_roughness_quality != RS::ENV_SSR_ROUGHNESS_QUALITY_DISABLED) { + RD::Uniform u_output(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssr_buffers.output_slices[v] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 2, u_output), 2); + } else { + RD::Uniform u_intermediate(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssr_buffers.intermediate })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 2, u_intermediate), 2); + } + + RD::Uniform u_scale_depth(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssr_buffers.depth_scaled })); + RD::Uniform u_scale_normal(RD::UNIFORM_TYPE_IMAGE, 1, Vector<RID>({ p_ssr_buffers.normal_scaled })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 3, u_scale_depth, u_scale_normal), 3); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ScreenSpaceReflectionScalePushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); + + RD::get_singleton()->compute_list_add_barrier(compute_list); + + RD::get_singleton()->draw_command_end_label(); + } + + { + RD::get_singleton()->draw_command_begin_label("SSR main"); + + ScreenSpaceReflectionPushConstant push_constant; + push_constant.view_index = v; + push_constant.camera_z_far = p_projections[v].get_z_far(); + push_constant.camera_z_near = p_projections[v].get_z_near(); + push_constant.orthogonal = p_projections[v].is_orthogonal(); + push_constant.screen_size[0] = p_screen_size.x; + push_constant.screen_size[1] = p_screen_size.y; + push_constant.curve_fade_in = p_fade_in; + push_constant.distance_fade = p_fade_out; + push_constant.num_steps = p_max_steps; + push_constant.depth_tolerance = p_tolerance; + push_constant.use_half_res = true; + push_constant.proj_info[0] = -2.0f / (p_screen_size.width * p_projections[v].columns[0][0]); + push_constant.proj_info[1] = -2.0f / (p_screen_size.height * p_projections[v].columns[1][1]); + push_constant.proj_info[2] = (1.0f - p_projections[v].columns[0][2]) / p_projections[v].columns[0][0]; + push_constant.proj_info[3] = (1.0f + p_projections[v].columns[1][2]) / p_projections[v].columns[1][1]; + + ScreenSpaceReflectionMode mode = (ssr_roughness_quality != RS::ENV_SSR_ROUGHNESS_QUALITY_DISABLED) ? SCREEN_SPACE_REFLECTION_ROUGH : SCREEN_SPACE_REFLECTION_NORMAL; + RID shader = ssr.shader.version_get_shader(ssr.shader_version, mode); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssr.pipelines[pipeline_specialization][mode]); + + RD::Uniform u_scene_data(RD::UNIFORM_TYPE_UNIFORM_BUFFER, 0, ssr.ubo); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 4, u_scene_data), 4); + + if (ssr_roughness_quality != RS::ENV_SSR_ROUGHNESS_QUALITY_DISABLED) { + // read from output slices (our scale wrote into these) + RD::Uniform u_output(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssr_buffers.output_slices[v] })); + RD::Uniform u_scale_depth(RD::UNIFORM_TYPE_IMAGE, 1, Vector<RID>({ p_ssr_buffers.depth_scaled })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_output, u_scale_depth), 0); + + // write to intermediate (our roughness pass will output into output slices) + RD::Uniform u_intermediate(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssr_buffers.intermediate })); + RD::Uniform u_blur_radius(RD::UNIFORM_TYPE_IMAGE, 1, Vector<RID>({ p_ssr_buffers.blur_radius[0] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_intermediate, u_blur_radius), 1); + } else { + // read from intermediate (our scale wrote into these) + RD::Uniform u_intermediate(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssr_buffers.intermediate })); + RD::Uniform u_scale_depth(RD::UNIFORM_TYPE_IMAGE, 1, Vector<RID>({ p_ssr_buffers.depth_scaled })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_intermediate, u_scale_depth), 0); + + // We are not performing our blur so go directly to output. + RD::Uniform u_output(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssr_buffers.output_slices[v] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_output), 1); + } + + RD::Uniform u_scale_normal(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssr_buffers.normal_scaled })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 2, u_scale_normal), 2); + + RD::Uniform u_metallic(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_metallic_slices[v] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 3, u_metallic), 3); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ScreenSpaceReflectionPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); + + RD::get_singleton()->draw_command_end_label(); + } + + if (ssr_roughness_quality != RS::ENV_SSR_ROUGHNESS_QUALITY_DISABLED) { + RD::get_singleton()->draw_command_begin_label("SSR filter"); + //blur + + RD::get_singleton()->compute_list_add_barrier(compute_list); + + ScreenSpaceReflectionFilterPushConstant push_constant; + push_constant.view_index = v; + push_constant.orthogonal = p_projections[v].is_orthogonal(); + push_constant.edge_tolerance = Math::sin(Math::deg_to_rad(15.0)); + push_constant.proj_info[0] = -2.0f / (p_screen_size.width * p_projections[v].columns[0][0]); + push_constant.proj_info[1] = -2.0f / (p_screen_size.height * p_projections[v].columns[1][1]); + push_constant.proj_info[2] = (1.0f - p_projections[v].columns[0][2]) / p_projections[v].columns[0][0]; + push_constant.proj_info[3] = (1.0f + p_projections[v].columns[1][2]) / p_projections[v].columns[1][1]; + push_constant.vertical = 0; + if (ssr_roughness_quality == RS::ENV_SSR_ROUGHNESS_QUALITY_LOW) { + push_constant.steps = p_max_steps / 3; + push_constant.increment = 3; + } else if (ssr_roughness_quality == RS::ENV_SSR_ROUGHNESS_QUALITY_MEDIUM) { + push_constant.steps = p_max_steps / 2; + push_constant.increment = 2; + } else { + push_constant.steps = p_max_steps; + push_constant.increment = 1; + } + + push_constant.screen_size[0] = p_screen_size.width; + push_constant.screen_size[1] = p_screen_size.height; + + // Horizontal pass + + SSRReflectionMode mode = SCREEN_SPACE_REFLECTION_FILTER_HORIZONTAL; + + RID shader = ssr_filter.shader.version_get_shader(ssr_filter.shader_version, mode); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssr_filter.pipelines[pipeline_specialization][mode]); + + RD::Uniform u_scene_data(RD::UNIFORM_TYPE_UNIFORM_BUFFER, 0, ssr.ubo); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 4, u_scene_data), 4); + + RD::Uniform u_intermediate(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssr_buffers.intermediate })); + RD::Uniform u_blur_radius(RD::UNIFORM_TYPE_IMAGE, 1, Vector<RID>({ p_ssr_buffers.blur_radius[0] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_intermediate, u_blur_radius), 0); + + RD::Uniform u_scale_normal(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssr_buffers.normal_scaled })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_scale_normal), 1); + + RD::Uniform u_output_blur(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssr_buffers.output_slices[v] })); + RD::Uniform u_blur_radius2(RD::UNIFORM_TYPE_IMAGE, 1, Vector<RID>({ p_ssr_buffers.blur_radius[1] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 2, u_output_blur, u_blur_radius2), 2); + + RD::Uniform u_scale_depth(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssr_buffers.depth_scaled })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 3, u_scale_depth), 3); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ScreenSpaceReflectionFilterPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + // Vertical pass + + mode = SCREEN_SPACE_REFLECTION_FILTER_VERTICAL; + shader = ssr_filter.shader.version_get_shader(ssr_filter.shader_version, mode); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssr_filter.pipelines[pipeline_specialization][mode]); + + push_constant.vertical = 1; + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_output_blur, u_blur_radius2), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_scale_normal), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 2, u_intermediate), 2); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 3, u_scale_depth), 3); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 4, u_scene_data), 4); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ScreenSpaceReflectionFilterPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); + + if (v != p_view_count - 1) { + RD::get_singleton()->compute_list_add_barrier(compute_list); + } + + RD::get_singleton()->draw_command_end_label(); + } + + RD::get_singleton()->draw_command_end_label(); + } + + RD::get_singleton()->compute_list_end(); +} + +void SSEffects::ssr_free(SSRRenderBuffers &p_ssr_buffers) { + for (uint32_t v = 0; v < RendererSceneRender::MAX_RENDER_VIEWS; v++) { + p_ssr_buffers.output_slices[v] = RID(); + } + + if (p_ssr_buffers.output.is_valid()) { + RD::get_singleton()->free(p_ssr_buffers.output); + p_ssr_buffers.output = RID(); + } + + if (p_ssr_buffers.intermediate.is_valid()) { + RD::get_singleton()->free(p_ssr_buffers.intermediate); + p_ssr_buffers.intermediate = RID(); + } + + if (p_ssr_buffers.blur_radius[0].is_valid()) { + RD::get_singleton()->free(p_ssr_buffers.blur_radius[0]); + RD::get_singleton()->free(p_ssr_buffers.blur_radius[1]); + p_ssr_buffers.blur_radius[0] = RID(); + p_ssr_buffers.blur_radius[1] = RID(); + } + + if (p_ssr_buffers.depth_scaled.is_valid()) { + RD::get_singleton()->free(p_ssr_buffers.depth_scaled); + p_ssr_buffers.depth_scaled = RID(); + RD::get_singleton()->free(p_ssr_buffers.normal_scaled); + p_ssr_buffers.normal_scaled = RID(); + } +} + +/* Subsurface scattering */ + +void SSEffects::sss_set_quality(RS::SubSurfaceScatteringQuality p_quality) { + sss_quality = p_quality; +} + +RS::SubSurfaceScatteringQuality SSEffects::sss_get_quality() const { + return sss_quality; +} + +void SSEffects::sss_set_scale(float p_scale, float p_depth_scale) { + sss_scale = p_scale; + sss_depth_scale = p_depth_scale; +} + +void SSEffects::sub_surface_scattering(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_diffuse, RID p_depth, const Projection &p_camera, const Size2i &p_screen_size) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + // Our intermediate buffer is only created if we haven't created it already. + RD::DataFormat format = p_render_buffers->get_base_data_format(); + uint32_t usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + uint32_t layers = 1; // We only need one layer, we're handling one view at a time + uint32_t mipmaps = 1; // Image::get_image_required_mipmaps(p_screen_size.x, p_screen_size.y, Image::FORMAT_RGBAH); + RID intermediate = p_render_buffers->create_texture(SNAME("SSR"), SNAME("intermediate"), format, usage_bits, RD::TEXTURE_SAMPLES_1, p_screen_size, layers, mipmaps); + + Plane p = p_camera.xform4(Plane(1, 0, -1, 1)); + p.normal /= p.d; + float unit_size = p.normal.x; + + { //scale color and depth to half + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + + sss.push_constant.camera_z_far = p_camera.get_z_far(); + sss.push_constant.camera_z_near = p_camera.get_z_near(); + sss.push_constant.orthogonal = p_camera.is_orthogonal(); + sss.push_constant.unit_size = unit_size; + sss.push_constant.screen_size[0] = p_screen_size.x; + sss.push_constant.screen_size[1] = p_screen_size.y; + sss.push_constant.vertical = false; + sss.push_constant.scale = sss_scale; + sss.push_constant.depth_scale = sss_depth_scale; + + RID shader = sss.shader.version_get_shader(sss.shader_version, sss_quality - 1); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sss.pipelines[sss_quality - 1]); + + RD::Uniform u_diffuse_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_diffuse })); + RD::Uniform u_diffuse(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_diffuse })); + RD::Uniform u_intermediate_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, intermediate })); + RD::Uniform u_intermediate(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ intermediate })); + RD::Uniform u_depth_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_depth })); + + // horizontal + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_diffuse_with_sampler), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_intermediate), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 2, u_depth_with_sampler), 2); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &sss.push_constant, sizeof(SubSurfaceScatteringPushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); + + RD::get_singleton()->compute_list_add_barrier(compute_list); + + // vertical + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_intermediate_with_sampler), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_diffuse), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 2, u_depth_with_sampler), 2); + + sss.push_constant.vertical = true; + RD::get_singleton()->compute_list_set_push_constant(compute_list, &sss.push_constant, sizeof(SubSurfaceScatteringPushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); + + RD::get_singleton()->compute_list_end(); + } +} diff --git a/servers/rendering/renderer_rd/effects/ss_effects.h b/servers/rendering/renderer_rd/effects/ss_effects.h new file mode 100644 index 0000000000..dfaf3881bb --- /dev/null +++ b/servers/rendering/renderer_rd/effects/ss_effects.h @@ -0,0 +1,552 @@ +/*************************************************************************/ +/* ss_effects.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef SS_EFFECTS_RD_H +#define SS_EFFECTS_RD_H + +#include "servers/rendering/renderer_rd/pipeline_cache_rd.h" +#include "servers/rendering/renderer_rd/shaders/effects/screen_space_reflection.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/screen_space_reflection_filter.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/screen_space_reflection_scale.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/ss_effects_downsample.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/ssao.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/ssao_blur.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/ssao_importance_map.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/ssao_interleave.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/ssil.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/ssil_blur.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/ssil_importance_map.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/ssil_interleave.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/subsurface_scattering.glsl.gen.h" +#include "servers/rendering/renderer_scene_render.h" +#include "servers/rendering_server.h" + +class RenderSceneBuffersRD; + +namespace RendererRD { + +class SSEffects { +private: + static SSEffects *singleton; + +public: + static SSEffects *get_singleton() { return singleton; } + + SSEffects(); + ~SSEffects(); + + /* SS Downsampler */ + + void downsample_depth(RID p_depth_buffer, const Vector<RID> &p_depth_mipmaps, bool p_invalidate_uniform_set, Size2i p_full_screen_size, const Projection &p_projection); + + /* SSIL */ + void ssil_set_quality(RS::EnvironmentSSILQuality p_quality, bool p_half_size, float p_adaptive_target, int p_blur_passes, float p_fadeout_from, float p_fadeout_to); + + struct SSILRenderBuffers { + bool half_size = false; + int buffer_width; + int buffer_height; + int half_buffer_width; + int half_buffer_height; + + RID ssil_final; + RID deinterleaved; + Vector<RID> deinterleaved_slices; + RID pong; + Vector<RID> pong_slices; + RID edges; + Vector<RID> edges_slices; + RID importance_map[2]; + RID depth_texture_view; + + RID last_frame; + Vector<RID> last_frame_slices; + + RID gather_uniform_set; + RID importance_map_uniform_set; + RID projection_uniform_set; + }; + + struct SSILSettings { + float radius = 1.0; + float intensity = 2.0; + float sharpness = 0.98; + float normal_rejection = 1.0; + + Size2i full_screen_size = Size2i(); + }; + + void ssil_allocate_buffers(SSILRenderBuffers &p_ssil_buffers, const SSILSettings &p_settings, RID p_linear_depth); + void screen_space_indirect_lighting(SSILRenderBuffers &p_ssil_buffers, RID p_normal_buffer, const Projection &p_projection, const Projection &p_last_projection, const SSILSettings &p_settings); + void ssil_free(SSILRenderBuffers &p_ssil_buffers); + + /* SSAO */ + void ssao_set_quality(RS::EnvironmentSSAOQuality p_quality, bool p_half_size, float p_adaptive_target, int p_blur_passes, float p_fadeout_from, float p_fadeout_to); + + struct SSAORenderBuffers { + bool half_size = false; + int buffer_width; + int buffer_height; + int half_buffer_width; + int half_buffer_height; + + RID ao_deinterleaved; + Vector<RID> ao_deinterleaved_slices; + RID ao_pong; + Vector<RID> ao_pong_slices; + RID ao_final; + RID importance_map[2]; + RID depth_texture_view; + + RID gather_uniform_set; + RID importance_map_uniform_set; + }; + + struct SSAOSettings { + float radius = 1.0; + float intensity = 2.0; + float power = 1.5; + float detail = 0.5; + float horizon = 0.06; + float sharpness = 0.98; + + Size2i full_screen_size = Size2i(); + }; + + void ssao_allocate_buffers(SSAORenderBuffers &p_ssao_buffers, const SSAOSettings &p_settings, RID p_linear_depth); + void generate_ssao(SSAORenderBuffers &p_ssao_buffers, RID p_normal_buffer, const Projection &p_projection, const SSAOSettings &p_settings); + void ssao_free(SSAORenderBuffers &p_ssao_buffers); + + /* Screen Space Reflection */ + void ssr_set_roughness_quality(RS::EnvironmentSSRRoughnessQuality p_quality); + + struct SSRRenderBuffers { + Size2i size; + RenderingServer::EnvironmentSSRRoughnessQuality roughness_quality = RenderingServer::ENV_SSR_ROUGHNESS_QUALITY_DISABLED; + + RID normal_scaled; + RID depth_scaled; + RID blur_radius[2]; + RID intermediate; + RID output; + RID output_slices[RendererSceneRender::MAX_RENDER_VIEWS]; + }; + + void ssr_allocate_buffers(SSRRenderBuffers &p_ssr_buffers, const RenderingDevice::DataFormat p_color_format, const Size2i &p_screen_size, const uint32_t p_view_count); + void screen_space_reflection(SSRRenderBuffers &p_ssr_buffers, const RID *p_diffuse_slices, const RID *p_normal_roughness_slices, const RID *p_metallic_slices, const RID *p_depth_slices, const Size2i &p_screen_size, int p_max_steps, float p_fade_in, float p_fade_out, float p_tolerance, const uint32_t p_view_count, const Projection *p_projections, const Vector3 *p_eye_offsets); + void ssr_free(SSRRenderBuffers &p_ssr_buffers); + + /* subsurface scattering */ + void sss_set_quality(RS::SubSurfaceScatteringQuality p_quality); + RS::SubSurfaceScatteringQuality sss_get_quality() const; + void sss_set_scale(float p_scale, float p_depth_scale); + + void sub_surface_scattering(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_diffuse, RID p_depth, const Projection &p_camera, const Size2i &p_screen_size); + +private: + /* Settings */ + + RS::EnvironmentSSAOQuality ssao_quality = RS::ENV_SSAO_QUALITY_MEDIUM; + bool ssao_half_size = false; + float ssao_adaptive_target = 0.5; + int ssao_blur_passes = 2; + float ssao_fadeout_from = 50.0; + float ssao_fadeout_to = 300.0; + + RS::EnvironmentSSILQuality ssil_quality = RS::ENV_SSIL_QUALITY_MEDIUM; + bool ssil_half_size = false; + float ssil_adaptive_target = 0.5; + int ssil_blur_passes = 4; + float ssil_fadeout_from = 50.0; + float ssil_fadeout_to = 300.0; + + RS::EnvironmentSSRRoughnessQuality ssr_roughness_quality = RS::ENV_SSR_ROUGHNESS_QUALITY_LOW; + + RS::SubSurfaceScatteringQuality sss_quality = RS::SUB_SURFACE_SCATTERING_QUALITY_MEDIUM; + float sss_scale = 0.05; + float sss_depth_scale = 0.01; + + /* SS Downsampler */ + + struct SSEffectsDownsamplePushConstant { + float pixel_size[2]; + float z_far; + float z_near; + uint32_t orthogonal; + float radius_sq; + uint32_t pad[2]; + }; + + enum SSEffectsMode { + SS_EFFECTS_DOWNSAMPLE, + SS_EFFECTS_DOWNSAMPLE_HALF_RES, + SS_EFFECTS_DOWNSAMPLE_MIPMAP, + SS_EFFECTS_DOWNSAMPLE_MIPMAP_HALF_RES, + SS_EFFECTS_DOWNSAMPLE_HALF, + SS_EFFECTS_DOWNSAMPLE_HALF_RES_HALF, + SS_EFFECTS_DOWNSAMPLE_FULL_MIPS, + SS_EFFECTS_MAX + }; + + struct SSEffectsGatherConstants { + float rotation_matrices[80]; //5 vec4s * 4 + }; + + struct SSEffectsShader { + SSEffectsDownsamplePushConstant downsample_push_constant; + SsEffectsDownsampleShaderRD downsample_shader; + RID downsample_shader_version; + RID downsample_uniform_set; + bool used_half_size_last_frame = false; + bool used_mips_last_frame = false; + bool used_full_mips_last_frame = false; + + RID gather_constants_buffer; + + RID mirror_sampler; + + RID pipelines[SS_EFFECTS_MAX]; + } ss_effects; + + /* SSIL */ + + enum SSILMode { + SSIL_GATHER, + SSIL_GATHER_BASE, + SSIL_GATHER_ADAPTIVE, + SSIL_GENERATE_IMPORTANCE_MAP, + SSIL_PROCESS_IMPORTANCE_MAPA, + SSIL_PROCESS_IMPORTANCE_MAPB, + SSIL_BLUR_PASS, + SSIL_BLUR_PASS_SMART, + SSIL_BLUR_PASS_WIDE, + SSIL_INTERLEAVE, + SSIL_INTERLEAVE_SMART, + SSIL_INTERLEAVE_HALF, + SSIL_MAX + }; + + struct SSILGatherPushConstant { + int32_t screen_size[2]; + int pass; + int quality; + + float half_screen_pixel_size[2]; + float half_screen_pixel_size_x025[2]; + + float NDC_to_view_mul[2]; + float NDC_to_view_add[2]; + + float pad2[2]; + float z_near; + float z_far; + + float radius; + float intensity; + int size_multiplier; + int pad; + + float fade_out_mul; + float fade_out_add; + float normal_rejection_amount; + float inv_radius_near_limit; + + uint32_t is_orthogonal; + float neg_inv_radius; + float load_counter_avg_div; + float adaptive_sample_limit; + + int32_t pass_coord_offset[2]; + float pass_uv_offset[2]; + }; + + struct SSILImportanceMapPushConstant { + float half_screen_pixel_size[2]; + float intensity; + float pad; + }; + + struct SSILBlurPushConstant { + float edge_sharpness; + float pad; + float half_screen_pixel_size[2]; + }; + + struct SSILInterleavePushConstant { + float inv_sharpness; + uint32_t size_modifier; + float pixel_size[2]; + }; + + struct SSILProjectionUniforms { + float inv_last_frame_projection_matrix[16]; + }; + + struct SSIL { + SSILGatherPushConstant gather_push_constant; + SsilShaderRD gather_shader; + RID gather_shader_version; + RID projection_uniform_buffer; + + SSILImportanceMapPushConstant importance_map_push_constant; + SsilImportanceMapShaderRD importance_map_shader; + RID importance_map_shader_version; + RID importance_map_load_counter; + RID counter_uniform_set; + + SSILBlurPushConstant blur_push_constant; + SsilBlurShaderRD blur_shader; + RID blur_shader_version; + + SSILInterleavePushConstant interleave_push_constant; + SsilInterleaveShaderRD interleave_shader; + RID interleave_shader_version; + + RID pipelines[SSIL_MAX]; + } ssil; + + void gather_ssil(RD::ComputeListID p_compute_list, const Vector<RID> p_ssil_slices, const Vector<RID> p_edges_slices, const SSILSettings &p_settings, bool p_adaptive_base_pass, RID p_gather_uniform_set, RID p_importance_map_uniform_set, RID p_projection_uniform_set); + + /* SSAO */ + + enum SSAOMode { + SSAO_GATHER, + SSAO_GATHER_BASE, + SSAO_GATHER_ADAPTIVE, + SSAO_GENERATE_IMPORTANCE_MAP, + SSAO_PROCESS_IMPORTANCE_MAPA, + SSAO_PROCESS_IMPORTANCE_MAPB, + SSAO_BLUR_PASS, + SSAO_BLUR_PASS_SMART, + SSAO_BLUR_PASS_WIDE, + SSAO_INTERLEAVE, + SSAO_INTERLEAVE_SMART, + SSAO_INTERLEAVE_HALF, + SSAO_MAX + }; + + struct SSAOGatherPushConstant { + int32_t screen_size[2]; + int pass; + int quality; + + float half_screen_pixel_size[2]; + int size_multiplier; + float detail_intensity; + + float NDC_to_view_mul[2]; + float NDC_to_view_add[2]; + + float pad[2]; + float half_screen_pixel_size_x025[2]; + + float radius; + float intensity; + float shadow_power; + float shadow_clamp; + + float fade_out_mul; + float fade_out_add; + float horizon_angle_threshold; + float inv_radius_near_limit; + + uint32_t is_orthogonal; + float neg_inv_radius; + float load_counter_avg_div; + float adaptive_sample_limit; + + int32_t pass_coord_offset[2]; + float pass_uv_offset[2]; + }; + + struct SSAOImportanceMapPushConstant { + float half_screen_pixel_size[2]; + float intensity; + float power; + }; + + struct SSAOBlurPushConstant { + float edge_sharpness; + float pad; + float half_screen_pixel_size[2]; + }; + + struct SSAOInterleavePushConstant { + float inv_sharpness; + uint32_t size_modifier; + float pixel_size[2]; + }; + + struct SSAO { + SSAOGatherPushConstant gather_push_constant; + SsaoShaderRD gather_shader; + RID gather_shader_version; + + SSAOImportanceMapPushConstant importance_map_push_constant; + SsaoImportanceMapShaderRD importance_map_shader; + RID importance_map_shader_version; + RID importance_map_load_counter; + RID counter_uniform_set; + + SSAOBlurPushConstant blur_push_constant; + SsaoBlurShaderRD blur_shader; + RID blur_shader_version; + + SSAOInterleavePushConstant interleave_push_constant; + SsaoInterleaveShaderRD interleave_shader; + RID interleave_shader_version; + + RID pipelines[SSAO_MAX]; + } ssao; + + void gather_ssao(RD::ComputeListID p_compute_list, const Vector<RID> p_ao_slices, const SSAOSettings &p_settings, bool p_adaptive_base_pass, RID p_gather_uniform_set, RID p_importance_map_uniform_set); + + /* Screen Space Reflection */ + + enum SSRShaderSpecializations { + SSR_MULTIVIEW = 1 << 0, + SSR_VARIATIONS = 2, + }; + + struct ScreenSpaceReflectionSceneData { + float projection[2][16]; + float inv_projection[2][16]; + float eye_offset[2][4]; + }; + + // SSR Scale + + struct ScreenSpaceReflectionScalePushConstant { + int32_t screen_size[2]; + float camera_z_near; + float camera_z_far; + + uint32_t orthogonal; + uint32_t filter; + uint32_t view_index; + uint32_t pad1; + }; + + struct ScreenSpaceReflectionScale { + ScreenSpaceReflectionScaleShaderRD shader; + RID shader_version; + RID pipelines[SSR_VARIATIONS]; + } ssr_scale; + + // SSR main + + enum ScreenSpaceReflectionMode { + SCREEN_SPACE_REFLECTION_NORMAL, + SCREEN_SPACE_REFLECTION_ROUGH, + SCREEN_SPACE_REFLECTION_MAX, + }; + + struct ScreenSpaceReflectionPushConstant { + float proj_info[4]; // 16 - 16 + + int32_t screen_size[2]; // 8 - 24 + float camera_z_near; // 4 - 28 + float camera_z_far; // 4 - 32 + + int32_t num_steps; // 4 - 36 + float depth_tolerance; // 4 - 40 + float distance_fade; // 4 - 44 + float curve_fade_in; // 4 - 48 + + uint32_t orthogonal; // 4 - 52 + float filter_mipmap_levels; // 4 - 56 + uint32_t use_half_res; // 4 - 60 + uint32_t view_index; // 4 - 64 + + // float projection[16]; // this is in our ScreenSpaceReflectionSceneData now + }; + + struct ScreenSpaceReflection { + ScreenSpaceReflectionShaderRD shader; + RID shader_version; + RID pipelines[SSR_VARIATIONS][SCREEN_SPACE_REFLECTION_MAX]; + + RID ubo; + } ssr; + + // SSR Filter + + struct ScreenSpaceReflectionFilterPushConstant { + float proj_info[4]; // 16 - 16 + + uint32_t orthogonal; // 4 - 20 + float edge_tolerance; // 4 - 24 + int32_t increment; // 4 - 28 + uint32_t view_index; // 4 - 32 + + int32_t screen_size[2]; // 8 - 40 + uint32_t vertical; // 4 - 44 + uint32_t steps; // 4 - 48 + }; + + enum SSRReflectionMode { + SCREEN_SPACE_REFLECTION_FILTER_HORIZONTAL, + SCREEN_SPACE_REFLECTION_FILTER_VERTICAL, + SCREEN_SPACE_REFLECTION_FILTER_MAX, + }; + + struct ScreenSpaceReflectionFilter { + ScreenSpaceReflectionFilterShaderRD shader; + RID shader_version; + RID pipelines[SSR_VARIATIONS][SCREEN_SPACE_REFLECTION_FILTER_MAX]; + } ssr_filter; + + /* Subsurface scattering */ + + struct SubSurfaceScatteringPushConstant { + int32_t screen_size[2]; + float camera_z_far; + float camera_z_near; + + uint32_t vertical; + uint32_t orthogonal; + float unit_size; + float scale; + + float depth_scale; + uint32_t pad[3]; + }; + + struct SubSurfaceScattering { + SubSurfaceScatteringPushConstant push_constant; + SubsurfaceScatteringShaderRD shader; + RID shader_version; + RID pipelines[3]; //3 quality levels + } sss; +}; + +} // namespace RendererRD + +#endif // SS_EFFECTS_RD_H diff --git a/servers/rendering/renderer_rd/effects/taa.cpp b/servers/rendering/renderer_rd/effects/taa.cpp new file mode 100644 index 0000000000..657385a509 --- /dev/null +++ b/servers/rendering/renderer_rd/effects/taa.cpp @@ -0,0 +1,138 @@ +/*************************************************************************/ +/* taa.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "taa.h" +#include "servers/rendering/renderer_rd/effects/copy_effects.h" +#include "servers/rendering/renderer_rd/storage_rd/material_storage.h" +#include "servers/rendering/renderer_rd/uniform_set_cache_rd.h" + +using namespace RendererRD; + +TAA::TAA() { + Vector<String> taa_modes; + taa_modes.push_back("\n#define MODE_TAA_RESOLVE"); + taa_shader.initialize(taa_modes); + shader_version = taa_shader.version_create(); + pipeline = RD::get_singleton()->compute_pipeline_create(taa_shader.version_get_shader(shader_version, 0)); +} + +TAA::~TAA() { + taa_shader.version_free(shader_version); +} + +void TAA::msaa_resolve(Ref<RenderSceneBuffersRD> p_render_buffers) { + if (!p_render_buffers->has_velocity_buffer(true)) { + // nothing to resolve + return; + } + + for (uint32_t v = 0; v < p_render_buffers->get_view_count(); v++) { + RID velocity_buffer_msaa = p_render_buffers->get_velocity_buffer(true, v); + RID velocity_buffer = p_render_buffers->get_velocity_buffer(false, v); + + RD::get_singleton()->texture_resolve_multisample(velocity_buffer_msaa, velocity_buffer); + } +} + +void TAA::resolve(RID p_frame, RID p_temp, RID p_depth, RID p_velocity, RID p_prev_velocity, RID p_history, Size2 p_resolution, float p_z_near, float p_z_far) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + RID shader = taa_shader.version_get_shader(shader_version, 0); + ERR_FAIL_COND(shader.is_null()); + + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + TAAResolvePushConstant push_constant; + memset(&push_constant, 0, sizeof(TAAResolvePushConstant)); + push_constant.resolution_width = p_resolution.width; + push_constant.resolution_height = p_resolution.height; + push_constant.disocclusion_threshold = 0.025f; + push_constant.disocclusion_scale = 10.0f; + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, pipeline); + + RD::Uniform u_frame_source(RD::UNIFORM_TYPE_IMAGE, 0, { p_frame }); + RD::Uniform u_depth(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 1, { default_sampler, p_depth }); + RD::Uniform u_velocity(RD::UNIFORM_TYPE_IMAGE, 2, { p_velocity }); + RD::Uniform u_prev_velocity(RD::UNIFORM_TYPE_IMAGE, 3, { p_prev_velocity }); + RD::Uniform u_history(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 4, { default_sampler, p_history }); + RD::Uniform u_frame_dest(RD::UNIFORM_TYPE_IMAGE, 5, { p_temp }); + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_frame_source, u_depth, u_velocity, u_prev_velocity, u_history, u_frame_dest), 0); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(TAAResolvePushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_resolution.width, p_resolution.height, 1); + RD::get_singleton()->compute_list_end(); +} + +void TAA::process(Ref<RenderSceneBuffersRD> p_render_buffers, RD::DataFormat p_format, float p_z_near, float p_z_far) { + CopyEffects *copy_effects = CopyEffects::get_singleton(); + + uint32_t view_count = p_render_buffers->get_view_count(); + Size2i internal_size = p_render_buffers->get_internal_size(); + Size2i target_size = p_render_buffers->get_target_size(); + + bool just_allocated = false; + if (!p_render_buffers->has_texture(SNAME("taa"), SNAME("history"))) { + uint32_t usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + + p_render_buffers->create_texture(SNAME("taa"), SNAME("history"), p_format, usage_bits); + p_render_buffers->create_texture(SNAME("taa"), SNAME("temp"), p_format, usage_bits); + + p_render_buffers->create_texture(SNAME("taa"), SNAME("prev_velocity"), RD::DATA_FORMAT_R16G16_SFLOAT, usage_bits); + + just_allocated = true; + } + + RD::get_singleton()->draw_command_begin_label("TAA"); + + for (uint32_t v = 0; v < view_count; v++) { + // Get our (cached) slices + RID internal_texture = p_render_buffers->get_internal_texture(v); + RID velocity_buffer = p_render_buffers->get_velocity_buffer(false, v); + RID taa_history = p_render_buffers->get_texture_slice(SNAME("taa"), SNAME("history"), v, 0); + RID taa_prev_velocity = p_render_buffers->get_texture_slice(SNAME("taa"), SNAME("prev_velocity"), v, 0); + + if (!just_allocated) { + RID depth_texture = p_render_buffers->get_depth_texture(v); + RID taa_temp = p_render_buffers->get_texture_slice(SNAME("taa"), SNAME("temp"), v, 0); + resolve(internal_texture, taa_temp, depth_texture, velocity_buffer, taa_prev_velocity, taa_history, Size2(internal_size.x, internal_size.y), p_z_near, p_z_far); + copy_effects->copy_to_rect(taa_temp, internal_texture, Rect2(0, 0, internal_size.x, internal_size.y)); + } + + copy_effects->copy_to_rect(internal_texture, taa_history, Rect2(0, 0, internal_size.x, internal_size.y)); + copy_effects->copy_to_rect(velocity_buffer, taa_prev_velocity, Rect2(0, 0, target_size.x, target_size.y)); + } + + RD::get_singleton()->draw_command_end_label(); +} diff --git a/servers/rendering/renderer_rd/effects/taa.h b/servers/rendering/renderer_rd/effects/taa.h new file mode 100644 index 0000000000..ce4af18866 --- /dev/null +++ b/servers/rendering/renderer_rd/effects/taa.h @@ -0,0 +1,68 @@ +/*************************************************************************/ +/* taa.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef TAA_RD_H +#define TAA_RD_H + +#include "servers/rendering/renderer_rd/pipeline_cache_rd.h" +#include "servers/rendering/renderer_rd/shaders/effects/taa_resolve.glsl.gen.h" +#include "servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.h" +#include "servers/rendering/renderer_scene_render.h" + +#include "servers/rendering_server.h" + +namespace RendererRD { + +class TAA { +public: + TAA(); + ~TAA(); + + void msaa_resolve(Ref<RenderSceneBuffersRD> p_render_buffers); + void process(Ref<RenderSceneBuffersRD> p_render_buffers, RD::DataFormat p_format, float p_z_near, float p_z_far); + +private: + struct TAAResolvePushConstant { + float resolution_width; + float resolution_height; + float disocclusion_threshold; + float disocclusion_scale; + }; + + TaaResolveShaderRD taa_shader; + RID shader_version; + RID pipeline; + + void resolve(RID p_frame, RID p_temp, RID p_depth, RID p_velocity, RID p_prev_velocity, RID p_history, Size2 p_resolution, float p_z_near, float p_z_far); +}; + +} // namespace RendererRD + +#endif // TAA_RD_H diff --git a/servers/rendering/renderer_rd/effects/tone_mapper.cpp b/servers/rendering/renderer_rd/effects/tone_mapper.cpp new file mode 100644 index 0000000000..3a47b1420b --- /dev/null +++ b/servers/rendering/renderer_rd/effects/tone_mapper.cpp @@ -0,0 +1,257 @@ +/*************************************************************************/ +/* tone_mapper.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "tone_mapper.h" +#include "servers/rendering/renderer_rd/renderer_compositor_rd.h" +#include "servers/rendering/renderer_rd/storage_rd/material_storage.h" +#include "servers/rendering/renderer_rd/uniform_set_cache_rd.h" + +using namespace RendererRD; + +ToneMapper::ToneMapper() { + { + // Initialize tonemapper + Vector<String> tonemap_modes; + tonemap_modes.push_back("\n"); + tonemap_modes.push_back("\n#define USE_GLOW_FILTER_BICUBIC\n"); + tonemap_modes.push_back("\n#define USE_1D_LUT\n"); + tonemap_modes.push_back("\n#define USE_GLOW_FILTER_BICUBIC\n#define USE_1D_LUT\n"); + tonemap_modes.push_back("\n#define SUBPASS\n"); + tonemap_modes.push_back("\n#define SUBPASS\n#define USE_1D_LUT\n"); + + // multiview versions of our shaders + tonemap_modes.push_back("\n#define MULTIVIEW\n"); + tonemap_modes.push_back("\n#define MULTIVIEW\n#define USE_GLOW_FILTER_BICUBIC\n"); + tonemap_modes.push_back("\n#define MULTIVIEW\n#define USE_1D_LUT\n"); + tonemap_modes.push_back("\n#define MULTIVIEW\n#define USE_GLOW_FILTER_BICUBIC\n#define USE_1D_LUT\n"); + tonemap_modes.push_back("\n#define MULTIVIEW\n#define SUBPASS\n"); + tonemap_modes.push_back("\n#define MULTIVIEW\n#define SUBPASS\n#define USE_1D_LUT\n"); + + tonemap.shader.initialize(tonemap_modes); + + if (!RendererCompositorRD::singleton->is_xr_enabled()) { + tonemap.shader.set_variant_enabled(TONEMAP_MODE_NORMAL_MULTIVIEW, false); + tonemap.shader.set_variant_enabled(TONEMAP_MODE_BICUBIC_GLOW_FILTER_MULTIVIEW, false); + tonemap.shader.set_variant_enabled(TONEMAP_MODE_1D_LUT_MULTIVIEW, false); + tonemap.shader.set_variant_enabled(TONEMAP_MODE_BICUBIC_GLOW_FILTER_1D_LUT_MULTIVIEW, false); + tonemap.shader.set_variant_enabled(TONEMAP_MODE_SUBPASS_MULTIVIEW, false); + tonemap.shader.set_variant_enabled(TONEMAP_MODE_SUBPASS_1D_LUT_MULTIVIEW, false); + } + + tonemap.shader_version = tonemap.shader.version_create(); + + for (int i = 0; i < TONEMAP_MODE_MAX; i++) { + if (tonemap.shader.is_variant_enabled(i)) { + tonemap.pipelines[i].setup(tonemap.shader.version_get_shader(tonemap.shader_version, i), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RD::PipelineColorBlendState::create_disabled(), 0); + } else { + tonemap.pipelines[i].clear(); + } + } + } +} + +ToneMapper::~ToneMapper() { + tonemap.shader.version_free(tonemap.shader_version); +} + +void ToneMapper::tonemapper(RID p_source_color, RID p_dst_framebuffer, const TonemapSettings &p_settings) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + memset(&tonemap.push_constant, 0, sizeof(TonemapPushConstant)); + + tonemap.push_constant.use_bcs = p_settings.use_bcs; + tonemap.push_constant.bcs[0] = p_settings.brightness; + tonemap.push_constant.bcs[1] = p_settings.contrast; + tonemap.push_constant.bcs[2] = p_settings.saturation; + + tonemap.push_constant.use_glow = p_settings.use_glow; + tonemap.push_constant.glow_intensity = p_settings.glow_intensity; + tonemap.push_constant.glow_map_strength = p_settings.glow_map_strength; + tonemap.push_constant.glow_levels[0] = p_settings.glow_levels[0]; // clean this up to just pass by pointer or something + tonemap.push_constant.glow_levels[1] = p_settings.glow_levels[1]; + tonemap.push_constant.glow_levels[2] = p_settings.glow_levels[2]; + tonemap.push_constant.glow_levels[3] = p_settings.glow_levels[3]; + tonemap.push_constant.glow_levels[4] = p_settings.glow_levels[4]; + tonemap.push_constant.glow_levels[5] = p_settings.glow_levels[5]; + tonemap.push_constant.glow_levels[6] = p_settings.glow_levels[6]; + tonemap.push_constant.glow_texture_size[0] = p_settings.glow_texture_size.x; + tonemap.push_constant.glow_texture_size[1] = p_settings.glow_texture_size.y; + tonemap.push_constant.glow_mode = p_settings.glow_mode; + + int mode = p_settings.glow_use_bicubic_upscale ? TONEMAP_MODE_BICUBIC_GLOW_FILTER : TONEMAP_MODE_NORMAL; + if (p_settings.use_1d_color_correction) { + mode += 2; + } + + tonemap.push_constant.tonemapper = p_settings.tonemap_mode; + tonemap.push_constant.use_auto_exposure = p_settings.use_auto_exposure; + tonemap.push_constant.exposure = p_settings.exposure; + tonemap.push_constant.white = p_settings.white; + tonemap.push_constant.auto_exposure_scale = p_settings.auto_exposure_scale; + tonemap.push_constant.luminance_multiplier = p_settings.luminance_multiplier; + + tonemap.push_constant.use_color_correction = p_settings.use_color_correction; + + tonemap.push_constant.use_fxaa = p_settings.use_fxaa; + tonemap.push_constant.use_debanding = p_settings.use_debanding; + tonemap.push_constant.pixel_size[0] = 1.0 / p_settings.texture_size.x; + tonemap.push_constant.pixel_size[1] = 1.0 / p_settings.texture_size.y; + + if (p_settings.view_count > 1) { + // Use MULTIVIEW versions + mode += 6; + } + + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + RID default_mipmap_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_source_color(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_source_color })); + + RD::Uniform u_exposure_texture; + u_exposure_texture.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; + u_exposure_texture.binding = 0; + u_exposure_texture.append_id(default_sampler); + u_exposure_texture.append_id(p_settings.exposure_texture); + + RD::Uniform u_glow_texture; + u_glow_texture.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; + u_glow_texture.binding = 0; + u_glow_texture.append_id(default_mipmap_sampler); + u_glow_texture.append_id(p_settings.glow_texture); + + RD::Uniform u_glow_map; + u_glow_map.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; + u_glow_map.binding = 1; + u_glow_map.append_id(default_mipmap_sampler); + u_glow_map.append_id(p_settings.glow_map); + + RD::Uniform u_color_correction_texture; + u_color_correction_texture.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; + u_color_correction_texture.binding = 0; + u_color_correction_texture.append_id(default_sampler); + u_color_correction_texture.append_id(p_settings.color_correction_texture); + + RID shader = tonemap.shader.version_get_shader(tonemap.shader_version, mode); + ERR_FAIL_COND(shader.is_null()); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dst_framebuffer, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, tonemap.pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dst_framebuffer), false, RD::get_singleton()->draw_list_get_current_pass())); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 0, u_source_color), 0); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 1, u_exposure_texture), 1); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 2, u_glow_texture, u_glow_map), 2); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 3, u_color_correction_texture), 3); + RD::get_singleton()->draw_list_bind_index_array(draw_list, material_storage->get_quad_index_array()); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &tonemap.push_constant, sizeof(TonemapPushConstant)); + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); +} + +void ToneMapper::tonemapper(RD::DrawListID p_subpass_draw_list, RID p_source_color, RD::FramebufferFormatID p_dst_format_id, const TonemapSettings &p_settings) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + memset(&tonemap.push_constant, 0, sizeof(TonemapPushConstant)); + + tonemap.push_constant.use_bcs = p_settings.use_bcs; + tonemap.push_constant.bcs[0] = p_settings.brightness; + tonemap.push_constant.bcs[1] = p_settings.contrast; + tonemap.push_constant.bcs[2] = p_settings.saturation; + + ERR_FAIL_COND_MSG(p_settings.use_glow, "Glow is not supported when using subpasses."); + tonemap.push_constant.use_glow = p_settings.use_glow; + + int mode = p_settings.use_1d_color_correction ? TONEMAP_MODE_SUBPASS_1D_LUT : TONEMAP_MODE_SUBPASS; + if (p_settings.view_count > 1) { + // Use MULTIVIEW versions + mode += 6; + } + + tonemap.push_constant.tonemapper = p_settings.tonemap_mode; + tonemap.push_constant.use_auto_exposure = p_settings.use_auto_exposure; + tonemap.push_constant.exposure = p_settings.exposure; + tonemap.push_constant.white = p_settings.white; + tonemap.push_constant.auto_exposure_scale = p_settings.auto_exposure_scale; + + tonemap.push_constant.use_color_correction = p_settings.use_color_correction; + + tonemap.push_constant.use_debanding = p_settings.use_debanding; + tonemap.push_constant.luminance_multiplier = p_settings.luminance_multiplier; + + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + RID default_mipmap_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_source_color; + u_source_color.uniform_type = RD::UNIFORM_TYPE_INPUT_ATTACHMENT; + u_source_color.binding = 0; + u_source_color.append_id(p_source_color); + + RD::Uniform u_exposure_texture; + u_exposure_texture.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; + u_exposure_texture.binding = 0; + u_exposure_texture.append_id(default_sampler); + u_exposure_texture.append_id(p_settings.exposure_texture); + + RD::Uniform u_glow_texture; + u_glow_texture.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; + u_glow_texture.binding = 0; + u_glow_texture.append_id(default_mipmap_sampler); + u_glow_texture.append_id(p_settings.glow_texture); + + RD::Uniform u_glow_map; + u_glow_map.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; + u_glow_map.binding = 1; + u_glow_map.append_id(default_mipmap_sampler); + u_glow_map.append_id(p_settings.glow_map); + + RD::Uniform u_color_correction_texture; + u_color_correction_texture.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; + u_color_correction_texture.binding = 0; + u_color_correction_texture.append_id(default_sampler); + u_color_correction_texture.append_id(p_settings.color_correction_texture); + + RID shader = tonemap.shader.version_get_shader(tonemap.shader_version, mode); + ERR_FAIL_COND(shader.is_null()); + + RD::get_singleton()->draw_list_bind_render_pipeline(p_subpass_draw_list, tonemap.pipelines[mode].get_render_pipeline(RD::INVALID_ID, p_dst_format_id, false, RD::get_singleton()->draw_list_get_current_pass())); + RD::get_singleton()->draw_list_bind_uniform_set(p_subpass_draw_list, uniform_set_cache->get_cache(shader, 0, u_source_color), 0); + RD::get_singleton()->draw_list_bind_uniform_set(p_subpass_draw_list, uniform_set_cache->get_cache(shader, 1, u_exposure_texture), 1); // should be set to a default texture, it's ignored + RD::get_singleton()->draw_list_bind_uniform_set(p_subpass_draw_list, uniform_set_cache->get_cache(shader, 2, u_glow_texture, u_glow_map), 2); // should be set to a default texture, it's ignored + RD::get_singleton()->draw_list_bind_uniform_set(p_subpass_draw_list, uniform_set_cache->get_cache(shader, 3, u_color_correction_texture), 3); + RD::get_singleton()->draw_list_bind_index_array(p_subpass_draw_list, material_storage->get_quad_index_array()); + + RD::get_singleton()->draw_list_set_push_constant(p_subpass_draw_list, &tonemap.push_constant, sizeof(TonemapPushConstant)); + RD::get_singleton()->draw_list_draw(p_subpass_draw_list, true); +} diff --git a/servers/rendering/renderer_rd/effects/tone_mapper.h b/servers/rendering/renderer_rd/effects/tone_mapper.h new file mode 100644 index 0000000000..e91118e241 --- /dev/null +++ b/servers/rendering/renderer_rd/effects/tone_mapper.h @@ -0,0 +1,152 @@ +/*************************************************************************/ +/* tone_mapper.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef TONE_MAPPER_RD_H +#define TONE_MAPPER_RD_H + +#include "servers/rendering/renderer_rd/pipeline_cache_rd.h" +#include "servers/rendering/renderer_rd/shaders/effects/tonemap.glsl.gen.h" +#include "servers/rendering/renderer_scene_render.h" + +#include "servers/rendering_server.h" + +namespace RendererRD { + +class ToneMapper { +private: + enum TonemapMode { + TONEMAP_MODE_NORMAL, + TONEMAP_MODE_BICUBIC_GLOW_FILTER, + TONEMAP_MODE_1D_LUT, + TONEMAP_MODE_BICUBIC_GLOW_FILTER_1D_LUT, + TONEMAP_MODE_SUBPASS, + TONEMAP_MODE_SUBPASS_1D_LUT, + + TONEMAP_MODE_NORMAL_MULTIVIEW, + TONEMAP_MODE_BICUBIC_GLOW_FILTER_MULTIVIEW, + TONEMAP_MODE_1D_LUT_MULTIVIEW, + TONEMAP_MODE_BICUBIC_GLOW_FILTER_1D_LUT_MULTIVIEW, + TONEMAP_MODE_SUBPASS_MULTIVIEW, + TONEMAP_MODE_SUBPASS_1D_LUT_MULTIVIEW, + + TONEMAP_MODE_MAX + }; + + struct TonemapPushConstant { + float bcs[3]; // 12 - 12 + uint32_t use_bcs; // 4 - 16 + + uint32_t use_glow; // 4 - 20 + uint32_t use_auto_exposure; // 4 - 24 + uint32_t use_color_correction; // 4 - 28 + uint32_t tonemapper; // 4 - 32 + + uint32_t glow_texture_size[2]; // 8 - 40 + float glow_intensity; // 4 - 44 + float glow_map_strength; // 4 - 48 + + uint32_t glow_mode; // 4 - 52 + float glow_levels[7]; // 28 - 80 + + float exposure; // 4 - 84 + float white; // 4 - 88 + float auto_exposure_scale; // 4 - 92 + float luminance_multiplier; // 4 - 96 + + float pixel_size[2]; // 8 - 104 + uint32_t use_fxaa; // 4 - 108 + uint32_t use_debanding; // 4 - 112 + }; + + /* tonemap actually writes to a framebuffer, which is + * better to do using the raster pipeline rather than + * compute, as that framebuffer might be in different formats + */ + struct Tonemap { + TonemapPushConstant push_constant; + TonemapShaderRD shader; + RID shader_version; + PipelineCacheRD pipelines[TONEMAP_MODE_MAX]; + } tonemap; + +public: + ToneMapper(); + ~ToneMapper(); + + struct TonemapSettings { + bool use_glow = false; + enum GlowMode { + GLOW_MODE_ADD, + GLOW_MODE_SCREEN, + GLOW_MODE_SOFTLIGHT, + GLOW_MODE_REPLACE, + GLOW_MODE_MIX + }; + + GlowMode glow_mode = GLOW_MODE_ADD; + float glow_intensity = 1.0; + float glow_map_strength = 0.0f; + float glow_levels[7] = { 0.0, 0.0, 1.0, 0.0, 1.0, 0.0, 0.0 }; + Vector2i glow_texture_size; + bool glow_use_bicubic_upscale = false; + RID glow_texture; + RID glow_map; + + RS::EnvironmentToneMapper tonemap_mode = RS::ENV_TONE_MAPPER_LINEAR; + float exposure = 1.0; + float white = 1.0; + + bool use_auto_exposure = false; + float auto_exposure_scale = 0.5; + RID exposure_texture; + float luminance_multiplier = 1.0; + + bool use_bcs = false; + float brightness = 1.0; + float contrast = 1.0; + float saturation = 1.0; + + bool use_color_correction = false; + bool use_1d_color_correction = false; + RID color_correction_texture; + + bool use_fxaa = false; + bool use_debanding = false; + Vector2i texture_size; + uint32_t view_count = 1; + }; + + void tonemapper(RID p_source_color, RID p_dst_framebuffer, const TonemapSettings &p_settings); + void tonemapper(RD::DrawListID p_subpass_draw_list, RID p_source_color, RD::FramebufferFormatID p_dst_format_id, const TonemapSettings &p_settings); +}; + +} // namespace RendererRD + +#endif // TONE_MAPPER_RD_H diff --git a/servers/rendering/renderer_rd/effects/vrs.cpp b/servers/rendering/renderer_rd/effects/vrs.cpp new file mode 100644 index 0000000000..5ff00aa94c --- /dev/null +++ b/servers/rendering/renderer_rd/effects/vrs.cpp @@ -0,0 +1,149 @@ +/*************************************************************************/ +/* vrs.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "vrs.h" +#include "../renderer_compositor_rd.h" +#include "../storage_rd/texture_storage.h" +#include "../uniform_set_cache_rd.h" +#include "servers/xr_server.h" + +using namespace RendererRD; + +VRS::VRS() { + { + Vector<String> vrs_modes; + vrs_modes.push_back("\n"); // VRS_DEFAULT + vrs_modes.push_back("\n#define MULTIVIEW\n"); // VRS_MULTIVIEW + + vrs_shader.shader.initialize(vrs_modes); + + if (!RendererCompositorRD::singleton->is_xr_enabled()) { + vrs_shader.shader.set_variant_enabled(VRS_MULTIVIEW, false); + } + + vrs_shader.shader_version = vrs_shader.shader.version_create(); + + //use additive + + for (int i = 0; i < VRS_MAX; i++) { + if (vrs_shader.shader.is_variant_enabled(i)) { + vrs_shader.pipelines[i].setup(vrs_shader.shader.version_get_shader(vrs_shader.shader_version, i), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RD::PipelineColorBlendState::create_disabled(), 0); + } else { + vrs_shader.pipelines[i].clear(); + } + } + } +} + +VRS::~VRS() { + vrs_shader.shader.version_free(vrs_shader.shader_version); +} + +void VRS::copy_vrs(RID p_source_rd_texture, RID p_dest_framebuffer, bool p_multiview) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + // setup our uniforms + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_source_rd_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_source_rd_texture })); + + VRSMode mode = p_multiview ? VRS_MULTIVIEW : VRS_DEFAULT; + + RID shader = vrs_shader.shader.version_get_shader(vrs_shader.shader_version, mode); + ERR_FAIL_COND(shader.is_null()); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD, Vector<Color>()); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, vrs_shader.pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dest_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 0, u_source_rd_texture), 0); + RD::get_singleton()->draw_list_bind_index_array(draw_list, material_storage->get_quad_index_array()); + // RD::get_singleton()->draw_list_set_push_constant(draw_list, &vrs_shader.push_constant, sizeof(VRSPushConstant)); + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); +} + +Size2i VRS::get_vrs_texture_size(const Size2i p_base_size) const { + // TODO we should find some way to store this properly, we're assuming 16x16 as this seems to be the standard but in our vrs_capacities we + // obtain a minimum and maximum size, and we should choose something within this range and then make sure that is consistently set when creating + // our frame buffer. Also it is important that we make the resulting size we calculate down below available to the end user so they know the size + // of the VRS buffer to supply. + Size2i texel_size = Size2i(16, 16); + + int width = p_base_size.x / texel_size.x; + if (p_base_size.x % texel_size.x != 0) { + width++; + } + int height = p_base_size.y / texel_size.y; + if (p_base_size.y % texel_size.y != 0) { + height++; + } + return Size2i(width, height); +} + +void VRS::update_vrs_texture(RID p_vrs_fb, RID p_render_target) { + TextureStorage *texture_storage = TextureStorage::get_singleton(); + RS::ViewportVRSMode vrs_mode = texture_storage->render_target_get_vrs_mode(p_render_target); + + if (vrs_mode != RS::VIEWPORT_VRS_DISABLED) { + RD::get_singleton()->draw_command_begin_label("VRS Setup"); + + // TODO figure out if image has changed since it was last copied so we can save some resources.. + + if (vrs_mode == RS::VIEWPORT_VRS_TEXTURE) { + RID vrs_texture = texture_storage->render_target_get_vrs_texture(p_render_target); + if (vrs_texture.is_valid()) { + RID rd_texture = texture_storage->texture_get_rd_texture(vrs_texture); + int layers = texture_storage->texture_get_layers(vrs_texture); + if (rd_texture.is_valid()) { + // Copy into our density buffer + copy_vrs(rd_texture, p_vrs_fb, layers > 1); + } + } + } else if (vrs_mode == RS::VIEWPORT_VRS_XR) { + Ref<XRInterface> interface = XRServer::get_singleton()->get_primary_interface(); + if (interface.is_valid()) { + RID vrs_texture = interface->get_vrs_texture(); + if (vrs_texture.is_valid()) { + RID rd_texture = texture_storage->texture_get_rd_texture(vrs_texture); + int layers = texture_storage->texture_get_layers(vrs_texture); + + if (rd_texture.is_valid()) { + // Copy into our density buffer + copy_vrs(rd_texture, p_vrs_fb, layers > 1); + } + } + } + } + + RD::get_singleton()->draw_command_end_label(); + } +} diff --git a/servers/rendering/renderer_rd/effects/vrs.h b/servers/rendering/renderer_rd/effects/vrs.h new file mode 100644 index 0000000000..7125c6455d --- /dev/null +++ b/servers/rendering/renderer_rd/effects/vrs.h @@ -0,0 +1,75 @@ +/*************************************************************************/ +/* vrs.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef VRS_RD_H +#define VRS_RD_H + +#include "servers/rendering/renderer_rd/pipeline_cache_rd.h" +#include "servers/rendering/renderer_rd/shaders/effects/vrs.glsl.gen.h" +#include "servers/rendering/renderer_scene_render.h" + +#include "servers/rendering_server.h" + +namespace RendererRD { + +class VRS { +private: + enum VRSMode { + VRS_DEFAULT, + VRS_MULTIVIEW, + VRS_MAX, + }; + + /* we have no push constant here (yet) + struct VRSPushConstant { + + }; + */ + + struct VRSShader { + // VRSPushConstant push_constant; + VrsShaderRD shader; + RID shader_version; + PipelineCacheRD pipelines[VRS_MAX]; + } vrs_shader; + +public: + VRS(); + ~VRS(); + + void copy_vrs(RID p_source_rd_texture, RID p_dest_framebuffer, bool p_multiview = false); + + Size2i get_vrs_texture_size(const Size2i p_base_size) const; + void update_vrs_texture(RID p_vrs_fb, RID p_render_target); +}; + +} // namespace RendererRD + +#endif // VRS_RD_H diff --git a/servers/rendering/renderer_rd/effects_rd.cpp b/servers/rendering/renderer_rd/effects_rd.cpp new file mode 100644 index 0000000000..b03415f2e3 --- /dev/null +++ b/servers/rendering/renderer_rd/effects_rd.cpp @@ -0,0 +1,380 @@ +/*************************************************************************/ +/* effects_rd.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "effects_rd.h" + +#include "core/config/project_settings.h" +#include "core/math/math_defs.h" +#include "core/os/os.h" + +#include "servers/rendering/renderer_rd/renderer_compositor_rd.h" +#include "thirdparty/misc/cubemap_coeffs.h" + +bool EffectsRD::get_prefer_raster_effects() { + return prefer_raster_effects; +} + +RID EffectsRD::_get_uniform_set_from_image(RID p_image) { + if (image_to_uniform_set_cache.has(p_image)) { + RID uniform_set = image_to_uniform_set_cache[p_image]; + if (RD::get_singleton()->uniform_set_is_valid(uniform_set)) { + return uniform_set; + } + } + Vector<RD::Uniform> uniforms; + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 0; + u.append_id(p_image); + uniforms.push_back(u); + //any thing with the same configuration (one texture in binding 0 for set 0), is good + RID uniform_set = RD::get_singleton()->uniform_set_create(uniforms, luminance_reduce.shader.version_get_shader(luminance_reduce.shader_version, 0), 1); + + image_to_uniform_set_cache[p_image] = uniform_set; + + return uniform_set; +} + +RID EffectsRD::_get_uniform_set_from_texture(RID p_texture, bool p_use_mipmaps) { + if (texture_to_uniform_set_cache.has(p_texture)) { + RID uniform_set = texture_to_uniform_set_cache[p_texture]; + if (RD::get_singleton()->uniform_set_is_valid(uniform_set)) { + return uniform_set; + } + } + + Vector<RD::Uniform> uniforms; + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; + u.binding = 0; + u.append_id(p_use_mipmaps ? default_mipmap_sampler : default_sampler); + u.append_id(p_texture); + uniforms.push_back(u); + // anything with the same configuration (one texture in binding 0 for set 0), is good + RID uniform_set = RD::get_singleton()->uniform_set_create(uniforms, luminance_reduce_raster.shader.version_get_shader(luminance_reduce_raster.shader_version, 0), 0); + + texture_to_uniform_set_cache[p_texture] = uniform_set; + + return uniform_set; +} + +RID EffectsRD::_get_compute_uniform_set_from_texture(RID p_texture, bool p_use_mipmaps) { + if (texture_to_compute_uniform_set_cache.has(p_texture)) { + RID uniform_set = texture_to_compute_uniform_set_cache[p_texture]; + if (RD::get_singleton()->uniform_set_is_valid(uniform_set)) { + return uniform_set; + } + } + + Vector<RD::Uniform> uniforms; + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; + u.binding = 0; + u.append_id(p_use_mipmaps ? default_mipmap_sampler : default_sampler); + u.append_id(p_texture); + uniforms.push_back(u); + //any thing with the same configuration (one texture in binding 0 for set 0), is good + RID uniform_set = RD::get_singleton()->uniform_set_create(uniforms, luminance_reduce.shader.version_get_shader(luminance_reduce.shader_version, 0), 0); + + texture_to_compute_uniform_set_cache[p_texture] = uniform_set; + + return uniform_set; +} + +void EffectsRD::luminance_reduction(RID p_source_texture, const Size2i p_source_size, const Vector<RID> p_reduce, RID p_prev_luminance, float p_min_luminance, float p_max_luminance, float p_adjust, bool p_set) { + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use compute version of luminance reduction with the mobile renderer."); + + luminance_reduce.push_constant.source_size[0] = p_source_size.x; + luminance_reduce.push_constant.source_size[1] = p_source_size.y; + luminance_reduce.push_constant.max_luminance = p_max_luminance; + luminance_reduce.push_constant.min_luminance = p_min_luminance; + luminance_reduce.push_constant.exposure_adjust = p_adjust; + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + + for (int i = 0; i < p_reduce.size(); i++) { + if (i == 0) { + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, luminance_reduce.pipelines[LUMINANCE_REDUCE_READ]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_texture), 0); + } else { + RD::get_singleton()->compute_list_add_barrier(compute_list); //needs barrier, wait until previous is done + + if (i == p_reduce.size() - 1 && !p_set) { + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, luminance_reduce.pipelines[LUMINANCE_REDUCE_WRITE]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_prev_luminance), 2); + } else { + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, luminance_reduce.pipelines[LUMINANCE_REDUCE]); + } + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_reduce[i - 1]), 0); + } + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_reduce[i]), 1); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &luminance_reduce.push_constant, sizeof(LuminanceReducePushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, luminance_reduce.push_constant.source_size[0], luminance_reduce.push_constant.source_size[1], 1); + + luminance_reduce.push_constant.source_size[0] = MAX(luminance_reduce.push_constant.source_size[0] / 8, 1); + luminance_reduce.push_constant.source_size[1] = MAX(luminance_reduce.push_constant.source_size[1] / 8, 1); + } + + RD::get_singleton()->compute_list_end(); +} + +void EffectsRD::luminance_reduction_raster(RID p_source_texture, const Size2i p_source_size, const Vector<RID> p_reduce, Vector<RID> p_fb, RID p_prev_luminance, float p_min_luminance, float p_max_luminance, float p_adjust, bool p_set) { + ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use raster version of luminance reduction with the clustered renderer."); + ERR_FAIL_COND_MSG(p_reduce.size() != p_fb.size(), "Incorrect frame buffer account for luminance reduction."); + + luminance_reduce_raster.push_constant.max_luminance = p_max_luminance; + luminance_reduce_raster.push_constant.min_luminance = p_min_luminance; + luminance_reduce_raster.push_constant.exposure_adjust = p_adjust; + + for (int i = 0; i < p_reduce.size(); i++) { + luminance_reduce_raster.push_constant.source_size[0] = i == 0 ? p_source_size.x : luminance_reduce_raster.push_constant.dest_size[0]; + luminance_reduce_raster.push_constant.source_size[1] = i == 0 ? p_source_size.y : luminance_reduce_raster.push_constant.dest_size[1]; + luminance_reduce_raster.push_constant.dest_size[0] = MAX(luminance_reduce_raster.push_constant.source_size[0] / 8, 1); + luminance_reduce_raster.push_constant.dest_size[1] = MAX(luminance_reduce_raster.push_constant.source_size[1] / 8, 1); + + bool final = !p_set && (luminance_reduce_raster.push_constant.dest_size[0] == 1) && (luminance_reduce_raster.push_constant.dest_size[1] == 1); + LuminanceReduceRasterMode mode = final ? LUMINANCE_REDUCE_FRAGMENT_FINAL : (i == 0 ? LUMINANCE_REDUCE_FRAGMENT_FIRST : LUMINANCE_REDUCE_FRAGMENT); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_fb[i], RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, luminance_reduce_raster.pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_fb[i]))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(i == 0 ? p_source_texture : p_reduce[i - 1]), 0); + if (final) { + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_prev_luminance), 1); + } + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &luminance_reduce_raster.push_constant, sizeof(LuminanceReduceRasterPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + } +} + +void EffectsRD::roughness_limit(RID p_source_normal, RID p_roughness, const Size2i &p_size, float p_curve) { + roughness_limiter.push_constant.screen_size[0] = p_size.x; + roughness_limiter.push_constant.screen_size[1] = p_size.y; + roughness_limiter.push_constant.curve = p_curve; + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, roughness_limiter.pipeline); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_normal), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_roughness), 1); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &roughness_limiter.push_constant, sizeof(RoughnessLimiterPushConstant)); //not used but set anyway + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_size.x, p_size.y, 1); + + RD::get_singleton()->compute_list_end(); +} + +void EffectsRD::sort_buffer(RID p_uniform_set, int p_size) { + Sort::PushConstant push_constant; + push_constant.total_elements = p_size; + + bool done = true; + + int numThreadGroups = ((p_size - 1) >> 9) + 1; + + if (numThreadGroups > 1) { + done = false; + } + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sort.pipelines[SORT_MODE_BLOCK]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, p_uniform_set, 1); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(Sort::PushConstant)); + RD::get_singleton()->compute_list_dispatch(compute_list, numThreadGroups, 1, 1); + + int presorted = 512; + + while (!done) { + RD::get_singleton()->compute_list_add_barrier(compute_list); + + done = true; + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sort.pipelines[SORT_MODE_STEP]); + + numThreadGroups = 0; + + if (p_size > presorted) { + if (p_size > presorted * 2) { + done = false; + } + + int pow2 = presorted; + while (pow2 < p_size) { + pow2 *= 2; + } + numThreadGroups = pow2 >> 9; + } + + unsigned int nMergeSize = presorted * 2; + + for (unsigned int nMergeSubSize = nMergeSize >> 1; nMergeSubSize > 256; nMergeSubSize = nMergeSubSize >> 1) { + push_constant.job_params[0] = nMergeSubSize; + if (nMergeSubSize == nMergeSize >> 1) { + push_constant.job_params[1] = (2 * nMergeSubSize - 1); + push_constant.job_params[2] = -1; + } else { + push_constant.job_params[1] = nMergeSubSize; + push_constant.job_params[2] = 1; + } + push_constant.job_params[3] = 0; + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(Sort::PushConstant)); + RD::get_singleton()->compute_list_dispatch(compute_list, numThreadGroups, 1, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + } + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sort.pipelines[SORT_MODE_INNER]); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(Sort::PushConstant)); + RD::get_singleton()->compute_list_dispatch(compute_list, numThreadGroups, 1, 1); + + presorted *= 2; + } + + RD::get_singleton()->compute_list_end(); +} + +EffectsRD::EffectsRD(bool p_prefer_raster_effects) { + prefer_raster_effects = p_prefer_raster_effects; + + if (prefer_raster_effects) { + Vector<String> luminance_reduce_modes; + luminance_reduce_modes.push_back("\n#define FIRST_PASS\n"); // LUMINANCE_REDUCE_FRAGMENT_FIRST + luminance_reduce_modes.push_back("\n"); // LUMINANCE_REDUCE_FRAGMENT + luminance_reduce_modes.push_back("\n#define FINAL_PASS\n"); // LUMINANCE_REDUCE_FRAGMENT_FINAL + + luminance_reduce_raster.shader.initialize(luminance_reduce_modes); + memset(&luminance_reduce_raster.push_constant, 0, sizeof(LuminanceReduceRasterPushConstant)); + luminance_reduce_raster.shader_version = luminance_reduce_raster.shader.version_create(); + + for (int i = 0; i < LUMINANCE_REDUCE_FRAGMENT_MAX; i++) { + luminance_reduce_raster.pipelines[i].setup(luminance_reduce_raster.shader.version_get_shader(luminance_reduce_raster.shader_version, i), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RD::PipelineColorBlendState::create_disabled(), 0); + } + } else { + // Initialize luminance_reduce + Vector<String> luminance_reduce_modes; + luminance_reduce_modes.push_back("\n#define READ_TEXTURE\n"); + luminance_reduce_modes.push_back("\n"); + luminance_reduce_modes.push_back("\n#define WRITE_LUMINANCE\n"); + + luminance_reduce.shader.initialize(luminance_reduce_modes); + + luminance_reduce.shader_version = luminance_reduce.shader.version_create(); + + for (int i = 0; i < LUMINANCE_REDUCE_MAX; i++) { + luminance_reduce.pipelines[i] = RD::get_singleton()->compute_pipeline_create(luminance_reduce.shader.version_get_shader(luminance_reduce.shader_version, i)); + } + + for (int i = 0; i < LUMINANCE_REDUCE_FRAGMENT_MAX; i++) { + luminance_reduce_raster.pipelines[i].clear(); + } + } + + if (!prefer_raster_effects) { + // Initialize roughness limiter + Vector<String> shader_modes; + shader_modes.push_back(""); + + roughness_limiter.shader.initialize(shader_modes); + + roughness_limiter.shader_version = roughness_limiter.shader.version_create(); + + roughness_limiter.pipeline = RD::get_singleton()->compute_pipeline_create(roughness_limiter.shader.version_get_shader(roughness_limiter.shader_version, 0)); + } + + { + Vector<String> sort_modes; + sort_modes.push_back("\n#define MODE_SORT_BLOCK\n"); + sort_modes.push_back("\n#define MODE_SORT_STEP\n"); + sort_modes.push_back("\n#define MODE_SORT_INNER\n"); + + sort.shader.initialize(sort_modes); + + sort.shader_version = sort.shader.version_create(); + + for (int i = 0; i < SORT_MODE_MAX; i++) { + sort.pipelines[i] = RD::get_singleton()->compute_pipeline_create(sort.shader.version_get_shader(sort.shader_version, i)); + } + } + + RD::SamplerState sampler; + sampler.mag_filter = RD::SAMPLER_FILTER_LINEAR; + sampler.min_filter = RD::SAMPLER_FILTER_LINEAR; + sampler.max_lod = 0; + + default_sampler = RD::get_singleton()->sampler_create(sampler); + RD::get_singleton()->set_resource_name(default_sampler, "Default Linear Sampler"); + + sampler.min_filter = RD::SAMPLER_FILTER_LINEAR; + sampler.mip_filter = RD::SAMPLER_FILTER_LINEAR; + sampler.max_lod = 1e20; + + default_mipmap_sampler = RD::get_singleton()->sampler_create(sampler); + RD::get_singleton()->set_resource_name(default_mipmap_sampler, "Default MipMap Sampler"); + + { //create index array for copy shaders + Vector<uint8_t> pv; + pv.resize(6 * 4); + { + uint8_t *w = pv.ptrw(); + int *p32 = (int *)w; + p32[0] = 0; + p32[1] = 1; + p32[2] = 2; + p32[3] = 0; + p32[4] = 2; + p32[5] = 3; + } + index_buffer = RD::get_singleton()->index_buffer_create(6, RenderingDevice::INDEX_BUFFER_FORMAT_UINT32, pv); + index_array = RD::get_singleton()->index_array_create(index_buffer, 0, 6); + } +} + +EffectsRD::~EffectsRD() { + RD::get_singleton()->free(default_sampler); + RD::get_singleton()->free(default_mipmap_sampler); + RD::get_singleton()->free(index_buffer); //array gets freed as dependency + + if (prefer_raster_effects) { + luminance_reduce_raster.shader.version_free(luminance_reduce_raster.shader_version); + } else { + luminance_reduce.shader.version_free(luminance_reduce.shader_version); + } + if (!prefer_raster_effects) { + roughness_limiter.shader.version_free(roughness_limiter.shader_version); + } + sort.shader.version_free(sort.shader_version); +} diff --git a/servers/rendering/renderer_rd/effects_rd.h b/servers/rendering/renderer_rd/effects_rd.h new file mode 100644 index 0000000000..b05af73cf3 --- /dev/null +++ b/servers/rendering/renderer_rd/effects_rd.h @@ -0,0 +1,184 @@ +/*************************************************************************/ +/* effects_rd.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef EFFECTS_RD_H +#define EFFECTS_RD_H + +#include "core/math/projection.h" +#include "servers/rendering/renderer_rd/pipeline_cache_rd.h" +#include "servers/rendering/renderer_rd/shaders/luminance_reduce.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/luminance_reduce_raster.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/roughness_limiter.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/sort.glsl.gen.h" +#include "servers/rendering/renderer_scene_render.h" + +#include "servers/rendering_server.h" + +class EffectsRD { +private: + bool prefer_raster_effects; + + enum LuminanceReduceMode { + LUMINANCE_REDUCE_READ, + LUMINANCE_REDUCE, + LUMINANCE_REDUCE_WRITE, + LUMINANCE_REDUCE_MAX + }; + + struct LuminanceReducePushConstant { + int32_t source_size[2]; + float max_luminance; + float min_luminance; + float exposure_adjust; + float pad[3]; + }; + + struct LuminanceReduce { + LuminanceReducePushConstant push_constant; + LuminanceReduceShaderRD shader; + RID shader_version; + RID pipelines[LUMINANCE_REDUCE_MAX]; + } luminance_reduce; + + enum LuminanceReduceRasterMode { + LUMINANCE_REDUCE_FRAGMENT_FIRST, + LUMINANCE_REDUCE_FRAGMENT, + LUMINANCE_REDUCE_FRAGMENT_FINAL, + LUMINANCE_REDUCE_FRAGMENT_MAX + }; + + struct LuminanceReduceRasterPushConstant { + int32_t source_size[2]; + int32_t dest_size[2]; + float exposure_adjust; + float min_luminance; + float max_luminance; + uint32_t pad1; + }; + + struct LuminanceReduceFragment { + LuminanceReduceRasterPushConstant push_constant; + LuminanceReduceRasterShaderRD shader; + RID shader_version; + PipelineCacheRD pipelines[LUMINANCE_REDUCE_FRAGMENT_MAX]; + } luminance_reduce_raster; + + struct RoughnessLimiterPushConstant { + int32_t screen_size[2]; + float curve; + uint32_t pad; + }; + + struct RoughnessLimiter { + RoughnessLimiterPushConstant push_constant; + RoughnessLimiterShaderRD shader; + RID shader_version; + RID pipeline; + + } roughness_limiter; + + enum SortMode { + SORT_MODE_BLOCK, + SORT_MODE_STEP, + SORT_MODE_INNER, + SORT_MODE_MAX + }; + + struct Sort { + struct PushConstant { + uint32_t total_elements; + uint32_t pad[3]; + int32_t job_params[4]; + }; + + SortShaderRD shader; + RID shader_version; + RID pipelines[SORT_MODE_MAX]; + } sort; + + RID default_sampler; + RID default_mipmap_sampler; + RID index_buffer; + RID index_array; + + HashMap<RID, RID> texture_to_uniform_set_cache; + HashMap<RID, RID> input_to_uniform_set_cache; + + HashMap<RID, RID> image_to_uniform_set_cache; + + struct TexturePair { + RID texture1; + RID texture2; + _FORCE_INLINE_ bool operator<(const TexturePair &p_pair) const { + if (texture1 == p_pair.texture1) { + return texture2 < p_pair.texture2; + } else { + return texture1 < p_pair.texture1; + } + } + }; + + struct TextureSamplerPair { + RID texture; + RID sampler; + _FORCE_INLINE_ bool operator<(const TextureSamplerPair &p_pair) const { + if (texture == p_pair.texture) { + return sampler < p_pair.sampler; + } else { + return texture < p_pair.texture; + } + } + }; + + RBMap<TexturePair, RID> texture_pair_to_uniform_set_cache; + RBMap<RID, RID> texture_to_compute_uniform_set_cache; + RBMap<TexturePair, RID> texture_pair_to_compute_uniform_set_cache; + RBMap<TexturePair, RID> image_pair_to_compute_uniform_set_cache; + RBMap<TextureSamplerPair, RID> texture_sampler_to_compute_uniform_set_cache; + + RID _get_uniform_set_from_image(RID p_texture); + RID _get_uniform_set_from_texture(RID p_texture, bool p_use_mipmaps = false); + RID _get_compute_uniform_set_from_texture(RID p_texture, bool p_use_mipmaps = false); + +public: + bool get_prefer_raster_effects(); + + void luminance_reduction(RID p_source_texture, const Size2i p_source_size, const Vector<RID> p_reduce, RID p_prev_luminance, float p_min_luminance, float p_max_luminance, float p_adjust, bool p_set = false); + void luminance_reduction_raster(RID p_source_texture, const Size2i p_source_size, const Vector<RID> p_reduce, Vector<RID> p_fb, RID p_prev_luminance, float p_min_luminance, float p_max_luminance, float p_adjust, bool p_set = false); + + void roughness_limit(RID p_source_normal, RID p_roughness, const Size2i &p_size, float p_curve); + + void sort_buffer(RID p_uniform_set, int p_size); + + EffectsRD(bool p_prefer_raster_effects); + ~EffectsRD(); +}; + +#endif // EFFECTS_RD_H diff --git a/servers/rendering/renderer_rd/environment/SCsub b/servers/rendering/renderer_rd/environment/SCsub new file mode 100644 index 0000000000..86681f9c74 --- /dev/null +++ b/servers/rendering/renderer_rd/environment/SCsub @@ -0,0 +1,5 @@ +#!/usr/bin/env python + +Import("env") + +env.add_source_files(env.servers_sources, "*.cpp") diff --git a/servers/rendering/renderer_rd/environment/fog.cpp b/servers/rendering/renderer_rd/environment/fog.cpp new file mode 100644 index 0000000000..74082906c4 --- /dev/null +++ b/servers/rendering/renderer_rd/environment/fog.cpp @@ -0,0 +1,1212 @@ +/*************************************************************************/ +/* fog.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "fog.h" + +#include "servers/rendering/renderer_rd/renderer_compositor_rd.h" +#include "servers/rendering/renderer_rd/storage_rd/material_storage.h" +#include "servers/rendering/renderer_rd/storage_rd/texture_storage.h" +#include "servers/rendering/rendering_server_default.h" + +using namespace RendererRD; + +Fog *Fog::singleton = nullptr; + +Fog::Fog() { + singleton = this; +} + +Fog::~Fog() { + singleton = nullptr; +} + +/* FOG VOLUMES */ + +RID Fog::fog_volume_allocate() { + return fog_volume_owner.allocate_rid(); +} + +void Fog::fog_volume_initialize(RID p_rid) { + fog_volume_owner.initialize_rid(p_rid, FogVolume()); +} + +void Fog::fog_volume_free(RID p_rid) { + FogVolume *fog_volume = fog_volume_owner.get_or_null(p_rid); + fog_volume->dependency.deleted_notify(p_rid); + fog_volume_owner.free(p_rid); +} + +Dependency *Fog::fog_volume_get_dependency(RID p_fog_volume) const { + FogVolume *fog_volume = fog_volume_owner.get_or_null(p_fog_volume); + ERR_FAIL_NULL_V(fog_volume, nullptr); + + return &fog_volume->dependency; +} + +void Fog::fog_volume_set_shape(RID p_fog_volume, RS::FogVolumeShape p_shape) { + FogVolume *fog_volume = fog_volume_owner.get_or_null(p_fog_volume); + ERR_FAIL_COND(!fog_volume); + + if (p_shape == fog_volume->shape) { + return; + } + + fog_volume->shape = p_shape; + fog_volume->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); +} + +void Fog::fog_volume_set_extents(RID p_fog_volume, const Vector3 &p_extents) { + FogVolume *fog_volume = fog_volume_owner.get_or_null(p_fog_volume); + ERR_FAIL_COND(!fog_volume); + + fog_volume->extents = p_extents; + fog_volume->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); +} + +void Fog::fog_volume_set_material(RID p_fog_volume, RID p_material) { + FogVolume *fog_volume = fog_volume_owner.get_or_null(p_fog_volume); + ERR_FAIL_COND(!fog_volume); + fog_volume->material = p_material; +} + +RID Fog::fog_volume_get_material(RID p_fog_volume) const { + FogVolume *fog_volume = fog_volume_owner.get_or_null(p_fog_volume); + ERR_FAIL_COND_V(!fog_volume, RID()); + + return fog_volume->material; +} + +RS::FogVolumeShape Fog::fog_volume_get_shape(RID p_fog_volume) const { + FogVolume *fog_volume = fog_volume_owner.get_or_null(p_fog_volume); + ERR_FAIL_COND_V(!fog_volume, RS::FOG_VOLUME_SHAPE_BOX); + + return fog_volume->shape; +} + +AABB Fog::fog_volume_get_aabb(RID p_fog_volume) const { + FogVolume *fog_volume = fog_volume_owner.get_or_null(p_fog_volume); + ERR_FAIL_COND_V(!fog_volume, AABB()); + + switch (fog_volume->shape) { + case RS::FOG_VOLUME_SHAPE_ELLIPSOID: + case RS::FOG_VOLUME_SHAPE_CONE: + case RS::FOG_VOLUME_SHAPE_CYLINDER: + case RS::FOG_VOLUME_SHAPE_BOX: { + AABB aabb; + aabb.position = -fog_volume->extents; + aabb.size = fog_volume->extents * 2; + return aabb; + } + default: { + // Need some size otherwise will get culled + return AABB(Vector3(-1, -1, -1), Vector3(2, 2, 2)); + } + } +} + +Vector3 Fog::fog_volume_get_extents(RID p_fog_volume) const { + const FogVolume *fog_volume = fog_volume_owner.get_or_null(p_fog_volume); + ERR_FAIL_COND_V(!fog_volume, Vector3()); + return fog_volume->extents; +} + +//////////////////////////////////////////////////////////////////////////////// +// Fog material + +bool Fog::FogMaterialData::update_parameters(const HashMap<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { + uniform_set_updated = true; + + return update_parameters_uniform_set(p_parameters, p_uniform_dirty, p_textures_dirty, shader_data->uniforms, shader_data->ubo_offsets.ptr(), shader_data->texture_uniforms, shader_data->default_texture_params, shader_data->ubo_size, uniform_set, Fog::get_singleton()->volumetric_fog.shader.version_get_shader(shader_data->version, 0), VolumetricFogShader::FogSet::FOG_SET_MATERIAL, true); +} + +Fog::FogMaterialData::~FogMaterialData() { + free_parameters_uniform_set(uniform_set); +} + +RendererRD::MaterialStorage::ShaderData *Fog::_create_fog_shader_func() { + FogShaderData *shader_data = memnew(FogShaderData); + return shader_data; +} + +RendererRD::MaterialStorage::ShaderData *Fog::_create_fog_shader_funcs() { + return Fog::get_singleton()->_create_fog_shader_func(); +}; + +RendererRD::MaterialStorage::MaterialData *Fog::_create_fog_material_func(FogShaderData *p_shader) { + FogMaterialData *material_data = memnew(FogMaterialData); + material_data->shader_data = p_shader; + //update will happen later anyway so do nothing. + return material_data; +} + +RendererRD::MaterialStorage::MaterialData *Fog::_create_fog_material_funcs(RendererRD::MaterialStorage::ShaderData *p_shader) { + return Fog::get_singleton()->_create_fog_material_func(static_cast<FogShaderData *>(p_shader)); +}; + +//////////////////////////////////////////////////////////////////////////////// +// FOG VOLUMES INSTANCE + +RID Fog::fog_volume_instance_create(RID p_fog_volume) { + FogVolumeInstance fvi; + fvi.volume = p_fog_volume; + return fog_volume_instance_owner.make_rid(fvi); +} + +void Fog::fog_instance_free(RID p_rid) { + fog_volume_instance_owner.free(p_rid); +} + +//////////////////////////////////////////////////////////////////////////////// +// Volumetric Fog Shader + +void Fog::init_fog_shader(uint32_t p_max_directional_lights, int p_roughness_layers, bool p_is_using_radiance_cubemap_array) { + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + + { + // Initialize local fog shader + Vector<String> volumetric_fog_modes; + volumetric_fog_modes.push_back(""); + volumetric_fog.shader.initialize(volumetric_fog_modes); + + material_storage->shader_set_data_request_function(RendererRD::MaterialStorage::SHADER_TYPE_FOG, _create_fog_shader_funcs); + material_storage->material_set_data_request_function(RendererRD::MaterialStorage::SHADER_TYPE_FOG, _create_fog_material_funcs); + volumetric_fog.volume_ubo = RD::get_singleton()->uniform_buffer_create(sizeof(VolumetricFogShader::VolumeUBO)); + } + + { + ShaderCompiler::DefaultIdentifierActions actions; + + actions.renames["TIME"] = "scene_params.time"; + actions.renames["PI"] = _MKSTR(Math_PI); + actions.renames["TAU"] = _MKSTR(Math_TAU); + actions.renames["E"] = _MKSTR(Math_E); + actions.renames["WORLD_POSITION"] = "world.xyz"; + actions.renames["OBJECT_POSITION"] = "params.position"; + actions.renames["UVW"] = "uvw"; + actions.renames["EXTENTS"] = "params.extents"; + actions.renames["ALBEDO"] = "albedo"; + actions.renames["DENSITY"] = "density"; + actions.renames["EMISSION"] = "emission"; + actions.renames["SDF"] = "sdf"; + + actions.usage_defines["SDF"] = "#define SDF_USED\n"; + actions.usage_defines["DENSITY"] = "#define DENSITY_USED\n"; + actions.usage_defines["ALBEDO"] = "#define ALBEDO_USED\n"; + actions.usage_defines["EMISSION"] = "#define EMISSION_USED\n"; + + actions.sampler_array_name = "material_samplers"; + actions.base_texture_binding_index = 1; + actions.texture_layout_set = VolumetricFogShader::FogSet::FOG_SET_MATERIAL; + actions.base_uniform_string = "material."; + + actions.default_filter = ShaderLanguage::FILTER_LINEAR_MIPMAP; + actions.default_repeat = ShaderLanguage::REPEAT_DISABLE; + actions.global_buffer_array_variable = "global_variables.data"; + + volumetric_fog.compiler.initialize(actions); + } + + { + // default material and shader for fog shader + volumetric_fog.default_shader = material_storage->shader_allocate(); + material_storage->shader_initialize(volumetric_fog.default_shader); + material_storage->shader_set_code(volumetric_fog.default_shader, R"( +// Default fog shader. + +shader_type fog; + +void fog() { +DENSITY = 1.0; +ALBEDO = vec3(1.0); +} +)"); + volumetric_fog.default_material = material_storage->material_allocate(); + material_storage->material_initialize(volumetric_fog.default_material); + material_storage->material_set_shader(volumetric_fog.default_material, volumetric_fog.default_shader); + + FogMaterialData *md = static_cast<FogMaterialData *>(material_storage->material_get_data(volumetric_fog.default_material, RendererRD::MaterialStorage::SHADER_TYPE_FOG)); + volumetric_fog.default_shader_rd = volumetric_fog.shader.version_get_shader(md->shader_data->version, 0); + + Vector<RD::Uniform> uniforms; + + { + Vector<RID> ids; + ids.resize(12); + RID *ids_ptr = ids.ptrw(); + ids_ptr[0] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[1] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[2] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[3] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[4] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[5] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[6] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[7] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[8] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[9] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[10] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[11] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + + RD::Uniform u(RD::UNIFORM_TYPE_SAMPLER, 1, ids); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 2; + u.append_id(RendererRD::MaterialStorage::get_singleton()->global_shader_uniforms_get_storage_buffer()); + uniforms.push_back(u); + } + + volumetric_fog.base_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, volumetric_fog.default_shader_rd, VolumetricFogShader::FogSet::FOG_SET_BASE); + } + { + String defines = "\n#define MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS " + itos(p_max_directional_lights) + "\n"; + defines += "\n#define MAX_SKY_LOD " + itos(p_roughness_layers - 1) + ".0\n"; + if (p_is_using_radiance_cubemap_array) { + defines += "\n#define USE_RADIANCE_CUBEMAP_ARRAY \n"; + } + Vector<String> volumetric_fog_modes; + volumetric_fog_modes.push_back("\n#define MODE_DENSITY\n"); + volumetric_fog_modes.push_back("\n#define MODE_DENSITY\n#define ENABLE_SDFGI\n"); + volumetric_fog_modes.push_back("\n#define MODE_FILTER\n"); + volumetric_fog_modes.push_back("\n#define MODE_FOG\n"); + volumetric_fog_modes.push_back("\n#define MODE_COPY\n"); + + volumetric_fog.process_shader.initialize(volumetric_fog_modes, defines); + volumetric_fog.process_shader_version = volumetric_fog.process_shader.version_create(); + for (int i = 0; i < VolumetricFogShader::VOLUMETRIC_FOG_PROCESS_SHADER_MAX; i++) { + volumetric_fog.process_pipelines[i] = RD::get_singleton()->compute_pipeline_create(volumetric_fog.process_shader.version_get_shader(volumetric_fog.process_shader_version, i)); + } + volumetric_fog.params_ubo = RD::get_singleton()->uniform_buffer_create(sizeof(VolumetricFogShader::ParamsUBO)); + } +} + +void Fog::free_fog_shader() { + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + + volumetric_fog.process_shader.version_free(volumetric_fog.process_shader_version); + RD::get_singleton()->free(volumetric_fog.volume_ubo); + RD::get_singleton()->free(volumetric_fog.params_ubo); + material_storage->shader_free(volumetric_fog.default_shader); + material_storage->material_free(volumetric_fog.default_material); +} + +void Fog::FogShaderData::set_path_hint(const String &p_path) { + path = p_path; +} + +void Fog::FogShaderData::set_code(const String &p_code) { + //compile + + code = p_code; + valid = false; + ubo_size = 0; + uniforms.clear(); + + if (code.is_empty()) { + return; //just invalid, but no error + } + + ShaderCompiler::GeneratedCode gen_code; + ShaderCompiler::IdentifierActions actions; + actions.entry_point_stages["fog"] = ShaderCompiler::STAGE_COMPUTE; + + uses_time = false; + + actions.usage_flag_pointers["TIME"] = &uses_time; + + actions.uniforms = &uniforms; + + Fog *fog_singleton = Fog::get_singleton(); + + Error err = fog_singleton->volumetric_fog.compiler.compile(RS::SHADER_FOG, code, &actions, path, gen_code); + ERR_FAIL_COND_MSG(err != OK, "Fog shader compilation failed."); + + if (version.is_null()) { + version = fog_singleton->volumetric_fog.shader.version_create(); + } + + fog_singleton->volumetric_fog.shader.version_set_compute_code(version, gen_code.code, gen_code.uniforms, gen_code.stage_globals[ShaderCompiler::STAGE_COMPUTE], gen_code.defines); + ERR_FAIL_COND(!fog_singleton->volumetric_fog.shader.version_is_valid(version)); + + ubo_size = gen_code.uniform_total_size; + ubo_offsets = gen_code.uniform_offsets; + texture_uniforms = gen_code.texture_uniforms; + + pipeline = RD::get_singleton()->compute_pipeline_create(fog_singleton->volumetric_fog.shader.version_get_shader(version, 0)); + + valid = true; +} + +void Fog::FogShaderData::set_default_texture_parameter(const StringName &p_name, RID p_texture, int p_index) { + if (!p_texture.is_valid()) { + if (default_texture_params.has(p_name) && default_texture_params[p_name].has(p_index)) { + default_texture_params[p_name].erase(p_index); + + if (default_texture_params[p_name].is_empty()) { + default_texture_params.erase(p_name); + } + } + } else { + if (!default_texture_params.has(p_name)) { + default_texture_params[p_name] = HashMap<int, RID>(); + } + default_texture_params[p_name][p_index] = p_texture; + } +} + +void Fog::FogShaderData::get_shader_uniform_list(List<PropertyInfo> *p_param_list) const { + RBMap<int, StringName> order; + + for (const KeyValue<StringName, ShaderLanguage::ShaderNode::Uniform> &E : uniforms) { + if (E.value.scope == ShaderLanguage::ShaderNode::Uniform::SCOPE_GLOBAL || E.value.scope == ShaderLanguage::ShaderNode::Uniform::SCOPE_INSTANCE) { + continue; + } + + if (E.value.texture_order >= 0) { + order[E.value.texture_order + 100000] = E.key; + } else { + order[E.value.order] = E.key; + } + } + + String last_group; + for (const KeyValue<int, StringName> &E : order) { + String group = uniforms[E.value].group; + if (!uniforms[E.value].subgroup.is_empty()) { + group += "::" + uniforms[E.value].subgroup; + } + + if (group != last_group) { + PropertyInfo pi; + pi.usage = PROPERTY_USAGE_GROUP; + pi.name = group; + p_param_list->push_back(pi); + + last_group = group; + } + + PropertyInfo pi = ShaderLanguage::uniform_to_property_info(uniforms[E.value]); + pi.name = E.value; + p_param_list->push_back(pi); + } +} + +void Fog::FogShaderData::get_instance_param_list(List<RendererMaterialStorage::InstanceShaderParam> *p_param_list) const { + for (const KeyValue<StringName, ShaderLanguage::ShaderNode::Uniform> &E : uniforms) { + if (E.value.scope != ShaderLanguage::ShaderNode::Uniform::SCOPE_INSTANCE) { + continue; + } + + RendererMaterialStorage::InstanceShaderParam p; + p.info = ShaderLanguage::uniform_to_property_info(E.value); + p.info.name = E.key; //supply name + p.index = E.value.instance_index; + p.default_value = ShaderLanguage::constant_value_to_variant(E.value.default_value, E.value.type, E.value.array_size, E.value.hint); + p_param_list->push_back(p); + } +} + +bool Fog::FogShaderData::is_parameter_texture(const StringName &p_param) const { + if (!uniforms.has(p_param)) { + return false; + } + + return uniforms[p_param].texture_order >= 0; +} + +bool Fog::FogShaderData::is_animated() const { + return false; +} + +bool Fog::FogShaderData::casts_shadows() const { + return false; +} + +Variant Fog::FogShaderData::get_default_parameter(const StringName &p_parameter) const { + if (uniforms.has(p_parameter)) { + ShaderLanguage::ShaderNode::Uniform uniform = uniforms[p_parameter]; + Vector<ShaderLanguage::ConstantNode::Value> default_value = uniform.default_value; + return ShaderLanguage::constant_value_to_variant(default_value, uniform.type, uniform.array_size, uniform.hint); + } + return Variant(); +} + +RS::ShaderNativeSourceCode Fog::FogShaderData::get_native_source_code() const { + Fog *fog_singleton = Fog::get_singleton(); + + return fog_singleton->volumetric_fog.shader.version_get_native_source_code(version); +} + +Fog::FogShaderData::~FogShaderData() { + Fog *fog_singleton = Fog::get_singleton(); + ERR_FAIL_COND(!fog_singleton); + //pipeline variants will clear themselves if shader is gone + if (version.is_valid()) { + fog_singleton->volumetric_fog.shader.version_free(version); + } +} + +//////////////////////////////////////////////////////////////////////////////// +// Volumetric Fog + +void Fog::VolumetricFog::init(const Vector3i &fog_size, RID p_sky_shader) { + width = fog_size.x; + height = fog_size.y; + depth = fog_size.z; + + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.width = fog_size.x; + tf.height = fog_size.y; + tf.depth = fog_size.z; + tf.texture_type = RD::TEXTURE_TYPE_3D; + tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT; + + light_density_map = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(light_density_map, "Fog light-density map"); + + tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; + + prev_light_density_map = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(prev_light_density_map, "Fog previous light-density map"); + RD::get_singleton()->texture_clear(prev_light_density_map, Color(0, 0, 0, 0), 0, 1, 0, 1); + + tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; + + fog_map = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(fog_map, "Fog map"); + +#if defined(MACOS_ENABLED) || defined(IOS_ENABLED) + Vector<uint8_t> dm; + dm.resize(fog_size.x * fog_size.y * fog_size.z * 4); + dm.fill(0); + + density_map = RD::get_singleton()->storage_buffer_create(dm.size(), dm); + RD::get_singleton()->set_resource_name(density_map, "Fog density map"); + light_map = RD::get_singleton()->storage_buffer_create(dm.size(), dm); + RD::get_singleton()->set_resource_name(light_map, "Fog light map"); + emissive_map = RD::get_singleton()->storage_buffer_create(dm.size(), dm); + RD::get_singleton()->set_resource_name(emissive_map, "Fog emissive map"); +#else + tf.format = RD::DATA_FORMAT_R32_UINT; + tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; + density_map = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(density_map, "Fog density map"); + RD::get_singleton()->texture_clear(density_map, Color(0, 0, 0, 0), 0, 1, 0, 1); + light_map = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(light_map, "Fog light map"); + RD::get_singleton()->texture_clear(light_map, Color(0, 0, 0, 0), 0, 1, 0, 1); + emissive_map = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(emissive_map, "Fog emissive map"); + RD::get_singleton()->texture_clear(emissive_map, Color(0, 0, 0, 0), 0, 1, 0, 1); +#endif + + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.binding = 0; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.append_id(fog_map); + uniforms.push_back(u); + } + + sky_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, p_sky_shader, RendererRD::SkyRD::SKY_SET_FOG); +} + +Fog::VolumetricFog::~VolumetricFog() { + RD::get_singleton()->free(prev_light_density_map); + RD::get_singleton()->free(light_density_map); + RD::get_singleton()->free(fog_map); + RD::get_singleton()->free(density_map); + RD::get_singleton()->free(light_map); + RD::get_singleton()->free(emissive_map); + + if (fog_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(fog_uniform_set)) { + RD::get_singleton()->free(fog_uniform_set); + } + if (process_uniform_set_density.is_valid() && RD::get_singleton()->uniform_set_is_valid(process_uniform_set_density)) { + RD::get_singleton()->free(process_uniform_set_density); + } + if (process_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(process_uniform_set)) { + RD::get_singleton()->free(process_uniform_set); + } + if (process_uniform_set2.is_valid() && RD::get_singleton()->uniform_set_is_valid(process_uniform_set2)) { + RD::get_singleton()->free(process_uniform_set2); + } + if (sdfgi_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(sdfgi_uniform_set)) { + RD::get_singleton()->free(sdfgi_uniform_set); + } + if (sky_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(sky_uniform_set)) { + RD::get_singleton()->free(sky_uniform_set); + } +} + +Vector3i Fog::_point_get_position_in_froxel_volume(const Vector3 &p_point, float fog_end, const Vector2 &fog_near_size, const Vector2 &fog_far_size, float volumetric_fog_detail_spread, const Vector3 &fog_size, const Transform3D &p_cam_transform) { + Vector3 view_position = p_cam_transform.affine_inverse().xform(p_point); + view_position.z = MIN(view_position.z, -0.01); // Clamp to the front of camera + Vector3 fog_position = Vector3(0, 0, 0); + + view_position.y = -view_position.y; + fog_position.z = -view_position.z / fog_end; + fog_position.x = (view_position.x / (2 * (fog_near_size.x * (1.0 - fog_position.z) + fog_far_size.x * fog_position.z))) + 0.5; + fog_position.y = (view_position.y / (2 * (fog_near_size.y * (1.0 - fog_position.z) + fog_far_size.y * fog_position.z))) + 0.5; + fog_position.z = Math::pow(float(fog_position.z), float(1.0 / volumetric_fog_detail_spread)); + fog_position = fog_position * fog_size - Vector3(0.5, 0.5, 0.5); + + fog_position.x = CLAMP(fog_position.x, 0.0, fog_size.x); + fog_position.y = CLAMP(fog_position.y, 0.0, fog_size.y); + fog_position.z = CLAMP(fog_position.z, 0.0, fog_size.z); + + return Vector3i(fog_position); +} + +void Fog::volumetric_fog_update(const VolumetricFogSettings &p_settings, const Projection &p_cam_projection, const Transform3D &p_cam_transform, const Transform3D &p_prev_cam_inv_transform, RID p_shadow_atlas, int p_directional_light_count, bool p_use_directional_shadows, int p_positional_light_count, int p_voxel_gi_count, const PagedArray<RID> &p_fog_volumes) { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + + RENDER_TIMESTAMP("> Volumetric Fog"); + RD::get_singleton()->draw_command_begin_label("Volumetric Fog"); + + Ref<VolumetricFog> fog = p_settings.vfog; + + if (p_fog_volumes.size() > 0) { + RD::get_singleton()->draw_command_begin_label("Render Volumetric Fog Volumes"); + + RENDER_TIMESTAMP("Render FogVolumes"); + + VolumetricFogShader::VolumeUBO params; + + Vector2 frustum_near_size = p_cam_projection.get_viewport_half_extents(); + Vector2 frustum_far_size = p_cam_projection.get_far_plane_half_extents(); + float z_near = p_cam_projection.get_z_near(); + float z_far = p_cam_projection.get_z_far(); + float fog_end = RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_length(p_settings.env); + + Vector2 fog_far_size = frustum_near_size.lerp(frustum_far_size, (fog_end - z_near) / (z_far - z_near)); + Vector2 fog_near_size; + if (p_cam_projection.is_orthogonal()) { + fog_near_size = fog_far_size; + } else { + fog_near_size = Vector2(); + } + + params.fog_frustum_size_begin[0] = fog_near_size.x; + params.fog_frustum_size_begin[1] = fog_near_size.y; + + params.fog_frustum_size_end[0] = fog_far_size.x; + params.fog_frustum_size_end[1] = fog_far_size.y; + + params.fog_frustum_end = fog_end; + params.z_near = z_near; + params.z_far = z_far; + params.time = p_settings.time; + + params.fog_volume_size[0] = fog->width; + params.fog_volume_size[1] = fog->height; + params.fog_volume_size[2] = fog->depth; + + params.use_temporal_reprojection = RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_temporal_reprojection(p_settings.env); + params.temporal_frame = RSG::rasterizer->get_frame_number() % VolumetricFog::MAX_TEMPORAL_FRAMES; + params.detail_spread = RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_detail_spread(p_settings.env); + params.temporal_blend = RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_temporal_reprojection_amount(p_settings.env); + + Transform3D to_prev_cam_view = p_prev_cam_inv_transform * p_cam_transform; + RendererRD::MaterialStorage::store_transform(to_prev_cam_view, params.to_prev_view); + RendererRD::MaterialStorage::store_transform(p_cam_transform, params.transform); + + RD::get_singleton()->buffer_update(volumetric_fog.volume_ubo, 0, sizeof(VolumetricFogShader::VolumeUBO), ¶ms, RD::BARRIER_MASK_COMPUTE); + + if (fog->fog_uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(fog->fog_uniform_set)) { + Vector<RD::Uniform> uniforms; + + { + RD::Uniform u; +#if defined(MACOS_ENABLED) || defined(IOS_ENABLED) + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; +#else + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; +#endif + u.binding = 1; + u.append_id(fog->emissive_map); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 2; + u.append_id(volumetric_fog.volume_ubo); + uniforms.push_back(u); + } + + { + RD::Uniform u; +#if defined(MACOS_ENABLED) || defined(IOS_ENABLED) + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; +#else + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; +#endif + u.binding = 3; + u.append_id(fog->density_map); + uniforms.push_back(u); + } + + { + RD::Uniform u; +#if defined(MACOS_ENABLED) || defined(IOS_ENABLED) + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; +#else + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; +#endif + u.binding = 4; + u.append_id(fog->light_map); + uniforms.push_back(u); + } + + fog->fog_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, volumetric_fog.default_shader_rd, VolumetricFogShader::FogSet::FOG_SET_UNIFORMS); + } + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + bool any_uses_time = false; + + for (int i = 0; i < (int)p_fog_volumes.size(); i++) { + FogVolumeInstance *fog_volume_instance = fog_volume_instance_owner.get_or_null(p_fog_volumes[i]); + ERR_FAIL_COND(!fog_volume_instance); + RID fog_volume = fog_volume_instance->volume; + + RID fog_material = RendererRD::Fog::get_singleton()->fog_volume_get_material(fog_volume); + + FogMaterialData *material = nullptr; + + if (fog_material.is_valid()) { + material = static_cast<FogMaterialData *>(material_storage->material_get_data(fog_material, RendererRD::MaterialStorage::SHADER_TYPE_FOG)); + if (!material || !material->shader_data->valid) { + material = nullptr; + } + } + + if (!material) { + fog_material = volumetric_fog.default_material; + material = static_cast<FogMaterialData *>(material_storage->material_get_data(fog_material, RendererRD::MaterialStorage::SHADER_TYPE_FOG)); + } + + ERR_FAIL_COND(!material); + + FogShaderData *shader_data = material->shader_data; + + ERR_FAIL_COND(!shader_data); + + any_uses_time |= shader_data->uses_time; + + Vector3i min = Vector3i(); + Vector3i max = Vector3i(); + Vector3i kernel_size = Vector3i(); + + Vector3 position = fog_volume_instance->transform.get_origin(); + RS::FogVolumeShape volume_type = RendererRD::Fog::get_singleton()->fog_volume_get_shape(fog_volume); + Vector3 extents = RendererRD::Fog::get_singleton()->fog_volume_get_extents(fog_volume); + + if (volume_type != RS::FOG_VOLUME_SHAPE_WORLD) { + // Local fog volume. + Vector3i points[8]; + Vector3 fog_size = Vector3(fog->width, fog->height, fog->depth); + float volumetric_fog_detail_spread = RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_detail_spread(p_settings.env); + points[0] = _point_get_position_in_froxel_volume(fog_volume_instance->transform.xform(Vector3(extents.x, extents.y, extents.z)), fog_end, fog_near_size, fog_far_size, volumetric_fog_detail_spread, fog_size, p_cam_transform); + points[1] = _point_get_position_in_froxel_volume(fog_volume_instance->transform.xform(Vector3(-extents.x, extents.y, extents.z)), fog_end, fog_near_size, fog_far_size, volumetric_fog_detail_spread, fog_size, p_cam_transform); + points[2] = _point_get_position_in_froxel_volume(fog_volume_instance->transform.xform(Vector3(extents.x, -extents.y, extents.z)), fog_end, fog_near_size, fog_far_size, volumetric_fog_detail_spread, fog_size, p_cam_transform); + points[3] = _point_get_position_in_froxel_volume(fog_volume_instance->transform.xform(Vector3(-extents.x, -extents.y, extents.z)), fog_end, fog_near_size, fog_far_size, volumetric_fog_detail_spread, fog_size, p_cam_transform); + points[4] = _point_get_position_in_froxel_volume(fog_volume_instance->transform.xform(Vector3(extents.x, extents.y, -extents.z)), fog_end, fog_near_size, fog_far_size, volumetric_fog_detail_spread, fog_size, p_cam_transform); + points[5] = _point_get_position_in_froxel_volume(fog_volume_instance->transform.xform(Vector3(-extents.x, extents.y, -extents.z)), fog_end, fog_near_size, fog_far_size, volumetric_fog_detail_spread, fog_size, p_cam_transform); + points[6] = _point_get_position_in_froxel_volume(fog_volume_instance->transform.xform(Vector3(extents.x, -extents.y, -extents.z)), fog_end, fog_near_size, fog_far_size, volumetric_fog_detail_spread, fog_size, p_cam_transform); + points[7] = _point_get_position_in_froxel_volume(fog_volume_instance->transform.xform(Vector3(-extents.x, -extents.y, -extents.z)), fog_end, fog_near_size, fog_far_size, volumetric_fog_detail_spread, fog_size, p_cam_transform); + + min = Vector3i(int32_t(fog->width) - 1, int32_t(fog->height) - 1, int32_t(fog->depth) - 1); + max = Vector3i(1, 1, 1); + + for (int j = 0; j < 8; j++) { + min = Vector3i(MIN(min.x, points[j].x), MIN(min.y, points[j].y), MIN(min.z, points[j].z)); + max = Vector3i(MAX(max.x, points[j].x), MAX(max.y, points[j].y), MAX(max.z, points[j].z)); + } + + kernel_size = max - min; + } else { + // Volume type global runs on all cells + extents = Vector3(fog->width, fog->height, fog->depth); + min = Vector3i(0, 0, 0); + kernel_size = Vector3i(int32_t(fog->width), int32_t(fog->height), int32_t(fog->depth)); + } + + if (kernel_size.x == 0 || kernel_size.y == 0 || kernel_size.z == 0) { + continue; + } + + VolumetricFogShader::FogPushConstant push_constant; + push_constant.position[0] = position.x; + push_constant.position[1] = position.y; + push_constant.position[2] = position.z; + push_constant.extents[0] = extents.x; + push_constant.extents[1] = extents.y; + push_constant.extents[2] = extents.z; + push_constant.corner[0] = min.x; + push_constant.corner[1] = min.y; + push_constant.corner[2] = min.z; + push_constant.shape = uint32_t(RendererRD::Fog::get_singleton()->fog_volume_get_shape(fog_volume)); + RendererRD::MaterialStorage::store_transform(fog_volume_instance->transform.affine_inverse(), push_constant.transform); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, shader_data->pipeline); + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, fog->fog_uniform_set, VolumetricFogShader::FogSet::FOG_SET_UNIFORMS); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(VolumetricFogShader::FogPushConstant)); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, volumetric_fog.base_uniform_set, VolumetricFogShader::FogSet::FOG_SET_BASE); + if (material->uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(material->uniform_set)) { // Material may not have a uniform set. + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, material->uniform_set, VolumetricFogShader::FogSet::FOG_SET_MATERIAL); + } + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, kernel_size.x, kernel_size.y, kernel_size.z); + } + if (any_uses_time || RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_temporal_reprojection(p_settings.env)) { + RenderingServerDefault::redraw_request(); + } + + RD::get_singleton()->draw_command_end_label(); + + RD::get_singleton()->compute_list_end(); + } + + if (fog->process_uniform_set_density.is_null() || !RD::get_singleton()->uniform_set_is_valid(fog->process_uniform_set_density)) { + //re create uniform set if needed + Vector<RD::Uniform> uniforms; + Vector<RD::Uniform> copy_uniforms; + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 1; + if (p_settings.shadow_atlas_depth.is_null()) { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_BLACK)); + } else { + u.append_id(p_settings.shadow_atlas_depth); + } + + uniforms.push_back(u); + copy_uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 2; + if (p_settings.directional_shadow_depth.is_valid()) { + u.append_id(p_settings.directional_shadow_depth); + } else { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_BLACK)); + } + uniforms.push_back(u); + copy_uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 3; + u.append_id(p_settings.omni_light_buffer); + uniforms.push_back(u); + copy_uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 4; + u.append_id(p_settings.spot_light_buffer); + uniforms.push_back(u); + copy_uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 5; + u.append_id(p_settings.directional_light_buffer); + uniforms.push_back(u); + copy_uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 6; + u.append_id(p_settings.cluster_builder->get_cluster_buffer()); + uniforms.push_back(u); + copy_uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 7; + u.append_id(material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED)); + uniforms.push_back(u); + copy_uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 8; + u.append_id(fog->light_density_map); + uniforms.push_back(u); + copy_uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 9; + u.append_id(fog->fog_map); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 9; + u.append_id(fog->prev_light_density_map); + copy_uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 10; + u.append_id(p_settings.shadow_sampler); + uniforms.push_back(u); + copy_uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 11; + u.append_id(p_settings.voxel_gi_buffer); + uniforms.push_back(u); + copy_uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 12; + for (int i = 0; i < RendererRD::GI::MAX_VOXEL_GI_INSTANCES; i++) { + u.append_id(p_settings.rbgi->voxel_gi_textures[i]); + } + uniforms.push_back(u); + copy_uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 13; + u.append_id(material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED)); + uniforms.push_back(u); + copy_uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 14; + u.append_id(volumetric_fog.params_ubo); + uniforms.push_back(u); + copy_uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 15; + u.append_id(fog->prev_light_density_map); + uniforms.push_back(u); + } + { + RD::Uniform u; +#if defined(MACOS_ENABLED) || defined(IOS_ENABLED) + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; +#else + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; +#endif + u.binding = 16; + u.append_id(fog->density_map); + uniforms.push_back(u); + } + { + RD::Uniform u; +#if defined(MACOS_ENABLED) || defined(IOS_ENABLED) + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; +#else + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; +#endif + u.binding = 17; + u.append_id(fog->light_map); + uniforms.push_back(u); + } + + { + RD::Uniform u; +#if defined(MACOS_ENABLED) || defined(IOS_ENABLED) + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; +#else + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; +#endif + u.binding = 18; + u.append_id(fog->emissive_map); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 19; + RID radiance_texture = texture_storage->texture_rd_get_default(p_settings.is_using_radiance_cubemap_array ? RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_BLACK : RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK); + RID sky_texture = RendererSceneRenderRD::get_singleton()->environment_get_sky(p_settings.env).is_valid() ? p_settings.sky->sky_get_radiance_texture_rd(RendererSceneRenderRD::get_singleton()->environment_get_sky(p_settings.env)) : RID(); + u.append_id(sky_texture.is_valid() ? sky_texture : radiance_texture); + uniforms.push_back(u); + } + + fog->copy_uniform_set = RD::get_singleton()->uniform_set_create(copy_uniforms, volumetric_fog.process_shader.version_get_shader(volumetric_fog.process_shader_version, VolumetricFogShader::VOLUMETRIC_FOG_PROCESS_SHADER_COPY), 0); + + fog->process_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, volumetric_fog.process_shader.version_get_shader(volumetric_fog.process_shader_version, VolumetricFogShader::VOLUMETRIC_FOG_PROCESS_SHADER_FOG), 0); + + RID aux7 = uniforms.write[7].get_id(0); + RID aux8 = uniforms.write[8].get_id(0); + + uniforms.write[7].set_id(0, aux8); + uniforms.write[8].set_id(0, aux7); + + fog->process_uniform_set2 = RD::get_singleton()->uniform_set_create(uniforms, volumetric_fog.process_shader.version_get_shader(volumetric_fog.process_shader_version, VolumetricFogShader::VOLUMETRIC_FOG_PROCESS_SHADER_FOG), 0); + + uniforms.remove_at(8); + uniforms.write[7].set_id(0, aux7); + fog->process_uniform_set_density = RD::get_singleton()->uniform_set_create(uniforms, volumetric_fog.process_shader.version_get_shader(volumetric_fog.process_shader_version, VolumetricFogShader::VOLUMETRIC_FOG_PROCESS_SHADER_DENSITY), 0); + } + + bool using_sdfgi = RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_gi_inject(p_settings.env) > 0.0001 && RendererSceneRenderRD::get_singleton()->environment_get_sdfgi_enabled(p_settings.env) && (p_settings.sdfgi.is_valid()); + + if (using_sdfgi) { + if (fog->sdfgi_uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(fog->sdfgi_uniform_set)) { + Vector<RD::Uniform> uniforms; + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 0; + u.append_id(p_settings.gi->sdfgi_ubo); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 1; + u.append_id(p_settings.sdfgi->ambient_texture); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 2; + u.append_id(p_settings.sdfgi->occlusion_texture); + uniforms.push_back(u); + } + + fog->sdfgi_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, volumetric_fog.process_shader.version_get_shader(volumetric_fog.process_shader_version, VolumetricFogShader::VOLUMETRIC_FOG_PROCESS_SHADER_DENSITY_WITH_SDFGI), 1); + } + } + + fog->length = RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_length(p_settings.env); + fog->spread = RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_detail_spread(p_settings.env); + + VolumetricFogShader::ParamsUBO params; + + Vector2 frustum_near_size = p_cam_projection.get_viewport_half_extents(); + Vector2 frustum_far_size = p_cam_projection.get_far_plane_half_extents(); + float z_near = p_cam_projection.get_z_near(); + float z_far = p_cam_projection.get_z_far(); + float fog_end = RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_length(p_settings.env); + + Vector2 fog_far_size = frustum_near_size.lerp(frustum_far_size, (fog_end - z_near) / (z_far - z_near)); + Vector2 fog_near_size; + if (p_cam_projection.is_orthogonal()) { + fog_near_size = fog_far_size; + } else { + fog_near_size = Vector2(); + } + + params.fog_frustum_size_begin[0] = fog_near_size.x; + params.fog_frustum_size_begin[1] = fog_near_size.y; + + params.fog_frustum_size_end[0] = fog_far_size.x; + params.fog_frustum_size_end[1] = fog_far_size.y; + + params.ambient_inject = RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_ambient_inject(p_settings.env) * RendererSceneRenderRD::get_singleton()->environment_get_ambient_light_energy(p_settings.env); + params.z_far = z_far; + + params.fog_frustum_end = fog_end; + + Color ambient_color = RendererSceneRenderRD::get_singleton()->environment_get_ambient_light(p_settings.env).srgb_to_linear(); + params.ambient_color[0] = ambient_color.r; + params.ambient_color[1] = ambient_color.g; + params.ambient_color[2] = ambient_color.b; + params.sky_contribution = RendererSceneRenderRD::get_singleton()->environment_get_ambient_sky_contribution(p_settings.env); + + params.fog_volume_size[0] = fog->width; + params.fog_volume_size[1] = fog->height; + params.fog_volume_size[2] = fog->depth; + + params.directional_light_count = p_directional_light_count; + + Color emission = RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_emission(p_settings.env).srgb_to_linear(); + params.base_emission[0] = emission.r * RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_emission_energy(p_settings.env); + params.base_emission[1] = emission.g * RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_emission_energy(p_settings.env); + params.base_emission[2] = emission.b * RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_emission_energy(p_settings.env); + params.base_density = RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_density(p_settings.env); + + Color base_scattering = RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_scattering(p_settings.env).srgb_to_linear(); + params.base_scattering[0] = base_scattering.r; + params.base_scattering[1] = base_scattering.g; + params.base_scattering[2] = base_scattering.b; + params.phase_g = RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_anisotropy(p_settings.env); + + params.detail_spread = RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_detail_spread(p_settings.env); + params.gi_inject = RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_gi_inject(p_settings.env); + + params.cam_rotation[0] = p_cam_transform.basis[0][0]; + params.cam_rotation[1] = p_cam_transform.basis[1][0]; + params.cam_rotation[2] = p_cam_transform.basis[2][0]; + params.cam_rotation[3] = 0; + params.cam_rotation[4] = p_cam_transform.basis[0][1]; + params.cam_rotation[5] = p_cam_transform.basis[1][1]; + params.cam_rotation[6] = p_cam_transform.basis[2][1]; + params.cam_rotation[7] = 0; + params.cam_rotation[8] = p_cam_transform.basis[0][2]; + params.cam_rotation[9] = p_cam_transform.basis[1][2]; + params.cam_rotation[10] = p_cam_transform.basis[2][2]; + params.cam_rotation[11] = 0; + params.filter_axis = 0; + params.max_voxel_gi_instances = RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_gi_inject(p_settings.env) > 0.001 ? p_voxel_gi_count : 0; + params.temporal_frame = RSG::rasterizer->get_frame_number() % VolumetricFog::MAX_TEMPORAL_FRAMES; + + Transform3D to_prev_cam_view = p_prev_cam_inv_transform * p_cam_transform; + RendererRD::MaterialStorage::store_transform(to_prev_cam_view, params.to_prev_view); + + params.use_temporal_reprojection = RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_temporal_reprojection(p_settings.env); + params.temporal_blend = RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_temporal_reprojection_amount(p_settings.env); + + { + uint32_t cluster_size = p_settings.cluster_builder->get_cluster_size(); + params.cluster_shift = get_shift_from_power_of_2(cluster_size); + + uint32_t cluster_screen_width = (p_settings.rb_size.x - 1) / cluster_size + 1; + uint32_t cluster_screen_height = (p_settings.rb_size.y - 1) / cluster_size + 1; + params.max_cluster_element_count_div_32 = p_settings.max_cluster_elements / 32; + params.cluster_type_size = cluster_screen_width * cluster_screen_height * (params.max_cluster_element_count_div_32 + 32); + params.cluster_width = cluster_screen_width; + + params.screen_size[0] = p_settings.rb_size.x; + params.screen_size[1] = p_settings.rb_size.y; + } + + Basis sky_transform = RendererSceneRenderRD::get_singleton()->environment_get_sky_orientation(p_settings.env); + sky_transform = sky_transform.inverse() * p_cam_transform.basis; + RendererRD::MaterialStorage::store_transform_3x3(sky_transform, params.radiance_inverse_xform); + + RD::get_singleton()->draw_command_begin_label("Render Volumetric Fog"); + + RENDER_TIMESTAMP("Render Fog"); + RD::get_singleton()->buffer_update(volumetric_fog.params_ubo, 0, sizeof(VolumetricFogShader::ParamsUBO), ¶ms, RD::BARRIER_MASK_COMPUTE); + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, volumetric_fog.process_pipelines[using_sdfgi ? VolumetricFogShader::VOLUMETRIC_FOG_PROCESS_SHADER_DENSITY_WITH_SDFGI : VolumetricFogShader::VOLUMETRIC_FOG_PROCESS_SHADER_DENSITY]); + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, fog->process_uniform_set_density, 0); + + if (using_sdfgi) { + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, fog->sdfgi_uniform_set, 1); + } + RD::get_singleton()->compute_list_dispatch_threads(compute_list, fog->width, fog->height, fog->depth); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + // Copy fog to history buffer + if (RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_temporal_reprojection(p_settings.env)) { + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, volumetric_fog.process_pipelines[VolumetricFogShader::VOLUMETRIC_FOG_PROCESS_SHADER_COPY]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, fog->copy_uniform_set, 0); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, fog->width, fog->height, fog->depth); + RD::get_singleton()->compute_list_add_barrier(compute_list); + } + RD::get_singleton()->draw_command_end_label(); + + if (p_settings.volumetric_fog_filter_active) { + RD::get_singleton()->draw_command_begin_label("Filter Fog"); + + RENDER_TIMESTAMP("Filter Fog"); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, volumetric_fog.process_pipelines[VolumetricFogShader::VOLUMETRIC_FOG_PROCESS_SHADER_FILTER]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, fog->process_uniform_set, 0); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, fog->width, fog->height, fog->depth); + + RD::get_singleton()->compute_list_end(); + //need restart for buffer update + + params.filter_axis = 1; + RD::get_singleton()->buffer_update(volumetric_fog.params_ubo, 0, sizeof(VolumetricFogShader::ParamsUBO), ¶ms); + + compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, volumetric_fog.process_pipelines[VolumetricFogShader::VOLUMETRIC_FOG_PROCESS_SHADER_FILTER]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, fog->process_uniform_set2, 0); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, fog->width, fog->height, fog->depth); + + RD::get_singleton()->compute_list_add_barrier(compute_list); + RD::get_singleton()->draw_command_end_label(); + } + + RENDER_TIMESTAMP("Integrate Fog"); + RD::get_singleton()->draw_command_begin_label("Integrate Fog"); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, volumetric_fog.process_pipelines[VolumetricFogShader::VOLUMETRIC_FOG_PROCESS_SHADER_FOG]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, fog->process_uniform_set, 0); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, fog->width, fog->height, 1); + + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_RASTER); + + RENDER_TIMESTAMP("< Volumetric Fog"); + RD::get_singleton()->draw_command_end_label(); + RD::get_singleton()->draw_command_end_label(); +} diff --git a/servers/rendering/renderer_rd/environment/fog.h b/servers/rendering/renderer_rd/environment/fog.h new file mode 100644 index 0000000000..0ade995758 --- /dev/null +++ b/servers/rendering/renderer_rd/environment/fog.h @@ -0,0 +1,360 @@ +/*************************************************************************/ +/* fog.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef FOG_RD_H +#define FOG_RD_H + +#include "core/templates/local_vector.h" +#include "core/templates/rid_owner.h" +#include "servers/rendering/environment/renderer_fog.h" +#include "servers/rendering/renderer_rd/cluster_builder_rd.h" +#include "servers/rendering/renderer_rd/environment/gi.h" +#include "servers/rendering/renderer_rd/shaders/environment/volumetric_fog.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/environment/volumetric_fog_process.glsl.gen.h" +#include "servers/rendering/renderer_rd/storage_rd/render_buffer_custom_data_rd.h" +#include "servers/rendering/storage/utilities.h" + +#define RB_SCOPE_FOG SNAME("Fog") + +namespace RendererRD { + +class Fog : public RendererFog { +private: + static Fog *singleton; + + /* FOG VOLUMES */ + + struct FogVolume { + RID material; + Vector3 extents = Vector3(1, 1, 1); + + RS::FogVolumeShape shape = RS::FOG_VOLUME_SHAPE_BOX; + + Dependency dependency; + }; + + mutable RID_Owner<FogVolume, true> fog_volume_owner; + + struct FogVolumeInstance { + RID volume; + Transform3D transform; + bool active = false; + }; + + mutable RID_Owner<FogVolumeInstance> fog_volume_instance_owner; + + /* Volumetric Fog */ + struct VolumetricFogShader { + enum FogSet { + FOG_SET_BASE, + FOG_SET_UNIFORMS, + FOG_SET_MATERIAL, + FOG_SET_MAX, + }; + + struct FogPushConstant { + float position[3]; + float pad; + + float extents[3]; + float pad2; + + int32_t corner[3]; + uint32_t shape; + + float transform[16]; + }; + + struct VolumeUBO { + float fog_frustum_size_begin[2]; + float fog_frustum_size_end[2]; + + float fog_frustum_end; + float z_near; + float z_far; + float time; + + int32_t fog_volume_size[3]; + uint32_t directional_light_count; + + uint32_t use_temporal_reprojection; + uint32_t temporal_frame; + float detail_spread; + float temporal_blend; + + float to_prev_view[16]; + float transform[16]; + }; + + ShaderCompiler compiler; + VolumetricFogShaderRD shader; + RID volume_ubo; + + RID default_shader; + RID default_material; + RID default_shader_rd; + + RID base_uniform_set; + + RID params_ubo; + + enum { + VOLUMETRIC_FOG_PROCESS_SHADER_DENSITY, + VOLUMETRIC_FOG_PROCESS_SHADER_DENSITY_WITH_SDFGI, + VOLUMETRIC_FOG_PROCESS_SHADER_FILTER, + VOLUMETRIC_FOG_PROCESS_SHADER_FOG, + VOLUMETRIC_FOG_PROCESS_SHADER_COPY, + VOLUMETRIC_FOG_PROCESS_SHADER_MAX, + }; + + struct ParamsUBO { + float fog_frustum_size_begin[2]; + float fog_frustum_size_end[2]; + + float fog_frustum_end; + float ambient_inject; + float z_far; + uint32_t filter_axis; + + float ambient_color[3]; + float sky_contribution; + + int32_t fog_volume_size[3]; + uint32_t directional_light_count; + + float base_emission[3]; + float base_density; + + float base_scattering[3]; + float phase_g; + + float detail_spread; + float gi_inject; + uint32_t max_voxel_gi_instances; + uint32_t cluster_type_size; + + float screen_size[2]; + uint32_t cluster_shift; + uint32_t cluster_width; + + uint32_t max_cluster_element_count_div_32; + uint32_t use_temporal_reprojection; + uint32_t temporal_frame; + float temporal_blend; + + float cam_rotation[12]; + float to_prev_view[16]; + float radiance_inverse_xform[12]; + }; + + VolumetricFogProcessShaderRD process_shader; + + RID process_shader_version; + RID process_pipelines[VOLUMETRIC_FOG_PROCESS_SHADER_MAX]; + + } volumetric_fog; + + Vector3i _point_get_position_in_froxel_volume(const Vector3 &p_point, float fog_end, const Vector2 &fog_near_size, const Vector2 &fog_far_size, float volumetric_fog_detail_spread, const Vector3 &fog_size, const Transform3D &p_cam_transform); + + struct FogShaderData : public RendererRD::MaterialStorage::ShaderData { + bool valid = false; + RID version; + + RID pipeline; + HashMap<StringName, ShaderLanguage::ShaderNode::Uniform> uniforms; + Vector<ShaderCompiler::GeneratedCode::Texture> texture_uniforms; + + Vector<uint32_t> ubo_offsets; + uint32_t ubo_size = 0; + + String path; + String code; + HashMap<StringName, HashMap<int, RID>> default_texture_params; + + bool uses_time = false; + + virtual void set_path_hint(const String &p_hint); + virtual void set_code(const String &p_Code); + virtual void set_default_texture_parameter(const StringName &p_name, RID p_texture, int p_index); + virtual void get_shader_uniform_list(List<PropertyInfo> *p_param_list) const; + virtual void get_instance_param_list(List<RendererMaterialStorage::InstanceShaderParam> *p_param_list) const; + virtual bool is_parameter_texture(const StringName &p_param) const; + virtual bool is_animated() const; + virtual bool casts_shadows() const; + virtual Variant get_default_parameter(const StringName &p_parameter) const; + virtual RS::ShaderNativeSourceCode get_native_source_code() const; + + FogShaderData() {} + virtual ~FogShaderData(); + }; + + struct FogMaterialData : public RendererRD::MaterialStorage::MaterialData { + FogShaderData *shader_data = nullptr; + RID uniform_set; + bool uniform_set_updated; + + virtual void set_render_priority(int p_priority) {} + virtual void set_next_pass(RID p_pass) {} + virtual bool update_parameters(const HashMap<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); + virtual ~FogMaterialData(); + }; + + RendererRD::MaterialStorage::ShaderData *_create_fog_shader_func(); + static RendererRD::MaterialStorage::ShaderData *_create_fog_shader_funcs(); + + RendererRD::MaterialStorage::MaterialData *_create_fog_material_func(FogShaderData *p_shader); + static RendererRD::MaterialStorage::MaterialData *_create_fog_material_funcs(RendererRD::MaterialStorage::ShaderData *p_shader); + +public: + static Fog *get_singleton() { return singleton; } + + Fog(); + ~Fog(); + + /* FOG VOLUMES */ + + bool owns_fog_volume(RID p_rid) { return fog_volume_owner.owns(p_rid); }; + + virtual RID fog_volume_allocate() override; + virtual void fog_volume_initialize(RID p_rid) override; + virtual void fog_volume_free(RID p_rid) override; + Dependency *fog_volume_get_dependency(RID p_fog_volume) const; + + virtual void fog_volume_set_shape(RID p_fog_volume, RS::FogVolumeShape p_shape) override; + virtual void fog_volume_set_extents(RID p_fog_volume, const Vector3 &p_extents) override; + virtual void fog_volume_set_material(RID p_fog_volume, RID p_material) override; + virtual RS::FogVolumeShape fog_volume_get_shape(RID p_fog_volume) const override; + RID fog_volume_get_material(RID p_fog_volume) const; + virtual AABB fog_volume_get_aabb(RID p_fog_volume) const override; + Vector3 fog_volume_get_extents(RID p_fog_volume) const; + + /* FOG VOLUMES INSTANCE */ + + bool owns_fog_volume_instance(RID p_rid) { return fog_volume_instance_owner.owns(p_rid); }; + + RID fog_volume_instance_create(RID p_fog_volume); + void fog_instance_free(RID p_rid); + + void fog_volume_instance_set_transform(RID p_fog_volume_instance, const Transform3D &p_transform) { + Fog::FogVolumeInstance *fvi = fog_volume_instance_owner.get_or_null(p_fog_volume_instance); + ERR_FAIL_COND(!fvi); + fvi->transform = p_transform; + } + + void fog_volume_instance_set_active(RID p_fog_volume_instance, bool p_active) { + Fog::FogVolumeInstance *fvi = fog_volume_instance_owner.get_or_null(p_fog_volume_instance); + ERR_FAIL_COND(!fvi); + fvi->active = p_active; + } + + RID fog_volume_instance_get_volume(RID p_fog_volume_instance) const { + Fog::FogVolumeInstance *fvi = fog_volume_instance_owner.get_or_null(p_fog_volume_instance); + ERR_FAIL_COND_V(!fvi, RID()); + return fvi->volume; + } + + Vector3 fog_volume_instance_get_position(RID p_fog_volume_instance) const { + Fog::FogVolumeInstance *fvi = fog_volume_instance_owner.get_or_null(p_fog_volume_instance); + ERR_FAIL_COND_V(!fvi, Vector3()); + return fvi->transform.get_origin(); + } + + /* Volumetric FOG */ + class VolumetricFog : public RenderBufferCustomDataRD { + GDCLASS(VolumetricFog, RenderBufferCustomDataRD) + + public: + enum { + MAX_TEMPORAL_FRAMES = 16 + }; + + uint32_t width = 0; + uint32_t height = 0; + uint32_t depth = 0; + + float length; + float spread; + + RID light_density_map; + RID prev_light_density_map; + RID fog_map; + RID density_map; + RID light_map; + RID emissive_map; + + RID fog_uniform_set; + RID copy_uniform_set; + RID process_uniform_set_density; + RID process_uniform_set; + RID process_uniform_set2; + RID sdfgi_uniform_set; + RID sky_uniform_set; + + int last_shadow_filter = -1; + + virtual void configure(RenderSceneBuffersRD *p_render_buffers) override{}; + virtual void free_data() override{}; + + void init(const Vector3i &fog_size, RID p_sky_shader); + ~VolumetricFog(); + }; + + void init_fog_shader(uint32_t p_max_directional_lights, int p_roughness_layers, bool p_is_using_radiance_cubemap_array); + void free_fog_shader(); + + struct VolumetricFogSettings { + Vector2i rb_size; + double time; + bool is_using_radiance_cubemap_array; + uint32_t max_cluster_elements; + bool volumetric_fog_filter_active; + RID shadow_sampler; + RID voxel_gi_buffer; + RID shadow_atlas_depth; + RID omni_light_buffer; + RID spot_light_buffer; + RID directional_shadow_depth; + RID directional_light_buffer; + + // Objects related to our render buffer + Ref<VolumetricFog> vfog; + ClusterBuilderRD *cluster_builder; + GI *gi; + Ref<GI::SDFGI> sdfgi; + Ref<GI::RenderBuffersGI> rbgi; + RID env; + SkyRD *sky; + }; + void volumetric_fog_update(const VolumetricFogSettings &p_settings, const Projection &p_cam_projection, const Transform3D &p_cam_transform, const Transform3D &p_prev_cam_inv_transform, RID p_shadow_atlas, int p_directional_light_count, bool p_use_directional_shadows, int p_positional_light_count, int p_voxel_gi_count, const PagedArray<RID> &p_fog_volumes); +}; + +} // namespace RendererRD + +#endif // FOG_RD_H diff --git a/servers/rendering/renderer_rd/environment/gi.cpp b/servers/rendering/renderer_rd/environment/gi.cpp new file mode 100644 index 0000000000..550fe27e4c --- /dev/null +++ b/servers/rendering/renderer_rd/environment/gi.cpp @@ -0,0 +1,4082 @@ +/*************************************************************************/ +/* gi.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "gi.h" + +#include "core/config/project_settings.h" +#include "servers/rendering/renderer_rd/renderer_compositor_rd.h" +#include "servers/rendering/renderer_rd/renderer_scene_render_rd.h" +#include "servers/rendering/renderer_rd/storage_rd/material_storage.h" +#include "servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.h" +#include "servers/rendering/renderer_rd/storage_rd/texture_storage.h" +#include "servers/rendering/rendering_server_default.h" + +using namespace RendererRD; + +const Vector3i GI::SDFGI::Cascade::DIRTY_ALL = Vector3i(0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF); + +GI *GI::singleton = nullptr; + +//////////////////////////////////////////////////////////////////////////////// +// VOXEL GI STORAGE + +RID GI::voxel_gi_allocate() { + return voxel_gi_owner.allocate_rid(); +} + +void GI::voxel_gi_free(RID p_voxel_gi) { + voxel_gi_allocate_data(p_voxel_gi, Transform3D(), AABB(), Vector3i(), Vector<uint8_t>(), Vector<uint8_t>(), Vector<uint8_t>(), Vector<int>()); //deallocate + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + voxel_gi->dependency.deleted_notify(p_voxel_gi); + voxel_gi_owner.free(p_voxel_gi); +} + +void GI::voxel_gi_initialize(RID p_voxel_gi) { + voxel_gi_owner.initialize_rid(p_voxel_gi, VoxelGI()); +} + +void GI::voxel_gi_allocate_data(RID p_voxel_gi, const Transform3D &p_to_cell_xform, const AABB &p_aabb, const Vector3i &p_octree_size, const Vector<uint8_t> &p_octree_cells, const Vector<uint8_t> &p_data_cells, const Vector<uint8_t> &p_distance_field, const Vector<int> &p_level_counts) { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND(!voxel_gi); + + if (voxel_gi->octree_buffer.is_valid()) { + RD::get_singleton()->free(voxel_gi->octree_buffer); + RD::get_singleton()->free(voxel_gi->data_buffer); + if (voxel_gi->sdf_texture.is_valid()) { + RD::get_singleton()->free(voxel_gi->sdf_texture); + } + + voxel_gi->sdf_texture = RID(); + voxel_gi->octree_buffer = RID(); + voxel_gi->data_buffer = RID(); + voxel_gi->octree_buffer_size = 0; + voxel_gi->data_buffer_size = 0; + voxel_gi->cell_count = 0; + } + + voxel_gi->to_cell_xform = p_to_cell_xform; + voxel_gi->bounds = p_aabb; + voxel_gi->octree_size = p_octree_size; + voxel_gi->level_counts = p_level_counts; + + if (p_octree_cells.size()) { + ERR_FAIL_COND(p_octree_cells.size() % 32 != 0); //cells size must be a multiple of 32 + + uint32_t cell_count = p_octree_cells.size() / 32; + + ERR_FAIL_COND(p_data_cells.size() != (int)cell_count * 16); //see that data size matches + + voxel_gi->cell_count = cell_count; + voxel_gi->octree_buffer = RD::get_singleton()->storage_buffer_create(p_octree_cells.size(), p_octree_cells); + voxel_gi->octree_buffer_size = p_octree_cells.size(); + voxel_gi->data_buffer = RD::get_singleton()->storage_buffer_create(p_data_cells.size(), p_data_cells); + voxel_gi->data_buffer_size = p_data_cells.size(); + + if (p_distance_field.size()) { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R8_UNORM; + tf.width = voxel_gi->octree_size.x; + tf.height = voxel_gi->octree_size.y; + tf.depth = voxel_gi->octree_size.z; + tf.texture_type = RD::TEXTURE_TYPE_3D; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT; + Vector<Vector<uint8_t>> s; + s.push_back(p_distance_field); + voxel_gi->sdf_texture = RD::get_singleton()->texture_create(tf, RD::TextureView(), s); + RD::get_singleton()->set_resource_name(voxel_gi->sdf_texture, "VoxelGI SDF Texture"); + } +#if 0 + { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R8_UNORM; + tf.width = voxel_gi->octree_size.x; + tf.height = voxel_gi->octree_size.y; + tf.depth = voxel_gi->octree_size.z; + tf.type = RD::TEXTURE_TYPE_3D; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; + tf.shareable_formats.push_back(RD::DATA_FORMAT_R8_UNORM); + tf.shareable_formats.push_back(RD::DATA_FORMAT_R8_UINT); + voxel_gi->sdf_texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(voxel_gi->sdf_texture, "VoxelGI SDF Texture"); + } + RID shared_tex; + { + RD::TextureView tv; + tv.format_override = RD::DATA_FORMAT_R8_UINT; + shared_tex = RD::get_singleton()->texture_create_shared(tv, voxel_gi->sdf_texture); + } + //update SDF texture + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 1; + u.append_id(voxel_gi->octree_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 2; + u.append_id(voxel_gi->data_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 3; + u.append_id(shared_tex); + uniforms.push_back(u); + } + + RID uniform_set = RD::get_singleton()->uniform_set_create(uniforms, voxel_gi_sdf_shader_version_shader, 0); + + { + uint32_t push_constant[4] = { 0, 0, 0, 0 }; + + for (int i = 0; i < voxel_gi->level_counts.size() - 1; i++) { + push_constant[0] += voxel_gi->level_counts[i]; + } + push_constant[1] = push_constant[0] + voxel_gi->level_counts[voxel_gi->level_counts.size() - 1]; + + print_line("offset: " + itos(push_constant[0])); + print_line("size: " + itos(push_constant[1])); + //create SDF + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, voxel_gi_sdf_shader_pipeline); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set, 0); + RD::get_singleton()->compute_list_set_push_constant(compute_list, push_constant, sizeof(uint32_t) * 4); + RD::get_singleton()->compute_list_dispatch(compute_list, voxel_gi->octree_size.x / 4, voxel_gi->octree_size.y / 4, voxel_gi->octree_size.z / 4); + RD::get_singleton()->compute_list_end(); + } + + RD::get_singleton()->free(uniform_set); + RD::get_singleton()->free(shared_tex); + } +#endif + } + + voxel_gi->version++; + voxel_gi->data_version++; + + voxel_gi->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); +} + +AABB GI::voxel_gi_get_bounds(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, AABB()); + + return voxel_gi->bounds; +} + +Vector3i GI::voxel_gi_get_octree_size(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, Vector3i()); + return voxel_gi->octree_size; +} + +Vector<uint8_t> GI::voxel_gi_get_octree_cells(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, Vector<uint8_t>()); + + if (voxel_gi->octree_buffer.is_valid()) { + return RD::get_singleton()->buffer_get_data(voxel_gi->octree_buffer); + } + return Vector<uint8_t>(); +} + +Vector<uint8_t> GI::voxel_gi_get_data_cells(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, Vector<uint8_t>()); + + if (voxel_gi->data_buffer.is_valid()) { + return RD::get_singleton()->buffer_get_data(voxel_gi->data_buffer); + } + return Vector<uint8_t>(); +} + +Vector<uint8_t> GI::voxel_gi_get_distance_field(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, Vector<uint8_t>()); + + if (voxel_gi->data_buffer.is_valid()) { + return RD::get_singleton()->texture_get_data(voxel_gi->sdf_texture, 0); + } + return Vector<uint8_t>(); +} + +Vector<int> GI::voxel_gi_get_level_counts(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, Vector<int>()); + + return voxel_gi->level_counts; +} + +Transform3D GI::voxel_gi_get_to_cell_xform(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, Transform3D()); + + return voxel_gi->to_cell_xform; +} + +void GI::voxel_gi_set_dynamic_range(RID p_voxel_gi, float p_range) { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND(!voxel_gi); + + voxel_gi->dynamic_range = p_range; + voxel_gi->version++; +} + +float GI::voxel_gi_get_dynamic_range(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, 0); + + return voxel_gi->dynamic_range; +} + +void GI::voxel_gi_set_propagation(RID p_voxel_gi, float p_range) { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND(!voxel_gi); + + voxel_gi->propagation = p_range; + voxel_gi->version++; +} + +float GI::voxel_gi_get_propagation(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, 0); + return voxel_gi->propagation; +} + +void GI::voxel_gi_set_energy(RID p_voxel_gi, float p_energy) { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND(!voxel_gi); + + voxel_gi->energy = p_energy; +} + +float GI::voxel_gi_get_energy(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, 0); + return voxel_gi->energy; +} + +void GI::voxel_gi_set_baked_exposure_normalization(RID p_voxel_gi, float p_baked_exposure) { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND(!voxel_gi); + + voxel_gi->baked_exposure = p_baked_exposure; +} + +float GI::voxel_gi_get_baked_exposure_normalization(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, 0); + return voxel_gi->baked_exposure; +} + +void GI::voxel_gi_set_bias(RID p_voxel_gi, float p_bias) { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND(!voxel_gi); + + voxel_gi->bias = p_bias; +} + +float GI::voxel_gi_get_bias(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, 0); + return voxel_gi->bias; +} + +void GI::voxel_gi_set_normal_bias(RID p_voxel_gi, float p_normal_bias) { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND(!voxel_gi); + + voxel_gi->normal_bias = p_normal_bias; +} + +float GI::voxel_gi_get_normal_bias(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, 0); + return voxel_gi->normal_bias; +} + +void GI::voxel_gi_set_interior(RID p_voxel_gi, bool p_enable) { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND(!voxel_gi); + + voxel_gi->interior = p_enable; +} + +void GI::voxel_gi_set_use_two_bounces(RID p_voxel_gi, bool p_enable) { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND(!voxel_gi); + + voxel_gi->use_two_bounces = p_enable; + voxel_gi->version++; +} + +bool GI::voxel_gi_is_using_two_bounces(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, false); + return voxel_gi->use_two_bounces; +} + +bool GI::voxel_gi_is_interior(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, 0); + return voxel_gi->interior; +} + +uint32_t GI::voxel_gi_get_version(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, 0); + return voxel_gi->version; +} + +uint32_t GI::voxel_gi_get_data_version(RID p_voxel_gi) { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, 0); + return voxel_gi->data_version; +} + +RID GI::voxel_gi_get_octree_buffer(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, RID()); + return voxel_gi->octree_buffer; +} + +RID GI::voxel_gi_get_data_buffer(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, RID()); + return voxel_gi->data_buffer; +} + +RID GI::voxel_gi_get_sdf_texture(RID p_voxel_gi) { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, RID()); + + return voxel_gi->sdf_texture; +} + +Dependency *GI::voxel_gi_get_dependency(RID p_voxel_gi) const { + VoxelGI *voxel_gi = voxel_gi_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND_V(!voxel_gi, nullptr); + + return &voxel_gi->dependency; +} + +//////////////////////////////////////////////////////////////////////////////// +// SDFGI + +void GI::SDFGI::create(RID p_env, const Vector3 &p_world_position, uint32_t p_requested_history_size, GI *p_gi) { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + + gi = p_gi; + num_cascades = RendererSceneRenderRD::get_singleton()->environment_get_sdfgi_cascades(p_env); + min_cell_size = RendererSceneRenderRD::get_singleton()->environment_get_sdfgi_min_cell_size(p_env); + uses_occlusion = RendererSceneRenderRD::get_singleton()->environment_get_sdfgi_use_occlusion(p_env); + y_scale_mode = RendererSceneRenderRD::get_singleton()->environment_get_sdfgi_y_scale(p_env); + static const float y_scale[3] = { 2.0, 1.5, 1.0 }; + y_mult = y_scale[y_scale_mode]; + cascades.resize(num_cascades); + probe_axis_count = SDFGI::PROBE_DIVISOR + 1; + solid_cell_ratio = gi->sdfgi_solid_cell_ratio; + solid_cell_count = uint32_t(float(cascade_size * cascade_size * cascade_size) * solid_cell_ratio); + + float base_cell_size = min_cell_size; + + RD::TextureFormat tf_sdf; + tf_sdf.format = RD::DATA_FORMAT_R8_UNORM; + tf_sdf.width = cascade_size; // Always 64x64 + tf_sdf.height = cascade_size; + tf_sdf.depth = cascade_size; + tf_sdf.texture_type = RD::TEXTURE_TYPE_3D; + tf_sdf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT; + + { + RD::TextureFormat tf_render = tf_sdf; + tf_render.format = RD::DATA_FORMAT_R16_UINT; + render_albedo = RD::get_singleton()->texture_create(tf_render, RD::TextureView()); + RD::get_singleton()->set_resource_name(render_albedo, "VoxelGI Render Albedo"); + tf_render.format = RD::DATA_FORMAT_R32_UINT; + render_emission = RD::get_singleton()->texture_create(tf_render, RD::TextureView()); + RD::get_singleton()->set_resource_name(render_emission, "VoxelGI Render Emission"); + render_emission_aniso = RD::get_singleton()->texture_create(tf_render, RD::TextureView()); + RD::get_singleton()->set_resource_name(render_emission_aniso, "VoxelGI Render Emission Aniso"); + + tf_render.format = RD::DATA_FORMAT_R8_UNORM; //at least its easy to visualize + + for (int i = 0; i < 8; i++) { + render_occlusion[i] = RD::get_singleton()->texture_create(tf_render, RD::TextureView()); + RD::get_singleton()->set_resource_name(render_occlusion[i], String("VoxelGI Render Occlusion ") + itos(i)); + } + + tf_render.format = RD::DATA_FORMAT_R32_UINT; + render_geom_facing = RD::get_singleton()->texture_create(tf_render, RD::TextureView()); + RD::get_singleton()->set_resource_name(render_geom_facing, "VoxelGI Render Geometry Facing"); + + tf_render.format = RD::DATA_FORMAT_R8G8B8A8_UINT; + render_sdf[0] = RD::get_singleton()->texture_create(tf_render, RD::TextureView()); + RD::get_singleton()->set_resource_name(render_sdf[0], "VoxelGI Render SDF 0"); + render_sdf[1] = RD::get_singleton()->texture_create(tf_render, RD::TextureView()); + RD::get_singleton()->set_resource_name(render_sdf[1], "VoxelGI Render SDF 1"); + + tf_render.width /= 2; + tf_render.height /= 2; + tf_render.depth /= 2; + + render_sdf_half[0] = RD::get_singleton()->texture_create(tf_render, RD::TextureView()); + RD::get_singleton()->set_resource_name(render_sdf_half[0], "VoxelGI Render SDF Half 0"); + render_sdf_half[1] = RD::get_singleton()->texture_create(tf_render, RD::TextureView()); + RD::get_singleton()->set_resource_name(render_sdf_half[1], "VoxelGI Render SDF Half 1"); + } + + RD::TextureFormat tf_occlusion = tf_sdf; + tf_occlusion.format = RD::DATA_FORMAT_R16_UINT; + tf_occlusion.shareable_formats.push_back(RD::DATA_FORMAT_R16_UINT); + tf_occlusion.shareable_formats.push_back(RD::DATA_FORMAT_R4G4B4A4_UNORM_PACK16); + tf_occlusion.depth *= cascades.size(); //use depth for occlusion slices + tf_occlusion.width *= 2; //use width for the other half + + RD::TextureFormat tf_light = tf_sdf; + tf_light.format = RD::DATA_FORMAT_R32_UINT; + tf_light.shareable_formats.push_back(RD::DATA_FORMAT_R32_UINT); + tf_light.shareable_formats.push_back(RD::DATA_FORMAT_E5B9G9R9_UFLOAT_PACK32); + + RD::TextureFormat tf_aniso0 = tf_sdf; + tf_aniso0.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + RD::TextureFormat tf_aniso1 = tf_sdf; + tf_aniso1.format = RD::DATA_FORMAT_R8G8_UNORM; + + int passes = nearest_shift(cascade_size) - 1; + + //store lightprobe SH + RD::TextureFormat tf_probes; + tf_probes.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf_probes.width = probe_axis_count * probe_axis_count; + tf_probes.height = probe_axis_count * SDFGI::SH_SIZE; + tf_probes.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT; + tf_probes.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; + + history_size = p_requested_history_size; + + RD::TextureFormat tf_probe_history = tf_probes; + tf_probe_history.format = RD::DATA_FORMAT_R16G16B16A16_SINT; //signed integer because SH are signed + tf_probe_history.array_layers = history_size; + + RD::TextureFormat tf_probe_average = tf_probes; + tf_probe_average.format = RD::DATA_FORMAT_R32G32B32A32_SINT; //signed integer because SH are signed + tf_probe_average.texture_type = RD::TEXTURE_TYPE_2D; + + lightprobe_history_scroll = RD::get_singleton()->texture_create(tf_probe_history, RD::TextureView()); + RD::get_singleton()->set_resource_name(lightprobe_history_scroll, "VoxelGI LightProbe History Scroll"); + lightprobe_average_scroll = RD::get_singleton()->texture_create(tf_probe_average, RD::TextureView()); + RD::get_singleton()->set_resource_name(lightprobe_average_scroll, "VoxelGI LightProbe Average Scroll"); + + { + //octahedral lightprobes + RD::TextureFormat tf_octprobes = tf_probes; + tf_octprobes.array_layers = cascades.size() * 2; + tf_octprobes.format = RD::DATA_FORMAT_R32_UINT; //pack well with RGBE + tf_octprobes.width = probe_axis_count * probe_axis_count * (SDFGI::LIGHTPROBE_OCT_SIZE + 2); + tf_octprobes.height = probe_axis_count * (SDFGI::LIGHTPROBE_OCT_SIZE + 2); + tf_octprobes.shareable_formats.push_back(RD::DATA_FORMAT_R32_UINT); + tf_octprobes.shareable_formats.push_back(RD::DATA_FORMAT_E5B9G9R9_UFLOAT_PACK32); + //lightprobe texture is an octahedral texture + + lightprobe_data = RD::get_singleton()->texture_create(tf_octprobes, RD::TextureView()); + RD::get_singleton()->set_resource_name(lightprobe_data, "VoxelGI LightProbe Data"); + RD::TextureView tv; + tv.format_override = RD::DATA_FORMAT_E5B9G9R9_UFLOAT_PACK32; + lightprobe_texture = RD::get_singleton()->texture_create_shared(tv, lightprobe_data); + + //texture handling ambient data, to integrate with volumetric foc + RD::TextureFormat tf_ambient = tf_probes; + tf_ambient.array_layers = cascades.size(); + tf_ambient.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; //pack well with RGBE + tf_ambient.width = probe_axis_count * probe_axis_count; + tf_ambient.height = probe_axis_count; + tf_ambient.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; + //lightprobe texture is an octahedral texture + ambient_texture = RD::get_singleton()->texture_create(tf_ambient, RD::TextureView()); + RD::get_singleton()->set_resource_name(ambient_texture, "VoxelGI Ambient Texture"); + } + + cascades_ubo = RD::get_singleton()->uniform_buffer_create(sizeof(SDFGI::Cascade::UBO) * SDFGI::MAX_CASCADES); + + occlusion_data = RD::get_singleton()->texture_create(tf_occlusion, RD::TextureView()); + RD::get_singleton()->set_resource_name(occlusion_data, "VoxelGI Occlusion Data"); + { + RD::TextureView tv; + tv.format_override = RD::DATA_FORMAT_R4G4B4A4_UNORM_PACK16; + occlusion_texture = RD::get_singleton()->texture_create_shared(tv, occlusion_data); + } + + for (uint32_t i = 0; i < cascades.size(); i++) { + SDFGI::Cascade &cascade = cascades[i]; + + /* 3D Textures */ + + cascade.sdf_tex = RD::get_singleton()->texture_create(tf_sdf, RD::TextureView()); + RD::get_singleton()->set_resource_name(cascade.sdf_tex, "VoxelGI Cascade SDF Texture"); + + cascade.light_data = RD::get_singleton()->texture_create(tf_light, RD::TextureView()); + RD::get_singleton()->set_resource_name(cascade.light_data, "VoxelGI Cascade Light Data"); + + cascade.light_aniso_0_tex = RD::get_singleton()->texture_create(tf_aniso0, RD::TextureView()); + RD::get_singleton()->set_resource_name(cascade.light_aniso_0_tex, "VoxelGI Cascade Light Aniso 0 Texture"); + cascade.light_aniso_1_tex = RD::get_singleton()->texture_create(tf_aniso1, RD::TextureView()); + RD::get_singleton()->set_resource_name(cascade.light_aniso_1_tex, "VoxelGI Cascade Light Aniso 1 Texture"); + + { + RD::TextureView tv; + tv.format_override = RD::DATA_FORMAT_E5B9G9R9_UFLOAT_PACK32; + cascade.light_tex = RD::get_singleton()->texture_create_shared(tv, cascade.light_data); + + RD::get_singleton()->texture_clear(cascade.light_tex, Color(0, 0, 0, 0), 0, 1, 0, 1); + RD::get_singleton()->texture_clear(cascade.light_aniso_0_tex, Color(0, 0, 0, 0), 0, 1, 0, 1); + RD::get_singleton()->texture_clear(cascade.light_aniso_1_tex, Color(0, 0, 0, 0), 0, 1, 0, 1); + } + + cascade.cell_size = base_cell_size; + Vector3 world_position = p_world_position; + world_position.y *= y_mult; + int32_t probe_cells = cascade_size / SDFGI::PROBE_DIVISOR; + Vector3 probe_size = Vector3(1, 1, 1) * cascade.cell_size * probe_cells; + Vector3i probe_pos = Vector3i((world_position / probe_size + Vector3(0.5, 0.5, 0.5)).floor()); + cascade.position = probe_pos * probe_cells; + + cascade.dirty_regions = SDFGI::Cascade::DIRTY_ALL; + + base_cell_size *= 2.0; + + /* Probe History */ + + cascade.lightprobe_history_tex = RD::get_singleton()->texture_create(tf_probe_history, RD::TextureView()); + RD::get_singleton()->set_resource_name(cascade.lightprobe_history_tex, "VoxelGI Cascade LightProbe History Texture"); + RD::get_singleton()->texture_clear(cascade.lightprobe_history_tex, Color(0, 0, 0, 0), 0, 1, 0, tf_probe_history.array_layers); //needs to be cleared for average to work + + cascade.lightprobe_average_tex = RD::get_singleton()->texture_create(tf_probe_average, RD::TextureView()); + RD::get_singleton()->set_resource_name(cascade.lightprobe_average_tex, "VoxelGI Cascade LightProbe Average Texture"); + RD::get_singleton()->texture_clear(cascade.lightprobe_average_tex, Color(0, 0, 0, 0), 0, 1, 0, 1); //needs to be cleared for average to work + + /* Buffers */ + + cascade.solid_cell_buffer = RD::get_singleton()->storage_buffer_create(sizeof(SDFGI::Cascade::SolidCell) * solid_cell_count); + cascade.solid_cell_dispatch_buffer = RD::get_singleton()->storage_buffer_create(sizeof(uint32_t) * 4, Vector<uint8_t>(), RD::STORAGE_BUFFER_USAGE_DISPATCH_INDIRECT); + cascade.lights_buffer = RD::get_singleton()->storage_buffer_create(sizeof(SDFGIShader::Light) * MAX(SDFGI::MAX_STATIC_LIGHTS, SDFGI::MAX_DYNAMIC_LIGHTS)); + { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 1; + u.append_id(render_sdf[(passes & 1) ? 1 : 0]); //if passes are even, we read from buffer 0, else we read from buffer 1 + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 2; + u.append_id(render_albedo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 3; + for (int j = 0; j < 8; j++) { + u.append_id(render_occlusion[j]); + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 4; + u.append_id(render_emission); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 5; + u.append_id(render_emission_aniso); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 6; + u.append_id(render_geom_facing); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 7; + u.append_id(cascade.sdf_tex); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 8; + u.append_id(occlusion_data); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 10; + u.append_id(cascade.solid_cell_dispatch_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 11; + u.append_id(cascade.solid_cell_buffer); + uniforms.push_back(u); + } + + cascade.sdf_store_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->sdfgi_shader.preprocess.version_get_shader(gi->sdfgi_shader.preprocess_shader, SDFGIShader::PRE_PROCESS_STORE), 0); + } + + { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 1; + u.append_id(render_albedo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 2; + u.append_id(render_geom_facing); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 3; + u.append_id(render_emission); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 4; + u.append_id(render_emission_aniso); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 5; + u.append_id(cascade.solid_cell_dispatch_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 6; + u.append_id(cascade.solid_cell_buffer); + uniforms.push_back(u); + } + + cascade.scroll_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->sdfgi_shader.preprocess.version_get_shader(gi->sdfgi_shader.preprocess_shader, SDFGIShader::PRE_PROCESS_SCROLL), 0); + } + { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 1; + for (int j = 0; j < 8; j++) { + u.append_id(render_occlusion[j]); + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 2; + u.append_id(occlusion_data); + uniforms.push_back(u); + } + + cascade.scroll_occlusion_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->sdfgi_shader.preprocess.version_get_shader(gi->sdfgi_shader.preprocess_shader, SDFGIShader::PRE_PROCESS_SCROLL_OCCLUSION), 0); + } + } + + //direct light + for (uint32_t i = 0; i < cascades.size(); i++) { + SDFGI::Cascade &cascade = cascades[i]; + + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.binding = 1; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + for (uint32_t j = 0; j < SDFGI::MAX_CASCADES; j++) { + if (j < cascades.size()) { + u.append_id(cascades[j].sdf_tex); + } else { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE)); + } + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 2; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.append_id(material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED)); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 3; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.append_id(cascade.solid_cell_dispatch_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 4; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.append_id(cascade.solid_cell_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 5; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.append_id(cascade.light_data); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 6; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.append_id(cascade.light_aniso_0_tex); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 7; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.append_id(cascade.light_aniso_1_tex); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 8; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.append_id(cascades_ubo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 9; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.append_id(cascade.lights_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 10; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.append_id(lightprobe_texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 11; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.append_id(occlusion_texture); + uniforms.push_back(u); + } + + cascade.sdf_direct_light_static_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->sdfgi_shader.direct_light.version_get_shader(gi->sdfgi_shader.direct_light_shader, SDFGIShader::DIRECT_LIGHT_MODE_STATIC), 0); + cascade.sdf_direct_light_dynamic_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->sdfgi_shader.direct_light.version_get_shader(gi->sdfgi_shader.direct_light_shader, SDFGIShader::DIRECT_LIGHT_MODE_DYNAMIC), 0); + } + + //preprocess initialize uniform set + { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 1; + u.append_id(render_albedo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 2; + u.append_id(render_sdf[0]); + uniforms.push_back(u); + } + + sdf_initialize_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->sdfgi_shader.preprocess.version_get_shader(gi->sdfgi_shader.preprocess_shader, SDFGIShader::PRE_PROCESS_JUMP_FLOOD_INITIALIZE), 0); + } + + { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 1; + u.append_id(render_albedo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 2; + u.append_id(render_sdf_half[0]); + uniforms.push_back(u); + } + + sdf_initialize_half_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->sdfgi_shader.preprocess.version_get_shader(gi->sdfgi_shader.preprocess_shader, SDFGIShader::PRE_PROCESS_JUMP_FLOOD_INITIALIZE_HALF), 0); + } + + //jump flood uniform set + { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 1; + u.append_id(render_sdf[0]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 2; + u.append_id(render_sdf[1]); + uniforms.push_back(u); + } + + jump_flood_uniform_set[0] = RD::get_singleton()->uniform_set_create(uniforms, gi->sdfgi_shader.preprocess.version_get_shader(gi->sdfgi_shader.preprocess_shader, SDFGIShader::PRE_PROCESS_JUMP_FLOOD), 0); + RID aux0 = uniforms.write[0].get_id(0); + RID aux1 = uniforms.write[1].get_id(0); + uniforms.write[0].set_id(0, aux1); + uniforms.write[1].set_id(0, aux0); + jump_flood_uniform_set[1] = RD::get_singleton()->uniform_set_create(uniforms, gi->sdfgi_shader.preprocess.version_get_shader(gi->sdfgi_shader.preprocess_shader, SDFGIShader::PRE_PROCESS_JUMP_FLOOD), 0); + } + //jump flood half uniform set + { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 1; + u.append_id(render_sdf_half[0]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 2; + u.append_id(render_sdf_half[1]); + uniforms.push_back(u); + } + + jump_flood_half_uniform_set[0] = RD::get_singleton()->uniform_set_create(uniforms, gi->sdfgi_shader.preprocess.version_get_shader(gi->sdfgi_shader.preprocess_shader, SDFGIShader::PRE_PROCESS_JUMP_FLOOD), 0); + RID aux0 = uniforms.write[0].get_id(0); + RID aux1 = uniforms.write[1].get_id(0); + uniforms.write[0].set_id(0, aux1); + uniforms.write[1].set_id(0, aux0); + jump_flood_half_uniform_set[1] = RD::get_singleton()->uniform_set_create(uniforms, gi->sdfgi_shader.preprocess.version_get_shader(gi->sdfgi_shader.preprocess_shader, SDFGIShader::PRE_PROCESS_JUMP_FLOOD), 0); + } + + //upscale half size sdf + { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 1; + u.append_id(render_albedo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 2; + u.append_id(render_sdf_half[(passes & 1) ? 0 : 1]); //reverse pass order because half size + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 3; + u.append_id(render_sdf[(passes & 1) ? 0 : 1]); //reverse pass order because it needs an extra JFA pass + uniforms.push_back(u); + } + + upscale_jfa_uniform_set_index = (passes & 1) ? 0 : 1; + sdf_upscale_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->sdfgi_shader.preprocess.version_get_shader(gi->sdfgi_shader.preprocess_shader, SDFGIShader::PRE_PROCESS_JUMP_FLOOD_UPSCALE), 0); + } + + //occlusion uniform set + { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 1; + u.append_id(render_albedo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 2; + for (int i = 0; i < 8; i++) { + u.append_id(render_occlusion[i]); + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 3; + u.append_id(render_geom_facing); + uniforms.push_back(u); + } + + occlusion_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->sdfgi_shader.preprocess.version_get_shader(gi->sdfgi_shader.preprocess_shader, SDFGIShader::PRE_PROCESS_OCCLUSION), 0); + } + + for (uint32_t i = 0; i < cascades.size(); i++) { + //integrate uniform + + Vector<RD::Uniform> uniforms; + + { + RD::Uniform u; + u.binding = 1; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + for (uint32_t j = 0; j < SDFGI::MAX_CASCADES; j++) { + if (j < cascades.size()) { + u.append_id(cascades[j].sdf_tex); + } else { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE)); + } + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 2; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + for (uint32_t j = 0; j < SDFGI::MAX_CASCADES; j++) { + if (j < cascades.size()) { + u.append_id(cascades[j].light_tex); + } else { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE)); + } + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 3; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + for (uint32_t j = 0; j < SDFGI::MAX_CASCADES; j++) { + if (j < cascades.size()) { + u.append_id(cascades[j].light_aniso_0_tex); + } else { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE)); + } + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 4; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + for (uint32_t j = 0; j < SDFGI::MAX_CASCADES; j++) { + if (j < cascades.size()) { + u.append_id(cascades[j].light_aniso_1_tex); + } else { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE)); + } + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 6; + u.append_id(material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED)); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 7; + u.append_id(cascades_ubo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 8; + u.append_id(lightprobe_data); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 9; + u.append_id(cascades[i].lightprobe_history_tex); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 10; + u.append_id(cascades[i].lightprobe_average_tex); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 11; + u.append_id(lightprobe_history_scroll); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 12; + u.append_id(lightprobe_average_scroll); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 13; + RID parent_average; + if (cascades.size() == 1) { + // If there is only one SDFGI cascade, we can't use the previous cascade for blending. + parent_average = cascades[i].lightprobe_average_tex; + } else if (i < cascades.size() - 1) { + parent_average = cascades[i + 1].lightprobe_average_tex; + } else { + parent_average = cascades[i - 1].lightprobe_average_tex; //to use something, but it won't be used + } + u.append_id(parent_average); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 14; + u.append_id(ambient_texture); + uniforms.push_back(u); + } + + cascades[i].integrate_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->sdfgi_shader.integrate.version_get_shader(gi->sdfgi_shader.integrate_shader, 0), 0); + } + + bounce_feedback = RendererSceneRenderRD::get_singleton()->environment_get_sdfgi_bounce_feedback(p_env); + energy = RendererSceneRenderRD::get_singleton()->environment_get_sdfgi_energy(p_env); + normal_bias = RendererSceneRenderRD::get_singleton()->environment_get_sdfgi_normal_bias(p_env); + probe_bias = RendererSceneRenderRD::get_singleton()->environment_get_sdfgi_probe_bias(p_env); + reads_sky = RendererSceneRenderRD::get_singleton()->environment_get_sdfgi_read_sky_light(p_env); +} + +void GI::SDFGI::free_data() { + // we don't free things here, we handle SDFGI differently at the moment destructing the object when it needs to change. +} + +GI::SDFGI::~SDFGI() { + for (uint32_t i = 0; i < cascades.size(); i++) { + const SDFGI::Cascade &c = cascades[i]; + RD::get_singleton()->free(c.light_data); + RD::get_singleton()->free(c.light_aniso_0_tex); + RD::get_singleton()->free(c.light_aniso_1_tex); + RD::get_singleton()->free(c.sdf_tex); + RD::get_singleton()->free(c.solid_cell_dispatch_buffer); + RD::get_singleton()->free(c.solid_cell_buffer); + RD::get_singleton()->free(c.lightprobe_history_tex); + RD::get_singleton()->free(c.lightprobe_average_tex); + RD::get_singleton()->free(c.lights_buffer); + } + + RD::get_singleton()->free(render_albedo); + RD::get_singleton()->free(render_emission); + RD::get_singleton()->free(render_emission_aniso); + + RD::get_singleton()->free(render_sdf[0]); + RD::get_singleton()->free(render_sdf[1]); + + RD::get_singleton()->free(render_sdf_half[0]); + RD::get_singleton()->free(render_sdf_half[1]); + + for (int i = 0; i < 8; i++) { + RD::get_singleton()->free(render_occlusion[i]); + } + + RD::get_singleton()->free(render_geom_facing); + + RD::get_singleton()->free(lightprobe_data); + RD::get_singleton()->free(lightprobe_history_scroll); + RD::get_singleton()->free(lightprobe_average_scroll); + RD::get_singleton()->free(occlusion_data); + RD::get_singleton()->free(ambient_texture); + + RD::get_singleton()->free(cascades_ubo); + + for (uint32_t v = 0; v < RendererSceneRender::MAX_RENDER_VIEWS; v++) { + if (RD::get_singleton()->uniform_set_is_valid(debug_uniform_set[v])) { + RD::get_singleton()->free(debug_uniform_set[v]); + } + debug_uniform_set[v] = RID(); + } + + if (RD::get_singleton()->uniform_set_is_valid(debug_probes_uniform_set)) { + RD::get_singleton()->free(debug_probes_uniform_set); + } + debug_probes_uniform_set = RID(); + + if (debug_probes_scene_data_ubo.is_valid()) { + RD::get_singleton()->free(debug_probes_scene_data_ubo); + debug_probes_scene_data_ubo = RID(); + } +} + +void GI::SDFGI::update(RID p_env, const Vector3 &p_world_position) { + bounce_feedback = RendererSceneRenderRD::get_singleton()->environment_get_sdfgi_bounce_feedback(p_env); + energy = RendererSceneRenderRD::get_singleton()->environment_get_sdfgi_energy(p_env); + normal_bias = RendererSceneRenderRD::get_singleton()->environment_get_sdfgi_normal_bias(p_env); + probe_bias = RendererSceneRenderRD::get_singleton()->environment_get_sdfgi_probe_bias(p_env); + reads_sky = RendererSceneRenderRD::get_singleton()->environment_get_sdfgi_read_sky_light(p_env); + + int32_t drag_margin = (cascade_size / SDFGI::PROBE_DIVISOR) / 2; + + for (uint32_t i = 0; i < cascades.size(); i++) { + SDFGI::Cascade &cascade = cascades[i]; + cascade.dirty_regions = Vector3i(); + + Vector3 probe_half_size = Vector3(1, 1, 1) * cascade.cell_size * float(cascade_size / SDFGI::PROBE_DIVISOR) * 0.5; + probe_half_size = Vector3(0, 0, 0); + + Vector3 world_position = p_world_position; + world_position.y *= y_mult; + Vector3i pos_in_cascade = Vector3i((world_position + probe_half_size) / cascade.cell_size); + + for (int j = 0; j < 3; j++) { + if (pos_in_cascade[j] < cascade.position[j]) { + while (pos_in_cascade[j] < (cascade.position[j] - drag_margin)) { + cascade.position[j] -= drag_margin * 2; + cascade.dirty_regions[j] += drag_margin * 2; + } + } else if (pos_in_cascade[j] > cascade.position[j]) { + while (pos_in_cascade[j] > (cascade.position[j] + drag_margin)) { + cascade.position[j] += drag_margin * 2; + cascade.dirty_regions[j] -= drag_margin * 2; + } + } + + if (cascade.dirty_regions[j] == 0) { + continue; // not dirty + } else if (uint32_t(ABS(cascade.dirty_regions[j])) >= cascade_size) { + //moved too much, just redraw everything (make all dirty) + cascade.dirty_regions = SDFGI::Cascade::DIRTY_ALL; + break; + } + } + + if (cascade.dirty_regions != Vector3i() && cascade.dirty_regions != SDFGI::Cascade::DIRTY_ALL) { + //see how much the total dirty volume represents from the total volume + uint32_t total_volume = cascade_size * cascade_size * cascade_size; + uint32_t safe_volume = 1; + for (int j = 0; j < 3; j++) { + safe_volume *= cascade_size - ABS(cascade.dirty_regions[j]); + } + uint32_t dirty_volume = total_volume - safe_volume; + if (dirty_volume > (safe_volume / 2)) { + //more than half the volume is dirty, make all dirty so its only rendered once + cascade.dirty_regions = SDFGI::Cascade::DIRTY_ALL; + } + } + } +} + +void GI::SDFGI::update_light() { + RD::get_singleton()->draw_command_begin_label("SDFGI Update dynamic Light"); + + /* Update dynamic light */ + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->sdfgi_shader.direct_light_pipeline[SDFGIShader::DIRECT_LIGHT_MODE_DYNAMIC]); + + SDFGIShader::DirectLightPushConstant push_constant; + + push_constant.grid_size[0] = cascade_size; + push_constant.grid_size[1] = cascade_size; + push_constant.grid_size[2] = cascade_size; + push_constant.max_cascades = cascades.size(); + push_constant.probe_axis_size = probe_axis_count; + push_constant.bounce_feedback = bounce_feedback; + push_constant.y_mult = y_mult; + push_constant.use_occlusion = uses_occlusion; + + for (uint32_t i = 0; i < cascades.size(); i++) { + SDFGI::Cascade &cascade = cascades[i]; + push_constant.light_count = cascade_dynamic_light_count[i]; + push_constant.cascade = i; + + if (cascades[i].all_dynamic_lights_dirty || gi->sdfgi_frames_to_update_light == RS::ENV_SDFGI_UPDATE_LIGHT_IN_1_FRAME) { + push_constant.process_offset = 0; + push_constant.process_increment = 1; + } else { + static const uint32_t frames_to_update_table[RS::ENV_SDFGI_UPDATE_LIGHT_MAX] = { + 1, 2, 4, 8, 16 + }; + + uint32_t frames_to_update = frames_to_update_table[gi->sdfgi_frames_to_update_light]; + + push_constant.process_offset = RSG::rasterizer->get_frame_number() % frames_to_update; + push_constant.process_increment = frames_to_update; + } + cascades[i].all_dynamic_lights_dirty = false; + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, cascade.sdf_direct_light_dynamic_uniform_set, 0); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDFGIShader::DirectLightPushConstant)); + RD::get_singleton()->compute_list_dispatch_indirect(compute_list, cascade.solid_cell_dispatch_buffer, 0); + } + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_COMPUTE); + RD::get_singleton()->draw_command_end_label(); +} + +void GI::SDFGI::update_probes(RID p_env, SkyRD::Sky *p_sky) { + RD::get_singleton()->draw_command_begin_label("SDFGI Update Probes"); + + SDFGIShader::IntegratePushConstant push_constant; + push_constant.grid_size[1] = cascade_size; + push_constant.grid_size[2] = cascade_size; + push_constant.grid_size[0] = cascade_size; + push_constant.max_cascades = cascades.size(); + push_constant.probe_axis_size = probe_axis_count; + push_constant.history_index = render_pass % history_size; + push_constant.history_size = history_size; + static const uint32_t ray_count[RS::ENV_SDFGI_RAY_COUNT_MAX] = { 4, 8, 16, 32, 64, 96, 128 }; + push_constant.ray_count = ray_count[gi->sdfgi_ray_count]; + push_constant.ray_bias = probe_bias; + push_constant.image_size[0] = probe_axis_count * probe_axis_count; + push_constant.image_size[1] = probe_axis_count; + push_constant.store_ambient_texture = RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_enabled(p_env); + + RID sky_uniform_set = gi->sdfgi_shader.integrate_default_sky_uniform_set; + push_constant.sky_mode = SDFGIShader::IntegratePushConstant::SKY_MODE_DISABLED; + push_constant.y_mult = y_mult; + + if (reads_sky && p_env.is_valid()) { + push_constant.sky_energy = RendererSceneRenderRD::get_singleton()->environment_get_bg_energy_multiplier(p_env); + + if (RendererSceneRenderRD::get_singleton()->environment_get_background(p_env) == RS::ENV_BG_CLEAR_COLOR) { + push_constant.sky_mode = SDFGIShader::IntegratePushConstant::SKY_MODE_COLOR; + Color c = RSG::texture_storage->get_default_clear_color().srgb_to_linear(); + push_constant.sky_color[0] = c.r; + push_constant.sky_color[1] = c.g; + push_constant.sky_color[2] = c.b; + } else if (RendererSceneRenderRD::get_singleton()->environment_get_background(p_env) == RS::ENV_BG_COLOR) { + push_constant.sky_mode = SDFGIShader::IntegratePushConstant::SKY_MODE_COLOR; + Color c = RendererSceneRenderRD::get_singleton()->environment_get_bg_color(p_env); + push_constant.sky_color[0] = c.r; + push_constant.sky_color[1] = c.g; + push_constant.sky_color[2] = c.b; + + } else if (RendererSceneRenderRD::get_singleton()->environment_get_background(p_env) == RS::ENV_BG_SKY) { + if (p_sky && p_sky->radiance.is_valid()) { + if (integrate_sky_uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(integrate_sky_uniform_set)) { + Vector<RD::Uniform> uniforms; + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 0; + u.append_id(p_sky->radiance); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 1; + u.append_id(RendererRD::MaterialStorage::get_singleton()->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED)); + uniforms.push_back(u); + } + + integrate_sky_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->sdfgi_shader.integrate.version_get_shader(gi->sdfgi_shader.integrate_shader, 0), 1); + } + sky_uniform_set = integrate_sky_uniform_set; + push_constant.sky_mode = SDFGIShader::IntegratePushConstant::SKY_MODE_SKY; + } + } + } + + render_pass++; + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(true); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->sdfgi_shader.integrate_pipeline[SDFGIShader::INTEGRATE_MODE_PROCESS]); + + int32_t probe_divisor = cascade_size / SDFGI::PROBE_DIVISOR; + for (uint32_t i = 0; i < cascades.size(); i++) { + push_constant.cascade = i; + push_constant.world_offset[0] = cascades[i].position.x / probe_divisor; + push_constant.world_offset[1] = cascades[i].position.y / probe_divisor; + push_constant.world_offset[2] = cascades[i].position.z / probe_divisor; + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, cascades[i].integrate_uniform_set, 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, sky_uniform_set, 1); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDFGIShader::IntegratePushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, probe_axis_count * probe_axis_count, probe_axis_count, 1); + } + + //end later after raster to avoid barriering on layout changes + //RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_NO_BARRIER); + + RD::get_singleton()->draw_command_end_label(); +} + +void GI::SDFGI::store_probes() { + RD::get_singleton()->barrier(RD::BARRIER_MASK_COMPUTE, RD::BARRIER_MASK_COMPUTE); + RD::get_singleton()->draw_command_begin_label("SDFGI Store Probes"); + + SDFGIShader::IntegratePushConstant push_constant; + push_constant.grid_size[1] = cascade_size; + push_constant.grid_size[2] = cascade_size; + push_constant.grid_size[0] = cascade_size; + push_constant.max_cascades = cascades.size(); + push_constant.probe_axis_size = probe_axis_count; + push_constant.history_index = render_pass % history_size; + push_constant.history_size = history_size; + static const uint32_t ray_count[RS::ENV_SDFGI_RAY_COUNT_MAX] = { 4, 8, 16, 32, 64, 96, 128 }; + push_constant.ray_count = ray_count[gi->sdfgi_ray_count]; + push_constant.ray_bias = probe_bias; + push_constant.image_size[0] = probe_axis_count * probe_axis_count; + push_constant.image_size[1] = probe_axis_count; + push_constant.store_ambient_texture = false; + + push_constant.sky_mode = 0; + push_constant.y_mult = y_mult; + + // Then store values into the lightprobe texture. Separating these steps has a small performance hit, but it allows for multiple bounces + RENDER_TIMESTAMP("Average SDFGI Probes"); + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->sdfgi_shader.integrate_pipeline[SDFGIShader::INTEGRATE_MODE_STORE]); + + //convert to octahedral to store + push_constant.image_size[0] *= SDFGI::LIGHTPROBE_OCT_SIZE; + push_constant.image_size[1] *= SDFGI::LIGHTPROBE_OCT_SIZE; + + for (uint32_t i = 0; i < cascades.size(); i++) { + push_constant.cascade = i; + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, cascades[i].integrate_uniform_set, 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, gi->sdfgi_shader.integrate_default_sky_uniform_set, 1); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDFGIShader::IntegratePushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, probe_axis_count * probe_axis_count * SDFGI::LIGHTPROBE_OCT_SIZE, probe_axis_count * SDFGI::LIGHTPROBE_OCT_SIZE, 1); + } + + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_COMPUTE); + + RD::get_singleton()->draw_command_end_label(); +} + +int GI::SDFGI::get_pending_region_data(int p_region, Vector3i &r_local_offset, Vector3i &r_local_size, AABB &r_bounds) const { + int dirty_count = 0; + for (uint32_t i = 0; i < cascades.size(); i++) { + const SDFGI::Cascade &c = cascades[i]; + + if (c.dirty_regions == SDFGI::Cascade::DIRTY_ALL) { + if (dirty_count == p_region) { + r_local_offset = Vector3i(); + r_local_size = Vector3i(1, 1, 1) * cascade_size; + + r_bounds.position = Vector3((Vector3i(1, 1, 1) * -int32_t(cascade_size >> 1) + c.position)) * c.cell_size * Vector3(1, 1.0 / y_mult, 1); + r_bounds.size = Vector3(r_local_size) * c.cell_size * Vector3(1, 1.0 / y_mult, 1); + return i; + } + dirty_count++; + } else { + for (int j = 0; j < 3; j++) { + if (c.dirty_regions[j] != 0) { + if (dirty_count == p_region) { + Vector3i from = Vector3i(0, 0, 0); + Vector3i to = Vector3i(1, 1, 1) * cascade_size; + + if (c.dirty_regions[j] > 0) { + //fill from the beginning + to[j] = c.dirty_regions[j]; + } else { + //fill from the end + from[j] = to[j] + c.dirty_regions[j]; + } + + for (int k = 0; k < j; k++) { + // "chip" away previous regions to avoid re-voxelizing the same thing + if (c.dirty_regions[k] > 0) { + from[k] += c.dirty_regions[k]; + } else if (c.dirty_regions[k] < 0) { + to[k] += c.dirty_regions[k]; + } + } + + r_local_offset = from; + r_local_size = to - from; + + r_bounds.position = Vector3(from + Vector3i(1, 1, 1) * -int32_t(cascade_size >> 1) + c.position) * c.cell_size * Vector3(1, 1.0 / y_mult, 1); + r_bounds.size = Vector3(r_local_size) * c.cell_size * Vector3(1, 1.0 / y_mult, 1); + + return i; + } + + dirty_count++; + } + } + } + } + return -1; +} + +void GI::SDFGI::update_cascades() { + //update cascades + SDFGI::Cascade::UBO cascade_data[SDFGI::MAX_CASCADES]; + int32_t probe_divisor = cascade_size / SDFGI::PROBE_DIVISOR; + + for (uint32_t i = 0; i < cascades.size(); i++) { + Vector3 pos = Vector3((Vector3i(1, 1, 1) * -int32_t(cascade_size >> 1) + cascades[i].position)) * cascades[i].cell_size; + + cascade_data[i].offset[0] = pos.x; + cascade_data[i].offset[1] = pos.y; + cascade_data[i].offset[2] = pos.z; + cascade_data[i].to_cell = 1.0 / cascades[i].cell_size; + cascade_data[i].probe_offset[0] = cascades[i].position.x / probe_divisor; + cascade_data[i].probe_offset[1] = cascades[i].position.y / probe_divisor; + cascade_data[i].probe_offset[2] = cascades[i].position.z / probe_divisor; + cascade_data[i].pad = 0; + } + + RD::get_singleton()->buffer_update(cascades_ubo, 0, sizeof(SDFGI::Cascade::UBO) * SDFGI::MAX_CASCADES, cascade_data, RD::BARRIER_MASK_COMPUTE); +} + +void GI::SDFGI::debug_draw(uint32_t p_view_count, const Projection *p_projections, const Transform3D &p_transform, int p_width, int p_height, RID p_render_target, RID p_texture, const Vector<RID> &p_texture_views) { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + RendererRD::CopyEffects *copy_effects = RendererRD::CopyEffects::get_singleton(); + + for (uint32_t v = 0; v < p_view_count; v++) { + if (!debug_uniform_set[v].is_valid() || !RD::get_singleton()->uniform_set_is_valid(debug_uniform_set[v])) { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.binding = 1; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + for (uint32_t i = 0; i < SDFGI::MAX_CASCADES; i++) { + if (i < cascades.size()) { + u.append_id(cascades[i].sdf_tex); + } else { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE)); + } + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 2; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + for (uint32_t i = 0; i < SDFGI::MAX_CASCADES; i++) { + if (i < cascades.size()) { + u.append_id(cascades[i].light_tex); + } else { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE)); + } + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 3; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + for (uint32_t i = 0; i < SDFGI::MAX_CASCADES; i++) { + if (i < cascades.size()) { + u.append_id(cascades[i].light_aniso_0_tex); + } else { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE)); + } + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 4; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + for (uint32_t i = 0; i < SDFGI::MAX_CASCADES; i++) { + if (i < cascades.size()) { + u.append_id(cascades[i].light_aniso_1_tex); + } else { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE)); + } + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 5; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.append_id(occlusion_texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 8; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.append_id(material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED)); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 9; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.append_id(cascades_ubo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 10; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.append_id(p_texture_views[v]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 11; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.append_id(lightprobe_texture); + uniforms.push_back(u); + } + debug_uniform_set[v] = RD::get_singleton()->uniform_set_create(uniforms, gi->sdfgi_shader.debug_shader_version, 0); + } + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->sdfgi_shader.debug_pipeline); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, debug_uniform_set[v], 0); + + SDFGIShader::DebugPushConstant push_constant; + push_constant.grid_size[0] = cascade_size; + push_constant.grid_size[1] = cascade_size; + push_constant.grid_size[2] = cascade_size; + push_constant.max_cascades = cascades.size(); + push_constant.screen_size[0] = p_width; + push_constant.screen_size[1] = p_height; + push_constant.y_mult = y_mult; + + push_constant.z_near = -p_projections[v].get_z_near(); + + for (int i = 0; i < 3; i++) { + for (int j = 0; j < 3; j++) { + push_constant.cam_basis[i][j] = p_transform.basis.rows[j][i]; + } + } + push_constant.cam_origin[0] = p_transform.origin[0]; + push_constant.cam_origin[1] = p_transform.origin[1]; + push_constant.cam_origin[2] = p_transform.origin[2]; + + // need to properly unproject for asymmetric projection matrices in stereo.. + Projection inv_projection = p_projections[v].inverse(); + for (int i = 0; i < 4; i++) { + for (int j = 0; j < 3; j++) { + push_constant.inv_projection[j][i] = inv_projection.columns[i][j]; + } + } + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDFGIShader::DebugPushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_width, p_height, 1); + RD::get_singleton()->compute_list_end(); + } + + Size2i rtsize = texture_storage->render_target_get_size(p_render_target); + copy_effects->copy_to_fb_rect(p_texture, texture_storage->render_target_get_rd_framebuffer(p_render_target), Rect2i(Point2i(), rtsize), true, false, false, false, RID(), p_view_count > 1); +} + +void GI::SDFGI::debug_probes(RID p_framebuffer, const uint32_t p_view_count, const Projection *p_camera_with_transforms, bool p_will_continue_color, bool p_will_continue_depth) { + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + + // setup scene data + { + SDFGIShader::DebugProbesSceneData scene_data; + + if (debug_probes_scene_data_ubo.is_null()) { + debug_probes_scene_data_ubo = RD::get_singleton()->uniform_buffer_create(sizeof(SDFGIShader::DebugProbesSceneData)); + } + + for (uint32_t v = 0; v < p_view_count; v++) { + RendererRD::MaterialStorage::store_camera(p_camera_with_transforms[v], scene_data.projection[v]); + } + + RD::get_singleton()->buffer_update(debug_probes_scene_data_ubo, 0, sizeof(SDFGIShader::DebugProbesSceneData), &scene_data, RD::BARRIER_MASK_RASTER); + } + + // setup push constant + SDFGIShader::DebugProbesPushConstant push_constant; + + //gen spheres from strips + uint32_t band_points = 16; + push_constant.band_power = 4; + push_constant.sections_in_band = ((band_points / 2) - 1); + push_constant.band_mask = band_points - 2; + push_constant.section_arc = Math_TAU / float(push_constant.sections_in_band); + push_constant.y_mult = y_mult; + + uint32_t total_points = push_constant.sections_in_band * band_points; + uint32_t total_probes = probe_axis_count * probe_axis_count * probe_axis_count; + + push_constant.grid_size[0] = cascade_size; + push_constant.grid_size[1] = cascade_size; + push_constant.grid_size[2] = cascade_size; + push_constant.cascade = 0; + + push_constant.probe_axis_size = probe_axis_count; + + if (!debug_probes_uniform_set.is_valid() || !RD::get_singleton()->uniform_set_is_valid(debug_probes_uniform_set)) { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.binding = 1; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.append_id(cascades_ubo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 2; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.append_id(lightprobe_texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 3; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.append_id(material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED)); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 4; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.append_id(occlusion_texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 5; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.append_id(debug_probes_scene_data_ubo); + uniforms.push_back(u); + } + + debug_probes_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->sdfgi_shader.debug_probes.version_get_shader(gi->sdfgi_shader.debug_probes_shader, 0), 0); + } + + SDFGIShader::ProbeDebugMode mode = p_view_count > 1 ? SDFGIShader::PROBE_DEBUG_PROBES_MULTIVIEW : SDFGIShader::PROBE_DEBUG_PROBES; + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_framebuffer, RD::INITIAL_ACTION_CONTINUE, p_will_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CONTINUE, p_will_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ); + RD::get_singleton()->draw_command_begin_label("Debug SDFGI"); + + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, gi->sdfgi_shader.debug_probes_pipeline[mode].get_render_pipeline(RD::INVALID_FORMAT_ID, RD::get_singleton()->framebuffer_get_format(p_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, debug_probes_uniform_set, 0); + RD::get_singleton()->draw_list_set_push_constant(draw_list, &push_constant, sizeof(SDFGIShader::DebugProbesPushConstant)); + RD::get_singleton()->draw_list_draw(draw_list, false, total_probes, total_points); + + if (gi->sdfgi_debug_probe_dir != Vector3()) { + uint32_t cascade = 0; + Vector3 offset = Vector3((Vector3i(1, 1, 1) * -int32_t(cascade_size >> 1) + cascades[cascade].position)) * cascades[cascade].cell_size * Vector3(1.0, 1.0 / y_mult, 1.0); + Vector3 probe_size = cascades[cascade].cell_size * (cascade_size / SDFGI::PROBE_DIVISOR) * Vector3(1.0, 1.0 / y_mult, 1.0); + Vector3 ray_from = gi->sdfgi_debug_probe_pos; + Vector3 ray_to = gi->sdfgi_debug_probe_pos + gi->sdfgi_debug_probe_dir * cascades[cascade].cell_size * Math::sqrt(3.0) * cascade_size; + float sphere_radius = 0.2; + float closest_dist = 1e20; + gi->sdfgi_debug_probe_enabled = false; + + Vector3i probe_from = cascades[cascade].position / (cascade_size / SDFGI::PROBE_DIVISOR); + for (int i = 0; i < (SDFGI::PROBE_DIVISOR + 1); i++) { + for (int j = 0; j < (SDFGI::PROBE_DIVISOR + 1); j++) { + for (int k = 0; k < (SDFGI::PROBE_DIVISOR + 1); k++) { + Vector3 pos = offset + probe_size * Vector3(i, j, k); + Vector3 res; + if (Geometry3D::segment_intersects_sphere(ray_from, ray_to, pos, sphere_radius, &res)) { + float d = ray_from.distance_to(res); + if (d < closest_dist) { + closest_dist = d; + gi->sdfgi_debug_probe_enabled = true; + gi->sdfgi_debug_probe_index = probe_from + Vector3i(i, j, k); + } + } + } + } + } + + gi->sdfgi_debug_probe_dir = Vector3(); + } + + if (gi->sdfgi_debug_probe_enabled) { + uint32_t cascade = 0; + uint32_t probe_cells = (cascade_size / SDFGI::PROBE_DIVISOR); + Vector3i probe_from = cascades[cascade].position / probe_cells; + Vector3i ofs = gi->sdfgi_debug_probe_index - probe_from; + if (ofs.x < 0 || ofs.y < 0 || ofs.z < 0) { + return; + } + if (ofs.x > SDFGI::PROBE_DIVISOR || ofs.y > SDFGI::PROBE_DIVISOR || ofs.z > SDFGI::PROBE_DIVISOR) { + return; + } + + uint32_t mult = (SDFGI::PROBE_DIVISOR + 1); + uint32_t index = ofs.z * mult * mult + ofs.y * mult + ofs.x; + + push_constant.probe_debug_index = index; + + uint32_t cell_count = probe_cells * 2 * probe_cells * 2 * probe_cells * 2; + + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, gi->sdfgi_shader.debug_probes_pipeline[p_view_count > 1 ? SDFGIShader::PROBE_DEBUG_VISIBILITY_MULTIVIEW : SDFGIShader::PROBE_DEBUG_VISIBILITY].get_render_pipeline(RD::INVALID_FORMAT_ID, RD::get_singleton()->framebuffer_get_format(p_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, debug_probes_uniform_set, 0); + RD::get_singleton()->draw_list_set_push_constant(draw_list, &push_constant, sizeof(SDFGIShader::DebugProbesPushConstant)); + RD::get_singleton()->draw_list_draw(draw_list, false, cell_count, total_points); + } + + RD::get_singleton()->draw_command_end_label(); + RD::get_singleton()->draw_list_end(); +} + +void GI::SDFGI::pre_process_gi(const Transform3D &p_transform, RenderDataRD *p_render_data) { + RendererRD::LightStorage *light_storage = RendererRD::LightStorage::get_singleton(); + /* Update general SDFGI Buffer */ + + SDFGIData sdfgi_data; + + sdfgi_data.grid_size[0] = cascade_size; + sdfgi_data.grid_size[1] = cascade_size; + sdfgi_data.grid_size[2] = cascade_size; + + sdfgi_data.max_cascades = cascades.size(); + sdfgi_data.probe_axis_size = probe_axis_count; + sdfgi_data.cascade_probe_size[0] = sdfgi_data.probe_axis_size - 1; //float version for performance + sdfgi_data.cascade_probe_size[1] = sdfgi_data.probe_axis_size - 1; + sdfgi_data.cascade_probe_size[2] = sdfgi_data.probe_axis_size - 1; + + float csize = cascade_size; + sdfgi_data.probe_to_uvw = 1.0 / float(sdfgi_data.cascade_probe_size[0]); + sdfgi_data.use_occlusion = uses_occlusion; + //sdfgi_data.energy = energy; + + sdfgi_data.y_mult = y_mult; + + float cascade_voxel_size = (csize / sdfgi_data.cascade_probe_size[0]); + float occlusion_clamp = (cascade_voxel_size - 0.5) / cascade_voxel_size; + sdfgi_data.occlusion_clamp[0] = occlusion_clamp; + sdfgi_data.occlusion_clamp[1] = occlusion_clamp; + sdfgi_data.occlusion_clamp[2] = occlusion_clamp; + sdfgi_data.normal_bias = (normal_bias / csize) * sdfgi_data.cascade_probe_size[0]; + + //vec2 tex_pixel_size = 1.0 / vec2(ivec2( (OCT_SIZE+2) * params.probe_axis_size * params.probe_axis_size, (OCT_SIZE+2) * params.probe_axis_size ) ); + //vec3 probe_uv_offset = (ivec3(OCT_SIZE+2,OCT_SIZE+2,(OCT_SIZE+2) * params.probe_axis_size)) * tex_pixel_size.xyx; + + uint32_t oct_size = SDFGI::LIGHTPROBE_OCT_SIZE; + + sdfgi_data.lightprobe_tex_pixel_size[0] = 1.0 / ((oct_size + 2) * sdfgi_data.probe_axis_size * sdfgi_data.probe_axis_size); + sdfgi_data.lightprobe_tex_pixel_size[1] = 1.0 / ((oct_size + 2) * sdfgi_data.probe_axis_size); + sdfgi_data.lightprobe_tex_pixel_size[2] = 1.0; + + sdfgi_data.energy = energy; + + sdfgi_data.lightprobe_uv_offset[0] = float(oct_size + 2) * sdfgi_data.lightprobe_tex_pixel_size[0]; + sdfgi_data.lightprobe_uv_offset[1] = float(oct_size + 2) * sdfgi_data.lightprobe_tex_pixel_size[1]; + sdfgi_data.lightprobe_uv_offset[2] = float((oct_size + 2) * sdfgi_data.probe_axis_size) * sdfgi_data.lightprobe_tex_pixel_size[0]; + + sdfgi_data.occlusion_renormalize[0] = 0.5; + sdfgi_data.occlusion_renormalize[1] = 1.0; + sdfgi_data.occlusion_renormalize[2] = 1.0 / float(sdfgi_data.max_cascades); + + int32_t probe_divisor = cascade_size / SDFGI::PROBE_DIVISOR; + + for (uint32_t i = 0; i < sdfgi_data.max_cascades; i++) { + SDFGIData::ProbeCascadeData &c = sdfgi_data.cascades[i]; + Vector3 pos = Vector3((Vector3i(1, 1, 1) * -int32_t(cascade_size >> 1) + cascades[i].position)) * cascades[i].cell_size; + Vector3 cam_origin = p_transform.origin; + cam_origin.y *= y_mult; + pos -= cam_origin; //make pos local to camera, to reduce numerical error + c.position[0] = pos.x; + c.position[1] = pos.y; + c.position[2] = pos.z; + c.to_probe = 1.0 / (float(cascade_size) * cascades[i].cell_size / float(probe_axis_count - 1)); + + Vector3i probe_ofs = cascades[i].position / probe_divisor; + c.probe_world_offset[0] = probe_ofs.x; + c.probe_world_offset[1] = probe_ofs.y; + c.probe_world_offset[2] = probe_ofs.z; + + c.to_cell = 1.0 / cascades[i].cell_size; + c.exposure_normalization = 1.0; + if (p_render_data->camera_attributes.is_valid()) { + float exposure_normalization = RSG::camera_attributes->camera_attributes_get_exposure_normalization_factor(p_render_data->camera_attributes); + c.exposure_normalization = exposure_normalization / cascades[i].baked_exposure_normalization; + } + } + + RD::get_singleton()->buffer_update(gi->sdfgi_ubo, 0, sizeof(SDFGIData), &sdfgi_data, RD::BARRIER_MASK_COMPUTE); + + /* Update dynamic lights in SDFGI cascades */ + + for (uint32_t i = 0; i < cascades.size(); i++) { + SDFGI::Cascade &cascade = cascades[i]; + + SDFGIShader::Light lights[SDFGI::MAX_DYNAMIC_LIGHTS]; + uint32_t idx = 0; + for (uint32_t j = 0; j < (uint32_t)p_render_data->sdfgi_update_data->directional_lights->size(); j++) { + if (idx == SDFGI::MAX_DYNAMIC_LIGHTS) { + break; + } + + RID light_instance = p_render_data->sdfgi_update_data->directional_lights->get(j); + ERR_CONTINUE(!light_storage->owns_light_instance(light_instance)); + + RID light = light_storage->light_instance_get_base_light(light_instance); + Transform3D light_transform = light_storage->light_instance_get_base_transform(light_instance); + + if (RSG::light_storage->light_directional_get_sky_mode(light) == RS::LIGHT_DIRECTIONAL_SKY_MODE_SKY_ONLY) { + continue; + } + + Vector3 dir = -light_transform.basis.get_column(Vector3::AXIS_Z); + dir.y *= y_mult; + dir.normalize(); + lights[idx].direction[0] = dir.x; + lights[idx].direction[1] = dir.y; + lights[idx].direction[2] = dir.z; + Color color = RSG::light_storage->light_get_color(light); + color = color.srgb_to_linear(); + lights[idx].color[0] = color.r; + lights[idx].color[1] = color.g; + lights[idx].color[2] = color.b; + lights[idx].type = RS::LIGHT_DIRECTIONAL; + lights[idx].energy = RSG::light_storage->light_get_param(light, RS::LIGHT_PARAM_ENERGY) * RSG::light_storage->light_get_param(light, RS::LIGHT_PARAM_INDIRECT_ENERGY); + if (RendererSceneRenderRD::get_singleton()->is_using_physical_light_units()) { + lights[idx].energy *= RSG::light_storage->light_get_param(light, RS::LIGHT_PARAM_INTENSITY); + } + + if (p_render_data->camera_attributes.is_valid()) { + lights[idx].energy *= RSG::camera_attributes->camera_attributes_get_exposure_normalization_factor(p_render_data->camera_attributes); + } + + lights[idx].has_shadow = RSG::light_storage->light_has_shadow(light); + + idx++; + } + + AABB cascade_aabb; + cascade_aabb.position = Vector3((Vector3i(1, 1, 1) * -int32_t(cascade_size >> 1) + cascade.position)) * cascade.cell_size; + cascade_aabb.size = Vector3(1, 1, 1) * cascade_size * cascade.cell_size; + + for (uint32_t j = 0; j < p_render_data->sdfgi_update_data->positional_light_count; j++) { + if (idx == SDFGI::MAX_DYNAMIC_LIGHTS) { + break; + } + + RID light_instance = p_render_data->sdfgi_update_data->positional_light_instances[j]; + ERR_CONTINUE(!light_storage->owns_light_instance(light_instance)); + + RID light = light_storage->light_instance_get_base_light(light_instance); + AABB light_aabb = light_storage->light_instance_get_base_aabb(light_instance); + Transform3D light_transform = light_storage->light_instance_get_base_transform(light_instance); + + uint32_t max_sdfgi_cascade = RSG::light_storage->light_get_max_sdfgi_cascade(light); + if (i > max_sdfgi_cascade) { + continue; + } + + if (!cascade_aabb.intersects(light_aabb)) { + continue; + } + + Vector3 dir = -light_transform.basis.get_column(Vector3::AXIS_Z); + //faster to not do this here + //dir.y *= y_mult; + //dir.normalize(); + lights[idx].direction[0] = dir.x; + lights[idx].direction[1] = dir.y; + lights[idx].direction[2] = dir.z; + Vector3 pos = light_transform.origin; + pos.y *= y_mult; + lights[idx].position[0] = pos.x; + lights[idx].position[1] = pos.y; + lights[idx].position[2] = pos.z; + Color color = RSG::light_storage->light_get_color(light); + color = color.srgb_to_linear(); + lights[idx].color[0] = color.r; + lights[idx].color[1] = color.g; + lights[idx].color[2] = color.b; + lights[idx].type = RSG::light_storage->light_get_type(light); + + lights[idx].energy = RSG::light_storage->light_get_param(light, RS::LIGHT_PARAM_ENERGY) * RSG::light_storage->light_get_param(light, RS::LIGHT_PARAM_INDIRECT_ENERGY); + if (RendererSceneRenderRD::get_singleton()->is_using_physical_light_units()) { + lights[idx].energy *= RSG::light_storage->light_get_param(light, RS::LIGHT_PARAM_INTENSITY); + + // Convert from Luminous Power to Luminous Intensity + if (lights[idx].type == RS::LIGHT_OMNI) { + lights[idx].energy *= 1.0 / (Math_PI * 4.0); + } else if (lights[idx].type == RS::LIGHT_SPOT) { + // Spot Lights are not physically accurate, Luminous Intensity should change in relation to the cone angle. + // We make this assumption to keep them easy to control. + lights[idx].energy *= 1.0 / Math_PI; + } + } + + if (p_render_data->camera_attributes.is_valid()) { + lights[idx].energy *= RSG::camera_attributes->camera_attributes_get_exposure_normalization_factor(p_render_data->camera_attributes); + } + + lights[idx].has_shadow = RSG::light_storage->light_has_shadow(light); + lights[idx].attenuation = RSG::light_storage->light_get_param(light, RS::LIGHT_PARAM_ATTENUATION); + lights[idx].radius = RSG::light_storage->light_get_param(light, RS::LIGHT_PARAM_RANGE); + lights[idx].cos_spot_angle = Math::cos(Math::deg_to_rad(RSG::light_storage->light_get_param(light, RS::LIGHT_PARAM_SPOT_ANGLE))); + lights[idx].inv_spot_attenuation = 1.0f / RSG::light_storage->light_get_param(light, RS::LIGHT_PARAM_SPOT_ATTENUATION); + + idx++; + } + + if (idx > 0) { + RD::get_singleton()->buffer_update(cascade.lights_buffer, 0, idx * sizeof(SDFGIShader::Light), lights, RD::BARRIER_MASK_COMPUTE); + } + + cascade_dynamic_light_count[i] = idx; + } +} + +void GI::SDFGI::render_region(Ref<RenderSceneBuffersRD> p_render_buffers, int p_region, const PagedArray<RenderGeometryInstance *> &p_instances, float p_exposure_normalization) { + //print_line("rendering region " + itos(p_region)); + ERR_FAIL_COND(p_render_buffers.is_null()); // we wouldn't be here if this failed but... + AABB bounds; + Vector3i from; + Vector3i size; + + int cascade_prev = get_pending_region_data(p_region - 1, from, size, bounds); + int cascade_next = get_pending_region_data(p_region + 1, from, size, bounds); + int cascade = get_pending_region_data(p_region, from, size, bounds); + ERR_FAIL_COND(cascade < 0); + + if (cascade_prev != cascade) { + //initialize render + RD::get_singleton()->texture_clear(render_albedo, Color(0, 0, 0, 0), 0, 1, 0, 1); + RD::get_singleton()->texture_clear(render_emission, Color(0, 0, 0, 0), 0, 1, 0, 1); + RD::get_singleton()->texture_clear(render_emission_aniso, Color(0, 0, 0, 0), 0, 1, 0, 1); + RD::get_singleton()->texture_clear(render_geom_facing, Color(0, 0, 0, 0), 0, 1, 0, 1); + } + + //print_line("rendering cascade " + itos(p_region) + " objects: " + itos(p_cull_count) + " bounds: " + bounds + " from: " + from + " size: " + size + " cell size: " + rtos(cascades[cascade].cell_size)); + RendererSceneRenderRD::get_singleton()->_render_sdfgi(p_render_buffers, from, size, bounds, p_instances, render_albedo, render_emission, render_emission_aniso, render_geom_facing, p_exposure_normalization); + + if (cascade_next != cascade) { + RD::get_singleton()->draw_command_begin_label("SDFGI Pre-Process Cascade"); + + RENDER_TIMESTAMP("> SDFGI Update SDF"); + //done rendering! must update SDF + //clear dispatch indirect data + + SDFGIShader::PreprocessPushConstant push_constant; + memset(&push_constant, 0, sizeof(SDFGIShader::PreprocessPushConstant)); + + RENDER_TIMESTAMP("SDFGI Scroll SDF"); + + //scroll + if (cascades[cascade].dirty_regions != SDFGI::Cascade::DIRTY_ALL) { + //for scroll + Vector3i dirty = cascades[cascade].dirty_regions; + push_constant.scroll[0] = dirty.x; + push_constant.scroll[1] = dirty.y; + push_constant.scroll[2] = dirty.z; + } else { + //for no scroll + push_constant.scroll[0] = 0; + push_constant.scroll[1] = 0; + push_constant.scroll[2] = 0; + } + + cascades[cascade].all_dynamic_lights_dirty = true; + cascades[cascade].baked_exposure_normalization = p_exposure_normalization; + + push_constant.grid_size = cascade_size; + push_constant.cascade = cascade; + + if (cascades[cascade].dirty_regions != SDFGI::Cascade::DIRTY_ALL) { + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + + //must pre scroll existing data because not all is dirty + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->sdfgi_shader.preprocess_pipeline[SDFGIShader::PRE_PROCESS_SCROLL]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, cascades[cascade].scroll_uniform_set, 0); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDFGIShader::PreprocessPushConstant)); + RD::get_singleton()->compute_list_dispatch_indirect(compute_list, cascades[cascade].solid_cell_dispatch_buffer, 0); + // no barrier do all together + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->sdfgi_shader.preprocess_pipeline[SDFGIShader::PRE_PROCESS_SCROLL_OCCLUSION]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, cascades[cascade].scroll_occlusion_uniform_set, 0); + + Vector3i dirty = cascades[cascade].dirty_regions; + Vector3i groups; + groups.x = cascade_size - ABS(dirty.x); + groups.y = cascade_size - ABS(dirty.y); + groups.z = cascade_size - ABS(dirty.z); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDFGIShader::PreprocessPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, groups.x, groups.y, groups.z); + + //no barrier, continue together + + { + //scroll probes and their history also + + SDFGIShader::IntegratePushConstant ipush_constant; + ipush_constant.grid_size[1] = cascade_size; + ipush_constant.grid_size[2] = cascade_size; + ipush_constant.grid_size[0] = cascade_size; + ipush_constant.max_cascades = cascades.size(); + ipush_constant.probe_axis_size = probe_axis_count; + ipush_constant.history_index = 0; + ipush_constant.history_size = history_size; + ipush_constant.ray_count = 0; + ipush_constant.ray_bias = 0; + ipush_constant.sky_mode = 0; + ipush_constant.sky_energy = 0; + ipush_constant.sky_color[0] = 0; + ipush_constant.sky_color[1] = 0; + ipush_constant.sky_color[2] = 0; + ipush_constant.y_mult = y_mult; + ipush_constant.store_ambient_texture = false; + + ipush_constant.image_size[0] = probe_axis_count * probe_axis_count; + ipush_constant.image_size[1] = probe_axis_count; + + int32_t probe_divisor = cascade_size / SDFGI::PROBE_DIVISOR; + ipush_constant.cascade = cascade; + ipush_constant.world_offset[0] = cascades[cascade].position.x / probe_divisor; + ipush_constant.world_offset[1] = cascades[cascade].position.y / probe_divisor; + ipush_constant.world_offset[2] = cascades[cascade].position.z / probe_divisor; + + ipush_constant.scroll[0] = dirty.x / probe_divisor; + ipush_constant.scroll[1] = dirty.y / probe_divisor; + ipush_constant.scroll[2] = dirty.z / probe_divisor; + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->sdfgi_shader.integrate_pipeline[SDFGIShader::INTEGRATE_MODE_SCROLL]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, cascades[cascade].integrate_uniform_set, 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, gi->sdfgi_shader.integrate_default_sky_uniform_set, 1); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ipush_constant, sizeof(SDFGIShader::IntegratePushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, probe_axis_count * probe_axis_count, probe_axis_count, 1); + + RD::get_singleton()->compute_list_add_barrier(compute_list); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->sdfgi_shader.integrate_pipeline[SDFGIShader::INTEGRATE_MODE_SCROLL_STORE]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, cascades[cascade].integrate_uniform_set, 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, gi->sdfgi_shader.integrate_default_sky_uniform_set, 1); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ipush_constant, sizeof(SDFGIShader::IntegratePushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, probe_axis_count * probe_axis_count, probe_axis_count, 1); + + RD::get_singleton()->compute_list_add_barrier(compute_list); + + if (bounce_feedback > 0.0) { + //multibounce requires this to be stored so direct light can read from it + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->sdfgi_shader.integrate_pipeline[SDFGIShader::INTEGRATE_MODE_STORE]); + + //convert to octahedral to store + ipush_constant.image_size[0] *= SDFGI::LIGHTPROBE_OCT_SIZE; + ipush_constant.image_size[1] *= SDFGI::LIGHTPROBE_OCT_SIZE; + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, cascades[cascade].integrate_uniform_set, 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, gi->sdfgi_shader.integrate_default_sky_uniform_set, 1); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ipush_constant, sizeof(SDFGIShader::IntegratePushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, probe_axis_count * probe_axis_count * SDFGI::LIGHTPROBE_OCT_SIZE, probe_axis_count * SDFGI::LIGHTPROBE_OCT_SIZE, 1); + } + } + + //ok finally barrier + RD::get_singleton()->compute_list_end(); + } + + //clear dispatch indirect data + uint32_t dispatch_indirct_data[4] = { 0, 0, 0, 0 }; + RD::get_singleton()->buffer_update(cascades[cascade].solid_cell_dispatch_buffer, 0, sizeof(uint32_t) * 4, dispatch_indirct_data); + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + + bool half_size = true; //much faster, very little difference + static const int optimized_jf_group_size = 8; + + if (half_size) { + push_constant.grid_size >>= 1; + + uint32_t cascade_half_size = cascade_size >> 1; + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->sdfgi_shader.preprocess_pipeline[SDFGIShader::PRE_PROCESS_JUMP_FLOOD_INITIALIZE_HALF]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, sdf_initialize_half_uniform_set, 0); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDFGIShader::PreprocessPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, cascade_half_size, cascade_half_size, cascade_half_size); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + //must start with regular jumpflood + + push_constant.half_size = true; + { + RENDER_TIMESTAMP("SDFGI Jump Flood (Half-Size)"); + + uint32_t s = cascade_half_size; + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->sdfgi_shader.preprocess_pipeline[SDFGIShader::PRE_PROCESS_JUMP_FLOOD]); + + int jf_us = 0; + //start with regular jump flood for very coarse reads, as this is impossible to optimize + while (s > 1) { + s /= 2; + push_constant.step_size = s; + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, jump_flood_half_uniform_set[jf_us], 0); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDFGIShader::PreprocessPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, cascade_half_size, cascade_half_size, cascade_half_size); + RD::get_singleton()->compute_list_add_barrier(compute_list); + jf_us = jf_us == 0 ? 1 : 0; + + if (cascade_half_size / (s / 2) >= optimized_jf_group_size) { + break; + } + } + + RENDER_TIMESTAMP("SDFGI Jump Flood Optimized (Half-Size)"); + + //continue with optimized jump flood for smaller reads + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->sdfgi_shader.preprocess_pipeline[SDFGIShader::PRE_PROCESS_JUMP_FLOOD_OPTIMIZED]); + while (s > 1) { + s /= 2; + push_constant.step_size = s; + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, jump_flood_half_uniform_set[jf_us], 0); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDFGIShader::PreprocessPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, cascade_half_size, cascade_half_size, cascade_half_size); + RD::get_singleton()->compute_list_add_barrier(compute_list); + jf_us = jf_us == 0 ? 1 : 0; + } + } + + // restore grid size for last passes + push_constant.grid_size = cascade_size; + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->sdfgi_shader.preprocess_pipeline[SDFGIShader::PRE_PROCESS_JUMP_FLOOD_UPSCALE]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, sdf_upscale_uniform_set, 0); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDFGIShader::PreprocessPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, cascade_size, cascade_size, cascade_size); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + //run one pass of fullsize jumpflood to fix up half size arctifacts + + push_constant.half_size = false; + push_constant.step_size = 1; + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->sdfgi_shader.preprocess_pipeline[SDFGIShader::PRE_PROCESS_JUMP_FLOOD_OPTIMIZED]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, jump_flood_uniform_set[upscale_jfa_uniform_set_index], 0); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDFGIShader::PreprocessPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, cascade_size, cascade_size, cascade_size); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + } else { + //full size jumpflood + RENDER_TIMESTAMP("SDFGI Jump Flood (Full-Size)"); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->sdfgi_shader.preprocess_pipeline[SDFGIShader::PRE_PROCESS_JUMP_FLOOD_INITIALIZE]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, sdf_initialize_uniform_set, 0); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDFGIShader::PreprocessPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, cascade_size, cascade_size, cascade_size); + + RD::get_singleton()->compute_list_add_barrier(compute_list); + + push_constant.half_size = false; + { + uint32_t s = cascade_size; + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->sdfgi_shader.preprocess_pipeline[SDFGIShader::PRE_PROCESS_JUMP_FLOOD]); + + int jf_us = 0; + //start with regular jump flood for very coarse reads, as this is impossible to optimize + while (s > 1) { + s /= 2; + push_constant.step_size = s; + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, jump_flood_uniform_set[jf_us], 0); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDFGIShader::PreprocessPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, cascade_size, cascade_size, cascade_size); + RD::get_singleton()->compute_list_add_barrier(compute_list); + jf_us = jf_us == 0 ? 1 : 0; + + if (cascade_size / (s / 2) >= optimized_jf_group_size) { + break; + } + } + + RENDER_TIMESTAMP("SDFGI Jump Flood Optimized (Full-Size)"); + + //continue with optimized jump flood for smaller reads + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->sdfgi_shader.preprocess_pipeline[SDFGIShader::PRE_PROCESS_JUMP_FLOOD_OPTIMIZED]); + while (s > 1) { + s /= 2; + push_constant.step_size = s; + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, jump_flood_uniform_set[jf_us], 0); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDFGIShader::PreprocessPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, cascade_size, cascade_size, cascade_size); + RD::get_singleton()->compute_list_add_barrier(compute_list); + jf_us = jf_us == 0 ? 1 : 0; + } + } + } + + RENDER_TIMESTAMP("SDFGI Occlusion"); + + // occlusion + { + uint32_t probe_size = cascade_size / SDFGI::PROBE_DIVISOR; + Vector3i probe_global_pos = cascades[cascade].position / probe_size; + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->sdfgi_shader.preprocess_pipeline[SDFGIShader::PRE_PROCESS_OCCLUSION]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, occlusion_uniform_set, 0); + for (int i = 0; i < 8; i++) { + //dispatch all at once for performance + Vector3i offset(i & 1, (i >> 1) & 1, (i >> 2) & 1); + + if ((probe_global_pos.x & 1) != 0) { + offset.x = (offset.x + 1) & 1; + } + if ((probe_global_pos.y & 1) != 0) { + offset.y = (offset.y + 1) & 1; + } + if ((probe_global_pos.z & 1) != 0) { + offset.z = (offset.z + 1) & 1; + } + push_constant.probe_offset[0] = offset.x; + push_constant.probe_offset[1] = offset.y; + push_constant.probe_offset[2] = offset.z; + push_constant.occlusion_index = i; + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDFGIShader::PreprocessPushConstant)); + + Vector3i groups = Vector3i(probe_size + 1, probe_size + 1, probe_size + 1) - offset; //if offset, it's one less probe per axis to compute + RD::get_singleton()->compute_list_dispatch(compute_list, groups.x, groups.y, groups.z); + } + RD::get_singleton()->compute_list_add_barrier(compute_list); + } + + RENDER_TIMESTAMP("SDFGI Store"); + + // store + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->sdfgi_shader.preprocess_pipeline[SDFGIShader::PRE_PROCESS_STORE]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, cascades[cascade].sdf_store_uniform_set, 0); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDFGIShader::PreprocessPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, cascade_size, cascade_size, cascade_size); + + RD::get_singleton()->compute_list_end(); + + //clear these textures, as they will have previous garbage on next draw + RD::get_singleton()->texture_clear(cascades[cascade].light_tex, Color(0, 0, 0, 0), 0, 1, 0, 1); + RD::get_singleton()->texture_clear(cascades[cascade].light_aniso_0_tex, Color(0, 0, 0, 0), 0, 1, 0, 1); + RD::get_singleton()->texture_clear(cascades[cascade].light_aniso_1_tex, Color(0, 0, 0, 0), 0, 1, 0, 1); + +#if 0 + Vector<uint8_t> data = RD::get_singleton()->texture_get_data(cascades[cascade].sdf, 0); + Ref<Image> img; + img.instantiate(); + for (uint32_t i = 0; i < cascade_size; i++) { + Vector<uint8_t> subarr = data.slice(128 * 128 * i, 128 * 128 * (i + 1)); + img->set_data(cascade_size, cascade_size, false, Image::FORMAT_L8, subarr); + img->save_png("res://cascade_sdf_" + itos(cascade) + "_" + itos(i) + ".png"); + } + + //finalize render and update sdf +#endif + +#if 0 + Vector<uint8_t> data = RD::get_singleton()->texture_get_data(render_albedo, 0); + Ref<Image> img; + img.instantiate(); + for (uint32_t i = 0; i < cascade_size; i++) { + Vector<uint8_t> subarr = data.slice(128 * 128 * i * 2, 128 * 128 * (i + 1) * 2); + img->createcascade_size, cascade_size, false, Image::FORMAT_RGB565, subarr); + img->convert(Image::FORMAT_RGBA8); + img->save_png("res://cascade_" + itos(cascade) + "_" + itos(i) + ".png"); + } + + //finalize render and update sdf +#endif + + RENDER_TIMESTAMP("< SDFGI Update SDF"); + RD::get_singleton()->draw_command_end_label(); + } +} + +void GI::SDFGI::render_static_lights(RenderDataRD *p_render_data, Ref<RenderSceneBuffersRD> p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const PagedArray<RID> *p_positional_light_cull_result) { + ERR_FAIL_COND(p_render_buffers.is_null()); // we wouldn't be here if this failed but... + + RendererRD::LightStorage *light_storage = RendererRD::LightStorage::get_singleton(); + + RD::get_singleton()->draw_command_begin_label("SDFGI Render Static Lights"); + + update_cascades(); + + SDFGIShader::Light lights[SDFGI::MAX_STATIC_LIGHTS]; + uint32_t light_count[SDFGI::MAX_STATIC_LIGHTS]; + + for (uint32_t i = 0; i < p_cascade_count; i++) { + ERR_CONTINUE(p_cascade_indices[i] >= cascades.size()); + + SDFGI::Cascade &cc = cascades[p_cascade_indices[i]]; + + { //fill light buffer + + AABB cascade_aabb; + cascade_aabb.position = Vector3((Vector3i(1, 1, 1) * -int32_t(cascade_size >> 1) + cc.position)) * cc.cell_size; + cascade_aabb.size = Vector3(1, 1, 1) * cascade_size * cc.cell_size; + + int idx = 0; + + for (uint32_t j = 0; j < (uint32_t)p_positional_light_cull_result[i].size(); j++) { + if (idx == SDFGI::MAX_STATIC_LIGHTS) { + break; + } + + RID light_instance = p_positional_light_cull_result[i][j]; + ERR_CONTINUE(!light_storage->owns_light_instance(light_instance)); + + RID light = light_storage->light_instance_get_base_light(light_instance); + AABB light_aabb = light_storage->light_instance_get_base_aabb(light_instance); + Transform3D light_transform = light_storage->light_instance_get_base_transform(light_instance); + + uint32_t max_sdfgi_cascade = RSG::light_storage->light_get_max_sdfgi_cascade(light); + if (p_cascade_indices[i] > max_sdfgi_cascade) { + continue; + } + + if (!cascade_aabb.intersects(light_aabb)) { + continue; + } + + lights[idx].type = RSG::light_storage->light_get_type(light); + + Vector3 dir = -light_transform.basis.get_column(Vector3::AXIS_Z); + if (lights[idx].type == RS::LIGHT_DIRECTIONAL) { + dir.y *= y_mult; //only makes sense for directional + dir.normalize(); + } + lights[idx].direction[0] = dir.x; + lights[idx].direction[1] = dir.y; + lights[idx].direction[2] = dir.z; + Vector3 pos = light_transform.origin; + pos.y *= y_mult; + lights[idx].position[0] = pos.x; + lights[idx].position[1] = pos.y; + lights[idx].position[2] = pos.z; + Color color = RSG::light_storage->light_get_color(light); + color = color.srgb_to_linear(); + lights[idx].color[0] = color.r; + lights[idx].color[1] = color.g; + lights[idx].color[2] = color.b; + + lights[idx].energy = RSG::light_storage->light_get_param(light, RS::LIGHT_PARAM_ENERGY) * RSG::light_storage->light_get_param(light, RS::LIGHT_PARAM_INDIRECT_ENERGY); + if (RendererSceneRenderRD::get_singleton()->is_using_physical_light_units()) { + lights[idx].energy *= RSG::light_storage->light_get_param(light, RS::LIGHT_PARAM_INTENSITY); + + // Convert from Luminous Power to Luminous Intensity + if (lights[idx].type == RS::LIGHT_OMNI) { + lights[idx].energy *= 1.0 / (Math_PI * 4.0); + } else if (lights[idx].type == RS::LIGHT_SPOT) { + // Spot Lights are not physically accurate, Luminous Intensity should change in relation to the cone angle. + // We make this assumption to keep them easy to control. + lights[idx].energy *= 1.0 / Math_PI; + } + } + + if (p_render_data->camera_attributes.is_valid()) { + lights[idx].energy *= RSG::camera_attributes->camera_attributes_get_exposure_normalization_factor(p_render_data->camera_attributes); + } + + lights[idx].has_shadow = RSG::light_storage->light_has_shadow(light); + lights[idx].attenuation = RSG::light_storage->light_get_param(light, RS::LIGHT_PARAM_ATTENUATION); + lights[idx].radius = RSG::light_storage->light_get_param(light, RS::LIGHT_PARAM_RANGE); + lights[idx].cos_spot_angle = Math::cos(Math::deg_to_rad(RSG::light_storage->light_get_param(light, RS::LIGHT_PARAM_SPOT_ANGLE))); + lights[idx].inv_spot_attenuation = 1.0f / RSG::light_storage->light_get_param(light, RS::LIGHT_PARAM_SPOT_ATTENUATION); + + idx++; + } + + if (idx > 0) { + RD::get_singleton()->buffer_update(cc.lights_buffer, 0, idx * sizeof(SDFGIShader::Light), lights); + } + + light_count[i] = idx; + } + } + + /* Static Lights */ + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->sdfgi_shader.direct_light_pipeline[SDFGIShader::DIRECT_LIGHT_MODE_STATIC]); + + SDFGIShader::DirectLightPushConstant dl_push_constant; + + dl_push_constant.grid_size[0] = cascade_size; + dl_push_constant.grid_size[1] = cascade_size; + dl_push_constant.grid_size[2] = cascade_size; + dl_push_constant.max_cascades = cascades.size(); + dl_push_constant.probe_axis_size = probe_axis_count; + dl_push_constant.bounce_feedback = 0.0; // this is static light, do not multibounce yet + dl_push_constant.y_mult = y_mult; + dl_push_constant.use_occlusion = uses_occlusion; + + //all must be processed + dl_push_constant.process_offset = 0; + dl_push_constant.process_increment = 1; + + for (uint32_t i = 0; i < p_cascade_count; i++) { + ERR_CONTINUE(p_cascade_indices[i] >= cascades.size()); + + SDFGI::Cascade &cc = cascades[p_cascade_indices[i]]; + + dl_push_constant.light_count = light_count[i]; + dl_push_constant.cascade = p_cascade_indices[i]; + + if (dl_push_constant.light_count > 0) { + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, cc.sdf_direct_light_static_uniform_set, 0); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &dl_push_constant, sizeof(SDFGIShader::DirectLightPushConstant)); + RD::get_singleton()->compute_list_dispatch_indirect(compute_list, cc.solid_cell_dispatch_buffer, 0); + } + } + + RD::get_singleton()->compute_list_end(); + + RD::get_singleton()->draw_command_end_label(); +} + +//////////////////////////////////////////////////////////////////////////////// +// VoxelGIInstance + +void GI::VoxelGIInstance::update(bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RenderGeometryInstance *> &p_dynamic_objects) { + RendererRD::LightStorage *light_storage = RendererRD::LightStorage::get_singleton(); + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + + uint32_t data_version = gi->voxel_gi_get_data_version(probe); + + // (RE)CREATE IF NEEDED + + if (last_probe_data_version != data_version) { + //need to re-create everything + free_resources(); + + Vector3i octree_size = gi->voxel_gi_get_octree_size(probe); + + if (octree_size != Vector3i()) { + //can create a 3D texture + Vector<int> levels = gi->voxel_gi_get_level_counts(probe); + + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + tf.width = octree_size.x; + tf.height = octree_size.y; + tf.depth = octree_size.z; + tf.texture_type = RD::TEXTURE_TYPE_3D; + tf.mipmaps = levels.size(); + + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; + + texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(texture, "VoxelGI Instance Texture"); + + RD::get_singleton()->texture_clear(texture, Color(0, 0, 0, 0), 0, levels.size(), 0, 1); + + { + int total_elements = 0; + for (int i = 0; i < levels.size(); i++) { + total_elements += levels[i]; + } + + write_buffer = RD::get_singleton()->storage_buffer_create(total_elements * 16); + } + + for (int i = 0; i < levels.size(); i++) { + VoxelGIInstance::Mipmap mipmap; + mipmap.texture = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), texture, 0, i, 1, RD::TEXTURE_SLICE_3D); + mipmap.level = levels.size() - i - 1; + mipmap.cell_offset = 0; + for (uint32_t j = 0; j < mipmap.level; j++) { + mipmap.cell_offset += levels[j]; + } + mipmap.cell_count = levels[mipmap.level]; + + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 1; + u.append_id(gi->voxel_gi_get_octree_buffer(probe)); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 2; + u.append_id(gi->voxel_gi_get_data_buffer(probe)); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 4; + u.append_id(write_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 9; + u.append_id(gi->voxel_gi_get_sdf_texture(probe)); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 10; + u.append_id(material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED)); + uniforms.push_back(u); + } + + { + Vector<RD::Uniform> copy_uniforms = uniforms; + if (i == 0) { + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 3; + u.append_id(gi->voxel_gi_lights_uniform); + copy_uniforms.push_back(u); + } + + mipmap.uniform_set = RD::get_singleton()->uniform_set_create(copy_uniforms, gi->voxel_gi_lighting_shader_version_shaders[VOXEL_GI_SHADER_VERSION_COMPUTE_LIGHT], 0); + + copy_uniforms = uniforms; //restore + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 5; + u.append_id(texture); + copy_uniforms.push_back(u); + } + mipmap.second_bounce_uniform_set = RD::get_singleton()->uniform_set_create(copy_uniforms, gi->voxel_gi_lighting_shader_version_shaders[VOXEL_GI_SHADER_VERSION_COMPUTE_SECOND_BOUNCE], 0); + } else { + mipmap.uniform_set = RD::get_singleton()->uniform_set_create(copy_uniforms, gi->voxel_gi_lighting_shader_version_shaders[VOXEL_GI_SHADER_VERSION_COMPUTE_MIPMAP], 0); + } + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 5; + u.append_id(mipmap.texture); + uniforms.push_back(u); + } + + mipmap.write_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->voxel_gi_lighting_shader_version_shaders[VOXEL_GI_SHADER_VERSION_WRITE_TEXTURE], 0); + + mipmaps.push_back(mipmap); + } + + { + uint32_t dynamic_map_size = MAX(MAX(octree_size.x, octree_size.y), octree_size.z); + uint32_t oversample = nearest_power_of_2_templated(4); + int mipmap_index = 0; + + while (mipmap_index < mipmaps.size()) { + VoxelGIInstance::DynamicMap dmap; + + if (oversample > 0) { + dmap.size = dynamic_map_size * (1 << oversample); + dmap.mipmap = -1; + oversample--; + } else { + dmap.size = dynamic_map_size >> mipmap_index; + dmap.mipmap = mipmap_index; + mipmap_index++; + } + + RD::TextureFormat dtf; + dtf.width = dmap.size; + dtf.height = dmap.size; + dtf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + dtf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT; + + if (dynamic_maps.size() == 0) { + dtf.usage_bits |= RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + } + dmap.texture = RD::get_singleton()->texture_create(dtf, RD::TextureView()); + RD::get_singleton()->set_resource_name(dmap.texture, "VoxelGI Instance DMap Texture"); + + if (dynamic_maps.size() == 0) { + // Render depth for first one. + // Use 16-bit depth when supported to improve performance. + dtf.format = RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_D16_UNORM, RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) ? RD::DATA_FORMAT_D16_UNORM : RD::DATA_FORMAT_X8_D24_UNORM_PACK32; + dtf.usage_bits = RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT; + dmap.fb_depth = RD::get_singleton()->texture_create(dtf, RD::TextureView()); + RD::get_singleton()->set_resource_name(dmap.fb_depth, "VoxelGI Instance DMap FB Depth"); + } + + //just use depth as-is + dtf.format = RD::DATA_FORMAT_R32_SFLOAT; + dtf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + + dmap.depth = RD::get_singleton()->texture_create(dtf, RD::TextureView()); + RD::get_singleton()->set_resource_name(dmap.depth, "VoxelGI Instance DMap Depth"); + + if (dynamic_maps.size() == 0) { + dtf.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + dtf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + dmap.albedo = RD::get_singleton()->texture_create(dtf, RD::TextureView()); + RD::get_singleton()->set_resource_name(dmap.albedo, "VoxelGI Instance DMap Albedo"); + dmap.normal = RD::get_singleton()->texture_create(dtf, RD::TextureView()); + RD::get_singleton()->set_resource_name(dmap.normal, "VoxelGI Instance DMap Normal"); + dmap.orm = RD::get_singleton()->texture_create(dtf, RD::TextureView()); + RD::get_singleton()->set_resource_name(dmap.orm, "VoxelGI Instance DMap ORM"); + + Vector<RID> fb; + fb.push_back(dmap.albedo); + fb.push_back(dmap.normal); + fb.push_back(dmap.orm); + fb.push_back(dmap.texture); //emission + fb.push_back(dmap.depth); + fb.push_back(dmap.fb_depth); + + dmap.fb = RD::get_singleton()->framebuffer_create(fb); + + { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 3; + u.append_id(gi->voxel_gi_lights_uniform); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 5; + u.append_id(dmap.albedo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 6; + u.append_id(dmap.normal); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 7; + u.append_id(dmap.orm); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 8; + u.append_id(dmap.fb_depth); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 9; + u.append_id(gi->voxel_gi_get_sdf_texture(probe)); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 10; + u.append_id(material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED)); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 11; + u.append_id(dmap.texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 12; + u.append_id(dmap.depth); + uniforms.push_back(u); + } + + dmap.uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->voxel_gi_lighting_shader_version_shaders[VOXEL_GI_SHADER_VERSION_DYNAMIC_OBJECT_LIGHTING], 0); + } + } else { + bool plot = dmap.mipmap >= 0; + bool write = dmap.mipmap < (mipmaps.size() - 1); + + Vector<RD::Uniform> uniforms; + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 5; + u.append_id(dynamic_maps[dynamic_maps.size() - 1].texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 6; + u.append_id(dynamic_maps[dynamic_maps.size() - 1].depth); + uniforms.push_back(u); + } + + if (write) { + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 7; + u.append_id(dmap.texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 8; + u.append_id(dmap.depth); + uniforms.push_back(u); + } + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 9; + u.append_id(gi->voxel_gi_get_sdf_texture(probe)); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 10; + u.append_id(material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED)); + uniforms.push_back(u); + } + + if (plot) { + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 11; + u.append_id(mipmaps[dmap.mipmap].texture); + uniforms.push_back(u); + } + } + + dmap.uniform_set = RD::get_singleton()->uniform_set_create( + uniforms, + gi->voxel_gi_lighting_shader_version_shaders[(write && plot) ? VOXEL_GI_SHADER_VERSION_DYNAMIC_SHRINK_WRITE_PLOT : (write ? VOXEL_GI_SHADER_VERSION_DYNAMIC_SHRINK_WRITE : VOXEL_GI_SHADER_VERSION_DYNAMIC_SHRINK_PLOT)], + 0); + } + + dynamic_maps.push_back(dmap); + } + } + } + + last_probe_data_version = data_version; + p_update_light_instances = true; //just in case + + RendererSceneRenderRD::get_singleton()->base_uniforms_changed(); + } + + // UDPDATE TIME + + if (has_dynamic_object_data) { + //if it has dynamic object data, it needs to be cleared + RD::get_singleton()->texture_clear(texture, Color(0, 0, 0, 0), 0, mipmaps.size(), 0, 1); + } + + uint32_t light_count = 0; + + if (p_update_light_instances || p_dynamic_objects.size() > 0) { + light_count = MIN(gi->voxel_gi_max_lights, (uint32_t)p_light_instances.size()); + + { + Transform3D to_cell = gi->voxel_gi_get_to_cell_xform(probe); + Transform3D to_probe_xform = to_cell * transform.affine_inverse(); + + //update lights + + for (uint32_t i = 0; i < light_count; i++) { + VoxelGILight &l = gi->voxel_gi_lights[i]; + RID light_instance = p_light_instances[i]; + RID light = light_storage->light_instance_get_base_light(light_instance); + + l.type = RSG::light_storage->light_get_type(light); + if (l.type == RS::LIGHT_DIRECTIONAL && RSG::light_storage->light_directional_get_sky_mode(light) == RS::LIGHT_DIRECTIONAL_SKY_MODE_SKY_ONLY) { + light_count--; + continue; + } + + l.attenuation = RSG::light_storage->light_get_param(light, RS::LIGHT_PARAM_ATTENUATION); + l.energy = RSG::light_storage->light_get_param(light, RS::LIGHT_PARAM_ENERGY) * RSG::light_storage->light_get_param(light, RS::LIGHT_PARAM_INDIRECT_ENERGY); + + if (RendererSceneRenderRD::get_singleton()->is_using_physical_light_units()) { + l.energy *= RSG::light_storage->light_get_param(light, RS::LIGHT_PARAM_INTENSITY); + + l.energy *= gi->voxel_gi_get_baked_exposure_normalization(probe); + + // Convert from Luminous Power to Luminous Intensity + if (l.type == RS::LIGHT_OMNI) { + l.energy *= 1.0 / (Math_PI * 4.0); + } else if (l.type == RS::LIGHT_SPOT) { + // Spot Lights are not physically accurate, Luminous Intensity should change in relation to the cone angle. + // We make this assumption to keep them easy to control. + l.energy *= 1.0 / Math_PI; + } + } + + l.radius = to_cell.basis.xform(Vector3(RSG::light_storage->light_get_param(light, RS::LIGHT_PARAM_RANGE), 0, 0)).length(); + Color color = RSG::light_storage->light_get_color(light).srgb_to_linear(); + l.color[0] = color.r; + l.color[1] = color.g; + l.color[2] = color.b; + + l.cos_spot_angle = Math::cos(Math::deg_to_rad(RSG::light_storage->light_get_param(light, RS::LIGHT_PARAM_SPOT_ANGLE))); + l.inv_spot_attenuation = 1.0f / RSG::light_storage->light_get_param(light, RS::LIGHT_PARAM_SPOT_ATTENUATION); + + Transform3D xform = light_storage->light_instance_get_base_transform(light_instance); + + Vector3 pos = to_probe_xform.xform(xform.origin); + Vector3 dir = to_probe_xform.basis.xform(-xform.basis.get_column(2)).normalized(); + + l.position[0] = pos.x; + l.position[1] = pos.y; + l.position[2] = pos.z; + + l.direction[0] = dir.x; + l.direction[1] = dir.y; + l.direction[2] = dir.z; + + l.has_shadow = RSG::light_storage->light_has_shadow(light); + } + + RD::get_singleton()->buffer_update(gi->voxel_gi_lights_uniform, 0, sizeof(VoxelGILight) * light_count, gi->voxel_gi_lights); + } + } + + if (has_dynamic_object_data || p_update_light_instances || p_dynamic_objects.size()) { + // PROCESS MIPMAPS + if (mipmaps.size()) { + //can update mipmaps + + Vector3i probe_size = gi->voxel_gi_get_octree_size(probe); + + VoxelGIPushConstant push_constant; + + push_constant.limits[0] = probe_size.x; + push_constant.limits[1] = probe_size.y; + push_constant.limits[2] = probe_size.z; + push_constant.stack_size = mipmaps.size(); + push_constant.emission_scale = 1.0; + push_constant.propagation = gi->voxel_gi_get_propagation(probe); + push_constant.dynamic_range = gi->voxel_gi_get_dynamic_range(probe); + push_constant.light_count = light_count; + push_constant.aniso_strength = 0; + + /* print_line("probe update to version " + itos(last_probe_version)); + print_line("propagation " + rtos(push_constant.propagation)); + print_line("dynrange " + rtos(push_constant.dynamic_range)); + */ + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + + int passes; + if (p_update_light_instances) { + passes = gi->voxel_gi_is_using_two_bounces(probe) ? 2 : 1; + } else { + passes = 1; //only re-blitting is necessary + } + int wg_size = 64; + int64_t wg_limit_x = (int64_t)RD::get_singleton()->limit_get(RD::LIMIT_MAX_COMPUTE_WORKGROUP_COUNT_X); + + for (int pass = 0; pass < passes; pass++) { + if (p_update_light_instances) { + for (int i = 0; i < mipmaps.size(); i++) { + if (i == 0) { + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->voxel_gi_lighting_shader_version_pipelines[pass == 0 ? VOXEL_GI_SHADER_VERSION_COMPUTE_LIGHT : VOXEL_GI_SHADER_VERSION_COMPUTE_SECOND_BOUNCE]); + } else if (i == 1) { + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->voxel_gi_lighting_shader_version_pipelines[VOXEL_GI_SHADER_VERSION_COMPUTE_MIPMAP]); + } + + if (pass == 1 || i > 0) { + RD::get_singleton()->compute_list_add_barrier(compute_list); //wait til previous step is done + } + if (pass == 0 || i > 0) { + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, mipmaps[i].uniform_set, 0); + } else { + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, mipmaps[i].second_bounce_uniform_set, 0); + } + + push_constant.cell_offset = mipmaps[i].cell_offset; + push_constant.cell_count = mipmaps[i].cell_count; + + int64_t wg_todo = (mipmaps[i].cell_count - 1) / wg_size + 1; + while (wg_todo) { + int64_t wg_count = MIN(wg_todo, wg_limit_x); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(VoxelGIPushConstant)); + RD::get_singleton()->compute_list_dispatch(compute_list, wg_count, 1, 1); + wg_todo -= wg_count; + push_constant.cell_offset += wg_count * wg_size; + } + } + + RD::get_singleton()->compute_list_add_barrier(compute_list); //wait til previous step is done + } + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->voxel_gi_lighting_shader_version_pipelines[VOXEL_GI_SHADER_VERSION_WRITE_TEXTURE]); + + for (int i = 0; i < mipmaps.size(); i++) { + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, mipmaps[i].write_uniform_set, 0); + + push_constant.cell_offset = mipmaps[i].cell_offset; + push_constant.cell_count = mipmaps[i].cell_count; + + int64_t wg_todo = (mipmaps[i].cell_count - 1) / wg_size + 1; + while (wg_todo) { + int64_t wg_count = MIN(wg_todo, wg_limit_x); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(VoxelGIPushConstant)); + RD::get_singleton()->compute_list_dispatch(compute_list, wg_count, 1, 1); + wg_todo -= wg_count; + push_constant.cell_offset += wg_count * wg_size; + } + } + } + + RD::get_singleton()->compute_list_end(); + } + } + + has_dynamic_object_data = false; //clear until dynamic object data is used again + + if (p_dynamic_objects.size() && dynamic_maps.size()) { + Vector3i octree_size = gi->voxel_gi_get_octree_size(probe); + int multiplier = dynamic_maps[0].size / MAX(MAX(octree_size.x, octree_size.y), octree_size.z); + + Transform3D oversample_scale; + oversample_scale.basis.scale(Vector3(multiplier, multiplier, multiplier)); + + Transform3D to_cell = oversample_scale * gi->voxel_gi_get_to_cell_xform(probe); + Transform3D to_world_xform = transform * to_cell.affine_inverse(); + Transform3D to_probe_xform = to_world_xform.affine_inverse(); + + AABB probe_aabb(Vector3(), octree_size); + + //this could probably be better parallelized in compute.. + for (int i = 0; i < (int)p_dynamic_objects.size(); i++) { + RenderGeometryInstance *instance = p_dynamic_objects[i]; + + //transform aabb to voxel_gi + AABB aabb = (to_probe_xform * instance->get_transform()).xform(instance->get_aabb()); + + //this needs to wrap to grid resolution to avoid jitter + //also extend margin a bit just in case + Vector3i begin = aabb.position - Vector3i(1, 1, 1); + Vector3i end = aabb.position + aabb.size + Vector3i(1, 1, 1); + + for (int j = 0; j < 3; j++) { + if ((end[j] - begin[j]) & 1) { + end[j]++; //for half extents split, it needs to be even + } + begin[j] = MAX(begin[j], 0); + end[j] = MIN(end[j], octree_size[j] * multiplier); + } + + //aabb = aabb.intersection(probe_aabb); //intersect + aabb.position = begin; + aabb.size = end - begin; + + //print_line("aabb: " + aabb); + + for (int j = 0; j < 6; j++) { + //if (j != 0 && j != 3) { + // continue; + //} + static const Vector3 render_z[6] = { + Vector3(1, 0, 0), + Vector3(0, 1, 0), + Vector3(0, 0, 1), + Vector3(-1, 0, 0), + Vector3(0, -1, 0), + Vector3(0, 0, -1), + }; + static const Vector3 render_up[6] = { + Vector3(0, 1, 0), + Vector3(0, 0, 1), + Vector3(0, 1, 0), + Vector3(0, 1, 0), + Vector3(0, 0, 1), + Vector3(0, 1, 0), + }; + + Vector3 render_dir = render_z[j]; + Vector3 up_dir = render_up[j]; + + Vector3 center = aabb.get_center(); + Transform3D xform; + xform.set_look_at(center - aabb.size * 0.5 * render_dir, center, up_dir); + + Vector3 x_dir = xform.basis.get_column(0).abs(); + int x_axis = int(Vector3(0, 1, 2).dot(x_dir)); + Vector3 y_dir = xform.basis.get_column(1).abs(); + int y_axis = int(Vector3(0, 1, 2).dot(y_dir)); + Vector3 z_dir = -xform.basis.get_column(2); + int z_axis = int(Vector3(0, 1, 2).dot(z_dir.abs())); + + Rect2i rect(aabb.position[x_axis], aabb.position[y_axis], aabb.size[x_axis], aabb.size[y_axis]); + bool x_flip = bool(Vector3(1, 1, 1).dot(xform.basis.get_column(0)) < 0); + bool y_flip = bool(Vector3(1, 1, 1).dot(xform.basis.get_column(1)) < 0); + bool z_flip = bool(Vector3(1, 1, 1).dot(xform.basis.get_column(2)) > 0); + + Projection cm; + cm.set_orthogonal(-rect.size.width / 2, rect.size.width / 2, -rect.size.height / 2, rect.size.height / 2, 0.0001, aabb.size[z_axis]); + + if (RendererSceneRenderRD::get_singleton()->cull_argument.size() == 0) { + RendererSceneRenderRD::get_singleton()->cull_argument.push_back(nullptr); + } + RendererSceneRenderRD::get_singleton()->cull_argument[0] = instance; + + float exposure_normalization = 1.0; + if (RendererSceneRenderRD::get_singleton()->is_using_physical_light_units()) { + exposure_normalization = gi->voxel_gi_get_baked_exposure_normalization(probe); + } + + RendererSceneRenderRD::get_singleton()->_render_material(to_world_xform * xform, cm, true, RendererSceneRenderRD::get_singleton()->cull_argument, dynamic_maps[0].fb, Rect2i(Vector2i(), rect.size), exposure_normalization); + + VoxelGIDynamicPushConstant push_constant; + memset(&push_constant, 0, sizeof(VoxelGIDynamicPushConstant)); + push_constant.limits[0] = octree_size.x; + push_constant.limits[1] = octree_size.y; + push_constant.limits[2] = octree_size.z; + push_constant.light_count = p_light_instances.size(); + push_constant.x_dir[0] = x_dir[0]; + push_constant.x_dir[1] = x_dir[1]; + push_constant.x_dir[2] = x_dir[2]; + push_constant.y_dir[0] = y_dir[0]; + push_constant.y_dir[1] = y_dir[1]; + push_constant.y_dir[2] = y_dir[2]; + push_constant.z_dir[0] = z_dir[0]; + push_constant.z_dir[1] = z_dir[1]; + push_constant.z_dir[2] = z_dir[2]; + push_constant.z_base = xform.origin[z_axis]; + push_constant.z_sign = (z_flip ? -1.0 : 1.0); + push_constant.pos_multiplier = float(1.0) / multiplier; + push_constant.dynamic_range = gi->voxel_gi_get_dynamic_range(probe); + push_constant.flip_x = x_flip; + push_constant.flip_y = y_flip; + push_constant.rect_pos[0] = rect.position[0]; + push_constant.rect_pos[1] = rect.position[1]; + push_constant.rect_size[0] = rect.size[0]; + push_constant.rect_size[1] = rect.size[1]; + push_constant.prev_rect_ofs[0] = 0; + push_constant.prev_rect_ofs[1] = 0; + push_constant.prev_rect_size[0] = 0; + push_constant.prev_rect_size[1] = 0; + push_constant.on_mipmap = false; + push_constant.propagation = gi->voxel_gi_get_propagation(probe); + push_constant.pad[0] = 0; + push_constant.pad[1] = 0; + push_constant.pad[2] = 0; + + //process lighting + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->voxel_gi_lighting_shader_version_pipelines[VOXEL_GI_SHADER_VERSION_DYNAMIC_OBJECT_LIGHTING]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, dynamic_maps[0].uniform_set, 0); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(VoxelGIDynamicPushConstant)); + RD::get_singleton()->compute_list_dispatch(compute_list, (rect.size.x - 1) / 8 + 1, (rect.size.y - 1) / 8 + 1, 1); + //print_line("rect: " + itos(i) + ": " + rect); + + for (int k = 1; k < dynamic_maps.size(); k++) { + // enlarge the rect if needed so all pixels fit when downscaled, + // this ensures downsampling is smooth and optimal because no pixels are left behind + + //x + if (rect.position.x & 1) { + rect.size.x++; + push_constant.prev_rect_ofs[0] = 1; //this is used to ensure reading is also optimal + } else { + push_constant.prev_rect_ofs[0] = 0; + } + if (rect.size.x & 1) { + rect.size.x++; + } + + rect.position.x >>= 1; + rect.size.x = MAX(1, rect.size.x >> 1); + + //y + if (rect.position.y & 1) { + rect.size.y++; + push_constant.prev_rect_ofs[1] = 1; + } else { + push_constant.prev_rect_ofs[1] = 0; + } + if (rect.size.y & 1) { + rect.size.y++; + } + + rect.position.y >>= 1; + rect.size.y = MAX(1, rect.size.y >> 1); + + //shrink limits to ensure plot does not go outside map + if (dynamic_maps[k].mipmap > 0) { + for (int l = 0; l < 3; l++) { + push_constant.limits[l] = MAX(1, push_constant.limits[l] >> 1); + } + } + + //print_line("rect: " + itos(i) + ": " + rect); + push_constant.rect_pos[0] = rect.position[0]; + push_constant.rect_pos[1] = rect.position[1]; + push_constant.prev_rect_size[0] = push_constant.rect_size[0]; + push_constant.prev_rect_size[1] = push_constant.rect_size[1]; + push_constant.rect_size[0] = rect.size[0]; + push_constant.rect_size[1] = rect.size[1]; + push_constant.on_mipmap = dynamic_maps[k].mipmap > 0; + + RD::get_singleton()->compute_list_add_barrier(compute_list); + + if (dynamic_maps[k].mipmap < 0) { + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->voxel_gi_lighting_shader_version_pipelines[VOXEL_GI_SHADER_VERSION_DYNAMIC_SHRINK_WRITE]); + } else if (k < dynamic_maps.size() - 1) { + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->voxel_gi_lighting_shader_version_pipelines[VOXEL_GI_SHADER_VERSION_DYNAMIC_SHRINK_WRITE_PLOT]); + } else { + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->voxel_gi_lighting_shader_version_pipelines[VOXEL_GI_SHADER_VERSION_DYNAMIC_SHRINK_PLOT]); + } + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, dynamic_maps[k].uniform_set, 0); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(VoxelGIDynamicPushConstant)); + RD::get_singleton()->compute_list_dispatch(compute_list, (rect.size.x - 1) / 8 + 1, (rect.size.y - 1) / 8 + 1, 1); + } + + RD::get_singleton()->compute_list_end(); + } + } + + has_dynamic_object_data = true; //clear until dynamic object data is used again + } + + last_probe_version = gi->voxel_gi_get_version(probe); +} + +void GI::VoxelGIInstance::free_resources() { + if (texture.is_valid()) { + RD::get_singleton()->free(texture); + RD::get_singleton()->free(write_buffer); + + texture = RID(); + write_buffer = RID(); + mipmaps.clear(); + } + + for (int i = 0; i < dynamic_maps.size(); i++) { + RD::get_singleton()->free(dynamic_maps[i].texture); + RD::get_singleton()->free(dynamic_maps[i].depth); + + // these only exist on the first level... + if (dynamic_maps[i].fb_depth.is_valid()) { + RD::get_singleton()->free(dynamic_maps[i].fb_depth); + } + if (dynamic_maps[i].albedo.is_valid()) { + RD::get_singleton()->free(dynamic_maps[i].albedo); + } + if (dynamic_maps[i].normal.is_valid()) { + RD::get_singleton()->free(dynamic_maps[i].normal); + } + if (dynamic_maps[i].orm.is_valid()) { + RD::get_singleton()->free(dynamic_maps[i].orm); + } + } + dynamic_maps.clear(); +} + +void GI::VoxelGIInstance::debug(RD::DrawListID p_draw_list, RID p_framebuffer, const Projection &p_camera_with_transform, bool p_lighting, bool p_emission, float p_alpha) { + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + + if (mipmaps.size() == 0) { + return; + } + + Projection cam_transform = (p_camera_with_transform * Projection(transform)) * Projection(gi->voxel_gi_get_to_cell_xform(probe).affine_inverse()); + + int level = 0; + Vector3i octree_size = gi->voxel_gi_get_octree_size(probe); + + VoxelGIDebugPushConstant push_constant; + push_constant.alpha = p_alpha; + push_constant.dynamic_range = gi->voxel_gi_get_dynamic_range(probe); + push_constant.cell_offset = mipmaps[level].cell_offset; + push_constant.level = level; + + push_constant.bounds[0] = octree_size.x >> level; + push_constant.bounds[1] = octree_size.y >> level; + push_constant.bounds[2] = octree_size.z >> level; + push_constant.pad = 0; + + for (int i = 0; i < 4; i++) { + for (int j = 0; j < 4; j++) { + push_constant.projection[i * 4 + j] = cam_transform.columns[i][j]; + } + } + + if (gi->voxel_gi_debug_uniform_set.is_valid()) { + RD::get_singleton()->free(gi->voxel_gi_debug_uniform_set); + } + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 1; + u.append_id(gi->voxel_gi_get_data_buffer(probe)); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 2; + u.append_id(texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 3; + u.append_id(material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED)); + uniforms.push_back(u); + } + + int cell_count; + if (!p_emission && p_lighting && has_dynamic_object_data) { + cell_count = push_constant.bounds[0] * push_constant.bounds[1] * push_constant.bounds[2]; + } else { + cell_count = mipmaps[level].cell_count; + } + + gi->voxel_gi_debug_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->voxel_gi_debug_shader_version_shaders[0], 0); + + int voxel_gi_debug_pipeline = VOXEL_GI_DEBUG_COLOR; + if (p_emission) { + voxel_gi_debug_pipeline = VOXEL_GI_DEBUG_EMISSION; + } else if (p_lighting) { + voxel_gi_debug_pipeline = has_dynamic_object_data ? VOXEL_GI_DEBUG_LIGHT_FULL : VOXEL_GI_DEBUG_LIGHT; + } + RD::get_singleton()->draw_list_bind_render_pipeline( + p_draw_list, + gi->voxel_gi_debug_shader_version_pipelines[voxel_gi_debug_pipeline].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(p_draw_list, gi->voxel_gi_debug_uniform_set, 0); + RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(VoxelGIDebugPushConstant)); + RD::get_singleton()->draw_list_draw(p_draw_list, false, cell_count, 36); +} + +//////////////////////////////////////////////////////////////////////////////// +// GI + +GI::GI() { + singleton = this; + + sdfgi_ray_count = RS::EnvironmentSDFGIRayCount(CLAMP(int32_t(GLOBAL_GET("rendering/global_illumination/sdfgi/probe_ray_count")), 0, int32_t(RS::ENV_SDFGI_RAY_COUNT_MAX - 1))); + sdfgi_frames_to_converge = RS::EnvironmentSDFGIFramesToConverge(CLAMP(int32_t(GLOBAL_GET("rendering/global_illumination/sdfgi/frames_to_converge")), 0, int32_t(RS::ENV_SDFGI_CONVERGE_MAX - 1))); + sdfgi_frames_to_update_light = RS::EnvironmentSDFGIFramesToUpdateLight(CLAMP(int32_t(GLOBAL_GET("rendering/global_illumination/sdfgi/frames_to_update_lights")), 0, int32_t(RS::ENV_SDFGI_UPDATE_LIGHT_MAX - 1))); +} + +GI::~GI() { + singleton = nullptr; +} + +void GI::init(SkyRD *p_sky) { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + + /* GI */ + + { + //kinda complicated to compute the amount of slots, we try to use as many as we can + + voxel_gi_lights = memnew_arr(VoxelGILight, voxel_gi_max_lights); + voxel_gi_lights_uniform = RD::get_singleton()->uniform_buffer_create(voxel_gi_max_lights * sizeof(VoxelGILight)); + voxel_gi_quality = RS::VoxelGIQuality(CLAMP(int(GLOBAL_GET("rendering/global_illumination/voxel_gi/quality")), 0, 1)); + + String defines = "\n#define MAX_LIGHTS " + itos(voxel_gi_max_lights) + "\n"; + + Vector<String> versions; + versions.push_back("\n#define MODE_COMPUTE_LIGHT\n"); + versions.push_back("\n#define MODE_SECOND_BOUNCE\n"); + versions.push_back("\n#define MODE_UPDATE_MIPMAPS\n"); + versions.push_back("\n#define MODE_WRITE_TEXTURE\n"); + versions.push_back("\n#define MODE_DYNAMIC\n#define MODE_DYNAMIC_LIGHTING\n"); + versions.push_back("\n#define MODE_DYNAMIC\n#define MODE_DYNAMIC_SHRINK\n#define MODE_DYNAMIC_SHRINK_WRITE\n"); + versions.push_back("\n#define MODE_DYNAMIC\n#define MODE_DYNAMIC_SHRINK\n#define MODE_DYNAMIC_SHRINK_PLOT\n"); + versions.push_back("\n#define MODE_DYNAMIC\n#define MODE_DYNAMIC_SHRINK\n#define MODE_DYNAMIC_SHRINK_PLOT\n#define MODE_DYNAMIC_SHRINK_WRITE\n"); + + voxel_gi_shader.initialize(versions, defines); + voxel_gi_lighting_shader_version = voxel_gi_shader.version_create(); + for (int i = 0; i < VOXEL_GI_SHADER_VERSION_MAX; i++) { + voxel_gi_lighting_shader_version_shaders[i] = voxel_gi_shader.version_get_shader(voxel_gi_lighting_shader_version, i); + voxel_gi_lighting_shader_version_pipelines[i] = RD::get_singleton()->compute_pipeline_create(voxel_gi_lighting_shader_version_shaders[i]); + } + } + + { + String defines; + Vector<String> versions; + versions.push_back("\n#define MODE_DEBUG_COLOR\n"); + versions.push_back("\n#define MODE_DEBUG_LIGHT\n"); + versions.push_back("\n#define MODE_DEBUG_EMISSION\n"); + versions.push_back("\n#define MODE_DEBUG_LIGHT\n#define MODE_DEBUG_LIGHT_FULL\n"); + + voxel_gi_debug_shader.initialize(versions, defines); + voxel_gi_debug_shader_version = voxel_gi_debug_shader.version_create(); + for (int i = 0; i < VOXEL_GI_DEBUG_MAX; i++) { + voxel_gi_debug_shader_version_shaders[i] = voxel_gi_debug_shader.version_get_shader(voxel_gi_debug_shader_version, i); + + RD::PipelineRasterizationState rs; + rs.cull_mode = RD::POLYGON_CULL_FRONT; + RD::PipelineDepthStencilState ds; + ds.enable_depth_test = true; + ds.enable_depth_write = true; + ds.depth_compare_operator = RD::COMPARE_OP_LESS_OR_EQUAL; + + voxel_gi_debug_shader_version_pipelines[i].setup(voxel_gi_debug_shader_version_shaders[i], RD::RENDER_PRIMITIVE_TRIANGLES, rs, RD::PipelineMultisampleState(), ds, RD::PipelineColorBlendState::create_disabled(), 0); + } + } + + /* SDGFI */ + + { + Vector<String> preprocess_modes; + preprocess_modes.push_back("\n#define MODE_SCROLL\n"); + preprocess_modes.push_back("\n#define MODE_SCROLL_OCCLUSION\n"); + preprocess_modes.push_back("\n#define MODE_INITIALIZE_JUMP_FLOOD\n"); + preprocess_modes.push_back("\n#define MODE_INITIALIZE_JUMP_FLOOD_HALF\n"); + preprocess_modes.push_back("\n#define MODE_JUMPFLOOD\n"); + preprocess_modes.push_back("\n#define MODE_JUMPFLOOD_OPTIMIZED\n"); + preprocess_modes.push_back("\n#define MODE_UPSCALE_JUMP_FLOOD\n"); + preprocess_modes.push_back("\n#define MODE_OCCLUSION\n"); + preprocess_modes.push_back("\n#define MODE_STORE\n"); + String defines = "\n#define OCCLUSION_SIZE " + itos(SDFGI::CASCADE_SIZE / SDFGI::PROBE_DIVISOR) + "\n"; + sdfgi_shader.preprocess.initialize(preprocess_modes, defines); + sdfgi_shader.preprocess_shader = sdfgi_shader.preprocess.version_create(); + for (int i = 0; i < SDFGIShader::PRE_PROCESS_MAX; i++) { + sdfgi_shader.preprocess_pipeline[i] = RD::get_singleton()->compute_pipeline_create(sdfgi_shader.preprocess.version_get_shader(sdfgi_shader.preprocess_shader, i)); + } + } + + { + //calculate tables + String defines = "\n#define OCT_SIZE " + itos(SDFGI::LIGHTPROBE_OCT_SIZE) + "\n"; + + Vector<String> direct_light_modes; + direct_light_modes.push_back("\n#define MODE_PROCESS_STATIC\n"); + direct_light_modes.push_back("\n#define MODE_PROCESS_DYNAMIC\n"); + sdfgi_shader.direct_light.initialize(direct_light_modes, defines); + sdfgi_shader.direct_light_shader = sdfgi_shader.direct_light.version_create(); + for (int i = 0; i < SDFGIShader::DIRECT_LIGHT_MODE_MAX; i++) { + sdfgi_shader.direct_light_pipeline[i] = RD::get_singleton()->compute_pipeline_create(sdfgi_shader.direct_light.version_get_shader(sdfgi_shader.direct_light_shader, i)); + } + } + + { + //calculate tables + String defines = "\n#define OCT_SIZE " + itos(SDFGI::LIGHTPROBE_OCT_SIZE) + "\n"; + defines += "\n#define SH_SIZE " + itos(SDFGI::SH_SIZE) + "\n"; + if (p_sky->sky_use_cubemap_array) { + defines += "\n#define USE_CUBEMAP_ARRAY\n"; + } + + Vector<String> integrate_modes; + integrate_modes.push_back("\n#define MODE_PROCESS\n"); + integrate_modes.push_back("\n#define MODE_STORE\n"); + integrate_modes.push_back("\n#define MODE_SCROLL\n"); + integrate_modes.push_back("\n#define MODE_SCROLL_STORE\n"); + sdfgi_shader.integrate.initialize(integrate_modes, defines); + sdfgi_shader.integrate_shader = sdfgi_shader.integrate.version_create(); + + for (int i = 0; i < SDFGIShader::INTEGRATE_MODE_MAX; i++) { + sdfgi_shader.integrate_pipeline[i] = RD::get_singleton()->compute_pipeline_create(sdfgi_shader.integrate.version_get_shader(sdfgi_shader.integrate_shader, i)); + } + + { + Vector<RD::Uniform> uniforms; + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 0; + if (p_sky->sky_use_cubemap_array) { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_WHITE)); + } else { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_CUBEMAP_WHITE)); + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 1; + u.append_id(material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED)); + uniforms.push_back(u); + } + + sdfgi_shader.integrate_default_sky_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.integrate.version_get_shader(sdfgi_shader.integrate_shader, 0), 1); + } + } + + //GK + { + //calculate tables + String defines = "\n#define SDFGI_OCT_SIZE " + itos(SDFGI::LIGHTPROBE_OCT_SIZE) + "\n"; + Vector<String> gi_modes; + + gi_modes.push_back("\n#define USE_VOXEL_GI_INSTANCES\n"); // MODE_VOXEL_GI + gi_modes.push_back("\n#define USE_SDFGI\n"); // MODE_SDFGI + gi_modes.push_back("\n#define USE_SDFGI\n\n#define USE_VOXEL_GI_INSTANCES\n"); // MODE_COMBINED + + shader.initialize(gi_modes, defines); + shader_version = shader.version_create(); + + Vector<RD::PipelineSpecializationConstant> specialization_constants; + + { + RD::PipelineSpecializationConstant sc; + sc.type = RD::PIPELINE_SPECIALIZATION_CONSTANT_TYPE_BOOL; + sc.constant_id = 0; // SHADER_SPECIALIZATION_HALF_RES + sc.bool_value = false; + specialization_constants.push_back(sc); + + sc.type = RD::PIPELINE_SPECIALIZATION_CONSTANT_TYPE_BOOL; + sc.constant_id = 1; // SHADER_SPECIALIZATION_USE_FULL_PROJECTION_MATRIX + sc.bool_value = false; + specialization_constants.push_back(sc); + + sc.type = RD::PIPELINE_SPECIALIZATION_CONSTANT_TYPE_BOOL; + sc.constant_id = 2; // SHADER_SPECIALIZATION_USE_VRS + sc.bool_value = false; + specialization_constants.push_back(sc); + } + + for (int v = 0; v < SHADER_SPECIALIZATION_VARIATIONS; v++) { + specialization_constants.ptrw()[0].bool_value = (v & SHADER_SPECIALIZATION_HALF_RES) ? true : false; + specialization_constants.ptrw()[1].bool_value = (v & SHADER_SPECIALIZATION_USE_FULL_PROJECTION_MATRIX) ? true : false; + specialization_constants.ptrw()[2].bool_value = (v & SHADER_SPECIALIZATION_USE_VRS) ? true : false; + for (int i = 0; i < MODE_MAX; i++) { + pipelines[v][i] = RD::get_singleton()->compute_pipeline_create(shader.version_get_shader(shader_version, i), specialization_constants); + } + } + + sdfgi_ubo = RD::get_singleton()->uniform_buffer_create(sizeof(SDFGIData)); + } + { + String defines = "\n#define OCT_SIZE " + itos(SDFGI::LIGHTPROBE_OCT_SIZE) + "\n"; + Vector<String> debug_modes; + debug_modes.push_back(""); + sdfgi_shader.debug.initialize(debug_modes, defines); + sdfgi_shader.debug_shader = sdfgi_shader.debug.version_create(); + sdfgi_shader.debug_shader_version = sdfgi_shader.debug.version_get_shader(sdfgi_shader.debug_shader, 0); + sdfgi_shader.debug_pipeline = RD::get_singleton()->compute_pipeline_create(sdfgi_shader.debug_shader_version); + } + { + String defines = "\n#define OCT_SIZE " + itos(SDFGI::LIGHTPROBE_OCT_SIZE) + "\n"; + + Vector<String> versions; + versions.push_back("\n#define MODE_PROBES\n"); + versions.push_back("\n#define MODE_PROBES\n#define USE_MULTIVIEW\n"); + versions.push_back("\n#define MODE_VISIBILITY\n"); + versions.push_back("\n#define MODE_VISIBILITY\n#define USE_MULTIVIEW\n"); + + sdfgi_shader.debug_probes.initialize(versions, defines); + + // TODO disable multiview versions if turned off + + sdfgi_shader.debug_probes_shader = sdfgi_shader.debug_probes.version_create(); + + { + RD::PipelineRasterizationState rs; + rs.cull_mode = RD::POLYGON_CULL_DISABLED; + RD::PipelineDepthStencilState ds; + ds.enable_depth_test = true; + ds.enable_depth_write = true; + ds.depth_compare_operator = RD::COMPARE_OP_LESS_OR_EQUAL; + for (int i = 0; i < SDFGIShader::PROBE_DEBUG_MAX; i++) { + // TODO check if version is enabled + + RID debug_probes_shader_version = sdfgi_shader.debug_probes.version_get_shader(sdfgi_shader.debug_probes_shader, i); + sdfgi_shader.debug_probes_pipeline[i].setup(debug_probes_shader_version, RD::RENDER_PRIMITIVE_TRIANGLE_STRIPS, rs, RD::PipelineMultisampleState(), ds, RD::PipelineColorBlendState::create_disabled(), 0); + } + } + } + default_voxel_gi_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(VoxelGIData) * MAX_VOXEL_GI_INSTANCES); + half_resolution = GLOBAL_GET("rendering/global_illumination/gi/use_half_resolution"); +} + +void GI::free() { + RD::get_singleton()->free(default_voxel_gi_buffer); + RD::get_singleton()->free(voxel_gi_lights_uniform); + RD::get_singleton()->free(sdfgi_ubo); + + voxel_gi_debug_shader.version_free(voxel_gi_debug_shader_version); + voxel_gi_shader.version_free(voxel_gi_lighting_shader_version); + shader.version_free(shader_version); + sdfgi_shader.debug_probes.version_free(sdfgi_shader.debug_probes_shader); + sdfgi_shader.debug.version_free(sdfgi_shader.debug_shader); + sdfgi_shader.direct_light.version_free(sdfgi_shader.direct_light_shader); + sdfgi_shader.integrate.version_free(sdfgi_shader.integrate_shader); + sdfgi_shader.preprocess.version_free(sdfgi_shader.preprocess_shader); + + if (voxel_gi_lights) { + memdelete_arr(voxel_gi_lights); + } +} + +Ref<GI::SDFGI> GI::create_sdfgi(RID p_env, const Vector3 &p_world_position, uint32_t p_requested_history_size) { + Ref<SDFGI> sdfgi; + sdfgi.instantiate(); + + sdfgi->create(p_env, p_world_position, p_requested_history_size, this); + + return sdfgi; +} + +void GI::setup_voxel_gi_instances(RenderDataRD *p_render_data, Ref<RenderSceneBuffersRD> p_render_buffers, const Transform3D &p_transform, const PagedArray<RID> &p_voxel_gi_instances, uint32_t &r_voxel_gi_instances_used) { + ERR_FAIL_COND(p_render_buffers.is_null()); + + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + ERR_FAIL_NULL(texture_storage); + + r_voxel_gi_instances_used = 0; + + Ref<RenderBuffersGI> rbgi = p_render_buffers->get_custom_data(RB_SCOPE_GI); + ERR_FAIL_COND(rbgi.is_null()); + + RID voxel_gi_buffer = rbgi->get_voxel_gi_buffer(); + VoxelGIData voxel_gi_data[MAX_VOXEL_GI_INSTANCES]; + + bool voxel_gi_instances_changed = false; + + Transform3D to_camera; + to_camera.origin = p_transform.origin; //only translation, make local + + for (int i = 0; i < MAX_VOXEL_GI_INSTANCES; i++) { + RID texture; + if (i < (int)p_voxel_gi_instances.size()) { + VoxelGIInstance *gipi = voxel_gi_instance_owner.get_or_null(p_voxel_gi_instances[i]); + + if (gipi) { + texture = gipi->texture; + VoxelGIData &gipd = voxel_gi_data[i]; + + RID base_probe = gipi->probe; + + Transform3D to_cell = voxel_gi_get_to_cell_xform(gipi->probe) * gipi->transform.affine_inverse() * to_camera; + + gipd.xform[0] = to_cell.basis.rows[0][0]; + gipd.xform[1] = to_cell.basis.rows[1][0]; + gipd.xform[2] = to_cell.basis.rows[2][0]; + gipd.xform[3] = 0; + gipd.xform[4] = to_cell.basis.rows[0][1]; + gipd.xform[5] = to_cell.basis.rows[1][1]; + gipd.xform[6] = to_cell.basis.rows[2][1]; + gipd.xform[7] = 0; + gipd.xform[8] = to_cell.basis.rows[0][2]; + gipd.xform[9] = to_cell.basis.rows[1][2]; + gipd.xform[10] = to_cell.basis.rows[2][2]; + gipd.xform[11] = 0; + gipd.xform[12] = to_cell.origin.x; + gipd.xform[13] = to_cell.origin.y; + gipd.xform[14] = to_cell.origin.z; + gipd.xform[15] = 1; + + Vector3 bounds = voxel_gi_get_octree_size(base_probe); + + gipd.bounds[0] = bounds.x; + gipd.bounds[1] = bounds.y; + gipd.bounds[2] = bounds.z; + + gipd.dynamic_range = voxel_gi_get_dynamic_range(base_probe) * voxel_gi_get_energy(base_probe); + gipd.bias = voxel_gi_get_bias(base_probe); + gipd.normal_bias = voxel_gi_get_normal_bias(base_probe); + gipd.blend_ambient = !voxel_gi_is_interior(base_probe); + gipd.mipmaps = gipi->mipmaps.size(); + gipd.exposure_normalization = 1.0; + if (p_render_data->camera_attributes.is_valid()) { + float exposure_normalization = RSG::camera_attributes->camera_attributes_get_exposure_normalization_factor(p_render_data->camera_attributes); + gipd.exposure_normalization = exposure_normalization / voxel_gi_get_baked_exposure_normalization(base_probe); + } + } + + r_voxel_gi_instances_used++; + } + + if (texture == RID()) { + texture = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE); + } + + if (texture != rbgi->voxel_gi_textures[i]) { + voxel_gi_instances_changed = true; + rbgi->voxel_gi_textures[i] = texture; + } + } + + if (voxel_gi_instances_changed) { + for (uint32_t v = 0; v < RendererSceneRender::MAX_RENDER_VIEWS; v++) { + if (RD::get_singleton()->uniform_set_is_valid(rbgi->uniform_set[v])) { + RD::get_singleton()->free(rbgi->uniform_set[v]); + } + rbgi->uniform_set[v] = RID(); + } + if (p_render_buffers->has_custom_data(RB_SCOPE_FOG)) { + Ref<Fog::VolumetricFog> fog = p_render_buffers->get_custom_data(RB_SCOPE_FOG); + + if (RD::get_singleton()->uniform_set_is_valid(fog->fog_uniform_set)) { + RD::get_singleton()->free(fog->fog_uniform_set); + RD::get_singleton()->free(fog->process_uniform_set); + RD::get_singleton()->free(fog->process_uniform_set2); + } + fog->fog_uniform_set = RID(); + fog->process_uniform_set = RID(); + fog->process_uniform_set2 = RID(); + } + } + + if (p_voxel_gi_instances.size() > 0) { + RD::get_singleton()->draw_command_begin_label("VoxelGIs Setup"); + + RD::get_singleton()->buffer_update(voxel_gi_buffer, 0, sizeof(VoxelGIData) * MIN((uint64_t)MAX_VOXEL_GI_INSTANCES, p_voxel_gi_instances.size()), voxel_gi_data, RD::BARRIER_MASK_COMPUTE); + + RD::get_singleton()->draw_command_end_label(); + } +} + +RID GI::RenderBuffersGI::get_voxel_gi_buffer() { + if (voxel_gi_buffer.is_null()) { + voxel_gi_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(GI::VoxelGIData) * GI::MAX_VOXEL_GI_INSTANCES); + } + return voxel_gi_buffer; +} + +void GI::RenderBuffersGI::free_data() { + for (uint32_t v = 0; v < RendererSceneRender::MAX_RENDER_VIEWS; v++) { + if (RD::get_singleton()->uniform_set_is_valid(uniform_set[v])) { + RD::get_singleton()->free(uniform_set[v]); + } + uniform_set[v] = RID(); + } + + if (scene_data_ubo.is_valid()) { + RD::get_singleton()->free(scene_data_ubo); + scene_data_ubo = RID(); + } + + if (voxel_gi_buffer.is_valid()) { + RD::get_singleton()->free(voxel_gi_buffer); + voxel_gi_buffer = RID(); + } +} + +void GI::process_gi(Ref<RenderSceneBuffersRD> p_render_buffers, const RID *p_normal_roughness_slices, RID p_voxel_gi_buffer, RID p_environment, uint32_t p_view_count, const Projection *p_projections, const Vector3 *p_eye_offsets, const Transform3D &p_cam_transform, const PagedArray<RID> &p_voxel_gi_instances) { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + + ERR_FAIL_COND_MSG(p_view_count > 2, "Maximum of 2 views supported for Processing GI."); + + RD::get_singleton()->draw_command_begin_label("GI Render"); + + ERR_FAIL_COND(p_render_buffers.is_null()); + + Ref<RenderBuffersGI> rbgi = p_render_buffers->get_custom_data(RB_SCOPE_GI); + ERR_FAIL_COND(rbgi.is_null()); + + Size2i internal_size = p_render_buffers->get_internal_size(); + + if (rbgi->using_half_size_gi != half_resolution) { + p_render_buffers->clear_context(RB_SCOPE_GI); + } + + if (!p_render_buffers->has_texture(RB_SCOPE_GI, RB_TEX_AMBIENT)) { + Size2i size = internal_size; + uint32_t usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + + if (half_resolution) { + size.x >>= 1; + size.y >>= 1; + } + + p_render_buffers->create_texture(RB_SCOPE_GI, RB_TEX_AMBIENT, RD::DATA_FORMAT_R16G16B16A16_SFLOAT, usage_bits, RD::TEXTURE_SAMPLES_1, size); + p_render_buffers->create_texture(RB_SCOPE_GI, RB_TEX_REFLECTION, RD::DATA_FORMAT_R16G16B16A16_SFLOAT, usage_bits, RD::TEXTURE_SAMPLES_1, size); + + rbgi->using_half_size_gi = half_resolution; + } + + // Setup our scene data + { + SceneData scene_data; + + if (rbgi->scene_data_ubo.is_null()) { + rbgi->scene_data_ubo = RD::get_singleton()->uniform_buffer_create(sizeof(SceneData)); + } + + for (uint32_t v = 0; v < p_view_count; v++) { + RendererRD::MaterialStorage::store_camera(p_projections[v].inverse(), scene_data.inv_projection[v]); + scene_data.eye_offset[v][0] = p_eye_offsets[v].x; + scene_data.eye_offset[v][1] = p_eye_offsets[v].y; + scene_data.eye_offset[v][2] = p_eye_offsets[v].z; + scene_data.eye_offset[v][3] = 0.0; + } + + // Note that we will be ignoring the origin of this transform. + RendererRD::MaterialStorage::store_transform(p_cam_transform, scene_data.cam_transform); + + scene_data.screen_size[0] = internal_size.x; + scene_data.screen_size[1] = internal_size.y; + + RD::get_singleton()->buffer_update(rbgi->scene_data_ubo, 0, sizeof(SceneData), &scene_data, RD::BARRIER_MASK_COMPUTE); + } + + // Now compute the contents of our buffers. + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(true); + + // Render each eye separately. + // We need to look into whether we can make our compute shader use Multiview but not sure that works or makes a difference.. + + // setup our push constant + + PushConstant push_constant; + + push_constant.max_voxel_gi_instances = MIN((uint64_t)MAX_VOXEL_GI_INSTANCES, p_voxel_gi_instances.size()); + push_constant.high_quality_vct = voxel_gi_quality == RS::VOXEL_GI_QUALITY_HIGH; + + // these should be the same for all views + push_constant.orthogonal = p_projections[0].is_orthogonal(); + push_constant.z_near = p_projections[0].get_z_near(); + push_constant.z_far = p_projections[0].get_z_far(); + + // these are only used if we have 1 view, else we use the projections in our scene data + push_constant.proj_info[0] = -2.0f / (internal_size.x * p_projections[0].columns[0][0]); + push_constant.proj_info[1] = -2.0f / (internal_size.y * p_projections[0].columns[1][1]); + push_constant.proj_info[2] = (1.0f - p_projections[0].columns[0][2]) / p_projections[0].columns[0][0]; + push_constant.proj_info[3] = (1.0f + p_projections[0].columns[1][2]) / p_projections[0].columns[1][1]; + + bool use_sdfgi = p_render_buffers->has_custom_data(RB_SCOPE_SDFGI); + bool use_voxel_gi_instances = push_constant.max_voxel_gi_instances > 0; + + Ref<SDFGI> sdfgi; + if (use_sdfgi) { + sdfgi = p_render_buffers->get_custom_data(RB_SCOPE_SDFGI); + } + + uint32_t pipeline_specialization = 0; + if (rbgi->using_half_size_gi) { + pipeline_specialization |= SHADER_SPECIALIZATION_HALF_RES; + } + if (p_view_count > 1) { + pipeline_specialization |= SHADER_SPECIALIZATION_USE_FULL_PROJECTION_MATRIX; + } + bool has_vrs_texture = p_render_buffers->has_texture(RB_SCOPE_VRS, RB_TEXTURE); + if (has_vrs_texture) { + pipeline_specialization |= SHADER_SPECIALIZATION_USE_VRS; + } + + Mode mode = (use_sdfgi && use_voxel_gi_instances) ? MODE_COMBINED : (use_sdfgi ? MODE_SDFGI : MODE_VOXEL_GI); + + for (uint32_t v = 0; v < p_view_count; v++) { + push_constant.view_index = v; + + // setup our uniform set + if (rbgi->uniform_set[v].is_null() || !RD::get_singleton()->uniform_set_is_valid(rbgi->uniform_set[v])) { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.binding = 1; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + for (uint32_t j = 0; j < SDFGI::MAX_CASCADES; j++) { + if (use_sdfgi && j < sdfgi->cascades.size()) { + u.append_id(sdfgi->cascades[j].sdf_tex); + } else { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE)); + } + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 2; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + for (uint32_t j = 0; j < SDFGI::MAX_CASCADES; j++) { + if (use_sdfgi && j < sdfgi->cascades.size()) { + u.append_id(sdfgi->cascades[j].light_tex); + } else { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE)); + } + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 3; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + for (uint32_t j = 0; j < SDFGI::MAX_CASCADES; j++) { + if (use_sdfgi && j < sdfgi->cascades.size()) { + u.append_id(sdfgi->cascades[j].light_aniso_0_tex); + } else { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE)); + } + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 4; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + for (uint32_t j = 0; j < SDFGI::MAX_CASCADES; j++) { + if (use_sdfgi && j < sdfgi->cascades.size()) { + u.append_id(sdfgi->cascades[j].light_aniso_1_tex); + } else { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE)); + } + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 5; + if (use_sdfgi) { + u.append_id(sdfgi->occlusion_texture); + } else { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE)); + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 6; + u.append_id(material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED)); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 7; + u.append_id(material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED)); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 9; + u.append_id(p_render_buffers->get_texture_slice(RB_SCOPE_GI, RB_TEX_AMBIENT, v, 0)); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 10; + u.append_id(p_render_buffers->get_texture_slice(RB_SCOPE_GI, RB_TEX_REFLECTION, v, 0)); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 11; + if (use_sdfgi) { + u.append_id(sdfgi->lightprobe_texture); + } else { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE)); + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 12; + u.append_id(p_render_buffers->get_depth_texture(v)); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 13; + u.append_id(p_normal_roughness_slices[v]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 14; + RID buffer = p_voxel_gi_buffer.is_valid() ? p_voxel_gi_buffer : texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_BLACK); + u.append_id(buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 15; + u.append_id(sdfgi_ubo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 16; + u.append_id(rbgi->get_voxel_gi_buffer()); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 17; + for (int i = 0; i < MAX_VOXEL_GI_INSTANCES; i++) { + u.append_id(rbgi->voxel_gi_textures[i]); + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 18; + u.append_id(rbgi->scene_data_ubo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 19; + RID buffer = has_vrs_texture ? p_render_buffers->get_texture_slice(RB_SCOPE_VRS, RB_TEXTURE, v, 0) : texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_VRS); + u.append_id(buffer); + uniforms.push_back(u); + } + + rbgi->uniform_set[v] = RD::get_singleton()->uniform_set_create(uniforms, shader.version_get_shader(shader_version, 0), 0); + } + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, pipelines[pipeline_specialization][mode]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rbgi->uniform_set[v], 0); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(PushConstant)); + + if (rbgi->using_half_size_gi) { + RD::get_singleton()->compute_list_dispatch_threads(compute_list, internal_size.x >> 1, internal_size.y >> 1, 1); + } else { + RD::get_singleton()->compute_list_dispatch_threads(compute_list, internal_size.x, internal_size.y, 1); + } + } + + //do barrier later to allow oeverlap + //RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_NO_BARRIER); //no barriers, let other compute, raster and transfer happen at the same time + RD::get_singleton()->draw_command_end_label(); +} + +RID GI::voxel_gi_instance_create(RID p_base) { + VoxelGIInstance voxel_gi; + voxel_gi.gi = this; + voxel_gi.probe = p_base; + RID rid = voxel_gi_instance_owner.make_rid(voxel_gi); + return rid; +} + +void GI::voxel_gi_instance_free(RID p_rid) { + GI::VoxelGIInstance *voxel_gi = voxel_gi_instance_owner.get_or_null(p_rid); + voxel_gi->free_resources(); + voxel_gi_instance_owner.free(p_rid); +} + +void GI::voxel_gi_instance_set_transform_to_data(RID p_probe, const Transform3D &p_xform) { + VoxelGIInstance *voxel_gi = voxel_gi_instance_owner.get_or_null(p_probe); + ERR_FAIL_COND(!voxel_gi); + + voxel_gi->transform = p_xform; +} + +bool GI::voxel_gi_needs_update(RID p_probe) const { + VoxelGIInstance *voxel_gi = voxel_gi_instance_owner.get_or_null(p_probe); + ERR_FAIL_COND_V(!voxel_gi, false); + + return voxel_gi->last_probe_version != voxel_gi_get_version(voxel_gi->probe); +} + +void GI::voxel_gi_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RenderGeometryInstance *> &p_dynamic_objects) { + VoxelGIInstance *voxel_gi = voxel_gi_instance_owner.get_or_null(p_probe); + ERR_FAIL_COND(!voxel_gi); + + voxel_gi->update(p_update_light_instances, p_light_instances, p_dynamic_objects); +} + +void GI::debug_voxel_gi(RID p_voxel_gi, RD::DrawListID p_draw_list, RID p_framebuffer, const Projection &p_camera_with_transform, bool p_lighting, bool p_emission, float p_alpha) { + VoxelGIInstance *voxel_gi = voxel_gi_instance_owner.get_or_null(p_voxel_gi); + ERR_FAIL_COND(!voxel_gi); + + voxel_gi->debug(p_draw_list, p_framebuffer, p_camera_with_transform, p_lighting, p_emission, p_alpha); +} diff --git a/servers/rendering/renderer_rd/environment/gi.h b/servers/rendering/renderer_rd/environment/gi.h new file mode 100644 index 0000000000..2182ca6a20 --- /dev/null +++ b/servers/rendering/renderer_rd/environment/gi.h @@ -0,0 +1,827 @@ +/*************************************************************************/ +/* gi.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef GI_RD_H +#define GI_RD_H + +#include "core/templates/local_vector.h" +#include "core/templates/rid_owner.h" +#include "servers/rendering/environment/renderer_gi.h" +#include "servers/rendering/renderer_compositor.h" +#include "servers/rendering/renderer_rd/environment/sky.h" +#include "servers/rendering/renderer_rd/shaders/environment/gi.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/environment/sdfgi_debug.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/environment/sdfgi_debug_probes.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/environment/sdfgi_direct_light.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/environment/sdfgi_integrate.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/environment/sdfgi_preprocess.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/environment/voxel_gi.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/environment/voxel_gi_debug.glsl.gen.h" +#include "servers/rendering/renderer_rd/storage_rd/render_buffer_custom_data_rd.h" +#include "servers/rendering/renderer_scene_render.h" +#include "servers/rendering/rendering_device.h" +#include "servers/rendering/storage/utilities.h" + +#define RB_SCOPE_GI SNAME("rbgi") +#define RB_SCOPE_SDFGI SNAME("sdfgi") + +#define RB_TEX_AMBIENT SNAME("ambient") +#define RB_TEX_REFLECTION SNAME("reflection") + +// Forward declare RenderDataRD and RendererSceneRenderRD so we can pass it into some of our methods, these classes are pretty tightly bound +struct RenderDataRD; +class RendererSceneRenderRD; + +namespace RendererRD { + +class GI : public RendererGI { +public: + /* VOXEL GI STORAGE */ + + struct VoxelGI { + RID octree_buffer; + RID data_buffer; + RID sdf_texture; + + uint32_t octree_buffer_size = 0; + uint32_t data_buffer_size = 0; + + Vector<int> level_counts; + + int cell_count = 0; + + Transform3D to_cell_xform; + AABB bounds; + Vector3i octree_size; + + float dynamic_range = 2.0; + float energy = 1.0; + float baked_exposure = 1.0; + float bias = 1.4; + float normal_bias = 0.0; + float propagation = 0.5; + bool interior = false; + bool use_two_bounces = true; + + uint32_t version = 1; + uint32_t data_version = 1; + + Dependency dependency; + }; + + /* VOXEL_GI INSTANCE */ + + //@TODO VoxelGIInstance is still directly used in the render code, we'll address this when we refactor the render code itself. + + struct VoxelGIInstance { + // access to our containers + GI *gi = nullptr; + + RID probe; + RID texture; + RID write_buffer; + + struct Mipmap { + RID texture; + RID uniform_set; + RID second_bounce_uniform_set; + RID write_uniform_set; + uint32_t level; + uint32_t cell_offset; + uint32_t cell_count; + }; + Vector<Mipmap> mipmaps; + + struct DynamicMap { + RID texture; //color normally, or emission on first pass + RID fb_depth; //actual depth buffer for the first pass, float depth for later passes + RID depth; //actual depth buffer for the first pass, float depth for later passes + RID normal; //normal buffer for the first pass + RID albedo; //emission buffer for the first pass + RID orm; //orm buffer for the first pass + RID fb; //used for rendering, only valid on first map + RID uniform_set; + uint32_t size; + int mipmap; // mipmap to write to, -1 if no mipmap assigned + }; + + Vector<DynamicMap> dynamic_maps; + + int slot = -1; + uint32_t last_probe_version = 0; + uint32_t last_probe_data_version = 0; + + //uint64_t last_pass = 0; + uint32_t render_index = 0; + + bool has_dynamic_object_data = false; + + Transform3D transform; + + void update(bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RenderGeometryInstance *> &p_dynamic_objects); + void debug(RD::DrawListID p_draw_list, RID p_framebuffer, const Projection &p_camera_with_transform, bool p_lighting, bool p_emission, float p_alpha); + void free_resources(); + }; + +private: + static GI *singleton; + + /* VOXEL GI STORAGE */ + + mutable RID_Owner<VoxelGI, true> voxel_gi_owner; + + /* VOXEL_GI INSTANCE */ + + mutable RID_Owner<VoxelGIInstance> voxel_gi_instance_owner; + + struct VoxelGILight { + uint32_t type; + float energy; + float radius; + float attenuation; + + float color[3]; + float cos_spot_angle; + + float position[3]; + float inv_spot_attenuation; + + float direction[3]; + uint32_t has_shadow; + }; + + struct VoxelGIPushConstant { + int32_t limits[3]; + uint32_t stack_size; + + float emission_scale; + float propagation; + float dynamic_range; + uint32_t light_count; + + uint32_t cell_offset; + uint32_t cell_count; + float aniso_strength; + uint32_t pad; + }; + + struct VoxelGIDynamicPushConstant { + int32_t limits[3]; + uint32_t light_count; + int32_t x_dir[3]; + float z_base; + int32_t y_dir[3]; + float z_sign; + int32_t z_dir[3]; + float pos_multiplier; + uint32_t rect_pos[2]; + uint32_t rect_size[2]; + uint32_t prev_rect_ofs[2]; + uint32_t prev_rect_size[2]; + uint32_t flip_x; + uint32_t flip_y; + float dynamic_range; + uint32_t on_mipmap; + float propagation; + float pad[3]; + }; + + VoxelGILight *voxel_gi_lights = nullptr; + uint32_t voxel_gi_max_lights = 32; + RID voxel_gi_lights_uniform; + + enum { + VOXEL_GI_SHADER_VERSION_COMPUTE_LIGHT, + VOXEL_GI_SHADER_VERSION_COMPUTE_SECOND_BOUNCE, + VOXEL_GI_SHADER_VERSION_COMPUTE_MIPMAP, + VOXEL_GI_SHADER_VERSION_WRITE_TEXTURE, + VOXEL_GI_SHADER_VERSION_DYNAMIC_OBJECT_LIGHTING, + VOXEL_GI_SHADER_VERSION_DYNAMIC_SHRINK_WRITE, + VOXEL_GI_SHADER_VERSION_DYNAMIC_SHRINK_PLOT, + VOXEL_GI_SHADER_VERSION_DYNAMIC_SHRINK_WRITE_PLOT, + VOXEL_GI_SHADER_VERSION_MAX + }; + + VoxelGiShaderRD voxel_gi_shader; + RID voxel_gi_lighting_shader_version; + RID voxel_gi_lighting_shader_version_shaders[VOXEL_GI_SHADER_VERSION_MAX]; + RID voxel_gi_lighting_shader_version_pipelines[VOXEL_GI_SHADER_VERSION_MAX]; + + enum { + VOXEL_GI_DEBUG_COLOR, + VOXEL_GI_DEBUG_LIGHT, + VOXEL_GI_DEBUG_EMISSION, + VOXEL_GI_DEBUG_LIGHT_FULL, + VOXEL_GI_DEBUG_MAX + }; + + struct VoxelGIDebugPushConstant { + float projection[16]; + uint32_t cell_offset; + float dynamic_range; + float alpha; + uint32_t level; + int32_t bounds[3]; + uint32_t pad; + }; + + VoxelGiDebugShaderRD voxel_gi_debug_shader; + RID voxel_gi_debug_shader_version; + RID voxel_gi_debug_shader_version_shaders[VOXEL_GI_DEBUG_MAX]; + PipelineCacheRD voxel_gi_debug_shader_version_pipelines[VOXEL_GI_DEBUG_MAX]; + RID voxel_gi_debug_uniform_set; + + /* SDFGI */ + + struct SDFGIShader { + enum SDFGIPreprocessShaderVersion { + PRE_PROCESS_SCROLL, + PRE_PROCESS_SCROLL_OCCLUSION, + PRE_PROCESS_JUMP_FLOOD_INITIALIZE, + PRE_PROCESS_JUMP_FLOOD_INITIALIZE_HALF, + PRE_PROCESS_JUMP_FLOOD, + PRE_PROCESS_JUMP_FLOOD_OPTIMIZED, + PRE_PROCESS_JUMP_FLOOD_UPSCALE, + PRE_PROCESS_OCCLUSION, + PRE_PROCESS_STORE, + PRE_PROCESS_MAX + }; + + struct PreprocessPushConstant { + int32_t scroll[3]; + int32_t grid_size; + + int32_t probe_offset[3]; + int32_t step_size; + + int32_t half_size; + uint32_t occlusion_index; + int32_t cascade; + uint32_t pad; + }; + + SdfgiPreprocessShaderRD preprocess; + RID preprocess_shader; + RID preprocess_pipeline[PRE_PROCESS_MAX]; + + struct DebugPushConstant { + float grid_size[3]; + uint32_t max_cascades; + + int32_t screen_size[2]; + float y_mult; + + float z_near; + + float inv_projection[3][4]; + float cam_basis[3][3]; + float cam_origin[3]; + }; + + SdfgiDebugShaderRD debug; + RID debug_shader; + RID debug_shader_version; + RID debug_pipeline; + + enum ProbeDebugMode { + PROBE_DEBUG_PROBES, + PROBE_DEBUG_PROBES_MULTIVIEW, + PROBE_DEBUG_VISIBILITY, + PROBE_DEBUG_VISIBILITY_MULTIVIEW, + PROBE_DEBUG_MAX + }; + + struct DebugProbesSceneData { + float projection[2][16]; + }; + + struct DebugProbesPushConstant { + uint32_t band_power; + uint32_t sections_in_band; + uint32_t band_mask; + float section_arc; + + float grid_size[3]; + uint32_t cascade; + + uint32_t pad; + float y_mult; + int32_t probe_debug_index; + int32_t probe_axis_size; + }; + + SdfgiDebugProbesShaderRD debug_probes; + RID debug_probes_shader; + RID debug_probes_shader_version; + + PipelineCacheRD debug_probes_pipeline[PROBE_DEBUG_MAX]; + + struct Light { + float color[3]; + float energy; + + float direction[3]; + uint32_t has_shadow; + + float position[3]; + float attenuation; + + uint32_t type; + float cos_spot_angle; + float inv_spot_attenuation; + float radius; + }; + + struct DirectLightPushConstant { + float grid_size[3]; + uint32_t max_cascades; + + uint32_t cascade; + uint32_t light_count; + uint32_t process_offset; + uint32_t process_increment; + + int32_t probe_axis_size; + float bounce_feedback; + float y_mult; + uint32_t use_occlusion; + }; + + enum { + DIRECT_LIGHT_MODE_STATIC, + DIRECT_LIGHT_MODE_DYNAMIC, + DIRECT_LIGHT_MODE_MAX + }; + SdfgiDirectLightShaderRD direct_light; + RID direct_light_shader; + RID direct_light_pipeline[DIRECT_LIGHT_MODE_MAX]; + + enum { + INTEGRATE_MODE_PROCESS, + INTEGRATE_MODE_STORE, + INTEGRATE_MODE_SCROLL, + INTEGRATE_MODE_SCROLL_STORE, + INTEGRATE_MODE_MAX + }; + struct IntegratePushConstant { + enum { + SKY_MODE_DISABLED, + SKY_MODE_COLOR, + SKY_MODE_SKY, + }; + + float grid_size[3]; + uint32_t max_cascades; + + uint32_t probe_axis_size; + uint32_t cascade; + uint32_t history_index; + uint32_t history_size; + + uint32_t ray_count; + float ray_bias; + int32_t image_size[2]; + + int32_t world_offset[3]; + uint32_t sky_mode; + + int32_t scroll[3]; + float sky_energy; + + float sky_color[3]; + float y_mult; + + uint32_t store_ambient_texture; + uint32_t pad[3]; + }; + + SdfgiIntegrateShaderRD integrate; + RID integrate_shader; + RID integrate_pipeline[INTEGRATE_MODE_MAX]; + + RID integrate_default_sky_uniform_set; + + } sdfgi_shader; + +public: + static GI *get_singleton() { return singleton; } + + /* GI */ + + enum { + MAX_VOXEL_GI_INSTANCES = 8 + }; + + // Struct for use in render buffer + class RenderBuffersGI : public RenderBufferCustomDataRD { + GDCLASS(RenderBuffersGI, RenderBufferCustomDataRD) + + private: + RID voxel_gi_buffer; + + public: + RID voxel_gi_textures[MAX_VOXEL_GI_INSTANCES]; + + RID full_buffer; + RID full_dispatch; + RID full_mask; + + /* GI buffers */ + bool using_half_size_gi = false; + + RID uniform_set[RendererSceneRender::MAX_RENDER_VIEWS]; + RID scene_data_ubo; + + RID get_voxel_gi_buffer(); + + virtual void configure(RenderSceneBuffersRD *p_render_buffers) override{}; + virtual void free_data() override; + }; + + /* VOXEL GI API */ + + bool owns_voxel_gi(RID p_rid) { return voxel_gi_owner.owns(p_rid); }; + + virtual RID voxel_gi_allocate() override; + virtual void voxel_gi_free(RID p_voxel_gi) override; + virtual void voxel_gi_initialize(RID p_voxel_gi) override; + + virtual void voxel_gi_allocate_data(RID p_voxel_gi, const Transform3D &p_to_cell_xform, const AABB &p_aabb, const Vector3i &p_octree_size, const Vector<uint8_t> &p_octree_cells, const Vector<uint8_t> &p_data_cells, const Vector<uint8_t> &p_distance_field, const Vector<int> &p_level_counts) override; + + virtual AABB voxel_gi_get_bounds(RID p_voxel_gi) const override; + virtual Vector3i voxel_gi_get_octree_size(RID p_voxel_gi) const override; + virtual Vector<uint8_t> voxel_gi_get_octree_cells(RID p_voxel_gi) const override; + virtual Vector<uint8_t> voxel_gi_get_data_cells(RID p_voxel_gi) const override; + virtual Vector<uint8_t> voxel_gi_get_distance_field(RID p_voxel_gi) const override; + + virtual Vector<int> voxel_gi_get_level_counts(RID p_voxel_gi) const override; + virtual Transform3D voxel_gi_get_to_cell_xform(RID p_voxel_gi) const override; + + virtual void voxel_gi_set_dynamic_range(RID p_voxel_gi, float p_range) override; + virtual float voxel_gi_get_dynamic_range(RID p_voxel_gi) const override; + + virtual void voxel_gi_set_propagation(RID p_voxel_gi, float p_range) override; + virtual float voxel_gi_get_propagation(RID p_voxel_gi) const override; + + virtual void voxel_gi_set_energy(RID p_voxel_gi, float p_energy) override; + virtual float voxel_gi_get_energy(RID p_voxel_gi) const override; + + virtual void voxel_gi_set_baked_exposure_normalization(RID p_voxel_gi, float p_baked_exposure) override; + virtual float voxel_gi_get_baked_exposure_normalization(RID p_voxel_gi) const override; + + virtual void voxel_gi_set_bias(RID p_voxel_gi, float p_bias) override; + virtual float voxel_gi_get_bias(RID p_voxel_gi) const override; + + virtual void voxel_gi_set_normal_bias(RID p_voxel_gi, float p_range) override; + virtual float voxel_gi_get_normal_bias(RID p_voxel_gi) const override; + + virtual void voxel_gi_set_interior(RID p_voxel_gi, bool p_enable) override; + virtual bool voxel_gi_is_interior(RID p_voxel_gi) const override; + + virtual void voxel_gi_set_use_two_bounces(RID p_voxel_gi, bool p_enable) override; + virtual bool voxel_gi_is_using_two_bounces(RID p_voxel_gi) const override; + + virtual uint32_t voxel_gi_get_version(RID p_probe) const override; + uint32_t voxel_gi_get_data_version(RID p_probe); + + RID voxel_gi_get_octree_buffer(RID p_voxel_gi) const; + RID voxel_gi_get_data_buffer(RID p_voxel_gi) const; + + RID voxel_gi_get_sdf_texture(RID p_voxel_gi); + + Dependency *voxel_gi_get_dependency(RID p_voxel_gi) const; + + /* VOXEL_GI INSTANCE */ + + _FORCE_INLINE_ RID voxel_gi_instance_get_texture(RID p_probe) { + VoxelGIInstance *voxel_gi = voxel_gi_instance_owner.get_or_null(p_probe); + ERR_FAIL_COND_V(!voxel_gi, RID()); + return voxel_gi->texture; + }; + + _FORCE_INLINE_ void voxel_gi_instance_set_render_index(RID p_probe, uint32_t p_index) { + VoxelGIInstance *voxel_gi = voxel_gi_instance_owner.get_or_null(p_probe); + ERR_FAIL_NULL(voxel_gi); + + voxel_gi->render_index = p_index; + }; + + bool voxel_gi_instance_owns(RID p_rid) const { + return voxel_gi_instance_owner.owns(p_rid); + } + + void voxel_gi_instance_free(RID p_rid); + + RS::VoxelGIQuality voxel_gi_quality = RS::VOXEL_GI_QUALITY_LOW; + + /* SDFGI */ + + class SDFGI : public RenderBufferCustomDataRD { + GDCLASS(SDFGI, RenderBufferCustomDataRD) + + public: + enum { + MAX_CASCADES = 8, + CASCADE_SIZE = 128, + PROBE_DIVISOR = 16, + ANISOTROPY_SIZE = 6, + MAX_DYNAMIC_LIGHTS = 128, + MAX_STATIC_LIGHTS = 1024, + LIGHTPROBE_OCT_SIZE = 6, + SH_SIZE = 16 + }; + + struct Cascade { + struct UBO { + float offset[3]; + float to_cell; + int32_t probe_offset[3]; + uint32_t pad; + float pad2[4]; + }; + + //cascade blocks are full-size for volume (128^3), half size for albedo/emission + RID sdf_tex; + RID light_tex; + RID light_aniso_0_tex; + RID light_aniso_1_tex; + + RID light_data; + RID light_aniso_0_data; + RID light_aniso_1_data; + + struct SolidCell { // this struct is unused, but remains as reference for size + uint32_t position; + uint32_t albedo; + uint32_t static_light; + uint32_t static_light_aniso; + }; + + RID solid_cell_dispatch_buffer; //buffer for indirect compute dispatch + RID solid_cell_buffer; + + RID lightprobe_history_tex; + RID lightprobe_average_tex; + + float cell_size; + Vector3i position; + + static const Vector3i DIRTY_ALL; + Vector3i dirty_regions; //(0,0,0 is not dirty, negative is refresh from the end, DIRTY_ALL is refresh all. + + RID sdf_store_uniform_set; + RID sdf_direct_light_static_uniform_set; + RID sdf_direct_light_dynamic_uniform_set; + RID scroll_uniform_set; + RID scroll_occlusion_uniform_set; + RID integrate_uniform_set; + RID lights_buffer; + + float baked_exposure_normalization = 1.0; + + bool all_dynamic_lights_dirty = true; + }; + + // access to our containers + GI *gi = nullptr; + + // used for rendering (voxelization) + RID render_albedo; + RID render_emission; + RID render_emission_aniso; + RID render_occlusion[8]; + RID render_geom_facing; + + RID render_sdf[2]; + RID render_sdf_half[2]; + + // used for ping pong processing in cascades + RID sdf_initialize_uniform_set; + RID sdf_initialize_half_uniform_set; + RID jump_flood_uniform_set[2]; + RID jump_flood_half_uniform_set[2]; + RID sdf_upscale_uniform_set; + int upscale_jfa_uniform_set_index; + RID occlusion_uniform_set; + + uint32_t cascade_size = 128; + + LocalVector<Cascade> cascades; + + RID lightprobe_texture; + RID lightprobe_data; + RID occlusion_texture; + RID occlusion_data; + RID ambient_texture; //integrates with volumetric fog + + RID lightprobe_history_scroll; //used for scrolling lightprobes + RID lightprobe_average_scroll; //used for scrolling lightprobes + + uint32_t history_size = 0; + float solid_cell_ratio = 0; + uint32_t solid_cell_count = 0; + + int num_cascades = 6; + float min_cell_size = 0; + uint32_t probe_axis_count = 0; //amount of probes per axis, this is an odd number because it encloses endpoints + + RID debug_uniform_set[RendererSceneRender::MAX_RENDER_VIEWS]; + RID debug_probes_scene_data_ubo; + RID debug_probes_uniform_set; + RID cascades_ubo; + + bool uses_occlusion = false; + float bounce_feedback = 0.5; + bool reads_sky = true; + float energy = 1.0; + float normal_bias = 1.1; + float probe_bias = 1.1; + RS::EnvironmentSDFGIYScale y_scale_mode = RS::ENV_SDFGI_Y_SCALE_75_PERCENT; + + float y_mult = 1.0; + + uint32_t render_pass = 0; + + int32_t cascade_dynamic_light_count[SDFGI::MAX_CASCADES]; //used dynamically + RID integrate_sky_uniform_set; + + virtual void configure(RenderSceneBuffersRD *p_render_buffers) override{}; + virtual void free_data() override; + ~SDFGI(); + + void create(RID p_env, const Vector3 &p_world_position, uint32_t p_requested_history_size, GI *p_gi); + void update(RID p_env, const Vector3 &p_world_position); + void update_light(); + void update_probes(RID p_env, RendererRD::SkyRD::Sky *p_sky); + void store_probes(); + int get_pending_region_data(int p_region, Vector3i &r_local_offset, Vector3i &r_local_size, AABB &r_bounds) const; + void update_cascades(); + + void debug_draw(uint32_t p_view_count, const Projection *p_projections, const Transform3D &p_transform, int p_width, int p_height, RID p_render_target, RID p_texture, const Vector<RID> &p_texture_views); + void debug_probes(RID p_framebuffer, const uint32_t p_view_count, const Projection *p_camera_with_transforms, bool p_will_continue_color, bool p_will_continue_depth); + + void pre_process_gi(const Transform3D &p_transform, RenderDataRD *p_render_data); + void render_region(Ref<RenderSceneBuffersRD> p_render_buffers, int p_region, const PagedArray<RenderGeometryInstance *> &p_instances, float p_exposure_normalization); + void render_static_lights(RenderDataRD *p_render_data, Ref<RenderSceneBuffersRD> p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const PagedArray<RID> *p_positional_light_cull_result); + }; + + RS::EnvironmentSDFGIRayCount sdfgi_ray_count = RS::ENV_SDFGI_RAY_COUNT_16; + RS::EnvironmentSDFGIFramesToConverge sdfgi_frames_to_converge = RS::ENV_SDFGI_CONVERGE_IN_30_FRAMES; + RS::EnvironmentSDFGIFramesToUpdateLight sdfgi_frames_to_update_light = RS::ENV_SDFGI_UPDATE_LIGHT_IN_4_FRAMES; + + float sdfgi_solid_cell_ratio = 0.25; + Vector3 sdfgi_debug_probe_pos; + Vector3 sdfgi_debug_probe_dir; + bool sdfgi_debug_probe_enabled = false; + Vector3i sdfgi_debug_probe_index; + + /* SDFGI UPDATE */ + + int sdfgi_get_lightprobe_octahedron_size() const { return SDFGI::LIGHTPROBE_OCT_SIZE; } + + struct SDFGIData { + float grid_size[3]; + uint32_t max_cascades; + + uint32_t use_occlusion; + int32_t probe_axis_size; + float probe_to_uvw; + float normal_bias; + + float lightprobe_tex_pixel_size[3]; + float energy; + + float lightprobe_uv_offset[3]; + float y_mult; + + float occlusion_clamp[3]; + uint32_t pad3; + + float occlusion_renormalize[3]; + uint32_t pad4; + + float cascade_probe_size[3]; + uint32_t pad5; + + struct ProbeCascadeData { + float position[3]; //offset of (0,0,0) in world coordinates + float to_probe; // 1/bounds * grid_size + int32_t probe_world_offset[3]; + float to_cell; // 1/bounds * grid_size + float pad[3]; + float exposure_normalization; + }; + + ProbeCascadeData cascades[SDFGI::MAX_CASCADES]; + }; + + struct VoxelGIData { + float xform[16]; // 64 - 64 + + float bounds[3]; // 12 - 76 + float dynamic_range; // 4 - 80 + + float bias; // 4 - 84 + float normal_bias; // 4 - 88 + uint32_t blend_ambient; // 4 - 92 + uint32_t mipmaps; // 4 - 96 + + float pad[3]; // 12 - 108 + float exposure_normalization; // 4 - 112 + }; + + struct SceneData { + float inv_projection[2][16]; + float cam_transform[16]; + float eye_offset[2][4]; + + int32_t screen_size[2]; + float pad1; + float pad2; + }; + + struct PushConstant { + uint32_t max_voxel_gi_instances; + uint32_t high_quality_vct; + uint32_t orthogonal; + uint32_t view_index; + + float proj_info[4]; + + float z_near; + float z_far; + float pad2; + float pad3; + }; + + RID sdfgi_ubo; + + enum Mode { + MODE_VOXEL_GI, + MODE_SDFGI, + MODE_COMBINED, + MODE_MAX + }; + + enum ShaderSpecializations { + SHADER_SPECIALIZATION_HALF_RES = 1 << 0, + SHADER_SPECIALIZATION_USE_FULL_PROJECTION_MATRIX = 1 << 1, + SHADER_SPECIALIZATION_USE_VRS = 1 << 2, + SHADER_SPECIALIZATION_VARIATIONS = 8, + }; + + RID default_voxel_gi_buffer; + + bool half_resolution = false; + GiShaderRD shader; + RID shader_version; + RID pipelines[SHADER_SPECIALIZATION_VARIATIONS][MODE_MAX]; + + GI(); + ~GI(); + + void init(RendererRD::SkyRD *p_sky); + void free(); + + Ref<SDFGI> create_sdfgi(RID p_env, const Vector3 &p_world_position, uint32_t p_requested_history_size); + + void setup_voxel_gi_instances(RenderDataRD *p_render_data, Ref<RenderSceneBuffersRD> p_render_buffers, const Transform3D &p_transform, const PagedArray<RID> &p_voxel_gi_instances, uint32_t &r_voxel_gi_instances_used); + void process_gi(Ref<RenderSceneBuffersRD> p_render_buffers, const RID *p_normal_roughness_slices, RID p_voxel_gi_buffer, RID p_environment, uint32_t p_view_count, const Projection *p_projections, const Vector3 *p_eye_offsets, const Transform3D &p_cam_transform, const PagedArray<RID> &p_voxel_gi_instances); + + RID voxel_gi_instance_create(RID p_base); + void voxel_gi_instance_set_transform_to_data(RID p_probe, const Transform3D &p_xform); + bool voxel_gi_needs_update(RID p_probe) const; + void voxel_gi_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RenderGeometryInstance *> &p_dynamic_objects); + void debug_voxel_gi(RID p_voxel_gi, RD::DrawListID p_draw_list, RID p_framebuffer, const Projection &p_camera_with_transform, bool p_lighting, bool p_emission, float p_alpha); +}; + +} // namespace RendererRD + +#endif // GI_RD_H diff --git a/servers/rendering/renderer_rd/environment/sky.cpp b/servers/rendering/renderer_rd/environment/sky.cpp new file mode 100644 index 0000000000..6940276040 --- /dev/null +++ b/servers/rendering/renderer_rd/environment/sky.cpp @@ -0,0 +1,1958 @@ +/*************************************************************************/ +/* sky.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "sky.h" +#include "core/config/project_settings.h" +#include "core/math/math_defs.h" +#include "servers/rendering/renderer_rd/effects/copy_effects.h" +#include "servers/rendering/renderer_rd/renderer_compositor_rd.h" +#include "servers/rendering/renderer_rd/renderer_scene_render_rd.h" +#include "servers/rendering/renderer_rd/storage_rd/material_storage.h" +#include "servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.h" +#include "servers/rendering/renderer_rd/storage_rd/texture_storage.h" +#include "servers/rendering/rendering_server_default.h" +#include "servers/rendering/rendering_server_globals.h" + +using namespace RendererRD; + +//////////////////////////////////////////////////////////////////////////////// +// SKY SHADER + +void SkyRD::SkyShaderData::set_path_hint(const String &p_path) { + path = p_path; +} + +void SkyRD::SkyShaderData::set_code(const String &p_code) { + //compile + + code = p_code; + valid = false; + ubo_size = 0; + uniforms.clear(); + + if (code.is_empty()) { + return; //just invalid, but no error + } + + ShaderCompiler::GeneratedCode gen_code; + ShaderCompiler::IdentifierActions actions; + actions.entry_point_stages["sky"] = ShaderCompiler::STAGE_FRAGMENT; + + uses_time = false; + uses_half_res = false; + uses_quarter_res = false; + uses_position = false; + uses_light = false; + + actions.render_mode_flags["use_half_res_pass"] = &uses_half_res; + actions.render_mode_flags["use_quarter_res_pass"] = &uses_quarter_res; + + actions.usage_flag_pointers["TIME"] = &uses_time; + actions.usage_flag_pointers["POSITION"] = &uses_position; + actions.usage_flag_pointers["LIGHT0_ENABLED"] = &uses_light; + actions.usage_flag_pointers["LIGHT0_ENERGY"] = &uses_light; + actions.usage_flag_pointers["LIGHT0_DIRECTION"] = &uses_light; + actions.usage_flag_pointers["LIGHT0_COLOR"] = &uses_light; + actions.usage_flag_pointers["LIGHT0_SIZE"] = &uses_light; + actions.usage_flag_pointers["LIGHT1_ENABLED"] = &uses_light; + actions.usage_flag_pointers["LIGHT1_ENERGY"] = &uses_light; + actions.usage_flag_pointers["LIGHT1_DIRECTION"] = &uses_light; + actions.usage_flag_pointers["LIGHT1_COLOR"] = &uses_light; + actions.usage_flag_pointers["LIGHT1_SIZE"] = &uses_light; + actions.usage_flag_pointers["LIGHT2_ENABLED"] = &uses_light; + actions.usage_flag_pointers["LIGHT2_ENERGY"] = &uses_light; + actions.usage_flag_pointers["LIGHT2_DIRECTION"] = &uses_light; + actions.usage_flag_pointers["LIGHT2_COLOR"] = &uses_light; + actions.usage_flag_pointers["LIGHT2_SIZE"] = &uses_light; + actions.usage_flag_pointers["LIGHT3_ENABLED"] = &uses_light; + actions.usage_flag_pointers["LIGHT3_ENERGY"] = &uses_light; + actions.usage_flag_pointers["LIGHT3_DIRECTION"] = &uses_light; + actions.usage_flag_pointers["LIGHT3_COLOR"] = &uses_light; + actions.usage_flag_pointers["LIGHT3_SIZE"] = &uses_light; + + actions.uniforms = &uniforms; + + // !BAS! Contemplate making `SkyShader sky` accessible from this struct or even part of this struct. + RendererSceneRenderRD *scene_singleton = static_cast<RendererSceneRenderRD *>(RendererSceneRenderRD::singleton); + + Error err = scene_singleton->sky.sky_shader.compiler.compile(RS::SHADER_SKY, code, &actions, path, gen_code); + ERR_FAIL_COND_MSG(err != OK, "Shader compilation failed."); + + if (version.is_null()) { + version = scene_singleton->sky.sky_shader.shader.version_create(); + } + +#if 0 + print_line("**compiling shader:"); + print_line("**defines:\n"); + for (int i = 0; i < gen_code.defines.size(); i++) { + print_line(gen_code.defines[i]); + } + + HashMap<String, String>::Iterator el = gen_code.code.begin(); + while (el) { + print_line("\n**code " + el->key + ":\n" + el->value); + ++el; + } + + print_line("\n**uniforms:\n" + gen_code.uniforms); + print_line("\n**vertex_globals:\n" + gen_code.stage_globals[ShaderCompiler::STAGE_VERTEX]); + print_line("\n**fragment_globals:\n" + gen_code.stage_globals[ShaderCompiler::STAGE_FRAGMENT]); +#endif + + scene_singleton->sky.sky_shader.shader.version_set_code(version, gen_code.code, gen_code.uniforms, gen_code.stage_globals[ShaderCompiler::STAGE_VERTEX], gen_code.stage_globals[ShaderCompiler::STAGE_FRAGMENT], gen_code.defines); + ERR_FAIL_COND(!scene_singleton->sky.sky_shader.shader.version_is_valid(version)); + + ubo_size = gen_code.uniform_total_size; + ubo_offsets = gen_code.uniform_offsets; + texture_uniforms = gen_code.texture_uniforms; + + //update pipelines + + for (int i = 0; i < SKY_VERSION_MAX; i++) { + RD::PipelineDepthStencilState depth_stencil_state; + depth_stencil_state.enable_depth_test = true; + depth_stencil_state.depth_compare_operator = RD::COMPARE_OP_LESS_OR_EQUAL; + + if (scene_singleton->sky.sky_shader.shader.is_variant_enabled(i)) { + RID shader_variant = scene_singleton->sky.sky_shader.shader.version_get_shader(version, i); + pipelines[i].setup(shader_variant, RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), depth_stencil_state, RD::PipelineColorBlendState::create_disabled(), 0); + } else { + pipelines[i].clear(); + } + } + + valid = true; +} + +void SkyRD::SkyShaderData::set_default_texture_parameter(const StringName &p_name, RID p_texture, int p_index) { + if (!p_texture.is_valid()) { + if (default_texture_params.has(p_name) && default_texture_params[p_name].has(p_index)) { + default_texture_params[p_name].erase(p_index); + + if (default_texture_params[p_name].is_empty()) { + default_texture_params.erase(p_name); + } + } + } else { + if (!default_texture_params.has(p_name)) { + default_texture_params[p_name] = HashMap<int, RID>(); + } + default_texture_params[p_name][p_index] = p_texture; + } +} + +void SkyRD::SkyShaderData::get_shader_uniform_list(List<PropertyInfo> *p_param_list) const { + HashMap<int, StringName> order; + + for (const KeyValue<StringName, ShaderLanguage::ShaderNode::Uniform> &E : uniforms) { + if (E.value.scope == ShaderLanguage::ShaderNode::Uniform::SCOPE_GLOBAL || E.value.scope == ShaderLanguage::ShaderNode::Uniform::SCOPE_INSTANCE) { + continue; + } + + if (E.value.texture_order >= 0) { + order[E.value.texture_order + 100000] = E.key; + } else { + order[E.value.order] = E.key; + } + } + String last_group; + for (const KeyValue<int, StringName> &E : order) { + String group = uniforms[E.value].group; + if (!uniforms[E.value].subgroup.is_empty()) { + group += "::" + uniforms[E.value].subgroup; + } + + if (group != last_group) { + PropertyInfo pi; + pi.usage = PROPERTY_USAGE_GROUP; + pi.name = group; + p_param_list->push_back(pi); + + last_group = group; + } + + PropertyInfo pi = ShaderLanguage::uniform_to_property_info(uniforms[E.value]); + pi.name = E.value; + p_param_list->push_back(pi); + } +} + +void SkyRD::SkyShaderData::get_instance_param_list(List<RendererMaterialStorage::InstanceShaderParam> *p_param_list) const { + for (const KeyValue<StringName, ShaderLanguage::ShaderNode::Uniform> &E : uniforms) { + if (E.value.scope != ShaderLanguage::ShaderNode::Uniform::SCOPE_INSTANCE) { + continue; + } + + RendererMaterialStorage::InstanceShaderParam p; + p.info = ShaderLanguage::uniform_to_property_info(E.value); + p.info.name = E.key; //supply name + p.index = E.value.instance_index; + p.default_value = ShaderLanguage::constant_value_to_variant(E.value.default_value, E.value.type, E.value.array_size, E.value.hint); + p_param_list->push_back(p); + } +} + +bool SkyRD::SkyShaderData::is_parameter_texture(const StringName &p_param) const { + if (!uniforms.has(p_param)) { + return false; + } + + return uniforms[p_param].texture_order >= 0; +} + +bool SkyRD::SkyShaderData::is_animated() const { + return false; +} + +bool SkyRD::SkyShaderData::casts_shadows() const { + return false; +} + +Variant SkyRD::SkyShaderData::get_default_parameter(const StringName &p_parameter) const { + if (uniforms.has(p_parameter)) { + ShaderLanguage::ShaderNode::Uniform uniform = uniforms[p_parameter]; + Vector<ShaderLanguage::ConstantNode::Value> default_value = uniform.default_value; + return ShaderLanguage::constant_value_to_variant(default_value, uniform.type, uniform.array_size, uniform.hint); + } + return Variant(); +} + +RS::ShaderNativeSourceCode SkyRD::SkyShaderData::get_native_source_code() const { + RendererSceneRenderRD *scene_singleton = static_cast<RendererSceneRenderRD *>(RendererSceneRenderRD::singleton); + + return scene_singleton->sky.sky_shader.shader.version_get_native_source_code(version); +} + +SkyRD::SkyShaderData::~SkyShaderData() { + RendererSceneRenderRD *scene_singleton = static_cast<RendererSceneRenderRD *>(RendererSceneRenderRD::singleton); + ERR_FAIL_COND(!scene_singleton); + //pipeline variants will clear themselves if shader is gone + if (version.is_valid()) { + scene_singleton->sky.sky_shader.shader.version_free(version); + } +} + +//////////////////////////////////////////////////////////////////////////////// +// Sky material + +bool SkyRD::SkyMaterialData::update_parameters(const HashMap<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { + RendererSceneRenderRD *scene_singleton = static_cast<RendererSceneRenderRD *>(RendererSceneRenderRD::singleton); + + uniform_set_updated = true; + + return update_parameters_uniform_set(p_parameters, p_uniform_dirty, p_textures_dirty, shader_data->uniforms, shader_data->ubo_offsets.ptr(), shader_data->texture_uniforms, shader_data->default_texture_params, shader_data->ubo_size, uniform_set, scene_singleton->sky.sky_shader.shader.version_get_shader(shader_data->version, 0), SKY_SET_MATERIAL, true); +} + +SkyRD::SkyMaterialData::~SkyMaterialData() { + free_parameters_uniform_set(uniform_set); +} + +//////////////////////////////////////////////////////////////////////////////// +// Render sky + +static _FORCE_INLINE_ void store_transform_3x3(const Basis &p_basis, float *p_array) { + p_array[0] = p_basis.rows[0][0]; + p_array[1] = p_basis.rows[1][0]; + p_array[2] = p_basis.rows[2][0]; + p_array[3] = 0; + p_array[4] = p_basis.rows[0][1]; + p_array[5] = p_basis.rows[1][1]; + p_array[6] = p_basis.rows[2][1]; + p_array[7] = 0; + p_array[8] = p_basis.rows[0][2]; + p_array[9] = p_basis.rows[1][2]; + p_array[10] = p_basis.rows[2][2]; + p_array[11] = 0; +} + +void SkyRD::_render_sky(RD::DrawListID p_list, float p_time, RID p_fb, PipelineCacheRD *p_pipeline, RID p_uniform_set, RID p_texture_set, uint32_t p_view_count, const Projection *p_projections, const Basis &p_orientation, const Vector3 &p_position, float p_luminance_multiplier) { + SkyPushConstant sky_push_constant; + + memset(&sky_push_constant, 0, sizeof(SkyPushConstant)); + + for (uint32_t v = 0; v < p_view_count; v++) { + // We only need key components of our projection matrix + sky_push_constant.projections[v][0] = p_projections[v].columns[2][0]; + sky_push_constant.projections[v][1] = p_projections[v].columns[0][0]; + sky_push_constant.projections[v][2] = p_projections[v].columns[2][1]; + sky_push_constant.projections[v][3] = p_projections[v].columns[1][1]; + } + sky_push_constant.position[0] = p_position.x; + sky_push_constant.position[1] = p_position.y; + sky_push_constant.position[2] = p_position.z; + sky_push_constant.time = p_time; + sky_push_constant.luminance_multiplier = p_luminance_multiplier; + store_transform_3x3(p_orientation, sky_push_constant.orientation); + + RenderingDevice::FramebufferFormatID fb_format = RD::get_singleton()->framebuffer_get_format(p_fb); + + RD::DrawListID draw_list = p_list; + + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, p_pipeline->get_render_pipeline(RD::INVALID_ID, fb_format, false, RD::get_singleton()->draw_list_get_current_pass())); + + // Update uniform sets. + { + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, sky_scene_state.uniform_set, 0); + if (p_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(p_uniform_set)) { // Material may not have a uniform set. + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, p_uniform_set, 1); + } + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, p_texture_set, 2); + // Fog uniform set can be invalidated before drawing, so validate at draw time + if (sky_scene_state.fog_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(sky_scene_state.fog_uniform_set)) { + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, sky_scene_state.fog_uniform_set, 3); + } else { + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, sky_scene_state.default_fog_uniform_set, 3); + } + } + + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &sky_push_constant, sizeof(SkyPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); +} + +//////////////////////////////////////////////////////////////////////////////// +// ReflectionData + +void SkyRD::ReflectionData::clear_reflection_data() { + layers.clear(); + radiance_base_cubemap = RID(); + if (downsampled_radiance_cubemap.is_valid()) { + RD::get_singleton()->free(downsampled_radiance_cubemap); + } + downsampled_radiance_cubemap = RID(); + downsampled_layer.mipmaps.clear(); + coefficient_buffer = RID(); +} + +void SkyRD::ReflectionData::update_reflection_data(int p_size, int p_mipmaps, bool p_use_array, RID p_base_cube, int p_base_layer, bool p_low_quality, int p_roughness_layers, RD::DataFormat p_texture_format) { + //recreate radiance and all data + + int mipmaps = p_mipmaps; + uint32_t w = p_size, h = p_size; + + EffectsRD *effects = RendererCompositorRD::singleton->get_effects(); + ERR_FAIL_NULL_MSG(effects, "Effects haven't been initialised"); + bool prefer_raster_effects = effects->get_prefer_raster_effects(); + + if (p_use_array) { + int num_layers = p_low_quality ? 8 : p_roughness_layers; + + for (int i = 0; i < num_layers; i++) { + ReflectionData::Layer layer; + uint32_t mmw = w; + uint32_t mmh = h; + layer.mipmaps.resize(mipmaps); + layer.views.resize(mipmaps); + for (int j = 0; j < mipmaps; j++) { + ReflectionData::Layer::Mipmap &mm = layer.mipmaps.write[j]; + mm.size.width = mmw; + mm.size.height = mmh; + for (int k = 0; k < 6; k++) { + mm.views[k] = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_base_cube, p_base_layer + i * 6 + k, j); + Vector<RID> fbtex; + fbtex.push_back(mm.views[k]); + mm.framebuffers[k] = RD::get_singleton()->framebuffer_create(fbtex); + } + + layer.views.write[j] = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_base_cube, p_base_layer + i * 6, j, 1, RD::TEXTURE_SLICE_CUBEMAP); + + mmw = MAX(1u, mmw >> 1); + mmh = MAX(1u, mmh >> 1); + } + + layers.push_back(layer); + } + + } else { + mipmaps = p_low_quality ? 8 : mipmaps; + //regular cubemap, lower quality (aliasing, less memory) + ReflectionData::Layer layer; + uint32_t mmw = w; + uint32_t mmh = h; + layer.mipmaps.resize(mipmaps); + layer.views.resize(mipmaps); + for (int j = 0; j < mipmaps; j++) { + ReflectionData::Layer::Mipmap &mm = layer.mipmaps.write[j]; + mm.size.width = mmw; + mm.size.height = mmh; + for (int k = 0; k < 6; k++) { + mm.views[k] = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_base_cube, p_base_layer + k, j); + Vector<RID> fbtex; + fbtex.push_back(mm.views[k]); + mm.framebuffers[k] = RD::get_singleton()->framebuffer_create(fbtex); + } + + layer.views.write[j] = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_base_cube, p_base_layer, j, 1, RD::TEXTURE_SLICE_CUBEMAP); + + mmw = MAX(1u, mmw >> 1); + mmh = MAX(1u, mmh >> 1); + } + + layers.push_back(layer); + } + + radiance_base_cubemap = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_base_cube, p_base_layer, 0, 1, RD::TEXTURE_SLICE_CUBEMAP); + RD::get_singleton()->set_resource_name(radiance_base_cubemap, "radiance base cubemap"); + + RD::TextureFormat tf; + tf.format = p_texture_format; + tf.width = p_low_quality ? 64 : p_size >> 1; // Always 64x64 when using REALTIME. + tf.height = p_low_quality ? 64 : p_size >> 1; + tf.texture_type = RD::TEXTURE_TYPE_CUBE; + tf.array_layers = 6; + tf.mipmaps = p_low_quality ? 7 : mipmaps - 1; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + + downsampled_radiance_cubemap = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(downsampled_radiance_cubemap, "downsampled radiance cubemap"); + { + uint32_t mmw = tf.width; + uint32_t mmh = tf.height; + downsampled_layer.mipmaps.resize(tf.mipmaps); + for (int j = 0; j < downsampled_layer.mipmaps.size(); j++) { + ReflectionData::DownsampleLayer::Mipmap &mm = downsampled_layer.mipmaps.write[j]; + mm.size.width = mmw; + mm.size.height = mmh; + mm.view = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), downsampled_radiance_cubemap, 0, j, 1, RD::TEXTURE_SLICE_CUBEMAP); + RD::get_singleton()->set_resource_name(mm.view, "Downsampled Radiance Cubemap Mip " + itos(j) + " "); + if (prefer_raster_effects) { + // we need a framebuffer for each side of our cubemap + + for (int k = 0; k < 6; k++) { + mm.views[k] = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), downsampled_radiance_cubemap, k, j); + RD::get_singleton()->set_resource_name(mm.view, "Downsampled Radiance Cubemap Mip: " + itos(j) + " Face: " + itos(k) + " "); + Vector<RID> fbtex; + fbtex.push_back(mm.views[k]); + mm.framebuffers[k] = RD::get_singleton()->framebuffer_create(fbtex); + } + } + + mmw = MAX(1u, mmw >> 1); + mmh = MAX(1u, mmh >> 1); + } + } +} + +void SkyRD::ReflectionData::create_reflection_fast_filter(bool p_use_arrays) { + RendererRD::CopyEffects *copy_effects = RendererRD::CopyEffects::get_singleton(); + ERR_FAIL_NULL_MSG(copy_effects, "Effects haven't been initialised"); + bool prefer_raster_effects = copy_effects->get_prefer_raster_effects(); + + if (prefer_raster_effects) { + RD::get_singleton()->draw_command_begin_label("Downsample radiance map"); + for (int k = 0; k < 6; k++) { + copy_effects->cubemap_downsample_raster(radiance_base_cubemap, downsampled_layer.mipmaps[0].framebuffers[k], k, downsampled_layer.mipmaps[0].size); + } + + for (int i = 1; i < downsampled_layer.mipmaps.size(); i++) { + for (int k = 0; k < 6; k++) { + copy_effects->cubemap_downsample_raster(downsampled_layer.mipmaps[i - 1].view, downsampled_layer.mipmaps[i].framebuffers[k], k, downsampled_layer.mipmaps[i].size); + } + } + RD::get_singleton()->draw_command_end_label(); // Downsample Radiance + + if (p_use_arrays) { + RD::get_singleton()->draw_command_begin_label("filter radiance map into array heads"); + for (int i = 0; i < layers.size(); i++) { + for (int k = 0; k < 6; k++) { + copy_effects->cubemap_filter_raster(downsampled_radiance_cubemap, layers[i].mipmaps[0].framebuffers[k], k, i); + } + } + } else { + RD::get_singleton()->draw_command_begin_label("filter radiance map into mipmaps directly"); + for (int j = 0; j < layers[0].mipmaps.size(); j++) { + for (int k = 0; k < 6; k++) { + copy_effects->cubemap_filter_raster(downsampled_radiance_cubemap, layers[0].mipmaps[j].framebuffers[k], k, j); + } + } + } + RD::get_singleton()->draw_command_end_label(); // Filter radiance + } else { + RD::get_singleton()->draw_command_begin_label("Downsample radiance map"); + copy_effects->cubemap_downsample(radiance_base_cubemap, downsampled_layer.mipmaps[0].view, downsampled_layer.mipmaps[0].size); + + for (int i = 1; i < downsampled_layer.mipmaps.size(); i++) { + copy_effects->cubemap_downsample(downsampled_layer.mipmaps[i - 1].view, downsampled_layer.mipmaps[i].view, downsampled_layer.mipmaps[i].size); + } + RD::get_singleton()->draw_command_end_label(); // Downsample Radiance + Vector<RID> views; + if (p_use_arrays) { + for (int i = 1; i < layers.size(); i++) { + views.push_back(layers[i].views[0]); + } + } else { + for (int i = 1; i < layers[0].views.size(); i++) { + views.push_back(layers[0].views[i]); + } + } + RD::get_singleton()->draw_command_begin_label("Fast filter radiance"); + copy_effects->cubemap_filter(downsampled_radiance_cubemap, views, p_use_arrays); + RD::get_singleton()->draw_command_end_label(); // Filter radiance + } +} + +void SkyRD::ReflectionData::create_reflection_importance_sample(bool p_use_arrays, int p_cube_side, int p_base_layer, uint32_t p_sky_ggx_samples_quality) { + RendererRD::CopyEffects *copy_effects = RendererRD::CopyEffects::get_singleton(); + ERR_FAIL_NULL_MSG(copy_effects, "Effects haven't been initialised"); + bool prefer_raster_effects = copy_effects->get_prefer_raster_effects(); + + if (prefer_raster_effects) { + if (p_base_layer == 1) { + RD::get_singleton()->draw_command_begin_label("Downsample radiance map"); + for (int k = 0; k < 6; k++) { + copy_effects->cubemap_downsample_raster(radiance_base_cubemap, downsampled_layer.mipmaps[0].framebuffers[k], k, downsampled_layer.mipmaps[0].size); + } + + for (int i = 1; i < downsampled_layer.mipmaps.size(); i++) { + for (int k = 0; k < 6; k++) { + copy_effects->cubemap_downsample_raster(downsampled_layer.mipmaps[i - 1].view, downsampled_layer.mipmaps[i].framebuffers[k], k, downsampled_layer.mipmaps[i].size); + } + } + RD::get_singleton()->draw_command_end_label(); // Downsample Radiance + } + + RD::get_singleton()->draw_command_begin_label("High Quality filter radiance"); + if (p_use_arrays) { + for (int k = 0; k < 6; k++) { + copy_effects->cubemap_roughness_raster( + downsampled_radiance_cubemap, + layers[p_base_layer].mipmaps[0].framebuffers[k], + k, + p_sky_ggx_samples_quality, + float(p_base_layer) / (layers.size() - 1.0), + layers[p_base_layer].mipmaps[0].size.x); + } + } else { + for (int k = 0; k < 6; k++) { + copy_effects->cubemap_roughness_raster( + downsampled_radiance_cubemap, + layers[0].mipmaps[p_base_layer].framebuffers[k], + k, + p_sky_ggx_samples_quality, + float(p_base_layer) / (layers[0].mipmaps.size() - 1.0), + layers[0].mipmaps[p_base_layer].size.x); + } + } + } else { + if (p_base_layer == 1) { + RD::get_singleton()->draw_command_begin_label("Downsample radiance map"); + copy_effects->cubemap_downsample(radiance_base_cubemap, downsampled_layer.mipmaps[0].view, downsampled_layer.mipmaps[0].size); + + for (int i = 1; i < downsampled_layer.mipmaps.size(); i++) { + copy_effects->cubemap_downsample(downsampled_layer.mipmaps[i - 1].view, downsampled_layer.mipmaps[i].view, downsampled_layer.mipmaps[i].size); + } + RD::get_singleton()->draw_command_end_label(); // Downsample Radiance + } + + RD::get_singleton()->draw_command_begin_label("High Quality filter radiance"); + if (p_use_arrays) { + copy_effects->cubemap_roughness(downsampled_radiance_cubemap, layers[p_base_layer].views[0], p_cube_side, p_sky_ggx_samples_quality, float(p_base_layer) / (layers.size() - 1.0), layers[p_base_layer].mipmaps[0].size.x); + } else { + copy_effects->cubemap_roughness( + downsampled_radiance_cubemap, + layers[0].views[p_base_layer], + p_cube_side, + p_sky_ggx_samples_quality, + float(p_base_layer) / (layers[0].mipmaps.size() - 1.0), + layers[0].mipmaps[p_base_layer].size.x); + } + } + RD::get_singleton()->draw_command_end_label(); // Filter radiance +} + +void SkyRD::ReflectionData::update_reflection_mipmaps(int p_start, int p_end) { + RendererRD::CopyEffects *copy_effects = RendererRD::CopyEffects::get_singleton(); + ERR_FAIL_NULL_MSG(copy_effects, "Effects haven't been initialised"); + bool prefer_raster_effects = copy_effects->get_prefer_raster_effects(); + + RD::get_singleton()->draw_command_begin_label("Update Radiance Cubemap Array Mipmaps"); + for (int i = p_start; i < p_end; i++) { + for (int j = 0; j < layers[i].views.size() - 1; j++) { + RID view = layers[i].views[j]; + Size2i size = layers[i].mipmaps[j + 1].size; + if (prefer_raster_effects) { + for (int k = 0; k < 6; k++) { + RID framebuffer = layers[i].mipmaps[j + 1].framebuffers[k]; + copy_effects->cubemap_downsample_raster(view, framebuffer, k, size); + } + } else { + RID texture = layers[i].views[j + 1]; + copy_effects->cubemap_downsample(view, texture, size); + } + } + } + RD::get_singleton()->draw_command_end_label(); +} + +//////////////////////////////////////////////////////////////////////////////// +// SkyRD::Sky + +void SkyRD::Sky::free() { + if (radiance.is_valid()) { + RD::get_singleton()->free(radiance); + radiance = RID(); + } + reflection.clear_reflection_data(); + + if (uniform_buffer.is_valid()) { + RD::get_singleton()->free(uniform_buffer); + uniform_buffer = RID(); + } + + if (half_res_pass.is_valid()) { + RD::get_singleton()->free(half_res_pass); + half_res_pass = RID(); + } + + if (quarter_res_pass.is_valid()) { + RD::get_singleton()->free(quarter_res_pass); + quarter_res_pass = RID(); + } + + if (material.is_valid()) { + RSG::material_storage->material_free(material); + material = RID(); + } +} + +RID SkyRD::Sky::get_textures(SkyTextureSetVersion p_version, RID p_default_shader_rd) { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + + if (texture_uniform_sets[p_version].is_valid() && RD::get_singleton()->uniform_set_is_valid(texture_uniform_sets[p_version])) { + return texture_uniform_sets[p_version]; + } + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 0; + if (radiance.is_valid() && p_version <= SKY_TEXTURE_SET_QUARTER_RES) { + u.append_id(radiance); + } else { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK)); + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 1; // half res + if (half_res_pass.is_valid() && p_version != SKY_TEXTURE_SET_HALF_RES && p_version != SKY_TEXTURE_SET_CUBEMAP_HALF_RES) { + if (p_version >= SKY_TEXTURE_SET_CUBEMAP) { + u.append_id(reflection.layers[0].views[1]); + } else { + u.append_id(half_res_pass); + } + } else { + if (p_version < SKY_TEXTURE_SET_CUBEMAP) { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_WHITE)); + } else { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK)); + } + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 2; // quarter res + if (quarter_res_pass.is_valid() && p_version != SKY_TEXTURE_SET_QUARTER_RES && p_version != SKY_TEXTURE_SET_CUBEMAP_QUARTER_RES) { + if (p_version >= SKY_TEXTURE_SET_CUBEMAP) { + u.append_id(reflection.layers[0].views[2]); + } else { + u.append_id(quarter_res_pass); + } + } else { + if (p_version < SKY_TEXTURE_SET_CUBEMAP) { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_WHITE)); + } else { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK)); + } + } + uniforms.push_back(u); + } + + texture_uniform_sets[p_version] = RD::get_singleton()->uniform_set_create(uniforms, p_default_shader_rd, SKY_SET_TEXTURES); + return texture_uniform_sets[p_version]; +} + +bool SkyRD::Sky::set_radiance_size(int p_radiance_size) { + ERR_FAIL_COND_V(p_radiance_size < 32 || p_radiance_size > 2048, false); + if (radiance_size == p_radiance_size) { + return false; + } + radiance_size = p_radiance_size; + + if (mode == RS::SKY_MODE_REALTIME && radiance_size != 256) { + WARN_PRINT("Realtime Skies can only use a radiance size of 256. Radiance size will be set to 256 internally."); + radiance_size = 256; + } + + if (radiance.is_valid()) { + RD::get_singleton()->free(radiance); + radiance = RID(); + } + reflection.clear_reflection_data(); + + return true; +} + +bool SkyRD::Sky::set_mode(RS::SkyMode p_mode) { + if (mode == p_mode) { + return false; + } + + mode = p_mode; + + if (mode == RS::SKY_MODE_REALTIME && radiance_size != 256) { + WARN_PRINT("Realtime Skies can only use a radiance size of 256. Radiance size will be set to 256 internally."); + set_radiance_size(256); + } + + if (radiance.is_valid()) { + RD::get_singleton()->free(radiance); + radiance = RID(); + } + reflection.clear_reflection_data(); + + return true; +} + +bool SkyRD::Sky::set_material(RID p_material) { + if (material == p_material) { + return false; + } + + material = p_material; + return true; +} + +Ref<Image> SkyRD::Sky::bake_panorama(float p_energy, int p_roughness_layers, const Size2i &p_size) { + if (radiance.is_valid()) { + RendererRD::CopyEffects *copy_effects = RendererRD::CopyEffects::get_singleton(); + + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R32G32B32A32_SFLOAT; // Could be RGBA16 + tf.width = p_size.width; + tf.height = p_size.height; + tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT; + + RID rad_tex = RD::get_singleton()->texture_create(tf, RD::TextureView()); + copy_effects->copy_cubemap_to_panorama(radiance, rad_tex, p_size, p_roughness_layers, reflection.layers.size() > 1); + Vector<uint8_t> data = RD::get_singleton()->texture_get_data(rad_tex, 0); + RD::get_singleton()->free(rad_tex); + + Ref<Image> img = Image::create_from_data(p_size.width, p_size.height, false, Image::FORMAT_RGBAF, data); + for (int i = 0; i < p_size.width; i++) { + for (int j = 0; j < p_size.height; j++) { + Color c = img->get_pixel(i, j); + c.r *= p_energy; + c.g *= p_energy; + c.b *= p_energy; + img->set_pixel(i, j, c); + } + } + return img; + } + + return Ref<Image>(); +} + +//////////////////////////////////////////////////////////////////////////////// +// SkyRD + +RendererRD::MaterialStorage::ShaderData *SkyRD::_create_sky_shader_func() { + SkyShaderData *shader_data = memnew(SkyShaderData); + return shader_data; +} + +RendererRD::MaterialStorage::ShaderData *SkyRD::_create_sky_shader_funcs() { + // !BAS! Why isn't _create_sky_shader_func not just static too? + return static_cast<RendererSceneRenderRD *>(RendererSceneRenderRD::singleton)->sky._create_sky_shader_func(); +}; + +RendererRD::MaterialStorage::MaterialData *SkyRD::_create_sky_material_func(SkyShaderData *p_shader) { + SkyMaterialData *material_data = memnew(SkyMaterialData); + material_data->shader_data = p_shader; + //update will happen later anyway so do nothing. + return material_data; +} + +RendererRD::MaterialStorage::MaterialData *SkyRD::_create_sky_material_funcs(RendererRD::MaterialStorage::ShaderData *p_shader) { + // !BAS! same here, we could just make _create_sky_material_func static? + return static_cast<RendererSceneRenderRD *>(RendererSceneRenderRD::singleton)->sky._create_sky_material_func(static_cast<SkyShaderData *>(p_shader)); +}; + +SkyRD::SkyRD() { + roughness_layers = GLOBAL_GET("rendering/reflections/sky_reflections/roughness_layers"); + sky_ggx_samples_quality = GLOBAL_GET("rendering/reflections/sky_reflections/ggx_samples"); + sky_use_cubemap_array = GLOBAL_GET("rendering/reflections/sky_reflections/texture_array_reflections"); +} + +void SkyRD::init() { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + + { + // Start with the directional lights for the sky + sky_scene_state.max_directional_lights = 4; + uint32_t directional_light_buffer_size = sky_scene_state.max_directional_lights * sizeof(SkyDirectionalLightData); + sky_scene_state.directional_lights = memnew_arr(SkyDirectionalLightData, sky_scene_state.max_directional_lights); + sky_scene_state.last_frame_directional_lights = memnew_arr(SkyDirectionalLightData, sky_scene_state.max_directional_lights); + sky_scene_state.last_frame_directional_light_count = sky_scene_state.max_directional_lights + 1; + sky_scene_state.directional_light_buffer = RD::get_singleton()->uniform_buffer_create(directional_light_buffer_size); + + String defines = "\n#define MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS " + itos(sky_scene_state.max_directional_lights) + "\n"; + + // Initialize sky + Vector<String> sky_modes; + sky_modes.push_back(""); // Full size + sky_modes.push_back("\n#define USE_HALF_RES_PASS\n"); // Half Res + sky_modes.push_back("\n#define USE_QUARTER_RES_PASS\n"); // Quarter res + sky_modes.push_back("\n#define USE_CUBEMAP_PASS\n"); // Cubemap + sky_modes.push_back("\n#define USE_CUBEMAP_PASS\n#define USE_HALF_RES_PASS\n"); // Half Res Cubemap + sky_modes.push_back("\n#define USE_CUBEMAP_PASS\n#define USE_QUARTER_RES_PASS\n"); // Quarter res Cubemap + + sky_modes.push_back("\n#define USE_MULTIVIEW\n"); // Full size multiview + sky_modes.push_back("\n#define USE_HALF_RES_PASS\n#define USE_MULTIVIEW\n"); // Half Res multiview + sky_modes.push_back("\n#define USE_QUARTER_RES_PASS\n#define USE_MULTIVIEW\n"); // Quarter res multiview + + sky_shader.shader.initialize(sky_modes, defines); + + if (!RendererCompositorRD::singleton->is_xr_enabled()) { + sky_shader.shader.set_variant_enabled(SKY_VERSION_BACKGROUND_MULTIVIEW, false); + sky_shader.shader.set_variant_enabled(SKY_VERSION_HALF_RES_MULTIVIEW, false); + sky_shader.shader.set_variant_enabled(SKY_VERSION_QUARTER_RES_MULTIVIEW, false); + } + } + + // register our shader funds + material_storage->shader_set_data_request_function(RendererRD::MaterialStorage::SHADER_TYPE_SKY, _create_sky_shader_funcs); + material_storage->material_set_data_request_function(RendererRD::MaterialStorage::SHADER_TYPE_SKY, _create_sky_material_funcs); + + { + ShaderCompiler::DefaultIdentifierActions actions; + + actions.renames["COLOR"] = "color"; + actions.renames["ALPHA"] = "alpha"; + actions.renames["EYEDIR"] = "cube_normal"; + actions.renames["POSITION"] = "params.position"; + actions.renames["SKY_COORDS"] = "panorama_coords"; + actions.renames["SCREEN_UV"] = "uv"; + actions.renames["FRAGCOORD"] = "gl_FragCoord"; + actions.renames["TIME"] = "params.time"; + actions.renames["PI"] = _MKSTR(Math_PI); + actions.renames["TAU"] = _MKSTR(Math_TAU); + actions.renames["E"] = _MKSTR(Math_E); + actions.renames["HALF_RES_COLOR"] = "half_res_color"; + actions.renames["QUARTER_RES_COLOR"] = "quarter_res_color"; + actions.renames["RADIANCE"] = "radiance"; + actions.renames["FOG"] = "custom_fog"; + actions.renames["LIGHT0_ENABLED"] = "directional_lights.data[0].enabled"; + actions.renames["LIGHT0_DIRECTION"] = "directional_lights.data[0].direction_energy.xyz"; + actions.renames["LIGHT0_ENERGY"] = "directional_lights.data[0].direction_energy.w"; + actions.renames["LIGHT0_COLOR"] = "directional_lights.data[0].color_size.xyz"; + actions.renames["LIGHT0_SIZE"] = "directional_lights.data[0].color_size.w"; + actions.renames["LIGHT1_ENABLED"] = "directional_lights.data[1].enabled"; + actions.renames["LIGHT1_DIRECTION"] = "directional_lights.data[1].direction_energy.xyz"; + actions.renames["LIGHT1_ENERGY"] = "directional_lights.data[1].direction_energy.w"; + actions.renames["LIGHT1_COLOR"] = "directional_lights.data[1].color_size.xyz"; + actions.renames["LIGHT1_SIZE"] = "directional_lights.data[1].color_size.w"; + actions.renames["LIGHT2_ENABLED"] = "directional_lights.data[2].enabled"; + actions.renames["LIGHT2_DIRECTION"] = "directional_lights.data[2].direction_energy.xyz"; + actions.renames["LIGHT2_ENERGY"] = "directional_lights.data[2].direction_energy.w"; + actions.renames["LIGHT2_COLOR"] = "directional_lights.data[2].color_size.xyz"; + actions.renames["LIGHT2_SIZE"] = "directional_lights.data[2].color_size.w"; + actions.renames["LIGHT3_ENABLED"] = "directional_lights.data[3].enabled"; + actions.renames["LIGHT3_DIRECTION"] = "directional_lights.data[3].direction_energy.xyz"; + actions.renames["LIGHT3_ENERGY"] = "directional_lights.data[3].direction_energy.w"; + actions.renames["LIGHT3_COLOR"] = "directional_lights.data[3].color_size.xyz"; + actions.renames["LIGHT3_SIZE"] = "directional_lights.data[3].color_size.w"; + actions.renames["AT_CUBEMAP_PASS"] = "AT_CUBEMAP_PASS"; + actions.renames["AT_HALF_RES_PASS"] = "AT_HALF_RES_PASS"; + actions.renames["AT_QUARTER_RES_PASS"] = "AT_QUARTER_RES_PASS"; + actions.custom_samplers["RADIANCE"] = "material_samplers[3]"; + actions.usage_defines["HALF_RES_COLOR"] = "\n#define USES_HALF_RES_COLOR\n"; + actions.usage_defines["QUARTER_RES_COLOR"] = "\n#define USES_QUARTER_RES_COLOR\n"; + actions.render_mode_defines["disable_fog"] = "#define DISABLE_FOG\n"; + actions.render_mode_defines["use_debanding"] = "#define USE_DEBANDING\n"; + + actions.sampler_array_name = "material_samplers"; + actions.base_texture_binding_index = 1; + actions.texture_layout_set = 1; + actions.base_uniform_string = "material."; + actions.base_varying_index = 10; + + actions.default_filter = ShaderLanguage::FILTER_LINEAR_MIPMAP; + actions.default_repeat = ShaderLanguage::REPEAT_ENABLE; + actions.global_buffer_array_variable = "global_shader_uniforms.data"; + + sky_shader.compiler.initialize(actions); + } + + { + // default material and shader for sky shader + sky_shader.default_shader = material_storage->shader_allocate(); + material_storage->shader_initialize(sky_shader.default_shader); + + material_storage->shader_set_code(sky_shader.default_shader, R"( +// Default sky shader. + +shader_type sky; + +void sky() { + COLOR = vec3(0.0); +} +)"); + + sky_shader.default_material = material_storage->material_allocate(); + material_storage->material_initialize(sky_shader.default_material); + + material_storage->material_set_shader(sky_shader.default_material, sky_shader.default_shader); + + SkyMaterialData *md = static_cast<SkyMaterialData *>(material_storage->material_get_data(sky_shader.default_material, RendererRD::MaterialStorage::SHADER_TYPE_SKY)); + sky_shader.default_shader_rd = sky_shader.shader.version_get_shader(md->shader_data->version, SKY_VERSION_BACKGROUND); + + sky_scene_state.uniform_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(SkySceneState::UBO)); + + Vector<RD::Uniform> uniforms; + + { + Vector<RID> ids; + ids.resize(12); + RID *ids_ptr = ids.ptrw(); + ids_ptr[0] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[1] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[2] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[3] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[4] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[5] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[6] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[7] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[8] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[9] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[10] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[11] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + + RD::Uniform u(RD::UNIFORM_TYPE_SAMPLER, 0, ids); + + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 1; + u.append_id(RendererRD::MaterialStorage::get_singleton()->global_shader_uniforms_get_storage_buffer()); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 2; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.append_id(sky_scene_state.uniform_buffer); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 3; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.append_id(sky_scene_state.directional_light_buffer); + uniforms.push_back(u); + } + + sky_scene_state.uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sky_shader.default_shader_rd, SKY_SET_UNIFORMS); + } + + { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.binding = 0; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID vfog = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE); + u.append_id(vfog); + uniforms.push_back(u); + } + + sky_scene_state.default_fog_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sky_shader.default_shader_rd, SKY_SET_FOG); + } + + { + // Need defaults for using fog with clear color + sky_scene_state.fog_shader = material_storage->shader_allocate(); + material_storage->shader_initialize(sky_scene_state.fog_shader); + + material_storage->shader_set_code(sky_scene_state.fog_shader, R"( +// Default clear color sky shader. + +shader_type sky; + +uniform vec4 clear_color; + +void sky() { + COLOR = clear_color.rgb; +} +)"); + sky_scene_state.fog_material = material_storage->material_allocate(); + material_storage->material_initialize(sky_scene_state.fog_material); + + material_storage->material_set_shader(sky_scene_state.fog_material, sky_scene_state.fog_shader); + + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 0; + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK)); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 1; + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_WHITE)); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 2; + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_WHITE)); + uniforms.push_back(u); + } + + sky_scene_state.fog_only_texture_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sky_shader.default_shader_rd, SKY_SET_TEXTURES); + } + + { //create index array for copy shaders + Vector<uint8_t> pv; + pv.resize(6 * 4); + { + uint8_t *w = pv.ptrw(); + int *p32 = (int *)w; + p32[0] = 0; + p32[1] = 1; + p32[2] = 2; + p32[3] = 0; + p32[4] = 2; + p32[5] = 3; + } + index_buffer = RD::get_singleton()->index_buffer_create(6, RenderingDevice::INDEX_BUFFER_FORMAT_UINT32, pv); + index_array = RD::get_singleton()->index_array_create(index_buffer, 0, 6); + } +} + +void SkyRD::set_texture_format(RD::DataFormat p_texture_format) { + texture_format = p_texture_format; +} + +SkyRD::~SkyRD() { + // cleanup anything created in init... + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + + SkyMaterialData *md = static_cast<SkyMaterialData *>(material_storage->material_get_data(sky_shader.default_material, RendererRD::MaterialStorage::SHADER_TYPE_SKY)); + sky_shader.shader.version_free(md->shader_data->version); + RD::get_singleton()->free(sky_scene_state.directional_light_buffer); + RD::get_singleton()->free(sky_scene_state.uniform_buffer); + memdelete_arr(sky_scene_state.directional_lights); + memdelete_arr(sky_scene_state.last_frame_directional_lights); + material_storage->shader_free(sky_shader.default_shader); + material_storage->material_free(sky_shader.default_material); + material_storage->shader_free(sky_scene_state.fog_shader); + material_storage->material_free(sky_scene_state.fog_material); + + if (RD::get_singleton()->uniform_set_is_valid(sky_scene_state.uniform_set)) { + RD::get_singleton()->free(sky_scene_state.uniform_set); + } + + if (RD::get_singleton()->uniform_set_is_valid(sky_scene_state.default_fog_uniform_set)) { + RD::get_singleton()->free(sky_scene_state.default_fog_uniform_set); + } + + if (RD::get_singleton()->uniform_set_is_valid(sky_scene_state.fog_only_texture_uniform_set)) { + RD::get_singleton()->free(sky_scene_state.fog_only_texture_uniform_set); + } + + RD::get_singleton()->free(index_buffer); //array gets freed as dependency +} + +void SkyRD::setup(RID p_env, Ref<RenderSceneBuffersRD> p_render_buffers, const PagedArray<RID> &p_lights, RID p_camera_attributes, const Projection &p_projection, const Transform3D &p_transform, const Size2i p_screen_size, RendererSceneRenderRD *p_scene_render) { + RendererRD::LightStorage *light_storage = RendererRD::LightStorage::get_singleton(); + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + ERR_FAIL_COND(p_env.is_null()); + + SkyMaterialData *material = nullptr; + Sky *sky = get_sky(RendererSceneRenderRD::get_singleton()->environment_get_sky(p_env)); + + RID sky_material; + + SkyShaderData *shader_data = nullptr; + + if (sky) { + sky_material = sky_get_material(RendererSceneRenderRD::get_singleton()->environment_get_sky(p_env)); + + if (sky_material.is_valid()) { + material = static_cast<SkyMaterialData *>(material_storage->material_get_data(sky_material, RendererRD::MaterialStorage::SHADER_TYPE_SKY)); + if (!material || !material->shader_data->valid) { + material = nullptr; + } + } + + if (!material) { + sky_material = sky_shader.default_material; + material = static_cast<SkyMaterialData *>(material_storage->material_get_data(sky_material, RendererRD::MaterialStorage::SHADER_TYPE_SKY)); + } + + ERR_FAIL_COND(!material); + + shader_data = material->shader_data; + + ERR_FAIL_COND(!shader_data); + + // Invalidate supbass buffers if screen size changes + if (sky->screen_size != p_screen_size) { + sky->screen_size = p_screen_size; + sky->screen_size.x = sky->screen_size.x < 4 ? 4 : sky->screen_size.x; + sky->screen_size.y = sky->screen_size.y < 4 ? 4 : sky->screen_size.y; + if (shader_data->uses_half_res) { + if (sky->half_res_pass.is_valid()) { + RD::get_singleton()->free(sky->half_res_pass); + sky->half_res_pass = RID(); + } + invalidate_sky(sky); + } + if (shader_data->uses_quarter_res) { + if (sky->quarter_res_pass.is_valid()) { + RD::get_singleton()->free(sky->quarter_res_pass); + sky->quarter_res_pass = RID(); + } + invalidate_sky(sky); + } + } + + // Create new subpass buffers if necessary + if ((shader_data->uses_half_res && sky->half_res_pass.is_null()) || + (shader_data->uses_quarter_res && sky->quarter_res_pass.is_null()) || + sky->radiance.is_null()) { + invalidate_sky(sky); + update_dirty_skys(); + } + + if (shader_data->uses_time && p_scene_render->time - sky->prev_time > 0.00001) { + sky->prev_time = p_scene_render->time; + sky->reflection.dirty = true; + RenderingServerDefault::redraw_request(); + } + + if (material != sky->prev_material) { + sky->prev_material = material; + sky->reflection.dirty = true; + } + + if (material->uniform_set_updated) { + material->uniform_set_updated = false; + sky->reflection.dirty = true; + } + + if (!p_transform.origin.is_equal_approx(sky->prev_position) && shader_data->uses_position) { + sky->prev_position = p_transform.origin; + sky->reflection.dirty = true; + } + + if (shader_data->uses_light) { + sky_scene_state.ubo.directional_light_count = 0; + // Run through the list of lights in the scene and pick out the Directional Lights. + // This can't be done in RenderSceneRenderRD::_setup lights because that needs to be called + // after the depth prepass, but this runs before the depth prepass + for (int i = 0; i < (int)p_lights.size(); i++) { + if (!light_storage->owns_light_instance(p_lights[i])) { + continue; + } + RID base = light_storage->light_instance_get_base_light(p_lights[i]); + + ERR_CONTINUE(base.is_null()); + + RS::LightType type = light_storage->light_get_type(base); + if (type == RS::LIGHT_DIRECTIONAL && light_storage->light_directional_get_sky_mode(base) != RS::LIGHT_DIRECTIONAL_SKY_MODE_LIGHT_ONLY) { + SkyDirectionalLightData &sky_light_data = sky_scene_state.directional_lights[sky_scene_state.ubo.directional_light_count]; + Transform3D light_transform = light_storage->light_instance_get_base_transform(p_lights[i]); + Vector3 world_direction = light_transform.basis.xform(Vector3(0, 0, 1)).normalized(); + + sky_light_data.direction[0] = world_direction.x; + sky_light_data.direction[1] = world_direction.y; + sky_light_data.direction[2] = world_direction.z; + + float sign = light_storage->light_is_negative(base) ? -1 : 1; + sky_light_data.energy = sign * light_storage->light_get_param(base, RS::LIGHT_PARAM_ENERGY); + + if (p_scene_render->is_using_physical_light_units()) { + sky_light_data.energy *= light_storage->light_get_param(base, RS::LIGHT_PARAM_INTENSITY); + } + + if (p_camera_attributes.is_valid()) { + sky_light_data.energy *= RSG::camera_attributes->camera_attributes_get_exposure_normalization_factor(p_camera_attributes); + } + + Color linear_col = light_storage->light_get_color(base).srgb_to_linear(); + sky_light_data.color[0] = linear_col.r; + sky_light_data.color[1] = linear_col.g; + sky_light_data.color[2] = linear_col.b; + + sky_light_data.enabled = true; + + float angular_diameter = light_storage->light_get_param(base, RS::LIGHT_PARAM_SIZE); + if (angular_diameter > 0.0) { + // I know tan(0) is 0, but let's not risk it with numerical precision. + // technically this will keep expanding until reaching the sun, but all we care + // is expand until we reach the radius of the near plane (there can't be more occluders than that) + angular_diameter = Math::tan(Math::deg_to_rad(angular_diameter)); + } else { + angular_diameter = 0.0; + } + sky_light_data.size = angular_diameter; + sky_scene_state.ubo.directional_light_count++; + if (sky_scene_state.ubo.directional_light_count >= sky_scene_state.max_directional_lights) { + break; + } + } + } + // Check whether the directional_light_buffer changes + bool light_data_dirty = false; + + // Light buffer is dirty if we have fewer or more lights + // If we have fewer lights, make sure that old lights are disabled + if (sky_scene_state.ubo.directional_light_count != sky_scene_state.last_frame_directional_light_count) { + light_data_dirty = true; + for (uint32_t i = sky_scene_state.ubo.directional_light_count; i < sky_scene_state.max_directional_lights; i++) { + sky_scene_state.directional_lights[i].enabled = false; + sky_scene_state.last_frame_directional_lights[i].enabled = false; + } + } + + if (!light_data_dirty) { + for (uint32_t i = 0; i < sky_scene_state.ubo.directional_light_count; i++) { + if (sky_scene_state.directional_lights[i].direction[0] != sky_scene_state.last_frame_directional_lights[i].direction[0] || + sky_scene_state.directional_lights[i].direction[1] != sky_scene_state.last_frame_directional_lights[i].direction[1] || + sky_scene_state.directional_lights[i].direction[2] != sky_scene_state.last_frame_directional_lights[i].direction[2] || + sky_scene_state.directional_lights[i].energy != sky_scene_state.last_frame_directional_lights[i].energy || + sky_scene_state.directional_lights[i].color[0] != sky_scene_state.last_frame_directional_lights[i].color[0] || + sky_scene_state.directional_lights[i].color[1] != sky_scene_state.last_frame_directional_lights[i].color[1] || + sky_scene_state.directional_lights[i].color[2] != sky_scene_state.last_frame_directional_lights[i].color[2] || + sky_scene_state.directional_lights[i].enabled != sky_scene_state.last_frame_directional_lights[i].enabled || + sky_scene_state.directional_lights[i].size != sky_scene_state.last_frame_directional_lights[i].size) { + light_data_dirty = true; + break; + } + } + } + + if (light_data_dirty) { + RD::get_singleton()->buffer_update(sky_scene_state.directional_light_buffer, 0, sizeof(SkyDirectionalLightData) * sky_scene_state.max_directional_lights, sky_scene_state.directional_lights); + + SkyDirectionalLightData *temp = sky_scene_state.last_frame_directional_lights; + sky_scene_state.last_frame_directional_lights = sky_scene_state.directional_lights; + sky_scene_state.directional_lights = temp; + sky_scene_state.last_frame_directional_light_count = sky_scene_state.ubo.directional_light_count; + sky->reflection.dirty = true; + } + } + } + + //setup fog variables + sky_scene_state.ubo.volumetric_fog_enabled = false; + if (p_render_buffers.is_valid()) { + if (p_render_buffers->has_custom_data(RB_SCOPE_FOG)) { + Ref<RendererRD::Fog::VolumetricFog> fog = p_render_buffers->get_custom_data(RB_SCOPE_FOG); + sky_scene_state.ubo.volumetric_fog_enabled = true; + + float fog_end = fog->length; + if (fog_end > 0.0) { + sky_scene_state.ubo.volumetric_fog_inv_length = 1.0 / fog_end; + } else { + sky_scene_state.ubo.volumetric_fog_inv_length = 1.0; + } + + float fog_detail_spread = fog->spread; //reverse lookup + if (fog_detail_spread > 0.0) { + sky_scene_state.ubo.volumetric_fog_detail_spread = 1.0 / fog_detail_spread; + } else { + sky_scene_state.ubo.volumetric_fog_detail_spread = 1.0; + } + + sky_scene_state.fog_uniform_set = fog->sky_uniform_set; + } + } + + sky_scene_state.ubo.z_far = p_projection.get_z_far(); + sky_scene_state.ubo.fog_enabled = RendererSceneRenderRD::get_singleton()->environment_get_fog_enabled(p_env); + sky_scene_state.ubo.fog_density = RendererSceneRenderRD::get_singleton()->environment_get_fog_density(p_env); + sky_scene_state.ubo.fog_aerial_perspective = RendererSceneRenderRD::get_singleton()->environment_get_fog_aerial_perspective(p_env); + Color fog_color = RendererSceneRenderRD::get_singleton()->environment_get_fog_light_color(p_env).srgb_to_linear(); + float fog_energy = RendererSceneRenderRD::get_singleton()->environment_get_fog_light_energy(p_env); + sky_scene_state.ubo.fog_light_color[0] = fog_color.r * fog_energy; + sky_scene_state.ubo.fog_light_color[1] = fog_color.g * fog_energy; + sky_scene_state.ubo.fog_light_color[2] = fog_color.b * fog_energy; + sky_scene_state.ubo.fog_sun_scatter = RendererSceneRenderRD::get_singleton()->environment_get_fog_sun_scatter(p_env); + + sky_scene_state.ubo.fog_sky_affect = RendererSceneRenderRD::get_singleton()->environment_get_fog_sky_affect(p_env); + sky_scene_state.ubo.volumetric_fog_sky_affect = RendererSceneRenderRD::get_singleton()->environment_get_volumetric_fog_sky_affect(p_env); + + RD::get_singleton()->buffer_update(sky_scene_state.uniform_buffer, 0, sizeof(SkySceneState::UBO), &sky_scene_state.ubo); +} + +void SkyRD::update(RID p_env, const Projection &p_projection, const Transform3D &p_transform, double p_time, float p_luminance_multiplier) { + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + ERR_FAIL_COND(p_env.is_null()); + + Sky *sky = get_sky(RendererSceneRenderRD::get_singleton()->environment_get_sky(p_env)); + ERR_FAIL_COND(!sky); + + RID sky_material = sky_get_material(RendererSceneRenderRD::get_singleton()->environment_get_sky(p_env)); + + SkyMaterialData *material = nullptr; + + if (sky_material.is_valid()) { + material = static_cast<SkyMaterialData *>(material_storage->material_get_data(sky_material, RendererRD::MaterialStorage::SHADER_TYPE_SKY)); + if (!material || !material->shader_data->valid) { + material = nullptr; + } + } + + if (!material) { + sky_material = sky_shader.default_material; + material = static_cast<SkyMaterialData *>(material_storage->material_get_data(sky_material, RendererRD::MaterialStorage::SHADER_TYPE_SKY)); + } + + ERR_FAIL_COND(!material); + + SkyShaderData *shader_data = material->shader_data; + + ERR_FAIL_COND(!shader_data); + + bool update_single_frame = sky->mode == RS::SKY_MODE_REALTIME || sky->mode == RS::SKY_MODE_QUALITY; + RS::SkyMode sky_mode = sky->mode; + + if (sky_mode == RS::SKY_MODE_AUTOMATIC) { + if (shader_data->uses_time || shader_data->uses_position) { + update_single_frame = true; + sky_mode = RS::SKY_MODE_REALTIME; + } else if (shader_data->uses_light || shader_data->ubo_size > 0) { + update_single_frame = false; + sky_mode = RS::SKY_MODE_INCREMENTAL; + } else { + update_single_frame = true; + sky_mode = RS::SKY_MODE_QUALITY; + } + } + + if (sky->processing_layer == 0 && sky_mode == RS::SKY_MODE_INCREMENTAL) { + // On the first frame after creating sky, rebuild in single frame + update_single_frame = true; + sky_mode = RS::SKY_MODE_QUALITY; + } + + int max_processing_layer = sky_use_cubemap_array ? sky->reflection.layers.size() : sky->reflection.layers[0].mipmaps.size(); + + // Update radiance cubemap + if (sky->reflection.dirty && (sky->processing_layer >= max_processing_layer || update_single_frame)) { + static const Vector3 view_normals[6] = { + Vector3(+1, 0, 0), + Vector3(-1, 0, 0), + Vector3(0, +1, 0), + Vector3(0, -1, 0), + Vector3(0, 0, +1), + Vector3(0, 0, -1) + }; + static const Vector3 view_up[6] = { + Vector3(0, -1, 0), + Vector3(0, -1, 0), + Vector3(0, 0, +1), + Vector3(0, 0, -1), + Vector3(0, -1, 0), + Vector3(0, -1, 0) + }; + + Projection cm; + cm.set_perspective(90, 1, 0.01, 10.0); + Projection correction; + correction.set_depth_correction(true); + cm = correction * cm; + + if (shader_data->uses_quarter_res) { + RD::get_singleton()->draw_command_begin_label("Render Sky to Quarter Res Cubemap"); + PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_CUBEMAP_QUARTER_RES]; + + Vector<Color> clear_colors; + clear_colors.push_back(Color(0.0, 0.0, 0.0)); + RD::DrawListID cubemap_draw_list; + + for (int i = 0; i < 6; i++) { + Basis local_view = Basis::looking_at(view_normals[i], view_up[i]); + RID texture_uniform_set = sky->get_textures(SKY_TEXTURE_SET_CUBEMAP_QUARTER_RES, sky_shader.default_shader_rd); + + cubemap_draw_list = RD::get_singleton()->draw_list_begin(sky->reflection.layers[0].mipmaps[2].framebuffers[i], RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + _render_sky(cubemap_draw_list, p_time, sky->reflection.layers[0].mipmaps[2].framebuffers[i], pipeline, material->uniform_set, texture_uniform_set, 1, &cm, local_view, p_transform.origin, p_luminance_multiplier); + RD::get_singleton()->draw_list_end(); + } + RD::get_singleton()->draw_command_end_label(); + } + + if (shader_data->uses_half_res) { + RD::get_singleton()->draw_command_begin_label("Render Sky to Half Res Cubemap"); + PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_CUBEMAP_HALF_RES]; + + Vector<Color> clear_colors; + clear_colors.push_back(Color(0.0, 0.0, 0.0)); + RD::DrawListID cubemap_draw_list; + + for (int i = 0; i < 6; i++) { + Basis local_view = Basis::looking_at(view_normals[i], view_up[i]); + RID texture_uniform_set = sky->get_textures(SKY_TEXTURE_SET_CUBEMAP_HALF_RES, sky_shader.default_shader_rd); + + cubemap_draw_list = RD::get_singleton()->draw_list_begin(sky->reflection.layers[0].mipmaps[1].framebuffers[i], RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + _render_sky(cubemap_draw_list, p_time, sky->reflection.layers[0].mipmaps[1].framebuffers[i], pipeline, material->uniform_set, texture_uniform_set, 1, &cm, local_view, p_transform.origin, p_luminance_multiplier); + RD::get_singleton()->draw_list_end(); + } + RD::get_singleton()->draw_command_end_label(); + } + + RD::DrawListID cubemap_draw_list; + PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_CUBEMAP]; + + RD::get_singleton()->draw_command_begin_label("Render Sky Cubemap"); + for (int i = 0; i < 6; i++) { + Basis local_view = Basis::looking_at(view_normals[i], view_up[i]); + RID texture_uniform_set = sky->get_textures(SKY_TEXTURE_SET_CUBEMAP, sky_shader.default_shader_rd); + + cubemap_draw_list = RD::get_singleton()->draw_list_begin(sky->reflection.layers[0].mipmaps[0].framebuffers[i], RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + _render_sky(cubemap_draw_list, p_time, sky->reflection.layers[0].mipmaps[0].framebuffers[i], pipeline, material->uniform_set, texture_uniform_set, 1, &cm, local_view, p_transform.origin, p_luminance_multiplier); + RD::get_singleton()->draw_list_end(); + } + RD::get_singleton()->draw_command_end_label(); + + if (sky_mode == RS::SKY_MODE_REALTIME) { + sky->reflection.create_reflection_fast_filter(sky_use_cubemap_array); + if (sky_use_cubemap_array) { + sky->reflection.update_reflection_mipmaps(0, sky->reflection.layers.size()); + } + } else { + if (update_single_frame) { + for (int i = 1; i < max_processing_layer; i++) { + sky->reflection.create_reflection_importance_sample(sky_use_cubemap_array, 10, i, sky_ggx_samples_quality); + } + if (sky_use_cubemap_array) { + sky->reflection.update_reflection_mipmaps(0, sky->reflection.layers.size()); + } + } else { + if (sky_use_cubemap_array) { + // Multi-Frame so just update the first array level + sky->reflection.update_reflection_mipmaps(0, 1); + } + } + sky->processing_layer = 1; + } + sky->baked_exposure = p_luminance_multiplier; + sky->reflection.dirty = false; + + } else { + if (sky_mode == RS::SKY_MODE_INCREMENTAL && sky->processing_layer < max_processing_layer) { + sky->reflection.create_reflection_importance_sample(sky_use_cubemap_array, 10, sky->processing_layer, sky_ggx_samples_quality); + + if (sky_use_cubemap_array) { + sky->reflection.update_reflection_mipmaps(sky->processing_layer, sky->processing_layer + 1); + } + + sky->processing_layer++; + } + } +} + +void SkyRD::draw(RID p_env, bool p_can_continue_color, bool p_can_continue_depth, RID p_fb, uint32_t p_view_count, const Projection *p_projections, const Transform3D &p_transform, double p_time, float p_luminance_multiplier) { + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + ERR_FAIL_COND(p_env.is_null()); + + ERR_FAIL_COND(p_view_count == 0); + ERR_FAIL_COND(p_view_count > RendererSceneRender::MAX_RENDER_VIEWS); + + Sky *sky = get_sky(RendererSceneRenderRD::get_singleton()->environment_get_sky(p_env)); + + SkyMaterialData *material = nullptr; + RID sky_material; + + RS::EnvironmentBG background = RendererSceneRenderRD::get_singleton()->environment_get_background(p_env); + + if (!(background == RS::ENV_BG_CLEAR_COLOR || background == RS::ENV_BG_COLOR) || sky) { + ERR_FAIL_COND(!sky); + sky_material = sky_get_material(RendererSceneRenderRD::get_singleton()->environment_get_sky(p_env)); + + if (sky_material.is_valid()) { + material = static_cast<SkyMaterialData *>(material_storage->material_get_data(sky_material, RendererRD::MaterialStorage::SHADER_TYPE_SKY)); + if (!material || !material->shader_data->valid) { + material = nullptr; + } + } + + if (!material) { + sky_material = sky_shader.default_material; + material = static_cast<SkyMaterialData *>(material_storage->material_get_data(sky_material, RendererRD::MaterialStorage::SHADER_TYPE_SKY)); + } + } + + if (background == RS::ENV_BG_CLEAR_COLOR || background == RS::ENV_BG_COLOR) { + sky_material = sky_scene_state.fog_material; + material = static_cast<SkyMaterialData *>(material_storage->material_get_data(sky_material, RendererRD::MaterialStorage::SHADER_TYPE_SKY)); + } + + ERR_FAIL_COND(!material); + + SkyShaderData *shader_data = material->shader_data; + + ERR_FAIL_COND(!shader_data); + + Basis sky_transform = RendererSceneRenderRD::get_singleton()->environment_get_sky_orientation(p_env); + sky_transform.invert(); + + float custom_fov = RendererSceneRenderRD::get_singleton()->environment_get_sky_custom_fov(p_env); + + // Camera + Projection camera; + uint32_t view_count = p_view_count; + const Projection *projections = p_projections; + + if (custom_fov) { + // With custom fov we don't support stereo... + float near_plane = p_projections[0].get_z_near(); + float far_plane = p_projections[0].get_z_far(); + float aspect = p_projections[0].get_aspect(); + + camera.set_perspective(custom_fov, aspect, near_plane, far_plane); + + view_count = 1; + projections = &camera; + } + + sky_transform = sky_transform * p_transform.basis; + + if (shader_data->uses_quarter_res) { + PipelineCacheRD *pipeline = &shader_data->pipelines[view_count > 1 ? SKY_VERSION_QUARTER_RES_MULTIVIEW : SKY_VERSION_QUARTER_RES]; + + RID texture_uniform_set = sky->get_textures(SKY_TEXTURE_SET_QUARTER_RES, sky_shader.default_shader_rd); + + Vector<Color> clear_colors; + clear_colors.push_back(Color(0.0, 0.0, 0.0)); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(sky->quarter_res_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_DISCARD, clear_colors); + _render_sky(draw_list, p_time, sky->quarter_res_framebuffer, pipeline, material->uniform_set, texture_uniform_set, view_count, projections, sky_transform, p_transform.origin, p_luminance_multiplier); + RD::get_singleton()->draw_list_end(); + } + + if (shader_data->uses_half_res) { + PipelineCacheRD *pipeline = &shader_data->pipelines[view_count > 1 ? SKY_VERSION_HALF_RES_MULTIVIEW : SKY_VERSION_HALF_RES]; + + RID texture_uniform_set = sky->get_textures(SKY_TEXTURE_SET_HALF_RES, sky_shader.default_shader_rd); + + Vector<Color> clear_colors; + clear_colors.push_back(Color(0.0, 0.0, 0.0)); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(sky->half_res_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_DISCARD, clear_colors); + _render_sky(draw_list, p_time, sky->half_res_framebuffer, pipeline, material->uniform_set, texture_uniform_set, view_count, projections, sky_transform, p_transform.origin, p_luminance_multiplier); + RD::get_singleton()->draw_list_end(); + } + + PipelineCacheRD *pipeline = &shader_data->pipelines[view_count > 1 ? SKY_VERSION_BACKGROUND_MULTIVIEW : SKY_VERSION_BACKGROUND]; + + RID texture_uniform_set; + if (sky) { + texture_uniform_set = sky->get_textures(SKY_TEXTURE_SET_BACKGROUND, sky_shader.default_shader_rd); + } else { + texture_uniform_set = sky_scene_state.fog_only_texture_uniform_set; + } + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_fb, RD::INITIAL_ACTION_CONTINUE, p_can_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CONTINUE, p_can_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ); + _render_sky(draw_list, p_time, p_fb, pipeline, material->uniform_set, texture_uniform_set, view_count, projections, sky_transform, p_transform.origin, p_luminance_multiplier); + RD::get_singleton()->draw_list_end(); +} + +void SkyRD::update_res_buffers(RID p_env, uint32_t p_view_count, const Projection *p_projections, const Transform3D &p_transform, double p_time, float p_luminance_multiplier) { + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + ERR_FAIL_COND(p_env.is_null()); + + ERR_FAIL_COND(p_view_count == 0); + ERR_FAIL_COND(p_view_count > RendererSceneRender::MAX_RENDER_VIEWS); + + Sky *sky = get_sky(RendererSceneRenderRD::get_singleton()->environment_get_sky(p_env)); + ERR_FAIL_COND(!sky); + + SkyMaterialData *material = nullptr; + RID sky_material; + + sky_material = sky_get_material(RendererSceneRenderRD::get_singleton()->environment_get_sky(p_env)); + + if (sky_material.is_valid()) { + material = static_cast<SkyMaterialData *>(material_storage->material_get_data(sky_material, RendererRD::MaterialStorage::SHADER_TYPE_SKY)); + if (!material || !material->shader_data->valid) { + material = nullptr; + } + } + + if (!material) { + sky_material = sky_shader.default_material; + material = static_cast<SkyMaterialData *>(material_storage->material_get_data(sky_material, RendererRD::MaterialStorage::SHADER_TYPE_SKY)); + } + + ERR_FAIL_COND(!material); + + SkyShaderData *shader_data = material->shader_data; + + ERR_FAIL_COND(!shader_data); + + Basis sky_transform = RendererSceneRenderRD::get_singleton()->environment_get_sky_orientation(p_env); + sky_transform.invert(); + + float custom_fov = RendererSceneRenderRD::get_singleton()->environment_get_sky_custom_fov(p_env); + + // Camera + Projection camera; + uint32_t view_count = p_view_count; + const Projection *projections = p_projections; + + if (custom_fov) { + // With custom fov we don't support stereo... + float near_plane = p_projections[0].get_z_near(); + float far_plane = p_projections[0].get_z_far(); + float aspect = p_projections[0].get_aspect(); + + camera.set_perspective(custom_fov, aspect, near_plane, far_plane); + + view_count = 1; + projections = &camera; + } + + sky_transform = p_transform.basis * sky_transform; + + if (shader_data->uses_quarter_res) { + PipelineCacheRD *pipeline = &shader_data->pipelines[view_count > 1 ? SKY_VERSION_QUARTER_RES_MULTIVIEW : SKY_VERSION_QUARTER_RES]; + + RID texture_uniform_set = sky->get_textures(SKY_TEXTURE_SET_QUARTER_RES, sky_shader.default_shader_rd); + + Vector<Color> clear_colors; + clear_colors.push_back(Color(0.0, 0.0, 0.0)); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(sky->quarter_res_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_DISCARD, clear_colors); + _render_sky(draw_list, p_time, sky->quarter_res_framebuffer, pipeline, material->uniform_set, texture_uniform_set, view_count, projections, sky_transform, p_transform.origin, p_luminance_multiplier); + RD::get_singleton()->draw_list_end(); + } + + if (shader_data->uses_half_res) { + PipelineCacheRD *pipeline = &shader_data->pipelines[view_count > 1 ? SKY_VERSION_HALF_RES_MULTIVIEW : SKY_VERSION_HALF_RES]; + + RID texture_uniform_set = sky->get_textures(SKY_TEXTURE_SET_HALF_RES, sky_shader.default_shader_rd); + + Vector<Color> clear_colors; + clear_colors.push_back(Color(0.0, 0.0, 0.0)); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(sky->half_res_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_DISCARD, clear_colors); + _render_sky(draw_list, p_time, sky->half_res_framebuffer, pipeline, material->uniform_set, texture_uniform_set, view_count, projections, sky_transform, p_transform.origin, p_luminance_multiplier); + RD::get_singleton()->draw_list_end(); + } +} + +void SkyRD::draw(RD::DrawListID p_draw_list, RID p_env, RID p_fb, uint32_t p_view_count, const Projection *p_projections, const Transform3D &p_transform, double p_time, float p_luminance_multiplier) { + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + ERR_FAIL_COND(p_env.is_null()); + + ERR_FAIL_COND(p_view_count == 0); + ERR_FAIL_COND(p_view_count > RendererSceneRender::MAX_RENDER_VIEWS); + + Sky *sky = get_sky(RendererSceneRenderRD::get_singleton()->environment_get_sky(p_env)); + + SkyMaterialData *material = nullptr; + RID sky_material; + + RS::EnvironmentBG background = RendererSceneRenderRD::get_singleton()->environment_get_background(p_env); + + if (!(background == RS::ENV_BG_CLEAR_COLOR || background == RS::ENV_BG_COLOR) || sky) { + ERR_FAIL_COND(!sky); + sky_material = sky_get_material(RendererSceneRenderRD::get_singleton()->environment_get_sky(p_env)); + + if (sky_material.is_valid()) { + material = static_cast<SkyMaterialData *>(material_storage->material_get_data(sky_material, RendererRD::MaterialStorage::SHADER_TYPE_SKY)); + if (!material || !material->shader_data->valid) { + material = nullptr; + } + } + + if (!material) { + sky_material = sky_shader.default_material; + material = static_cast<SkyMaterialData *>(material_storage->material_get_data(sky_material, RendererRD::MaterialStorage::SHADER_TYPE_SKY)); + } + } + + if (background == RS::ENV_BG_CLEAR_COLOR || background == RS::ENV_BG_COLOR) { + sky_material = sky_scene_state.fog_material; + material = static_cast<SkyMaterialData *>(material_storage->material_get_data(sky_material, RendererRD::MaterialStorage::SHADER_TYPE_SKY)); + } + + ERR_FAIL_COND(!material); + + SkyShaderData *shader_data = material->shader_data; + + ERR_FAIL_COND(!shader_data); + + Basis sky_transform = RendererSceneRenderRD::get_singleton()->environment_get_sky_orientation(p_env); + sky_transform.invert(); + + float custom_fov = RendererSceneRenderRD::get_singleton()->environment_get_sky_custom_fov(p_env); + + // Camera + Projection camera; + uint32_t view_count = p_view_count; + const Projection *projections = p_projections; + + if (custom_fov) { + // With custom fov we don't support stereo... + float near_plane = p_projections[0].get_z_near(); + float far_plane = p_projections[0].get_z_far(); + float aspect = p_projections[0].get_aspect(); + + camera.set_perspective(custom_fov, aspect, near_plane, far_plane); + + view_count = 1; + projections = &camera; + } + + sky_transform = p_transform.basis * sky_transform; + + PipelineCacheRD *pipeline = &shader_data->pipelines[view_count > 1 ? SKY_VERSION_BACKGROUND_MULTIVIEW : SKY_VERSION_BACKGROUND]; + + RID texture_uniform_set; + if (sky) { + texture_uniform_set = sky->get_textures(SKY_TEXTURE_SET_BACKGROUND, sky_shader.default_shader_rd); + } else { + texture_uniform_set = sky_scene_state.fog_only_texture_uniform_set; + } + + _render_sky(p_draw_list, p_time, p_fb, pipeline, material->uniform_set, texture_uniform_set, view_count, projections, sky_transform, p_transform.origin, p_luminance_multiplier); +} + +void SkyRD::invalidate_sky(Sky *p_sky) { + if (!p_sky->dirty) { + p_sky->dirty = true; + p_sky->dirty_list = dirty_sky_list; + dirty_sky_list = p_sky; + } +} + +void SkyRD::update_dirty_skys() { + Sky *sky = dirty_sky_list; + + while (sky) { + bool texture_set_dirty = false; + //update sky configuration if texture is missing + + if (sky->radiance.is_null()) { + int mipmaps = Image::get_image_required_mipmaps(sky->radiance_size, sky->radiance_size, Image::FORMAT_RGBAH) + 1; + + uint32_t w = sky->radiance_size, h = sky->radiance_size; + int layers = roughness_layers; + if (sky->mode == RS::SKY_MODE_REALTIME) { + layers = 8; + if (roughness_layers != 8) { + WARN_PRINT("When using REALTIME skies, roughness_layers should be set to 8 in the project settings for best quality reflections"); + } + } + + if (sky_use_cubemap_array) { + //array (higher quality, 6 times more memory) + RD::TextureFormat tf; + tf.array_layers = layers * 6; + tf.format = texture_format; + tf.texture_type = RD::TEXTURE_TYPE_CUBE_ARRAY; + tf.mipmaps = mipmaps; + tf.width = w; + tf.height = h; + tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + + sky->radiance = RD::get_singleton()->texture_create(tf, RD::TextureView()); + + sky->reflection.update_reflection_data(sky->radiance_size, mipmaps, true, sky->radiance, 0, sky->mode == RS::SKY_MODE_REALTIME, roughness_layers, texture_format); + + } else { + //regular cubemap, lower quality (aliasing, less memory) + RD::TextureFormat tf; + tf.array_layers = 6; + tf.format = texture_format; + tf.texture_type = RD::TEXTURE_TYPE_CUBE; + tf.mipmaps = MIN(mipmaps, layers); + tf.width = w; + tf.height = h; + tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + + sky->radiance = RD::get_singleton()->texture_create(tf, RD::TextureView()); + + sky->reflection.update_reflection_data(sky->radiance_size, MIN(mipmaps, layers), false, sky->radiance, 0, sky->mode == RS::SKY_MODE_REALTIME, roughness_layers, texture_format); + } + texture_set_dirty = true; + } + + // Create subpass buffers if they haven't been created already + if (sky->half_res_pass.is_null() && !RD::get_singleton()->texture_is_valid(sky->half_res_pass) && sky->screen_size.x >= 4 && sky->screen_size.y >= 4) { + RD::TextureFormat tformat; + tformat.format = texture_format; + tformat.width = sky->screen_size.x / 2; + tformat.height = sky->screen_size.y / 2; + tformat.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + tformat.texture_type = RD::TEXTURE_TYPE_2D; + + sky->half_res_pass = RD::get_singleton()->texture_create(tformat, RD::TextureView()); + Vector<RID> texs; + texs.push_back(sky->half_res_pass); + sky->half_res_framebuffer = RD::get_singleton()->framebuffer_create(texs); + texture_set_dirty = true; + } + + if (sky->quarter_res_pass.is_null() && !RD::get_singleton()->texture_is_valid(sky->quarter_res_pass) && sky->screen_size.x >= 4 && sky->screen_size.y >= 4) { + RD::TextureFormat tformat; + tformat.format = texture_format; + tformat.width = sky->screen_size.x / 4; + tformat.height = sky->screen_size.y / 4; + tformat.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + tformat.texture_type = RD::TEXTURE_TYPE_2D; + + sky->quarter_res_pass = RD::get_singleton()->texture_create(tformat, RD::TextureView()); + Vector<RID> texs; + texs.push_back(sky->quarter_res_pass); + sky->quarter_res_framebuffer = RD::get_singleton()->framebuffer_create(texs); + texture_set_dirty = true; + } + + if (texture_set_dirty) { + for (int i = 0; i < SKY_TEXTURE_SET_MAX; i++) { + if (sky->texture_uniform_sets[i].is_valid() && RD::get_singleton()->uniform_set_is_valid(sky->texture_uniform_sets[i])) { + RD::get_singleton()->free(sky->texture_uniform_sets[i]); + sky->texture_uniform_sets[i] = RID(); + } + } + } + + sky->reflection.dirty = true; + sky->processing_layer = 0; + + Sky *next = sky->dirty_list; + sky->dirty_list = nullptr; + sky->dirty = false; + sky = next; + } + + dirty_sky_list = nullptr; +} + +RID SkyRD::sky_get_material(RID p_sky) const { + Sky *sky = get_sky(p_sky); + ERR_FAIL_COND_V(!sky, RID()); + + return sky->material; +} + +float SkyRD::sky_get_baked_exposure(RID p_sky) const { + Sky *sky = get_sky(p_sky); + ERR_FAIL_COND_V(!sky, 1.0); + + return sky->baked_exposure; +} + +RID SkyRD::allocate_sky_rid() { + return sky_owner.allocate_rid(); +} + +void SkyRD::initialize_sky_rid(RID p_rid) { + sky_owner.initialize_rid(p_rid, Sky()); +} + +SkyRD::Sky *SkyRD::get_sky(RID p_sky) const { + return sky_owner.get_or_null(p_sky); +} + +void SkyRD::free_sky(RID p_sky) { + Sky *sky = get_sky(p_sky); + ERR_FAIL_COND(!sky); + + sky->free(); + sky_owner.free(p_sky); +} + +void SkyRD::sky_set_radiance_size(RID p_sky, int p_radiance_size) { + Sky *sky = get_sky(p_sky); + ERR_FAIL_COND(!sky); + + if (sky->set_radiance_size(p_radiance_size)) { + invalidate_sky(sky); + } +} + +void SkyRD::sky_set_mode(RID p_sky, RS::SkyMode p_mode) { + Sky *sky = get_sky(p_sky); + ERR_FAIL_COND(!sky); + + if (sky->set_mode(p_mode)) { + invalidate_sky(sky); + } +} + +void SkyRD::sky_set_material(RID p_sky, RID p_material) { + Sky *sky = get_sky(p_sky); + ERR_FAIL_COND(!sky); + + if (sky->set_material(p_material)) { + invalidate_sky(sky); + } +} + +Ref<Image> SkyRD::sky_bake_panorama(RID p_sky, float p_energy, bool p_bake_irradiance, const Size2i &p_size) { + Sky *sky = get_sky(p_sky); + ERR_FAIL_COND_V(!sky, Ref<Image>()); + + update_dirty_skys(); + + return sky->bake_panorama(p_energy, p_bake_irradiance ? roughness_layers : 0, p_size); +} + +RID SkyRD::sky_get_radiance_texture_rd(RID p_sky) const { + Sky *sky = get_sky(p_sky); + ERR_FAIL_COND_V(!sky, RID()); + + return sky->radiance; +} diff --git a/servers/rendering/renderer_rd/environment/sky.h b/servers/rendering/renderer_rd/environment/sky.h new file mode 100644 index 0000000000..45c4f9bda7 --- /dev/null +++ b/servers/rendering/renderer_rd/environment/sky.h @@ -0,0 +1,328 @@ +/*************************************************************************/ +/* sky.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef SKY_RD_H +#define SKY_RD_H + +#include "core/templates/rid_owner.h" +#include "servers/rendering/renderer_compositor.h" +#include "servers/rendering/renderer_rd/pipeline_cache_rd.h" +#include "servers/rendering/renderer_rd/shaders/environment/sky.glsl.gen.h" +#include "servers/rendering/renderer_rd/storage_rd/material_storage.h" +#include "servers/rendering/renderer_scene_render.h" +#include "servers/rendering/rendering_device.h" +#include "servers/rendering/shader_compiler.h" + +// Forward declare RendererSceneRenderRD so we can pass it into some of our methods, these classes are pretty tightly bound +class RendererSceneRenderRD; +class RenderSceneBuffersRD; + +namespace RendererRD { + +class SkyRD { +public: + enum SkySet { + SKY_SET_UNIFORMS, + SKY_SET_MATERIAL, + SKY_SET_TEXTURES, + SKY_SET_FOG, + SKY_SET_MAX + }; + + // Skys need less info from Directional Lights than the normal shaders + struct SkyDirectionalLightData { + float direction[3]; + float energy; + float color[3]; + float size; + uint32_t enabled; + uint32_t pad[3]; + }; + +private: + RD::DataFormat texture_format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + + RID index_buffer; + RID index_array; + + enum SkyTextureSetVersion { + SKY_TEXTURE_SET_BACKGROUND, + SKY_TEXTURE_SET_HALF_RES, + SKY_TEXTURE_SET_QUARTER_RES, + SKY_TEXTURE_SET_CUBEMAP, + SKY_TEXTURE_SET_CUBEMAP_HALF_RES, + SKY_TEXTURE_SET_CUBEMAP_QUARTER_RES, + SKY_TEXTURE_SET_MAX + }; + + enum SkyVersion { + SKY_VERSION_BACKGROUND, + SKY_VERSION_HALF_RES, + SKY_VERSION_QUARTER_RES, + SKY_VERSION_CUBEMAP, + SKY_VERSION_CUBEMAP_HALF_RES, + SKY_VERSION_CUBEMAP_QUARTER_RES, + + SKY_VERSION_BACKGROUND_MULTIVIEW, + SKY_VERSION_HALF_RES_MULTIVIEW, + SKY_VERSION_QUARTER_RES_MULTIVIEW, + + SKY_VERSION_MAX + }; + + struct SkyPushConstant { + float orientation[12]; // 48 - 48 + float projections[RendererSceneRender::MAX_RENDER_VIEWS][4]; // 2 x 16 - 80 + float position[3]; // 12 - 92 + float time; // 4 - 96 + float pad[3]; // 12 - 108 + float luminance_multiplier; // 4 - 112 + // 128 is the max size of a push constant. We can replace "pad" but we can't add any more. + }; + + struct SkyShaderData : public RendererRD::MaterialStorage::ShaderData { + bool valid = false; + RID version; + + PipelineCacheRD pipelines[SKY_VERSION_MAX]; + HashMap<StringName, ShaderLanguage::ShaderNode::Uniform> uniforms; + Vector<ShaderCompiler::GeneratedCode::Texture> texture_uniforms; + + Vector<uint32_t> ubo_offsets; + uint32_t ubo_size = 0; + + String path; + String code; + HashMap<StringName, HashMap<int, RID>> default_texture_params; + + bool uses_time = false; + bool uses_position = false; + bool uses_half_res = false; + bool uses_quarter_res = false; + bool uses_light = false; + + virtual void set_code(const String &p_Code); + virtual void set_path_hint(const String &p_hint); + virtual void set_default_texture_parameter(const StringName &p_name, RID p_texture, int p_index); + virtual void get_shader_uniform_list(List<PropertyInfo> *p_param_list) const; + virtual void get_instance_param_list(List<RendererMaterialStorage::InstanceShaderParam> *p_param_list) const; + virtual bool is_parameter_texture(const StringName &p_param) const; + virtual bool is_animated() const; + virtual bool casts_shadows() const; + virtual Variant get_default_parameter(const StringName &p_parameter) const; + virtual RS::ShaderNativeSourceCode get_native_source_code() const; + + SkyShaderData() {} + virtual ~SkyShaderData(); + }; + + void _render_sky(RD::DrawListID p_list, float p_time, RID p_fb, PipelineCacheRD *p_pipeline, RID p_uniform_set, RID p_texture_set, uint32_t p_view_count, const Projection *p_projections, const Basis &p_orientation, const Vector3 &p_position, float p_luminance_multiplier); + +public: + struct SkySceneState { + struct UBO { + uint32_t volumetric_fog_enabled; // 4 - 4 + float volumetric_fog_inv_length; // 4 - 8 + float volumetric_fog_detail_spread; // 4 - 12 + float volumetric_fog_sky_affect; // 4 - 16 + + uint32_t fog_enabled; // 4 - 20 + float fog_sky_affect; // 4 - 24 + float fog_density; // 4 - 28 + float fog_sun_scatter; // 4 - 32 + + float fog_light_color[3]; // 12 - 44 + float fog_aerial_perspective; // 4 - 48 + + float z_far; // 4 - 52 + uint32_t directional_light_count; // 4 - 56 + uint32_t pad1; // 4 - 60 + uint32_t pad2; // 4 - 64 + }; + + UBO ubo; + + SkyDirectionalLightData *directional_lights = nullptr; + SkyDirectionalLightData *last_frame_directional_lights = nullptr; + uint32_t max_directional_lights; + uint32_t last_frame_directional_light_count; + RID directional_light_buffer; + RID uniform_set; + RID uniform_buffer; + RID fog_uniform_set; + RID default_fog_uniform_set; + + RID fog_shader; + RID fog_material; + RID fog_only_texture_uniform_set; + } sky_scene_state; + + struct ReflectionData { + struct Layer { + struct Mipmap { + RID framebuffers[6]; + RID views[6]; + Size2i size; + }; + Vector<Mipmap> mipmaps; //per-face view + Vector<RID> views; // per-cubemap view + }; + + struct DownsampleLayer { + struct Mipmap { + RID view; + Size2i size; + + // for mobile only + RID views[6]; + RID framebuffers[6]; + }; + Vector<Mipmap> mipmaps; + }; + + RID radiance_base_cubemap; //cubemap for first layer, first cubemap + RID downsampled_radiance_cubemap; + DownsampleLayer downsampled_layer; + RID coefficient_buffer; + + bool dirty = true; + + Vector<Layer> layers; + + void clear_reflection_data(); + void update_reflection_data(int p_size, int p_mipmaps, bool p_use_array, RID p_base_cube, int p_base_layer, bool p_low_quality, int p_roughness_layers, RD::DataFormat p_texture_format); + void create_reflection_fast_filter(bool p_use_arrays); + void create_reflection_importance_sample(bool p_use_arrays, int p_cube_side, int p_base_layer, uint32_t p_sky_ggx_samples_quality); + void update_reflection_mipmaps(int p_start, int p_end); + }; + + /* Sky shader */ + + struct SkyShader { + SkyShaderRD shader; + ShaderCompiler compiler; + + RID default_shader; + RID default_material; + RID default_shader_rd; + } sky_shader; + + struct SkyMaterialData : public RendererRD::MaterialStorage::MaterialData { + SkyShaderData *shader_data = nullptr; + RID uniform_set; + bool uniform_set_updated; + + virtual void set_render_priority(int p_priority) {} + virtual void set_next_pass(RID p_pass) {} + virtual bool update_parameters(const HashMap<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); + virtual ~SkyMaterialData(); + }; + + struct Sky { + RID radiance; + RID half_res_pass; + RID half_res_framebuffer; + RID quarter_res_pass; + RID quarter_res_framebuffer; + Size2i screen_size; + + RID texture_uniform_sets[SKY_TEXTURE_SET_MAX]; + RID uniform_set; + + RID material; + RID uniform_buffer; + + int radiance_size = 256; + + RS::SkyMode mode = RS::SKY_MODE_AUTOMATIC; + + ReflectionData reflection; + bool dirty = false; + int processing_layer = 0; + Sky *dirty_list = nullptr; + float baked_exposure = 1.0; + + //State to track when radiance cubemap needs updating + SkyMaterialData *prev_material = nullptr; + Vector3 prev_position; + float prev_time; + + void free(); + + RID get_textures(SkyTextureSetVersion p_version, RID p_default_shader_rd); + bool set_radiance_size(int p_radiance_size); + bool set_mode(RS::SkyMode p_mode); + bool set_material(RID p_material); + Ref<Image> bake_panorama(float p_energy, int p_roughness_layers, const Size2i &p_size); + }; + + uint32_t sky_ggx_samples_quality; + bool sky_use_cubemap_array; + Sky *dirty_sky_list = nullptr; + mutable RID_Owner<Sky, true> sky_owner; + int roughness_layers; + + RendererRD::MaterialStorage::ShaderData *_create_sky_shader_func(); + static RendererRD::MaterialStorage::ShaderData *_create_sky_shader_funcs(); + + RendererRD::MaterialStorage::MaterialData *_create_sky_material_func(SkyShaderData *p_shader); + static RendererRD::MaterialStorage::MaterialData *_create_sky_material_funcs(RendererRD::MaterialStorage::ShaderData *p_shader); + + SkyRD(); + void init(); + void set_texture_format(RD::DataFormat p_texture_format); + ~SkyRD(); + + void setup(RID p_env, Ref<RenderSceneBuffersRD> p_render_buffers, const PagedArray<RID> &p_lights, RID p_camera_attributes, const Projection &p_projection, const Transform3D &p_transform, const Size2i p_screen_size, RendererSceneRenderRD *p_scene_render); + void update(RID p_env, const Projection &p_projection, const Transform3D &p_transform, double p_time, float p_luminance_multiplier = 1.0); + void draw(RID p_env, bool p_can_continue_color, bool p_can_continue_depth, RID p_fb, uint32_t p_view_count, const Projection *p_projections, const Transform3D &p_transform, double p_time, float p_luminance_multiplier = 1.0); // only called by clustered renderer + void update_res_buffers(RID p_env, uint32_t p_view_count, const Projection *p_projections, const Transform3D &p_transform, double p_time, float p_luminance_multiplier = 1.0); + void draw(RD::DrawListID p_draw_list, RID p_env, RID p_fb, uint32_t p_view_count, const Projection *p_projections, const Transform3D &p_transform, double p_time, float p_luminance_multiplier = 1.0); + + void invalidate_sky(Sky *p_sky); + void update_dirty_skys(); + + RID sky_get_material(RID p_sky) const; + RID sky_get_radiance_texture_rd(RID p_sky) const; + float sky_get_baked_exposure(RID p_sky) const; + + RID allocate_sky_rid(); + void initialize_sky_rid(RID p_rid); + Sky *get_sky(RID p_sky) const; + void free_sky(RID p_sky); + void sky_set_radiance_size(RID p_sky, int p_radiance_size); + void sky_set_mode(RID p_sky, RS::SkyMode p_mode); + void sky_set_material(RID p_sky, RID p_material); + Ref<Image> sky_bake_panorama(RID p_sky, float p_energy, bool p_bake_irradiance, const Size2i &p_size); +}; + +} // namespace RendererRD + +#endif // SKY_RD_H diff --git a/servers/rendering/renderer_rd/forward_clustered/SCsub b/servers/rendering/renderer_rd/forward_clustered/SCsub new file mode 100644 index 0000000000..86681f9c74 --- /dev/null +++ b/servers/rendering/renderer_rd/forward_clustered/SCsub @@ -0,0 +1,5 @@ +#!/usr/bin/env python + +Import("env") + +env.add_source_files(env.servers_sources, "*.cpp") diff --git a/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp new file mode 100644 index 0000000000..ceed287184 --- /dev/null +++ b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp @@ -0,0 +1,3885 @@ +/*************************************************************************/ +/* render_forward_clustered.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "render_forward_clustered.h" +#include "core/config/project_settings.h" +#include "servers/rendering/renderer_rd/framebuffer_cache_rd.h" +#include "servers/rendering/renderer_rd/renderer_compositor_rd.h" +#include "servers/rendering/renderer_rd/storage_rd/light_storage.h" +#include "servers/rendering/renderer_rd/storage_rd/mesh_storage.h" +#include "servers/rendering/renderer_rd/storage_rd/particles_storage.h" +#include "servers/rendering/renderer_rd/storage_rd/texture_storage.h" +#include "servers/rendering/renderer_rd/uniform_set_cache_rd.h" +#include "servers/rendering/rendering_device.h" +#include "servers/rendering/rendering_server_default.h" + +using namespace RendererSceneRenderImplementation; + +void RenderForwardClustered::RenderBufferDataForwardClustered::ensure_specular() { + ERR_FAIL_NULL(render_buffers); + + if (!render_buffers->has_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_SPECULAR)) { + RD::DataFormat format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + uint32_t usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + if (render_buffers->get_msaa_3d() != RS::VIEWPORT_MSAA_DISABLED) { + usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; + } else { + usage_bits |= RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + } + + render_buffers->create_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_SPECULAR, format, usage_bits); + + if (render_buffers->get_msaa_3d() != RS::VIEWPORT_MSAA_DISABLED) { + usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT; + render_buffers->create_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_SPECULAR_MSAA, format, usage_bits, texture_samples); + } + } +} + +void RenderForwardClustered::RenderBufferDataForwardClustered::ensure_normal_roughness_texture() { + ERR_FAIL_NULL(render_buffers); + + if (!render_buffers->has_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_ROUGHNESS)) { + RD::DataFormat format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + uint32_t usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + + if (render_buffers->get_msaa_3d() != RS::VIEWPORT_MSAA_DISABLED) { + usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; + } else { + usage_bits |= RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + } + + render_buffers->create_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_ROUGHNESS, format, usage_bits); + + if (render_buffers->get_msaa_3d() != RS::VIEWPORT_MSAA_DISABLED) { + usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; + render_buffers->create_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_ROUGHNESS_MSAA, format, usage_bits, texture_samples); + } + } +} + +void RenderForwardClustered::RenderBufferDataForwardClustered::ensure_voxelgi() { + ERR_FAIL_NULL(render_buffers); + + if (!render_buffers->has_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_VOXEL_GI)) { + RD::DataFormat format = RD::DATA_FORMAT_R8G8_UINT; + uint32_t usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + if (render_buffers->get_msaa_3d() == RS::VIEWPORT_MSAA_DISABLED) { + usage_bits |= RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + } + + render_buffers->create_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_VOXEL_GI, format, usage_bits); + + if (render_buffers->get_msaa_3d() != RS::VIEWPORT_MSAA_DISABLED) { + usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + render_buffers->create_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_VOXEL_GI_MSAA, format, usage_bits, texture_samples); + } + } +} + +void RenderForwardClustered::RenderBufferDataForwardClustered::free_data() { + // JIC, should already have been cleared + if (render_buffers) { + render_buffers->clear_context(RB_SCOPE_FORWARD_CLUSTERED); + } + + if (cluster_builder) { + memdelete(cluster_builder); + cluster_builder = nullptr; + } + + if (!render_sdfgi_uniform_set.is_null() && RD::get_singleton()->uniform_set_is_valid(render_sdfgi_uniform_set)) { + RD::get_singleton()->free(render_sdfgi_uniform_set); + } + + if (ss_effects_data.linear_depth.is_valid()) { + RD::get_singleton()->free(ss_effects_data.linear_depth); + ss_effects_data.linear_depth = RID(); + ss_effects_data.linear_depth_slices.clear(); + } + + if (ss_effects_data.downsample_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(ss_effects_data.downsample_uniform_set)) { + RD::get_singleton()->free(ss_effects_data.downsample_uniform_set); + ss_effects_data.downsample_uniform_set = RID(); + } + + RenderForwardClustered::get_singleton()->get_ss_effects()->ssao_free(ss_effects_data.ssao); + RenderForwardClustered::get_singleton()->get_ss_effects()->ssil_free(ss_effects_data.ssil); + RenderForwardClustered::get_singleton()->get_ss_effects()->ssr_free(ss_effects_data.ssr); +} + +void RenderForwardClustered::RenderBufferDataForwardClustered::configure(RenderSceneBuffersRD *p_render_buffers) { + if (render_buffers) { + // JIC + free_data(); + } + + render_buffers = p_render_buffers; + ERR_FAIL_NULL(render_buffers); + + RS::ViewportMSAA msaa_3d = render_buffers->get_msaa_3d(); + + if (msaa_3d != RS::VIEWPORT_MSAA_DISABLED) { + RD::DataFormat format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + uint32_t usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; + + const RD::TextureSamples ts[RS::VIEWPORT_MSAA_MAX] = { + RD::TEXTURE_SAMPLES_1, + RD::TEXTURE_SAMPLES_2, + RD::TEXTURE_SAMPLES_4, + RD::TEXTURE_SAMPLES_8, + }; + + texture_samples = ts[msaa_3d]; + + p_render_buffers->create_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_COLOR_MSAA, format, usage_bits, texture_samples); + + format = RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_D24_UNORM_S8_UINT, RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT) ? RD::DATA_FORMAT_D24_UNORM_S8_UINT : RD::DATA_FORMAT_D32_SFLOAT_S8_UINT; + usage_bits = RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; + + p_render_buffers->create_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_DEPTH_MSAA, format, usage_bits, texture_samples); + } + + if (cluster_builder == nullptr) { + cluster_builder = memnew(ClusterBuilderRD); + } + cluster_builder->set_shared(RenderForwardClustered::get_singleton()->get_cluster_builder_shared()); + + RID sampler = RendererRD::MaterialStorage::get_singleton()->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + cluster_builder->setup(p_render_buffers->get_internal_size(), p_render_buffers->get_max_cluster_elements(), p_render_buffers->get_depth_texture(), sampler, p_render_buffers->get_internal_texture()); +} + +RID RenderForwardClustered::RenderBufferDataForwardClustered::get_color_only_fb() { + ERR_FAIL_NULL_V(render_buffers, RID()); + + bool use_msaa = render_buffers->get_msaa_3d() != RS::VIEWPORT_MSAA_DISABLED; + + RID color = use_msaa ? render_buffers->get_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_COLOR_MSAA) : render_buffers->get_internal_texture(); + RID depth = use_msaa ? render_buffers->get_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_DEPTH_MSAA) : render_buffers->get_depth_texture(); + + if (render_buffers->has_texture(RB_SCOPE_VRS, RB_TEXTURE)) { + RID vrs_texture = render_buffers->get_texture(RB_SCOPE_VRS, RB_TEXTURE); + + return FramebufferCacheRD::get_singleton()->get_cache_multiview(render_buffers->get_view_count(), color, depth, vrs_texture); + } else { + return FramebufferCacheRD::get_singleton()->get_cache_multiview(render_buffers->get_view_count(), color, depth); + } +} + +RID RenderForwardClustered::RenderBufferDataForwardClustered::get_color_pass_fb(uint32_t p_color_pass_flags) { + ERR_FAIL_NULL_V(render_buffers, RID()); + bool use_msaa = render_buffers->get_msaa_3d() != RS::VIEWPORT_MSAA_DISABLED; + + int v_count = (p_color_pass_flags & COLOR_PASS_FLAG_MULTIVIEW) ? render_buffers->get_view_count() : 1; + RID color = use_msaa ? render_buffers->get_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_COLOR_MSAA) : render_buffers->get_internal_texture(); + + RID specular; + if (p_color_pass_flags & COLOR_PASS_FLAG_SEPARATE_SPECULAR) { + ensure_specular(); + specular = render_buffers->get_texture(RB_SCOPE_FORWARD_CLUSTERED, use_msaa ? RB_TEX_SPECULAR_MSAA : RB_TEX_SPECULAR); + } + + RID velocity_buffer; + if (p_color_pass_flags & COLOR_PASS_FLAG_MOTION_VECTORS) { + render_buffers->ensure_velocity(); + velocity_buffer = render_buffers->get_velocity_buffer(use_msaa); + } + + RID depth = use_msaa ? render_buffers->get_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_DEPTH_MSAA) : render_buffers->get_depth_texture(); + + if (render_buffers->has_texture(RB_SCOPE_VRS, RB_TEXTURE)) { + RID vrs_texture = render_buffers->get_texture(RB_SCOPE_VRS, RB_TEXTURE); + + return FramebufferCacheRD::get_singleton()->get_cache_multiview(v_count, color, specular, velocity_buffer, depth, vrs_texture); + } else { + return FramebufferCacheRD::get_singleton()->get_cache_multiview(v_count, color, specular, velocity_buffer, depth); + } +} + +RID RenderForwardClustered::RenderBufferDataForwardClustered::get_depth_fb(DepthFrameBufferType p_type) { + ERR_FAIL_NULL_V(render_buffers, RID()); + bool use_msaa = render_buffers->get_msaa_3d() != RS::VIEWPORT_MSAA_DISABLED; + + RID depth = use_msaa ? render_buffers->get_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_DEPTH_MSAA) : render_buffers->get_depth_texture(); + + switch (p_type) { + case DEPTH_FB: { + return FramebufferCacheRD::get_singleton()->get_cache_multiview(render_buffers->get_view_count(), depth); + } break; + case DEPTH_FB_ROUGHNESS: { + ensure_normal_roughness_texture(); + + RID normal_roughness_buffer = render_buffers->get_texture(RB_SCOPE_FORWARD_CLUSTERED, use_msaa ? RB_TEX_ROUGHNESS_MSAA : RB_TEX_ROUGHNESS); + + return FramebufferCacheRD::get_singleton()->get_cache_multiview(render_buffers->get_view_count(), depth, normal_roughness_buffer); + } break; + case DEPTH_FB_ROUGHNESS_VOXELGI: { + ensure_normal_roughness_texture(); + ensure_voxelgi(); + + RID normal_roughness_buffer = render_buffers->get_texture(RB_SCOPE_FORWARD_CLUSTERED, use_msaa ? RB_TEX_ROUGHNESS_MSAA : RB_TEX_ROUGHNESS); + RID voxelgi_buffer = render_buffers->get_texture(RB_SCOPE_FORWARD_CLUSTERED, use_msaa ? RB_TEX_VOXEL_GI_MSAA : RB_TEX_VOXEL_GI); + + return FramebufferCacheRD::get_singleton()->get_cache_multiview(render_buffers->get_view_count(), depth, normal_roughness_buffer, voxelgi_buffer); + } break; + default: { + ERR_FAIL_V(RID()); + } break; + } +} + +RID RenderForwardClustered::RenderBufferDataForwardClustered::get_specular_only_fb() { + bool use_msaa = render_buffers->get_msaa_3d() != RS::VIEWPORT_MSAA_DISABLED; + + RID specular = render_buffers->get_texture(RB_SCOPE_FORWARD_CLUSTERED, use_msaa ? RB_TEX_SPECULAR_MSAA : RB_TEX_SPECULAR); + + return FramebufferCacheRD::get_singleton()->get_cache_multiview(render_buffers->get_view_count(), specular); +} + +void RenderForwardClustered::setup_render_buffer_data(Ref<RenderSceneBuffersRD> p_render_buffers) { + Ref<RenderBufferDataForwardClustered> data; + data.instantiate(); + p_render_buffers->set_custom_data(RB_SCOPE_FORWARD_CLUSTERED, data); + + Ref<RendererRD::GI::RenderBuffersGI> rbgi; + rbgi.instantiate(); + p_render_buffers->set_custom_data(RB_SCOPE_GI, rbgi); +} + +bool RenderForwardClustered::free(RID p_rid) { + if (RendererSceneRenderRD::free(p_rid)) { + return true; + } + return false; +} + +/// RENDERING /// + +template <RenderForwardClustered::PassMode p_pass_mode, uint32_t p_color_pass_flags> +void RenderForwardClustered::_render_list_template(RenderingDevice::DrawListID p_draw_list, RenderingDevice::FramebufferFormatID p_framebuffer_Format, RenderListParameters *p_params, uint32_t p_from_element, uint32_t p_to_element) { + RendererRD::MeshStorage *mesh_storage = RendererRD::MeshStorage::get_singleton(); + RD::DrawListID draw_list = p_draw_list; + RD::FramebufferFormatID framebuffer_format = p_framebuffer_Format; + + //global scope bindings + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, render_base_uniform_set, SCENE_UNIFORM_SET); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, p_params->render_pass_uniform_set, RENDER_PASS_UNIFORM_SET); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, scene_shader.default_vec4_xform_uniform_set, TRANSFORMS_UNIFORM_SET); + + RID prev_material_uniform_set; + + RID prev_vertex_array_rd; + RID prev_index_array_rd; + RID prev_pipeline_rd; + RID prev_xforms_uniform_set; + + bool shadow_pass = (p_pass_mode == PASS_MODE_SHADOW) || (p_pass_mode == PASS_MODE_SHADOW_DP); + + SceneState::PushConstant push_constant; + + if constexpr (p_pass_mode == PASS_MODE_DEPTH_MATERIAL) { + push_constant.uv_offset = Math::make_half_float(p_params->uv_offset.y) << 16; + push_constant.uv_offset |= Math::make_half_float(p_params->uv_offset.x); + } else { + push_constant.uv_offset = 0; + } + + bool should_request_redraw = false; + + for (uint32_t i = p_from_element; i < p_to_element; i++) { + const GeometryInstanceSurfaceDataCache *surf = p_params->elements[i]; + const RenderElementInfo &element_info = p_params->element_info[i]; + + if ((p_pass_mode == PASS_MODE_COLOR && !(p_color_pass_flags & COLOR_PASS_FLAG_TRANSPARENT)) && !(surf->flags & GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE)) { + continue; // Objects with "Depth-prepass" transparency are included in both render lists, but should only be rendered in the transparent pass + } + + if (surf->owner->instance_count == 0) { + continue; + } + + push_constant.base_index = i + p_params->element_offset; + + RID material_uniform_set; + SceneShaderForwardClustered::ShaderData *shader; + void *mesh_surface; + + if (shadow_pass || p_pass_mode == PASS_MODE_DEPTH) { //regular depth pass can use these too + material_uniform_set = surf->material_uniform_set_shadow; + shader = surf->shader_shadow; + mesh_surface = surf->surface_shadow; + + } else { +#ifdef DEBUG_ENABLED + if (unlikely(get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_LIGHTING)) { + material_uniform_set = scene_shader.default_material_uniform_set; + shader = scene_shader.default_material_shader_ptr; + } else if (unlikely(get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_OVERDRAW)) { + material_uniform_set = scene_shader.overdraw_material_uniform_set; + shader = scene_shader.overdraw_material_shader_ptr; + } else { +#endif + material_uniform_set = surf->material_uniform_set; + shader = surf->shader; +#ifdef DEBUG_ENABLED + } +#endif + mesh_surface = surf->surface; + } + + if (!mesh_surface) { + continue; + } + + //request a redraw if one of the shaders uses TIME + if (shader->uses_time) { + should_request_redraw = true; + } + + //find cull variant + SceneShaderForwardClustered::ShaderData::CullVariant cull_variant; + + if (p_pass_mode == PASS_MODE_DEPTH_MATERIAL || p_pass_mode == PASS_MODE_SDF || ((p_pass_mode == PASS_MODE_SHADOW || p_pass_mode == PASS_MODE_SHADOW_DP) && surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_DOUBLE_SIDED_SHADOWS)) { + cull_variant = SceneShaderForwardClustered::ShaderData::CULL_VARIANT_DOUBLE_SIDED; + } else { + bool mirror = surf->owner->mirror; + if (p_params->reverse_cull) { + mirror = !mirror; + } + cull_variant = mirror ? SceneShaderForwardClustered::ShaderData::CULL_VARIANT_REVERSED : SceneShaderForwardClustered::ShaderData::CULL_VARIANT_NORMAL; + } + + RS::PrimitiveType primitive = surf->primitive; + RID xforms_uniform_set = surf->owner->transforms_uniform_set; + + SceneShaderForwardClustered::PipelineVersion pipeline_version = SceneShaderForwardClustered::PIPELINE_VERSION_MAX; // Assigned to silence wrong -Wmaybe-initialized. + uint32_t pipeline_color_pass_flags = 0; + uint32_t pipeline_specialization = 0; + + if constexpr (p_pass_mode == PASS_MODE_COLOR) { + if (element_info.uses_softshadow) { + pipeline_specialization |= SceneShaderForwardClustered::SHADER_SPECIALIZATION_SOFT_SHADOWS; + } + if (element_info.uses_projector) { + pipeline_specialization |= SceneShaderForwardClustered::SHADER_SPECIALIZATION_PROJECTOR; + } + + if (p_params->use_directional_soft_shadow) { + pipeline_specialization |= SceneShaderForwardClustered::SHADER_SPECIALIZATION_DIRECTIONAL_SOFT_SHADOWS; + } + } + + switch (p_pass_mode) { + case PASS_MODE_COLOR: { + if (element_info.uses_lightmap) { + pipeline_color_pass_flags |= SceneShaderForwardClustered::PIPELINE_COLOR_PASS_FLAG_LIGHTMAP; + } else { + if (element_info.uses_forward_gi) { + pipeline_specialization |= SceneShaderForwardClustered::SHADER_SPECIALIZATION_FORWARD_GI; + } + } + + if constexpr ((p_color_pass_flags & COLOR_PASS_FLAG_SEPARATE_SPECULAR) != 0) { + pipeline_color_pass_flags |= SceneShaderForwardClustered::PIPELINE_COLOR_PASS_FLAG_SEPARATE_SPECULAR; + } + + if constexpr ((p_color_pass_flags & COLOR_PASS_FLAG_MOTION_VECTORS) != 0) { + pipeline_color_pass_flags |= SceneShaderForwardClustered::PIPELINE_COLOR_PASS_FLAG_MOTION_VECTORS; + } + + if constexpr ((p_color_pass_flags & COLOR_PASS_FLAG_TRANSPARENT) != 0) { + pipeline_color_pass_flags |= SceneShaderForwardClustered::PIPELINE_COLOR_PASS_FLAG_TRANSPARENT; + } + + if constexpr ((p_color_pass_flags & COLOR_PASS_FLAG_MULTIVIEW) != 0) { + pipeline_color_pass_flags |= SceneShaderForwardClustered::PIPELINE_COLOR_PASS_FLAG_MULTIVIEW; + } + + pipeline_version = SceneShaderForwardClustered::PIPELINE_VERSION_COLOR_PASS; + } break; + case PASS_MODE_SHADOW: + case PASS_MODE_DEPTH: { + pipeline_version = p_params->view_count > 1 ? SceneShaderForwardClustered::PIPELINE_VERSION_DEPTH_PASS_MULTIVIEW : SceneShaderForwardClustered::PIPELINE_VERSION_DEPTH_PASS; + } break; + case PASS_MODE_SHADOW_DP: { + ERR_FAIL_COND_MSG(p_params->view_count > 1, "Multiview not supported for shadow DP pass"); + pipeline_version = SceneShaderForwardClustered::PIPELINE_VERSION_DEPTH_PASS_DP; + } break; + case PASS_MODE_DEPTH_NORMAL_ROUGHNESS: { + pipeline_version = p_params->view_count > 1 ? SceneShaderForwardClustered::PIPELINE_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_MULTIVIEW : SceneShaderForwardClustered::PIPELINE_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS; + } break; + case PASS_MODE_DEPTH_NORMAL_ROUGHNESS_VOXEL_GI: { + pipeline_version = p_params->view_count > 1 ? SceneShaderForwardClustered::PIPELINE_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_VOXEL_GI_MULTIVIEW : SceneShaderForwardClustered::PIPELINE_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_VOXEL_GI; + } break; + case PASS_MODE_DEPTH_MATERIAL: { + ERR_FAIL_COND_MSG(p_params->view_count > 1, "Multiview not supported for material pass"); + pipeline_version = SceneShaderForwardClustered::PIPELINE_VERSION_DEPTH_PASS_WITH_MATERIAL; + } break; + case PASS_MODE_SDF: { + // Note, SDF is prepared in world space, this shouldn't be a multiview buffer even when stereoscopic rendering is used. + ERR_FAIL_COND_MSG(p_params->view_count > 1, "Multiview not supported for SDF pass"); + pipeline_version = SceneShaderForwardClustered::PIPELINE_VERSION_DEPTH_PASS_WITH_SDF; + } break; + } + + PipelineCacheRD *pipeline = nullptr; + + if constexpr (p_pass_mode == PASS_MODE_COLOR) { + pipeline = &shader->color_pipelines[cull_variant][primitive][pipeline_color_pass_flags]; + } else { + pipeline = &shader->pipelines[cull_variant][primitive][pipeline_version]; + } + + RD::VertexFormatID vertex_format = -1; + RID vertex_array_rd; + RID index_array_rd; + + //skeleton and blend shape + if (surf->owner->mesh_instance.is_valid()) { + mesh_storage->mesh_instance_surface_get_vertex_arrays_and_format(surf->owner->mesh_instance, surf->surface_index, pipeline->get_vertex_input_mask(), vertex_array_rd, vertex_format); + } else { + mesh_storage->mesh_surface_get_vertex_arrays_and_format(mesh_surface, pipeline->get_vertex_input_mask(), vertex_array_rd, vertex_format); + } + + index_array_rd = mesh_storage->mesh_surface_get_index_array(mesh_surface, element_info.lod_index); + + if (prev_vertex_array_rd != vertex_array_rd) { + RD::get_singleton()->draw_list_bind_vertex_array(draw_list, vertex_array_rd); + prev_vertex_array_rd = vertex_array_rd; + } + + if (prev_index_array_rd != index_array_rd) { + if (index_array_rd.is_valid()) { + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array_rd); + } + prev_index_array_rd = index_array_rd; + } + + RID pipeline_rd = pipeline->get_render_pipeline(vertex_format, framebuffer_format, p_params->force_wireframe, 0, pipeline_specialization); + + if (pipeline_rd != prev_pipeline_rd) { + // checking with prev shader does not make so much sense, as + // the pipeline may still be different. + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, pipeline_rd); + prev_pipeline_rd = pipeline_rd; + } + + if (xforms_uniform_set.is_valid() && prev_xforms_uniform_set != xforms_uniform_set) { + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, xforms_uniform_set, TRANSFORMS_UNIFORM_SET); + prev_xforms_uniform_set = xforms_uniform_set; + } + + if (material_uniform_set != prev_material_uniform_set) { + // Update uniform set. + if (material_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(material_uniform_set)) { // Material may not have a uniform set. + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, material_uniform_set, MATERIAL_UNIFORM_SET); + } + + prev_material_uniform_set = material_uniform_set; + } + + if ((surf->owner->base_flags & (INSTANCE_DATA_FLAG_MULTIMESH | INSTANCE_DATA_FLAG_PARTICLES)) == INSTANCE_DATA_FLAG_MULTIMESH) { + mesh_storage->_multimesh_get_motion_vectors_offsets(surf->owner->data->base, push_constant.multimesh_motion_vectors_current_offset, push_constant.multimesh_motion_vectors_previous_offset); + } else { + push_constant.multimesh_motion_vectors_current_offset = 0; + push_constant.multimesh_motion_vectors_previous_offset = 0; + } + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &push_constant, sizeof(SceneState::PushConstant)); + + uint32_t instance_count = surf->owner->instance_count > 1 ? surf->owner->instance_count : element_info.repeat; + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_PARTICLE_TRAILS) { + instance_count /= surf->owner->trail_steps; + } + + RD::get_singleton()->draw_list_draw(draw_list, index_array_rd.is_valid(), instance_count); + i += element_info.repeat - 1; //skip equal elements + } + + // Make the actual redraw request + if (should_request_redraw) { + RenderingServerDefault::redraw_request(); + } +} + +void RenderForwardClustered::_render_list(RenderingDevice::DrawListID p_draw_list, RenderingDevice::FramebufferFormatID p_framebuffer_Format, RenderListParameters *p_params, uint32_t p_from_element, uint32_t p_to_element) { + //use template for faster performance (pass mode comparisons are inlined) + + switch (p_params->pass_mode) { +#define VALID_FLAG_COMBINATION(f) \ + case f: { \ + _render_list_template<PASS_MODE_COLOR, f>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); \ + } break; + + case PASS_MODE_COLOR: { + switch (p_params->color_pass_flags) { + VALID_FLAG_COMBINATION(0); + VALID_FLAG_COMBINATION(COLOR_PASS_FLAG_TRANSPARENT); + VALID_FLAG_COMBINATION(COLOR_PASS_FLAG_TRANSPARENT | COLOR_PASS_FLAG_MULTIVIEW); + VALID_FLAG_COMBINATION(COLOR_PASS_FLAG_TRANSPARENT | COLOR_PASS_FLAG_MOTION_VECTORS); + VALID_FLAG_COMBINATION(COLOR_PASS_FLAG_SEPARATE_SPECULAR); + VALID_FLAG_COMBINATION(COLOR_PASS_FLAG_SEPARATE_SPECULAR | COLOR_PASS_FLAG_MULTIVIEW); + VALID_FLAG_COMBINATION(COLOR_PASS_FLAG_SEPARATE_SPECULAR | COLOR_PASS_FLAG_MOTION_VECTORS); + VALID_FLAG_COMBINATION(COLOR_PASS_FLAG_MULTIVIEW); + VALID_FLAG_COMBINATION(COLOR_PASS_FLAG_MULTIVIEW | COLOR_PASS_FLAG_MOTION_VECTORS); + VALID_FLAG_COMBINATION(COLOR_PASS_FLAG_MOTION_VECTORS); + default: { + ERR_FAIL_MSG("Invalid color pass flag combination " + itos(p_params->color_pass_flags)); + } + } + + } break; + case PASS_MODE_SHADOW: { + _render_list_template<PASS_MODE_SHADOW>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + case PASS_MODE_SHADOW_DP: { + _render_list_template<PASS_MODE_SHADOW_DP>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + case PASS_MODE_DEPTH: { + _render_list_template<PASS_MODE_DEPTH>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + case PASS_MODE_DEPTH_NORMAL_ROUGHNESS: { + _render_list_template<PASS_MODE_DEPTH_NORMAL_ROUGHNESS>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + case PASS_MODE_DEPTH_NORMAL_ROUGHNESS_VOXEL_GI: { + _render_list_template<PASS_MODE_DEPTH_NORMAL_ROUGHNESS_VOXEL_GI>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + case PASS_MODE_DEPTH_MATERIAL: { + _render_list_template<PASS_MODE_DEPTH_MATERIAL>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + case PASS_MODE_SDF: { + _render_list_template<PASS_MODE_SDF>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + } +} + +void RenderForwardClustered::_render_list_thread_function(uint32_t p_thread, RenderListParameters *p_params) { + uint32_t render_total = p_params->element_count; + uint32_t total_threads = WorkerThreadPool::get_singleton()->get_thread_count(); + uint32_t render_from = p_thread * render_total / total_threads; + uint32_t render_to = (p_thread + 1 == total_threads) ? render_total : ((p_thread + 1) * render_total / total_threads); + _render_list(thread_draw_lists[p_thread], p_params->framebuffer_format, p_params, render_from, render_to); +} + +void RenderForwardClustered::_render_list_with_threads(RenderListParameters *p_params, RID p_framebuffer, RD::InitialAction p_initial_color_action, RD::FinalAction p_final_color_action, RD::InitialAction p_initial_depth_action, RD::FinalAction p_final_depth_action, const Vector<Color> &p_clear_color_values, float p_clear_depth, uint32_t p_clear_stencil, const Rect2 &p_region, const Vector<RID> &p_storage_textures) { + RD::FramebufferFormatID fb_format = RD::get_singleton()->framebuffer_get_format(p_framebuffer); + p_params->framebuffer_format = fb_format; + + if ((uint32_t)p_params->element_count > render_list_thread_threshold && false) { // secondary command buffers need more testing at this time + //multi threaded + thread_draw_lists.resize(WorkerThreadPool::get_singleton()->get_thread_count()); + RD::get_singleton()->draw_list_begin_split(p_framebuffer, thread_draw_lists.size(), thread_draw_lists.ptr(), p_initial_color_action, p_final_color_action, p_initial_depth_action, p_final_depth_action, p_clear_color_values, p_clear_depth, p_clear_stencil, p_region, p_storage_textures); + WorkerThreadPool::GroupID group_task = WorkerThreadPool::get_singleton()->add_template_group_task(this, &RenderForwardClustered::_render_list_thread_function, p_params, thread_draw_lists.size(), -1, true, SNAME("ForwardClusteredRenderList")); + WorkerThreadPool::get_singleton()->wait_for_group_task_completion(group_task); + RD::get_singleton()->draw_list_end(p_params->barrier); + } else { + //single threaded + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_framebuffer, p_initial_color_action, p_final_color_action, p_initial_depth_action, p_final_depth_action, p_clear_color_values, p_clear_depth, p_clear_stencil, p_region, p_storage_textures); + _render_list(draw_list, fb_format, p_params, 0, p_params->element_count); + RD::get_singleton()->draw_list_end(p_params->barrier); + } +} + +void RenderForwardClustered::_setup_environment(const RenderDataRD *p_render_data, bool p_no_fog, const Size2i &p_screen_size, bool p_flip_y, const Color &p_default_bg_color, bool p_opaque_render_buffers, bool p_pancake_shadows, int p_index) { + RendererRD::LightStorage *light_storage = RendererRD::LightStorage::get_singleton(); + + Ref<RenderSceneBuffersRD> rd = p_render_data->render_buffers; + RID env = is_environment(p_render_data->environment) ? p_render_data->environment : RID(); + RID reflection_probe_instance = p_render_data->reflection_probe.is_valid() ? light_storage->reflection_probe_instance_get_probe(p_render_data->reflection_probe) : RID(); + + // May do this earlier in RenderSceneRenderRD::render_scene + if (p_index >= (int)scene_state.uniform_buffers.size()) { + uint32_t from = scene_state.uniform_buffers.size(); + scene_state.uniform_buffers.resize(p_index + 1); + for (uint32_t i = from; i < scene_state.uniform_buffers.size(); i++) { + scene_state.uniform_buffers[i] = p_render_data->scene_data->create_uniform_buffer(); + } + } + + p_render_data->scene_data->update_ubo(scene_state.uniform_buffers[p_index], get_debug_draw_mode(), env, reflection_probe_instance, p_render_data->camera_attributes, p_flip_y, p_pancake_shadows, p_screen_size, p_default_bg_color, _render_buffers_get_luminance_multiplier(), p_opaque_render_buffers); + + // now do implementation UBO + + scene_state.ubo.cluster_shift = get_shift_from_power_of_2(p_render_data->cluster_size); + scene_state.ubo.max_cluster_element_count_div_32 = p_render_data->cluster_max_elements / 32; + { + uint32_t cluster_screen_width = (p_screen_size.width - 1) / p_render_data->cluster_size + 1; + uint32_t cluster_screen_height = (p_screen_size.height - 1) / p_render_data->cluster_size + 1; + scene_state.ubo.cluster_type_size = cluster_screen_width * cluster_screen_height * (scene_state.ubo.max_cluster_element_count_div_32 + 32); + scene_state.ubo.cluster_width = cluster_screen_width; + } + + scene_state.ubo.gi_upscale_for_msaa = false; + scene_state.ubo.volumetric_fog_enabled = false; + + if (rd.is_valid()) { + if (rd->get_msaa_3d() != RS::VIEWPORT_MSAA_DISABLED) { + scene_state.ubo.gi_upscale_for_msaa = true; + } + + if (rd->has_custom_data(RB_SCOPE_FOG)) { + Ref<RendererRD::Fog::VolumetricFog> fog = rd->get_custom_data(RB_SCOPE_FOG); + + scene_state.ubo.volumetric_fog_enabled = true; + float fog_end = fog->length; + if (fog_end > 0.0) { + scene_state.ubo.volumetric_fog_inv_length = 1.0 / fog_end; + } else { + scene_state.ubo.volumetric_fog_inv_length = 1.0; + } + + float fog_detail_spread = fog->spread; //reverse lookup + if (fog_detail_spread > 0.0) { + scene_state.ubo.volumetric_fog_detail_spread = 1.0 / fog_detail_spread; + } else { + scene_state.ubo.volumetric_fog_detail_spread = 1.0; + } + } + } + + if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_UNSHADED) { + scene_state.ubo.ss_effects_flags = 0; + } else if (is_environment(p_render_data->environment)) { + scene_state.ubo.ssao_ao_affect = environment_get_ssao_ao_channel_affect(p_render_data->environment); + scene_state.ubo.ssao_light_affect = environment_get_ssao_direct_light_affect(p_render_data->environment); + uint32_t ss_flags = 0; + if (p_opaque_render_buffers) { + ss_flags |= environment_get_ssao_enabled(p_render_data->environment) ? 1 : 0; + ss_flags |= environment_get_ssil_enabled(p_render_data->environment) ? 2 : 0; + } + scene_state.ubo.ss_effects_flags = ss_flags; + } else { + scene_state.ubo.ss_effects_flags = 0; + } + + if (p_index >= (int)scene_state.implementation_uniform_buffers.size()) { + uint32_t from = scene_state.implementation_uniform_buffers.size(); + scene_state.implementation_uniform_buffers.resize(p_index + 1); + for (uint32_t i = from; i < scene_state.implementation_uniform_buffers.size(); i++) { + scene_state.implementation_uniform_buffers[i] = RD::get_singleton()->uniform_buffer_create(sizeof(SceneState::UBO)); + } + } + + RD::get_singleton()->buffer_update(scene_state.implementation_uniform_buffers[p_index], 0, sizeof(SceneState::UBO), &scene_state.ubo, RD::BARRIER_MASK_RASTER); +} + +void RenderForwardClustered::_update_instance_data_buffer(RenderListType p_render_list) { + if (scene_state.instance_data[p_render_list].size() > 0) { + if (scene_state.instance_buffer[p_render_list] == RID() || scene_state.instance_buffer_size[p_render_list] < scene_state.instance_data[p_render_list].size()) { + if (scene_state.instance_buffer[p_render_list] != RID()) { + RD::get_singleton()->free(scene_state.instance_buffer[p_render_list]); + } + uint32_t new_size = nearest_power_of_2_templated(MAX(uint64_t(INSTANCE_DATA_BUFFER_MIN_SIZE), scene_state.instance_data[p_render_list].size())); + scene_state.instance_buffer[p_render_list] = RD::get_singleton()->storage_buffer_create(new_size * sizeof(SceneState::InstanceData)); + scene_state.instance_buffer_size[p_render_list] = new_size; + } + RD::get_singleton()->buffer_update(scene_state.instance_buffer[p_render_list], 0, sizeof(SceneState::InstanceData) * scene_state.instance_data[p_render_list].size(), scene_state.instance_data[p_render_list].ptr(), RD::BARRIER_MASK_RASTER); + } +} +void RenderForwardClustered::_fill_instance_data(RenderListType p_render_list, int *p_render_info, uint32_t p_offset, int32_t p_max_elements, bool p_update_buffer) { + RenderList *rl = &render_list[p_render_list]; + uint32_t element_total = p_max_elements >= 0 ? uint32_t(p_max_elements) : rl->elements.size(); + + scene_state.instance_data[p_render_list].resize(p_offset + element_total); + rl->element_info.resize(p_offset + element_total); + + if (p_render_info) { + p_render_info[RS::VIEWPORT_RENDER_INFO_OBJECTS_IN_FRAME] += element_total; + } + uint64_t frame = RSG::rasterizer->get_frame_number(); + uint32_t repeats = 0; + GeometryInstanceSurfaceDataCache *prev_surface = nullptr; + for (uint32_t i = 0; i < element_total; i++) { + GeometryInstanceSurfaceDataCache *surface = rl->elements[i + p_offset]; + GeometryInstanceForwardClustered *inst = surface->owner; + + SceneState::InstanceData &instance_data = scene_state.instance_data[p_render_list][i + p_offset]; + + if (inst->prev_transform_dirty && frame > inst->prev_transform_change_frame + 1 && inst->prev_transform_change_frame) { + inst->prev_transform = inst->transform; + inst->prev_transform_dirty = false; + } + + if (inst->store_transform_cache) { + RendererRD::MaterialStorage::store_transform(inst->transform, instance_data.transform); + RendererRD::MaterialStorage::store_transform(inst->prev_transform, instance_data.prev_transform); + } else { + RendererRD::MaterialStorage::store_transform(Transform3D(), instance_data.transform); + RendererRD::MaterialStorage::store_transform(Transform3D(), instance_data.prev_transform); + } + + instance_data.flags = inst->flags_cache; + instance_data.gi_offset = inst->gi_offset_cache; + instance_data.layer_mask = inst->layer_mask; + instance_data.instance_uniforms_ofs = uint32_t(inst->shader_uniforms_offset); + instance_data.lightmap_uv_scale[0] = inst->lightmap_uv_scale.position.x; + instance_data.lightmap_uv_scale[1] = inst->lightmap_uv_scale.position.y; + instance_data.lightmap_uv_scale[2] = inst->lightmap_uv_scale.size.x; + instance_data.lightmap_uv_scale[3] = inst->lightmap_uv_scale.size.y; + +#ifdef REAL_T_IS_DOUBLE + // Split the origin into two components, the float approximation and the missing precision + // In the shader we will combine these back together to restore the lost precision. + RendererRD::MaterialStorage::split_double(inst->transform.origin.x, &instance_data.transform[12], &instance_data.transform[3]); + RendererRD::MaterialStorage::split_double(inst->transform.origin.y, &instance_data.transform[13], &instance_data.transform[7]); + RendererRD::MaterialStorage::split_double(inst->transform.origin.z, &instance_data.transform[14], &instance_data.transform[11]); +#endif + + bool cant_repeat = instance_data.flags & INSTANCE_DATA_FLAG_MULTIMESH || inst->mesh_instance.is_valid(); + + if (prev_surface != nullptr && !cant_repeat && prev_surface->sort.sort_key1 == surface->sort.sort_key1 && prev_surface->sort.sort_key2 == surface->sort.sort_key2 && inst->mirror == prev_surface->owner->mirror && repeats < RenderElementInfo::MAX_REPEATS) { + //this element is the same as the previous one, count repeats to draw it using instancing + repeats++; + } else { + if (repeats > 0) { + for (uint32_t j = 1; j <= repeats; j++) { + rl->element_info[p_offset + i - j].repeat = j; + } + } + repeats = 1; + if (p_render_info) { + p_render_info[RS::VIEWPORT_RENDER_INFO_DRAW_CALLS_IN_FRAME]++; + } + } + + RenderElementInfo &element_info = rl->element_info[p_offset + i]; + + element_info.lod_index = surface->sort.lod_index; + element_info.uses_forward_gi = surface->sort.uses_forward_gi; + element_info.uses_lightmap = surface->sort.uses_lightmap; + element_info.uses_softshadow = surface->sort.uses_softshadow; + element_info.uses_projector = surface->sort.uses_projector; + + if (cant_repeat) { + prev_surface = nullptr; + } else { + prev_surface = surface; + } + } + + if (repeats > 0) { + for (uint32_t j = 1; j <= repeats; j++) { + rl->element_info[p_offset + element_total - j].repeat = j; + } + } + + if (p_update_buffer) { + _update_instance_data_buffer(p_render_list); + } +} + +_FORCE_INLINE_ static uint32_t _indices_to_primitives(RS::PrimitiveType p_primitive, uint32_t p_indices) { + static const uint32_t divisor[RS::PRIMITIVE_MAX] = { 1, 2, 1, 3, 1 }; + static const uint32_t subtractor[RS::PRIMITIVE_MAX] = { 0, 0, 1, 0, 1 }; + return (p_indices - subtractor[p_primitive]) / divisor[p_primitive]; +} +void RenderForwardClustered::_fill_render_list(RenderListType p_render_list, const RenderDataRD *p_render_data, PassMode p_pass_mode, uint32_t p_color_pass_flags = 0, bool p_using_sdfgi, bool p_using_opaque_gi, bool p_append) { + RendererRD::MeshStorage *mesh_storage = RendererRD::MeshStorage::get_singleton(); + + if (p_render_list == RENDER_LIST_OPAQUE) { + scene_state.used_sss = false; + scene_state.used_screen_texture = false; + scene_state.used_normal_texture = false; + scene_state.used_depth_texture = false; + } + uint32_t lightmap_captures_used = 0; + + Plane near_plane = Plane(-p_render_data->scene_data->cam_transform.basis.get_column(Vector3::AXIS_Z), p_render_data->scene_data->cam_transform.origin); + near_plane.d += p_render_data->scene_data->cam_projection.get_z_near(); + float z_max = p_render_data->scene_data->cam_projection.get_z_far() - p_render_data->scene_data->cam_projection.get_z_near(); + + RenderList *rl = &render_list[p_render_list]; + _update_dirty_geometry_instances(); + + if (!p_append) { + rl->clear(); + if (p_render_list == RENDER_LIST_OPAQUE) { + render_list[RENDER_LIST_ALPHA].clear(); //opaque fills alpha too + } + } + + //fill list + + for (int i = 0; i < (int)p_render_data->instances->size(); i++) { + GeometryInstanceForwardClustered *inst = static_cast<GeometryInstanceForwardClustered *>((*p_render_data->instances)[i]); + + Vector3 support_min = inst->transformed_aabb.get_support(-near_plane.normal); + inst->depth = near_plane.distance_to(support_min); + uint32_t depth_layer = CLAMP(int(inst->depth * 16 / z_max), 0, 15); + + uint32_t flags = inst->base_flags; //fill flags if appropriate + + if (inst->non_uniform_scale) { + flags |= INSTANCE_DATA_FLAGS_NON_UNIFORM_SCALE; + } + bool uses_lightmap = false; + bool uses_gi = false; + float fade_alpha = 1.0; + + if (inst->fade_near || inst->fade_far) { + float fade_dist = inst->transform.origin.distance_to(p_render_data->scene_data->cam_transform.origin); + // Use `smoothstep()` to make opacity changes more gradual and less noticeable to the player. + if (inst->fade_far && fade_dist > inst->fade_far_begin) { + fade_alpha = Math::smoothstep(0.0f, 1.0f, 1.0f - (fade_dist - inst->fade_far_begin) / (inst->fade_far_end - inst->fade_far_begin)); + } else if (inst->fade_near && fade_dist < inst->fade_near_end) { + fade_alpha = Math::smoothstep(0.0f, 1.0f, (fade_dist - inst->fade_near_begin) / (inst->fade_near_end - inst->fade_near_begin)); + } + } + + fade_alpha *= inst->force_alpha * inst->parent_fade_alpha; + + flags = (flags & ~INSTANCE_DATA_FLAGS_FADE_MASK) | (uint32_t(fade_alpha * 255.0) << INSTANCE_DATA_FLAGS_FADE_SHIFT); + + if (p_render_list == RENDER_LIST_OPAQUE) { + // Setup GI + if (inst->lightmap_instance.is_valid()) { + int32_t lightmap_cull_index = -1; + for (uint32_t j = 0; j < scene_state.lightmaps_used; j++) { + if (scene_state.lightmap_ids[j] == inst->lightmap_instance) { + lightmap_cull_index = j; + break; + } + } + if (lightmap_cull_index >= 0) { + inst->gi_offset_cache = inst->lightmap_slice_index << 16; + inst->gi_offset_cache |= lightmap_cull_index; + flags |= INSTANCE_DATA_FLAG_USE_LIGHTMAP; + if (scene_state.lightmap_has_sh[lightmap_cull_index]) { + flags |= INSTANCE_DATA_FLAG_USE_SH_LIGHTMAP; + } + uses_lightmap = true; + } else { + inst->gi_offset_cache = 0xFFFFFFFF; + } + + } else if (inst->lightmap_sh) { + if (lightmap_captures_used < scene_state.max_lightmap_captures) { + const Color *src_capture = inst->lightmap_sh->sh; + LightmapCaptureData &lcd = scene_state.lightmap_captures[lightmap_captures_used]; + for (int j = 0; j < 9; j++) { + lcd.sh[j * 4 + 0] = src_capture[j].r; + lcd.sh[j * 4 + 1] = src_capture[j].g; + lcd.sh[j * 4 + 2] = src_capture[j].b; + lcd.sh[j * 4 + 3] = src_capture[j].a; + } + flags |= INSTANCE_DATA_FLAG_USE_LIGHTMAP_CAPTURE; + inst->gi_offset_cache = lightmap_captures_used; + lightmap_captures_used++; + uses_lightmap = true; + } + + } else { + if (p_using_opaque_gi) { + flags |= INSTANCE_DATA_FLAG_USE_GI_BUFFERS; + } + + if (inst->voxel_gi_instances[0].is_valid()) { + uint32_t probe0_index = 0xFFFF; + uint32_t probe1_index = 0xFFFF; + + for (uint32_t j = 0; j < scene_state.voxelgis_used; j++) { + if (scene_state.voxelgi_ids[j] == inst->voxel_gi_instances[0]) { + probe0_index = j; + } else if (scene_state.voxelgi_ids[j] == inst->voxel_gi_instances[1]) { + probe1_index = j; + } + } + + if (probe0_index == 0xFFFF && probe1_index != 0xFFFF) { + //0 must always exist if a probe exists + SWAP(probe0_index, probe1_index); + } + + inst->gi_offset_cache = probe0_index | (probe1_index << 16); + flags |= INSTANCE_DATA_FLAG_USE_VOXEL_GI; + uses_gi = true; + } else { + if (p_using_sdfgi && inst->can_sdfgi) { + flags |= INSTANCE_DATA_FLAG_USE_SDFGI; + uses_gi = true; + } + inst->gi_offset_cache = 0xFFFFFFFF; + } + } + } + inst->flags_cache = flags; + + GeometryInstanceSurfaceDataCache *surf = inst->surface_caches; + + while (surf) { + surf->sort.uses_forward_gi = 0; + surf->sort.uses_lightmap = 0; + + // LOD + + if (p_render_data->scene_data->screen_mesh_lod_threshold > 0.0 && mesh_storage->mesh_surface_has_lod(surf->surface)) { + // Get the LOD support points on the mesh AABB. + Vector3 lod_support_min = inst->transformed_aabb.get_support(p_render_data->scene_data->cam_transform.basis.get_column(Vector3::AXIS_Z)); + Vector3 lod_support_max = inst->transformed_aabb.get_support(-p_render_data->scene_data->cam_transform.basis.get_column(Vector3::AXIS_Z)); + + // Get the distances to those points on the AABB from the camera origin. + float distance_min = (float)p_render_data->scene_data->cam_transform.origin.distance_to(lod_support_min); + float distance_max = (float)p_render_data->scene_data->cam_transform.origin.distance_to(lod_support_max); + + float distance = 0.0; + + if (distance_min * distance_max < 0.0) { + //crossing plane + distance = 0.0; + } else if (distance_min >= 0.0) { + distance = distance_min; + } else if (distance_max <= 0.0) { + distance = -distance_max; + } + + if (p_render_data->scene_data->cam_orthogonal) { + distance = 1.0; + } + + uint32_t indices = 0; + surf->sort.lod_index = mesh_storage->mesh_surface_get_lod(surf->surface, inst->lod_model_scale * inst->lod_bias, distance * p_render_data->scene_data->lod_distance_multiplier, p_render_data->scene_data->screen_mesh_lod_threshold, indices); + if (p_render_data->render_info) { + indices = _indices_to_primitives(surf->primitive, indices); + if (p_render_list == RENDER_LIST_OPAQUE) { //opaque + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_VISIBLE][RS::VIEWPORT_RENDER_INFO_PRIMITIVES_IN_FRAME] += indices; + } else if (p_render_list == RENDER_LIST_SECONDARY) { //shadow + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_SHADOW][RS::VIEWPORT_RENDER_INFO_PRIMITIVES_IN_FRAME] += indices; + } + } + } else { + surf->sort.lod_index = 0; + if (p_render_data->render_info) { + uint32_t to_draw = mesh_storage->mesh_surface_get_vertices_drawn_count(surf->surface); + to_draw = _indices_to_primitives(surf->primitive, to_draw); + to_draw *= inst->instance_count; + if (p_render_list == RENDER_LIST_OPAQUE) { //opaque + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_VISIBLE][RS::VIEWPORT_RENDER_INFO_PRIMITIVES_IN_FRAME] += mesh_storage->mesh_surface_get_vertices_drawn_count(surf->surface); + } else if (p_render_list == RENDER_LIST_SECONDARY) { //shadow + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_SHADOW][RS::VIEWPORT_RENDER_INFO_PRIMITIVES_IN_FRAME] += mesh_storage->mesh_surface_get_vertices_drawn_count(surf->surface); + } + } + } + + // ADD Element + if (p_pass_mode == PASS_MODE_COLOR) { +#ifdef DEBUG_ENABLED + bool force_alpha = unlikely(get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_OVERDRAW); +#else + bool force_alpha = false; +#endif + + if (fade_alpha < 0.999) { + force_alpha = true; + } + + if (!force_alpha && (surf->flags & (GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH | GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE))) { + rl->add_element(surf); + } + if (force_alpha || (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_PASS_ALPHA)) { + render_list[RENDER_LIST_ALPHA].add_element(surf); + if (uses_gi) { + surf->sort.uses_forward_gi = 1; + } + } + + if (uses_lightmap) { + surf->sort.uses_lightmap = 1; + } + + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_SUBSURFACE_SCATTERING) { + scene_state.used_sss = true; + } + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_SCREEN_TEXTURE) { + scene_state.used_screen_texture = true; + } + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_NORMAL_TEXTURE) { + scene_state.used_normal_texture = true; + } + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_DEPTH_TEXTURE) { + scene_state.used_depth_texture = true; + } + + if (p_color_pass_flags & COLOR_PASS_FLAG_MOTION_VECTORS) { + if ((flags & (INSTANCE_DATA_FLAG_MULTIMESH | INSTANCE_DATA_FLAG_PARTICLES)) == INSTANCE_DATA_FLAG_MULTIMESH && RendererRD::MeshStorage::get_singleton()->_multimesh_enable_motion_vectors(inst->data->base)) { + inst->transforms_uniform_set = mesh_storage->multimesh_get_3d_uniform_set(inst->data->base, scene_shader.default_shader_rd, TRANSFORMS_UNIFORM_SET); + } + } + + } else if (p_pass_mode == PASS_MODE_SHADOW || p_pass_mode == PASS_MODE_SHADOW_DP) { + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_PASS_SHADOW) { + rl->add_element(surf); + } + } else { + if (surf->flags & (GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH | GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE)) { + rl->add_element(surf); + } + } + + surf->sort.depth_layer = depth_layer; + + surf = surf->next; + } + } + + if (p_render_list == RENDER_LIST_OPAQUE && lightmap_captures_used) { + RD::get_singleton()->buffer_update(scene_state.lightmap_capture_buffer, 0, sizeof(LightmapCaptureData) * lightmap_captures_used, scene_state.lightmap_captures, RD::BARRIER_MASK_RASTER); + } +} + +void RenderForwardClustered::_setup_voxelgis(const PagedArray<RID> &p_voxelgis) { + scene_state.voxelgis_used = MIN(p_voxelgis.size(), uint32_t(MAX_VOXEL_GI_INSTANCESS)); + for (uint32_t i = 0; i < scene_state.voxelgis_used; i++) { + scene_state.voxelgi_ids[i] = p_voxelgis[i]; + } +} + +void RenderForwardClustered::_setup_lightmaps(const RenderDataRD *p_render_data, const PagedArray<RID> &p_lightmaps, const Transform3D &p_cam_transform) { + RendererRD::LightStorage *light_storage = RendererRD::LightStorage::get_singleton(); + + scene_state.lightmaps_used = 0; + for (int i = 0; i < (int)p_lightmaps.size(); i++) { + if (i >= (int)scene_state.max_lightmaps) { + break; + } + + RID lightmap = light_storage->lightmap_instance_get_lightmap(p_lightmaps[i]); + + Basis to_lm = light_storage->lightmap_instance_get_transform(p_lightmaps[i]).basis.inverse() * p_cam_transform.basis; + to_lm = to_lm.inverse().transposed(); //will transform normals + RendererRD::MaterialStorage::store_transform_3x3(to_lm, scene_state.lightmaps[i].normal_xform); + scene_state.lightmaps[i].exposure_normalization = 1.0; + if (p_render_data->camera_attributes.is_valid()) { + float baked_exposure = light_storage->lightmap_get_baked_exposure_normalization(lightmap); + float enf = RSG::camera_attributes->camera_attributes_get_exposure_normalization_factor(p_render_data->camera_attributes); + scene_state.lightmaps[i].exposure_normalization = enf / baked_exposure; + } + + scene_state.lightmap_ids[i] = p_lightmaps[i]; + scene_state.lightmap_has_sh[i] = light_storage->lightmap_uses_spherical_harmonics(lightmap); + + scene_state.lightmaps_used++; + } + if (scene_state.lightmaps_used > 0) { + RD::get_singleton()->buffer_update(scene_state.lightmap_buffer, 0, sizeof(LightmapData) * scene_state.lightmaps_used, scene_state.lightmaps, RD::BARRIER_MASK_RASTER); + } +} + +/* SDFGI */ + +void RenderForwardClustered::_update_sdfgi(RenderDataRD *p_render_data) { + Ref<RenderSceneBuffersRD> rb; + if (p_render_data && p_render_data->render_buffers.is_valid()) { + rb = p_render_data->render_buffers; + } + + if (rb.is_valid() && rb->has_custom_data(RB_SCOPE_SDFGI)) { + Ref<RendererRD::GI::SDFGI> sdfgi = rb->get_custom_data(RB_SCOPE_SDFGI); + float exposure_normalization = 1.0; + + if (p_render_data->camera_attributes.is_valid()) { + exposure_normalization = RSG::camera_attributes->camera_attributes_get_exposure_normalization_factor(p_render_data->camera_attributes); + } + for (int i = 0; i < p_render_data->render_sdfgi_region_count; i++) { + sdfgi->render_region(rb, p_render_data->render_sdfgi_regions[i].region, p_render_data->render_sdfgi_regions[i].instances, exposure_normalization); + } + if (p_render_data->sdfgi_update_data->update_static) { + sdfgi->render_static_lights(p_render_data, rb, p_render_data->sdfgi_update_data->static_cascade_count, p_render_data->sdfgi_update_data->static_cascade_indices, p_render_data->sdfgi_update_data->static_positional_lights); + } + } +} + +/* Debug */ + +void RenderForwardClustered::_debug_draw_cluster(Ref<RenderSceneBuffersRD> p_render_buffers) { + if (p_render_buffers.is_valid() && current_cluster_builder != nullptr) { + RS::ViewportDebugDraw dd = get_debug_draw_mode(); + + if (dd == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_OMNI_LIGHTS || dd == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_SPOT_LIGHTS || dd == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_DECALS || dd == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_REFLECTION_PROBES) { + ClusterBuilderRD::ElementType elem_type = ClusterBuilderRD::ELEMENT_TYPE_MAX; + switch (dd) { + case RS::VIEWPORT_DEBUG_DRAW_CLUSTER_OMNI_LIGHTS: + elem_type = ClusterBuilderRD::ELEMENT_TYPE_OMNI_LIGHT; + break; + case RS::VIEWPORT_DEBUG_DRAW_CLUSTER_SPOT_LIGHTS: + elem_type = ClusterBuilderRD::ELEMENT_TYPE_SPOT_LIGHT; + break; + case RS::VIEWPORT_DEBUG_DRAW_CLUSTER_DECALS: + elem_type = ClusterBuilderRD::ELEMENT_TYPE_DECAL; + break; + case RS::VIEWPORT_DEBUG_DRAW_CLUSTER_REFLECTION_PROBES: + elem_type = ClusterBuilderRD::ELEMENT_TYPE_REFLECTION_PROBE; + break; + default: { + } + } + current_cluster_builder->debug(elem_type); + } + } +} + +//////////////////////////////////////////////////////////////////////////////// +// FOG SHADER + +void RenderForwardClustered::_update_volumetric_fog(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_environment, const Projection &p_cam_projection, const Transform3D &p_cam_transform, const Transform3D &p_prev_cam_inv_transform, RID p_shadow_atlas, int p_directional_light_count, bool p_use_directional_shadows, int p_positional_light_count, int p_voxel_gi_count, const PagedArray<RID> &p_fog_volumes) { + ERR_FAIL_COND(p_render_buffers.is_null()); + + Ref<RenderBufferDataForwardClustered> rb_data = p_render_buffers->get_custom_data(RB_SCOPE_FORWARD_CLUSTERED); + + ERR_FAIL_COND(!p_render_buffers->has_custom_data(RB_SCOPE_GI)); + Ref<RendererRD::GI::RenderBuffersGI> rbgi = p_render_buffers->get_custom_data(RB_SCOPE_GI); + + Ref<RendererRD::GI::SDFGI> sdfgi; + if (p_render_buffers->has_custom_data(RB_SCOPE_SDFGI)) { + sdfgi = p_render_buffers->get_custom_data(RB_SCOPE_SDFGI); + } + + Size2i size = p_render_buffers->get_internal_size(); + float ratio = float(size.x) / float((size.x + size.y) / 2); + uint32_t target_width = uint32_t(float(get_volumetric_fog_size()) * ratio); + uint32_t target_height = uint32_t(float(get_volumetric_fog_size()) / ratio); + + if (p_render_buffers->has_custom_data(RB_SCOPE_FOG)) { + Ref<RendererRD::Fog::VolumetricFog> fog = p_render_buffers->get_custom_data(RB_SCOPE_FOG); + //validate + if (p_environment.is_null() || !environment_get_volumetric_fog_enabled(p_environment) || fog->width != target_width || fog->height != target_height || fog->depth != get_volumetric_fog_depth()) { + p_render_buffers->set_custom_data(RB_SCOPE_FOG, Ref<RenderBufferCustomDataRD>()); + } + } + + if (p_environment.is_null() || !environment_get_volumetric_fog_enabled(p_environment)) { + //no reason to enable or update, bye + return; + } + + if (p_environment.is_valid() && environment_get_volumetric_fog_enabled(p_environment) && !p_render_buffers->has_custom_data(RB_SCOPE_FOG)) { + //required volumetric fog but not existing, create + Ref<RendererRD::Fog::VolumetricFog> fog; + + fog.instantiate(); + fog->init(Vector3i(target_width, target_height, get_volumetric_fog_depth()), sky.sky_shader.default_shader_rd); + + p_render_buffers->set_custom_data(RB_SCOPE_FOG, fog); + } + + if (p_render_buffers->has_custom_data(RB_SCOPE_FOG)) { + Ref<RendererRD::Fog::VolumetricFog> fog = p_render_buffers->get_custom_data(RB_SCOPE_FOG); + + RendererRD::Fog::VolumetricFogSettings settings; + settings.rb_size = size; + settings.time = time; + settings.is_using_radiance_cubemap_array = is_using_radiance_cubemap_array(); + settings.max_cluster_elements = RendererRD::LightStorage::get_singleton()->get_max_cluster_elements(); + settings.volumetric_fog_filter_active = get_volumetric_fog_filter_active(); + + settings.shadow_sampler = shadow_sampler; + settings.shadow_atlas_depth = RendererRD::LightStorage::get_singleton()->owns_shadow_atlas(p_shadow_atlas) ? RendererRD::LightStorage::get_singleton()->shadow_atlas_get_texture(p_shadow_atlas) : RID(); + settings.voxel_gi_buffer = rbgi->get_voxel_gi_buffer(); + settings.omni_light_buffer = RendererRD::LightStorage::get_singleton()->get_omni_light_buffer(); + settings.spot_light_buffer = RendererRD::LightStorage::get_singleton()->get_spot_light_buffer(); + settings.directional_shadow_depth = RendererRD::LightStorage::get_singleton()->directional_shadow_get_texture(); + settings.directional_light_buffer = RendererRD::LightStorage::get_singleton()->get_directional_light_buffer(); + + settings.vfog = fog; + settings.cluster_builder = rb_data->cluster_builder; + settings.rbgi = rbgi; + settings.sdfgi = sdfgi; + settings.env = p_environment; + settings.sky = &sky; + settings.gi = &gi; + + RendererRD::Fog::get_singleton()->volumetric_fog_update(settings, p_cam_projection, p_cam_transform, p_prev_cam_inv_transform, p_shadow_atlas, p_directional_light_count, p_use_directional_shadows, p_positional_light_count, p_voxel_gi_count, p_fog_volumes); + } +} + +/* Lighting */ + +void RenderForwardClustered::setup_added_reflection_probe(const Transform3D &p_transform, const Vector3 &p_half_extents) { + if (current_cluster_builder != nullptr) { + current_cluster_builder->add_box(ClusterBuilderRD::BOX_TYPE_REFLECTION_PROBE, p_transform, p_half_extents); + } +} + +void RenderForwardClustered::setup_added_light(const RS::LightType p_type, const Transform3D &p_transform, float p_radius, float p_spot_aperture) { + if (current_cluster_builder != nullptr) { + current_cluster_builder->add_light(p_type == RS::LIGHT_SPOT ? ClusterBuilderRD::LIGHT_TYPE_SPOT : ClusterBuilderRD::LIGHT_TYPE_OMNI, p_transform, p_radius, p_spot_aperture); + } +} + +void RenderForwardClustered::setup_added_decal(const Transform3D &p_transform, const Vector3 &p_half_extents) { + if (current_cluster_builder != nullptr) { + current_cluster_builder->add_box(ClusterBuilderRD::BOX_TYPE_DECAL, p_transform, p_half_extents); + } +} + +/* Render scene */ + +void RenderForwardClustered::_process_ssao(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_environment, RID p_normal_buffer, const Projection &p_projection) { + ERR_FAIL_NULL(ss_effects); + ERR_FAIL_COND(p_render_buffers.is_null()); + ERR_FAIL_COND(p_environment.is_null()); + + Ref<RenderBufferDataForwardClustered> rb_data = p_render_buffers->get_custom_data(RB_SCOPE_FORWARD_CLUSTERED); + ERR_FAIL_COND(rb_data.is_null()); + + RENDER_TIMESTAMP("Process SSAO"); + + RendererRD::SSEffects::SSAOSettings settings; + settings.radius = environment_get_ssao_radius(p_environment); + settings.intensity = environment_get_ssao_intensity(p_environment); + settings.power = environment_get_ssao_power(p_environment); + settings.detail = environment_get_ssao_detail(p_environment); + settings.horizon = environment_get_ssao_horizon(p_environment); + settings.sharpness = environment_get_ssao_sharpness(p_environment); + settings.full_screen_size = p_render_buffers->get_internal_size(); + + ss_effects->ssao_allocate_buffers(rb_data->ss_effects_data.ssao, settings, rb_data->ss_effects_data.linear_depth); + ss_effects->generate_ssao(rb_data->ss_effects_data.ssao, p_normal_buffer, p_projection, settings); +} + +void RenderForwardClustered::_process_ssil(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_environment, RID p_normal_buffer, const Projection &p_projection, const Transform3D &p_transform) { + ERR_FAIL_NULL(ss_effects); + ERR_FAIL_COND(p_render_buffers.is_null()); + ERR_FAIL_COND(p_environment.is_null()); + + Ref<RenderBufferDataForwardClustered> rb_data = p_render_buffers->get_custom_data(RB_SCOPE_FORWARD_CLUSTERED); + ERR_FAIL_COND(rb_data.is_null()); + + RENDER_TIMESTAMP("Process SSIL"); + + RendererRD::SSEffects::SSILSettings settings; + settings.radius = environment_get_ssil_radius(p_environment); + settings.intensity = environment_get_ssil_intensity(p_environment); + settings.sharpness = environment_get_ssil_sharpness(p_environment); + settings.normal_rejection = environment_get_ssil_normal_rejection(p_environment); + settings.full_screen_size = p_render_buffers->get_internal_size(); + + Projection correction; + correction.set_depth_correction(true); + Projection projection = correction * p_projection; + Transform3D transform = p_transform; + transform.set_origin(Vector3(0.0, 0.0, 0.0)); + Projection last_frame_projection = rb_data->ss_effects_data.last_frame_projection * Projection(rb_data->ss_effects_data.last_frame_transform.affine_inverse()) * Projection(transform) * projection.inverse(); + + ss_effects->ssil_allocate_buffers(rb_data->ss_effects_data.ssil, settings, rb_data->ss_effects_data.linear_depth); + ss_effects->screen_space_indirect_lighting(rb_data->ss_effects_data.ssil, p_normal_buffer, p_projection, last_frame_projection, settings); + rb_data->ss_effects_data.last_frame_projection = projection; + rb_data->ss_effects_data.last_frame_transform = transform; +} + +void RenderForwardClustered::_copy_framebuffer_to_ssil(Ref<RenderSceneBuffersRD> p_render_buffers) { + ERR_FAIL_COND(p_render_buffers.is_null()); + + Ref<RenderBufferDataForwardClustered> rb_data = p_render_buffers->get_custom_data(RB_SCOPE_FORWARD_CLUSTERED); + ERR_FAIL_COND(rb_data.is_null()); + + if (rb_data->ss_effects_data.ssil.last_frame.is_valid()) { + Size2i size = p_render_buffers->get_internal_size(); + RID texture = p_render_buffers->get_internal_texture(); + copy_effects->copy_to_rect(texture, rb_data->ss_effects_data.ssil.last_frame, Rect2i(0, 0, size.x, size.y)); + + int width = size.x; + int height = size.y; + for (int i = 0; i < rb_data->ss_effects_data.ssil.last_frame_slices.size() - 1; i++) { + width = MAX(1, width >> 1); + height = MAX(1, height >> 1); + copy_effects->make_mipmap(rb_data->ss_effects_data.ssil.last_frame_slices[i], rb_data->ss_effects_data.ssil.last_frame_slices[i + 1], Size2i(width, height)); + } + } +} + +void RenderForwardClustered::_pre_opaque_render(RenderDataRD *p_render_data, bool p_use_ssao, bool p_use_ssil, bool p_use_gi, const RID *p_normal_roughness_slices, RID p_voxel_gi_buffer) { + // Render shadows while GI is rendering, due to how barriers are handled, this should happen at the same time + RendererRD::LightStorage *light_storage = RendererRD::LightStorage::get_singleton(); + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + + Ref<RenderSceneBuffersRD> rb = p_render_data->render_buffers; + Ref<RenderBufferDataForwardClustered> rb_data; + if (rb.is_valid()) { + rb_data = rb->get_custom_data(RB_SCOPE_FORWARD_CLUSTERED); + } + + if (rb.is_valid() && p_use_gi && rb->has_custom_data(RB_SCOPE_SDFGI)) { + Ref<RendererRD::GI::SDFGI> sdfgi = rb->get_custom_data(RB_SCOPE_SDFGI); + sdfgi->store_probes(); + } + + p_render_data->cube_shadows.clear(); + p_render_data->shadows.clear(); + p_render_data->directional_shadows.clear(); + + Plane camera_plane(-p_render_data->scene_data->cam_transform.basis.get_column(Vector3::AXIS_Z), p_render_data->scene_data->cam_transform.origin); + float lod_distance_multiplier = p_render_data->scene_data->cam_projection.get_lod_multiplier(); + { + for (int i = 0; i < p_render_data->render_shadow_count; i++) { + RID li = p_render_data->render_shadows[i].light; + RID base = light_storage->light_instance_get_base_light(li); + + if (light_storage->light_get_type(base) == RS::LIGHT_DIRECTIONAL) { + p_render_data->directional_shadows.push_back(i); + } else if (light_storage->light_get_type(base) == RS::LIGHT_OMNI && light_storage->light_omni_get_shadow_mode(base) == RS::LIGHT_OMNI_SHADOW_CUBE) { + p_render_data->cube_shadows.push_back(i); + } else { + p_render_data->shadows.push_back(i); + } + } + + //cube shadows are rendered in their own way + for (uint32_t i = 0; i < p_render_data->cube_shadows.size(); i++) { + _render_shadow_pass(p_render_data->render_shadows[p_render_data->cube_shadows[i]].light, p_render_data->shadow_atlas, p_render_data->render_shadows[p_render_data->cube_shadows[i]].pass, p_render_data->render_shadows[p_render_data->cube_shadows[i]].instances, camera_plane, lod_distance_multiplier, p_render_data->scene_data->screen_mesh_lod_threshold, true, true, true, p_render_data->render_info); + } + + if (p_render_data->directional_shadows.size()) { + //open the pass for directional shadows + light_storage->update_directional_shadow_atlas(); + RD::get_singleton()->draw_list_begin(light_storage->direction_shadow_get_fb(), RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_CONTINUE); + RD::get_singleton()->draw_list_end(); + } + } + + // Render GI + + bool render_shadows = p_render_data->directional_shadows.size() || p_render_data->shadows.size(); + bool render_gi = rb.is_valid() && p_use_gi; + + if (render_shadows && render_gi) { + RENDER_TIMESTAMP("Render GI + Render Shadows (Parallel)"); + } else if (render_shadows) { + RENDER_TIMESTAMP("Render Shadows"); + } else if (render_gi) { + RENDER_TIMESTAMP("Render GI"); + } + + //prepare shadow rendering + if (render_shadows) { + _render_shadow_begin(); + + //render directional shadows + for (uint32_t i = 0; i < p_render_data->directional_shadows.size(); i++) { + _render_shadow_pass(p_render_data->render_shadows[p_render_data->directional_shadows[i]].light, p_render_data->shadow_atlas, p_render_data->render_shadows[p_render_data->directional_shadows[i]].pass, p_render_data->render_shadows[p_render_data->directional_shadows[i]].instances, camera_plane, lod_distance_multiplier, p_render_data->scene_data->screen_mesh_lod_threshold, false, i == p_render_data->directional_shadows.size() - 1, false, p_render_data->render_info); + } + //render positional shadows + for (uint32_t i = 0; i < p_render_data->shadows.size(); i++) { + _render_shadow_pass(p_render_data->render_shadows[p_render_data->shadows[i]].light, p_render_data->shadow_atlas, p_render_data->render_shadows[p_render_data->shadows[i]].pass, p_render_data->render_shadows[p_render_data->shadows[i]].instances, camera_plane, lod_distance_multiplier, p_render_data->scene_data->screen_mesh_lod_threshold, i == 0, i == p_render_data->shadows.size() - 1, true, p_render_data->render_info); + } + + _render_shadow_process(); + } + + //start GI + if (render_gi) { + gi.process_gi(rb, p_normal_roughness_slices, p_voxel_gi_buffer, p_render_data->environment, p_render_data->scene_data->view_count, p_render_data->scene_data->view_projection, p_render_data->scene_data->view_eye_offset, p_render_data->scene_data->cam_transform, *p_render_data->voxel_gi_instances); + } + + //Do shadow rendering (in parallel with GI) + if (render_shadows) { + _render_shadow_end(RD::BARRIER_MASK_NO_BARRIER); + } + + if (render_gi) { + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_NO_BARRIER); //use a later barrier + } + + if (rb_data.is_valid() && ss_effects) { + if (p_use_ssao || p_use_ssil) { + Size2i size = rb->get_internal_size(); + + bool invalidate_uniform_set = false; + if (rb_data->ss_effects_data.linear_depth.is_null()) { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R16_SFLOAT; + tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; + tf.width = (size.x + 1) / 2; + tf.height = (size.y + 1) / 2; + tf.mipmaps = 5; + tf.array_layers = 4; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + rb_data->ss_effects_data.linear_depth = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(rb_data->ss_effects_data.linear_depth, "SS Effects Depth"); + for (uint32_t i = 0; i < tf.mipmaps; i++) { + RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), rb_data->ss_effects_data.linear_depth, 0, i, 1, RD::TEXTURE_SLICE_2D_ARRAY); + rb_data->ss_effects_data.linear_depth_slices.push_back(slice); + RD::get_singleton()->set_resource_name(slice, "SS Effects Depth Mip " + itos(i) + " "); + } + invalidate_uniform_set = true; + } + + RID depth_texture = rb->get_depth_texture(); + ss_effects->downsample_depth(depth_texture, rb_data->ss_effects_data.linear_depth_slices, invalidate_uniform_set, size, p_render_data->scene_data->cam_projection); + } + + if (p_use_ssao) { + // TODO make these proper stereo + _process_ssao(rb, p_render_data->environment, p_normal_roughness_slices[0], p_render_data->scene_data->cam_projection); + } + + if (p_use_ssil) { + // TODO make these proper stereo + _process_ssil(rb, p_render_data->environment, p_normal_roughness_slices[0], p_render_data->scene_data->cam_projection, p_render_data->scene_data->cam_transform); + } + } + + //full barrier here, we need raster, transfer and compute and it depends from the previous work + RD::get_singleton()->barrier(RD::BARRIER_MASK_ALL, RD::BARRIER_MASK_ALL); + + if (current_cluster_builder) { + current_cluster_builder->begin(p_render_data->scene_data->cam_transform, p_render_data->scene_data->cam_projection, !p_render_data->reflection_probe.is_valid()); + } + + bool using_shadows = true; + + if (p_render_data->reflection_probe.is_valid()) { + if (!RSG::light_storage->reflection_probe_renders_shadows(light_storage->reflection_probe_instance_get_probe(p_render_data->reflection_probe))) { + using_shadows = false; + } + } else { + //do not render reflections when rendering a reflection probe + light_storage->update_reflection_probe_buffer(p_render_data, *p_render_data->reflection_probes, p_render_data->scene_data->cam_transform.affine_inverse(), p_render_data->environment); + } + + uint32_t directional_light_count = 0; + uint32_t positional_light_count = 0; + light_storage->update_light_buffers(p_render_data, *p_render_data->lights, p_render_data->scene_data->cam_transform, p_render_data->shadow_atlas, using_shadows, directional_light_count, positional_light_count, p_render_data->directional_light_soft_shadows); + texture_storage->update_decal_buffer(*p_render_data->decals, p_render_data->scene_data->cam_transform.affine_inverse()); + + p_render_data->directional_light_count = directional_light_count; + + if (current_cluster_builder) { + current_cluster_builder->bake_cluster(); + } + + if (rb.is_valid()) { + bool directional_shadows = RendererRD::LightStorage::get_singleton()->has_directional_shadows(directional_light_count); + _update_volumetric_fog(rb, p_render_data->environment, p_render_data->scene_data->cam_projection, p_render_data->scene_data->cam_transform, p_render_data->scene_data->prev_cam_transform.affine_inverse(), p_render_data->shadow_atlas, directional_light_count, directional_shadows, positional_light_count, p_render_data->voxel_gi_count, *p_render_data->fog_volumes); + } +} + +void RenderForwardClustered::_process_ssr(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_dest_framebuffer, const RID *p_normal_slices, RID p_specular_buffer, const RID *p_metallic_slices, RID p_environment, const Projection *p_projections, const Vector3 *p_eye_offsets, bool p_use_additive) { + ERR_FAIL_NULL(ss_effects); + ERR_FAIL_COND(p_render_buffers.is_null()); + + Ref<RenderBufferDataForwardClustered> rb_data = p_render_buffers->get_custom_data(RB_SCOPE_FORWARD_CLUSTERED); + ERR_FAIL_COND(rb_data.is_null()); + + Size2i internal_size = p_render_buffers->get_internal_size(); + bool can_use_effects = internal_size.x >= 8 && internal_size.y >= 8; + uint32_t view_count = p_render_buffers->get_view_count(); + + if (!can_use_effects) { + //just copy + copy_effects->merge_specular(p_dest_framebuffer, p_specular_buffer, p_use_additive ? RID() : p_render_buffers->get_internal_texture(), RID(), view_count); + return; + } + + ERR_FAIL_COND(p_environment.is_null()); + ERR_FAIL_COND(!environment_get_ssr_enabled(p_environment)); + + Size2i half_size = Size2i(internal_size.x / 2, internal_size.y / 2); + ss_effects->ssr_allocate_buffers(rb_data->ss_effects_data.ssr, _render_buffers_get_color_format(), half_size, view_count); + + RID texture_slices[RendererSceneRender::MAX_RENDER_VIEWS]; + RID depth_slices[RendererSceneRender::MAX_RENDER_VIEWS]; + for (uint32_t v = 0; v < view_count; v++) { + texture_slices[v] = p_render_buffers->get_internal_texture(v); + depth_slices[v] = p_render_buffers->get_depth_texture(v); + } + ss_effects->screen_space_reflection(rb_data->ss_effects_data.ssr, texture_slices, p_normal_slices, p_metallic_slices, depth_slices, half_size, environment_get_ssr_max_steps(p_environment), environment_get_ssr_fade_in(p_environment), environment_get_ssr_fade_out(p_environment), environment_get_ssr_depth_tolerance(p_environment), view_count, p_projections, p_eye_offsets); + copy_effects->merge_specular(p_dest_framebuffer, p_specular_buffer, p_use_additive ? RID() : p_render_buffers->get_internal_texture(), rb_data->ss_effects_data.ssr.output, view_count); +} + +void RenderForwardClustered::_process_sss(Ref<RenderSceneBuffersRD> p_render_buffers, const Projection &p_camera) { + ERR_FAIL_COND(p_render_buffers.is_null()); + + Size2i internal_size = p_render_buffers->get_internal_size(); + bool can_use_effects = internal_size.x >= 8 && internal_size.y >= 8; + + if (!can_use_effects) { + //just copy + return; + } + + p_render_buffers->allocate_blur_textures(); + + for (uint32_t v = 0; v < p_render_buffers->get_view_count(); v++) { + RID internal_texture = p_render_buffers->get_internal_texture(v); + RID depth_texture = p_render_buffers->get_depth_texture(v); + ss_effects->sub_surface_scattering(p_render_buffers, internal_texture, depth_texture, p_camera, internal_size); + } +} + +void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Color &p_default_bg_color) { + RendererRD::LightStorage *light_storage = RendererRD::LightStorage::get_singleton(); + + Ref<RenderSceneBuffersRD> rb; + Ref<RenderBufferDataForwardClustered> rb_data; + if (p_render_data && p_render_data->render_buffers.is_valid()) { + rb = p_render_data->render_buffers; + rb_data = rb->get_custom_data(RB_SCOPE_FORWARD_CLUSTERED); + } + static const int texture_multisamples[RS::VIEWPORT_MSAA_MAX] = { 1, 2, 4, 8 }; + + //first of all, make a new render pass + //fill up ubo + + RENDER_TIMESTAMP("Prepare 3D Scene"); + + // sdfgi first + _update_sdfgi(p_render_data); + + // assign render indices to voxel_gi_instances + for (uint32_t i = 0; i < (uint32_t)p_render_data->voxel_gi_instances->size(); i++) { + RID voxel_gi_instance = (*p_render_data->voxel_gi_instances)[i]; + gi.voxel_gi_instance_set_render_index(voxel_gi_instance, i); + } + + // obtain cluster builder + if (rb_data.is_valid()) { + current_cluster_builder = rb_data->cluster_builder; + } else if (light_storage->owns_reflection_probe_instance(p_render_data->reflection_probe)) { + current_cluster_builder = light_storage->reflection_probe_instance_get_cluster_builder(p_render_data->reflection_probe, &cluster_builder_shared); + + if (p_render_data->camera_attributes.is_valid()) { + light_storage->reflection_probe_set_baked_exposure(light_storage->reflection_probe_instance_get_probe(p_render_data->reflection_probe), RSG::camera_attributes->camera_attributes_get_exposure_normalization_factor(p_render_data->camera_attributes)); + } + } else { + ERR_PRINT("No render buffer nor reflection atlas, bug"); //should never happen, will crash + current_cluster_builder = nullptr; + } + + p_render_data->voxel_gi_count = 0; + + if (rb.is_valid()) { + if (rb->has_custom_data(RB_SCOPE_SDFGI)) { + Ref<RendererRD::GI::SDFGI> sdfgi = rb->get_custom_data(RB_SCOPE_SDFGI); + if (sdfgi.is_valid()) { + sdfgi->update_cascades(); + sdfgi->pre_process_gi(p_render_data->scene_data->cam_transform, p_render_data); + sdfgi->update_light(); + } + } + + gi.setup_voxel_gi_instances(p_render_data, p_render_data->render_buffers, p_render_data->scene_data->cam_transform, *p_render_data->voxel_gi_instances, p_render_data->voxel_gi_count); + } + + if (current_cluster_builder != nullptr) { + p_render_data->cluster_buffer = current_cluster_builder->get_cluster_buffer(); + p_render_data->cluster_size = current_cluster_builder->get_cluster_size(); + p_render_data->cluster_max_elements = current_cluster_builder->get_max_cluster_elements(); + } + + _update_vrs(rb); + + RENDER_TIMESTAMP("Setup 3D Scene"); + + // check if we need motion vectors + if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_MOTION_VECTORS) { + p_render_data->scene_data->calculate_motion_vectors = true; + } else if (rb.is_valid() && rb->get_use_taa()) { + p_render_data->scene_data->calculate_motion_vectors = true; + } else { + p_render_data->scene_data->calculate_motion_vectors = false; + } + + //p_render_data->scene_data->subsurface_scatter_width = subsurface_scatter_size; + p_render_data->scene_data->directional_light_count = 0; + p_render_data->scene_data->opaque_prepass_threshold = 0.99f; + + Size2i screen_size; + RID color_framebuffer; + RID color_only_framebuffer; + RID depth_framebuffer; + + PassMode depth_pass_mode = PASS_MODE_DEPTH; + uint32_t color_pass_flags = 0; + Vector<Color> depth_pass_clear; + bool using_separate_specular = false; + bool using_ssr = false; + bool using_sdfgi = false; + bool using_voxelgi = false; + bool reverse_cull = false; + bool using_ssil = p_render_data->environment.is_valid() && environment_get_ssil_enabled(p_render_data->environment); + + if (rb.is_valid()) { + screen_size = rb->get_internal_size(); + + if (rb->get_use_taa() || get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_MOTION_VECTORS) { + color_pass_flags |= COLOR_PASS_FLAG_MOTION_VECTORS; + } + + if (p_render_data->voxel_gi_instances->size() > 0) { + using_voxelgi = true; + } + + if (p_render_data->environment.is_null() && using_voxelgi) { + depth_pass_mode = PASS_MODE_DEPTH_NORMAL_ROUGHNESS_VOXEL_GI; + } else if (p_render_data->environment.is_valid() && (environment_get_ssr_enabled(p_render_data->environment) || environment_get_sdfgi_enabled(p_render_data->environment) || using_voxelgi)) { + if (environment_get_sdfgi_enabled(p_render_data->environment)) { + depth_pass_mode = using_voxelgi ? PASS_MODE_DEPTH_NORMAL_ROUGHNESS_VOXEL_GI : PASS_MODE_DEPTH_NORMAL_ROUGHNESS; // also voxelgi + using_sdfgi = true; + } else { + depth_pass_mode = using_voxelgi ? PASS_MODE_DEPTH_NORMAL_ROUGHNESS_VOXEL_GI : PASS_MODE_DEPTH_NORMAL_ROUGHNESS; + } + if (environment_get_ssr_enabled(p_render_data->environment)) { + using_separate_specular = true; + using_ssr = true; + color_pass_flags |= COLOR_PASS_FLAG_SEPARATE_SPECULAR; + } + } else if (p_render_data->environment.is_valid() && (environment_get_ssao_enabled(p_render_data->environment) || using_ssil || get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_NORMAL_BUFFER)) { + depth_pass_mode = PASS_MODE_DEPTH_NORMAL_ROUGHNESS; + } + + switch (depth_pass_mode) { + case PASS_MODE_DEPTH: { + depth_framebuffer = rb_data->get_depth_fb(); + } break; + case PASS_MODE_DEPTH_NORMAL_ROUGHNESS: { + depth_framebuffer = rb_data->get_depth_fb(RenderBufferDataForwardClustered::DEPTH_FB_ROUGHNESS); + depth_pass_clear.push_back(Color(0.5, 0.5, 0.5, 0)); + } break; + case PASS_MODE_DEPTH_NORMAL_ROUGHNESS_VOXEL_GI: { + depth_framebuffer = rb_data->get_depth_fb(RenderBufferDataForwardClustered::DEPTH_FB_ROUGHNESS_VOXELGI); + depth_pass_clear.push_back(Color(0.5, 0.5, 0.5, 0)); + depth_pass_clear.push_back(Color(0, 0, 0, 0)); + } break; + default: { + }; + } + + if (p_render_data->scene_data->view_count > 1) { + color_pass_flags |= COLOR_PASS_FLAG_MULTIVIEW; + } + + color_framebuffer = rb_data->get_color_pass_fb(color_pass_flags); + color_only_framebuffer = rb_data->get_color_only_fb(); + } else if (p_render_data->reflection_probe.is_valid()) { + uint32_t resolution = light_storage->reflection_probe_instance_get_resolution(p_render_data->reflection_probe); + screen_size.x = resolution; + screen_size.y = resolution; + + color_framebuffer = light_storage->reflection_probe_instance_get_framebuffer(p_render_data->reflection_probe, p_render_data->reflection_probe_pass); + color_only_framebuffer = color_framebuffer; + depth_framebuffer = light_storage->reflection_probe_instance_get_depth_framebuffer(p_render_data->reflection_probe, p_render_data->reflection_probe_pass); + + if (light_storage->reflection_probe_is_interior(light_storage->reflection_probe_instance_get_probe(p_render_data->reflection_probe))) { + p_render_data->environment = RID(); //no environment on interiors + } + + reverse_cull = true; // for some reason our views are inverted + } else { + ERR_FAIL(); //bug? + } + + p_render_data->scene_data->emissive_exposure_normalization = -1.0; + + RD::get_singleton()->draw_command_begin_label("Render Setup"); + + _setup_lightmaps(p_render_data, *p_render_data->lightmaps, p_render_data->scene_data->cam_transform); + _setup_voxelgis(*p_render_data->voxel_gi_instances); + _setup_environment(p_render_data, p_render_data->reflection_probe.is_valid(), screen_size, !p_render_data->reflection_probe.is_valid(), p_default_bg_color, false); + + _update_render_base_uniform_set(); //may have changed due to the above (light buffer enlarged, as an example) + + _fill_render_list(RENDER_LIST_OPAQUE, p_render_data, PASS_MODE_COLOR, color_pass_flags, using_sdfgi, using_sdfgi || using_voxelgi); + render_list[RENDER_LIST_OPAQUE].sort_by_key(); + render_list[RENDER_LIST_ALPHA].sort_by_reverse_depth_and_priority(); + _fill_instance_data(RENDER_LIST_OPAQUE, p_render_data->render_info ? p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_VISIBLE] : (int *)nullptr); + _fill_instance_data(RENDER_LIST_ALPHA); + + RD::get_singleton()->draw_command_end_label(); + + bool using_sss = rb_data.is_valid() && scene_state.used_sss && ss_effects->sss_get_quality() != RS::SUB_SURFACE_SCATTERING_QUALITY_DISABLED; + + if (using_sss && !using_separate_specular) { + using_separate_specular = true; + color_pass_flags |= COLOR_PASS_FLAG_SEPARATE_SPECULAR; + color_framebuffer = rb_data->get_color_pass_fb(color_pass_flags); + } + RID radiance_texture; + bool draw_sky = false; + bool draw_sky_fog_only = false; + // We invert luminance_multiplier for sky so that we can combine it with exposure value. + float sky_energy_multiplier = 1.0 / _render_buffers_get_luminance_multiplier(); + + Color clear_color; + bool keep_color = false; + + if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_OVERDRAW) { + clear_color = Color(0, 0, 0, 1); //in overdraw mode, BG should always be black + } else if (is_environment(p_render_data->environment)) { + RS::EnvironmentBG bg_mode = environment_get_background(p_render_data->environment); + float bg_energy_multiplier = environment_get_bg_energy_multiplier(p_render_data->environment); + bg_energy_multiplier *= environment_get_bg_intensity(p_render_data->environment); + + if (p_render_data->camera_attributes.is_valid()) { + bg_energy_multiplier *= RSG::camera_attributes->camera_attributes_get_exposure_normalization_factor(p_render_data->camera_attributes); + } + + switch (bg_mode) { + case RS::ENV_BG_CLEAR_COLOR: { + clear_color = p_default_bg_color; + clear_color.r *= bg_energy_multiplier; + clear_color.g *= bg_energy_multiplier; + clear_color.b *= bg_energy_multiplier; + if ((rb.is_valid() && rb->has_custom_data(RB_SCOPE_FOG)) || environment_get_fog_enabled(p_render_data->environment)) { + draw_sky_fog_only = true; + RendererRD::MaterialStorage::get_singleton()->material_set_param(sky.sky_scene_state.fog_material, "clear_color", Variant(clear_color.srgb_to_linear())); + } + } break; + case RS::ENV_BG_COLOR: { + clear_color = environment_get_bg_color(p_render_data->environment); + clear_color.r *= bg_energy_multiplier; + clear_color.g *= bg_energy_multiplier; + clear_color.b *= bg_energy_multiplier; + if ((rb.is_valid() && rb->has_custom_data(RB_SCOPE_FOG)) || environment_get_fog_enabled(p_render_data->environment)) { + draw_sky_fog_only = true; + RendererRD::MaterialStorage::get_singleton()->material_set_param(sky.sky_scene_state.fog_material, "clear_color", Variant(clear_color.srgb_to_linear())); + } + } break; + case RS::ENV_BG_SKY: { + draw_sky = true; + } break; + case RS::ENV_BG_CANVAS: { + keep_color = true; + } break; + case RS::ENV_BG_KEEP: { + keep_color = true; + } break; + case RS::ENV_BG_CAMERA_FEED: { + } break; + default: { + } + } + // setup sky if used for ambient, reflections, or background + if (draw_sky || draw_sky_fog_only || environment_get_reflection_source(p_render_data->environment) == RS::ENV_REFLECTION_SOURCE_SKY || environment_get_ambient_source(p_render_data->environment) == RS::ENV_AMBIENT_SOURCE_SKY) { + RENDER_TIMESTAMP("Setup Sky"); + RD::get_singleton()->draw_command_begin_label("Setup Sky"); + Projection projection = p_render_data->scene_data->cam_projection; + if (p_render_data->reflection_probe.is_valid()) { + Projection correction; + correction.set_depth_correction(true); + projection = correction * p_render_data->scene_data->cam_projection; + } + + sky.setup(p_render_data->environment, rb, *p_render_data->lights, p_render_data->camera_attributes, projection, p_render_data->scene_data->cam_transform, screen_size, this); + + sky_energy_multiplier *= bg_energy_multiplier; + + RID sky_rid = environment_get_sky(p_render_data->environment); + if (sky_rid.is_valid()) { + sky.update(p_render_data->environment, projection, p_render_data->scene_data->cam_transform, time, sky_energy_multiplier); + radiance_texture = sky.sky_get_radiance_texture_rd(sky_rid); + } else { + // do not try to draw sky if invalid + draw_sky = false; + } + RD::get_singleton()->draw_command_end_label(); + } + } else { + clear_color = p_default_bg_color; + } + + bool debug_voxelgis = get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_VOXEL_GI_ALBEDO || get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_VOXEL_GI_LIGHTING || get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_VOXEL_GI_EMISSION; + bool debug_sdfgi_probes = get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_SDFGI_PROBES; + bool depth_pre_pass = bool(GLOBAL_GET("rendering/driver/depth_prepass/enable")) && depth_framebuffer.is_valid(); + + bool using_ssao = depth_pre_pass && rb.is_valid() && p_render_data->environment.is_valid() && environment_get_ssao_enabled(p_render_data->environment); + bool continue_depth = false; + if (depth_pre_pass) { //depth pre pass + + bool needs_pre_resolve = _needs_post_prepass_render(p_render_data, using_sdfgi || using_voxelgi); + if (needs_pre_resolve) { + RENDER_TIMESTAMP("GI + Render Depth Pre-Pass (Parallel)"); + } else { + RENDER_TIMESTAMP("Render Depth Pre-Pass"); + } + if (needs_pre_resolve) { + //pre clear the depth framebuffer, as AMD (and maybe others?) use compute for it, and barrier other compute shaders. + RD::get_singleton()->draw_list_begin(depth_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_CONTINUE, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_CONTINUE, depth_pass_clear); + RD::get_singleton()->draw_list_end(); + //start compute processes here, so they run at the same time as depth pre-pass + _post_prepass_render(p_render_data, using_sdfgi || using_voxelgi); + } + + RD::get_singleton()->draw_command_begin_label("Render Depth Pre-Pass"); + + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_OPAQUE, nullptr, RID()); + + bool finish_depth = using_ssao || using_sdfgi || using_voxelgi; + RenderListParameters render_list_params(render_list[RENDER_LIST_OPAQUE].elements.ptr(), render_list[RENDER_LIST_OPAQUE].element_info.ptr(), render_list[RENDER_LIST_OPAQUE].elements.size(), reverse_cull, depth_pass_mode, 0, rb_data.is_null(), p_render_data->directional_light_soft_shadows, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->scene_data->lod_distance_multiplier, p_render_data->scene_data->screen_mesh_lod_threshold, p_render_data->scene_data->view_count); + _render_list_with_threads(&render_list_params, depth_framebuffer, needs_pre_resolve ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, needs_pre_resolve ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_CLEAR, finish_depth ? RD::FINAL_ACTION_READ : RD::FINAL_ACTION_CONTINUE, needs_pre_resolve ? Vector<Color>() : depth_pass_clear); + + RD::get_singleton()->draw_command_end_label(); + + if (needs_pre_resolve) { + _pre_resolve_render(p_render_data, using_sdfgi || using_voxelgi); + } + + if (rb.is_valid() && rb->get_msaa_3d() != RS::VIEWPORT_MSAA_DISABLED) { + RENDER_TIMESTAMP("Resolve Depth Pre-Pass (MSAA)"); + RD::get_singleton()->draw_command_begin_label("Resolve Depth Pre-Pass (MSAA)"); + if (depth_pass_mode == PASS_MODE_DEPTH_NORMAL_ROUGHNESS || depth_pass_mode == PASS_MODE_DEPTH_NORMAL_ROUGHNESS_VOXEL_GI) { + if (needs_pre_resolve) { + RD::get_singleton()->barrier(RD::BARRIER_MASK_RASTER, RD::BARRIER_MASK_COMPUTE); + } + for (uint32_t v = 0; v < rb->get_view_count(); v++) { + resolve_effects->resolve_gi(rb_data->get_depth_msaa(v), rb_data->get_normal_roughness_msaa(v), using_voxelgi ? rb_data->get_voxelgi_msaa(v) : RID(), rb->get_depth_texture(v), rb_data->get_normal_roughness(v), using_voxelgi ? rb_data->get_voxelgi(v) : RID(), rb->get_internal_size(), texture_multisamples[rb->get_msaa_3d()]); + } + } else if (finish_depth) { + for (uint32_t v = 0; v < rb->get_view_count(); v++) { + resolve_effects->resolve_depth(rb_data->get_depth_msaa(v), rb->get_depth_texture(v), rb->get_internal_size(), texture_multisamples[rb->get_msaa_3d()]); + } + } + RD::get_singleton()->draw_command_end_label(); + } + + continue_depth = !finish_depth; + } + + RID normal_roughness_views[RendererSceneRender::MAX_RENDER_VIEWS]; + if (rb_data.is_valid() && rb_data->has_normal_roughness()) { + for (uint32_t v = 0; v < rb->get_view_count(); v++) { + normal_roughness_views[v] = rb_data->get_normal_roughness(v); + } + } + _pre_opaque_render(p_render_data, using_ssao, using_ssil, using_sdfgi || using_voxelgi, normal_roughness_views, rb_data.is_valid() && rb_data->has_voxelgi() ? rb_data->get_voxelgi() : RID()); + + RD::get_singleton()->draw_command_begin_label("Render Opaque Pass"); + + p_render_data->scene_data->directional_light_count = p_render_data->directional_light_count; + p_render_data->scene_data->opaque_prepass_threshold = 0.0f; + + _setup_environment(p_render_data, p_render_data->reflection_probe.is_valid(), screen_size, !p_render_data->reflection_probe.is_valid(), p_default_bg_color, rb.is_valid()); + + RENDER_TIMESTAMP("Render Opaque Pass"); + + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_OPAQUE, p_render_data, radiance_texture, true); + + bool can_continue_color = !scene_state.used_screen_texture && !using_ssr && !using_sss; + bool can_continue_depth = !scene_state.used_depth_texture && !using_ssr && !using_sss; + + { + bool will_continue_color = (can_continue_color || draw_sky || draw_sky_fog_only || debug_voxelgis || debug_sdfgi_probes); + bool will_continue_depth = (can_continue_depth || draw_sky || draw_sky_fog_only || debug_voxelgis || debug_sdfgi_probes); + + Vector<Color> c; + { + Color cc = clear_color.srgb_to_linear(); + if (using_separate_specular || rb_data.is_valid()) { + cc.a = 0; //subsurf scatter must be 0 + } + c.push_back(cc); + + if (rb_data.is_valid()) { + c.push_back(Color(0, 0, 0, 0)); // Separate specular + c.push_back(Color(0, 0, 0, 0)); // Motion vectors + } + } + + RenderListParameters render_list_params(render_list[RENDER_LIST_OPAQUE].elements.ptr(), render_list[RENDER_LIST_OPAQUE].element_info.ptr(), render_list[RENDER_LIST_OPAQUE].elements.size(), reverse_cull, PASS_MODE_COLOR, color_pass_flags, rb_data.is_null(), p_render_data->directional_light_soft_shadows, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->scene_data->lod_distance_multiplier, p_render_data->scene_data->screen_mesh_lod_threshold, p_render_data->scene_data->view_count); + _render_list_with_threads(&render_list_params, color_framebuffer, keep_color ? RD::INITIAL_ACTION_KEEP : RD::INITIAL_ACTION_CLEAR, will_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, depth_pre_pass ? (continue_depth ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP) : RD::INITIAL_ACTION_CLEAR, will_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, c, 1.0, 0); + if (will_continue_color && using_separate_specular) { + // close the specular framebuffer, as it's no longer used + RD::get_singleton()->draw_list_begin(rb_data->get_specular_only_fb(), RD::INITIAL_ACTION_CONTINUE, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CONTINUE, RD::FINAL_ACTION_CONTINUE); + RD::get_singleton()->draw_list_end(); + } + } + + RD::get_singleton()->draw_command_end_label(); + + if (debug_voxelgis) { + //debug voxelgis + bool will_continue_color = (can_continue_color || draw_sky || draw_sky_fog_only); + bool will_continue_depth = (can_continue_depth || draw_sky || draw_sky_fog_only); + + Projection dc; + dc.set_depth_correction(true); + Projection cm = (dc * p_render_data->scene_data->cam_projection) * Projection(p_render_data->scene_data->cam_transform.affine_inverse()); + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(color_only_framebuffer, RD::INITIAL_ACTION_CONTINUE, will_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CONTINUE, will_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ); + RD::get_singleton()->draw_command_begin_label("Debug VoxelGIs"); + for (int i = 0; i < (int)p_render_data->voxel_gi_instances->size(); i++) { + gi.debug_voxel_gi((*p_render_data->voxel_gi_instances)[i], draw_list, color_only_framebuffer, cm, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_VOXEL_GI_LIGHTING, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_VOXEL_GI_EMISSION, 1.0); + } + RD::get_singleton()->draw_command_end_label(); + RD::get_singleton()->draw_list_end(); + } + + if (debug_sdfgi_probes) { + //debug sdfgi + bool will_continue_color = (can_continue_color || draw_sky || draw_sky_fog_only); + bool will_continue_depth = (can_continue_depth || draw_sky || draw_sky_fog_only); + + Projection dc; + dc.set_depth_correction(true); + Projection cms[RendererSceneRender::MAX_RENDER_VIEWS]; + for (uint32_t v = 0; v < p_render_data->scene_data->view_count; v++) { + cms[v] = (dc * p_render_data->scene_data->view_projection[v]) * Projection(p_render_data->scene_data->cam_transform.affine_inverse()); + } + _debug_sdfgi_probes(rb, color_only_framebuffer, p_render_data->scene_data->view_count, cms, will_continue_color, will_continue_depth); + } + + if (draw_sky || draw_sky_fog_only) { + RENDER_TIMESTAMP("Render Sky"); + + RD::get_singleton()->draw_command_begin_label("Draw Sky"); + + if (p_render_data->reflection_probe.is_valid()) { + Projection correction; + correction.set_depth_correction(true); + Projection projection = correction * p_render_data->scene_data->cam_projection; + sky.draw(p_render_data->environment, can_continue_color, can_continue_depth, color_only_framebuffer, 1, &projection, p_render_data->scene_data->cam_transform, time, sky_energy_multiplier); + } else { + sky.draw(p_render_data->environment, can_continue_color, can_continue_depth, color_only_framebuffer, p_render_data->scene_data->view_count, p_render_data->scene_data->view_projection, p_render_data->scene_data->cam_transform, time, sky_energy_multiplier); + } + RD::get_singleton()->draw_command_end_label(); + } + + if (rb.is_valid() && !can_continue_color && rb->get_msaa_3d() != RS::VIEWPORT_MSAA_DISABLED) { + // Handle views individual, might want to look at rewriting our resolve to do both layers in one pass. + for (uint32_t v = 0; v < rb->get_view_count(); v++) { + RD::get_singleton()->texture_resolve_multisample(rb_data->get_color_msaa(v), rb->get_internal_texture(v)); + } + if (using_separate_specular) { + for (uint32_t v = 0; v < rb->get_view_count(); v++) { + RD::get_singleton()->texture_resolve_multisample(rb_data->get_specular_msaa(v), rb_data->get_specular(v)); + } + } + } + + if (rb.is_valid() && !can_continue_depth && rb->get_msaa_3d() != RS::VIEWPORT_MSAA_DISABLED) { + for (uint32_t v = 0; v < rb->get_view_count(); v++) { + resolve_effects->resolve_depth(rb_data->get_depth_msaa(v), rb->get_depth_texture(v), rb->get_internal_size(), texture_multisamples[rb->get_msaa_3d()]); + } + } + + if (using_separate_specular) { + if (using_sss) { + RENDER_TIMESTAMP("Sub-Surface Scattering"); + RD::get_singleton()->draw_command_begin_label("Process Sub-Surface Scattering"); + _process_sss(rb, p_render_data->scene_data->cam_projection); + RD::get_singleton()->draw_command_end_label(); + } + + if (using_ssr) { + RENDER_TIMESTAMP("Screen-Space Reflections"); + RD::get_singleton()->draw_command_begin_label("Process Screen-Space Reflections"); + RID specular_views[RendererSceneRender::MAX_RENDER_VIEWS]; + for (uint32_t v = 0; v < p_render_data->scene_data->view_count; v++) { + specular_views[v] = rb_data->get_specular(v); + } + _process_ssr(rb, color_only_framebuffer, normal_roughness_views, rb_data->get_specular(), specular_views, p_render_data->environment, p_render_data->scene_data->view_projection, p_render_data->scene_data->view_eye_offset, rb->get_msaa_3d() == RS::VIEWPORT_MSAA_DISABLED); + RD::get_singleton()->draw_command_end_label(); + } else { + //just mix specular back + RENDER_TIMESTAMP("Merge Specular"); + copy_effects->merge_specular(color_only_framebuffer, rb_data->get_specular(), rb->get_msaa_3d() == RS::VIEWPORT_MSAA_DISABLED ? RID() : rb->get_internal_texture(), RID(), p_render_data->scene_data->view_count); + } + } + + if (scene_state.used_screen_texture) { + // Copy screen texture to backbuffer so we can read from it + _render_buffers_copy_screen_texture(p_render_data); + } + + if (scene_state.used_depth_texture) { + // Copy depth texture to backbuffer so we can read from it + _render_buffers_copy_depth_texture(p_render_data); + } + + RENDER_TIMESTAMP("Render 3D Transparent Pass"); + + RD::get_singleton()->draw_command_begin_label("Render 3D Transparent Pass"); + + rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_ALPHA, p_render_data, radiance_texture, true); + + _setup_environment(p_render_data, p_render_data->reflection_probe.is_valid(), screen_size, !p_render_data->reflection_probe.is_valid(), p_default_bg_color, false); + + { + uint32_t transparent_color_pass_flags = (color_pass_flags | COLOR_PASS_FLAG_TRANSPARENT) & ~(COLOR_PASS_FLAG_SEPARATE_SPECULAR); + RID alpha_framebuffer = rb_data.is_valid() ? rb_data->get_color_pass_fb(transparent_color_pass_flags) : color_only_framebuffer; + RenderListParameters render_list_params(render_list[RENDER_LIST_ALPHA].elements.ptr(), render_list[RENDER_LIST_ALPHA].element_info.ptr(), render_list[RENDER_LIST_ALPHA].elements.size(), false, PASS_MODE_COLOR, transparent_color_pass_flags, rb_data.is_null(), p_render_data->directional_light_soft_shadows, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->scene_data->lod_distance_multiplier, p_render_data->scene_data->screen_mesh_lod_threshold, p_render_data->scene_data->view_count); + _render_list_with_threads(&render_list_params, alpha_framebuffer, can_continue_color ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, can_continue_depth ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ); + } + + RD::get_singleton()->draw_command_end_label(); + + RENDER_TIMESTAMP("Resolve"); + + RD::get_singleton()->draw_command_begin_label("Resolve"); + + if (rb.is_valid() && rb->get_msaa_3d() != RS::VIEWPORT_MSAA_DISABLED) { + for (uint32_t v = 0; v < rb->get_view_count(); v++) { + RD::get_singleton()->texture_resolve_multisample(rb_data->get_color_msaa(v), rb->get_internal_texture(v)); + resolve_effects->resolve_depth(rb_data->get_depth_msaa(v), rb->get_depth_texture(v), rb->get_internal_size(), texture_multisamples[rb->get_msaa_3d()]); + } + if (taa && rb->get_use_taa()) { + taa->msaa_resolve(rb); + } + } + + RD::get_singleton()->draw_command_end_label(); + + RD::get_singleton()->draw_command_begin_label("Copy framebuffer for SSIL"); + if (using_ssil) { + RENDER_TIMESTAMP("Copy Final Framebuffer (SSIL)"); + _copy_framebuffer_to_ssil(rb); + } + RD::get_singleton()->draw_command_end_label(); + + if (rb.is_valid() && taa && rb->get_use_taa()) { + RENDER_TIMESTAMP("TAA") + taa->process(rb, _render_buffers_get_color_format(), p_render_data->scene_data->z_near, p_render_data->scene_data->z_far); + } + + if (rb.is_valid()) { + _debug_draw_cluster(rb); + + RENDER_TIMESTAMP("Tonemap"); + + _render_buffers_post_process_and_tonemap(p_render_data); + } + + if (rb.is_valid()) { + _render_buffers_debug_draw(rb, p_render_data->shadow_atlas, p_render_data->occluder_debug_tex); + + if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_SDFGI && rb->has_custom_data(RB_SCOPE_SDFGI)) { + Ref<RendererRD::GI::SDFGI> sdfgi = rb->get_custom_data(RB_SCOPE_SDFGI); + Vector<RID> view_rids; + + // SDFGI renders at internal resolution, need to check if our debug correctly supports outputting upscaled. + Size2i size = rb->get_internal_size(); + RID source_texture = rb->get_internal_texture(); + for (uint32_t v = 0; v < rb->get_view_count(); v++) { + view_rids.push_back(rb->get_internal_texture(v)); + } + + sdfgi->debug_draw(p_render_data->scene_data->view_count, p_render_data->scene_data->view_projection, p_render_data->scene_data->cam_transform, size.x, size.y, rb->get_render_target(), source_texture, view_rids); + } + } +} + +void RenderForwardClustered::_render_buffers_debug_draw(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_shadow_atlas, RID p_occlusion_buffer) { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + ERR_FAIL_COND(p_render_buffers.is_null()); + + Ref<RenderBufferDataForwardClustered> rb_data = p_render_buffers->get_custom_data(RB_SCOPE_FORWARD_CLUSTERED); + ERR_FAIL_COND(rb_data.is_null()); + + RendererSceneRenderRD::_render_buffers_debug_draw(p_render_buffers, p_shadow_atlas, p_occlusion_buffer); + + RID render_target = p_render_buffers->get_render_target(); + + if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_SSAO && rb_data->ss_effects_data.ssao.ao_final.is_valid()) { + Size2i rtsize = texture_storage->render_target_get_size(render_target); + copy_effects->copy_to_fb_rect(rb_data->ss_effects_data.ssao.ao_final, texture_storage->render_target_get_rd_framebuffer(render_target), Rect2(Vector2(), rtsize), false, true); + } + + if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_SSIL && rb_data->ss_effects_data.ssil.ssil_final.is_valid()) { + Size2i rtsize = texture_storage->render_target_get_size(render_target); + copy_effects->copy_to_fb_rect(rb_data->ss_effects_data.ssil.ssil_final, texture_storage->render_target_get_rd_framebuffer(render_target), Rect2(Vector2(), rtsize), false, false); + } + + if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_BUFFER && p_render_buffers->has_texture(RB_SCOPE_GI, RB_TEX_AMBIENT)) { + Size2i rtsize = texture_storage->render_target_get_size(render_target); + RID ambient_texture = p_render_buffers->get_texture(RB_SCOPE_GI, RB_TEX_AMBIENT); + RID reflection_texture = p_render_buffers->get_texture(RB_SCOPE_GI, RB_TEX_REFLECTION); + copy_effects->copy_to_fb_rect(ambient_texture, texture_storage->render_target_get_rd_framebuffer(render_target), Rect2(Vector2(), rtsize), false, false, false, true, reflection_texture, p_render_buffers->get_view_count() > 1); + } +} + +void RenderForwardClustered::_render_shadow_pass(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<RenderGeometryInstance *> &p_instances, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_mesh_lod_threshold, bool p_open_pass, bool p_close_pass, bool p_clear_region, RenderingMethod::RenderInfo *p_render_info) { + RendererRD::LightStorage *light_storage = RendererRD::LightStorage::get_singleton(); + + ERR_FAIL_COND(!light_storage->owns_light_instance(p_light)); + + RID base = light_storage->light_instance_get_base_light(p_light); + + Rect2i atlas_rect; + uint32_t atlas_size = 1; + RID atlas_fb; + + bool using_dual_paraboloid = false; + bool using_dual_paraboloid_flip = false; + Vector2i dual_paraboloid_offset; + RID render_fb; + RID render_texture; + float zfar; + + bool use_pancake = false; + bool render_cubemap = false; + bool finalize_cubemap = false; + + bool flip_y = false; + + Projection light_projection; + Transform3D light_transform; + + if (light_storage->light_get_type(base) == RS::LIGHT_DIRECTIONAL) { + //set pssm stuff + uint64_t last_scene_shadow_pass = light_storage->light_instance_get_shadow_pass(p_light); + if (last_scene_shadow_pass != get_scene_pass()) { + light_storage->light_instance_set_directional_rect(p_light, light_storage->get_directional_shadow_rect()); + light_storage->directional_shadow_increase_current_light(); + light_storage->light_instance_set_shadow_pass(p_light, get_scene_pass()); + } + + use_pancake = light_storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_PANCAKE_SIZE) > 0; + light_projection = light_storage->light_instance_get_shadow_camera(p_light, p_pass); + light_transform = light_storage->light_instance_get_shadow_transform(p_light, p_pass); + + atlas_rect = light_storage->light_instance_get_directional_rect(p_light); + + if (light_storage->light_directional_get_shadow_mode(base) == RS::LIGHT_DIRECTIONAL_SHADOW_PARALLEL_4_SPLITS) { + atlas_rect.size.width /= 2; + atlas_rect.size.height /= 2; + + if (p_pass == 1) { + atlas_rect.position.x += atlas_rect.size.width; + } else if (p_pass == 2) { + atlas_rect.position.y += atlas_rect.size.height; + } else if (p_pass == 3) { + atlas_rect.position += atlas_rect.size; + } + } else if (light_storage->light_directional_get_shadow_mode(base) == RS::LIGHT_DIRECTIONAL_SHADOW_PARALLEL_2_SPLITS) { + atlas_rect.size.height /= 2; + + if (p_pass == 0) { + } else { + atlas_rect.position.y += atlas_rect.size.height; + } + } + + float directional_shadow_size = light_storage->directional_shadow_get_size(); + Rect2 atlas_rect_norm = atlas_rect; + atlas_rect_norm.position /= directional_shadow_size; + atlas_rect_norm.size /= directional_shadow_size; + light_storage->light_instance_set_directional_shadow_atlas_rect(p_light, p_pass, atlas_rect_norm); + + zfar = RSG::light_storage->light_get_param(base, RS::LIGHT_PARAM_RANGE); + + render_fb = light_storage->direction_shadow_get_fb(); + render_texture = RID(); + flip_y = true; + + } else { + //set from shadow atlas + + ERR_FAIL_COND(!light_storage->owns_shadow_atlas(p_shadow_atlas)); + ERR_FAIL_COND(!light_storage->shadow_atlas_owns_light_instance(p_shadow_atlas, p_light)); + + RSG::light_storage->shadow_atlas_update(p_shadow_atlas); + + uint32_t key = light_storage->shadow_atlas_get_light_instance_key(p_shadow_atlas, p_light); + + uint32_t quadrant = (key >> RendererRD::LightStorage::QUADRANT_SHIFT) & 0x3; + uint32_t shadow = key & RendererRD::LightStorage::SHADOW_INDEX_MASK; + uint32_t subdivision = light_storage->shadow_atlas_get_quadrant_subdivision(p_shadow_atlas, quadrant); + + ERR_FAIL_INDEX((int)shadow, light_storage->shadow_atlas_get_quadrant_shadow_size(p_shadow_atlas, quadrant)); + + uint32_t shadow_atlas_size = light_storage->shadow_atlas_get_size(p_shadow_atlas); + uint32_t quadrant_size = shadow_atlas_size >> 1; + + atlas_rect.position.x = (quadrant & 1) * quadrant_size; + atlas_rect.position.y = (quadrant >> 1) * quadrant_size; + + uint32_t shadow_size = (quadrant_size / subdivision); + atlas_rect.position.x += (shadow % subdivision) * shadow_size; + atlas_rect.position.y += (shadow / subdivision) * shadow_size; + + atlas_rect.size.width = shadow_size; + atlas_rect.size.height = shadow_size; + + zfar = light_storage->light_get_param(base, RS::LIGHT_PARAM_RANGE); + + if (light_storage->light_get_type(base) == RS::LIGHT_OMNI) { + bool wrap = (shadow + 1) % subdivision == 0; + dual_paraboloid_offset = wrap ? Vector2i(1 - subdivision, 1) : Vector2i(1, 0); + + if (light_storage->light_omni_get_shadow_mode(base) == RS::LIGHT_OMNI_SHADOW_CUBE) { + render_texture = light_storage->get_cubemap(shadow_size / 2); + render_fb = light_storage->get_cubemap_fb(shadow_size / 2, p_pass); + + light_projection = light_storage->light_instance_get_shadow_camera(p_light, p_pass); + light_transform = light_storage->light_instance_get_shadow_transform(p_light, p_pass); + render_cubemap = true; + finalize_cubemap = p_pass == 5; + atlas_fb = light_storage->shadow_atlas_get_fb(p_shadow_atlas); + + atlas_size = shadow_atlas_size; + + if (p_pass == 0) { + _render_shadow_begin(); + } + + } else { + atlas_rect.position.x += 1; + atlas_rect.position.y += 1; + atlas_rect.size.x -= 2; + atlas_rect.size.y -= 2; + + atlas_rect.position += p_pass * atlas_rect.size * dual_paraboloid_offset; + + light_projection = light_storage->light_instance_get_shadow_camera(p_light, 0); + light_transform = light_storage->light_instance_get_shadow_transform(p_light, 0); + + using_dual_paraboloid = true; + using_dual_paraboloid_flip = p_pass == 1; + render_fb = light_storage->shadow_atlas_get_fb(p_shadow_atlas); + flip_y = true; + } + + } else if (light_storage->light_get_type(base) == RS::LIGHT_SPOT) { + light_projection = light_storage->light_instance_get_shadow_camera(p_light, 0); + light_transform = light_storage->light_instance_get_shadow_transform(p_light, 0); + + render_fb = light_storage->shadow_atlas_get_fb(p_shadow_atlas); + + flip_y = true; + } + } + + if (render_cubemap) { + //rendering to cubemap + _render_shadow_append(render_fb, p_instances, light_projection, light_transform, zfar, 0, 0, false, false, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_mesh_lod_threshold, Rect2(), false, true, true, true, p_render_info); + if (finalize_cubemap) { + _render_shadow_process(); + _render_shadow_end(); + //reblit + Rect2 atlas_rect_norm = atlas_rect; + atlas_rect_norm.position /= float(atlas_size); + atlas_rect_norm.size /= float(atlas_size); + copy_effects->copy_cubemap_to_dp(render_texture, atlas_fb, atlas_rect_norm, atlas_rect.size, light_projection.get_z_near(), light_projection.get_z_far(), false); + atlas_rect_norm.position += Vector2(dual_paraboloid_offset) * atlas_rect_norm.size; + copy_effects->copy_cubemap_to_dp(render_texture, atlas_fb, atlas_rect_norm, atlas_rect.size, light_projection.get_z_near(), light_projection.get_z_far(), true); + + //restore transform so it can be properly used + light_storage->light_instance_set_shadow_transform(p_light, Projection(), light_storage->light_instance_get_base_transform(p_light), zfar, 0, 0, 0); + } + + } else { + //render shadow + _render_shadow_append(render_fb, p_instances, light_projection, light_transform, zfar, 0, 0, using_dual_paraboloid, using_dual_paraboloid_flip, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_mesh_lod_threshold, atlas_rect, flip_y, p_clear_region, p_open_pass, p_close_pass, p_render_info); + } +} + +void RenderForwardClustered::_render_shadow_begin() { + scene_state.shadow_passes.clear(); + RD::get_singleton()->draw_command_begin_label("Shadow Setup"); + _update_render_base_uniform_set(); + + render_list[RENDER_LIST_SECONDARY].clear(); + scene_state.instance_data[RENDER_LIST_SECONDARY].clear(); +} + +void RenderForwardClustered::_render_shadow_append(RID p_framebuffer, const PagedArray<RenderGeometryInstance *> &p_instances, const Projection &p_projection, const Transform3D &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_mesh_lod_threshold, const Rect2i &p_rect, bool p_flip_y, bool p_clear_region, bool p_begin, bool p_end, RenderingMethod::RenderInfo *p_render_info) { + uint32_t shadow_pass_index = scene_state.shadow_passes.size(); + + SceneState::ShadowPass shadow_pass; + + RenderSceneDataRD scene_data; + scene_data.cam_projection = p_projection; + scene_data.cam_transform = p_transform; + scene_data.view_projection[0] = p_projection; + scene_data.z_far = p_zfar; + scene_data.z_near = 0.0; + scene_data.lod_distance_multiplier = p_lod_distance_multiplier; + scene_data.dual_paraboloid_side = p_use_dp_flip ? -1 : 1; + scene_data.opaque_prepass_threshold = 0.1f; + + RenderDataRD render_data; + render_data.scene_data = &scene_data; + render_data.cluster_size = 1; + render_data.cluster_max_elements = 32; + render_data.instances = &p_instances; + render_data.render_info = p_render_info; + + _setup_environment(&render_data, true, Vector2(1, 1), !p_flip_y, Color(), false, p_use_pancake, shadow_pass_index); + + if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_DISABLE_LOD) { + scene_data.screen_mesh_lod_threshold = 0.0; + } else { + scene_data.screen_mesh_lod_threshold = p_screen_mesh_lod_threshold; + } + + PassMode pass_mode = p_use_dp ? PASS_MODE_SHADOW_DP : PASS_MODE_SHADOW; + + uint32_t render_list_from = render_list[RENDER_LIST_SECONDARY].elements.size(); + _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode, 0, false, false, true); + uint32_t render_list_size = render_list[RENDER_LIST_SECONDARY].elements.size() - render_list_from; + render_list[RENDER_LIST_SECONDARY].sort_by_key_range(render_list_from, render_list_size); + _fill_instance_data(RENDER_LIST_SECONDARY, p_render_info ? p_render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_SHADOW] : (int *)nullptr, render_list_from, render_list_size, false); + + { + //regular forward for now + bool flip_cull = p_use_dp_flip; + if (p_flip_y) { + flip_cull = !flip_cull; + } + + shadow_pass.element_from = render_list_from; + shadow_pass.element_count = render_list_size; + shadow_pass.flip_cull = flip_cull; + shadow_pass.pass_mode = pass_mode; + + shadow_pass.rp_uniform_set = RID(); //will be filled later when instance buffer is complete + shadow_pass.camera_plane = p_camera_plane; + shadow_pass.screen_mesh_lod_threshold = scene_data.screen_mesh_lod_threshold; + shadow_pass.lod_distance_multiplier = scene_data.lod_distance_multiplier; + + shadow_pass.framebuffer = p_framebuffer; + shadow_pass.initial_depth_action = p_begin ? (p_clear_region ? RD::INITIAL_ACTION_CLEAR_REGION : RD::INITIAL_ACTION_CLEAR) : (p_clear_region ? RD::INITIAL_ACTION_CLEAR_REGION_CONTINUE : RD::INITIAL_ACTION_CONTINUE); + shadow_pass.final_depth_action = p_end ? RD::FINAL_ACTION_READ : RD::FINAL_ACTION_CONTINUE; + shadow_pass.rect = p_rect; + + scene_state.shadow_passes.push_back(shadow_pass); + } +} + +void RenderForwardClustered::_render_shadow_process() { + _update_instance_data_buffer(RENDER_LIST_SECONDARY); + //render shadows one after the other, so this can be done un-barriered and the driver can optimize (as well as allow us to run compute at the same time) + + for (uint32_t i = 0; i < scene_state.shadow_passes.size(); i++) { + //render passes need to be configured after instance buffer is done, since they need the latest version + SceneState::ShadowPass &shadow_pass = scene_state.shadow_passes[i]; + shadow_pass.rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, nullptr, RID(), false, i); + } + + RD::get_singleton()->draw_command_end_label(); +} +void RenderForwardClustered::_render_shadow_end(uint32_t p_barrier) { + RD::get_singleton()->draw_command_begin_label("Shadow Render"); + + for (uint32_t i = 0; i < scene_state.shadow_passes.size(); i++) { + SceneState::ShadowPass &shadow_pass = scene_state.shadow_passes[i]; + RenderListParameters render_list_parameters(render_list[RENDER_LIST_SECONDARY].elements.ptr() + shadow_pass.element_from, render_list[RENDER_LIST_SECONDARY].element_info.ptr() + shadow_pass.element_from, shadow_pass.element_count, shadow_pass.flip_cull, shadow_pass.pass_mode, 0, true, false, shadow_pass.rp_uniform_set, false, Vector2(), shadow_pass.lod_distance_multiplier, shadow_pass.screen_mesh_lod_threshold, 1, shadow_pass.element_from, RD::BARRIER_MASK_NO_BARRIER); + _render_list_with_threads(&render_list_parameters, shadow_pass.framebuffer, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, shadow_pass.initial_depth_action, shadow_pass.final_depth_action, Vector<Color>(), 1.0, 0, shadow_pass.rect); + } + + if (p_barrier != RD::BARRIER_MASK_NO_BARRIER) { + RD::get_singleton()->barrier(RD::BARRIER_MASK_RASTER, p_barrier); + } + RD::get_singleton()->draw_command_end_label(); +} + +void RenderForwardClustered::_render_particle_collider_heightfield(RID p_fb, const Transform3D &p_cam_transform, const Projection &p_cam_projection, const PagedArray<RenderGeometryInstance *> &p_instances) { + RENDER_TIMESTAMP("Setup GPUParticlesCollisionHeightField3D"); + + RD::get_singleton()->draw_command_begin_label("Render Collider Heightfield"); + + RenderSceneDataRD scene_data; + scene_data.cam_projection = p_cam_projection; + scene_data.cam_transform = p_cam_transform; + scene_data.view_projection[0] = p_cam_projection; + scene_data.z_near = 0.0; + scene_data.z_far = p_cam_projection.get_z_far(); + scene_data.dual_paraboloid_side = 0; + scene_data.opaque_prepass_threshold = 0.0; + + RenderDataRD render_data; + render_data.scene_data = &scene_data; + render_data.cluster_size = 1; + render_data.cluster_max_elements = 32; + render_data.instances = &p_instances; + + _update_render_base_uniform_set(); + + _setup_environment(&render_data, true, Vector2(1, 1), true, Color(), false, false); + + PassMode pass_mode = PASS_MODE_SHADOW; + + _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode); + render_list[RENDER_LIST_SECONDARY].sort_by_key(); + _fill_instance_data(RENDER_LIST_SECONDARY); + + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, nullptr, RID()); + + RENDER_TIMESTAMP("Render Collider Heightfield"); + + { + //regular forward for now + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), false, pass_mode, 0, true, false, rp_uniform_set); + _render_list_with_threads(&render_list_params, p_fb, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ); + } + RD::get_singleton()->draw_command_end_label(); +} + +void RenderForwardClustered::_render_material(const Transform3D &p_cam_transform, const Projection &p_cam_projection, bool p_cam_orthogonal, const PagedArray<RenderGeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region, float p_exposure_normalization) { + RENDER_TIMESTAMP("Setup Rendering 3D Material"); + + RD::get_singleton()->draw_command_begin_label("Render 3D Material"); + + RenderSceneDataRD scene_data; + scene_data.cam_projection = p_cam_projection; + scene_data.cam_transform = p_cam_transform; + scene_data.view_projection[0] = p_cam_projection; + scene_data.dual_paraboloid_side = 0; + scene_data.material_uv2_mode = false; + scene_data.opaque_prepass_threshold = 0.0f; + scene_data.emissive_exposure_normalization = p_exposure_normalization; + + RenderDataRD render_data; + render_data.scene_data = &scene_data; + render_data.cluster_size = 1; + render_data.cluster_max_elements = 32; + render_data.instances = &p_instances; + + _update_render_base_uniform_set(); + + _setup_environment(&render_data, true, Vector2(1, 1), false, Color()); + + PassMode pass_mode = PASS_MODE_DEPTH_MATERIAL; + _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode); + render_list[RENDER_LIST_SECONDARY].sort_by_key(); + _fill_instance_data(RENDER_LIST_SECONDARY); + + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, nullptr, RID()); + + RENDER_TIMESTAMP("Render 3D Material"); + + { + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, 0, true, false, rp_uniform_set); + //regular forward for now + Vector<Color> clear = { + Color(0, 0, 0, 0), + Color(0, 0, 0, 0), + Color(0, 0, 0, 0), + Color(0, 0, 0, 0), + Color(0, 0, 0, 0) + }; + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, clear, 1.0, 0, p_region); + _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(p_framebuffer), &render_list_params, 0, render_list_params.element_count); + RD::get_singleton()->draw_list_end(); + } + + RD::get_singleton()->draw_command_end_label(); +} + +void RenderForwardClustered::_render_uv2(const PagedArray<RenderGeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) { + RENDER_TIMESTAMP("Setup Rendering UV2"); + + RD::get_singleton()->draw_command_begin_label("Render UV2"); + + RenderSceneDataRD scene_data; + scene_data.dual_paraboloid_side = 0; + scene_data.material_uv2_mode = true; + scene_data.opaque_prepass_threshold = 0.0; + scene_data.emissive_exposure_normalization = -1.0; + + RenderDataRD render_data; + render_data.scene_data = &scene_data; + render_data.cluster_size = 1; + render_data.cluster_max_elements = 32; + render_data.instances = &p_instances; + + _update_render_base_uniform_set(); + + _setup_environment(&render_data, true, Vector2(1, 1), false, Color()); + + PassMode pass_mode = PASS_MODE_DEPTH_MATERIAL; + _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode); + render_list[RENDER_LIST_SECONDARY].sort_by_key(); + _fill_instance_data(RENDER_LIST_SECONDARY); + + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, nullptr, RID()); + + RENDER_TIMESTAMP("Render 3D Material"); + + { + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, 0, true, false, rp_uniform_set, true); + //regular forward for now + Vector<Color> clear = { + Color(0, 0, 0, 0), + Color(0, 0, 0, 0), + Color(0, 0, 0, 0), + Color(0, 0, 0, 0), + Color(0, 0, 0, 0) + }; + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, clear, 1.0, 0, p_region); + + const int uv_offset_count = 9; + static const Vector2 uv_offsets[uv_offset_count] = { + Vector2(-1, 1), + Vector2(1, 1), + Vector2(1, -1), + Vector2(-1, -1), + Vector2(-1, 0), + Vector2(1, 0), + Vector2(0, -1), + Vector2(0, 1), + Vector2(0, 0), + + }; + + for (int i = 0; i < uv_offset_count; i++) { + Vector2 ofs = uv_offsets[i]; + ofs.x /= p_region.size.width; + ofs.y /= p_region.size.height; + render_list_params.uv_offset = ofs; + _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(p_framebuffer), &render_list_params, 0, render_list_params.element_count); //first wireframe, for pseudo conservative + } + render_list_params.uv_offset = Vector2(); + render_list_params.force_wireframe = false; + _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(p_framebuffer), &render_list_params, 0, render_list_params.element_count); //second regular triangles + + RD::get_singleton()->draw_list_end(); + } + + RD::get_singleton()->draw_command_end_label(); +} + +void RenderForwardClustered::_render_sdfgi(Ref<RenderSceneBuffersRD> p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<RenderGeometryInstance *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture, float p_exposure_normalization) { + RENDER_TIMESTAMP("Render SDFGI"); + + RD::get_singleton()->draw_command_begin_label("Render SDFGI Voxel"); + + RenderSceneDataRD scene_data; + + RenderDataRD render_data; + render_data.scene_data = &scene_data; + render_data.cluster_size = 1; + render_data.cluster_max_elements = 32; + render_data.instances = &p_instances; + + _update_render_base_uniform_set(); + + PassMode pass_mode = PASS_MODE_SDF; + _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode); + render_list[RENDER_LIST_SECONDARY].sort_by_key(); + _fill_instance_data(RENDER_LIST_SECONDARY); + + Vector3 half_extents = p_bounds.size * 0.5; + Vector3 center = p_bounds.position + half_extents; + + Vector<RID> sbs = { + p_albedo_texture, + p_emission_texture, + p_emission_aniso_texture, + p_geom_facing_texture + }; + + //print_line("re-render " + p_from + " - " + p_size + " bounds " + p_bounds); + for (int i = 0; i < 3; i++) { + scene_state.ubo.sdf_offset[i] = p_from[i]; + scene_state.ubo.sdf_size[i] = p_size[i]; + } + + for (int i = 0; i < 3; i++) { + Vector3 axis; + axis[i] = 1.0; + Vector3 up, right; + int right_axis = (i + 1) % 3; + int up_axis = (i + 2) % 3; + up[up_axis] = 1.0; + right[right_axis] = 1.0; + + Size2i fb_size; + fb_size.x = p_size[right_axis]; + fb_size.y = p_size[up_axis]; + + scene_data.cam_transform.origin = center + axis * half_extents; + scene_data.cam_transform.basis.set_column(0, right); + scene_data.cam_transform.basis.set_column(1, up); + scene_data.cam_transform.basis.set_column(2, axis); + + //print_line("pass: " + itos(i) + " xform " + scene_data.cam_transform); + + float h_size = half_extents[right_axis]; + float v_size = half_extents[up_axis]; + float d_size = half_extents[i] * 2.0; + scene_data.cam_projection.set_orthogonal(-h_size, h_size, -v_size, v_size, 0, d_size); + //print_line("pass: " + itos(i) + " cam hsize: " + rtos(h_size) + " vsize: " + rtos(v_size) + " dsize " + rtos(d_size)); + + Transform3D to_bounds; + to_bounds.origin = p_bounds.position; + to_bounds.basis.scale(p_bounds.size); + + RendererRD::MaterialStorage::store_transform(to_bounds.affine_inverse() * scene_data.cam_transform, scene_state.ubo.sdf_to_bounds); + + scene_data.emissive_exposure_normalization = p_exposure_normalization; + _setup_environment(&render_data, true, Vector2(1, 1), false, Color()); + + RID rp_uniform_set = _setup_sdfgi_render_pass_uniform_set(p_albedo_texture, p_emission_texture, p_emission_aniso_texture, p_geom_facing_texture); + + HashMap<Size2i, RID>::Iterator E = sdfgi_framebuffer_size_cache.find(fb_size); + if (!E) { + RID fb = RD::get_singleton()->framebuffer_create_empty(fb_size); + E = sdfgi_framebuffer_size_cache.insert(fb_size, fb); + } + + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, 0, true, false, rp_uniform_set, false); + _render_list_with_threads(&render_list_params, E->value, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, Vector<Color>(), 1.0, 0, Rect2(), sbs); + } + + RD::get_singleton()->draw_command_end_label(); +} + +void RenderForwardClustered::base_uniforms_changed() { + if (!render_base_uniform_set.is_null() && RD::get_singleton()->uniform_set_is_valid(render_base_uniform_set)) { + RD::get_singleton()->free(render_base_uniform_set); + } + render_base_uniform_set = RID(); +} + +void RenderForwardClustered::_update_render_base_uniform_set() { + RendererRD::LightStorage *light_storage = RendererRD::LightStorage::get_singleton(); + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + + if (render_base_uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(render_base_uniform_set) || (lightmap_texture_array_version != light_storage->lightmap_array_get_version()) || base_uniform_set_updated) { + base_uniform_set_updated = false; + + if (render_base_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(render_base_uniform_set)) { + RD::get_singleton()->free(render_base_uniform_set); + } + + lightmap_texture_array_version = light_storage->lightmap_array_get_version(); + + Vector<RD::Uniform> uniforms; + + { + Vector<RID> ids; + ids.resize(12); + RID *ids_ptr = ids.ptrw(); + ids_ptr[0] = material_storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[1] = material_storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[2] = material_storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[3] = material_storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[4] = material_storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[5] = material_storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[6] = material_storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[7] = material_storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[8] = material_storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[9] = material_storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[10] = material_storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[11] = material_storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + + RD::Uniform u(RD::UNIFORM_TYPE_SAMPLER, 1, ids); + + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 2; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.append_id(scene_shader.shadow_sampler); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 3; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + RID sampler; + switch (decals_get_filter()) { + case RS::DECAL_FILTER_NEAREST: { + sampler = material_storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_LINEAR: { + sampler = material_storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_NEAREST_MIPMAPS: { + sampler = material_storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_LINEAR_MIPMAPS: { + sampler = material_storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_NEAREST_MIPMAPS_ANISOTROPIC: { + sampler = material_storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_LINEAR_MIPMAPS_ANISOTROPIC: { + sampler = material_storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + } + + u.append_id(sampler); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 4; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + RID sampler; + switch (light_projectors_get_filter()) { + case RS::LIGHT_PROJECTOR_FILTER_NEAREST: { + sampler = material_storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_LINEAR: { + sampler = material_storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_NEAREST_MIPMAPS: { + sampler = material_storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS: { + sampler = material_storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_NEAREST_MIPMAPS_ANISOTROPIC: { + sampler = material_storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS_ANISOTROPIC: { + sampler = material_storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + } + + u.append_id(sampler); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 5; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.append_id(RendererRD::LightStorage::get_singleton()->get_omni_light_buffer()); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 6; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.append_id(RendererRD::LightStorage::get_singleton()->get_spot_light_buffer()); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 7; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.append_id(RendererRD::LightStorage::get_singleton()->get_reflection_probe_buffer()); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 8; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.append_id(RendererRD::LightStorage::get_singleton()->get_directional_light_buffer()); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 9; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.append_id(scene_state.lightmap_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 10; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.append_id(scene_state.lightmap_capture_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 11; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID decal_atlas = RendererRD::TextureStorage::get_singleton()->decal_atlas_get_texture(); + u.append_id(decal_atlas); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 12; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID decal_atlas = RendererRD::TextureStorage::get_singleton()->decal_atlas_get_texture_srgb(); + u.append_id(decal_atlas); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 13; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.append_id(RendererRD::TextureStorage::get_singleton()->get_decal_buffer()); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 14; + u.append_id(RendererRD::MaterialStorage::get_singleton()->global_shader_uniforms_get_storage_buffer()); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 15; + u.append_id(sdfgi_get_ubo()); + uniforms.push_back(u); + } + + render_base_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, scene_shader.default_shader_rd, SCENE_UNIFORM_SET); + } +} + +RID RenderForwardClustered::_setup_render_pass_uniform_set(RenderListType p_render_list, const RenderDataRD *p_render_data, RID p_radiance_texture, bool p_use_directional_shadow_atlas, int p_index) { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + RendererRD::LightStorage *light_storage = RendererRD::LightStorage::get_singleton(); + + Ref<RenderSceneBuffersRD> rb; // handy for not having to fully type out p_render_data->render_buffers all the time... + Ref<RenderBufferDataForwardClustered> rb_data; + if (p_render_data && p_render_data->render_buffers.is_valid()) { + rb = p_render_data->render_buffers; + rb_data = rb->get_custom_data(RB_SCOPE_FORWARD_CLUSTERED); + } + + //default render buffer and scene state uniform set + + Vector<RD::Uniform> uniforms; + + { + RD::Uniform u; + u.binding = 0; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.append_id(scene_state.uniform_buffers[p_index]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 1; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.append_id(scene_state.implementation_uniform_buffers[p_index]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 2; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + RID instance_buffer = scene_state.instance_buffer[p_render_list]; + if (instance_buffer == RID()) { + instance_buffer = scene_shader.default_vec4_xform_buffer; // any buffer will do since its not used + } + u.append_id(instance_buffer); + uniforms.push_back(u); + } + { + RID radiance_texture; + if (p_radiance_texture.is_valid()) { + radiance_texture = p_radiance_texture; + } else { + radiance_texture = texture_storage->texture_rd_get_default(is_using_radiance_cubemap_array() ? RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_BLACK : RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK); + } + RD::Uniform u; + u.binding = 3; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.append_id(radiance_texture); + uniforms.push_back(u); + } + { + RID ref_texture = (p_render_data && p_render_data->reflection_atlas.is_valid()) ? light_storage->reflection_atlas_get_texture(p_render_data->reflection_atlas) : RID(); + RD::Uniform u; + u.binding = 4; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + if (ref_texture.is_valid()) { + u.append_id(ref_texture); + } else { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_BLACK)); + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 5; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID texture; + if (p_render_data && p_render_data->shadow_atlas.is_valid()) { + texture = RendererRD::LightStorage::get_singleton()->shadow_atlas_get_texture(p_render_data->shadow_atlas); + } + if (!texture.is_valid()) { + texture = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_DEPTH); + } + u.append_id(texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 6; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + if (p_use_directional_shadow_atlas && RendererRD::LightStorage::get_singleton()->directional_shadow_get_texture().is_valid()) { + u.append_id(RendererRD::LightStorage::get_singleton()->directional_shadow_get_texture()); + } else { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_DEPTH)); + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 7; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + + RID default_tex = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE); + for (uint32_t i = 0; i < scene_state.max_lightmaps; i++) { + if (p_render_data && i < p_render_data->lightmaps->size()) { + RID base = light_storage->lightmap_instance_get_lightmap((*p_render_data->lightmaps)[i]); + RID texture = light_storage->lightmap_get_texture(base); + RID rd_texture = texture_storage->texture_get_rd_texture(texture); + u.append_id(rd_texture); + } else { + u.append_id(default_tex); + } + } + + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 8; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID default_tex = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE); + for (int i = 0; i < MAX_VOXEL_GI_INSTANCESS; i++) { + if (p_render_data && i < (int)p_render_data->voxel_gi_instances->size()) { + RID tex = gi.voxel_gi_instance_get_texture((*p_render_data->voxel_gi_instances)[i]); + if (!tex.is_valid()) { + tex = default_tex; + } + u.append_id(tex); + } else { + u.append_id(default_tex); + } + } + + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 9; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + RID cb = (p_render_data && p_render_data->cluster_buffer.is_valid()) ? p_render_data->cluster_buffer : scene_shader.default_vec4_xform_buffer; + u.append_id(cb); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 10; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID texture; + if (rb.is_valid() && rb->has_texture(RB_SCOPE_BUFFERS, RB_TEX_BACK_DEPTH)) { + texture = rb->get_texture(RB_SCOPE_BUFFERS, RB_TEX_BACK_DEPTH); + } else { + texture = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_DEPTH); + } + u.append_id(texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 11; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID bbt = rb_data.is_valid() ? rb->get_back_buffer_texture() : RID(); + RID texture = bbt.is_valid() ? bbt : texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_BLACK); + u.append_id(texture); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 12; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID texture = rb_data.is_valid() && rb_data->has_normal_roughness() ? rb_data->get_normal_roughness() : texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_NORMAL); + u.append_id(texture); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 13; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID aot = rb_data.is_valid() ? rb_data->get_ao_texture() : RID(); + RID texture = aot.is_valid() ? aot : texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_BLACK); + u.append_id(texture); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 14; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID texture = rb_data.is_valid() && rb->has_texture(RB_SCOPE_GI, RB_TEX_AMBIENT) ? rb->get_texture(RB_SCOPE_GI, RB_TEX_AMBIENT) : texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_BLACK); + u.append_id(texture); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 15; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID texture = rb_data.is_valid() && rb->has_texture(RB_SCOPE_GI, RB_TEX_REFLECTION) ? rb->get_texture(RB_SCOPE_GI, RB_TEX_REFLECTION) : texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_BLACK); + u.append_id(texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 16; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID t; + if (rb.is_valid() && rb->has_custom_data(RB_SCOPE_SDFGI)) { + Ref<RendererRD::GI::SDFGI> sdfgi = rb->get_custom_data(RB_SCOPE_SDFGI); + t = sdfgi->lightprobe_texture; + } + if (t.is_null()) { + t = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE); + } + u.append_id(t); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 17; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID t; + if (rb.is_valid() && rb->has_custom_data(RB_SCOPE_SDFGI)) { + Ref<RendererRD::GI::SDFGI> sdfgi = rb->get_custom_data(RB_SCOPE_SDFGI); + t = sdfgi->occlusion_texture; + } + if (t.is_null()) { + t = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE); + } + u.append_id(t); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 18; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + RID voxel_gi; + if (rb.is_valid() && rb->has_custom_data(RB_SCOPE_GI)) { + Ref<RendererRD::GI::RenderBuffersGI> rbgi = rb->get_custom_data(RB_SCOPE_GI); + voxel_gi = rbgi->get_voxel_gi_buffer(); + } + u.append_id(voxel_gi.is_valid() ? voxel_gi : render_buffers_get_default_voxel_gi_buffer()); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 19; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID vfog = RID(); + if (rb_data.is_valid() && rb->has_custom_data(RB_SCOPE_FOG)) { + Ref<RendererRD::Fog::VolumetricFog> fog = rb->get_custom_data(RB_SCOPE_FOG); + vfog = fog->fog_map; + if (vfog.is_null()) { + vfog = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE); + } + } else { + vfog = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE); + } + u.append_id(vfog); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 20; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID ssil = rb_data.is_valid() ? rb_data->get_ssil_texture() : RID(); + RID texture = ssil.is_valid() ? ssil : texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_BLACK); + u.append_id(texture); + uniforms.push_back(u); + } + + return UniformSetCacheRD::get_singleton()->get_cache_vec(scene_shader.default_shader_rd, RENDER_PASS_UNIFORM_SET, uniforms); +} + +RID RenderForwardClustered::_setup_sdfgi_render_pass_uniform_set(RID p_albedo_texture, RID p_emission_texture, RID p_emission_aniso_texture, RID p_geom_facing_texture) { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + Vector<RD::Uniform> uniforms; + + { + RD::Uniform u; + u.binding = 0; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.append_id(scene_state.uniform_buffers[0]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 1; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.append_id(scene_state.implementation_uniform_buffers[0]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 2; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + RID instance_buffer = scene_state.instance_buffer[RENDER_LIST_SECONDARY]; + if (instance_buffer == RID()) { + instance_buffer = scene_shader.default_vec4_xform_buffer; // any buffer will do since its not used + } + u.append_id(instance_buffer); + uniforms.push_back(u); + } + { + // No radiance texture. + RID radiance_texture = texture_storage->texture_rd_get_default(is_using_radiance_cubemap_array() ? RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_BLACK : RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK); + RD::Uniform u; + u.binding = 3; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.append_id(radiance_texture); + uniforms.push_back(u); + } + + { + // No reflection atlas. + RID ref_texture = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_BLACK); + RD::Uniform u; + u.binding = 4; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.append_id(ref_texture); + uniforms.push_back(u); + } + + { + // No shadow atlas. + RD::Uniform u; + u.binding = 5; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID texture = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_DEPTH); + u.append_id(texture); + uniforms.push_back(u); + } + + { + // No directional shadow atlas. + RD::Uniform u; + u.binding = 6; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID texture = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_DEPTH); + u.append_id(texture); + uniforms.push_back(u); + } + + { + // No Lightmaps + RD::Uniform u; + u.binding = 7; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + + RID default_tex = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE); + for (uint32_t i = 0; i < scene_state.max_lightmaps; i++) { + u.append_id(default_tex); + } + + uniforms.push_back(u); + } + + { + // No VoxelGIs + RD::Uniform u; + u.binding = 8; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + + RID default_tex = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE); + for (int i = 0; i < MAX_VOXEL_GI_INSTANCESS; i++) { + u.append_id(default_tex); + } + + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 9; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + RID cb = scene_shader.default_vec4_xform_buffer; + u.append_id(cb); + uniforms.push_back(u); + } + + // actual sdfgi stuff + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 10; + u.append_id(p_albedo_texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 11; + u.append_id(p_emission_texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 12; + u.append_id(p_emission_aniso_texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 13; + u.append_id(p_geom_facing_texture); + uniforms.push_back(u); + } + + return UniformSetCacheRD::get_singleton()->get_cache_vec(scene_shader.default_shader_sdfgi_rd, RENDER_PASS_UNIFORM_SET, uniforms); +} + +RID RenderForwardClustered::_render_buffers_get_normal_texture(Ref<RenderSceneBuffersRD> p_render_buffers) { + Ref<RenderBufferDataForwardClustered> rb_data = p_render_buffers->get_custom_data(RB_SCOPE_FORWARD_CLUSTERED); + + return p_render_buffers->get_msaa_3d() == RS::VIEWPORT_MSAA_DISABLED ? rb_data->get_normal_roughness() : rb_data->get_normal_roughness_msaa(); +} + +RID RenderForwardClustered::_render_buffers_get_velocity_texture(Ref<RenderSceneBuffersRD> p_render_buffers) { + return p_render_buffers->get_velocity_buffer(p_render_buffers->get_msaa_3d() != RS::VIEWPORT_MSAA_DISABLED); +} + +void RenderForwardClustered::environment_set_ssao_quality(RS::EnvironmentSSAOQuality p_quality, bool p_half_size, float p_adaptive_target, int p_blur_passes, float p_fadeout_from, float p_fadeout_to) { + ss_effects->ssao_set_quality(p_quality, p_half_size, p_adaptive_target, p_blur_passes, p_fadeout_from, p_fadeout_to); +} + +void RenderForwardClustered::environment_set_ssil_quality(RS::EnvironmentSSILQuality p_quality, bool p_half_size, float p_adaptive_target, int p_blur_passes, float p_fadeout_from, float p_fadeout_to) { + ss_effects->ssil_set_quality(p_quality, p_half_size, p_adaptive_target, p_blur_passes, p_fadeout_from, p_fadeout_to); +} + +void RenderForwardClustered::environment_set_ssr_roughness_quality(RS::EnvironmentSSRRoughnessQuality p_quality) { + ss_effects->ssr_set_roughness_quality(p_quality); +} + +void RenderForwardClustered::sub_surface_scattering_set_quality(RS::SubSurfaceScatteringQuality p_quality) { + ss_effects->sss_set_quality(p_quality); +} + +void RenderForwardClustered::sub_surface_scattering_set_scale(float p_scale, float p_depth_scale) { + ss_effects->sss_set_scale(p_scale, p_depth_scale); +} + +RenderForwardClustered *RenderForwardClustered::singleton = nullptr; + +void RenderForwardClustered::GeometryInstanceForwardClustered::_mark_dirty() { + if (dirty_list_element.in_list()) { + return; + } + + //clear surface caches + GeometryInstanceSurfaceDataCache *surf = surface_caches; + + while (surf) { + GeometryInstanceSurfaceDataCache *next = surf->next; + RenderForwardClustered::get_singleton()->geometry_instance_surface_alloc.free(surf); + surf = next; + } + + surface_caches = nullptr; + + RenderForwardClustered::get_singleton()->geometry_instance_dirty_list.add(&dirty_list_element); +} + +void RenderForwardClustered::_geometry_instance_add_surface_with_material(GeometryInstanceForwardClustered *ginstance, uint32_t p_surface, SceneShaderForwardClustered::MaterialData *p_material, uint32_t p_material_id, uint32_t p_shader_id, RID p_mesh) { + RendererRD::MeshStorage *mesh_storage = RendererRD::MeshStorage::get_singleton(); + + bool has_read_screen_alpha = p_material->shader_data->uses_screen_texture || p_material->shader_data->uses_depth_texture || p_material->shader_data->uses_normal_texture; + bool has_base_alpha = (p_material->shader_data->uses_alpha && !p_material->shader_data->uses_alpha_clip) || has_read_screen_alpha; + bool has_blend_alpha = p_material->shader_data->uses_blend_alpha; + bool has_alpha = has_base_alpha || has_blend_alpha; + + uint32_t flags = 0; + + if (p_material->shader_data->uses_sss) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_SUBSURFACE_SCATTERING; + } + + if (p_material->shader_data->uses_screen_texture) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_SCREEN_TEXTURE; + } + + if (p_material->shader_data->uses_depth_texture) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_DEPTH_TEXTURE; + } + + if (p_material->shader_data->uses_normal_texture) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_NORMAL_TEXTURE; + } + + if (ginstance->data->cast_double_sided_shadows) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_DOUBLE_SIDED_SHADOWS; + } + + if (has_alpha || has_read_screen_alpha || p_material->shader_data->depth_draw == SceneShaderForwardClustered::ShaderData::DEPTH_DRAW_DISABLED || p_material->shader_data->depth_test == SceneShaderForwardClustered::ShaderData::DEPTH_TEST_DISABLED) { + //material is only meant for alpha pass + flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_ALPHA; + if (p_material->shader_data->uses_depth_pre_pass && !(p_material->shader_data->depth_draw == SceneShaderForwardClustered::ShaderData::DEPTH_DRAW_DISABLED || p_material->shader_data->depth_test == SceneShaderForwardClustered::ShaderData::DEPTH_TEST_DISABLED)) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH; + flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_SHADOW; + } + } else { + flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE; + flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH; + flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_SHADOW; + } + + if (p_material->shader_data->uses_particle_trails) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_PARTICLE_TRAILS; + } + + SceneShaderForwardClustered::MaterialData *material_shadow = nullptr; + void *surface_shadow = nullptr; + if (!p_material->shader_data->uses_particle_trails && !p_material->shader_data->writes_modelview_or_projection && !p_material->shader_data->uses_vertex && !p_material->shader_data->uses_position && !p_material->shader_data->uses_discard && !p_material->shader_data->uses_depth_pre_pass && !p_material->shader_data->uses_alpha_clip && p_material->shader_data->cull_mode == SceneShaderForwardClustered::ShaderData::CULL_BACK) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_SHARED_SHADOW_MATERIAL; + material_shadow = static_cast<SceneShaderForwardClustered::MaterialData *>(RendererRD::MaterialStorage::get_singleton()->material_get_data(scene_shader.default_material, RendererRD::MaterialStorage::SHADER_TYPE_3D)); + + RID shadow_mesh = mesh_storage->mesh_get_shadow_mesh(p_mesh); + + if (shadow_mesh.is_valid()) { + surface_shadow = mesh_storage->mesh_get_surface(shadow_mesh, p_surface); + } + + } else { + material_shadow = p_material; + } + + GeometryInstanceSurfaceDataCache *sdcache = geometry_instance_surface_alloc.alloc(); + + sdcache->flags = flags; + + sdcache->shader = p_material->shader_data; + sdcache->material_uniform_set = p_material->uniform_set; + sdcache->surface = mesh_storage->mesh_get_surface(p_mesh, p_surface); + sdcache->primitive = mesh_storage->mesh_surface_get_primitive(sdcache->surface); + sdcache->surface_index = p_surface; + + if (ginstance->data->dirty_dependencies) { + RSG::utilities->base_update_dependency(p_mesh, &ginstance->data->dependency_tracker); + } + + //shadow + sdcache->shader_shadow = material_shadow->shader_data; + sdcache->material_uniform_set_shadow = material_shadow->uniform_set; + + sdcache->surface_shadow = surface_shadow ? surface_shadow : sdcache->surface; + + sdcache->owner = ginstance; + + sdcache->next = ginstance->surface_caches; + ginstance->surface_caches = sdcache; + + //sortkey + + sdcache->sort.sort_key1 = 0; + sdcache->sort.sort_key2 = 0; + + sdcache->sort.surface_index = p_surface; + sdcache->sort.material_id_low = p_material_id & 0xFFFF; + sdcache->sort.material_id_hi = p_material_id >> 16; + sdcache->sort.shader_id = p_shader_id; + sdcache->sort.geometry_id = p_mesh.get_local_index(); //only meshes can repeat anyway + sdcache->sort.uses_forward_gi = ginstance->can_sdfgi; + sdcache->sort.priority = p_material->priority; + sdcache->sort.uses_projector = ginstance->using_projectors; + sdcache->sort.uses_softshadow = ginstance->using_softshadows; +} + +void RenderForwardClustered::_geometry_instance_add_surface_with_material_chain(GeometryInstanceForwardClustered *ginstance, uint32_t p_surface, SceneShaderForwardClustered::MaterialData *p_material, RID p_mat_src, RID p_mesh) { + SceneShaderForwardClustered::MaterialData *material = p_material; + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + + _geometry_instance_add_surface_with_material(ginstance, p_surface, material, p_mat_src.get_local_index(), material_storage->material_get_shader_id(p_mat_src), p_mesh); + + while (material->next_pass.is_valid()) { + RID next_pass = material->next_pass; + material = static_cast<SceneShaderForwardClustered::MaterialData *>(material_storage->material_get_data(next_pass, RendererRD::MaterialStorage::SHADER_TYPE_3D)); + if (!material || !material->shader_data->valid) { + break; + } + if (ginstance->data->dirty_dependencies) { + material_storage->material_update_dependency(next_pass, &ginstance->data->dependency_tracker); + } + _geometry_instance_add_surface_with_material(ginstance, p_surface, material, next_pass.get_local_index(), material_storage->material_get_shader_id(next_pass), p_mesh); + } +} + +void RenderForwardClustered::_geometry_instance_add_surface(GeometryInstanceForwardClustered *ginstance, uint32_t p_surface, RID p_material, RID p_mesh) { + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + RID m_src; + + m_src = ginstance->data->material_override.is_valid() ? ginstance->data->material_override : p_material; + + SceneShaderForwardClustered::MaterialData *material = nullptr; + + if (m_src.is_valid()) { + material = static_cast<SceneShaderForwardClustered::MaterialData *>(material_storage->material_get_data(m_src, RendererRD::MaterialStorage::SHADER_TYPE_3D)); + if (!material || !material->shader_data->valid) { + material = nullptr; + } + } + + if (material) { + if (ginstance->data->dirty_dependencies) { + material_storage->material_update_dependency(m_src, &ginstance->data->dependency_tracker); + } + } else { + material = static_cast<SceneShaderForwardClustered::MaterialData *>(material_storage->material_get_data(scene_shader.default_material, RendererRD::MaterialStorage::SHADER_TYPE_3D)); + m_src = scene_shader.default_material; + } + + ERR_FAIL_COND(!material); + + _geometry_instance_add_surface_with_material_chain(ginstance, p_surface, material, m_src, p_mesh); + + if (ginstance->data->material_overlay.is_valid()) { + m_src = ginstance->data->material_overlay; + + material = static_cast<SceneShaderForwardClustered::MaterialData *>(material_storage->material_get_data(m_src, RendererRD::MaterialStorage::SHADER_TYPE_3D)); + if (material && material->shader_data->valid) { + if (ginstance->data->dirty_dependencies) { + material_storage->material_update_dependency(m_src, &ginstance->data->dependency_tracker); + } + + _geometry_instance_add_surface_with_material_chain(ginstance, p_surface, material, m_src, p_mesh); + } + } +} + +void RenderForwardClustered::_geometry_instance_update(RenderGeometryInstance *p_geometry_instance) { + RendererRD::MeshStorage *mesh_storage = RendererRD::MeshStorage::get_singleton(); + RendererRD::ParticlesStorage *particles_storage = RendererRD::ParticlesStorage::get_singleton(); + GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_geometry_instance); + + if (ginstance->data->dirty_dependencies) { + ginstance->data->dependency_tracker.update_begin(); + } + + //add geometry for drawing + switch (ginstance->data->base_type) { + case RS::INSTANCE_MESH: { + const RID *materials = nullptr; + uint32_t surface_count; + RID mesh = ginstance->data->base; + + materials = mesh_storage->mesh_get_surface_count_and_materials(mesh, surface_count); + if (materials) { + //if no materials, no surfaces. + const RID *inst_materials = ginstance->data->surface_materials.ptr(); + uint32_t surf_mat_count = ginstance->data->surface_materials.size(); + + for (uint32_t j = 0; j < surface_count; j++) { + RID material = (j < surf_mat_count && inst_materials[j].is_valid()) ? inst_materials[j] : materials[j]; + _geometry_instance_add_surface(ginstance, j, material, mesh); + } + } + + ginstance->instance_count = 1; + + } break; + + case RS::INSTANCE_MULTIMESH: { + RID mesh = mesh_storage->multimesh_get_mesh(ginstance->data->base); + if (mesh.is_valid()) { + const RID *materials = nullptr; + uint32_t surface_count; + + materials = mesh_storage->mesh_get_surface_count_and_materials(mesh, surface_count); + if (materials) { + for (uint32_t j = 0; j < surface_count; j++) { + _geometry_instance_add_surface(ginstance, j, materials[j], mesh); + } + } + + ginstance->instance_count = mesh_storage->multimesh_get_instances_to_draw(ginstance->data->base); + } + + } break; +#if 0 + case RS::INSTANCE_IMMEDIATE: { + RasterizerStorageGLES3::Immediate *immediate = storage->immediate_owner.get_or_null(inst->base); + ERR_CONTINUE(!immediate); + + _add_geometry(immediate, inst, nullptr, -1, p_depth_pass, p_shadow_pass); + + } break; +#endif + case RS::INSTANCE_PARTICLES: { + int draw_passes = particles_storage->particles_get_draw_passes(ginstance->data->base); + + for (int j = 0; j < draw_passes; j++) { + RID mesh = particles_storage->particles_get_draw_pass_mesh(ginstance->data->base, j); + if (!mesh.is_valid()) { + continue; + } + + const RID *materials = nullptr; + uint32_t surface_count; + + materials = mesh_storage->mesh_get_surface_count_and_materials(mesh, surface_count); + if (materials) { + for (uint32_t k = 0; k < surface_count; k++) { + _geometry_instance_add_surface(ginstance, k, materials[k], mesh); + } + } + } + + ginstance->instance_count = particles_storage->particles_get_amount(ginstance->data->base, ginstance->trail_steps); + + } break; + + default: { + } + } + + //Fill push constant + + ginstance->base_flags = 0; + + bool store_transform = true; + + if (ginstance->data->base_type == RS::INSTANCE_MULTIMESH) { + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH; + if (mesh_storage->multimesh_get_transform_format(ginstance->data->base) == RS::MULTIMESH_TRANSFORM_2D) { + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_FORMAT_2D; + } + if (mesh_storage->multimesh_uses_colors(ginstance->data->base)) { + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_HAS_COLOR; + } + if (mesh_storage->multimesh_uses_custom_data(ginstance->data->base)) { + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_HAS_CUSTOM_DATA; + } + + ginstance->transforms_uniform_set = mesh_storage->multimesh_get_3d_uniform_set(ginstance->data->base, scene_shader.default_shader_rd, TRANSFORMS_UNIFORM_SET); + + } else if (ginstance->data->base_type == RS::INSTANCE_PARTICLES) { + ginstance->base_flags |= INSTANCE_DATA_FLAG_PARTICLES; + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH; + + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_HAS_COLOR; + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_HAS_CUSTOM_DATA; + + //for particles, stride is the trail size + ginstance->base_flags |= (ginstance->trail_steps << INSTANCE_DATA_FLAGS_PARTICLE_TRAIL_SHIFT); + + if (!particles_storage->particles_is_using_local_coords(ginstance->data->base)) { + store_transform = false; + } + ginstance->transforms_uniform_set = particles_storage->particles_get_instance_buffer_uniform_set(ginstance->data->base, scene_shader.default_shader_rd, TRANSFORMS_UNIFORM_SET); + + } else if (ginstance->data->base_type == RS::INSTANCE_MESH) { + if (mesh_storage->skeleton_is_valid(ginstance->data->skeleton)) { + ginstance->transforms_uniform_set = mesh_storage->skeleton_get_3d_uniform_set(ginstance->data->skeleton, scene_shader.default_shader_rd, TRANSFORMS_UNIFORM_SET); + if (ginstance->data->dirty_dependencies) { + mesh_storage->skeleton_update_dependency(ginstance->data->skeleton, &ginstance->data->dependency_tracker); + } + } + } + + ginstance->store_transform_cache = store_transform; + ginstance->can_sdfgi = false; + + if (!RendererRD::LightStorage::get_singleton()->lightmap_instance_is_valid(ginstance->lightmap_instance)) { + if (ginstance->voxel_gi_instances[0].is_null() && (ginstance->data->use_baked_light || ginstance->data->use_dynamic_gi)) { + ginstance->can_sdfgi = true; + } + } + + if (ginstance->data->dirty_dependencies) { + ginstance->data->dependency_tracker.update_end(); + ginstance->data->dirty_dependencies = false; + } + + ginstance->dirty_list_element.remove_from_list(); +} + +void RenderForwardClustered::_update_dirty_geometry_instances() { + while (geometry_instance_dirty_list.first()) { + _geometry_instance_update(geometry_instance_dirty_list.first()->self()); + } +} + +void RenderForwardClustered::_geometry_instance_dependency_changed(Dependency::DependencyChangedNotification p_notification, DependencyTracker *p_tracker) { + switch (p_notification) { + case Dependency::DEPENDENCY_CHANGED_MATERIAL: + case Dependency::DEPENDENCY_CHANGED_MESH: + case Dependency::DEPENDENCY_CHANGED_PARTICLES: + case Dependency::DEPENDENCY_CHANGED_MULTIMESH: + case Dependency::DEPENDENCY_CHANGED_SKELETON_DATA: { + static_cast<RenderGeometryInstance *>(p_tracker->userdata)->_mark_dirty(); + } break; + case Dependency::DEPENDENCY_CHANGED_MULTIMESH_VISIBLE_INSTANCES: { + GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_tracker->userdata); + if (ginstance->data->base_type == RS::INSTANCE_MULTIMESH) { + ginstance->instance_count = RendererRD::MeshStorage::get_singleton()->multimesh_get_instances_to_draw(ginstance->data->base); + } + } break; + default: { + //rest of notifications of no interest + } break; + } +} +void RenderForwardClustered::_geometry_instance_dependency_deleted(const RID &p_dependency, DependencyTracker *p_tracker) { + static_cast<RenderGeometryInstance *>(p_tracker->userdata)->_mark_dirty(); +} + +RenderGeometryInstance *RenderForwardClustered::geometry_instance_create(RID p_base) { + RS::InstanceType type = RSG::utilities->get_base_type(p_base); + ERR_FAIL_COND_V(!((1 << type) & RS::INSTANCE_GEOMETRY_MASK), nullptr); + + GeometryInstanceForwardClustered *ginstance = geometry_instance_alloc.alloc(); + ginstance->data = memnew(GeometryInstanceForwardClustered::Data); + + ginstance->data->base = p_base; + ginstance->data->base_type = type; + ginstance->data->dependency_tracker.userdata = ginstance; + ginstance->data->dependency_tracker.changed_callback = _geometry_instance_dependency_changed; + ginstance->data->dependency_tracker.deleted_callback = _geometry_instance_dependency_deleted; + + ginstance->_mark_dirty(); + + return ginstance; +} + +void RenderForwardClustered::GeometryInstanceForwardClustered::set_transform(const Transform3D &p_transform, const AABB &p_aabb, const AABB &p_transformed_aabbb) { + uint64_t frame = RSG::rasterizer->get_frame_number(); + if (frame != prev_transform_change_frame) { + prev_transform = transform; + prev_transform_change_frame = frame; + prev_transform_dirty = true; + } + + RenderGeometryInstanceBase::set_transform(p_transform, p_aabb, p_transformed_aabbb); +} + +void RenderForwardClustered::GeometryInstanceForwardClustered::set_use_lightmap(RID p_lightmap_instance, const Rect2 &p_lightmap_uv_scale, int p_lightmap_slice_index) { + lightmap_instance = p_lightmap_instance; + lightmap_uv_scale = p_lightmap_uv_scale; + lightmap_slice_index = p_lightmap_slice_index; + + _mark_dirty(); +} + +void RenderForwardClustered::GeometryInstanceForwardClustered::set_lightmap_capture(const Color *p_sh9) { + if (p_sh9) { + if (lightmap_sh == nullptr) { + lightmap_sh = RenderForwardClustered::get_singleton()->geometry_instance_lightmap_sh.alloc(); + } + + memcpy(lightmap_sh->sh, p_sh9, sizeof(Color) * 9); + } else { + if (lightmap_sh != nullptr) { + RenderForwardClustered::get_singleton()->geometry_instance_lightmap_sh.free(lightmap_sh); + lightmap_sh = nullptr; + } + } + _mark_dirty(); +} + +void RenderForwardClustered::geometry_instance_free(RenderGeometryInstance *p_geometry_instance) { + GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + if (ginstance->lightmap_sh != nullptr) { + geometry_instance_lightmap_sh.free(ginstance->lightmap_sh); + } + GeometryInstanceSurfaceDataCache *surf = ginstance->surface_caches; + while (surf) { + GeometryInstanceSurfaceDataCache *next = surf->next; + geometry_instance_surface_alloc.free(surf); + surf = next; + } + memdelete(ginstance->data); + geometry_instance_alloc.free(ginstance); +} + +uint32_t RenderForwardClustered::geometry_instance_get_pair_mask() { + return (1 << RS::INSTANCE_VOXEL_GI); +} + +void RenderForwardClustered::GeometryInstanceForwardClustered::pair_voxel_gi_instances(const RID *p_voxel_gi_instances, uint32_t p_voxel_gi_instance_count) { + if (p_voxel_gi_instance_count > 0) { + voxel_gi_instances[0] = p_voxel_gi_instances[0]; + } else { + voxel_gi_instances[0] = RID(); + } + + if (p_voxel_gi_instance_count > 1) { + voxel_gi_instances[1] = p_voxel_gi_instances[1]; + } else { + voxel_gi_instances[1] = RID(); + } +} + +void RenderForwardClustered::GeometryInstanceForwardClustered::set_softshadow_projector_pairing(bool p_softshadow, bool p_projector) { + using_projectors = p_projector; + using_softshadows = p_softshadow; + _mark_dirty(); +} + +void RenderForwardClustered::_update_shader_quality_settings() { + Vector<RD::PipelineSpecializationConstant> spec_constants; + + RD::PipelineSpecializationConstant sc; + sc.type = RD::PIPELINE_SPECIALIZATION_CONSTANT_TYPE_INT; + + sc.constant_id = SPEC_CONSTANT_SOFT_SHADOW_SAMPLES; + sc.int_value = soft_shadow_samples_get(); + + spec_constants.push_back(sc); + + sc.constant_id = SPEC_CONSTANT_PENUMBRA_SHADOW_SAMPLES; + sc.int_value = penumbra_shadow_samples_get(); + + spec_constants.push_back(sc); + + sc.constant_id = SPEC_CONSTANT_DIRECTIONAL_SOFT_SHADOW_SAMPLES; + sc.int_value = directional_soft_shadow_samples_get(); + + spec_constants.push_back(sc); + + sc.constant_id = SPEC_CONSTANT_DIRECTIONAL_PENUMBRA_SHADOW_SAMPLES; + sc.int_value = directional_penumbra_shadow_samples_get(); + + spec_constants.push_back(sc); + + sc.type = RD::PIPELINE_SPECIALIZATION_CONSTANT_TYPE_BOOL; + sc.constant_id = SPEC_CONSTANT_DECAL_FILTER; + sc.bool_value = decals_get_filter() == RS::DECAL_FILTER_NEAREST_MIPMAPS || + decals_get_filter() == RS::DECAL_FILTER_LINEAR_MIPMAPS || + decals_get_filter() == RS::DECAL_FILTER_NEAREST_MIPMAPS_ANISOTROPIC || + decals_get_filter() == RS::DECAL_FILTER_LINEAR_MIPMAPS_ANISOTROPIC; + + spec_constants.push_back(sc); + + sc.constant_id = SPEC_CONSTANT_PROJECTOR_FILTER; + sc.bool_value = light_projectors_get_filter() == RS::LIGHT_PROJECTOR_FILTER_NEAREST_MIPMAPS || + light_projectors_get_filter() == RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS || + light_projectors_get_filter() == RS::LIGHT_PROJECTOR_FILTER_NEAREST_MIPMAPS_ANISOTROPIC || + light_projectors_get_filter() == RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS_ANISOTROPIC; + + spec_constants.push_back(sc); + + scene_shader.set_default_specialization_constants(spec_constants); + + base_uniforms_changed(); //also need this +} + +RenderForwardClustered::RenderForwardClustered() { + singleton = this; + + /* SCENE SHADER */ + + { + String defines; + defines += "\n#define MAX_ROUGHNESS_LOD " + itos(get_roughness_layers() - 1) + ".0\n"; + if (is_using_radiance_cubemap_array()) { + defines += "\n#define USE_RADIANCE_CUBEMAP_ARRAY \n"; + } + defines += "\n#define SDFGI_OCT_SIZE " + itos(gi.sdfgi_get_lightprobe_octahedron_size()) + "\n"; + defines += "\n#define MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS " + itos(MAX_DIRECTIONAL_LIGHTS) + "\n"; + + { + //lightmaps + scene_state.max_lightmaps = MAX_LIGHTMAPS; + defines += "\n#define MAX_LIGHTMAP_TEXTURES " + itos(scene_state.max_lightmaps) + "\n"; + defines += "\n#define MAX_LIGHTMAPS " + itos(scene_state.max_lightmaps) + "\n"; + + scene_state.lightmap_buffer = RD::get_singleton()->storage_buffer_create(sizeof(LightmapData) * scene_state.max_lightmaps); + } + { + //captures + scene_state.max_lightmap_captures = 2048; + scene_state.lightmap_captures = memnew_arr(LightmapCaptureData, scene_state.max_lightmap_captures); + scene_state.lightmap_capture_buffer = RD::get_singleton()->storage_buffer_create(sizeof(LightmapCaptureData) * scene_state.max_lightmap_captures); + } + { + defines += "\n#define MATERIAL_UNIFORM_SET " + itos(MATERIAL_UNIFORM_SET) + "\n"; + } +#ifdef REAL_T_IS_DOUBLE + { + defines += "\n#define USE_DOUBLE_PRECISION \n"; + } +#endif + + scene_shader.init(defines); + } + + /* shadow sampler */ + { + RD::SamplerState sampler; + sampler.mag_filter = RD::SAMPLER_FILTER_NEAREST; + sampler.min_filter = RD::SAMPLER_FILTER_NEAREST; + sampler.enable_compare = true; + sampler.compare_op = RD::COMPARE_OP_LESS; + shadow_sampler = RD::get_singleton()->sampler_create(sampler); + } + + render_list_thread_threshold = GLOBAL_GET("rendering/limits/forward_renderer/threaded_render_minimum_instances"); + + _update_shader_quality_settings(); + + resolve_effects = memnew(RendererRD::Resolve()); + taa = memnew(RendererRD::TAA); + ss_effects = memnew(RendererRD::SSEffects); +} + +RenderForwardClustered::~RenderForwardClustered() { + if (ss_effects != nullptr) { + memdelete(ss_effects); + ss_effects = nullptr; + } + + if (taa != nullptr) { + memdelete(taa); + taa = nullptr; + } + + if (resolve_effects != nullptr) { + memdelete(resolve_effects); + resolve_effects = nullptr; + } + + RD::get_singleton()->free(shadow_sampler); + RSG::light_storage->directional_shadow_atlas_set_size(0); + + { + for (uint32_t i = 0; i < scene_state.uniform_buffers.size(); i++) { + RD::get_singleton()->free(scene_state.uniform_buffers[i]); + } + for (uint32_t i = 0; i < scene_state.implementation_uniform_buffers.size(); i++) { + RD::get_singleton()->free(scene_state.implementation_uniform_buffers[i]); + } + RD::get_singleton()->free(scene_state.lightmap_buffer); + RD::get_singleton()->free(scene_state.lightmap_capture_buffer); + for (uint32_t i = 0; i < RENDER_LIST_MAX; i++) { + if (scene_state.instance_buffer[i] != RID()) { + RD::get_singleton()->free(scene_state.instance_buffer[i]); + } + } + memdelete_arr(scene_state.lightmap_captures); + } + + while (sdfgi_framebuffer_size_cache.begin()) { + RD::get_singleton()->free(sdfgi_framebuffer_size_cache.begin()->value); + sdfgi_framebuffer_size_cache.remove(sdfgi_framebuffer_size_cache.begin()); + } +} diff --git a/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h new file mode 100644 index 0000000000..670eb93e20 --- /dev/null +++ b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h @@ -0,0 +1,664 @@ +/*************************************************************************/ +/* render_forward_clustered.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef RENDER_FORWARD_CLUSTERED_H +#define RENDER_FORWARD_CLUSTERED_H + +#include "core/templates/paged_allocator.h" +#include "servers/rendering/renderer_rd/cluster_builder_rd.h" +#include "servers/rendering/renderer_rd/effects/resolve.h" +#include "servers/rendering/renderer_rd/effects/ss_effects.h" +#include "servers/rendering/renderer_rd/effects/taa.h" +#include "servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.h" +#include "servers/rendering/renderer_rd/pipeline_cache_rd.h" +#include "servers/rendering/renderer_rd/renderer_scene_render_rd.h" +#include "servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered.glsl.gen.h" +#include "servers/rendering/renderer_rd/storage_rd/utilities.h" + +#define RB_SCOPE_FORWARD_CLUSTERED SNAME("forward_clustered") + +#define RB_TEX_SPECULAR SNAME("specular") +#define RB_TEX_SPECULAR_MSAA SNAME("specular_msaa") +#define RB_TEX_ROUGHNESS SNAME("normal_roughnesss") +#define RB_TEX_ROUGHNESS_MSAA SNAME("normal_roughnesss_msaa") +#define RB_TEX_VOXEL_GI SNAME("voxel_gi") +#define RB_TEX_VOXEL_GI_MSAA SNAME("voxel_gi_msaa") + +namespace RendererSceneRenderImplementation { + +class RenderForwardClustered : public RendererSceneRenderRD { + friend SceneShaderForwardClustered; + + enum { + SCENE_UNIFORM_SET = 0, + RENDER_PASS_UNIFORM_SET = 1, + TRANSFORMS_UNIFORM_SET = 2, + MATERIAL_UNIFORM_SET = 3 + }; + + enum { + SPEC_CONSTANT_SOFT_SHADOW_SAMPLES = 6, + SPEC_CONSTANT_PENUMBRA_SHADOW_SAMPLES = 7, + SPEC_CONSTANT_DIRECTIONAL_SOFT_SHADOW_SAMPLES = 8, + SPEC_CONSTANT_DIRECTIONAL_PENUMBRA_SHADOW_SAMPLES = 9, + SPEC_CONSTANT_DECAL_FILTER = 10, + SPEC_CONSTANT_PROJECTOR_FILTER = 11, + }; + + enum { + SDFGI_MAX_CASCADES = 8, + MAX_VOXEL_GI_INSTANCESS = 8, + MAX_LIGHTMAPS = 8, + MAX_VOXEL_GI_INSTANCESS_PER_INSTANCE = 2, + INSTANCE_DATA_BUFFER_MIN_SIZE = 4096 + }; + + enum RenderListType { + RENDER_LIST_OPAQUE, //used for opaque objects + RENDER_LIST_ALPHA, //used for transparent objects + RENDER_LIST_SECONDARY, //used for shadows and other objects + RENDER_LIST_MAX + }; + + /* Scene Shader */ + + SceneShaderForwardClustered scene_shader; + + /* Framebuffer */ + + class RenderBufferDataForwardClustered : public RenderBufferCustomDataRD { + GDCLASS(RenderBufferDataForwardClustered, RenderBufferCustomDataRD) + + private: + RenderSceneBuffersRD *render_buffers = nullptr; + RD::TextureSamples texture_samples = RD::TEXTURE_SAMPLES_1; + + public: + ClusterBuilderRD *cluster_builder = nullptr; + + struct SSEffectsData { + RID linear_depth; + Vector<RID> linear_depth_slices; + + RID downsample_uniform_set; + + Projection last_frame_projection; + Transform3D last_frame_transform; + + RendererRD::SSEffects::SSAORenderBuffers ssao; + RendererRD::SSEffects::SSILRenderBuffers ssil; + RendererRD::SSEffects::SSRRenderBuffers ssr; + } ss_effects_data; + + enum DepthFrameBufferType { + DEPTH_FB, + DEPTH_FB_ROUGHNESS, + DEPTH_FB_ROUGHNESS_VOXELGI + }; + + RID render_sdfgi_uniform_set; + + RID get_color_msaa() const { return render_buffers->get_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_COLOR_MSAA); } + RID get_color_msaa(uint32_t p_layer) { return render_buffers->get_texture_slice(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_COLOR_MSAA, p_layer, 0); } + + RID get_depth_msaa() const { return render_buffers->get_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_DEPTH_MSAA); } + RID get_depth_msaa(uint32_t p_layer) { return render_buffers->get_texture_slice(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_DEPTH_MSAA, p_layer, 0); } + + void ensure_specular(); + bool has_specular() const { return render_buffers->has_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_SPECULAR); } + RID get_specular() const { return render_buffers->get_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_SPECULAR); } + RID get_specular(uint32_t p_layer) { return render_buffers->get_texture_slice(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_SPECULAR, p_layer, 0); } + RID get_specular_msaa(uint32_t p_layer) { return render_buffers->get_texture_slice(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_SPECULAR_MSAA, p_layer, 0); } + + void ensure_normal_roughness_texture(); + bool has_normal_roughness() const { return render_buffers->has_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_ROUGHNESS); } + RID get_normal_roughness() const { return render_buffers->get_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_ROUGHNESS); } + RID get_normal_roughness(uint32_t p_layer) { return render_buffers->get_texture_slice(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_ROUGHNESS, p_layer, 0); } + RID get_normal_roughness_msaa() const { return render_buffers->get_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_ROUGHNESS_MSAA); } + RID get_normal_roughness_msaa(uint32_t p_layer) { return render_buffers->get_texture_slice(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_ROUGHNESS_MSAA, p_layer, 0); } + + void ensure_voxelgi(); + bool has_voxelgi() const { return render_buffers->has_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_VOXEL_GI); } + RID get_voxelgi() const { return render_buffers->get_texture(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_VOXEL_GI); } + RID get_voxelgi(uint32_t p_layer) { return render_buffers->get_texture_slice(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_VOXEL_GI, p_layer, 0); } + RID get_voxelgi_msaa(uint32_t p_layer) { return render_buffers->get_texture_slice(RB_SCOPE_FORWARD_CLUSTERED, RB_TEX_VOXEL_GI_MSAA, p_layer, 0); } + + RID get_color_only_fb(); + RID get_color_pass_fb(uint32_t p_color_pass_flags); + RID get_depth_fb(DepthFrameBufferType p_type = DEPTH_FB); + RID get_specular_only_fb(); + + RID get_ao_texture() const { return ss_effects_data.ssao.ao_final; } + RID get_ssil_texture() const { return ss_effects_data.ssil.ssil_final; } + + virtual void configure(RenderSceneBuffersRD *p_render_buffers) override; + virtual void free_data() override; + }; + + virtual void setup_render_buffer_data(Ref<RenderSceneBuffersRD> p_render_buffers) override; + + RID render_base_uniform_set; + + uint64_t lightmap_texture_array_version = 0xFFFFFFFF; + + bool base_uniform_set_updated = false; + void _update_render_base_uniform_set(); + RID _setup_sdfgi_render_pass_uniform_set(RID p_albedo_texture, RID p_emission_texture, RID p_emission_aniso_texture, RID p_geom_facing_texture); + RID _setup_render_pass_uniform_set(RenderListType p_render_list, const RenderDataRD *p_render_data, RID p_radiance_texture, bool p_use_directional_shadow_atlas = false, int p_index = 0); + + enum PassMode { + PASS_MODE_COLOR, + PASS_MODE_SHADOW, + PASS_MODE_SHADOW_DP, + PASS_MODE_DEPTH, + PASS_MODE_DEPTH_NORMAL_ROUGHNESS, + PASS_MODE_DEPTH_NORMAL_ROUGHNESS_VOXEL_GI, + PASS_MODE_DEPTH_MATERIAL, + PASS_MODE_SDF, + }; + + enum ColorPassFlags { + COLOR_PASS_FLAG_TRANSPARENT = 1 << 0, + COLOR_PASS_FLAG_SEPARATE_SPECULAR = 1 << 1, + COLOR_PASS_FLAG_MULTIVIEW = 1 << 2, + COLOR_PASS_FLAG_MOTION_VECTORS = 1 << 3, + }; + + struct GeometryInstanceSurfaceDataCache; + struct RenderElementInfo; + + struct RenderListParameters { + GeometryInstanceSurfaceDataCache **elements = nullptr; + RenderElementInfo *element_info = nullptr; + int element_count = 0; + bool reverse_cull = false; + PassMode pass_mode = PASS_MODE_COLOR; + uint32_t color_pass_flags = 0; + bool no_gi = false; + uint32_t view_count = 1; + RID render_pass_uniform_set; + bool force_wireframe = false; + Vector2 uv_offset; + float lod_distance_multiplier = 0.0; + float screen_mesh_lod_threshold = 0.0; + RD::FramebufferFormatID framebuffer_format = 0; + uint32_t element_offset = 0; + uint32_t barrier = RD::BARRIER_MASK_ALL; + bool use_directional_soft_shadow = false; + + RenderListParameters(GeometryInstanceSurfaceDataCache **p_elements, RenderElementInfo *p_element_info, int p_element_count, bool p_reverse_cull, PassMode p_pass_mode, uint32_t p_color_pass_flags, bool p_no_gi, bool p_use_directional_soft_shadows, RID p_render_pass_uniform_set, bool p_force_wireframe = false, const Vector2 &p_uv_offset = Vector2(), float p_lod_distance_multiplier = 0.0, float p_screen_mesh_lod_threshold = 0.0, uint32_t p_view_count = 1, uint32_t p_element_offset = 0, uint32_t p_barrier = RD::BARRIER_MASK_ALL) { + elements = p_elements; + element_info = p_element_info; + element_count = p_element_count; + reverse_cull = p_reverse_cull; + pass_mode = p_pass_mode; + color_pass_flags = p_color_pass_flags; + no_gi = p_no_gi; + view_count = p_view_count; + render_pass_uniform_set = p_render_pass_uniform_set; + force_wireframe = p_force_wireframe; + uv_offset = p_uv_offset; + lod_distance_multiplier = p_lod_distance_multiplier; + screen_mesh_lod_threshold = p_screen_mesh_lod_threshold; + element_offset = p_element_offset; + barrier = p_barrier; + use_directional_soft_shadow = p_use_directional_soft_shadows; + } + }; + + struct LightmapData { + float normal_xform[12]; + float pad[3]; + float exposure_normalization; + }; + + struct LightmapCaptureData { + float sh[9 * 4]; + }; + + // When changing any of these enums, remember to change the corresponding enums in the shader files as well. + enum { + INSTANCE_DATA_FLAGS_NON_UNIFORM_SCALE = 1 << 4, + INSTANCE_DATA_FLAG_USE_GI_BUFFERS = 1 << 5, + INSTANCE_DATA_FLAG_USE_SDFGI = 1 << 6, + INSTANCE_DATA_FLAG_USE_LIGHTMAP_CAPTURE = 1 << 7, + INSTANCE_DATA_FLAG_USE_LIGHTMAP = 1 << 8, + INSTANCE_DATA_FLAG_USE_SH_LIGHTMAP = 1 << 9, + INSTANCE_DATA_FLAG_USE_VOXEL_GI = 1 << 10, + INSTANCE_DATA_FLAG_PARTICLES = 1 << 11, + INSTANCE_DATA_FLAG_MULTIMESH = 1 << 12, + INSTANCE_DATA_FLAG_MULTIMESH_FORMAT_2D = 1 << 13, + INSTANCE_DATA_FLAG_MULTIMESH_HAS_COLOR = 1 << 14, + INSTANCE_DATA_FLAG_MULTIMESH_HAS_CUSTOM_DATA = 1 << 15, + INSTANCE_DATA_FLAGS_PARTICLE_TRAIL_SHIFT = 16, + INSTANCE_DATA_FLAGS_PARTICLE_TRAIL_MASK = 0xFF, + INSTANCE_DATA_FLAGS_FADE_SHIFT = 24, + INSTANCE_DATA_FLAGS_FADE_MASK = 0xFFUL << INSTANCE_DATA_FLAGS_FADE_SHIFT + }; + + struct SceneState { + // This struct is loaded into Set 1 - Binding 1, populated at start of rendering a frame, must match with shader code + struct UBO { + uint32_t cluster_shift; + uint32_t cluster_width; + uint32_t cluster_type_size; + uint32_t max_cluster_element_count_div_32; + + uint32_t ss_effects_flags; + float ssao_light_affect; + float ssao_ao_affect; + uint32_t pad1; + + float sdf_to_bounds[16]; + + int32_t sdf_offset[3]; + uint32_t pad2; + + int32_t sdf_size[3]; + uint32_t gi_upscale_for_msaa; + + uint32_t volumetric_fog_enabled; + float volumetric_fog_inv_length; + float volumetric_fog_detail_spread; + uint32_t volumetric_fog_pad; + }; + + struct PushConstant { + uint32_t base_index; // + uint32_t uv_offset; //packed + uint32_t multimesh_motion_vectors_current_offset; + uint32_t multimesh_motion_vectors_previous_offset; + }; + + struct InstanceData { + float transform[16]; + float prev_transform[16]; + uint32_t flags; + uint32_t instance_uniforms_ofs; //base offset in global buffer for instance variables + uint32_t gi_offset; //GI information when using lightmapping (VCT or lightmap index) + uint32_t layer_mask; + float lightmap_uv_scale[4]; + }; + + UBO ubo; + + LocalVector<RID> uniform_buffers; + LocalVector<RID> implementation_uniform_buffers; + + LightmapData lightmaps[MAX_LIGHTMAPS]; + RID lightmap_ids[MAX_LIGHTMAPS]; + bool lightmap_has_sh[MAX_LIGHTMAPS]; + uint32_t lightmaps_used = 0; + uint32_t max_lightmaps; + RID lightmap_buffer; + + RID instance_buffer[RENDER_LIST_MAX]; + uint32_t instance_buffer_size[RENDER_LIST_MAX] = { 0, 0, 0 }; + LocalVector<InstanceData> instance_data[RENDER_LIST_MAX]; + + LightmapCaptureData *lightmap_captures = nullptr; + uint32_t max_lightmap_captures; + RID lightmap_capture_buffer; + + RID voxelgi_ids[MAX_VOXEL_GI_INSTANCESS]; + uint32_t voxelgis_used = 0; + + bool used_screen_texture = false; + bool used_normal_texture = false; + bool used_depth_texture = false; + bool used_sss = false; + + struct ShadowPass { + uint32_t element_from; + uint32_t element_count; + bool flip_cull; + PassMode pass_mode; + + RID rp_uniform_set; + Plane camera_plane; + float lod_distance_multiplier; + float screen_mesh_lod_threshold; + + RID framebuffer; + RD::InitialAction initial_depth_action; + RD::FinalAction final_depth_action; + Rect2i rect; + }; + + LocalVector<ShadowPass> shadow_passes; + + } scene_state; + + static RenderForwardClustered *singleton; + + void _setup_environment(const RenderDataRD *p_render_data, bool p_no_fog, const Size2i &p_screen_size, bool p_flip_y, const Color &p_default_bg_color, bool p_opaque_render_buffers = false, bool p_pancake_shadows = false, int p_index = 0); + void _setup_voxelgis(const PagedArray<RID> &p_voxelgis); + void _setup_lightmaps(const RenderDataRD *p_render_data, const PagedArray<RID> &p_lightmaps, const Transform3D &p_cam_transform); + + struct RenderElementInfo { + enum { MAX_REPEATS = (1 << 20) - 1 }; + uint32_t repeat : 20; + uint32_t uses_projector : 1; + uint32_t uses_softshadow : 1; + uint32_t uses_lightmap : 1; + uint32_t uses_forward_gi : 1; + uint32_t lod_index : 8; + }; + + template <PassMode p_pass_mode, uint32_t p_color_pass_flags = 0> + _FORCE_INLINE_ void _render_list_template(RenderingDevice::DrawListID p_draw_list, RenderingDevice::FramebufferFormatID p_framebuffer_Format, RenderListParameters *p_params, uint32_t p_from_element, uint32_t p_to_element); + + void _render_list(RenderingDevice::DrawListID p_draw_list, RenderingDevice::FramebufferFormatID p_framebuffer_Format, RenderListParameters *p_params, uint32_t p_from_element, uint32_t p_to_element); + + LocalVector<RD::DrawListID> thread_draw_lists; + void _render_list_thread_function(uint32_t p_thread, RenderListParameters *p_params); + void _render_list_with_threads(RenderListParameters *p_params, RID p_framebuffer, RD::InitialAction p_initial_color_action, RD::FinalAction p_final_color_action, RD::InitialAction p_initial_depth_action, RD::FinalAction p_final_depth_action, const Vector<Color> &p_clear_color_values = Vector<Color>(), float p_clear_depth = 1.0, uint32_t p_clear_stencil = 0, const Rect2 &p_region = Rect2(), const Vector<RID> &p_storage_textures = Vector<RID>()); + + uint32_t render_list_thread_threshold = 500; + + void _update_instance_data_buffer(RenderListType p_render_list); + void _fill_instance_data(RenderListType p_render_list, int *p_render_info = nullptr, uint32_t p_offset = 0, int32_t p_max_elements = -1, bool p_update_buffer = true); + void _fill_render_list(RenderListType p_render_list, const RenderDataRD *p_render_data, PassMode p_pass_mode, uint32_t p_color_pass_flags, bool p_using_sdfgi = false, bool p_using_opaque_gi = false, bool p_append = false); + + HashMap<Size2i, RID> sdfgi_framebuffer_size_cache; + + struct GeometryInstanceData; + class GeometryInstanceForwardClustered; + + struct GeometryInstanceLightmapSH { + Color sh[9]; + }; + + // Cached data for drawing surfaces + struct GeometryInstanceSurfaceDataCache { + enum { + FLAG_PASS_DEPTH = 1, + FLAG_PASS_OPAQUE = 2, + FLAG_PASS_ALPHA = 4, + FLAG_PASS_SHADOW = 8, + FLAG_USES_SHARED_SHADOW_MATERIAL = 128, + FLAG_USES_SUBSURFACE_SCATTERING = 2048, + FLAG_USES_SCREEN_TEXTURE = 4096, + FLAG_USES_DEPTH_TEXTURE = 8192, + FLAG_USES_NORMAL_TEXTURE = 16384, + FLAG_USES_DOUBLE_SIDED_SHADOWS = 32768, + FLAG_USES_PARTICLE_TRAILS = 65536, + }; + + union { + struct { + uint64_t lod_index : 8; + uint64_t surface_index : 8; + uint64_t geometry_id : 32; + uint64_t material_id_low : 16; + + uint64_t material_id_hi : 16; + uint64_t shader_id : 32; + uint64_t uses_softshadow : 1; + uint64_t uses_projector : 1; + uint64_t uses_forward_gi : 1; + uint64_t uses_lightmap : 1; + uint64_t depth_layer : 4; + uint64_t priority : 8; + }; + struct { + uint64_t sort_key1; + uint64_t sort_key2; + }; + } sort; + + RS::PrimitiveType primitive = RS::PRIMITIVE_MAX; + uint32_t flags = 0; + uint32_t surface_index = 0; + + void *surface = nullptr; + RID material_uniform_set; + SceneShaderForwardClustered::ShaderData *shader = nullptr; + + void *surface_shadow = nullptr; + RID material_uniform_set_shadow; + SceneShaderForwardClustered::ShaderData *shader_shadow = nullptr; + + GeometryInstanceSurfaceDataCache *next = nullptr; + GeometryInstanceForwardClustered *owner = nullptr; + }; + + class GeometryInstanceForwardClustered : public RenderGeometryInstanceBase { + public: + // lightmap + RID lightmap_instance; + Rect2 lightmap_uv_scale; + uint32_t lightmap_slice_index; + GeometryInstanceLightmapSH *lightmap_sh = nullptr; + + //used during rendering + + uint32_t gi_offset_cache = 0; + bool store_transform_cache = true; + RID transforms_uniform_set; + uint32_t instance_count = 0; + uint32_t trail_steps = 1; + bool can_sdfgi = false; + bool using_projectors = false; + bool using_softshadows = false; + + //used during setup + uint64_t prev_transform_change_frame = 0xFFFFFFFF; + bool prev_transform_dirty = true; + Transform3D prev_transform; + RID voxel_gi_instances[MAX_VOXEL_GI_INSTANCESS_PER_INSTANCE]; + GeometryInstanceSurfaceDataCache *surface_caches = nullptr; + SelfList<GeometryInstanceForwardClustered> dirty_list_element; + + GeometryInstanceForwardClustered() : + dirty_list_element(this) {} + + virtual void _mark_dirty() override; + + virtual void set_transform(const Transform3D &p_transform, const AABB &p_aabb, const AABB &p_transformed_aabbb) override; + virtual void set_use_lightmap(RID p_lightmap_instance, const Rect2 &p_lightmap_uv_scale, int p_lightmap_slice_index) override; + virtual void set_lightmap_capture(const Color *p_sh9) override; + + virtual void pair_light_instances(const RID *p_light_instances, uint32_t p_light_instance_count) override {} + virtual void pair_reflection_probe_instances(const RID *p_reflection_probe_instances, uint32_t p_reflection_probe_instance_count) override {} + virtual void pair_decal_instances(const RID *p_decal_instances, uint32_t p_decal_instance_count) override {} + virtual void pair_voxel_gi_instances(const RID *p_voxel_gi_instances, uint32_t p_voxel_gi_instance_count) override; + + virtual void set_softshadow_projector_pairing(bool p_softshadow, bool p_projector) override; + }; + + static void _geometry_instance_dependency_changed(Dependency::DependencyChangedNotification p_notification, DependencyTracker *p_tracker); + static void _geometry_instance_dependency_deleted(const RID &p_dependency, DependencyTracker *p_tracker); + + SelfList<GeometryInstanceForwardClustered>::List geometry_instance_dirty_list; + + PagedAllocator<GeometryInstanceForwardClustered> geometry_instance_alloc; + PagedAllocator<GeometryInstanceSurfaceDataCache> geometry_instance_surface_alloc; + PagedAllocator<GeometryInstanceLightmapSH> geometry_instance_lightmap_sh; + + void _geometry_instance_add_surface_with_material(GeometryInstanceForwardClustered *ginstance, uint32_t p_surface, SceneShaderForwardClustered::MaterialData *p_material, uint32_t p_material_id, uint32_t p_shader_id, RID p_mesh); + void _geometry_instance_add_surface_with_material_chain(GeometryInstanceForwardClustered *ginstance, uint32_t p_surface, SceneShaderForwardClustered::MaterialData *p_material, RID p_mat_src, RID p_mesh); + void _geometry_instance_add_surface(GeometryInstanceForwardClustered *ginstance, uint32_t p_surface, RID p_material, RID p_mesh); + void _geometry_instance_update(RenderGeometryInstance *p_geometry_instance); + void _update_dirty_geometry_instances(); + + /* Render List */ + + struct RenderList { + LocalVector<GeometryInstanceSurfaceDataCache *> elements; + LocalVector<RenderElementInfo> element_info; + + void clear() { + elements.clear(); + element_info.clear(); + } + + //should eventually be replaced by radix + + struct SortByKey { + _FORCE_INLINE_ bool operator()(const GeometryInstanceSurfaceDataCache *A, const GeometryInstanceSurfaceDataCache *B) const { + return (A->sort.sort_key2 == B->sort.sort_key2) ? (A->sort.sort_key1 < B->sort.sort_key1) : (A->sort.sort_key2 < B->sort.sort_key2); + } + }; + + void sort_by_key() { + SortArray<GeometryInstanceSurfaceDataCache *, SortByKey> sorter; + sorter.sort(elements.ptr(), elements.size()); + } + + void sort_by_key_range(uint32_t p_from, uint32_t p_size) { + SortArray<GeometryInstanceSurfaceDataCache *, SortByKey> sorter; + sorter.sort(elements.ptr() + p_from, p_size); + } + + struct SortByDepth { + _FORCE_INLINE_ bool operator()(const GeometryInstanceSurfaceDataCache *A, const GeometryInstanceSurfaceDataCache *B) const { + return (A->owner->depth < B->owner->depth); + } + }; + + void sort_by_depth() { //used for shadows + + SortArray<GeometryInstanceSurfaceDataCache *, SortByDepth> sorter; + sorter.sort(elements.ptr(), elements.size()); + } + + struct SortByReverseDepthAndPriority { + _FORCE_INLINE_ bool operator()(const GeometryInstanceSurfaceDataCache *A, const GeometryInstanceSurfaceDataCache *B) const { + return (A->sort.priority == B->sort.priority) ? (A->owner->depth > B->owner->depth) : (A->sort.priority < B->sort.priority); + } + }; + + void sort_by_reverse_depth_and_priority() { //used for alpha + + SortArray<GeometryInstanceSurfaceDataCache *, SortByReverseDepthAndPriority> sorter; + sorter.sort(elements.ptr(), elements.size()); + } + + _FORCE_INLINE_ void add_element(GeometryInstanceSurfaceDataCache *p_element) { + elements.push_back(p_element); + } + }; + + RenderList render_list[RENDER_LIST_MAX]; + + virtual void _update_shader_quality_settings() override; + + /* Effects */ + + RendererRD::Resolve *resolve_effects = nullptr; + RendererRD::TAA *taa = nullptr; + RendererRD::SSEffects *ss_effects = nullptr; + + /* Cluster builder */ + + ClusterBuilderSharedDataRD cluster_builder_shared; + ClusterBuilderRD *current_cluster_builder = nullptr; + + /* SDFGI */ + void _update_sdfgi(RenderDataRD *p_render_data); + + /* Volumetric fog */ + RID shadow_sampler; + + void _update_volumetric_fog(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_environment, const Projection &p_cam_projection, const Transform3D &p_cam_transform, const Transform3D &p_prev_cam_inv_transform, RID p_shadow_atlas, int p_directional_light_count, bool p_use_directional_shadows, int p_positional_light_count, int p_voxel_gi_count, const PagedArray<RID> &p_fog_volumes); + + /* Render shadows */ + + void _render_shadow_pass(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<RenderGeometryInstance *> &p_instances, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0, float p_screen_mesh_lod_threshold = 0.0, bool p_open_pass = true, bool p_close_pass = true, bool p_clear_region = true, RenderingMethod::RenderInfo *p_render_info = nullptr); + void _render_shadow_begin(); + void _render_shadow_append(RID p_framebuffer, const PagedArray<RenderGeometryInstance *> &p_instances, const Projection &p_projection, const Transform3D &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_mesh_lod_threshold = 0.0, const Rect2i &p_rect = Rect2i(), bool p_flip_y = false, bool p_clear_region = true, bool p_begin = true, bool p_end = true, RenderingMethod::RenderInfo *p_render_info = nullptr); + void _render_shadow_process(); + void _render_shadow_end(uint32_t p_barrier = RD::BARRIER_MASK_ALL); + + /* Render Scene */ + void _process_ssao(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_environment, RID p_normal_buffer, const Projection &p_projection); + void _process_ssil(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_environment, RID p_normal_buffer, const Projection &p_projection, const Transform3D &p_transform); + void _copy_framebuffer_to_ssil(Ref<RenderSceneBuffersRD> p_render_buffers); + void _pre_opaque_render(RenderDataRD *p_render_data, bool p_use_ssao, bool p_use_ssil, bool p_use_gi, const RID *p_normal_roughness_slices, RID p_voxel_gi_buffer); + void _process_ssr(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_dest_framebuffer, const RID *p_normal_buffer_slices, RID p_specular_buffer, const RID *p_metallic_slices, RID p_environment, const Projection *p_projections, const Vector3 *p_eye_offsets, bool p_use_additive); + void _process_sss(Ref<RenderSceneBuffersRD> p_render_buffers, const Projection &p_camera); + + /* Debug */ + void _debug_draw_cluster(Ref<RenderSceneBuffersRD> p_render_buffers); + +protected: + /* setup */ + + virtual RID _render_buffers_get_normal_texture(Ref<RenderSceneBuffersRD> p_render_buffers) override; + virtual RID _render_buffers_get_velocity_texture(Ref<RenderSceneBuffersRD> p_render_buffers) override; + + virtual void environment_set_ssao_quality(RS::EnvironmentSSAOQuality p_quality, bool p_half_size, float p_adaptive_target, int p_blur_passes, float p_fadeout_from, float p_fadeout_to) override; + virtual void environment_set_ssil_quality(RS::EnvironmentSSILQuality p_quality, bool p_half_size, float p_adaptive_target, int p_blur_passes, float p_fadeout_from, float p_fadeout_to) override; + virtual void environment_set_ssr_roughness_quality(RS::EnvironmentSSRRoughnessQuality p_quality) override; + + virtual void sub_surface_scattering_set_quality(RS::SubSurfaceScatteringQuality p_quality) override; + virtual void sub_surface_scattering_set_scale(float p_scale, float p_depth_scale) override; + + /* Rendering */ + + virtual void _render_scene(RenderDataRD *p_render_data, const Color &p_default_bg_color) override; + virtual void _render_buffers_debug_draw(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_shadow_atlas, RID p_occlusion_buffer) override; + + virtual void _render_material(const Transform3D &p_cam_transform, const Projection &p_cam_projection, bool p_cam_orthogonal, const PagedArray<RenderGeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region, float p_exposure_normalization) override; + virtual void _render_uv2(const PagedArray<RenderGeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) override; + virtual void _render_sdfgi(Ref<RenderSceneBuffersRD> p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<RenderGeometryInstance *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture, float p_exposure_normalization) override; + virtual void _render_particle_collider_heightfield(RID p_fb, const Transform3D &p_cam_transform, const Projection &p_cam_projection, const PagedArray<RenderGeometryInstance *> &p_instances) override; + +public: + static RenderForwardClustered *get_singleton() { return singleton; } + + ClusterBuilderSharedDataRD *get_cluster_builder_shared() { return &cluster_builder_shared; } + RendererRD::SSEffects *get_ss_effects() { return ss_effects; } + + /* callback from updating our lighting UBOs, used to populate cluster builder */ + virtual void setup_added_reflection_probe(const Transform3D &p_transform, const Vector3 &p_half_extents) override; + virtual void setup_added_light(const RS::LightType p_type, const Transform3D &p_transform, float p_radius, float p_spot_aperture) override; + virtual void setup_added_decal(const Transform3D &p_transform, const Vector3 &p_half_extents) override; + + virtual void base_uniforms_changed() override; + _FORCE_INLINE_ virtual void update_uniform_sets() override { + base_uniform_set_updated = true; + _update_render_base_uniform_set(); + } + + virtual RenderGeometryInstance *geometry_instance_create(RID p_base) override; + virtual void geometry_instance_free(RenderGeometryInstance *p_geometry_instance) override; + + virtual uint32_t geometry_instance_get_pair_mask() override; + + virtual bool free(RID p_rid) override; + + RenderForwardClustered(); + ~RenderForwardClustered(); +}; +} // namespace RendererSceneRenderImplementation + +#endif // RENDER_FORWARD_CLUSTERED_H diff --git a/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.cpp b/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.cpp new file mode 100644 index 0000000000..c1b23af82f --- /dev/null +++ b/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.cpp @@ -0,0 +1,899 @@ +/*************************************************************************/ +/* scene_shader_forward_clustered.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "scene_shader_forward_clustered.h" +#include "core/config/project_settings.h" +#include "core/math/math_defs.h" +#include "render_forward_clustered.h" +#include "servers/rendering/renderer_rd/renderer_compositor_rd.h" +#include "servers/rendering/renderer_rd/storage_rd/material_storage.h" + +using namespace RendererSceneRenderImplementation; + +void SceneShaderForwardClustered::ShaderData::set_path_hint(const String &p_path) { + path = p_path; +} + +void SceneShaderForwardClustered::ShaderData::set_code(const String &p_code) { + //compile + + code = p_code; + valid = false; + ubo_size = 0; + uniforms.clear(); + uses_screen_texture = false; + + if (code.is_empty()) { + return; //just invalid, but no error + } + + ShaderCompiler::GeneratedCode gen_code; + + int blend_mode = BLEND_MODE_MIX; + int depth_testi = DEPTH_TEST_ENABLED; + int alpha_antialiasing_mode = ALPHA_ANTIALIASING_OFF; + int cull_modei = CULL_BACK; + + uses_point_size = false; + uses_alpha = false; + uses_alpha_clip = false; + uses_blend_alpha = false; + uses_depth_pre_pass = false; + uses_discard = false; + uses_roughness = false; + uses_normal = false; + bool wireframe = false; + + unshaded = false; + uses_vertex = false; + uses_position = false; + uses_sss = false; + uses_transmittance = false; + uses_screen_texture = false; + uses_depth_texture = false; + uses_normal_texture = false; + uses_time = false; + writes_modelview_or_projection = false; + uses_world_coordinates = false; + uses_particle_trails = false; + + int depth_drawi = DEPTH_DRAW_OPAQUE; + + ShaderCompiler::IdentifierActions actions; + actions.entry_point_stages["vertex"] = ShaderCompiler::STAGE_VERTEX; + actions.entry_point_stages["fragment"] = ShaderCompiler::STAGE_FRAGMENT; + actions.entry_point_stages["light"] = ShaderCompiler::STAGE_FRAGMENT; + + actions.render_mode_values["blend_add"] = Pair<int *, int>(&blend_mode, BLEND_MODE_ADD); + actions.render_mode_values["blend_mix"] = Pair<int *, int>(&blend_mode, BLEND_MODE_MIX); + actions.render_mode_values["blend_sub"] = Pair<int *, int>(&blend_mode, BLEND_MODE_SUB); + actions.render_mode_values["blend_mul"] = Pair<int *, int>(&blend_mode, BLEND_MODE_MUL); + + actions.render_mode_values["alpha_to_coverage"] = Pair<int *, int>(&alpha_antialiasing_mode, ALPHA_ANTIALIASING_ALPHA_TO_COVERAGE); + actions.render_mode_values["alpha_to_coverage_and_one"] = Pair<int *, int>(&alpha_antialiasing_mode, ALPHA_ANTIALIASING_ALPHA_TO_COVERAGE_AND_TO_ONE); + + actions.render_mode_values["depth_draw_never"] = Pair<int *, int>(&depth_drawi, DEPTH_DRAW_DISABLED); + actions.render_mode_values["depth_draw_opaque"] = Pair<int *, int>(&depth_drawi, DEPTH_DRAW_OPAQUE); + actions.render_mode_values["depth_draw_always"] = Pair<int *, int>(&depth_drawi, DEPTH_DRAW_ALWAYS); + + actions.render_mode_values["depth_test_disabled"] = Pair<int *, int>(&depth_testi, DEPTH_TEST_DISABLED); + + actions.render_mode_values["cull_disabled"] = Pair<int *, int>(&cull_modei, CULL_DISABLED); + actions.render_mode_values["cull_front"] = Pair<int *, int>(&cull_modei, CULL_FRONT); + actions.render_mode_values["cull_back"] = Pair<int *, int>(&cull_modei, CULL_BACK); + + actions.render_mode_flags["unshaded"] = &unshaded; + actions.render_mode_flags["wireframe"] = &wireframe; + actions.render_mode_flags["particle_trails"] = &uses_particle_trails; + + actions.usage_flag_pointers["ALPHA"] = &uses_alpha; + actions.usage_flag_pointers["ALPHA_SCISSOR_THRESHOLD"] = &uses_alpha_clip; + actions.render_mode_flags["depth_prepass_alpha"] = &uses_depth_pre_pass; + + actions.usage_flag_pointers["SSS_STRENGTH"] = &uses_sss; + actions.usage_flag_pointers["SSS_TRANSMITTANCE_DEPTH"] = &uses_transmittance; + + actions.usage_flag_pointers["SCREEN_TEXTURE"] = &uses_screen_texture; + actions.usage_flag_pointers["DEPTH_TEXTURE"] = &uses_depth_texture; + actions.usage_flag_pointers["NORMAL_TEXTURE"] = &uses_normal_texture; + actions.usage_flag_pointers["DISCARD"] = &uses_discard; + actions.usage_flag_pointers["TIME"] = &uses_time; + actions.usage_flag_pointers["ROUGHNESS"] = &uses_roughness; + actions.usage_flag_pointers["NORMAL"] = &uses_normal; + actions.usage_flag_pointers["NORMAL_MAP"] = &uses_normal; + + actions.usage_flag_pointers["POINT_SIZE"] = &uses_point_size; + actions.usage_flag_pointers["POINT_COORD"] = &uses_point_size; + + actions.write_flag_pointers["MODELVIEW_MATRIX"] = &writes_modelview_or_projection; + actions.write_flag_pointers["PROJECTION_MATRIX"] = &writes_modelview_or_projection; + actions.write_flag_pointers["VERTEX"] = &uses_vertex; + actions.write_flag_pointers["POSITION"] = &uses_position; + + actions.uniforms = &uniforms; + + SceneShaderForwardClustered *shader_singleton = (SceneShaderForwardClustered *)SceneShaderForwardClustered::singleton; + Error err = shader_singleton->compiler.compile(RS::SHADER_SPATIAL, code, &actions, path, gen_code); + ERR_FAIL_COND_MSG(err != OK, "Shader compilation failed."); + + if (version.is_null()) { + version = shader_singleton->shader.version_create(); + } + + depth_draw = DepthDraw(depth_drawi); + depth_test = DepthTest(depth_testi); + cull_mode = Cull(cull_modei); + uses_screen_texture_mipmaps = gen_code.uses_screen_texture_mipmaps; + +#if 0 + print_line("**compiling shader:"); + print_line("**defines:\n"); + for (int i = 0; i < gen_code.defines.size(); i++) { + print_line(gen_code.defines[i]); + } + + HashMap<String, String>::Iterator el = gen_code.code.begin(); + while (el) { + print_line("\n**code " + el->key + ":\n" + el->value); + ++el; + } + + print_line("\n**uniforms:\n" + gen_code.uniforms); + print_line("\n**vertex_globals:\n" + gen_code.stage_globals[ShaderCompiler::STAGE_VERTEX]); + print_line("\n**fragment_globals:\n" + gen_code.stage_globals[ShaderCompiler::STAGE_FRAGMENT]); +#endif + shader_singleton->shader.version_set_code(version, gen_code.code, gen_code.uniforms, gen_code.stage_globals[ShaderCompiler::STAGE_VERTEX], gen_code.stage_globals[ShaderCompiler::STAGE_FRAGMENT], gen_code.defines); + ERR_FAIL_COND(!shader_singleton->shader.version_is_valid(version)); + + ubo_size = gen_code.uniform_total_size; + ubo_offsets = gen_code.uniform_offsets; + texture_uniforms = gen_code.texture_uniforms; + + //blend modes + + // if any form of Alpha Antialiasing is enabled, set the blend mode to alpha to coverage + if (alpha_antialiasing_mode != ALPHA_ANTIALIASING_OFF) { + blend_mode = BLEND_MODE_ALPHA_TO_COVERAGE; + } + + RD::PipelineColorBlendState::Attachment blend_attachment; + + switch (blend_mode) { + case BLEND_MODE_MIX: { + blend_attachment.enable_blend = true; + blend_attachment.alpha_blend_op = RD::BLEND_OP_ADD; + blend_attachment.color_blend_op = RD::BLEND_OP_ADD; + blend_attachment.src_color_blend_factor = RD::BLEND_FACTOR_SRC_ALPHA; + blend_attachment.dst_color_blend_factor = RD::BLEND_FACTOR_ONE_MINUS_SRC_ALPHA; + blend_attachment.src_alpha_blend_factor = RD::BLEND_FACTOR_ONE; + blend_attachment.dst_alpha_blend_factor = RD::BLEND_FACTOR_ONE_MINUS_SRC_ALPHA; + + } break; + case BLEND_MODE_ADD: { + blend_attachment.enable_blend = true; + blend_attachment.alpha_blend_op = RD::BLEND_OP_ADD; + blend_attachment.color_blend_op = RD::BLEND_OP_ADD; + blend_attachment.src_color_blend_factor = RD::BLEND_FACTOR_SRC_ALPHA; + blend_attachment.dst_color_blend_factor = RD::BLEND_FACTOR_ONE; + blend_attachment.src_alpha_blend_factor = RD::BLEND_FACTOR_SRC_ALPHA; + blend_attachment.dst_alpha_blend_factor = RD::BLEND_FACTOR_ONE; + uses_blend_alpha = true; //force alpha used because of blend + + } break; + case BLEND_MODE_SUB: { + blend_attachment.enable_blend = true; + blend_attachment.alpha_blend_op = RD::BLEND_OP_SUBTRACT; + blend_attachment.color_blend_op = RD::BLEND_OP_SUBTRACT; + blend_attachment.src_color_blend_factor = RD::BLEND_FACTOR_SRC_ALPHA; + blend_attachment.dst_color_blend_factor = RD::BLEND_FACTOR_ONE; + blend_attachment.src_alpha_blend_factor = RD::BLEND_FACTOR_SRC_ALPHA; + blend_attachment.dst_alpha_blend_factor = RD::BLEND_FACTOR_ONE; + uses_blend_alpha = true; //force alpha used because of blend + + } break; + case BLEND_MODE_MUL: { + blend_attachment.enable_blend = true; + blend_attachment.alpha_blend_op = RD::BLEND_OP_ADD; + blend_attachment.color_blend_op = RD::BLEND_OP_ADD; + blend_attachment.src_color_blend_factor = RD::BLEND_FACTOR_DST_COLOR; + blend_attachment.dst_color_blend_factor = RD::BLEND_FACTOR_ZERO; + blend_attachment.src_alpha_blend_factor = RD::BLEND_FACTOR_DST_ALPHA; + blend_attachment.dst_alpha_blend_factor = RD::BLEND_FACTOR_ZERO; + uses_blend_alpha = true; //force alpha used because of blend + } break; + case BLEND_MODE_ALPHA_TO_COVERAGE: { + blend_attachment.enable_blend = true; + blend_attachment.alpha_blend_op = RD::BLEND_OP_ADD; + blend_attachment.color_blend_op = RD::BLEND_OP_ADD; + blend_attachment.src_color_blend_factor = RD::BLEND_FACTOR_SRC_ALPHA; + blend_attachment.dst_color_blend_factor = RD::BLEND_FACTOR_ONE_MINUS_SRC_ALPHA; + blend_attachment.src_alpha_blend_factor = RD::BLEND_FACTOR_ONE; + blend_attachment.dst_alpha_blend_factor = RD::BLEND_FACTOR_ZERO; + } + } + + // Color pass -> attachment 0: Color/Diffuse, attachment 1: Separate Specular, attachment 2: Motion Vectors + RD::PipelineColorBlendState blend_state_color_blend; + blend_state_color_blend.attachments = { blend_attachment, RD::PipelineColorBlendState::Attachment(), RD::PipelineColorBlendState::Attachment() }; + RD::PipelineColorBlendState blend_state_color_opaque = RD::PipelineColorBlendState::create_disabled(3); + RD::PipelineColorBlendState blend_state_depth_normal_roughness = RD::PipelineColorBlendState::create_disabled(1); + RD::PipelineColorBlendState blend_state_depth_normal_roughness_giprobe = RD::PipelineColorBlendState::create_disabled(2); + + //update pipelines + + RD::PipelineDepthStencilState depth_stencil_state; + + if (depth_test != DEPTH_TEST_DISABLED) { + depth_stencil_state.enable_depth_test = true; + depth_stencil_state.depth_compare_operator = RD::COMPARE_OP_LESS_OR_EQUAL; + depth_stencil_state.enable_depth_write = depth_draw != DEPTH_DRAW_DISABLED ? true : false; + } + + for (int i = 0; i < CULL_VARIANT_MAX; i++) { + RD::PolygonCullMode cull_mode_rd_table[CULL_VARIANT_MAX][3] = { + { RD::POLYGON_CULL_DISABLED, RD::POLYGON_CULL_FRONT, RD::POLYGON_CULL_BACK }, + { RD::POLYGON_CULL_DISABLED, RD::POLYGON_CULL_BACK, RD::POLYGON_CULL_FRONT }, + { RD::POLYGON_CULL_DISABLED, RD::POLYGON_CULL_DISABLED, RD::POLYGON_CULL_DISABLED } + }; + + RD::PolygonCullMode cull_mode_rd = cull_mode_rd_table[i][cull_mode]; + + for (int j = 0; j < RS::PRIMITIVE_MAX; j++) { + RD::RenderPrimitive primitive_rd_table[RS::PRIMITIVE_MAX] = { + RD::RENDER_PRIMITIVE_POINTS, + RD::RENDER_PRIMITIVE_LINES, + RD::RENDER_PRIMITIVE_LINESTRIPS, + RD::RENDER_PRIMITIVE_TRIANGLES, + RD::RENDER_PRIMITIVE_TRIANGLE_STRIPS, + }; + + RD::RenderPrimitive primitive_rd = uses_point_size ? RD::RENDER_PRIMITIVE_POINTS : primitive_rd_table[j]; + + for (int k = 0; k < PIPELINE_VERSION_MAX; k++) { + ShaderVersion shader_version; + static const ShaderVersion shader_version_table[PIPELINE_VERSION_MAX] = { + SHADER_VERSION_DEPTH_PASS, + SHADER_VERSION_DEPTH_PASS_DP, + SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS, + SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_VOXEL_GI, + SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL, + SHADER_VERSION_DEPTH_PASS_WITH_SDF, + SHADER_VERSION_DEPTH_PASS_MULTIVIEW, + SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_MULTIVIEW, + SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_VOXEL_GI_MULTIVIEW, + SHADER_VERSION_COLOR_PASS, + }; + + shader_version = shader_version_table[k]; + + if (!static_cast<SceneShaderForwardClustered *>(singleton)->shader.is_variant_enabled(shader_version)) { + continue; + } + RD::PipelineRasterizationState raster_state; + raster_state.cull_mode = cull_mode_rd; + raster_state.wireframe = wireframe; + + if (k == PIPELINE_VERSION_COLOR_PASS) { + for (int l = 0; l < PIPELINE_COLOR_PASS_FLAG_COUNT; l++) { + if (!shader_singleton->valid_color_pass_pipelines.has(l)) { + continue; + } + + RD::PipelineColorBlendState blend_state; + RD::PipelineDepthStencilState depth_stencil = depth_stencil_state; + RD::PipelineMultisampleState multisample_state; + + int shader_flags = 0; + if (l & PIPELINE_COLOR_PASS_FLAG_TRANSPARENT) { + if (alpha_antialiasing_mode == ALPHA_ANTIALIASING_ALPHA_TO_COVERAGE) { + multisample_state.enable_alpha_to_coverage = true; + } else if (alpha_antialiasing_mode == ALPHA_ANTIALIASING_ALPHA_TO_COVERAGE_AND_TO_ONE) { + multisample_state.enable_alpha_to_coverage = true; + multisample_state.enable_alpha_to_one = true; + } + + blend_state = blend_state_color_blend; + + if (depth_draw == DEPTH_DRAW_OPAQUE) { + depth_stencil.enable_depth_write = false; //alpha does not draw depth + } + } else { + blend_state = blend_state_color_opaque; + + if (l & PIPELINE_COLOR_PASS_FLAG_SEPARATE_SPECULAR) { + shader_flags |= SHADER_COLOR_PASS_FLAG_SEPARATE_SPECULAR; + } + } + + if (l & PIPELINE_COLOR_PASS_FLAG_MOTION_VECTORS) { + shader_flags |= SHADER_COLOR_PASS_FLAG_MOTION_VECTORS; + } + + if (l & PIPELINE_COLOR_PASS_FLAG_LIGHTMAP) { + shader_flags |= SHADER_COLOR_PASS_FLAG_LIGHTMAP; + } + + if (l & PIPELINE_COLOR_PASS_FLAG_MULTIVIEW) { + shader_flags |= SHADER_COLOR_PASS_FLAG_MULTIVIEW; + } + + int variant = shader_version + shader_flags; + RID shader_variant = shader_singleton->shader.version_get_shader(version, variant); + color_pipelines[i][j][l].setup(shader_variant, primitive_rd, raster_state, multisample_state, depth_stencil, blend_state, 0, singleton->default_specialization_constants); + } + } else { + RD::PipelineColorBlendState blend_state; + RD::PipelineDepthStencilState depth_stencil = depth_stencil_state; + RD::PipelineMultisampleState multisample_state; + + if (k == PIPELINE_VERSION_DEPTH_PASS || k == PIPELINE_VERSION_DEPTH_PASS_DP || k == PIPELINE_VERSION_DEPTH_PASS_MULTIVIEW) { + //none, leave empty + } else if (k == PIPELINE_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS || k == PIPELINE_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_MULTIVIEW) { + blend_state = blend_state_depth_normal_roughness; + } else if (k == PIPELINE_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_VOXEL_GI || k == PIPELINE_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_VOXEL_GI_MULTIVIEW) { + blend_state = blend_state_depth_normal_roughness_giprobe; + } else if (k == PIPELINE_VERSION_DEPTH_PASS_WITH_MATERIAL) { + blend_state = RD::PipelineColorBlendState::create_disabled(5); //writes to normal and roughness in opaque way + } else if (k == PIPELINE_VERSION_DEPTH_PASS_WITH_SDF) { + blend_state = RD::PipelineColorBlendState(); //no color targets for SDF + } + + RID shader_variant = shader_singleton->shader.version_get_shader(version, shader_version); + pipelines[i][j][k].setup(shader_variant, primitive_rd, raster_state, multisample_state, depth_stencil, blend_state, 0, singleton->default_specialization_constants); + } + } + } + } + + valid = true; +} + +void SceneShaderForwardClustered::ShaderData::set_default_texture_parameter(const StringName &p_name, RID p_texture, int p_index) { + if (!p_texture.is_valid()) { + if (default_texture_params.has(p_name) && default_texture_params[p_name].has(p_index)) { + default_texture_params[p_name].erase(p_index); + + if (default_texture_params[p_name].is_empty()) { + default_texture_params.erase(p_name); + } + } + } else { + if (!default_texture_params.has(p_name)) { + default_texture_params[p_name] = HashMap<int, RID>(); + } + default_texture_params[p_name][p_index] = p_texture; + } +} + +void SceneShaderForwardClustered::ShaderData::get_shader_uniform_list(List<PropertyInfo> *p_param_list) const { + HashMap<int, StringName> order; + + for (const KeyValue<StringName, ShaderLanguage::ShaderNode::Uniform> &E : uniforms) { + if (E.value.scope != ShaderLanguage::ShaderNode::Uniform::SCOPE_LOCAL || + E.value.hint == ShaderLanguage::ShaderNode::Uniform::HINT_SCREEN_TEXTURE || + E.value.hint == ShaderLanguage::ShaderNode::Uniform::HINT_NORMAL_ROUGHNESS_TEXTURE || + E.value.hint == ShaderLanguage::ShaderNode::Uniform::HINT_DEPTH_TEXTURE) { + // Don't expose any of these. + continue; + } + + if (E.value.texture_order >= 0) { + order[E.value.texture_order + 100000] = E.key; + } else { + order[E.value.order] = E.key; + } + } + + String last_group; + for (const KeyValue<int, StringName> &E : order) { + String group = uniforms[E.value].group; + if (!uniforms[E.value].subgroup.is_empty()) { + group += "::" + uniforms[E.value].subgroup; + } + + if (group != last_group) { + PropertyInfo pi; + pi.usage = PROPERTY_USAGE_GROUP; + pi.name = group; + p_param_list->push_back(pi); + + last_group = group; + } + + PropertyInfo pi = ShaderLanguage::uniform_to_property_info(uniforms[E.value]); + pi.name = E.value; + p_param_list->push_back(pi); + } +} + +void SceneShaderForwardClustered::ShaderData::get_instance_param_list(List<RendererMaterialStorage::InstanceShaderParam> *p_param_list) const { + for (const KeyValue<StringName, ShaderLanguage::ShaderNode::Uniform> &E : uniforms) { + if (E.value.scope != ShaderLanguage::ShaderNode::Uniform::SCOPE_INSTANCE) { + continue; + } + + RendererMaterialStorage::InstanceShaderParam p; + p.info = ShaderLanguage::uniform_to_property_info(E.value); + p.info.name = E.key; //supply name + p.index = E.value.instance_index; + p.default_value = ShaderLanguage::constant_value_to_variant(E.value.default_value, E.value.type, E.value.array_size, E.value.hint); + p_param_list->push_back(p); + } +} + +bool SceneShaderForwardClustered::ShaderData::is_parameter_texture(const StringName &p_param) const { + if (!uniforms.has(p_param)) { + return false; + } + + return uniforms[p_param].texture_order >= 0; +} + +bool SceneShaderForwardClustered::ShaderData::is_animated() const { + return false; +} + +bool SceneShaderForwardClustered::ShaderData::casts_shadows() const { + return false; +} + +Variant SceneShaderForwardClustered::ShaderData::get_default_parameter(const StringName &p_parameter) const { + if (uniforms.has(p_parameter)) { + ShaderLanguage::ShaderNode::Uniform uniform = uniforms[p_parameter]; + Vector<ShaderLanguage::ConstantNode::Value> default_value = uniform.default_value; + return ShaderLanguage::constant_value_to_variant(default_value, uniform.type, uniform.array_size, uniform.hint); + } + return Variant(); +} + +RS::ShaderNativeSourceCode SceneShaderForwardClustered::ShaderData::get_native_source_code() const { + SceneShaderForwardClustered *shader_singleton = (SceneShaderForwardClustered *)SceneShaderForwardClustered::singleton; + + return shader_singleton->shader.version_get_native_source_code(version); +} + +SceneShaderForwardClustered::ShaderData::ShaderData() : + shader_list_element(this) { +} + +SceneShaderForwardClustered::ShaderData::~ShaderData() { + SceneShaderForwardClustered *shader_singleton = (SceneShaderForwardClustered *)SceneShaderForwardClustered::singleton; + ERR_FAIL_COND(!shader_singleton); + //pipeline variants will clear themselves if shader is gone + if (version.is_valid()) { + shader_singleton->shader.version_free(version); + } +} + +RendererRD::MaterialStorage::ShaderData *SceneShaderForwardClustered::_create_shader_func() { + ShaderData *shader_data = memnew(ShaderData); + singleton->shader_list.add(&shader_data->shader_list_element); + return shader_data; +} + +void SceneShaderForwardClustered::MaterialData::set_render_priority(int p_priority) { + priority = p_priority - RS::MATERIAL_RENDER_PRIORITY_MIN; //8 bits +} + +void SceneShaderForwardClustered::MaterialData::set_next_pass(RID p_pass) { + next_pass = p_pass; +} + +bool SceneShaderForwardClustered::MaterialData::update_parameters(const HashMap<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { + SceneShaderForwardClustered *shader_singleton = (SceneShaderForwardClustered *)SceneShaderForwardClustered::singleton; + + return update_parameters_uniform_set(p_parameters, p_uniform_dirty, p_textures_dirty, shader_data->uniforms, shader_data->ubo_offsets.ptr(), shader_data->texture_uniforms, shader_data->default_texture_params, shader_data->ubo_size, uniform_set, shader_singleton->shader.version_get_shader(shader_data->version, 0), RenderForwardClustered::MATERIAL_UNIFORM_SET, true, RD::BARRIER_MASK_RASTER); +} + +SceneShaderForwardClustered::MaterialData::~MaterialData() { + free_parameters_uniform_set(uniform_set); +} + +RendererRD::MaterialStorage::MaterialData *SceneShaderForwardClustered::_create_material_func(ShaderData *p_shader) { + MaterialData *material_data = memnew(MaterialData); + material_data->shader_data = p_shader; + //update will happen later anyway so do nothing. + return material_data; +} + +SceneShaderForwardClustered *SceneShaderForwardClustered::singleton = nullptr; + +SceneShaderForwardClustered::SceneShaderForwardClustered() { + // there should be only one of these, contained within our RenderFM singleton. + singleton = this; +} + +SceneShaderForwardClustered::~SceneShaderForwardClustered() { + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + + RD::get_singleton()->free(default_vec4_xform_buffer); + RD::get_singleton()->free(shadow_sampler); + + material_storage->shader_free(overdraw_material_shader); + material_storage->shader_free(default_shader); + + material_storage->material_free(overdraw_material); + material_storage->material_free(default_material); +} + +void SceneShaderForwardClustered::init(const String p_defines) { + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + + { + Vector<String> shader_versions; + shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n"); // SHADER_VERSION_DEPTH_PASS + shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_DUAL_PARABOLOID\n"); // SHADER_VERSION_DEPTH_PASS_DP + shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_RENDER_NORMAL_ROUGHNESS\n"); // SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS + shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_RENDER_NORMAL_ROUGHNESS\n#define MODE_RENDER_VOXEL_GI\n"); // SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_VOXEL_GI + shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_RENDER_MATERIAL\n"); // SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL + shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_RENDER_SDF\n"); // SHADER_VERSION_DEPTH_PASS_WITH_SDF + shader_versions.push_back("\n#define USE_MULTIVIEW\n#define MODE_RENDER_DEPTH\n"); // SHADER_VERSION_DEPTH_PASS_MULTIVIEW + shader_versions.push_back("\n#define USE_MULTIVIEW\n#define MODE_RENDER_DEPTH\n#define MODE_RENDER_NORMAL_ROUGHNESS\n"); // SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_MULTIVIEW + shader_versions.push_back("\n#define USE_MULTIVIEW\n#define MODE_RENDER_DEPTH\n#define MODE_RENDER_NORMAL_ROUGHNESS\n#define MODE_RENDER_VOXEL_GI\n"); // SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_VOXEL_GI_MULTIVIEW + + Vector<String> color_pass_flags = { + "\n#define MODE_SEPARATE_SPECULAR\n", // SHADER_COLOR_PASS_FLAG_SEPARATE_SPECULAR + "\n#define USE_LIGHTMAP\n", // SHADER_COLOR_PASS_FLAG_LIGHTMAP + "\n#define USE_MULTIVIEW\n", // SHADER_COLOR_PASS_FLAG_MULTIVIEW + "\n#define MOTION_VECTORS\n", // SHADER_COLOR_PASS_FLAG_MOTION_VECTORS + }; + + for (int i = 0; i < SHADER_COLOR_PASS_FLAG_COUNT; i++) { + String version = ""; + for (int j = 0; (1 << j) < SHADER_COLOR_PASS_FLAG_COUNT; j += 1) { + if ((1 << j) & i) { + version += color_pass_flags[j]; + } + } + shader_versions.push_back(version); + } + + shader.initialize(shader_versions, p_defines); + + if (!RendererCompositorRD::singleton->is_xr_enabled()) { + shader.set_variant_enabled(SHADER_VERSION_DEPTH_PASS_MULTIVIEW, false); + shader.set_variant_enabled(SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_MULTIVIEW, false); + shader.set_variant_enabled(SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_VOXEL_GI_MULTIVIEW, false); + // TODO Add a way to enable/disable color pass flags + } + } + + valid_color_pass_pipelines.insert(0); + + valid_color_pass_pipelines.insert(PIPELINE_COLOR_PASS_FLAG_TRANSPARENT); + valid_color_pass_pipelines.insert(PIPELINE_COLOR_PASS_FLAG_TRANSPARENT | PIPELINE_COLOR_PASS_FLAG_LIGHTMAP); + valid_color_pass_pipelines.insert(PIPELINE_COLOR_PASS_FLAG_TRANSPARENT | PIPELINE_COLOR_PASS_FLAG_MULTIVIEW); + valid_color_pass_pipelines.insert(PIPELINE_COLOR_PASS_FLAG_TRANSPARENT | PIPELINE_COLOR_PASS_FLAG_MOTION_VECTORS); + valid_color_pass_pipelines.insert(PIPELINE_COLOR_PASS_FLAG_TRANSPARENT | PIPELINE_COLOR_PASS_FLAG_LIGHTMAP | PIPELINE_COLOR_PASS_FLAG_MULTIVIEW); + valid_color_pass_pipelines.insert(PIPELINE_COLOR_PASS_FLAG_TRANSPARENT | PIPELINE_COLOR_PASS_FLAG_LIGHTMAP | PIPELINE_COLOR_PASS_FLAG_MOTION_VECTORS); + valid_color_pass_pipelines.insert(PIPELINE_COLOR_PASS_FLAG_TRANSPARENT | PIPELINE_COLOR_PASS_FLAG_MULTIVIEW | PIPELINE_COLOR_PASS_FLAG_MOTION_VECTORS); + valid_color_pass_pipelines.insert(PIPELINE_COLOR_PASS_FLAG_TRANSPARENT | PIPELINE_COLOR_PASS_FLAG_LIGHTMAP | PIPELINE_COLOR_PASS_FLAG_MULTIVIEW | PIPELINE_COLOR_PASS_FLAG_MOTION_VECTORS); + + valid_color_pass_pipelines.insert(PIPELINE_COLOR_PASS_FLAG_SEPARATE_SPECULAR); + valid_color_pass_pipelines.insert(PIPELINE_COLOR_PASS_FLAG_SEPARATE_SPECULAR | PIPELINE_COLOR_PASS_FLAG_LIGHTMAP); + valid_color_pass_pipelines.insert(PIPELINE_COLOR_PASS_FLAG_SEPARATE_SPECULAR | PIPELINE_COLOR_PASS_FLAG_MULTIVIEW); + valid_color_pass_pipelines.insert(PIPELINE_COLOR_PASS_FLAG_SEPARATE_SPECULAR | PIPELINE_COLOR_PASS_FLAG_MOTION_VECTORS); + valid_color_pass_pipelines.insert(PIPELINE_COLOR_PASS_FLAG_SEPARATE_SPECULAR | PIPELINE_COLOR_PASS_FLAG_LIGHTMAP | PIPELINE_COLOR_PASS_FLAG_MULTIVIEW); + valid_color_pass_pipelines.insert(PIPELINE_COLOR_PASS_FLAG_SEPARATE_SPECULAR | PIPELINE_COLOR_PASS_FLAG_LIGHTMAP | PIPELINE_COLOR_PASS_FLAG_MOTION_VECTORS); + valid_color_pass_pipelines.insert(PIPELINE_COLOR_PASS_FLAG_SEPARATE_SPECULAR | PIPELINE_COLOR_PASS_FLAG_LIGHTMAP | PIPELINE_COLOR_PASS_FLAG_MULTIVIEW | PIPELINE_COLOR_PASS_FLAG_MOTION_VECTORS); + + valid_color_pass_pipelines.insert(PIPELINE_COLOR_PASS_FLAG_LIGHTMAP); + valid_color_pass_pipelines.insert(PIPELINE_COLOR_PASS_FLAG_LIGHTMAP | PIPELINE_COLOR_PASS_FLAG_MULTIVIEW); + valid_color_pass_pipelines.insert(PIPELINE_COLOR_PASS_FLAG_LIGHTMAP | PIPELINE_COLOR_PASS_FLAG_MOTION_VECTORS); + valid_color_pass_pipelines.insert(PIPELINE_COLOR_PASS_FLAG_LIGHTMAP | PIPELINE_COLOR_PASS_FLAG_MULTIVIEW | PIPELINE_COLOR_PASS_FLAG_MOTION_VECTORS); + + valid_color_pass_pipelines.insert(PIPELINE_COLOR_PASS_FLAG_MULTIVIEW); + valid_color_pass_pipelines.insert(PIPELINE_COLOR_PASS_FLAG_MULTIVIEW | PIPELINE_COLOR_PASS_FLAG_MOTION_VECTORS); + + valid_color_pass_pipelines.insert(PIPELINE_COLOR_PASS_FLAG_MOTION_VECTORS); + + material_storage->shader_set_data_request_function(RendererRD::MaterialStorage::SHADER_TYPE_3D, _create_shader_funcs); + material_storage->material_set_data_request_function(RendererRD::MaterialStorage::SHADER_TYPE_3D, _create_material_funcs); + + { + //shader compiler + ShaderCompiler::DefaultIdentifierActions actions; + + actions.renames["MODEL_MATRIX"] = "read_model_matrix"; + actions.renames["MODEL_NORMAL_MATRIX"] = "model_normal_matrix"; + actions.renames["VIEW_MATRIX"] = "scene_data.view_matrix"; + actions.renames["INV_VIEW_MATRIX"] = "inv_view_matrix"; + actions.renames["PROJECTION_MATRIX"] = "projection_matrix"; + actions.renames["INV_PROJECTION_MATRIX"] = "inv_projection_matrix"; + actions.renames["MODELVIEW_MATRIX"] = "modelview"; + actions.renames["MODELVIEW_NORMAL_MATRIX"] = "modelview_normal"; + + actions.renames["VERTEX"] = "vertex"; + actions.renames["NORMAL"] = "normal"; + actions.renames["TANGENT"] = "tangent"; + actions.renames["BINORMAL"] = "binormal"; + actions.renames["POSITION"] = "position"; + actions.renames["UV"] = "uv_interp"; + actions.renames["UV2"] = "uv2_interp"; + actions.renames["COLOR"] = "color_interp"; + actions.renames["POINT_SIZE"] = "gl_PointSize"; + actions.renames["INSTANCE_ID"] = "gl_InstanceIndex"; + actions.renames["VERTEX_ID"] = "gl_VertexIndex"; + + actions.renames["ALPHA_SCISSOR_THRESHOLD"] = "alpha_scissor_threshold"; + actions.renames["ALPHA_HASH_SCALE"] = "alpha_hash_scale"; + actions.renames["ALPHA_ANTIALIASING_EDGE"] = "alpha_antialiasing_edge"; + actions.renames["ALPHA_TEXTURE_COORDINATE"] = "alpha_texture_coordinate"; + + //builtins + + actions.renames["TIME"] = "global_time"; + actions.renames["PI"] = _MKSTR(Math_PI); + actions.renames["TAU"] = _MKSTR(Math_TAU); + actions.renames["E"] = _MKSTR(Math_E); + actions.renames["VIEWPORT_SIZE"] = "scene_data.viewport_size"; + + actions.renames["FRAGCOORD"] = "gl_FragCoord"; + actions.renames["FRONT_FACING"] = "gl_FrontFacing"; + actions.renames["NORMAL_MAP"] = "normal_map"; + actions.renames["NORMAL_MAP_DEPTH"] = "normal_map_depth"; + actions.renames["ALBEDO"] = "albedo"; + actions.renames["ALPHA"] = "alpha"; + actions.renames["METALLIC"] = "metallic"; + actions.renames["SPECULAR"] = "specular"; + actions.renames["ROUGHNESS"] = "roughness"; + actions.renames["RIM"] = "rim"; + actions.renames["RIM_TINT"] = "rim_tint"; + actions.renames["CLEARCOAT"] = "clearcoat"; + actions.renames["CLEARCOAT_ROUGHNESS"] = "clearcoat_roughness"; + actions.renames["ANISOTROPY"] = "anisotropy"; + actions.renames["ANISOTROPY_FLOW"] = "anisotropy_flow"; + actions.renames["SSS_STRENGTH"] = "sss_strength"; + actions.renames["SSS_TRANSMITTANCE_COLOR"] = "transmittance_color"; + actions.renames["SSS_TRANSMITTANCE_DEPTH"] = "transmittance_depth"; + actions.renames["SSS_TRANSMITTANCE_BOOST"] = "transmittance_boost"; + actions.renames["BACKLIGHT"] = "backlight"; + actions.renames["AO"] = "ao"; + actions.renames["AO_LIGHT_AFFECT"] = "ao_light_affect"; + actions.renames["EMISSION"] = "emission"; + actions.renames["POINT_COORD"] = "gl_PointCoord"; + actions.renames["INSTANCE_CUSTOM"] = "instance_custom"; + actions.renames["SCREEN_UV"] = "screen_uv"; + actions.renames["SCREEN_TEXTURE"] = "color_buffer"; + actions.renames["DEPTH_TEXTURE"] = "depth_buffer"; + actions.renames["NORMAL_ROUGHNESS_TEXTURE"] = "normal_roughness_buffer"; + actions.renames["DEPTH"] = "gl_FragDepth"; + actions.renames["OUTPUT_IS_SRGB"] = "true"; + actions.renames["FOG"] = "fog"; + actions.renames["RADIANCE"] = "custom_radiance"; + actions.renames["IRRADIANCE"] = "custom_irradiance"; + actions.renames["BONE_INDICES"] = "bone_attrib"; + actions.renames["BONE_WEIGHTS"] = "weight_attrib"; + actions.renames["CUSTOM0"] = "custom0_attrib"; + actions.renames["CUSTOM1"] = "custom1_attrib"; + actions.renames["CUSTOM2"] = "custom2_attrib"; + actions.renames["CUSTOM3"] = "custom3_attrib"; + actions.renames["OUTPUT_IS_SRGB"] = "SHADER_IS_SRGB"; + + actions.renames["NODE_POSITION_WORLD"] = "read_model_matrix[3].xyz"; + actions.renames["CAMERA_POSITION_WORLD"] = "scene_data.inv_view_matrix[3].xyz"; + actions.renames["CAMERA_DIRECTION_WORLD"] = "scene_data.view_matrix[3].xyz"; + actions.renames["NODE_POSITION_VIEW"] = "(read_model_matrix * scene_data.view_matrix)[3].xyz"; + + actions.renames["VIEW_INDEX"] = "ViewIndex"; + actions.renames["VIEW_MONO_LEFT"] = "0"; + actions.renames["VIEW_RIGHT"] = "1"; + + //for light + actions.renames["VIEW"] = "view"; + actions.renames["LIGHT_COLOR"] = "light_color"; + actions.renames["LIGHT"] = "light"; + actions.renames["ATTENUATION"] = "attenuation"; + actions.renames["DIFFUSE_LIGHT"] = "diffuse_light"; + actions.renames["SPECULAR_LIGHT"] = "specular_light"; + + actions.usage_defines["NORMAL"] = "#define NORMAL_USED\n"; + actions.usage_defines["TANGENT"] = "#define TANGENT_USED\n"; + actions.usage_defines["BINORMAL"] = "@TANGENT"; + actions.usage_defines["RIM"] = "#define LIGHT_RIM_USED\n"; + actions.usage_defines["RIM_TINT"] = "@RIM"; + actions.usage_defines["CLEARCOAT"] = "#define LIGHT_CLEARCOAT_USED\n"; + actions.usage_defines["CLEARCOAT_ROUGHNESS"] = "@CLEARCOAT"; + actions.usage_defines["ANISOTROPY"] = "#define LIGHT_ANISOTROPY_USED\n"; + actions.usage_defines["ANISOTROPY_FLOW"] = "@ANISOTROPY"; + actions.usage_defines["AO"] = "#define AO_USED\n"; + actions.usage_defines["AO_LIGHT_AFFECT"] = "#define AO_USED\n"; + actions.usage_defines["UV"] = "#define UV_USED\n"; + actions.usage_defines["UV2"] = "#define UV2_USED\n"; + actions.usage_defines["BONE_INDICES"] = "#define BONES_USED\n"; + actions.usage_defines["BONE_WEIGHTS"] = "#define WEIGHTS_USED\n"; + actions.usage_defines["CUSTOM0"] = "#define CUSTOM0_USED\n"; + actions.usage_defines["CUSTOM1"] = "#define CUSTOM1_USED\n"; + actions.usage_defines["CUSTOM2"] = "#define CUSTOM2_USED\n"; + actions.usage_defines["CUSTOM3"] = "#define CUSTOM3_USED\n"; + actions.usage_defines["NORMAL_MAP"] = "#define NORMAL_MAP_USED\n"; + actions.usage_defines["NORMAL_MAP_DEPTH"] = "@NORMAL_MAP"; + actions.usage_defines["COLOR"] = "#define COLOR_USED\n"; + actions.usage_defines["INSTANCE_CUSTOM"] = "#define ENABLE_INSTANCE_CUSTOM\n"; + actions.usage_defines["POSITION"] = "#define OVERRIDE_POSITION\n"; + + actions.usage_defines["ALPHA_SCISSOR_THRESHOLD"] = "#define ALPHA_SCISSOR_USED\n"; + actions.usage_defines["ALPHA_HASH_SCALE"] = "#define ALPHA_HASH_USED\n"; + actions.usage_defines["ALPHA_ANTIALIASING_EDGE"] = "#define ALPHA_ANTIALIASING_EDGE_USED\n"; + actions.usage_defines["ALPHA_TEXTURE_COORDINATE"] = "@ALPHA_ANTIALIASING_EDGE"; + + actions.usage_defines["SSS_STRENGTH"] = "#define ENABLE_SSS\n"; + actions.usage_defines["SSS_TRANSMITTANCE_DEPTH"] = "#define ENABLE_TRANSMITTANCE\n"; + actions.usage_defines["BACKLIGHT"] = "#define LIGHT_BACKLIGHT_USED\n"; + actions.usage_defines["SCREEN_TEXTURE"] = "#define SCREEN_TEXTURE_USED\n"; + actions.usage_defines["SCREEN_UV"] = "#define SCREEN_UV_USED\n"; + + actions.usage_defines["DIFFUSE_LIGHT"] = "#define USE_LIGHT_SHADER_CODE\n"; + actions.usage_defines["SPECULAR_LIGHT"] = "#define USE_LIGHT_SHADER_CODE\n"; + + actions.usage_defines["FOG"] = "#define CUSTOM_FOG_USED\n"; + actions.usage_defines["RADIANCE"] = "#define CUSTOM_RADIANCE_USED\n"; + actions.usage_defines["IRRADIANCE"] = "#define CUSTOM_IRRADIANCE_USED\n"; + + actions.usage_defines["MODEL_MATRIX"] = "#define MODEL_MATRIX_USED\n"; + + actions.render_mode_defines["skip_vertex_transform"] = "#define SKIP_TRANSFORM_USED\n"; + actions.render_mode_defines["world_vertex_coords"] = "#define VERTEX_WORLD_COORDS_USED\n"; + actions.render_mode_defines["ensure_correct_normals"] = "#define ENSURE_CORRECT_NORMALS\n"; + actions.render_mode_defines["cull_front"] = "#define DO_SIDE_CHECK\n"; + actions.render_mode_defines["cull_disabled"] = "#define DO_SIDE_CHECK\n"; + actions.render_mode_defines["particle_trails"] = "#define USE_PARTICLE_TRAILS\n"; + actions.render_mode_defines["depth_draw_opaque"] = "#define USE_OPAQUE_PREPASS\n"; + + bool force_lambert = GLOBAL_GET("rendering/shading/overrides/force_lambert_over_burley"); + + if (!force_lambert) { + actions.render_mode_defines["diffuse_burley"] = "#define DIFFUSE_BURLEY\n"; + } + + actions.render_mode_defines["diffuse_lambert_wrap"] = "#define DIFFUSE_LAMBERT_WRAP\n"; + actions.render_mode_defines["diffuse_toon"] = "#define DIFFUSE_TOON\n"; + + actions.render_mode_defines["sss_mode_skin"] = "#define SSS_MODE_SKIN\n"; + + actions.render_mode_defines["specular_schlick_ggx"] = "#define SPECULAR_SCHLICK_GGX\n"; + + actions.custom_samplers["SCREEN_TEXTURE"] = "material_samplers[3]"; // linear filter with mipmaps + actions.custom_samplers["DEPTH_TEXTURE"] = "material_samplers[3]"; + actions.custom_samplers["NORMAL_ROUGHNESS_TEXTURE"] = "material_samplers[1]"; // linear filter + + actions.render_mode_defines["specular_toon"] = "#define SPECULAR_TOON\n"; + actions.render_mode_defines["specular_disabled"] = "#define SPECULAR_DISABLED\n"; + actions.render_mode_defines["shadows_disabled"] = "#define SHADOWS_DISABLED\n"; + actions.render_mode_defines["ambient_light_disabled"] = "#define AMBIENT_LIGHT_DISABLED\n"; + actions.render_mode_defines["shadow_to_opacity"] = "#define USE_SHADOW_TO_OPACITY\n"; + actions.render_mode_defines["unshaded"] = "#define MODE_UNSHADED\n"; + + actions.sampler_array_name = "material_samplers"; + actions.base_texture_binding_index = 1; + actions.texture_layout_set = RenderForwardClustered::MATERIAL_UNIFORM_SET; + actions.base_uniform_string = "material."; + actions.base_varying_index = 11; + + actions.default_filter = ShaderLanguage::FILTER_LINEAR_MIPMAP; + actions.default_repeat = ShaderLanguage::REPEAT_ENABLE; + actions.global_buffer_array_variable = "global_shader_uniforms.data"; + actions.instance_uniform_index_variable = "instances.data[instance_index_interp].instance_uniforms_ofs"; + + compiler.initialize(actions); + } + + { + //default material and shader + default_shader = material_storage->shader_allocate(); + material_storage->shader_initialize(default_shader); + material_storage->shader_set_code(default_shader, R"( +// Default 3D material shader (clustered). + +shader_type spatial; + +void vertex() { + ROUGHNESS = 0.8; +} + +void fragment() { + ALBEDO = vec3(0.6); + ROUGHNESS = 0.8; + METALLIC = 0.2; +} +)"); + default_material = material_storage->material_allocate(); + material_storage->material_initialize(default_material); + material_storage->material_set_shader(default_material, default_shader); + + MaterialData *md = static_cast<MaterialData *>(material_storage->material_get_data(default_material, RendererRD::MaterialStorage::SHADER_TYPE_3D)); + default_shader_rd = shader.version_get_shader(md->shader_data->version, SHADER_VERSION_COLOR_PASS); + default_shader_sdfgi_rd = shader.version_get_shader(md->shader_data->version, SHADER_VERSION_DEPTH_PASS_WITH_SDF); + + default_material_shader_ptr = md->shader_data; + default_material_uniform_set = md->uniform_set; + } + + { + overdraw_material_shader = material_storage->shader_allocate(); + material_storage->shader_initialize(overdraw_material_shader); + // Use relatively low opacity so that more "layers" of overlapping objects can be distinguished. + material_storage->shader_set_code(overdraw_material_shader, R"( +// 3D editor Overdraw debug draw mode shader (clustered). + +shader_type spatial; + +render_mode blend_add, unshaded; + +void fragment() { + ALBEDO = vec3(0.4, 0.8, 0.8); + ALPHA = 0.1; +} +)"); + overdraw_material = material_storage->material_allocate(); + material_storage->material_initialize(overdraw_material); + material_storage->material_set_shader(overdraw_material, overdraw_material_shader); + + MaterialData *md = static_cast<MaterialData *>(material_storage->material_get_data(overdraw_material, RendererRD::MaterialStorage::SHADER_TYPE_3D)); + overdraw_material_shader_ptr = md->shader_data; + overdraw_material_uniform_set = md->uniform_set; + } + + { + default_vec4_xform_buffer = RD::get_singleton()->storage_buffer_create(256); + Vector<RD::Uniform> uniforms; + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.append_id(default_vec4_xform_buffer); + u.binding = 0; + uniforms.push_back(u); + + default_vec4_xform_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, default_shader_rd, RenderForwardClustered::TRANSFORMS_UNIFORM_SET); + } + { + RD::SamplerState sampler; + sampler.mag_filter = RD::SAMPLER_FILTER_LINEAR; + sampler.min_filter = RD::SAMPLER_FILTER_LINEAR; + sampler.enable_compare = true; + sampler.compare_op = RD::COMPARE_OP_LESS; + shadow_sampler = RD::get_singleton()->sampler_create(sampler); + } +} + +void SceneShaderForwardClustered::set_default_specialization_constants(const Vector<RD::PipelineSpecializationConstant> &p_constants) { + default_specialization_constants = p_constants; + for (SelfList<ShaderData> *E = shader_list.first(); E; E = E->next()) { + for (int i = 0; i < ShaderData::CULL_VARIANT_MAX; i++) { + for (int j = 0; j < RS::PRIMITIVE_MAX; j++) { + for (int k = 0; k < SHADER_VERSION_MAX; k++) { + E->self()->pipelines[i][j][k].update_specialization_constants(default_specialization_constants); + } + for (int k = 0; k < PIPELINE_COLOR_PASS_FLAG_COUNT; k++) { + E->self()->color_pipelines[i][j][k].update_specialization_constants(default_specialization_constants); + } + } + } + } +} diff --git a/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.h b/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.h new file mode 100644 index 0000000000..a9a9fa94de --- /dev/null +++ b/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.h @@ -0,0 +1,257 @@ +/*************************************************************************/ +/* scene_shader_forward_clustered.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef SCENE_SHADER_FORWARD_CLUSTERED_H +#define SCENE_SHADER_FORWARD_CLUSTERED_H + +#include "servers/rendering/renderer_rd/renderer_scene_render_rd.h" +#include "servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered.glsl.gen.h" + +namespace RendererSceneRenderImplementation { + +class SceneShaderForwardClustered { +private: + static SceneShaderForwardClustered *singleton; + +public: + enum ShaderVersion { + SHADER_VERSION_DEPTH_PASS, + SHADER_VERSION_DEPTH_PASS_DP, + SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS, + SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_VOXEL_GI, + SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL, + SHADER_VERSION_DEPTH_PASS_WITH_SDF, + SHADER_VERSION_DEPTH_PASS_MULTIVIEW, + SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_MULTIVIEW, + SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_VOXEL_GI_MULTIVIEW, + SHADER_VERSION_COLOR_PASS, + SHADER_VERSION_MAX + }; + + enum ShaderColorPassFlags { + SHADER_COLOR_PASS_FLAG_SEPARATE_SPECULAR = 1 << 0, + SHADER_COLOR_PASS_FLAG_LIGHTMAP = 1 << 1, + SHADER_COLOR_PASS_FLAG_MULTIVIEW = 1 << 2, + SHADER_COLOR_PASS_FLAG_MOTION_VECTORS = 1 << 3, + SHADER_COLOR_PASS_FLAG_COUNT = 1 << 4 + }; + + enum PipelineVersion { + PIPELINE_VERSION_DEPTH_PASS, + PIPELINE_VERSION_DEPTH_PASS_DP, + PIPELINE_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS, + PIPELINE_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_VOXEL_GI, + PIPELINE_VERSION_DEPTH_PASS_WITH_MATERIAL, + PIPELINE_VERSION_DEPTH_PASS_WITH_SDF, + PIPELINE_VERSION_DEPTH_PASS_MULTIVIEW, + PIPELINE_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_MULTIVIEW, + PIPELINE_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_VOXEL_GI_MULTIVIEW, + PIPELINE_VERSION_COLOR_PASS, + PIPELINE_VERSION_MAX + }; + + enum PipelineColorPassFlags { + PIPELINE_COLOR_PASS_FLAG_TRANSPARENT = 1 << 0, + PIPELINE_COLOR_PASS_FLAG_SEPARATE_SPECULAR = 1 << 1, + PIPELINE_COLOR_PASS_FLAG_LIGHTMAP = 1 << 2, + PIPELINE_COLOR_PASS_FLAG_MULTIVIEW = 1 << 3, + PIPELINE_COLOR_PASS_FLAG_MOTION_VECTORS = 1 << 4, + PIPELINE_COLOR_PASS_FLAG_COUNT = 1 << 5, + }; + + enum ShaderSpecializations { + SHADER_SPECIALIZATION_FORWARD_GI = 1 << 0, + SHADER_SPECIALIZATION_PROJECTOR = 1 << 1, + SHADER_SPECIALIZATION_SOFT_SHADOWS = 1 << 2, + SHADER_SPECIALIZATION_DIRECTIONAL_SOFT_SHADOWS = 1 << 3, + }; + + struct ShaderData : public RendererRD::MaterialStorage::ShaderData { + enum BlendMode { //used internally + BLEND_MODE_MIX, + BLEND_MODE_ADD, + BLEND_MODE_SUB, + BLEND_MODE_MUL, + BLEND_MODE_ALPHA_TO_COVERAGE + }; + + enum DepthDraw { + DEPTH_DRAW_DISABLED, + DEPTH_DRAW_OPAQUE, + DEPTH_DRAW_ALWAYS + }; + + enum DepthTest { + DEPTH_TEST_DISABLED, + DEPTH_TEST_ENABLED + }; + + enum Cull { + CULL_DISABLED, + CULL_FRONT, + CULL_BACK + }; + + enum CullVariant { + CULL_VARIANT_NORMAL, + CULL_VARIANT_REVERSED, + CULL_VARIANT_DOUBLE_SIDED, + CULL_VARIANT_MAX + + }; + + enum AlphaAntiAliasing { + ALPHA_ANTIALIASING_OFF, + ALPHA_ANTIALIASING_ALPHA_TO_COVERAGE, + ALPHA_ANTIALIASING_ALPHA_TO_COVERAGE_AND_TO_ONE + }; + + bool valid = false; + RID version; + uint32_t vertex_input_mask = 0; + PipelineCacheRD pipelines[CULL_VARIANT_MAX][RS::PRIMITIVE_MAX][PIPELINE_VERSION_MAX]; + PipelineCacheRD color_pipelines[CULL_VARIANT_MAX][RS::PRIMITIVE_MAX][PIPELINE_COLOR_PASS_FLAG_COUNT]; + + String path; + + HashMap<StringName, ShaderLanguage::ShaderNode::Uniform> uniforms; + Vector<ShaderCompiler::GeneratedCode::Texture> texture_uniforms; + + Vector<uint32_t> ubo_offsets; + uint32_t ubo_size = 0; + + String code; + HashMap<StringName, HashMap<int, RID>> default_texture_params; + + DepthDraw depth_draw; + DepthTest depth_test; + + bool uses_point_size = false; + bool uses_alpha = false; + bool uses_blend_alpha = false; + bool uses_alpha_clip = false; + bool uses_depth_pre_pass = false; + bool uses_discard = false; + bool uses_roughness = false; + bool uses_normal = false; + bool uses_particle_trails = false; + + bool unshaded = false; + bool uses_vertex = false; + bool uses_position = false; + bool uses_sss = false; + bool uses_transmittance = false; + bool uses_screen_texture = false; + bool uses_depth_texture = false; + bool uses_normal_texture = false; + bool uses_time = false; + bool writes_modelview_or_projection = false; + bool uses_world_coordinates = false; + bool uses_screen_texture_mipmaps = false; + Cull cull_mode = CULL_DISABLED; + + uint64_t last_pass = 0; + uint32_t index = 0; + + virtual void set_code(const String &p_Code); + virtual void set_path_hint(const String &p_path); + virtual void set_default_texture_parameter(const StringName &p_name, RID p_texture, int p_index); + virtual void get_shader_uniform_list(List<PropertyInfo> *p_param_list) const; + void get_instance_param_list(List<RendererMaterialStorage::InstanceShaderParam> *p_param_list) const; + + virtual bool is_parameter_texture(const StringName &p_param) const; + virtual bool is_animated() const; + virtual bool casts_shadows() const; + virtual Variant get_default_parameter(const StringName &p_parameter) const; + virtual RS::ShaderNativeSourceCode get_native_source_code() const; + + SelfList<ShaderData> shader_list_element; + ShaderData(); + virtual ~ShaderData(); + }; + + SelfList<ShaderData>::List shader_list; + + RendererRD::MaterialStorage::ShaderData *_create_shader_func(); + static RendererRD::MaterialStorage::ShaderData *_create_shader_funcs() { + return static_cast<SceneShaderForwardClustered *>(singleton)->_create_shader_func(); + } + + struct MaterialData : public RendererRD::MaterialStorage::MaterialData { + ShaderData *shader_data = nullptr; + RID uniform_set; + uint64_t last_pass = 0; + uint32_t index = 0; + RID next_pass; + uint8_t priority; + virtual void set_render_priority(int p_priority); + virtual void set_next_pass(RID p_pass); + virtual bool update_parameters(const HashMap<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); + virtual ~MaterialData(); + }; + + RendererRD::MaterialStorage::MaterialData *_create_material_func(ShaderData *p_shader); + static RendererRD::MaterialStorage::MaterialData *_create_material_funcs(RendererRD::MaterialStorage::ShaderData *p_shader) { + return static_cast<SceneShaderForwardClustered *>(singleton)->_create_material_func(static_cast<ShaderData *>(p_shader)); + } + + SceneForwardClusteredShaderRD shader; + ShaderCompiler compiler; + + RID default_shader; + RID default_material; + RID overdraw_material_shader; + RID overdraw_material; + RID default_shader_rd; + RID default_shader_sdfgi_rd; + + RID default_vec4_xform_buffer; + RID default_vec4_xform_uniform_set; + + RID shadow_sampler; + + RID default_material_uniform_set; + ShaderData *default_material_shader_ptr = nullptr; + + RID overdraw_material_uniform_set; + ShaderData *overdraw_material_shader_ptr = nullptr; + + Vector<RD::PipelineSpecializationConstant> default_specialization_constants; + HashSet<uint32_t> valid_color_pass_pipelines; + SceneShaderForwardClustered(); + ~SceneShaderForwardClustered(); + + void init(const String p_defines); + void set_default_specialization_constants(const Vector<RD::PipelineSpecializationConstant> &p_constants); +}; + +} // namespace RendererSceneRenderImplementation + +#endif // SCENE_SHADER_FORWARD_CLUSTERED_H diff --git a/servers/rendering/renderer_rd/forward_mobile/SCsub b/servers/rendering/renderer_rd/forward_mobile/SCsub new file mode 100644 index 0000000000..86681f9c74 --- /dev/null +++ b/servers/rendering/renderer_rd/forward_mobile/SCsub @@ -0,0 +1,5 @@ +#!/usr/bin/env python + +Import("env") + +env.add_source_files(env.servers_sources, "*.cpp") diff --git a/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.cpp b/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.cpp new file mode 100644 index 0000000000..f76d016ae7 --- /dev/null +++ b/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.cpp @@ -0,0 +1,2781 @@ +/*************************************************************************/ +/* render_forward_mobile.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "render_forward_mobile.h" +#include "core/config/project_settings.h" +#include "servers/rendering/renderer_rd/storage_rd/light_storage.h" +#include "servers/rendering/renderer_rd/storage_rd/mesh_storage.h" +#include "servers/rendering/renderer_rd/storage_rd/particles_storage.h" +#include "servers/rendering/renderer_rd/storage_rd/texture_storage.h" +#include "servers/rendering/rendering_device.h" +#include "servers/rendering/rendering_server_default.h" + +using namespace RendererSceneRenderImplementation; + +RendererRD::ForwardID RenderForwardMobile::ForwardIDStorageMobile::allocate_forward_id(RendererRD::ForwardIDType p_type) { + int32_t index = -1; + for (uint32_t i = 0; i < forward_id_allocators[p_type].allocations.size(); i++) { + if (forward_id_allocators[p_type].allocations[i] == false) { + index = i; + break; + } + } + + if (index == -1) { + index = forward_id_allocators[p_type].allocations.size(); + forward_id_allocators[p_type].allocations.push_back(true); + forward_id_allocators[p_type].map.push_back(0xFF); + } else { + forward_id_allocators[p_type].allocations[index] = true; + } + + return index; +} +void RenderForwardMobile::ForwardIDStorageMobile::free_forward_id(RendererRD::ForwardIDType p_type, RendererRD::ForwardID p_id) { + ERR_FAIL_INDEX(p_id, (RendererRD::ForwardID)forward_id_allocators[p_type].allocations.size()); + forward_id_allocators[p_type].allocations[p_id] = false; +} + +void RenderForwardMobile::ForwardIDStorageMobile::map_forward_id(RendererRD::ForwardIDType p_type, RendererRD::ForwardID p_id, uint32_t p_index) { + forward_id_allocators[p_type].map[p_id] = p_index; +} + +void RenderForwardMobile::ForwardIDStorageMobile::fill_push_constant_instance_indices(GeometryInstanceForwardMobile::PushConstant *p_push_constant, uint32_t &spec_constants, const GeometryInstanceForwardMobile *p_instance) { + // first zero out our indices + + p_push_constant->omni_lights[0] = 0xFFFF; + p_push_constant->omni_lights[1] = 0xFFFF; + + p_push_constant->spot_lights[0] = 0xFFFF; + p_push_constant->spot_lights[1] = 0xFFFF; + + p_push_constant->decals[0] = 0xFFFF; + p_push_constant->decals[1] = 0xFFFF; + + p_push_constant->reflection_probes[0] = 0xFFFF; + p_push_constant->reflection_probes[1] = 0xFFFF; + + if (p_instance->omni_light_count == 0) { + spec_constants |= 1 << SPEC_CONSTANT_DISABLE_OMNI_LIGHTS; + } + if (p_instance->spot_light_count == 0) { + spec_constants |= 1 << SPEC_CONSTANT_DISABLE_SPOT_LIGHTS; + } + if (p_instance->reflection_probe_count == 0) { + spec_constants |= 1 << SPEC_CONSTANT_DISABLE_REFLECTION_PROBES; + } + if (p_instance->decals_count == 0) { + spec_constants |= 1 << SPEC_CONSTANT_DISABLE_DECALS; + } + + for (uint32_t i = 0; i < MAX_RDL_CULL; i++) { + uint32_t ofs = i < 4 ? 0 : 1; + uint32_t shift = (i & 0x3) << 3; + uint32_t mask = ~(0xFF << shift); + if (i < p_instance->omni_light_count) { + p_push_constant->omni_lights[ofs] &= mask; + p_push_constant->omni_lights[ofs] |= uint32_t(forward_id_allocators[RendererRD::FORWARD_ID_TYPE_OMNI_LIGHT].map[p_instance->omni_lights[i]]) << shift; + } + if (i < p_instance->spot_light_count) { + p_push_constant->spot_lights[ofs] &= mask; + p_push_constant->spot_lights[ofs] |= uint32_t(forward_id_allocators[RendererRD::FORWARD_ID_TYPE_SPOT_LIGHT].map[p_instance->spot_lights[i]]) << shift; + } + if (i < p_instance->decals_count) { + p_push_constant->decals[ofs] &= mask; + p_push_constant->decals[ofs] |= uint32_t(forward_id_allocators[RendererRD::FORWARD_ID_TYPE_DECAL].map[p_instance->decals[i]]) << shift; + } + if (i < p_instance->reflection_probe_count) { + p_push_constant->reflection_probes[ofs] &= mask; + p_push_constant->reflection_probes[ofs] |= uint32_t(forward_id_allocators[RendererRD::FORWARD_ID_TYPE_REFLECTION_PROBE].map[p_instance->reflection_probes[i]]) << shift; + } + } +} + +/* Render buffer */ + +void RenderForwardMobile::RenderBufferDataForwardMobile::free_data() { + // this should already be done but JIC.. + if (render_buffers) { + render_buffers->clear_context(RB_SCOPE_MOBILE); + } +} + +void RenderForwardMobile::RenderBufferDataForwardMobile::configure(RenderSceneBuffersRD *p_render_buffers) { + if (render_buffers) { + // JIC + free_data(); + } + + render_buffers = p_render_buffers; + ERR_FAIL_NULL(render_buffers); // Huh? really? + + RS::ViewportMSAA msaa_3d = render_buffers->get_msaa_3d(); + if (msaa_3d != RS::VIEWPORT_MSAA_DISABLED) { + // Create our MSAA textures... + + RD::DataFormat format = render_buffers->get_base_data_format(); + uint32_t usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; + + const RD::TextureSamples ts[RS::VIEWPORT_MSAA_MAX] = { + RD::TEXTURE_SAMPLES_1, + RD::TEXTURE_SAMPLES_2, + RD::TEXTURE_SAMPLES_4, + RD::TEXTURE_SAMPLES_8, + }; + + texture_samples = ts[msaa_3d]; + + render_buffers->create_texture(RB_SCOPE_MOBILE, RB_TEX_COLOR_MSAA, format, usage_bits, texture_samples); + + format = RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_D24_UNORM_S8_UINT, RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) ? RD::DATA_FORMAT_D24_UNORM_S8_UINT : RD::DATA_FORMAT_D32_SFLOAT_S8_UINT; + usage_bits = RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; + + render_buffers->create_texture(RB_SCOPE_MOBILE, RB_TEX_DEPTH_MSAA, format, usage_bits, texture_samples); + } +} + +RID RenderForwardMobile::RenderBufferDataForwardMobile::get_color_fbs(FramebufferConfigType p_config_type) { + ERR_FAIL_NULL_V(render_buffers, RID()); + + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + ERR_FAIL_NULL_V(texture_storage, RID()); + + // We use our framebuffer cache here instead of building these in RenderBufferDataForwardMobile::configure + // This approach ensures we only build the framebuffers we actually need for this viewport. + // In the (near) future this means that if we cycle through a texture chain for our render target, we'll also support + // this. + + RS::ViewportMSAA msaa_3d = render_buffers->get_msaa_3d(); + bool use_msaa = msaa_3d != RS::VIEWPORT_MSAA_DISABLED; + + uint32_t view_count = render_buffers->get_view_count(); + + RID vrs_texture; + if (render_buffers->has_texture(RB_SCOPE_VRS, RB_TEXTURE)) { + vrs_texture = render_buffers->get_texture(RB_SCOPE_VRS, RB_TEXTURE); + } + + Vector<RID> textures; + int color_buffer_id = 0; + textures.push_back(use_msaa ? get_color_msaa() : render_buffers->get_internal_texture()); // 0 - color buffer + textures.push_back(use_msaa ? get_depth_msaa() : render_buffers->get_depth_texture()); // 1 - depth buffer + if (vrs_texture.is_valid()) { + textures.push_back(vrs_texture); // 2 - vrs texture + } + if (use_msaa) { + color_buffer_id = textures.size(); + textures.push_back(render_buffers->get_internal_texture()); // color buffer for resolve + + // TODO add support for resolving depth buffer!!! + } + + // Now define our subpasses + Vector<RD::FramebufferPass> passes; + + // Define our base pass, we'll be re-using this + RD::FramebufferPass pass; + pass.color_attachments.push_back(0); + pass.depth_attachment = 1; + if (vrs_texture.is_valid()) { + pass.vrs_attachment = 2; + } + + switch (p_config_type) { + case FB_CONFIG_ONE_PASS: { + // just one pass + if (use_msaa) { + // Add resolve + pass.resolve_attachments.push_back(color_buffer_id); + } + passes.push_back(pass); + + return FramebufferCacheRD::get_singleton()->get_cache_multipass(textures, passes, view_count); + } break; + case FB_CONFIG_TWO_SUBPASSES: { + // - opaque pass + passes.push_back(pass); + + // - add sky pass + if (use_msaa) { + // add resolve + pass.resolve_attachments.push_back(color_buffer_id); + } + passes.push_back(pass); + + return FramebufferCacheRD::get_singleton()->get_cache_multipass(textures, passes, view_count); + } break; + case FB_CONFIG_THREE_SUBPASSES: { + // - opaque pass + passes.push_back(pass); + + // - add sky pass + passes.push_back(pass); + + // - add alpha pass + if (use_msaa) { + // add resolve + pass.resolve_attachments.push_back(color_buffer_id); + } + passes.push_back(pass); + + return FramebufferCacheRD::get_singleton()->get_cache_multipass(textures, passes, view_count); + } break; + case FB_CONFIG_FOUR_SUBPASSES: { + Size2i target_size = render_buffers->get_target_size(); + Size2i internal_size = render_buffers->get_internal_size(); + + // can't do our blit pass if resolutions don't match + ERR_FAIL_COND_V(target_size != internal_size, RID()); + + // - opaque pass + passes.push_back(pass); + + // - add sky pass + passes.push_back(pass); + + // - add alpha pass + if (use_msaa) { + // add resolve + pass.resolve_attachments.push_back(color_buffer_id); + } + passes.push_back(pass); + + // - add blit to 2D pass + RID render_target = render_buffers->get_render_target(); + ERR_FAIL_COND_V(render_target.is_null(), RID()); + RID target_buffer = texture_storage->render_target_get_rd_texture(render_target); + ERR_FAIL_COND_V(target_buffer.is_null(), RID()); + + int target_buffer_id = textures.size(); + textures.push_back(target_buffer); // target buffer + + RD::FramebufferPass blit_pass; + blit_pass.input_attachments.push_back(color_buffer_id); // Read from our (resolved) color buffer + blit_pass.color_attachments.push_back(target_buffer_id); // Write into our target buffer + // this doesn't need VRS + passes.push_back(blit_pass); + + return FramebufferCacheRD::get_singleton()->get_cache_multipass(textures, passes, view_count); + } break; + default: + break; + }; + + return RID(); +} + +RID RenderForwardMobile::reflection_probe_create_framebuffer(RID p_color, RID p_depth) { + // Our attachments + Vector<RID> fb; + fb.push_back(p_color); // 0 + fb.push_back(p_depth); // 1 + + // Now define our subpasses + Vector<RD::FramebufferPass> passes; + RD::FramebufferPass pass; + + // re-using the same attachments + pass.color_attachments.push_back(0); + pass.depth_attachment = 1; + + // - opaque pass + passes.push_back(pass); + + // - sky pass + passes.push_back(pass); + + // - alpha pass + passes.push_back(pass); + + return RD::get_singleton()->framebuffer_create_multipass(fb, passes); +} + +void RenderForwardMobile::setup_render_buffer_data(Ref<RenderSceneBuffersRD> p_render_buffers) { + Ref<RenderBufferDataForwardMobile> data; + data.instantiate(); + + p_render_buffers->set_custom_data(RB_SCOPE_MOBILE, data); +} + +bool RenderForwardMobile::free(RID p_rid) { + if (RendererSceneRenderRD::free(p_rid)) { + return true; + } + return false; +} + +/* Render functions */ + +float RenderForwardMobile::_render_buffers_get_luminance_multiplier() { + // On mobile renderer we need to multiply source colors by 2 due to using a UNORM buffer + // and multiplying by the output color during 3D rendering by 0.5 + return 2.0; +} + +RD::DataFormat RenderForwardMobile::_render_buffers_get_color_format() { + // Using 32bit buffers enables AFBC on mobile devices which should have a definite performance improvement (MALI G710 and newer support this on 64bit RTs) + return RD::DATA_FORMAT_A2B10G10R10_UNORM_PACK32; +} + +bool RenderForwardMobile::_render_buffers_can_be_storage() { + // Using 32bit buffers enables AFBC on mobile devices which should have a definite performance improvement (MALI G710 and newer support this on 64bit RTs) + // Doesn't support storage + return false; +} + +RID RenderForwardMobile::_setup_render_pass_uniform_set(RenderListType p_render_list, const RenderDataRD *p_render_data, RID p_radiance_texture, bool p_use_directional_shadow_atlas, int p_index) { + RendererRD::LightStorage *light_storage = RendererRD::LightStorage::get_singleton(); + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + + //there should always be enough uniform buffers for render passes, otherwise bugs + ERR_FAIL_INDEX_V(p_index, (int)scene_state.uniform_buffers.size(), RID()); + + Ref<RenderBufferDataForwardMobile> rb_data; + Ref<RenderSceneBuffersRD> rb; + if (p_render_data && p_render_data->render_buffers.is_valid()) { + rb = p_render_data->render_buffers; + rb_data = rb->get_custom_data(RB_SCOPE_MOBILE); + } + + // default render buffer and scene state uniform set + // loaded into set 1 + + Vector<RD::Uniform> uniforms; + + { + RD::Uniform u; + u.binding = 0; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.append_id(scene_state.uniform_buffers[p_index]); + uniforms.push_back(u); + } + + { + RID radiance_texture; + if (p_radiance_texture.is_valid()) { + radiance_texture = p_radiance_texture; + } else { + radiance_texture = texture_storage->texture_rd_get_default(is_using_radiance_cubemap_array() ? RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_BLACK : RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK); + } + RD::Uniform u; + u.binding = 2; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.append_id(radiance_texture); + uniforms.push_back(u); + } + + { + RID ref_texture = (p_render_data && p_render_data->reflection_atlas.is_valid()) ? light_storage->reflection_atlas_get_texture(p_render_data->reflection_atlas) : RID(); + RD::Uniform u; + u.binding = 3; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + if (ref_texture.is_valid()) { + u.append_id(ref_texture); + } else { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_BLACK)); + } + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 4; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID texture; + if (p_render_data && p_render_data->shadow_atlas.is_valid()) { + texture = light_storage->shadow_atlas_get_texture(p_render_data->shadow_atlas); + } + if (!texture.is_valid()) { + texture = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_DEPTH); + } + u.append_id(texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 5; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + if (p_use_directional_shadow_atlas && light_storage->directional_shadow_get_texture().is_valid()) { + u.append_id(light_storage->directional_shadow_get_texture()); + } else { + u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_DEPTH)); + } + uniforms.push_back(u); + } + + /* we have limited ability to keep textures like this so we're moving this to a set we change before drawing geometry and just pushing the needed texture in */ + { + RD::Uniform u; + u.binding = 6; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + + RID default_tex = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE); + for (uint32_t i = 0; i < scene_state.max_lightmaps; i++) { + if (p_render_data && i < p_render_data->lightmaps->size()) { + RID base = light_storage->lightmap_instance_get_lightmap((*p_render_data->lightmaps)[i]); + RID texture = light_storage->lightmap_get_texture(base); + RID rd_texture = texture_storage->texture_get_rd_texture(texture); + u.append_id(rd_texture); + } else { + u.append_id(default_tex); + } + } + + uniforms.push_back(u); + } + + /* + { + RD::Uniform u; + u.binding = 7; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.ids.resize(MAX_VOXEL_GI_INSTANCESS); + RID default_tex = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE); + for (int i = 0; i < MAX_VOXEL_GI_INSTANCESS; i++) { + if (i < (int)p_voxel_gi_instances.size()) { + RID tex = gi.voxel_gi_instance_get_texture(p_voxel_gi_instances[i]); + if (!tex.is_valid()) { + tex = default_tex; + } + u.ids.write[i] = tex; + } else { + u.ids.write[i] = default_tex; + } + } + + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 8; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + RID cb = p_cluster_buffer.is_valid() ? p_cluster_buffer : default_vec4_xform_buffer; + u.append_id(cb); + uniforms.push_back(u); + } + */ + + { + RD::Uniform u; + u.binding = 9; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID texture; + if (rb.is_valid() && rb->has_texture(RB_SCOPE_BUFFERS, RB_TEX_BACK_DEPTH)) { + texture = rb->get_texture(RB_SCOPE_BUFFERS, RB_TEX_BACK_DEPTH); + } else { + texture = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_DEPTH); + } + u.append_id(texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 10; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID bbt = rb_data.is_valid() ? rb->get_back_buffer_texture() : RID(); + RID texture = bbt.is_valid() ? bbt : texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_BLACK); + u.append_id(texture); + uniforms.push_back(u); + } + + if (p_index >= (int)render_pass_uniform_sets.size()) { + render_pass_uniform_sets.resize(p_index + 1); + } + + if (render_pass_uniform_sets[p_index].is_valid() && RD::get_singleton()->uniform_set_is_valid(render_pass_uniform_sets[p_index])) { + RD::get_singleton()->free(render_pass_uniform_sets[p_index]); + } + + render_pass_uniform_sets[p_index] = RD::get_singleton()->uniform_set_create(uniforms, scene_shader.default_shader_rd, RENDER_PASS_UNIFORM_SET); + return render_pass_uniform_sets[p_index]; +} + +void RenderForwardMobile::_setup_lightmaps(const RenderDataRD *p_render_data, const PagedArray<RID> &p_lightmaps, const Transform3D &p_cam_transform) { + RendererRD::LightStorage *light_storage = RendererRD::LightStorage::get_singleton(); + + // This probably needs to change... + scene_state.lightmaps_used = 0; + for (int i = 0; i < (int)p_lightmaps.size(); i++) { + if (i >= (int)scene_state.max_lightmaps) { + break; + } + + RID lightmap = light_storage->lightmap_instance_get_lightmap(p_lightmaps[i]); + + Basis to_lm = light_storage->lightmap_instance_get_transform(p_lightmaps[i]).basis.inverse() * p_cam_transform.basis; + to_lm = to_lm.inverse().transposed(); //will transform normals + RendererRD::MaterialStorage::store_transform_3x3(to_lm, scene_state.lightmaps[i].normal_xform); + scene_state.lightmaps[i].exposure_normalization = 1.0; + if (p_render_data->camera_attributes.is_valid()) { + float baked_exposure = light_storage->lightmap_get_baked_exposure_normalization(lightmap); + float enf = RSG::camera_attributes->camera_attributes_get_exposure_normalization_factor(p_render_data->camera_attributes); + scene_state.lightmaps[i].exposure_normalization = enf / baked_exposure; + } + + scene_state.lightmap_ids[i] = p_lightmaps[i]; + scene_state.lightmap_has_sh[i] = light_storage->lightmap_uses_spherical_harmonics(lightmap); + + scene_state.lightmaps_used++; + } + if (scene_state.lightmaps_used > 0) { + RD::get_singleton()->buffer_update(scene_state.lightmap_buffer, 0, sizeof(LightmapData) * scene_state.lightmaps_used, scene_state.lightmaps, RD::BARRIER_MASK_RASTER); + } +} + +void RenderForwardMobile::_pre_opaque_render(RenderDataRD *p_render_data) { + RendererRD::LightStorage *light_storage = RendererRD::LightStorage::get_singleton(); + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + + p_render_data->cube_shadows.clear(); + p_render_data->shadows.clear(); + p_render_data->directional_shadows.clear(); + + Plane camera_plane(-p_render_data->scene_data->cam_transform.basis.get_column(Vector3::AXIS_Z), p_render_data->scene_data->cam_transform.origin); + float lod_distance_multiplier = p_render_data->scene_data->cam_projection.get_lod_multiplier(); + { + for (int i = 0; i < p_render_data->render_shadow_count; i++) { + RID li = p_render_data->render_shadows[i].light; + RID base = light_storage->light_instance_get_base_light(li); + + if (light_storage->light_get_type(base) == RS::LIGHT_DIRECTIONAL) { + p_render_data->directional_shadows.push_back(i); + } else if (light_storage->light_get_type(base) == RS::LIGHT_OMNI && light_storage->light_omni_get_shadow_mode(base) == RS::LIGHT_OMNI_SHADOW_CUBE) { + p_render_data->cube_shadows.push_back(i); + } else { + p_render_data->shadows.push_back(i); + } + } + + //cube shadows are rendered in their own way + for (uint32_t i = 0; i < p_render_data->cube_shadows.size(); i++) { + _render_shadow_pass(p_render_data->render_shadows[p_render_data->cube_shadows[i]].light, p_render_data->shadow_atlas, p_render_data->render_shadows[p_render_data->cube_shadows[i]].pass, p_render_data->render_shadows[p_render_data->cube_shadows[i]].instances, camera_plane, lod_distance_multiplier, p_render_data->scene_data->screen_mesh_lod_threshold, true, true, true, p_render_data->render_info); + } + + if (p_render_data->directional_shadows.size()) { + //open the pass for directional shadows + light_storage->update_directional_shadow_atlas(); + RD::get_singleton()->draw_list_begin(light_storage->direction_shadow_get_fb(), RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_CONTINUE); + RD::get_singleton()->draw_list_end(); + } + } + + bool render_shadows = p_render_data->directional_shadows.size() || p_render_data->shadows.size(); + + if (render_shadows) { + RENDER_TIMESTAMP("Render Shadows"); + } + + //prepare shadow rendering + if (render_shadows) { + _render_shadow_begin(); + + //render directional shadows + for (uint32_t i = 0; i < p_render_data->directional_shadows.size(); i++) { + _render_shadow_pass(p_render_data->render_shadows[p_render_data->directional_shadows[i]].light, p_render_data->shadow_atlas, p_render_data->render_shadows[p_render_data->directional_shadows[i]].pass, p_render_data->render_shadows[p_render_data->directional_shadows[i]].instances, camera_plane, lod_distance_multiplier, p_render_data->scene_data->screen_mesh_lod_threshold, false, i == p_render_data->directional_shadows.size() - 1, false, p_render_data->render_info); + } + //render positional shadows + for (uint32_t i = 0; i < p_render_data->shadows.size(); i++) { + _render_shadow_pass(p_render_data->render_shadows[p_render_data->shadows[i]].light, p_render_data->shadow_atlas, p_render_data->render_shadows[p_render_data->shadows[i]].pass, p_render_data->render_shadows[p_render_data->shadows[i]].instances, camera_plane, lod_distance_multiplier, p_render_data->scene_data->screen_mesh_lod_threshold, i == 0, i == p_render_data->shadows.size() - 1, true, p_render_data->render_info); + } + + _render_shadow_process(); + + _render_shadow_end(RD::BARRIER_MASK_NO_BARRIER); + } + + //full barrier here, we need raster, transfer and compute and it depends from the previous work + RD::get_singleton()->barrier(RD::BARRIER_MASK_ALL, RD::BARRIER_MASK_ALL); + + bool using_shadows = true; + + if (p_render_data->reflection_probe.is_valid()) { + if (!RSG::light_storage->reflection_probe_renders_shadows(light_storage->reflection_probe_instance_get_probe(p_render_data->reflection_probe))) { + using_shadows = false; + } + } else { + //do not render reflections when rendering a reflection probe + light_storage->update_reflection_probe_buffer(p_render_data, *p_render_data->reflection_probes, p_render_data->scene_data->cam_transform.affine_inverse(), p_render_data->environment); + } + + uint32_t directional_light_count = 0; + uint32_t positional_light_count = 0; + light_storage->update_light_buffers(p_render_data, *p_render_data->lights, p_render_data->scene_data->cam_transform, p_render_data->shadow_atlas, using_shadows, directional_light_count, positional_light_count, p_render_data->directional_light_soft_shadows); + texture_storage->update_decal_buffer(*p_render_data->decals, p_render_data->scene_data->cam_transform.affine_inverse()); + + p_render_data->directional_light_count = directional_light_count; +} + +void RenderForwardMobile::_render_scene(RenderDataRD *p_render_data, const Color &p_default_bg_color) { + RendererRD::LightStorage *light_storage = RendererRD::LightStorage::get_singleton(); + + Ref<RenderSceneBuffersRD> rb; + Ref<RenderBufferDataForwardMobile> rb_data; + if (p_render_data->render_buffers.is_valid()) { + rb = p_render_data->render_buffers; + rb_data = rb->get_custom_data(RB_SCOPE_MOBILE); + } + + RENDER_TIMESTAMP("Prepare 3D Scene"); + + _update_vrs(rb); + + RENDER_TIMESTAMP("Setup 3D Scene"); + + /* TODO + // check if we need motion vectors + if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_MOTION_VECTORS) { + p_render_data->scene_data->calculate_motion_vectors = true; + } else if (render target has velocity override) { // TODO + p_render_data->scene_data->calculate_motion_vectors = true; + } else { + p_render_data->scene_data->calculate_motion_vectors = false; + } + */ + p_render_data->scene_data->calculate_motion_vectors = false; // for now, not yet supported... + + p_render_data->scene_data->directional_light_count = 0; + p_render_data->scene_data->opaque_prepass_threshold = 0.0; + + // We can only use our full subpass approach if we're: + // - not reading from SCREEN_TEXTURE/DEPTH_TEXTURE + // - not using ssr/sss (currently not supported) + // - not using glow or other post effects (can't do 4th subpass) + // - rendering to a half sized render buffer (can't do 4th subpass) + // We'll need to restrict how far we're going with subpasses based on this. + + Size2i screen_size; + RID framebuffer; + bool reverse_cull = false; + bool using_subpass_transparent = true; + bool using_subpass_post_process = true; + + // fill our render lists early so we can find out if we use various features + _fill_render_list(RENDER_LIST_OPAQUE, p_render_data, PASS_MODE_COLOR); + render_list[RENDER_LIST_OPAQUE].sort_by_key(); + render_list[RENDER_LIST_ALPHA].sort_by_reverse_depth_and_priority(); + _fill_element_info(RENDER_LIST_OPAQUE); + _fill_element_info(RENDER_LIST_ALPHA); + + if (p_render_data->render_info) { + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_VISIBLE][RS::VIEWPORT_RENDER_INFO_DRAW_CALLS_IN_FRAME] = p_render_data->instances->size(); + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_VISIBLE][RS::VIEWPORT_RENDER_INFO_OBJECTS_IN_FRAME] = p_render_data->instances->size(); + } + + if (rb_data.is_valid()) { + // setup rendering to render buffer + screen_size = p_render_data->render_buffers->get_internal_size(); + + if (rb_data->get_color_fbs(RenderBufferDataForwardMobile::FB_CONFIG_FOUR_SUBPASSES).is_null()) { + // can't do blit subpass + using_subpass_post_process = false; + } else if (p_render_data->environment.is_valid() && (environment_get_glow_enabled(p_render_data->environment) || RSG::camera_attributes->camera_attributes_uses_auto_exposure(p_render_data->camera_attributes) || RSG::camera_attributes->camera_attributes_uses_dof(p_render_data->camera_attributes))) { + // can't do blit subpass + using_subpass_post_process = false; + } + + if (scene_state.used_screen_texture || scene_state.used_depth_texture) { + // can't use our last two subpasses + using_subpass_transparent = false; + using_subpass_post_process = false; + } + + if (using_subpass_post_process) { + // all as subpasses + framebuffer = rb_data->get_color_fbs(RenderBufferDataForwardMobile::FB_CONFIG_FOUR_SUBPASSES); + } else if (using_subpass_transparent) { + // our tonemap pass is separate + framebuffer = rb_data->get_color_fbs(RenderBufferDataForwardMobile::FB_CONFIG_THREE_SUBPASSES); + } else { + // only opaque and sky as subpasses + framebuffer = rb_data->get_color_fbs(RenderBufferDataForwardMobile::FB_CONFIG_TWO_SUBPASSES); + } + } else if (p_render_data->reflection_probe.is_valid()) { + uint32_t resolution = light_storage->reflection_probe_instance_get_resolution(p_render_data->reflection_probe); + screen_size.x = resolution; + screen_size.y = resolution; + + framebuffer = light_storage->reflection_probe_instance_get_framebuffer(p_render_data->reflection_probe, p_render_data->reflection_probe_pass); + + if (light_storage->reflection_probe_is_interior(light_storage->reflection_probe_instance_get_probe(p_render_data->reflection_probe))) { + p_render_data->environment = RID(); //no environment on interiors + } + + reverse_cull = true; + using_subpass_transparent = true; // we ignore our screen/depth texture here + using_subpass_post_process = false; // not applicable at all for reflection probes. + } else { + ERR_FAIL(); //bug? + } + + p_render_data->scene_data->emissive_exposure_normalization = -1.0; + + RD::get_singleton()->draw_command_begin_label("Render Setup"); + + _setup_lightmaps(p_render_data, *p_render_data->lightmaps, p_render_data->scene_data->cam_transform); + _setup_environment(p_render_data, p_render_data->reflection_probe.is_valid(), screen_size, !p_render_data->reflection_probe.is_valid(), p_default_bg_color, false); + + _update_render_base_uniform_set(); //may have changed due to the above (light buffer enlarged, as an example) + + RD::get_singleton()->draw_command_end_label(); // Render Setup + + // setup environment + RID radiance_texture; + bool draw_sky = false; + bool draw_sky_fog_only = false; + // We invert luminance_multiplier for sky so that we can combine it with exposure value. + float sky_energy_multiplier = 1.0 / _render_buffers_get_luminance_multiplier(); + + Color clear_color = p_default_bg_color; + bool keep_color = false; + + if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_OVERDRAW) { + clear_color = Color(0, 0, 0, 1); //in overdraw mode, BG should always be black + } else if (is_environment(p_render_data->environment)) { + RS::EnvironmentBG bg_mode = environment_get_background(p_render_data->environment); + float bg_energy_multiplier = environment_get_bg_energy_multiplier(p_render_data->environment); + bg_energy_multiplier *= environment_get_bg_intensity(p_render_data->environment); + + if (p_render_data->camera_attributes.is_valid()) { + bg_energy_multiplier *= RSG::camera_attributes->camera_attributes_get_exposure_normalization_factor(p_render_data->camera_attributes); + } + + switch (bg_mode) { + case RS::ENV_BG_CLEAR_COLOR: { + clear_color = p_default_bg_color; + clear_color.r *= bg_energy_multiplier; + clear_color.g *= bg_energy_multiplier; + clear_color.b *= bg_energy_multiplier; + /* + if (p_render_data->render_buffers->has_custom_data(RB_SCOPE_FOG) || environment_get_fog_enabled(p_render_data->environment)) { + draw_sky_fog_only = true; + RendererRD::MaterialStorage::get_singleton()->material_set_param(sky.sky_scene_state.fog_material, "clear_color", Variant(clear_color.srgb_to_linear())); + } + */ + } break; + case RS::ENV_BG_COLOR: { + clear_color = environment_get_bg_color(p_render_data->environment); + clear_color.r *= bg_energy_multiplier; + clear_color.g *= bg_energy_multiplier; + clear_color.b *= bg_energy_multiplier; + /* + if (p_render_data->render_buffers->has_custom_data(RB_SCOPE_FOG) || environment_get_fog_enabled(p_render_data->environment)) { + draw_sky_fog_only = true; + RendererRD::MaterialStorage::get_singleton()->material_set_param(sky.sky_scene_state.fog_material, "clear_color", Variant(clear_color.srgb_to_linear())); + } + */ + } break; + case RS::ENV_BG_SKY: { + draw_sky = true; + } break; + case RS::ENV_BG_CANVAS: { + keep_color = true; + } break; + case RS::ENV_BG_KEEP: { + keep_color = true; + } break; + case RS::ENV_BG_CAMERA_FEED: { + } break; + default: { + } + } + // setup sky if used for ambient, reflections, or background + if (draw_sky || draw_sky_fog_only || environment_get_reflection_source(p_render_data->environment) == RS::ENV_REFLECTION_SOURCE_SKY || environment_get_ambient_source(p_render_data->environment) == RS::ENV_AMBIENT_SOURCE_SKY) { + RENDER_TIMESTAMP("Setup Sky"); + RD::get_singleton()->draw_command_begin_label("Setup Sky"); + Projection projection = p_render_data->scene_data->cam_projection; + if (p_render_data->reflection_probe.is_valid()) { + Projection correction; + correction.set_depth_correction(true); + projection = correction * p_render_data->scene_data->cam_projection; + } + + sky.setup(p_render_data->environment, p_render_data->render_buffers, *p_render_data->lights, p_render_data->camera_attributes, projection, p_render_data->scene_data->cam_transform, screen_size, this); + + sky_energy_multiplier *= bg_energy_multiplier; + + RID sky_rid = environment_get_sky(p_render_data->environment); + if (sky_rid.is_valid()) { + sky.update(p_render_data->environment, projection, p_render_data->scene_data->cam_transform, time, sky_energy_multiplier); + radiance_texture = sky.sky_get_radiance_texture_rd(sky_rid); + } else { + // do not try to draw sky if invalid + draw_sky = false; + } + RD::get_singleton()->draw_command_end_label(); // Setup Sky + } + } else { + clear_color = p_default_bg_color; + } + + // update sky buffers (if required) + if (draw_sky || draw_sky_fog_only) { + // !BAS! @TODO See if we can limit doing some things double and maybe even move this into _pre_opaque_render + // and change Forward Clustered in the same way as we have here (but without using subpasses) + RENDER_TIMESTAMP("Setup Sky Resolution Buffers"); + + RD::get_singleton()->draw_command_begin_label("Setup Sky Resolution Buffers"); + + if (p_render_data->reflection_probe.is_valid()) { + Projection correction; + correction.set_depth_correction(true); + Projection projection = correction * p_render_data->scene_data->cam_projection; + sky.update_res_buffers(p_render_data->environment, 1, &projection, p_render_data->scene_data->cam_transform, time, sky_energy_multiplier); + } else { + sky.update_res_buffers(p_render_data->environment, p_render_data->scene_data->view_count, p_render_data->scene_data->view_projection, p_render_data->scene_data->cam_transform, time, sky_energy_multiplier); + } + + RD::get_singleton()->draw_command_end_label(); // Setup Sky resolution buffers + } + + _pre_opaque_render(p_render_data); + + uint32_t spec_constant_base_flags = 0; + + { + //figure out spec constants + + if (p_render_data->directional_light_count > 0) { + if (p_render_data->directional_light_soft_shadows) { + spec_constant_base_flags |= 1 << SPEC_CONSTANT_USING_DIRECTIONAL_SOFT_SHADOWS; + } + } else { + spec_constant_base_flags |= 1 << SPEC_CONSTANT_DISABLE_DIRECTIONAL_LIGHTS; + } + + if (!is_environment(p_render_data->environment) || !environment_get_fog_enabled(p_render_data->environment)) { + spec_constant_base_flags |= 1 << SPEC_CONSTANT_DISABLE_FOG; + } + } + { + if (rb_data.is_valid()) { + RD::get_singleton()->draw_command_begin_label("Render 3D Pass"); + } else { + RD::get_singleton()->draw_command_begin_label("Render Reflection Probe Pass"); + } + + // opaque pass + + RD::get_singleton()->draw_command_begin_label("Render Opaque Subpass"); + + p_render_data->scene_data->directional_light_count = p_render_data->directional_light_count; + + _setup_environment(p_render_data, p_render_data->reflection_probe.is_valid(), screen_size, !p_render_data->reflection_probe.is_valid(), p_default_bg_color, p_render_data->render_buffers.is_valid()); + + if (using_subpass_transparent && using_subpass_post_process) { + RENDER_TIMESTAMP("Render Opaque + Transparent + Tonemap"); + } else if (using_subpass_transparent) { + RENDER_TIMESTAMP("Render Opaque + Transparent"); + } else { + RENDER_TIMESTAMP("Render Opaque"); + } + + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_OPAQUE, p_render_data, radiance_texture, true); + + bool can_continue_color = !using_subpass_transparent && !scene_state.used_screen_texture; + bool can_continue_depth = !using_subpass_transparent && !scene_state.used_depth_texture; + + { + // regular forward for now + Vector<Color> c; + c.push_back(clear_color.srgb_to_linear()); // our render buffer + if (rb_data.is_valid()) { + if (p_render_data->render_buffers->get_msaa_3d() != RS::VIEWPORT_MSAA_DISABLED) { + c.push_back(clear_color.srgb_to_linear()); // our resolve buffer + } + if (using_subpass_post_process) { + c.push_back(Color()); // our 2D buffer we're copying into + } + } + + RD::FramebufferFormatID fb_format = RD::get_singleton()->framebuffer_get_format(framebuffer); + RenderListParameters render_list_params(render_list[RENDER_LIST_OPAQUE].elements.ptr(), render_list[RENDER_LIST_OPAQUE].element_info.ptr(), render_list[RENDER_LIST_OPAQUE].elements.size(), reverse_cull, PASS_MODE_COLOR, rp_uniform_set, spec_constant_base_flags, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->scene_data->lod_distance_multiplier, p_render_data->scene_data->screen_mesh_lod_threshold, p_render_data->scene_data->view_count); + render_list_params.framebuffer_format = fb_format; + if ((uint32_t)render_list_params.element_count > render_list_thread_threshold && false) { + // secondary command buffers need more testing at this time + //multi threaded + thread_draw_lists.resize(WorkerThreadPool::get_singleton()->get_thread_count()); + RD::get_singleton()->draw_list_begin_split(framebuffer, thread_draw_lists.size(), thread_draw_lists.ptr(), keep_color ? RD::INITIAL_ACTION_KEEP : RD::INITIAL_ACTION_CLEAR, can_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, can_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, c, 1.0, 0); + + WorkerThreadPool::GroupID group_task = WorkerThreadPool::get_singleton()->add_template_group_task(this, &RenderForwardMobile::_render_list_thread_function, &render_list_params, thread_draw_lists.size(), -1, true, SNAME("ForwardMobileRenderList")); + WorkerThreadPool::get_singleton()->wait_for_group_task_completion(group_task); + + } else { + //single threaded + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(framebuffer, keep_color ? RD::INITIAL_ACTION_KEEP : RD::INITIAL_ACTION_CLEAR, can_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, can_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, c, 1.0, 0); + _render_list(draw_list, fb_format, &render_list_params, 0, render_list_params.element_count); + } + } + + RD::get_singleton()->draw_command_end_label(); //Render Opaque Subpass + + if (draw_sky || draw_sky_fog_only) { + RD::get_singleton()->draw_command_begin_label("Draw Sky Subpass"); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_switch_to_next_pass(); + + if (p_render_data->reflection_probe.is_valid()) { + Projection correction; + correction.set_depth_correction(true); + Projection projection = correction * p_render_data->scene_data->cam_projection; + sky.draw(draw_list, p_render_data->environment, framebuffer, 1, &projection, p_render_data->scene_data->cam_transform, time, sky_energy_multiplier); + } else { + sky.draw(draw_list, p_render_data->environment, framebuffer, p_render_data->scene_data->view_count, p_render_data->scene_data->view_projection, p_render_data->scene_data->cam_transform, time, sky_energy_multiplier); + } + + RD::get_singleton()->draw_command_end_label(); // Draw Sky Subpass + + // note, if MSAA is used in 2-subpass approach we should get an automatic resolve here + } else { + // switch to subpass but we do nothing here so basically we skip (though this should trigger resolve with 2-subpass MSAA). + RD::get_singleton()->draw_list_switch_to_next_pass(); + } + + if (!using_subpass_transparent) { + // We're done with our subpasses so end our container pass + RD::get_singleton()->draw_list_end(RD::BARRIER_MASK_ALL); + + RD::get_singleton()->draw_command_end_label(); // Render 3D Pass / Render Reflection Probe Pass + } + + if (scene_state.used_screen_texture) { + // Copy screen texture to backbuffer so we can read from it + _render_buffers_copy_screen_texture(p_render_data); + } + + if (scene_state.used_depth_texture) { + // Copy depth texture to backbuffer so we can read from it + _render_buffers_copy_depth_texture(p_render_data); + } + + // transparent pass + + RD::get_singleton()->draw_command_begin_label("Render Transparent Subpass"); + + rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_ALPHA, p_render_data, radiance_texture, true); + + if (using_subpass_transparent) { + RD::FramebufferFormatID fb_format = RD::get_singleton()->framebuffer_get_format(framebuffer); + RenderListParameters render_list_params(render_list[RENDER_LIST_ALPHA].elements.ptr(), render_list[RENDER_LIST_ALPHA].element_info.ptr(), render_list[RENDER_LIST_ALPHA].elements.size(), reverse_cull, PASS_MODE_COLOR_TRANSPARENT, rp_uniform_set, spec_constant_base_flags, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->scene_data->lod_distance_multiplier, p_render_data->scene_data->screen_mesh_lod_threshold, p_render_data->scene_data->view_count); + render_list_params.framebuffer_format = fb_format; + if ((uint32_t)render_list_params.element_count > render_list_thread_threshold && false) { + // secondary command buffers need more testing at this time + //multi threaded + thread_draw_lists.resize(WorkerThreadPool::get_singleton()->get_thread_count()); + RD::get_singleton()->draw_list_switch_to_next_pass_split(thread_draw_lists.size(), thread_draw_lists.ptr()); + render_list_params.subpass = RD::get_singleton()->draw_list_get_current_pass(); + WorkerThreadPool::GroupID group_task = WorkerThreadPool::get_singleton()->add_template_group_task(this, &RenderForwardMobile::_render_list_thread_function, &render_list_params, thread_draw_lists.size(), -1, true, SNAME("ForwardMobileRenderSubpass")); + WorkerThreadPool::get_singleton()->wait_for_group_task_completion(group_task); + + } else { + //single threaded + RD::DrawListID draw_list = RD::get_singleton()->draw_list_switch_to_next_pass(); + render_list_params.subpass = RD::get_singleton()->draw_list_get_current_pass(); + _render_list(draw_list, fb_format, &render_list_params, 0, render_list_params.element_count); + } + + RD::get_singleton()->draw_command_end_label(); // Render Transparent Subpass + + // note if we are using MSAA we should get an automatic resolve through our subpass configuration. + + // blit to tonemap + if (rb_data.is_valid() && using_subpass_post_process) { + _post_process_subpass(p_render_data->render_buffers->get_internal_texture(), framebuffer, p_render_data); + } + + RD::get_singleton()->draw_command_end_label(); // Render 3D Pass / Render Reflection Probe Pass + + RD::get_singleton()->draw_list_end(RD::BARRIER_MASK_ALL); + } else { + RENDER_TIMESTAMP("Render Transparent"); + + framebuffer = rb_data->get_color_fbs(RenderBufferDataForwardMobile::FB_CONFIG_ONE_PASS); + + // this may be needed if we re-introduced steps that change info, not sure which do so in the previous implementation + // _setup_environment(p_render_data, p_render_data->reflection_probe.is_valid(), screen_size, !p_render_data->reflection_probe.is_valid(), p_default_bg_color, false); + + RD::FramebufferFormatID fb_format = RD::get_singleton()->framebuffer_get_format(framebuffer); + RenderListParameters render_list_params(render_list[RENDER_LIST_ALPHA].elements.ptr(), render_list[RENDER_LIST_ALPHA].element_info.ptr(), render_list[RENDER_LIST_ALPHA].elements.size(), reverse_cull, PASS_MODE_COLOR, rp_uniform_set, spec_constant_base_flags, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->scene_data->lod_distance_multiplier, p_render_data->scene_data->screen_mesh_lod_threshold, p_render_data->scene_data->view_count); + render_list_params.framebuffer_format = fb_format; + if ((uint32_t)render_list_params.element_count > render_list_thread_threshold && false) { + // secondary command buffers need more testing at this time + //multi threaded + thread_draw_lists.resize(WorkerThreadPool::get_singleton()->get_thread_count()); + RD::get_singleton()->draw_list_begin_split(framebuffer, thread_draw_lists.size(), thread_draw_lists.ptr(), can_continue_color ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, can_continue_depth ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ); + WorkerThreadPool::GroupID group_task = WorkerThreadPool::get_singleton()->add_template_group_task(this, &RenderForwardMobile::_render_list_thread_function, &render_list_params, thread_draw_lists.size(), -1, true, SNAME("ForwardMobileRenderSubpass")); + WorkerThreadPool::get_singleton()->wait_for_group_task_completion(group_task); + + RD::get_singleton()->draw_list_end(RD::BARRIER_MASK_ALL); + } else { + //single threaded + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(framebuffer, can_continue_color ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, can_continue_depth ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ); + _render_list(draw_list, fb_format, &render_list_params, 0, render_list_params.element_count); + RD::get_singleton()->draw_list_end(RD::BARRIER_MASK_ALL); + } + + RD::get_singleton()->draw_command_end_label(); // Render Transparent Subpass + } + } + + if (rb_data.is_valid() && !using_subpass_post_process) { + RD::get_singleton()->draw_command_begin_label("Post process pass"); + + // If we need extra effects we do this in its own pass + RENDER_TIMESTAMP("Tonemap"); + + _render_buffers_post_process_and_tonemap(p_render_data); + + RD::get_singleton()->draw_command_end_label(); // Post process pass + } + + if (rb_data.is_valid()) { + _disable_clear_request(p_render_data); + } + + if (rb.is_valid()) { + _render_buffers_debug_draw(rb, p_render_data->shadow_atlas, p_render_data->occluder_debug_tex); + } +} + +/* these are being called from RendererSceneRenderRD::_pre_opaque_render */ + +void RenderForwardMobile::_render_shadow_pass(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<RenderGeometryInstance *> &p_instances, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_mesh_lod_threshold, bool p_open_pass, bool p_close_pass, bool p_clear_region, RenderingMethod::RenderInfo *p_render_info) { + RendererRD::LightStorage *light_storage = RendererRD::LightStorage::get_singleton(); + + ERR_FAIL_COND(!light_storage->owns_light_instance(p_light)); + + RID base = light_storage->light_instance_get_base_light(p_light); + + Rect2i atlas_rect; + uint32_t atlas_size = 1; + RID atlas_fb; + + bool using_dual_paraboloid = false; + bool using_dual_paraboloid_flip = false; + Vector2i dual_paraboloid_offset; + RID render_fb; + RID render_texture; + float zfar; + + bool use_pancake = false; + bool render_cubemap = false; + bool finalize_cubemap = false; + + bool flip_y = false; + + Projection light_projection; + Transform3D light_transform; + + if (light_storage->light_get_type(base) == RS::LIGHT_DIRECTIONAL) { + //set pssm stuff + uint64_t last_scene_shadow_pass = light_storage->light_instance_get_shadow_pass(p_light); + if (last_scene_shadow_pass != get_scene_pass()) { + light_storage->light_instance_set_directional_rect(p_light, light_storage->get_directional_shadow_rect()); + light_storage->directional_shadow_increase_current_light(); + light_storage->light_instance_set_shadow_pass(p_light, get_scene_pass()); + } + + use_pancake = light_storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_PANCAKE_SIZE) > 0; + light_projection = light_storage->light_instance_get_shadow_camera(p_light, p_pass); + light_transform = light_storage->light_instance_get_shadow_transform(p_light, p_pass); + + atlas_rect = light_storage->light_instance_get_directional_rect(p_light); + + if (light_storage->light_directional_get_shadow_mode(base) == RS::LIGHT_DIRECTIONAL_SHADOW_PARALLEL_4_SPLITS) { + atlas_rect.size.width /= 2; + atlas_rect.size.height /= 2; + + if (p_pass == 1) { + atlas_rect.position.x += atlas_rect.size.width; + } else if (p_pass == 2) { + atlas_rect.position.y += atlas_rect.size.height; + } else if (p_pass == 3) { + atlas_rect.position += atlas_rect.size; + } + } else if (light_storage->light_directional_get_shadow_mode(base) == RS::LIGHT_DIRECTIONAL_SHADOW_PARALLEL_2_SPLITS) { + atlas_rect.size.height /= 2; + + if (p_pass == 0) { + } else { + atlas_rect.position.y += atlas_rect.size.height; + } + } + + float directional_shadow_size = light_storage->directional_shadow_get_size(); + Rect2 atlas_rect_norm = atlas_rect; + atlas_rect_norm.position /= directional_shadow_size; + atlas_rect_norm.size /= directional_shadow_size; + light_storage->light_instance_set_directional_shadow_atlas_rect(p_light, p_pass, atlas_rect_norm); + + zfar = RSG::light_storage->light_get_param(base, RS::LIGHT_PARAM_RANGE); + + render_fb = light_storage->direction_shadow_get_fb(); + render_texture = RID(); + flip_y = true; + + } else { + //set from shadow atlas + + ERR_FAIL_COND(!light_storage->owns_shadow_atlas(p_shadow_atlas)); + ERR_FAIL_COND(!light_storage->shadow_atlas_owns_light_instance(p_shadow_atlas, p_light)); + + RSG::light_storage->shadow_atlas_update(p_shadow_atlas); + + uint32_t key = light_storage->shadow_atlas_get_light_instance_key(p_shadow_atlas, p_light); + + uint32_t quadrant = (key >> RendererRD::LightStorage::QUADRANT_SHIFT) & 0x3; + uint32_t shadow = key & RendererRD::LightStorage::SHADOW_INDEX_MASK; + uint32_t subdivision = light_storage->shadow_atlas_get_quadrant_subdivision(p_shadow_atlas, quadrant); + + ERR_FAIL_INDEX((int)shadow, light_storage->shadow_atlas_get_quadrant_shadow_size(p_shadow_atlas, quadrant)); + + uint32_t shadow_atlas_size = light_storage->shadow_atlas_get_size(p_shadow_atlas); + uint32_t quadrant_size = shadow_atlas_size >> 1; + + atlas_rect.position.x = (quadrant & 1) * quadrant_size; + atlas_rect.position.y = (quadrant >> 1) * quadrant_size; + + uint32_t shadow_size = (quadrant_size / subdivision); + atlas_rect.position.x += (shadow % subdivision) * shadow_size; + atlas_rect.position.y += (shadow / subdivision) * shadow_size; + + atlas_rect.size.width = shadow_size; + atlas_rect.size.height = shadow_size; + + zfar = light_storage->light_get_param(base, RS::LIGHT_PARAM_RANGE); + + if (light_storage->light_get_type(base) == RS::LIGHT_OMNI) { + bool wrap = (shadow + 1) % subdivision == 0; + dual_paraboloid_offset = wrap ? Vector2i(1 - subdivision, 1) : Vector2i(1, 0); + + if (light_storage->light_omni_get_shadow_mode(base) == RS::LIGHT_OMNI_SHADOW_CUBE) { + render_texture = light_storage->get_cubemap(shadow_size / 2); + render_fb = light_storage->get_cubemap_fb(shadow_size / 2, p_pass); + + light_projection = light_storage->light_instance_get_shadow_camera(p_light, p_pass); + light_transform = light_storage->light_instance_get_shadow_transform(p_light, p_pass); + render_cubemap = true; + finalize_cubemap = p_pass == 5; + atlas_fb = light_storage->shadow_atlas_get_fb(p_shadow_atlas); + + atlas_size = shadow_atlas_size; + + if (p_pass == 0) { + _render_shadow_begin(); + } + + } else { + atlas_rect.position.x += 1; + atlas_rect.position.y += 1; + atlas_rect.size.x -= 2; + atlas_rect.size.y -= 2; + + atlas_rect.position += p_pass * atlas_rect.size * dual_paraboloid_offset; + + light_projection = light_storage->light_instance_get_shadow_camera(p_light, 0); + light_transform = light_storage->light_instance_get_shadow_transform(p_light, 0); + + using_dual_paraboloid = true; + using_dual_paraboloid_flip = p_pass == 1; + render_fb = light_storage->shadow_atlas_get_fb(p_shadow_atlas); + flip_y = true; + } + + } else if (light_storage->light_get_type(base) == RS::LIGHT_SPOT) { + light_projection = light_storage->light_instance_get_shadow_camera(p_light, 0); + light_transform = light_storage->light_instance_get_shadow_transform(p_light, 0); + + render_fb = light_storage->shadow_atlas_get_fb(p_shadow_atlas); + + flip_y = true; + } + } + + if (render_cubemap) { + //rendering to cubemap + _render_shadow_append(render_fb, p_instances, light_projection, light_transform, zfar, 0, 0, false, false, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_mesh_lod_threshold, Rect2(), false, true, true, true, p_render_info); + if (finalize_cubemap) { + _render_shadow_process(); + _render_shadow_end(); + //reblit + Rect2 atlas_rect_norm = atlas_rect; + atlas_rect_norm.position /= float(atlas_size); + atlas_rect_norm.size /= float(atlas_size); + copy_effects->copy_cubemap_to_dp(render_texture, atlas_fb, atlas_rect_norm, atlas_rect.size, light_projection.get_z_near(), light_projection.get_z_far(), false); + atlas_rect_norm.position += Vector2(dual_paraboloid_offset) * atlas_rect_norm.size; + copy_effects->copy_cubemap_to_dp(render_texture, atlas_fb, atlas_rect_norm, atlas_rect.size, light_projection.get_z_near(), light_projection.get_z_far(), true); + + //restore transform so it can be properly used + light_storage->light_instance_set_shadow_transform(p_light, Projection(), light_storage->light_instance_get_base_transform(p_light), zfar, 0, 0, 0); + } + + } else { + //render shadow + _render_shadow_append(render_fb, p_instances, light_projection, light_transform, zfar, 0, 0, using_dual_paraboloid, using_dual_paraboloid_flip, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_mesh_lod_threshold, atlas_rect, flip_y, p_clear_region, p_open_pass, p_close_pass, p_render_info); + } +} + +void RenderForwardMobile::_render_shadow_begin() { + scene_state.shadow_passes.clear(); + RD::get_singleton()->draw_command_begin_label("Shadow Setup"); + _update_render_base_uniform_set(); + + render_list[RENDER_LIST_SECONDARY].clear(); +} + +void RenderForwardMobile::_render_shadow_append(RID p_framebuffer, const PagedArray<RenderGeometryInstance *> &p_instances, const Projection &p_projection, const Transform3D &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_mesh_lod_threshold, const Rect2i &p_rect, bool p_flip_y, bool p_clear_region, bool p_begin, bool p_end, RenderingMethod::RenderInfo *p_render_info) { + uint32_t shadow_pass_index = scene_state.shadow_passes.size(); + + SceneState::ShadowPass shadow_pass; + + if (p_render_info) { + p_render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_SHADOW][RS::VIEWPORT_RENDER_INFO_DRAW_CALLS_IN_FRAME] = p_instances.size(); + p_render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_SHADOW][RS::VIEWPORT_RENDER_INFO_OBJECTS_IN_FRAME] = p_instances.size(); + } + + RenderSceneDataRD scene_data; + scene_data.cam_projection = p_projection; + scene_data.cam_transform = p_transform; + scene_data.view_projection[0] = p_projection; + scene_data.z_near = 0.0; + scene_data.z_far = p_zfar; + scene_data.lod_distance_multiplier = p_lod_distance_multiplier; + scene_data.dual_paraboloid_side = p_use_dp_flip ? -1 : 1; + scene_data.opaque_prepass_threshold = 0.1; + + RenderDataRD render_data; + render_data.scene_data = &scene_data; + render_data.instances = &p_instances; + render_data.render_info = p_render_info; + + _setup_environment(&render_data, true, Vector2(1, 1), !p_flip_y, Color(), false, p_use_pancake, shadow_pass_index); + + if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_DISABLE_LOD) { + scene_data.screen_mesh_lod_threshold = 0.0; + } else { + scene_data.screen_mesh_lod_threshold = p_screen_mesh_lod_threshold; + } + + PassMode pass_mode = p_use_dp ? PASS_MODE_SHADOW_DP : PASS_MODE_SHADOW; + + uint32_t render_list_from = render_list[RENDER_LIST_SECONDARY].elements.size(); + _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode, true); + uint32_t render_list_size = render_list[RENDER_LIST_SECONDARY].elements.size() - render_list_from; + render_list[RENDER_LIST_SECONDARY].sort_by_key_range(render_list_from, render_list_size); + _fill_element_info(RENDER_LIST_SECONDARY, render_list_from, render_list_size); + + { + //regular forward for now + bool flip_cull = p_use_dp_flip; + if (p_flip_y) { + flip_cull = !flip_cull; + } + + shadow_pass.element_from = render_list_from; + shadow_pass.element_count = render_list_size; + shadow_pass.flip_cull = flip_cull; + shadow_pass.pass_mode = pass_mode; + + shadow_pass.rp_uniform_set = RID(); //will be filled later when instance buffer is complete + shadow_pass.camera_plane = p_camera_plane; + shadow_pass.screen_mesh_lod_threshold = scene_data.screen_mesh_lod_threshold; + shadow_pass.lod_distance_multiplier = scene_data.lod_distance_multiplier; + + shadow_pass.framebuffer = p_framebuffer; + shadow_pass.initial_depth_action = p_begin ? (p_clear_region ? RD::INITIAL_ACTION_CLEAR_REGION : RD::INITIAL_ACTION_CLEAR) : (p_clear_region ? RD::INITIAL_ACTION_CLEAR_REGION_CONTINUE : RD::INITIAL_ACTION_CONTINUE); + shadow_pass.final_depth_action = p_end ? RD::FINAL_ACTION_READ : RD::FINAL_ACTION_CONTINUE; + shadow_pass.rect = p_rect; + + scene_state.shadow_passes.push_back(shadow_pass); + } +} + +void RenderForwardMobile::_render_shadow_process() { + //render shadows one after the other, so this can be done un-barriered and the driver can optimize (as well as allow us to run compute at the same time) + + for (uint32_t i = 0; i < scene_state.shadow_passes.size(); i++) { + //render passes need to be configured after instance buffer is done, since they need the latest version + SceneState::ShadowPass &shadow_pass = scene_state.shadow_passes[i]; + shadow_pass.rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, nullptr, RID(), false, i); + } + + RD::get_singleton()->draw_command_end_label(); +} + +void RenderForwardMobile::_render_shadow_end(uint32_t p_barrier) { + RD::get_singleton()->draw_command_begin_label("Shadow Render"); + + for (uint32_t i = 0; i < scene_state.shadow_passes.size(); i++) { + SceneState::ShadowPass &shadow_pass = scene_state.shadow_passes[i]; + RenderListParameters render_list_parameters(render_list[RENDER_LIST_SECONDARY].elements.ptr() + shadow_pass.element_from, render_list[RENDER_LIST_SECONDARY].element_info.ptr() + shadow_pass.element_from, shadow_pass.element_count, shadow_pass.flip_cull, shadow_pass.pass_mode, shadow_pass.rp_uniform_set, 0, false, Vector2(), shadow_pass.lod_distance_multiplier, shadow_pass.screen_mesh_lod_threshold, 1, shadow_pass.element_from, RD::BARRIER_MASK_NO_BARRIER); + _render_list_with_threads(&render_list_parameters, shadow_pass.framebuffer, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, shadow_pass.initial_depth_action, shadow_pass.final_depth_action, Vector<Color>(), 1.0, 0, shadow_pass.rect); + } + + if (p_barrier != RD::BARRIER_MASK_NO_BARRIER) { + RD::get_singleton()->barrier(RD::BARRIER_MASK_RASTER, p_barrier); + } + RD::get_singleton()->draw_command_end_label(); +} + +/* */ + +void RenderForwardMobile::_render_material(const Transform3D &p_cam_transform, const Projection &p_cam_projection, bool p_cam_orthogonal, const PagedArray<RenderGeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region, float p_exposure_normalization) { + RENDER_TIMESTAMP("Setup Rendering 3D Material"); + + RD::get_singleton()->draw_command_begin_label("Render 3D Material"); + + _update_render_base_uniform_set(); + + RenderSceneDataRD scene_data; + scene_data.cam_projection = p_cam_projection; + scene_data.cam_transform = p_cam_transform; + scene_data.view_projection[0] = p_cam_projection; + scene_data.dual_paraboloid_side = 0; + scene_data.material_uv2_mode = false; + scene_data.opaque_prepass_threshold = 0.0f; + scene_data.emissive_exposure_normalization = p_exposure_normalization; + + RenderDataRD render_data; + render_data.scene_data = &scene_data; + render_data.instances = &p_instances; + + _setup_environment(&render_data, true, Vector2(1, 1), false, Color()); + + PassMode pass_mode = PASS_MODE_DEPTH_MATERIAL; + _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode); + render_list[RENDER_LIST_SECONDARY].sort_by_key(); + _fill_element_info(RENDER_LIST_SECONDARY); + + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, nullptr, RID()); + + RENDER_TIMESTAMP("Render 3D Material"); + + { + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, rp_uniform_set, 0); + //regular forward for now + Vector<Color> clear = { + Color(0, 0, 0, 0), + Color(0, 0, 0, 0), + Color(0, 0, 0, 0), + Color(0, 0, 0, 0), + Color(0, 0, 0, 0) + }; + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, clear, 1.0, 0, p_region); + _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(p_framebuffer), &render_list_params, 0, render_list_params.element_count); + RD::get_singleton()->draw_list_end(); + } + + RD::get_singleton()->draw_command_end_label(); +} + +void RenderForwardMobile::_render_uv2(const PagedArray<RenderGeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) { + RENDER_TIMESTAMP("Setup Rendering UV2"); + + RD::get_singleton()->draw_command_begin_label("Render UV2"); + + _update_render_base_uniform_set(); + + RenderSceneDataRD scene_data; + scene_data.dual_paraboloid_side = 0; + scene_data.material_uv2_mode = true; + scene_data.emissive_exposure_normalization = -1.0; + + RenderDataRD render_data; + render_data.scene_data = &scene_data; + render_data.instances = &p_instances; + + _setup_environment(&render_data, true, Vector2(1, 1), false, Color()); + + PassMode pass_mode = PASS_MODE_DEPTH_MATERIAL; + _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode); + render_list[RENDER_LIST_SECONDARY].sort_by_key(); + _fill_element_info(RENDER_LIST_SECONDARY); + + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, nullptr, RID()); + + RENDER_TIMESTAMP("Render 3D Material"); + + { + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, rp_uniform_set, true, false); + //regular forward for now + Vector<Color> clear = { + Color(0, 0, 0, 0), + Color(0, 0, 0, 0), + Color(0, 0, 0, 0), + Color(0, 0, 0, 0), + Color(0, 0, 0, 0) + }; + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, clear, 1.0, 0, p_region); + + const int uv_offset_count = 9; + static const Vector2 uv_offsets[uv_offset_count] = { + Vector2(-1, 1), + Vector2(1, 1), + Vector2(1, -1), + Vector2(-1, -1), + Vector2(-1, 0), + Vector2(1, 0), + Vector2(0, -1), + Vector2(0, 1), + Vector2(0, 0), + + }; + + for (int i = 0; i < uv_offset_count; i++) { + Vector2 ofs = uv_offsets[i]; + ofs.x /= p_region.size.width; + ofs.y /= p_region.size.height; + render_list_params.uv_offset = ofs; + _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(p_framebuffer), &render_list_params, 0, render_list_params.element_count); //first wireframe, for pseudo conservative + } + render_list_params.uv_offset = Vector2(); + _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(p_framebuffer), &render_list_params, 0, render_list_params.element_count); //second regular triangles + + RD::get_singleton()->draw_list_end(); + } + + RD::get_singleton()->draw_command_end_label(); +} + +void RenderForwardMobile::_render_sdfgi(Ref<RenderSceneBuffersRD> p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<RenderGeometryInstance *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture, float p_exposure_normalization) { + // we don't do SDFGI in low end.. +} + +void RenderForwardMobile::_render_particle_collider_heightfield(RID p_fb, const Transform3D &p_cam_transform, const Projection &p_cam_projection, const PagedArray<RenderGeometryInstance *> &p_instances) { + RENDER_TIMESTAMP("Setup GPUParticlesCollisionHeightField3D"); + + RD::get_singleton()->draw_command_begin_label("Render Collider Heightfield"); + + _update_render_base_uniform_set(); + + RenderSceneDataRD scene_data; + scene_data.cam_projection = p_cam_projection; + scene_data.cam_transform = p_cam_transform; + scene_data.view_projection[0] = p_cam_projection; + scene_data.z_near = 0.0; + scene_data.z_far = p_cam_projection.get_z_far(); + scene_data.dual_paraboloid_side = 0; + scene_data.opaque_prepass_threshold = 0.0; + + RenderDataRD render_data; + render_data.scene_data = &scene_data; + render_data.instances = &p_instances; + + _setup_environment(&render_data, true, Vector2(1, 1), true, Color(), false, false); + + PassMode pass_mode = PASS_MODE_SHADOW; + + _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode); + render_list[RENDER_LIST_SECONDARY].sort_by_key(); + _fill_element_info(RENDER_LIST_SECONDARY); + + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, nullptr, RID()); + + RENDER_TIMESTAMP("Render Collider Heightfield"); + + { + //regular forward for now + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), false, pass_mode, rp_uniform_set, 0); + _render_list_with_threads(&render_list_params, p_fb, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ); + } + RD::get_singleton()->draw_command_end_label(); +} + +void RenderForwardMobile::base_uniforms_changed() { + if (!render_base_uniform_set.is_null() && RD::get_singleton()->uniform_set_is_valid(render_base_uniform_set)) { + RD::get_singleton()->free(render_base_uniform_set); + } + render_base_uniform_set = RID(); +} + +void RenderForwardMobile::_update_render_base_uniform_set() { + RendererRD::LightStorage *light_storage = RendererRD::LightStorage::get_singleton(); + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + + if (render_base_uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(render_base_uniform_set) || (lightmap_texture_array_version != light_storage->lightmap_array_get_version())) { + if (render_base_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(render_base_uniform_set)) { + RD::get_singleton()->free(render_base_uniform_set); + } + + // This is all loaded into set 0 + + lightmap_texture_array_version = light_storage->lightmap_array_get_version(); + + Vector<RD::Uniform> uniforms; + + { + Vector<RID> ids; + ids.resize(12); + RID *ids_ptr = ids.ptrw(); + ids_ptr[0] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[1] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[2] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[3] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[4] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[5] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[6] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[7] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[8] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[9] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[10] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[11] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + + RD::Uniform u(RD::UNIFORM_TYPE_SAMPLER, 1, ids); + + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 2; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.append_id(scene_shader.shadow_sampler); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 3; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + RID sampler; + switch (decals_get_filter()) { + case RS::DECAL_FILTER_NEAREST: { + sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_LINEAR: { + sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_NEAREST_MIPMAPS: { + sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_LINEAR_MIPMAPS: { + sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_NEAREST_MIPMAPS_ANISOTROPIC: { + sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_LINEAR_MIPMAPS_ANISOTROPIC: { + sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + } + + u.append_id(sampler); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 4; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + RID sampler; + switch (light_projectors_get_filter()) { + case RS::LIGHT_PROJECTOR_FILTER_NEAREST: { + sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_LINEAR: { + sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_NEAREST_MIPMAPS: { + sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS: { + sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_NEAREST_MIPMAPS_ANISOTROPIC: { + sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS_ANISOTROPIC: { + sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + } + + u.append_id(sampler); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 5; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.append_id(RendererRD::LightStorage::get_singleton()->get_omni_light_buffer()); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 6; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.append_id(RendererRD::LightStorage::get_singleton()->get_spot_light_buffer()); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 7; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.append_id(RendererRD::LightStorage::get_singleton()->get_reflection_probe_buffer()); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 8; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.append_id(RendererRD::LightStorage::get_singleton()->get_directional_light_buffer()); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 9; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.append_id(scene_state.lightmap_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 10; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.append_id(scene_state.lightmap_capture_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 11; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID decal_atlas = RendererRD::TextureStorage::get_singleton()->decal_atlas_get_texture(); + u.append_id(decal_atlas); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 12; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID decal_atlas = RendererRD::TextureStorage::get_singleton()->decal_atlas_get_texture_srgb(); + u.append_id(decal_atlas); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 13; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.append_id(RendererRD::TextureStorage::get_singleton()->get_decal_buffer()); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 14; + u.append_id(RendererRD::MaterialStorage::get_singleton()->global_shader_uniforms_get_storage_buffer()); + uniforms.push_back(u); + } + + render_base_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, scene_shader.default_shader_rd, SCENE_UNIFORM_SET); + } +} + +RID RenderForwardMobile::_render_buffers_get_normal_texture(Ref<RenderSceneBuffersRD> p_render_buffers) { + return RID(); +} + +RID RenderForwardMobile::_render_buffers_get_velocity_texture(Ref<RenderSceneBuffersRD> p_render_buffers) { + return RID(); +} + +_FORCE_INLINE_ static uint32_t _indices_to_primitives(RS::PrimitiveType p_primitive, uint32_t p_indices) { + static const uint32_t divisor[RS::PRIMITIVE_MAX] = { 1, 2, 1, 3, 1 }; + static const uint32_t subtractor[RS::PRIMITIVE_MAX] = { 0, 0, 1, 0, 1 }; + return (p_indices - subtractor[p_primitive]) / divisor[p_primitive]; +} + +void RenderForwardMobile::_fill_render_list(RenderListType p_render_list, const RenderDataRD *p_render_data, PassMode p_pass_mode, bool p_append) { + RendererRD::MeshStorage *mesh_storage = RendererRD::MeshStorage::get_singleton(); + + if (p_render_list == RENDER_LIST_OPAQUE) { + scene_state.used_sss = false; + scene_state.used_screen_texture = false; + scene_state.used_normal_texture = false; + scene_state.used_depth_texture = false; + } + uint32_t lightmap_captures_used = 0; + + Plane near_plane(-p_render_data->scene_data->cam_transform.basis.get_column(Vector3::AXIS_Z), p_render_data->scene_data->cam_transform.origin); + near_plane.d += p_render_data->scene_data->cam_projection.get_z_near(); + float z_max = p_render_data->scene_data->cam_projection.get_z_far() - p_render_data->scene_data->cam_projection.get_z_near(); + + RenderList *rl = &render_list[p_render_list]; + + // Parse any updates on our geometry, updates surface caches and such + _update_dirty_geometry_instances(); + + if (!p_append) { + rl->clear(); + if (p_render_list == RENDER_LIST_OPAQUE) { + render_list[RENDER_LIST_ALPHA].clear(); //opaque fills alpha too + } + } + + //fill list + + for (int i = 0; i < (int)p_render_data->instances->size(); i++) { + GeometryInstanceForwardMobile *inst = static_cast<GeometryInstanceForwardMobile *>((*p_render_data->instances)[i]); + + Vector3 support_min = inst->transformed_aabb.get_support(-near_plane.normal); + inst->depth = near_plane.distance_to(support_min); + uint32_t depth_layer = CLAMP(int(inst->depth * 16 / z_max), 0, 15); + + uint32_t flags = inst->base_flags; //fill flags if appropriate + + if (inst->non_uniform_scale) { + flags |= INSTANCE_DATA_FLAGS_NON_UNIFORM_SCALE; + } + + bool uses_lightmap = false; + // bool uses_gi = false; + + if (p_render_list == RENDER_LIST_OPAQUE) { + if (inst->lightmap_instance.is_valid()) { + int32_t lightmap_cull_index = -1; + for (uint32_t j = 0; j < scene_state.lightmaps_used; j++) { + if (scene_state.lightmap_ids[j] == inst->lightmap_instance) { + lightmap_cull_index = j; + break; + } + } + if (lightmap_cull_index >= 0) { + inst->gi_offset_cache = inst->lightmap_slice_index << 16; + inst->gi_offset_cache |= lightmap_cull_index; + flags |= INSTANCE_DATA_FLAG_USE_LIGHTMAP; + if (scene_state.lightmap_has_sh[lightmap_cull_index]) { + flags |= INSTANCE_DATA_FLAG_USE_SH_LIGHTMAP; + } + uses_lightmap = true; + } else { + inst->gi_offset_cache = 0xFFFFFFFF; + } + + } else if (inst->lightmap_sh) { + if (lightmap_captures_used < scene_state.max_lightmap_captures) { + const Color *src_capture = inst->lightmap_sh->sh; + LightmapCaptureData &lcd = scene_state.lightmap_captures[lightmap_captures_used]; + for (int j = 0; j < 9; j++) { + lcd.sh[j * 4 + 0] = src_capture[j].r; + lcd.sh[j * 4 + 1] = src_capture[j].g; + lcd.sh[j * 4 + 2] = src_capture[j].b; + lcd.sh[j * 4 + 3] = src_capture[j].a; + } + flags |= INSTANCE_DATA_FLAG_USE_LIGHTMAP_CAPTURE; + inst->gi_offset_cache = lightmap_captures_used; + lightmap_captures_used++; + uses_lightmap = true; + } + } + } + inst->flags_cache = flags; + + GeometryInstanceSurfaceDataCache *surf = inst->surface_caches; + + while (surf) { + surf->sort.uses_lightmap = 0; + + // LOD + + if (p_render_data->scene_data->screen_mesh_lod_threshold > 0.0 && mesh_storage->mesh_surface_has_lod(surf->surface)) { + // Get the LOD support points on the mesh AABB. + Vector3 lod_support_min = inst->transformed_aabb.get_support(p_render_data->scene_data->cam_transform.basis.get_column(Vector3::AXIS_Z)); + Vector3 lod_support_max = inst->transformed_aabb.get_support(-p_render_data->scene_data->cam_transform.basis.get_column(Vector3::AXIS_Z)); + + // Get the distances to those points on the AABB from the camera origin. + float distance_min = (float)p_render_data->scene_data->cam_transform.origin.distance_to(lod_support_min); + float distance_max = (float)p_render_data->scene_data->cam_transform.origin.distance_to(lod_support_max); + + float distance = 0.0; + + if (distance_min * distance_max < 0.0) { + //crossing plane + distance = 0.0; + } else if (distance_min >= 0.0) { + distance = distance_min; + } else if (distance_max <= 0.0) { + distance = -distance_max; + } + + if (p_render_data->scene_data->cam_orthogonal) { + distance = 1.0; + } + + uint32_t indices = 0; + surf->lod_index = mesh_storage->mesh_surface_get_lod(surf->surface, inst->lod_model_scale * inst->lod_bias, distance * p_render_data->scene_data->lod_distance_multiplier, p_render_data->scene_data->screen_mesh_lod_threshold, indices); + if (p_render_data->render_info) { + indices = _indices_to_primitives(surf->primitive, indices); + if (p_render_list == RENDER_LIST_OPAQUE) { //opaque + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_VISIBLE][RS::VIEWPORT_RENDER_INFO_PRIMITIVES_IN_FRAME] += indices; + } else if (p_render_list == RENDER_LIST_SECONDARY) { //shadow + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_SHADOW][RS::VIEWPORT_RENDER_INFO_PRIMITIVES_IN_FRAME] += indices; + } + } + } else { + surf->lod_index = 0; + if (p_render_data->render_info) { + uint32_t to_draw = mesh_storage->mesh_surface_get_vertices_drawn_count(surf->surface); + to_draw = _indices_to_primitives(surf->primitive, to_draw); + to_draw *= inst->instance_count; + if (p_render_list == RENDER_LIST_OPAQUE) { //opaque + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_VISIBLE][RS::VIEWPORT_RENDER_INFO_PRIMITIVES_IN_FRAME] += mesh_storage->mesh_surface_get_vertices_drawn_count(surf->surface); + } else if (p_render_list == RENDER_LIST_SECONDARY) { //shadow + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_SHADOW][RS::VIEWPORT_RENDER_INFO_PRIMITIVES_IN_FRAME] += mesh_storage->mesh_surface_get_vertices_drawn_count(surf->surface); + } + } + } + + // ADD Element + if (p_pass_mode == PASS_MODE_COLOR || p_pass_mode == PASS_MODE_COLOR_TRANSPARENT) { +#ifdef DEBUG_ENABLED + bool force_alpha = unlikely(get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_OVERDRAW); +#else + bool force_alpha = false; +#endif + if (!force_alpha && (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE)) { + rl->add_element(surf); + } + if (force_alpha || (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_PASS_ALPHA)) { + render_list[RENDER_LIST_ALPHA].add_element(surf); + } + + if (uses_lightmap) { + surf->sort.uses_lightmap = 1; // This needs to become our lightmap index but we'll do that in a separate PR. + } + + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_SUBSURFACE_SCATTERING) { + scene_state.used_sss = true; + } + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_SCREEN_TEXTURE) { + scene_state.used_screen_texture = true; + } + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_NORMAL_TEXTURE) { + scene_state.used_normal_texture = true; + } + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_DEPTH_TEXTURE) { + scene_state.used_depth_texture = true; + } + + } else if (p_pass_mode == PASS_MODE_SHADOW || p_pass_mode == PASS_MODE_SHADOW_DP) { + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_PASS_SHADOW) { + rl->add_element(surf); + } + } else { + if (surf->flags & (GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH | GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE)) { + rl->add_element(surf); + } + } + + surf->sort.depth_layer = depth_layer; + + surf = surf->next; + } + } +} + +void RenderForwardMobile::_setup_environment(const RenderDataRD *p_render_data, bool p_no_fog, const Size2i &p_screen_size, bool p_flip_y, const Color &p_default_bg_color, bool p_opaque_render_buffers, bool p_pancake_shadows, int p_index) { + Ref<RenderSceneBuffersRD> rd = p_render_data->render_buffers; + RID env = is_environment(p_render_data->environment) ? p_render_data->environment : RID(); + RID reflection_probe_instance = p_render_data->reflection_probe.is_valid() ? RendererRD::LightStorage::get_singleton()->reflection_probe_instance_get_probe(p_render_data->reflection_probe) : RID(); + + // May do this earlier in RenderSceneRenderRD::render_scene + if (p_index >= (int)scene_state.uniform_buffers.size()) { + uint32_t from = scene_state.uniform_buffers.size(); + scene_state.uniform_buffers.resize(p_index + 1); + for (uint32_t i = from; i < scene_state.uniform_buffers.size(); i++) { + scene_state.uniform_buffers[i] = p_render_data->scene_data->create_uniform_buffer(); + } + } + + p_render_data->scene_data->update_ubo(scene_state.uniform_buffers[p_index], get_debug_draw_mode(), env, reflection_probe_instance, p_render_data->camera_attributes, p_flip_y, p_pancake_shadows, p_screen_size, p_default_bg_color, _render_buffers_get_luminance_multiplier(), p_opaque_render_buffers); +} + +void RenderForwardMobile::_fill_element_info(RenderListType p_render_list, uint32_t p_offset, int32_t p_max_elements) { + RenderList *rl = &render_list[p_render_list]; + uint32_t element_total = p_max_elements >= 0 ? uint32_t(p_max_elements) : rl->elements.size(); + + rl->element_info.resize(p_offset + element_total); + + for (uint32_t i = 0; i < element_total; i++) { + GeometryInstanceSurfaceDataCache *surface = rl->elements[i + p_offset]; + RenderElementInfo &element_info = rl->element_info[p_offset + i]; + + element_info.lod_index = surface->lod_index; + element_info.uses_lightmap = surface->sort.uses_lightmap; + } +} + +/// RENDERING /// + +void RenderForwardMobile::_render_list(RenderingDevice::DrawListID p_draw_list, RenderingDevice::FramebufferFormatID p_framebuffer_Format, RenderListParameters *p_params, uint32_t p_from_element, uint32_t p_to_element) { + //use template for faster performance (pass mode comparisons are inlined) + + switch (p_params->pass_mode) { + case PASS_MODE_COLOR: { + _render_list_template<PASS_MODE_COLOR>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + case PASS_MODE_COLOR_TRANSPARENT: { + _render_list_template<PASS_MODE_COLOR_TRANSPARENT>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + case PASS_MODE_SHADOW: { + _render_list_template<PASS_MODE_SHADOW>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + case PASS_MODE_SHADOW_DP: { + _render_list_template<PASS_MODE_SHADOW_DP>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + case PASS_MODE_DEPTH_MATERIAL: { + _render_list_template<PASS_MODE_DEPTH_MATERIAL>(p_draw_list, p_framebuffer_Format, p_params, p_from_element, p_to_element); + } break; + } +} + +void RenderForwardMobile::_render_list_thread_function(uint32_t p_thread, RenderListParameters *p_params) { + uint32_t render_total = p_params->element_count; + uint32_t total_threads = WorkerThreadPool::get_singleton()->get_thread_count(); + uint32_t render_from = p_thread * render_total / total_threads; + uint32_t render_to = (p_thread + 1 == total_threads) ? render_total : ((p_thread + 1) * render_total / total_threads); + _render_list(thread_draw_lists[p_thread], p_params->framebuffer_format, p_params, render_from, render_to); +} + +void RenderForwardMobile::_render_list_with_threads(RenderListParameters *p_params, RID p_framebuffer, RD::InitialAction p_initial_color_action, RD::FinalAction p_final_color_action, RD::InitialAction p_initial_depth_action, RD::FinalAction p_final_depth_action, const Vector<Color> &p_clear_color_values, float p_clear_depth, uint32_t p_clear_stencil, const Rect2 &p_region, const Vector<RID> &p_storage_textures) { + RD::FramebufferFormatID fb_format = RD::get_singleton()->framebuffer_get_format(p_framebuffer); + p_params->framebuffer_format = fb_format; + + if ((uint32_t)p_params->element_count > render_list_thread_threshold && false) { // secondary command buffers need more testing at this time + //multi threaded + thread_draw_lists.resize(WorkerThreadPool::get_singleton()->get_thread_count()); + RD::get_singleton()->draw_list_begin_split(p_framebuffer, thread_draw_lists.size(), thread_draw_lists.ptr(), p_initial_color_action, p_final_color_action, p_initial_depth_action, p_final_depth_action, p_clear_color_values, p_clear_depth, p_clear_stencil, p_region, p_storage_textures); + WorkerThreadPool::GroupID group_task = WorkerThreadPool::get_singleton()->add_template_group_task(this, &RenderForwardMobile::_render_list_thread_function, p_params, thread_draw_lists.size(), -1, true, SNAME("ForwardMobileRenderSubpass")); + WorkerThreadPool::get_singleton()->wait_for_group_task_completion(group_task); + + RD::get_singleton()->draw_list_end(p_params->barrier); + } else { + //single threaded + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_framebuffer, p_initial_color_action, p_final_color_action, p_initial_depth_action, p_final_depth_action, p_clear_color_values, p_clear_depth, p_clear_stencil, p_region, p_storage_textures); + _render_list(draw_list, fb_format, p_params, 0, p_params->element_count); + RD::get_singleton()->draw_list_end(p_params->barrier); + } +} + +template <RenderForwardMobile::PassMode p_pass_mode> +void RenderForwardMobile::_render_list_template(RenderingDevice::DrawListID p_draw_list, RenderingDevice::FramebufferFormatID p_framebuffer_Format, RenderListParameters *p_params, uint32_t p_from_element, uint32_t p_to_element) { + RendererRD::MeshStorage *mesh_storage = RendererRD::MeshStorage::get_singleton(); + + RD::DrawListID draw_list = p_draw_list; + RD::FramebufferFormatID framebuffer_format = p_framebuffer_Format; + + //global scope bindings + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, render_base_uniform_set, SCENE_UNIFORM_SET); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, p_params->render_pass_uniform_set, RENDER_PASS_UNIFORM_SET); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, scene_shader.default_vec4_xform_uniform_set, TRANSFORMS_UNIFORM_SET); + + RID prev_material_uniform_set; + + RID prev_vertex_array_rd; + RID prev_index_array_rd; + RID prev_pipeline_rd; + RID prev_xforms_uniform_set; + + bool shadow_pass = (p_params->pass_mode == PASS_MODE_SHADOW) || (p_params->pass_mode == PASS_MODE_SHADOW_DP); + + for (uint32_t i = p_from_element; i < p_to_element; i++) { + const GeometryInstanceSurfaceDataCache *surf = p_params->elements[i]; + const RenderElementInfo &element_info = p_params->element_info[i]; + const GeometryInstanceForwardMobile *inst = surf->owner; + + if (inst->instance_count == 0) { + continue; + } + + uint32_t base_spec_constants = p_params->spec_constant_base_flags; + + // GeometryInstanceForwardMobile::PushConstant push_constant = inst->push_constant; + GeometryInstanceForwardMobile::PushConstant push_constant; + + if (inst->store_transform_cache) { + RendererRD::MaterialStorage::store_transform(inst->transform, push_constant.transform); + } else { + RendererRD::MaterialStorage::store_transform(Transform3D(), push_constant.transform); + } + +#ifdef REAL_T_IS_DOUBLE + // Split the origin into two components, the float approximation and the missing precision + // In the shader we will combine these back together to restore the lost precision. + RendererRD::MaterialStorage::split_double(inst->transform.origin.x, &push_constant.transform[12], &push_constant.transform[3]); + RendererRD::MaterialStorage::split_double(inst->transform.origin.y, &push_constant.transform[13], &push_constant.transform[7]); + RendererRD::MaterialStorage::split_double(inst->transform.origin.z, &push_constant.transform[14], &push_constant.transform[11]); +#endif + + push_constant.flags = inst->flags_cache; + push_constant.gi_offset = inst->gi_offset_cache; + push_constant.layer_mask = inst->layer_mask; + push_constant.instance_uniforms_ofs = uint32_t(inst->shader_uniforms_offset); + + if (p_params->pass_mode == PASS_MODE_DEPTH_MATERIAL) { + // abuse lightmap_uv_scale[0] here, should not be needed here + push_constant.lightmap_uv_scale[0] = p_params->uv_offset.x; + push_constant.lightmap_uv_scale[1] = p_params->uv_offset.y; + } else { + push_constant.lightmap_uv_scale[0] = inst->lightmap_uv_scale.position.x; + push_constant.lightmap_uv_scale[1] = inst->lightmap_uv_scale.position.y; + push_constant.lightmap_uv_scale[2] = inst->lightmap_uv_scale.size.x; + push_constant.lightmap_uv_scale[3] = inst->lightmap_uv_scale.size.y; + }; + + RID material_uniform_set; + SceneShaderForwardMobile::ShaderData *shader; + void *mesh_surface; + + if (shadow_pass) { + material_uniform_set = surf->material_uniform_set_shadow; + shader = surf->shader_shadow; + mesh_surface = surf->surface_shadow; + + } else { + if (inst->use_projector) { + base_spec_constants |= 1 << SPEC_CONSTANT_USING_PROJECTOR; + } + if (inst->use_soft_shadow) { + base_spec_constants |= 1 << SPEC_CONSTANT_USING_SOFT_SHADOWS; + } + forward_id_storage_mobile->fill_push_constant_instance_indices(&push_constant, base_spec_constants, inst); + +#ifdef DEBUG_ENABLED + if (unlikely(get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_LIGHTING)) { + material_uniform_set = scene_shader.default_material_uniform_set; + shader = scene_shader.default_material_shader_ptr; + } else if (unlikely(get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_OVERDRAW)) { + material_uniform_set = scene_shader.overdraw_material_uniform_set; + shader = scene_shader.overdraw_material_shader_ptr; + } else { +#endif + material_uniform_set = surf->material_uniform_set; + shader = surf->shader; +#ifdef DEBUG_ENABLED + } +#endif + mesh_surface = surf->surface; + } + + if (!mesh_surface) { + continue; + } + + //find cull variant + SceneShaderForwardMobile::ShaderData::CullVariant cull_variant; + + if (p_params->pass_mode == PASS_MODE_DEPTH_MATERIAL || ((p_params->pass_mode == PASS_MODE_SHADOW || p_params->pass_mode == PASS_MODE_SHADOW_DP) && surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_DOUBLE_SIDED_SHADOWS)) { + cull_variant = SceneShaderForwardMobile::ShaderData::CULL_VARIANT_DOUBLE_SIDED; + } else { + bool mirror = surf->owner->mirror; + if (p_params->reverse_cull) { + mirror = !mirror; + } + cull_variant = mirror ? SceneShaderForwardMobile::ShaderData::CULL_VARIANT_REVERSED : SceneShaderForwardMobile::ShaderData::CULL_VARIANT_NORMAL; + } + + RS::PrimitiveType primitive = surf->primitive; + RID xforms_uniform_set = surf->owner->transforms_uniform_set; + + SceneShaderForwardMobile::ShaderVersion shader_version = SceneShaderForwardMobile::SHADER_VERSION_MAX; // Assigned to silence wrong -Wmaybe-initialized. + + switch (p_params->pass_mode) { + case PASS_MODE_COLOR: + case PASS_MODE_COLOR_TRANSPARENT: { + if (element_info.uses_lightmap) { + shader_version = p_params->view_count > 1 ? SceneShaderForwardMobile::SHADER_VERSION_LIGHTMAP_COLOR_PASS_MULTIVIEW : SceneShaderForwardMobile::SHADER_VERSION_LIGHTMAP_COLOR_PASS; + } else { + shader_version = p_params->view_count > 1 ? SceneShaderForwardMobile::SHADER_VERSION_COLOR_PASS_MULTIVIEW : SceneShaderForwardMobile::SHADER_VERSION_COLOR_PASS; + } + } break; + case PASS_MODE_SHADOW: { + shader_version = p_params->view_count > 1 ? SceneShaderForwardMobile::SHADER_VERSION_SHADOW_PASS_MULTIVIEW : SceneShaderForwardMobile::SHADER_VERSION_SHADOW_PASS; + } break; + case PASS_MODE_SHADOW_DP: { + ERR_FAIL_COND_MSG(p_params->view_count > 1, "Multiview not supported for shadow DP pass"); + shader_version = SceneShaderForwardMobile::SHADER_VERSION_SHADOW_PASS_DP; + } break; + case PASS_MODE_DEPTH_MATERIAL: { + ERR_FAIL_COND_MSG(p_params->view_count > 1, "Multiview not supported for material pass"); + shader_version = SceneShaderForwardMobile::SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL; + } break; + } + + PipelineCacheRD *pipeline = nullptr; + + pipeline = &shader->pipelines[cull_variant][primitive][shader_version]; + + RD::VertexFormatID vertex_format = -1; + RID vertex_array_rd; + RID index_array_rd; + + //skeleton and blend shape + if (surf->owner->mesh_instance.is_valid()) { + mesh_storage->mesh_instance_surface_get_vertex_arrays_and_format(surf->owner->mesh_instance, surf->surface_index, pipeline->get_vertex_input_mask(), vertex_array_rd, vertex_format); + } else { + mesh_storage->mesh_surface_get_vertex_arrays_and_format(mesh_surface, pipeline->get_vertex_input_mask(), vertex_array_rd, vertex_format); + } + + index_array_rd = mesh_storage->mesh_surface_get_index_array(mesh_surface, element_info.lod_index); + + if (prev_vertex_array_rd != vertex_array_rd) { + RD::get_singleton()->draw_list_bind_vertex_array(draw_list, vertex_array_rd); + prev_vertex_array_rd = vertex_array_rd; + } + + if (prev_index_array_rd != index_array_rd) { + if (index_array_rd.is_valid()) { + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array_rd); + } + prev_index_array_rd = index_array_rd; + } + + RID pipeline_rd = pipeline->get_render_pipeline(vertex_format, framebuffer_format, p_params->force_wireframe, p_params->subpass, base_spec_constants); + + if (pipeline_rd != prev_pipeline_rd) { + // checking with prev shader does not make so much sense, as + // the pipeline may still be different. + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, pipeline_rd); + prev_pipeline_rd = pipeline_rd; + } + + if (xforms_uniform_set.is_valid() && prev_xforms_uniform_set != xforms_uniform_set) { + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, xforms_uniform_set, TRANSFORMS_UNIFORM_SET); + prev_xforms_uniform_set = xforms_uniform_set; + } + + if (material_uniform_set != prev_material_uniform_set) { + // Update uniform set. + if (material_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(material_uniform_set)) { // Material may not have a uniform set. + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, material_uniform_set, MATERIAL_UNIFORM_SET); + } + + prev_material_uniform_set = material_uniform_set; + } + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &push_constant, sizeof(GeometryInstanceForwardMobile::PushConstant)); + + uint32_t instance_count = surf->owner->instance_count > 1 ? surf->owner->instance_count : 1; + if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_PARTICLE_TRAILS) { + instance_count /= surf->owner->trail_steps; + } + + RD::get_singleton()->draw_list_draw(draw_list, index_array_rd.is_valid(), instance_count); + } +} + +/* Geometry instance */ + +RenderGeometryInstance *RenderForwardMobile::geometry_instance_create(RID p_base) { + RS::InstanceType type = RSG::utilities->get_base_type(p_base); + ERR_FAIL_COND_V(!((1 << type) & RS::INSTANCE_GEOMETRY_MASK), nullptr); + + GeometryInstanceForwardMobile *ginstance = geometry_instance_alloc.alloc(); + ginstance->data = memnew(GeometryInstanceForwardMobile::Data); + + ginstance->data->base = p_base; + ginstance->data->base_type = type; + + ginstance->_mark_dirty(); + + return ginstance; +} + +void RenderForwardMobile::GeometryInstanceForwardMobile::set_use_lightmap(RID p_lightmap_instance, const Rect2 &p_lightmap_uv_scale, int p_lightmap_slice_index) { + lightmap_instance = p_lightmap_instance; + lightmap_uv_scale = p_lightmap_uv_scale; + lightmap_slice_index = p_lightmap_slice_index; + + _mark_dirty(); +} + +void RenderForwardMobile::GeometryInstanceForwardMobile::set_lightmap_capture(const Color *p_sh9) { + if (p_sh9) { + if (lightmap_sh == nullptr) { + lightmap_sh = RenderForwardMobile::get_singleton()->geometry_instance_lightmap_sh.alloc(); + } + + memcpy(lightmap_sh->sh, p_sh9, sizeof(Color) * 9); + } else { + if (lightmap_sh != nullptr) { + RenderForwardMobile::get_singleton()->geometry_instance_lightmap_sh.free(lightmap_sh); + lightmap_sh = nullptr; + } + } + _mark_dirty(); +} + +void RenderForwardMobile::geometry_instance_free(RenderGeometryInstance *p_geometry_instance) { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + if (ginstance->lightmap_sh != nullptr) { + geometry_instance_lightmap_sh.free(ginstance->lightmap_sh); + } + GeometryInstanceSurfaceDataCache *surf = ginstance->surface_caches; + while (surf) { + GeometryInstanceSurfaceDataCache *next = surf->next; + geometry_instance_surface_alloc.free(surf); + surf = next; + } + memdelete(ginstance->data); + geometry_instance_alloc.free(ginstance); +} + +uint32_t RenderForwardMobile::geometry_instance_get_pair_mask() { + return ((1 << RS::INSTANCE_LIGHT) + (1 << RS::INSTANCE_REFLECTION_PROBE) + (1 << RS::INSTANCE_DECAL)); +} + +void RenderForwardMobile::GeometryInstanceForwardMobile::pair_light_instances(const RID *p_light_instances, uint32_t p_light_instance_count) { + omni_light_count = 0; + spot_light_count = 0; + + for (uint32_t i = 0; i < p_light_instance_count; i++) { + RS::LightType type = RendererRD::LightStorage::get_singleton()->light_instance_get_type(p_light_instances[i]); + switch (type) { + case RS::LIGHT_OMNI: { + if (omni_light_count < (uint32_t)MAX_RDL_CULL) { + omni_lights[omni_light_count] = RendererRD::LightStorage::get_singleton()->light_instance_get_forward_id(p_light_instances[i]); + omni_light_count++; + } + } break; + case RS::LIGHT_SPOT: { + if (spot_light_count < (uint32_t)MAX_RDL_CULL) { + spot_lights[spot_light_count] = RendererRD::LightStorage::get_singleton()->light_instance_get_forward_id(p_light_instances[i]); + spot_light_count++; + } + } break; + default: + break; + } + } +} + +void RenderForwardMobile::GeometryInstanceForwardMobile::pair_reflection_probe_instances(const RID *p_reflection_probe_instances, uint32_t p_reflection_probe_instance_count) { + reflection_probe_count = p_reflection_probe_instance_count < (uint32_t)MAX_RDL_CULL ? p_reflection_probe_instance_count : (uint32_t)MAX_RDL_CULL; + for (uint32_t i = 0; i < reflection_probe_count; i++) { + reflection_probes[i] = RendererRD::LightStorage::get_singleton()->reflection_probe_instance_get_forward_id(p_reflection_probe_instances[i]); + } +} + +void RenderForwardMobile::GeometryInstanceForwardMobile::pair_decal_instances(const RID *p_decal_instances, uint32_t p_decal_instance_count) { + decals_count = p_decal_instance_count < (uint32_t)MAX_RDL_CULL ? p_decal_instance_count : (uint32_t)MAX_RDL_CULL; + for (uint32_t i = 0; i < decals_count; i++) { + decals[i] = RendererRD::TextureStorage::get_singleton()->decal_instance_get_forward_id(p_decal_instances[i]); + } +} + +void RenderForwardMobile::GeometryInstanceForwardMobile::set_softshadow_projector_pairing(bool p_softshadow, bool p_projector) { + use_projector = p_projector; + use_soft_shadow = p_softshadow; +} + +void RenderForwardMobile::GeometryInstanceForwardMobile::_mark_dirty() { + if (dirty_list_element.in_list()) { + return; + } + + //clear surface caches + GeometryInstanceSurfaceDataCache *surf = surface_caches; + + while (surf) { + GeometryInstanceSurfaceDataCache *next = surf->next; + RenderForwardMobile::get_singleton()->geometry_instance_surface_alloc.free(surf); + surf = next; + } + + surface_caches = nullptr; + + RenderForwardMobile::get_singleton()->geometry_instance_dirty_list.add(&dirty_list_element); +} + +void RenderForwardMobile::_geometry_instance_add_surface_with_material(GeometryInstanceForwardMobile *ginstance, uint32_t p_surface, SceneShaderForwardMobile::MaterialData *p_material, uint32_t p_material_id, uint32_t p_shader_id, RID p_mesh) { + RendererRD::MeshStorage *mesh_storage = RendererRD::MeshStorage::get_singleton(); + + bool has_read_screen_alpha = p_material->shader_data->uses_screen_texture || p_material->shader_data->uses_depth_texture || p_material->shader_data->uses_normal_texture; + bool has_base_alpha = ((p_material->shader_data->uses_alpha && !p_material->shader_data->uses_alpha_clip) || has_read_screen_alpha); + bool has_blend_alpha = p_material->shader_data->uses_blend_alpha; + bool has_alpha = has_base_alpha || has_blend_alpha; + + uint32_t flags = 0; + + if (p_material->shader_data->uses_sss) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_SUBSURFACE_SCATTERING; + } + + if (p_material->shader_data->uses_screen_texture) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_SCREEN_TEXTURE; + } + + if (p_material->shader_data->uses_depth_texture) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_DEPTH_TEXTURE; + } + + if (p_material->shader_data->uses_normal_texture) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_NORMAL_TEXTURE; + } + + if (ginstance->data->cast_double_sided_shadows) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_DOUBLE_SIDED_SHADOWS; + } + + if (has_alpha || has_read_screen_alpha || p_material->shader_data->depth_draw == SceneShaderForwardMobile::ShaderData::DEPTH_DRAW_DISABLED || p_material->shader_data->depth_test == SceneShaderForwardMobile::ShaderData::DEPTH_TEST_DISABLED) { + //material is only meant for alpha pass + flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_ALPHA; + if (p_material->shader_data->uses_depth_pre_pass && !(p_material->shader_data->depth_draw == SceneShaderForwardMobile::ShaderData::DEPTH_DRAW_DISABLED || p_material->shader_data->depth_test == SceneShaderForwardMobile::ShaderData::DEPTH_TEST_DISABLED)) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH; + flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_SHADOW; + } + } else { + flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE; + flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH; + flags |= GeometryInstanceSurfaceDataCache::FLAG_PASS_SHADOW; + } + + if (p_material->shader_data->uses_particle_trails) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_PARTICLE_TRAILS; + } + + SceneShaderForwardMobile::MaterialData *material_shadow = nullptr; + void *surface_shadow = nullptr; + if (!p_material->shader_data->uses_particle_trails && !p_material->shader_data->writes_modelview_or_projection && !p_material->shader_data->uses_vertex && !p_material->shader_data->uses_discard && !p_material->shader_data->uses_depth_pre_pass && !p_material->shader_data->uses_alpha_clip) { + flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_SHARED_SHADOW_MATERIAL; + material_shadow = static_cast<SceneShaderForwardMobile::MaterialData *>(RendererRD::MaterialStorage::get_singleton()->material_get_data(scene_shader.default_material, RendererRD::MaterialStorage::SHADER_TYPE_3D)); + + RID shadow_mesh = mesh_storage->mesh_get_shadow_mesh(p_mesh); + + if (shadow_mesh.is_valid()) { + surface_shadow = mesh_storage->mesh_get_surface(shadow_mesh, p_surface); + } + + } else { + material_shadow = p_material; + } + + GeometryInstanceSurfaceDataCache *sdcache = geometry_instance_surface_alloc.alloc(); + + sdcache->flags = flags; + + sdcache->shader = p_material->shader_data; + sdcache->material_uniform_set = p_material->uniform_set; + sdcache->surface = mesh_storage->mesh_get_surface(p_mesh, p_surface); + sdcache->primitive = mesh_storage->mesh_surface_get_primitive(sdcache->surface); + sdcache->surface_index = p_surface; + + if (ginstance->data->dirty_dependencies) { + RSG::utilities->base_update_dependency(p_mesh, &ginstance->data->dependency_tracker); + } + + //shadow + sdcache->shader_shadow = material_shadow->shader_data; + sdcache->material_uniform_set_shadow = material_shadow->uniform_set; + + sdcache->surface_shadow = surface_shadow ? surface_shadow : sdcache->surface; + + sdcache->owner = ginstance; + + sdcache->next = ginstance->surface_caches; + ginstance->surface_caches = sdcache; + + //sortkey + + sdcache->sort.sort_key1 = 0; + sdcache->sort.sort_key2 = 0; + + sdcache->sort.surface_index = p_surface; + sdcache->sort.material_id_low = p_material_id & 0x0000FFFF; + sdcache->sort.material_id_hi = p_material_id >> 16; + sdcache->sort.shader_id = p_shader_id; + sdcache->sort.geometry_id = p_mesh.get_local_index(); + // sdcache->sort.uses_forward_gi = ginstance->can_sdfgi; + sdcache->sort.priority = p_material->priority; +} + +void RenderForwardMobile::_geometry_instance_add_surface_with_material_chain(GeometryInstanceForwardMobile *ginstance, uint32_t p_surface, SceneShaderForwardMobile::MaterialData *p_material, RID p_mat_src, RID p_mesh) { + SceneShaderForwardMobile::MaterialData *material = p_material; + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + + _geometry_instance_add_surface_with_material(ginstance, p_surface, material, p_mat_src.get_local_index(), material_storage->material_get_shader_id(p_mat_src), p_mesh); + + while (material->next_pass.is_valid()) { + RID next_pass = material->next_pass; + material = static_cast<SceneShaderForwardMobile::MaterialData *>(material_storage->material_get_data(next_pass, RendererRD::MaterialStorage::SHADER_TYPE_3D)); + if (!material || !material->shader_data->valid) { + break; + } + if (ginstance->data->dirty_dependencies) { + material_storage->material_update_dependency(next_pass, &ginstance->data->dependency_tracker); + } + _geometry_instance_add_surface_with_material(ginstance, p_surface, material, next_pass.get_local_index(), material_storage->material_get_shader_id(next_pass), p_mesh); + } +} + +void RenderForwardMobile::_geometry_instance_add_surface(GeometryInstanceForwardMobile *ginstance, uint32_t p_surface, RID p_material, RID p_mesh) { + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + RID m_src; + + m_src = ginstance->data->material_override.is_valid() ? ginstance->data->material_override : p_material; + + SceneShaderForwardMobile::MaterialData *material = nullptr; + + if (m_src.is_valid()) { + material = static_cast<SceneShaderForwardMobile::MaterialData *>(material_storage->material_get_data(m_src, RendererRD::MaterialStorage::SHADER_TYPE_3D)); + if (!material || !material->shader_data->valid) { + material = nullptr; + } + } + + if (material) { + if (ginstance->data->dirty_dependencies) { + material_storage->material_update_dependency(m_src, &ginstance->data->dependency_tracker); + } + } else { + material = static_cast<SceneShaderForwardMobile::MaterialData *>(material_storage->material_get_data(scene_shader.default_material, RendererRD::MaterialStorage::SHADER_TYPE_3D)); + m_src = scene_shader.default_material; + } + + ERR_FAIL_COND(!material); + + _geometry_instance_add_surface_with_material_chain(ginstance, p_surface, material, m_src, p_mesh); + + if (ginstance->data->material_overlay.is_valid()) { + m_src = ginstance->data->material_overlay; + + material = static_cast<SceneShaderForwardMobile::MaterialData *>(material_storage->material_get_data(m_src, RendererRD::MaterialStorage::SHADER_TYPE_3D)); + if (material && material->shader_data->valid) { + if (ginstance->data->dirty_dependencies) { + material_storage->material_update_dependency(m_src, &ginstance->data->dependency_tracker); + } + + _geometry_instance_add_surface_with_material_chain(ginstance, p_surface, material, m_src, p_mesh); + } + } +} + +void RenderForwardMobile::_geometry_instance_update(RenderGeometryInstance *p_geometry_instance) { + RendererRD::MeshStorage *mesh_storage = RendererRD::MeshStorage::get_singleton(); + RendererRD::ParticlesStorage *particles_storage = RendererRD::ParticlesStorage::get_singleton(); + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_geometry_instance); + + if (ginstance->data->dirty_dependencies) { + ginstance->data->dependency_tracker.update_begin(); + } + + //add geometry for drawing + switch (ginstance->data->base_type) { + case RS::INSTANCE_MESH: { + const RID *materials = nullptr; + uint32_t surface_count; + RID mesh = ginstance->data->base; + + materials = mesh_storage->mesh_get_surface_count_and_materials(mesh, surface_count); + if (materials) { + //if no materials, no surfaces. + const RID *inst_materials = ginstance->data->surface_materials.ptr(); + uint32_t surf_mat_count = ginstance->data->surface_materials.size(); + + for (uint32_t j = 0; j < surface_count; j++) { + RID material = (j < surf_mat_count && inst_materials[j].is_valid()) ? inst_materials[j] : materials[j]; + _geometry_instance_add_surface(ginstance, j, material, mesh); + } + } + + ginstance->instance_count = 1; + + } break; + + case RS::INSTANCE_MULTIMESH: { + RID mesh = mesh_storage->multimesh_get_mesh(ginstance->data->base); + if (mesh.is_valid()) { + const RID *materials = nullptr; + uint32_t surface_count; + + materials = mesh_storage->mesh_get_surface_count_and_materials(mesh, surface_count); + if (materials) { + for (uint32_t j = 0; j < surface_count; j++) { + _geometry_instance_add_surface(ginstance, j, materials[j], mesh); + } + } + + ginstance->instance_count = mesh_storage->multimesh_get_instances_to_draw(ginstance->data->base); + } + + } break; +#if 0 + case RS::INSTANCE_IMMEDIATE: { + RasterizerStorageGLES3::Immediate *immediate = storage->immediate_owner.get_or_null(inst->base); + ERR_CONTINUE(!immediate); + + _add_geometry(immediate, inst, nullptr, -1, p_depth_pass, p_shadow_pass); + + } break; +#endif + case RS::INSTANCE_PARTICLES: { + int draw_passes = particles_storage->particles_get_draw_passes(ginstance->data->base); + + for (int j = 0; j < draw_passes; j++) { + RID mesh = particles_storage->particles_get_draw_pass_mesh(ginstance->data->base, j); + if (!mesh.is_valid()) { + continue; + } + + const RID *materials = nullptr; + uint32_t surface_count; + + materials = mesh_storage->mesh_get_surface_count_and_materials(mesh, surface_count); + if (materials) { + for (uint32_t k = 0; k < surface_count; k++) { + _geometry_instance_add_surface(ginstance, k, materials[k], mesh); + } + } + } + + ginstance->instance_count = particles_storage->particles_get_amount(ginstance->data->base, ginstance->trail_steps); + + } break; + + default: { + } + } + + //Fill push constant + + bool store_transform = true; + ginstance->base_flags = 0; + + if (ginstance->data->base_type == RS::INSTANCE_MULTIMESH) { + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH; + if (mesh_storage->multimesh_get_transform_format(ginstance->data->base) == RS::MULTIMESH_TRANSFORM_2D) { + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_FORMAT_2D; + } + if (mesh_storage->multimesh_uses_colors(ginstance->data->base)) { + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_HAS_COLOR; + } + if (mesh_storage->multimesh_uses_custom_data(ginstance->data->base)) { + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_HAS_CUSTOM_DATA; + } + + ginstance->transforms_uniform_set = mesh_storage->multimesh_get_3d_uniform_set(ginstance->data->base, scene_shader.default_shader_rd, TRANSFORMS_UNIFORM_SET); + + } else if (ginstance->data->base_type == RS::INSTANCE_PARTICLES) { + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH; + if (false) { // 2D particles + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_FORMAT_2D; + } + + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_HAS_COLOR; + ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH_HAS_CUSTOM_DATA; + + //for particles, stride is the trail size + ginstance->base_flags |= (ginstance->trail_steps << INSTANCE_DATA_FLAGS_PARTICLE_TRAIL_SHIFT); + + if (!particles_storage->particles_is_using_local_coords(ginstance->data->base)) { + store_transform = false; + } + ginstance->transforms_uniform_set = particles_storage->particles_get_instance_buffer_uniform_set(ginstance->data->base, scene_shader.default_shader_rd, TRANSFORMS_UNIFORM_SET); + + } else if (ginstance->data->base_type == RS::INSTANCE_MESH) { + if (mesh_storage->skeleton_is_valid(ginstance->data->skeleton)) { + ginstance->transforms_uniform_set = mesh_storage->skeleton_get_3d_uniform_set(ginstance->data->skeleton, scene_shader.default_shader_rd, TRANSFORMS_UNIFORM_SET); + if (ginstance->data->dirty_dependencies) { + mesh_storage->skeleton_update_dependency(ginstance->data->skeleton, &ginstance->data->dependency_tracker); + } + } + } + + ginstance->store_transform_cache = store_transform; + + if (ginstance->data->dirty_dependencies) { + ginstance->data->dependency_tracker.update_end(); + ginstance->data->dirty_dependencies = false; + } + + ginstance->dirty_list_element.remove_from_list(); +} + +void RenderForwardMobile::_update_dirty_geometry_instances() { + while (geometry_instance_dirty_list.first()) { + _geometry_instance_update(geometry_instance_dirty_list.first()->self()); + } +} + +void RenderForwardMobile::_geometry_instance_dependency_changed(Dependency::DependencyChangedNotification p_notification, DependencyTracker *p_tracker) { + switch (p_notification) { + case Dependency::DEPENDENCY_CHANGED_MATERIAL: + case Dependency::DEPENDENCY_CHANGED_MESH: + case Dependency::DEPENDENCY_CHANGED_PARTICLES: + case Dependency::DEPENDENCY_CHANGED_MULTIMESH: + case Dependency::DEPENDENCY_CHANGED_SKELETON_DATA: { + static_cast<RenderGeometryInstance *>(p_tracker->userdata)->_mark_dirty(); + } break; + case Dependency::DEPENDENCY_CHANGED_MULTIMESH_VISIBLE_INSTANCES: { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_tracker->userdata); + if (ginstance->data->base_type == RS::INSTANCE_MULTIMESH) { + ginstance->instance_count = RendererRD::MeshStorage::get_singleton()->multimesh_get_instances_to_draw(ginstance->data->base); + } + } break; + default: { + //rest of notifications of no interest + } break; + } +} +void RenderForwardMobile::_geometry_instance_dependency_deleted(const RID &p_dependency, DependencyTracker *p_tracker) { + static_cast<RenderGeometryInstance *>(p_tracker->userdata)->_mark_dirty(); +} + +/* misc */ + +bool RenderForwardMobile::is_dynamic_gi_supported() const { + return false; +} + +bool RenderForwardMobile::is_volumetric_supported() const { + return false; +} + +uint32_t RenderForwardMobile::get_max_elements() const { + return 256; +} + +RenderForwardMobile *RenderForwardMobile::singleton = nullptr; + +void RenderForwardMobile::_update_shader_quality_settings() { + Vector<RD::PipelineSpecializationConstant> spec_constants; + + RD::PipelineSpecializationConstant sc; + sc.type = RD::PIPELINE_SPECIALIZATION_CONSTANT_TYPE_INT; + + sc.constant_id = SPEC_CONSTANT_SOFT_SHADOW_SAMPLES; + sc.int_value = soft_shadow_samples_get(); + + spec_constants.push_back(sc); + + sc.constant_id = SPEC_CONSTANT_PENUMBRA_SHADOW_SAMPLES; + sc.int_value = penumbra_shadow_samples_get(); + + spec_constants.push_back(sc); + + sc.constant_id = SPEC_CONSTANT_DIRECTIONAL_SOFT_SHADOW_SAMPLES; + sc.int_value = directional_soft_shadow_samples_get(); + + spec_constants.push_back(sc); + + sc.constant_id = SPEC_CONSTANT_DIRECTIONAL_PENUMBRA_SHADOW_SAMPLES; + sc.int_value = directional_penumbra_shadow_samples_get(); + + spec_constants.push_back(sc); + + sc.type = RD::PIPELINE_SPECIALIZATION_CONSTANT_TYPE_BOOL; + sc.constant_id = SPEC_CONSTANT_DECAL_USE_MIPMAPS; + sc.bool_value = decals_get_filter() == RS::DECAL_FILTER_NEAREST_MIPMAPS || + decals_get_filter() == RS::DECAL_FILTER_LINEAR_MIPMAPS || + decals_get_filter() == RS::DECAL_FILTER_NEAREST_MIPMAPS_ANISOTROPIC || + decals_get_filter() == RS::DECAL_FILTER_LINEAR_MIPMAPS_ANISOTROPIC; + + spec_constants.push_back(sc); + + sc.constant_id = SPEC_CONSTANT_PROJECTOR_USE_MIPMAPS; + sc.bool_value = light_projectors_get_filter() == RS::LIGHT_PROJECTOR_FILTER_NEAREST_MIPMAPS || + light_projectors_get_filter() == RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS || + light_projectors_get_filter() == RS::LIGHT_PROJECTOR_FILTER_NEAREST_MIPMAPS_ANISOTROPIC || + light_projectors_get_filter() == RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS_ANISOTROPIC; + + spec_constants.push_back(sc); + + scene_shader.set_default_specialization_constants(spec_constants); + + base_uniforms_changed(); //also need this +} + +RenderForwardMobile::RenderForwardMobile() { + singleton = this; + + sky.set_texture_format(_render_buffers_get_color_format()); + + String defines; + + defines += "\n#define MAX_ROUGHNESS_LOD " + itos(get_roughness_layers() - 1) + ".0\n"; + if (is_using_radiance_cubemap_array()) { + defines += "\n#define USE_RADIANCE_CUBEMAP_ARRAY \n"; + } + // defines += "\n#define SDFGI_OCT_SIZE " + itos(gi.sdfgi_get_lightprobe_octahedron_size()) + "\n"; + defines += "\n#define MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS " + itos(MAX_DIRECTIONAL_LIGHTS) + "\n"; + + { + //lightmaps + scene_state.max_lightmaps = 2; + defines += "\n#define MAX_LIGHTMAP_TEXTURES " + itos(scene_state.max_lightmaps) + "\n"; + defines += "\n#define MAX_LIGHTMAPS " + itos(scene_state.max_lightmaps) + "\n"; + + scene_state.lightmap_buffer = RD::get_singleton()->storage_buffer_create(sizeof(LightmapData) * scene_state.max_lightmaps); + } + { + //captures + scene_state.max_lightmap_captures = 2048; + scene_state.lightmap_captures = memnew_arr(LightmapCaptureData, scene_state.max_lightmap_captures); + scene_state.lightmap_capture_buffer = RD::get_singleton()->storage_buffer_create(sizeof(LightmapCaptureData) * scene_state.max_lightmap_captures); + } + { + defines += "\n#define MATERIAL_UNIFORM_SET " + itos(MATERIAL_UNIFORM_SET) + "\n"; + } +#ifdef REAL_T_IS_DOUBLE + { + defines += "\n#define USE_DOUBLE_PRECISION \n"; + } +#endif + + scene_shader.init(defines); + + // !BAS! maybe we need a mobile version of this setting? + render_list_thread_threshold = GLOBAL_GET("rendering/limits/forward_renderer/threaded_render_minimum_instances"); + + _update_shader_quality_settings(); +} + +RenderForwardMobile::~RenderForwardMobile() { + RSG::light_storage->directional_shadow_atlas_set_size(0); + + //clear base uniform set if still valid + for (uint32_t i = 0; i < render_pass_uniform_sets.size(); i++) { + if (render_pass_uniform_sets[i].is_valid() && RD::get_singleton()->uniform_set_is_valid(render_pass_uniform_sets[i])) { + RD::get_singleton()->free(render_pass_uniform_sets[i]); + } + } + + { + for (uint32_t i = 0; i < scene_state.uniform_buffers.size(); i++) { + RD::get_singleton()->free(scene_state.uniform_buffers[i]); + } + RD::get_singleton()->free(scene_state.lightmap_buffer); + RD::get_singleton()->free(scene_state.lightmap_capture_buffer); + memdelete_arr(scene_state.lightmap_captures); + } +} diff --git a/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.h b/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.h new file mode 100644 index 0000000000..ce64d805b7 --- /dev/null +++ b/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.h @@ -0,0 +1,590 @@ +/*************************************************************************/ +/* render_forward_mobile.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef RENDER_FORWARD_MOBILE_H +#define RENDER_FORWARD_MOBILE_H + +#include "core/templates/paged_allocator.h" +#include "servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.h" +#include "servers/rendering/renderer_rd/pipeline_cache_rd.h" +#include "servers/rendering/renderer_rd/renderer_scene_render_rd.h" +#include "servers/rendering/renderer_rd/storage_rd/utilities.h" + +#define RB_SCOPE_MOBILE SNAME("mobile") + +namespace RendererSceneRenderImplementation { + +class RenderForwardMobile : public RendererSceneRenderRD { + friend SceneShaderForwardMobile; + +protected: + struct GeometryInstanceSurfaceDataCache; + +private: + static RenderForwardMobile *singleton; + + /* Scene Shader */ + + enum { + SCENE_UNIFORM_SET = 0, + RENDER_PASS_UNIFORM_SET = 1, + TRANSFORMS_UNIFORM_SET = 2, + MATERIAL_UNIFORM_SET = 3 + }; + + enum { + + SPEC_CONSTANT_USING_PROJECTOR = 0, + SPEC_CONSTANT_USING_SOFT_SHADOWS = 1, + SPEC_CONSTANT_USING_DIRECTIONAL_SOFT_SHADOWS = 2, + + SPEC_CONSTANT_SOFT_SHADOW_SAMPLES = 3, + SPEC_CONSTANT_PENUMBRA_SHADOW_SAMPLES = 4, + SPEC_CONSTANT_DIRECTIONAL_SOFT_SHADOW_SAMPLES = 5, + SPEC_CONSTANT_DIRECTIONAL_PENUMBRA_SHADOW_SAMPLES = 6, + + SPEC_CONSTANT_DECAL_USE_MIPMAPS = 7, + SPEC_CONSTANT_PROJECTOR_USE_MIPMAPS = 8, + + SPEC_CONSTANT_DISABLE_OMNI_LIGHTS = 9, + SPEC_CONSTANT_DISABLE_SPOT_LIGHTS = 10, + SPEC_CONSTANT_DISABLE_REFLECTION_PROBES = 11, + SPEC_CONSTANT_DISABLE_DIRECTIONAL_LIGHTS = 12, + + SPEC_CONSTANT_DISABLE_DECALS = 13, + SPEC_CONSTANT_DISABLE_FOG = 14, + + }; + + enum { + MAX_LIGHTMAPS = 8, + MAX_RDL_CULL = 8, // maximum number of reflection probes, decals or lights we can cull per geometry instance + INSTANCE_DATA_BUFFER_MIN_SIZE = 4096 + }; + + enum RenderListType { + RENDER_LIST_OPAQUE, //used for opaque objects + RENDER_LIST_ALPHA, //used for transparent objects + RENDER_LIST_SECONDARY, //used for shadows and other objects + RENDER_LIST_MAX + }; + + /* Scene Shader */ + + SceneShaderForwardMobile scene_shader; + + /* Render Buffer */ + + class RenderBufferDataForwardMobile : public RenderBufferCustomDataRD { + GDCLASS(RenderBufferDataForwardMobile, RenderBufferCustomDataRD); + + public: + // We can have: + // - 4 subpasses combining the full render cycle + // - 3 subpasses + 1 normal pass for tonemapping/glow/dof/etc (using fb for 2D buffer) + // - 2 subpasses + 1 normal pass for transparent + 1 normal pass for tonemapping/glow/dof/etc (using fb for 2D buffer) + enum FramebufferConfigType { + FB_CONFIG_ONE_PASS, // Single pass frame buffer for alpha pass + FB_CONFIG_TWO_SUBPASSES, // Opaque + Sky sub pass + FB_CONFIG_THREE_SUBPASSES, // Opaque + Sky + Alpha sub pass + FB_CONFIG_FOUR_SUBPASSES, // Opaque + Sky + Alpha sub pass + Tonemap pass + FB_CONFIG_MAX + }; + + RID get_color_msaa() const { return render_buffers->get_texture(RB_SCOPE_MOBILE, RB_TEX_COLOR_MSAA); } + RID get_color_msaa(uint32_t p_layer) { return render_buffers->get_texture_slice(RB_SCOPE_MOBILE, RB_TEX_COLOR_MSAA, p_layer, 0); } + + RID get_depth_msaa() const { return render_buffers->get_texture(RB_SCOPE_MOBILE, RB_TEX_DEPTH_MSAA); } + RID get_depth_msaa(uint32_t p_layer) { return render_buffers->get_texture_slice(RB_SCOPE_MOBILE, RB_TEX_DEPTH_MSAA, p_layer, 0); } + + RID get_color_fbs(FramebufferConfigType p_config_type); + virtual void free_data() override; + virtual void configure(RenderSceneBuffersRD *p_render_buffers) override; + + private: + RenderSceneBuffersRD *render_buffers = nullptr; + RD::TextureSamples texture_samples = RD::TEXTURE_SAMPLES_1; + }; + + virtual void setup_render_buffer_data(Ref<RenderSceneBuffersRD> p_render_buffers) override; + + /* Rendering */ + + enum PassMode { + PASS_MODE_COLOR, + // PASS_MODE_COLOR_SPECULAR, + PASS_MODE_COLOR_TRANSPARENT, + PASS_MODE_SHADOW, + PASS_MODE_SHADOW_DP, + // PASS_MODE_DEPTH, + // PASS_MODE_DEPTH_NORMAL_ROUGHNESS, + // PASS_MODE_DEPTH_NORMAL_ROUGHNESS_VOXEL_GI, + PASS_MODE_DEPTH_MATERIAL, + // PASS_MODE_SDF, + }; + + struct RenderElementInfo; + + struct RenderListParameters { + GeometryInstanceSurfaceDataCache **elements = nullptr; + RenderElementInfo *element_info = nullptr; + int element_count = 0; + bool reverse_cull = false; + PassMode pass_mode = PASS_MODE_COLOR; + // bool no_gi = false; + uint32_t view_count = 1; + RID render_pass_uniform_set; + bool force_wireframe = false; + Vector2 uv_offset; + uint32_t spec_constant_base_flags = 0; + float lod_distance_multiplier = 0.0; + float screen_mesh_lod_threshold = 0.0; + RD::FramebufferFormatID framebuffer_format = 0; + uint32_t element_offset = 0; + uint32_t barrier = RD::BARRIER_MASK_ALL; + uint32_t subpass = 0; + + RenderListParameters(GeometryInstanceSurfaceDataCache **p_elements, RenderElementInfo *p_element_info, int p_element_count, bool p_reverse_cull, PassMode p_pass_mode, RID p_render_pass_uniform_set, uint32_t p_spec_constant_base_flags = 0, bool p_force_wireframe = false, const Vector2 &p_uv_offset = Vector2(), float p_lod_distance_multiplier = 0.0, float p_screen_mesh_lod_threshold = 0.0, uint32_t p_view_count = 1, uint32_t p_element_offset = 0, uint32_t p_barrier = RD::BARRIER_MASK_ALL) { + elements = p_elements; + element_info = p_element_info; + element_count = p_element_count; + reverse_cull = p_reverse_cull; + pass_mode = p_pass_mode; + // no_gi = p_no_gi; + view_count = p_view_count; + render_pass_uniform_set = p_render_pass_uniform_set; + force_wireframe = p_force_wireframe; + uv_offset = p_uv_offset; + lod_distance_multiplier = p_lod_distance_multiplier; + screen_mesh_lod_threshold = p_screen_mesh_lod_threshold; + element_offset = p_element_offset; + barrier = p_barrier; + spec_constant_base_flags = p_spec_constant_base_flags; + } + }; + + /* Render shadows */ + + void _render_shadow_pass(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<RenderGeometryInstance *> &p_instances, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0, float p_screen_mesh_lod_threshold = 0.0, bool p_open_pass = true, bool p_close_pass = true, bool p_clear_region = true, RenderingMethod::RenderInfo *p_render_info = nullptr); + void _render_shadow_begin(); + void _render_shadow_append(RID p_framebuffer, const PagedArray<RenderGeometryInstance *> &p_instances, const Projection &p_projection, const Transform3D &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_mesh_lod_threshold = 0.0, const Rect2i &p_rect = Rect2i(), bool p_flip_y = false, bool p_clear_region = true, bool p_begin = true, bool p_end = true, RenderingMethod::RenderInfo *p_render_info = nullptr); + void _render_shadow_process(); + void _render_shadow_end(uint32_t p_barrier = RD::BARRIER_MASK_ALL); + + /* Render Scene */ + + RID _setup_render_pass_uniform_set(RenderListType p_render_list, const RenderDataRD *p_render_data, RID p_radiance_texture, bool p_use_directional_shadow_atlas = false, int p_index = 0); + void _pre_opaque_render(RenderDataRD *p_render_data); + + uint64_t lightmap_texture_array_version = 0xFFFFFFFF; + + void _update_render_base_uniform_set(); + + void _fill_render_list(RenderListType p_render_list, const RenderDataRD *p_render_data, PassMode p_pass_mode, bool p_append = false); + void _fill_element_info(RenderListType p_render_list, uint32_t p_offset = 0, int32_t p_max_elements = -1); + // void _update_instance_data_buffer(RenderListType p_render_list); + + void _setup_environment(const RenderDataRD *p_render_data, bool p_no_fog, const Size2i &p_screen_size, bool p_flip_y, const Color &p_default_bg_color, bool p_opaque_render_buffers = false, bool p_pancake_shadows = false, int p_index = 0); + void _setup_lightmaps(const RenderDataRD *p_render_data, const PagedArray<RID> &p_lightmaps, const Transform3D &p_cam_transform); + + RID render_base_uniform_set; + LocalVector<RID> render_pass_uniform_sets; + + /* Light map */ + + struct LightmapData { + float normal_xform[12]; + float pad[3]; + float exposure_normalization; + }; + + struct LightmapCaptureData { + float sh[9 * 4]; + }; + + /* Scene state */ + + struct SceneState { + LocalVector<RID> uniform_buffers; + + // !BAS! We need to change lightmaps, we're not going to do this with a buffer but pushing the used lightmap in + LightmapData lightmaps[MAX_LIGHTMAPS]; + RID lightmap_ids[MAX_LIGHTMAPS]; + bool lightmap_has_sh[MAX_LIGHTMAPS]; + uint32_t lightmaps_used = 0; + uint32_t max_lightmaps; + RID lightmap_buffer; + + LightmapCaptureData *lightmap_captures = nullptr; + uint32_t max_lightmap_captures; + RID lightmap_capture_buffer; + + bool used_screen_texture = false; + bool used_normal_texture = false; + bool used_depth_texture = false; + bool used_sss = false; + + struct ShadowPass { + uint32_t element_from; + uint32_t element_count; + bool flip_cull; + PassMode pass_mode; + + RID rp_uniform_set; + Plane camera_plane; + float lod_distance_multiplier; + float screen_mesh_lod_threshold; + + RID framebuffer; + RD::InitialAction initial_depth_action; + RD::FinalAction final_depth_action; + Rect2i rect; + }; + + LocalVector<ShadowPass> shadow_passes; + } scene_state; + + /* Render List */ + + // !BAS! Render list can probably be reused between clustered and mobile? + struct RenderList { + LocalVector<GeometryInstanceSurfaceDataCache *> elements; + LocalVector<RenderElementInfo> element_info; + + void clear() { + elements.clear(); + element_info.clear(); + } + + //should eventually be replaced by radix + + struct SortByKey { + _FORCE_INLINE_ bool operator()(const GeometryInstanceSurfaceDataCache *A, const GeometryInstanceSurfaceDataCache *B) const { + return (A->sort.sort_key2 == B->sort.sort_key2) ? (A->sort.sort_key1 < B->sort.sort_key1) : (A->sort.sort_key2 < B->sort.sort_key2); + } + }; + + void sort_by_key() { + SortArray<GeometryInstanceSurfaceDataCache *, SortByKey> sorter; + sorter.sort(elements.ptr(), elements.size()); + } + + void sort_by_key_range(uint32_t p_from, uint32_t p_size) { + SortArray<GeometryInstanceSurfaceDataCache *, SortByKey> sorter; + sorter.sort(elements.ptr() + p_from, p_size); + } + + struct SortByDepth { + _FORCE_INLINE_ bool operator()(const GeometryInstanceSurfaceDataCache *A, const GeometryInstanceSurfaceDataCache *B) const { + return (A->owner->depth < B->owner->depth); + } + }; + + void sort_by_depth() { //used for shadows + + SortArray<GeometryInstanceSurfaceDataCache *, SortByDepth> sorter; + sorter.sort(elements.ptr(), elements.size()); + } + + struct SortByReverseDepthAndPriority { + _FORCE_INLINE_ bool operator()(const GeometryInstanceSurfaceDataCache *A, const GeometryInstanceSurfaceDataCache *B) const { + return (A->sort.priority == B->sort.priority) ? (A->owner->depth > B->owner->depth) : (A->sort.priority < B->sort.priority); + } + }; + + void sort_by_reverse_depth_and_priority() { //used for alpha + + SortArray<GeometryInstanceSurfaceDataCache *, SortByReverseDepthAndPriority> sorter; + sorter.sort(elements.ptr(), elements.size()); + } + + _FORCE_INLINE_ void add_element(GeometryInstanceSurfaceDataCache *p_element) { + elements.push_back(p_element); + } + }; + + struct RenderElementInfo { + uint32_t uses_lightmap : 1; + uint32_t lod_index : 8; + uint32_t reserved : 23; + }; + + template <PassMode p_pass_mode> + _FORCE_INLINE_ void _render_list_template(RenderingDevice::DrawListID p_draw_list, RenderingDevice::FramebufferFormatID p_framebuffer_Format, RenderListParameters *p_params, uint32_t p_from_element, uint32_t p_to_element); + + void _render_list(RenderingDevice::DrawListID p_draw_list, RenderingDevice::FramebufferFormatID p_framebuffer_Format, RenderListParameters *p_params, uint32_t p_from_element, uint32_t p_to_element); + + LocalVector<RD::DrawListID> thread_draw_lists; + void _render_list_thread_function(uint32_t p_thread, RenderListParameters *p_params); + void _render_list_with_threads(RenderListParameters *p_params, RID p_framebuffer, RD::InitialAction p_initial_color_action, RD::FinalAction p_final_color_action, RD::InitialAction p_initial_depth_action, RD::FinalAction p_final_depth_action, const Vector<Color> &p_clear_color_values = Vector<Color>(), float p_clear_depth = 1.0, uint32_t p_clear_stencil = 0, const Rect2 &p_region = Rect2(), const Vector<RID> &p_storage_textures = Vector<RID>()); + + uint32_t render_list_thread_threshold = 500; + + RenderList render_list[RENDER_LIST_MAX]; + +protected: + /* setup */ + virtual void _update_shader_quality_settings() override; + + virtual float _render_buffers_get_luminance_multiplier() override; + virtual RD::DataFormat _render_buffers_get_color_format() override; + virtual bool _render_buffers_can_be_storage() override; + + virtual RID _render_buffers_get_normal_texture(Ref<RenderSceneBuffersRD> p_render_buffers) override; + virtual RID _render_buffers_get_velocity_texture(Ref<RenderSceneBuffersRD> p_render_buffers) override; + + virtual void environment_set_ssao_quality(RS::EnvironmentSSAOQuality p_quality, bool p_half_size, float p_adaptive_target, int p_blur_passes, float p_fadeout_from, float p_fadeout_to) override{}; + virtual void environment_set_ssil_quality(RS::EnvironmentSSILQuality p_quality, bool p_half_size, float p_adaptive_target, int p_blur_passes, float p_fadeout_from, float p_fadeout_to) override{}; + virtual void environment_set_ssr_roughness_quality(RS::EnvironmentSSRRoughnessQuality p_quality) override{}; + + virtual void sub_surface_scattering_set_quality(RS::SubSurfaceScatteringQuality p_quality) override{}; + virtual void sub_surface_scattering_set_scale(float p_scale, float p_depth_scale) override{}; + + /* Geometry instance */ + + class GeometryInstanceForwardMobile; + + // When changing any of these enums, remember to change the corresponding enums in the shader files as well. + enum { + INSTANCE_DATA_FLAGS_NON_UNIFORM_SCALE = 1 << 4, + INSTANCE_DATA_FLAG_USE_GI_BUFFERS = 1 << 5, + INSTANCE_DATA_FLAG_USE_SDFGI = 1 << 6, + INSTANCE_DATA_FLAG_USE_LIGHTMAP_CAPTURE = 1 << 7, + INSTANCE_DATA_FLAG_USE_LIGHTMAP = 1 << 8, + INSTANCE_DATA_FLAG_USE_SH_LIGHTMAP = 1 << 9, + INSTANCE_DATA_FLAG_USE_VOXEL_GI = 1 << 10, + INSTANCE_DATA_FLAG_PARTICLES = 1 << 11, + INSTANCE_DATA_FLAG_MULTIMESH = 1 << 12, + INSTANCE_DATA_FLAG_MULTIMESH_FORMAT_2D = 1 << 13, + INSTANCE_DATA_FLAG_MULTIMESH_HAS_COLOR = 1 << 14, + INSTANCE_DATA_FLAG_MULTIMESH_HAS_CUSTOM_DATA = 1 << 15, + INSTANCE_DATA_FLAGS_PARTICLE_TRAIL_SHIFT = 16, + INSTANCE_DATA_FLAGS_PARTICLE_TRAIL_MASK = 0xFF, + }; + + struct GeometryInstanceLightmapSH { + Color sh[9]; + }; + + // Cached data for drawing surfaces + struct GeometryInstanceSurfaceDataCache { + enum { + FLAG_PASS_DEPTH = 1, + FLAG_PASS_OPAQUE = 2, + FLAG_PASS_ALPHA = 4, + FLAG_PASS_SHADOW = 8, + FLAG_USES_SHARED_SHADOW_MATERIAL = 128, + FLAG_USES_SUBSURFACE_SCATTERING = 2048, + FLAG_USES_SCREEN_TEXTURE = 4096, + FLAG_USES_DEPTH_TEXTURE = 8192, + FLAG_USES_NORMAL_TEXTURE = 16384, + FLAG_USES_DOUBLE_SIDED_SHADOWS = 32768, + FLAG_USES_PARTICLE_TRAILS = 65536, + }; + + union { + struct { + // !BAS! CHECK BITS!!! + + uint64_t surface_index : 10; + uint64_t geometry_id : 32; + uint64_t material_id_low : 16; + + uint64_t material_id_hi : 16; + uint64_t shader_id : 32; + uint64_t uses_lightmap : 4; // sort by lightmap id here, not whether its yes/no (is 4 bits enough?) + uint64_t depth_layer : 4; + uint64_t priority : 8; + + // uint64_t lod_index : 8; // no need to sort on LOD + // uint64_t uses_forward_gi : 1; // no GI here, remove + }; + struct { + uint64_t sort_key1; + uint64_t sort_key2; + }; + } sort; + + RS::PrimitiveType primitive = RS::PRIMITIVE_MAX; + uint32_t flags = 0; + uint32_t surface_index = 0; + uint32_t lod_index = 0; + + void *surface = nullptr; + RID material_uniform_set; + SceneShaderForwardMobile::ShaderData *shader = nullptr; + + void *surface_shadow = nullptr; + RID material_uniform_set_shadow; + SceneShaderForwardMobile::ShaderData *shader_shadow = nullptr; + + GeometryInstanceSurfaceDataCache *next = nullptr; + GeometryInstanceForwardMobile *owner = nullptr; + }; + + class GeometryInstanceForwardMobile : public RenderGeometryInstanceBase { + public: + // this structure maps to our push constant in our shader and is populated right before our draw call + struct PushConstant { + float transform[16]; + uint32_t flags; + uint32_t instance_uniforms_ofs; //base offset in global buffer for instance variables + uint32_t gi_offset; //GI information when using lightmapping (VCT or lightmap index) + uint32_t layer_mask = 1; + float lightmap_uv_scale[4]; // doubles as uv_offset when needed + uint32_t reflection_probes[2]; // packed reflection probes + uint32_t omni_lights[2]; // packed omni lights + uint32_t spot_lights[2]; // packed spot lights + uint32_t decals[2]; // packed spot lights + }; + + // PushConstant push_constant; // we populate this from our instance data + + //used during rendering + RID transforms_uniform_set; + bool use_projector = false; + bool use_soft_shadow = false; + bool store_transform_cache = true; // if true we copy our transform into our PushConstant, if false we use our transforms UBO and clear our PushConstants transform + uint32_t instance_count = 0; + uint32_t trail_steps = 1; + + // lightmap + uint32_t gi_offset_cache = 0; // !BAS! Should rename this to lightmap_offset_cache, in forward clustered this was shared between gi and lightmap + RID lightmap_instance; + Rect2 lightmap_uv_scale; + uint32_t lightmap_slice_index; + GeometryInstanceLightmapSH *lightmap_sh = nullptr; + + // culled light info + uint32_t reflection_probe_count = 0; + RendererRD::ForwardID reflection_probes[MAX_RDL_CULL]; + uint32_t omni_light_count = 0; + RendererRD::ForwardID omni_lights[MAX_RDL_CULL]; + uint32_t spot_light_count = 0; + RendererRD::ForwardID spot_lights[MAX_RDL_CULL]; + uint32_t decals_count = 0; + RendererRD::ForwardID decals[MAX_RDL_CULL]; + + GeometryInstanceSurfaceDataCache *surface_caches = nullptr; + + // do we use this? + SelfList<GeometryInstanceForwardMobile> dirty_list_element; + + GeometryInstanceForwardMobile() : + dirty_list_element(this) {} + + virtual void _mark_dirty() override; + + virtual void set_use_lightmap(RID p_lightmap_instance, const Rect2 &p_lightmap_uv_scale, int p_lightmap_slice_index) override; + virtual void set_lightmap_capture(const Color *p_sh9) override; + + virtual void pair_light_instances(const RID *p_light_instances, uint32_t p_light_instance_count) override; + virtual void pair_reflection_probe_instances(const RID *p_reflection_probe_instances, uint32_t p_reflection_probe_instance_count) override; + virtual void pair_decal_instances(const RID *p_decal_instances, uint32_t p_decal_instance_count) override; + virtual void pair_voxel_gi_instances(const RID *p_voxel_gi_instances, uint32_t p_voxel_gi_instance_count) override {} + + virtual void set_softshadow_projector_pairing(bool p_softshadow, bool p_projector) override; + }; + + /* Rendering */ + + virtual void _render_scene(RenderDataRD *p_render_data, const Color &p_default_bg_color) override; + + virtual void _render_material(const Transform3D &p_cam_transform, const Projection &p_cam_projection, bool p_cam_orthogonal, const PagedArray<RenderGeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region, float p_exposure_normalization) override; + virtual void _render_uv2(const PagedArray<RenderGeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) override; + virtual void _render_sdfgi(Ref<RenderSceneBuffersRD> p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<RenderGeometryInstance *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture, float p_exposure_normalization) override; + virtual void _render_particle_collider_heightfield(RID p_fb, const Transform3D &p_cam_transform, const Projection &p_cam_projection, const PagedArray<RenderGeometryInstance *> &p_instances) override; + + /* Forward ID */ + + class ForwardIDStorageMobile : public RendererRD::ForwardIDStorage { + public: + struct ForwardIDAllocator { + LocalVector<bool> allocations; + LocalVector<uint8_t> map; + }; + + ForwardIDAllocator forward_id_allocators[RendererRD::FORWARD_ID_MAX]; + + public: + virtual RendererRD::ForwardID allocate_forward_id(RendererRD::ForwardIDType p_type) override; + virtual void free_forward_id(RendererRD::ForwardIDType p_type, RendererRD::ForwardID p_id) override; + virtual void map_forward_id(RendererRD::ForwardIDType p_type, RendererRD::ForwardID p_id, uint32_t p_index) override; + virtual bool uses_forward_ids() const override { return true; } + + void fill_push_constant_instance_indices(GeometryInstanceForwardMobile::PushConstant *p_push_constant, uint32_t &spec_constants, const GeometryInstanceForwardMobile *p_instance); + }; + + ForwardIDStorageMobile *forward_id_storage_mobile = nullptr; + + virtual RendererRD::ForwardIDStorage *create_forward_id_storage() override { + forward_id_storage_mobile = memnew(ForwardIDStorageMobile); + return forward_id_storage_mobile; + } + +public: + static RenderForwardMobile *get_singleton() { return singleton; } + + virtual RID reflection_probe_create_framebuffer(RID p_color, RID p_depth) override; + + static void _geometry_instance_dependency_changed(Dependency::DependencyChangedNotification p_notification, DependencyTracker *p_tracker); + static void _geometry_instance_dependency_deleted(const RID &p_dependency, DependencyTracker *p_tracker); + + SelfList<GeometryInstanceForwardMobile>::List geometry_instance_dirty_list; + + PagedAllocator<GeometryInstanceForwardMobile> geometry_instance_alloc; + PagedAllocator<GeometryInstanceSurfaceDataCache> geometry_instance_surface_alloc; + PagedAllocator<GeometryInstanceLightmapSH> geometry_instance_lightmap_sh; + + void _geometry_instance_add_surface_with_material(GeometryInstanceForwardMobile *ginstance, uint32_t p_surface, SceneShaderForwardMobile::MaterialData *p_material, uint32_t p_material_id, uint32_t p_shader_id, RID p_mesh); + void _geometry_instance_add_surface_with_material_chain(GeometryInstanceForwardMobile *ginstance, uint32_t p_surface, SceneShaderForwardMobile::MaterialData *p_material, RID p_mat_src, RID p_mesh); + void _geometry_instance_add_surface(GeometryInstanceForwardMobile *ginstance, uint32_t p_surface, RID p_material, RID p_mesh); + void _geometry_instance_update(RenderGeometryInstance *p_geometry_instance); + void _update_dirty_geometry_instances(); + + virtual RenderGeometryInstance *geometry_instance_create(RID p_base) override; + virtual void geometry_instance_free(RenderGeometryInstance *p_geometry_instance) override; + + virtual uint32_t geometry_instance_get_pair_mask() override; + + virtual bool free(RID p_rid) override; + + virtual void base_uniforms_changed() override; + + virtual bool is_dynamic_gi_supported() const override; + virtual bool is_volumetric_supported() const override; + virtual uint32_t get_max_elements() const override; + + RenderForwardMobile(); + ~RenderForwardMobile(); +}; +} // namespace RendererSceneRenderImplementation + +#endif // RENDER_FORWARD_MOBILE_H diff --git a/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.cpp b/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.cpp new file mode 100644 index 0000000000..02bd30d32d --- /dev/null +++ b/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.cpp @@ -0,0 +1,805 @@ +/*************************************************************************/ +/* scene_shader_forward_mobile.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "scene_shader_forward_mobile.h" +#include "core/config/project_settings.h" +#include "core/math/math_defs.h" +#include "render_forward_mobile.h" +#include "servers/rendering/renderer_rd/renderer_compositor_rd.h" +#include "servers/rendering/renderer_rd/storage_rd/material_storage.h" + +using namespace RendererSceneRenderImplementation; + +/* ShaderData */ + +void SceneShaderForwardMobile::ShaderData::set_path_hint(const String &p_path) { + path = p_path; +} + +void SceneShaderForwardMobile::ShaderData::set_code(const String &p_code) { + //compile + + code = p_code; + valid = false; + ubo_size = 0; + uniforms.clear(); + uses_screen_texture = false; + + if (code.is_empty()) { + return; //just invalid, but no error + } + + ShaderCompiler::GeneratedCode gen_code; + + int blend_mode = BLEND_MODE_MIX; + int depth_testi = DEPTH_TEST_ENABLED; + int alpha_antialiasing_mode = ALPHA_ANTIALIASING_OFF; + int cull = CULL_BACK; + + uses_point_size = false; + uses_alpha = false; + uses_alpha_clip = false; + uses_blend_alpha = false; + uses_depth_pre_pass = false; + uses_discard = false; + uses_roughness = false; + uses_normal = false; + bool wireframe = false; + + unshaded = false; + uses_vertex = false; + uses_sss = false; + uses_transmittance = false; + uses_screen_texture = false; + uses_depth_texture = false; + uses_normal_texture = false; + uses_time = false; + writes_modelview_or_projection = false; + uses_world_coordinates = false; + uses_particle_trails = false; + + int depth_drawi = DEPTH_DRAW_OPAQUE; + + ShaderCompiler::IdentifierActions actions; + actions.entry_point_stages["vertex"] = ShaderCompiler::STAGE_VERTEX; + actions.entry_point_stages["fragment"] = ShaderCompiler::STAGE_FRAGMENT; + actions.entry_point_stages["light"] = ShaderCompiler::STAGE_FRAGMENT; + + actions.render_mode_values["blend_add"] = Pair<int *, int>(&blend_mode, BLEND_MODE_ADD); + actions.render_mode_values["blend_mix"] = Pair<int *, int>(&blend_mode, BLEND_MODE_MIX); + actions.render_mode_values["blend_sub"] = Pair<int *, int>(&blend_mode, BLEND_MODE_SUB); + actions.render_mode_values["blend_mul"] = Pair<int *, int>(&blend_mode, BLEND_MODE_MUL); + + actions.render_mode_values["alpha_to_coverage"] = Pair<int *, int>(&alpha_antialiasing_mode, ALPHA_ANTIALIASING_ALPHA_TO_COVERAGE); + actions.render_mode_values["alpha_to_coverage_and_one"] = Pair<int *, int>(&alpha_antialiasing_mode, ALPHA_ANTIALIASING_ALPHA_TO_COVERAGE_AND_TO_ONE); + + actions.render_mode_values["depth_draw_never"] = Pair<int *, int>(&depth_drawi, DEPTH_DRAW_DISABLED); + actions.render_mode_values["depth_draw_opaque"] = Pair<int *, int>(&depth_drawi, DEPTH_DRAW_OPAQUE); + actions.render_mode_values["depth_draw_always"] = Pair<int *, int>(&depth_drawi, DEPTH_DRAW_ALWAYS); + + actions.render_mode_values["depth_test_disabled"] = Pair<int *, int>(&depth_testi, DEPTH_TEST_DISABLED); + + actions.render_mode_values["cull_disabled"] = Pair<int *, int>(&cull, CULL_DISABLED); + actions.render_mode_values["cull_front"] = Pair<int *, int>(&cull, CULL_FRONT); + actions.render_mode_values["cull_back"] = Pair<int *, int>(&cull, CULL_BACK); + + actions.render_mode_flags["unshaded"] = &unshaded; + actions.render_mode_flags["wireframe"] = &wireframe; + actions.render_mode_flags["particle_trails"] = &uses_particle_trails; + + actions.usage_flag_pointers["ALPHA"] = &uses_alpha; + actions.usage_flag_pointers["ALPHA_SCISSOR_THRESHOLD"] = &uses_alpha_clip; + actions.render_mode_flags["depth_prepass_alpha"] = &uses_depth_pre_pass; + + // actions.usage_flag_pointers["SSS_STRENGTH"] = &uses_sss; + // actions.usage_flag_pointers["SSS_TRANSMITTANCE_DEPTH"] = &uses_transmittance; + + actions.usage_flag_pointers["SCREEN_TEXTURE"] = &uses_screen_texture; + actions.usage_flag_pointers["DEPTH_TEXTURE"] = &uses_depth_texture; + actions.usage_flag_pointers["NORMAL_TEXTURE"] = &uses_normal_texture; + actions.usage_flag_pointers["DISCARD"] = &uses_discard; + actions.usage_flag_pointers["TIME"] = &uses_time; + actions.usage_flag_pointers["ROUGHNESS"] = &uses_roughness; + actions.usage_flag_pointers["NORMAL"] = &uses_normal; + actions.usage_flag_pointers["NORMAL_MAP"] = &uses_normal; + + actions.usage_flag_pointers["POINT_SIZE"] = &uses_point_size; + actions.usage_flag_pointers["POINT_COORD"] = &uses_point_size; + + actions.write_flag_pointers["MODELVIEW_MATRIX"] = &writes_modelview_or_projection; + actions.write_flag_pointers["PROJECTION_MATRIX"] = &writes_modelview_or_projection; + actions.write_flag_pointers["VERTEX"] = &uses_vertex; + + actions.uniforms = &uniforms; + + SceneShaderForwardMobile *shader_singleton = (SceneShaderForwardMobile *)SceneShaderForwardMobile::singleton; + + Error err = shader_singleton->compiler.compile(RS::SHADER_SPATIAL, code, &actions, path, gen_code); + ERR_FAIL_COND_MSG(err != OK, "Shader compilation failed."); + + if (version.is_null()) { + version = shader_singleton->shader.version_create(); + } + + depth_draw = DepthDraw(depth_drawi); + depth_test = DepthTest(depth_testi); + +#if 0 + print_line("**compiling shader:"); + print_line("**defines:\n"); + for (int i = 0; i < gen_code.defines.size(); i++) { + print_line(gen_code.defines[i]); + } + + HashMap<String, String>::Iterator el = gen_code.code.begin(); + while (el) { + print_line("\n**code " + el->key + ":\n" + el->value); + ++el; + } + + print_line("\n**uniforms:\n" + gen_code.uniforms); + print_line("\n**vertex_globals:\n" + gen_code.stage_globals[ShaderCompiler::STAGE_VERTEX]); + print_line("\n**fragment_globals:\n" + gen_code.stage_globals[ShaderCompiler::STAGE_FRAGMENT]); +#endif + + shader_singleton->shader.version_set_code(version, gen_code.code, gen_code.uniforms, gen_code.stage_globals[ShaderCompiler::STAGE_VERTEX], gen_code.stage_globals[ShaderCompiler::STAGE_FRAGMENT], gen_code.defines); + ERR_FAIL_COND(!shader_singleton->shader.version_is_valid(version)); + + ubo_size = gen_code.uniform_total_size; + ubo_offsets = gen_code.uniform_offsets; + texture_uniforms = gen_code.texture_uniforms; + + //blend modes + + // if any form of Alpha Antialiasing is enabled, set the blend mode to alpha to coverage + if (alpha_antialiasing_mode != ALPHA_ANTIALIASING_OFF) { + blend_mode = BLEND_MODE_ALPHA_TO_COVERAGE; + } + + RD::PipelineColorBlendState::Attachment blend_attachment; + + switch (blend_mode) { + case BLEND_MODE_MIX: { + blend_attachment.enable_blend = true; + blend_attachment.alpha_blend_op = RD::BLEND_OP_ADD; + blend_attachment.color_blend_op = RD::BLEND_OP_ADD; + blend_attachment.src_color_blend_factor = RD::BLEND_FACTOR_SRC_ALPHA; + blend_attachment.dst_color_blend_factor = RD::BLEND_FACTOR_ONE_MINUS_SRC_ALPHA; + blend_attachment.src_alpha_blend_factor = RD::BLEND_FACTOR_ONE; + blend_attachment.dst_alpha_blend_factor = RD::BLEND_FACTOR_ONE_MINUS_SRC_ALPHA; + + } break; + case BLEND_MODE_ADD: { + blend_attachment.enable_blend = true; + blend_attachment.alpha_blend_op = RD::BLEND_OP_ADD; + blend_attachment.color_blend_op = RD::BLEND_OP_ADD; + blend_attachment.src_color_blend_factor = RD::BLEND_FACTOR_SRC_ALPHA; + blend_attachment.dst_color_blend_factor = RD::BLEND_FACTOR_ONE; + blend_attachment.src_alpha_blend_factor = RD::BLEND_FACTOR_SRC_ALPHA; + blend_attachment.dst_alpha_blend_factor = RD::BLEND_FACTOR_ONE; + uses_blend_alpha = true; //force alpha used because of blend + + } break; + case BLEND_MODE_SUB: { + blend_attachment.enable_blend = true; + blend_attachment.alpha_blend_op = RD::BLEND_OP_SUBTRACT; + blend_attachment.color_blend_op = RD::BLEND_OP_SUBTRACT; + blend_attachment.src_color_blend_factor = RD::BLEND_FACTOR_SRC_ALPHA; + blend_attachment.dst_color_blend_factor = RD::BLEND_FACTOR_ONE; + blend_attachment.src_alpha_blend_factor = RD::BLEND_FACTOR_SRC_ALPHA; + blend_attachment.dst_alpha_blend_factor = RD::BLEND_FACTOR_ONE; + uses_blend_alpha = true; //force alpha used because of blend + + } break; + case BLEND_MODE_MUL: { + blend_attachment.enable_blend = true; + blend_attachment.alpha_blend_op = RD::BLEND_OP_ADD; + blend_attachment.color_blend_op = RD::BLEND_OP_ADD; + blend_attachment.src_color_blend_factor = RD::BLEND_FACTOR_DST_COLOR; + blend_attachment.dst_color_blend_factor = RD::BLEND_FACTOR_ZERO; + blend_attachment.src_alpha_blend_factor = RD::BLEND_FACTOR_DST_ALPHA; + blend_attachment.dst_alpha_blend_factor = RD::BLEND_FACTOR_ZERO; + uses_blend_alpha = true; //force alpha used because of blend + } break; + case BLEND_MODE_ALPHA_TO_COVERAGE: { + blend_attachment.enable_blend = true; + blend_attachment.alpha_blend_op = RD::BLEND_OP_ADD; + blend_attachment.color_blend_op = RD::BLEND_OP_ADD; + blend_attachment.src_color_blend_factor = RD::BLEND_FACTOR_SRC_ALPHA; + blend_attachment.dst_color_blend_factor = RD::BLEND_FACTOR_ONE_MINUS_SRC_ALPHA; + blend_attachment.src_alpha_blend_factor = RD::BLEND_FACTOR_ONE; + blend_attachment.dst_alpha_blend_factor = RD::BLEND_FACTOR_ZERO; + } + } + + RD::PipelineColorBlendState blend_state_blend; + blend_state_blend.attachments.push_back(blend_attachment); + RD::PipelineColorBlendState blend_state_opaque = RD::PipelineColorBlendState::create_disabled(1); + RD::PipelineColorBlendState blend_state_opaque_specular = RD::PipelineColorBlendState::create_disabled(2); + RD::PipelineColorBlendState blend_state_depth_normal_roughness = RD::PipelineColorBlendState::create_disabled(1); + RD::PipelineColorBlendState blend_state_depth_normal_roughness_giprobe = RD::PipelineColorBlendState::create_disabled(2); + + //update pipelines + + RD::PipelineDepthStencilState depth_stencil_state; + + if (depth_test != DEPTH_TEST_DISABLED) { + depth_stencil_state.enable_depth_test = true; + depth_stencil_state.depth_compare_operator = RD::COMPARE_OP_LESS_OR_EQUAL; + depth_stencil_state.enable_depth_write = depth_draw != DEPTH_DRAW_DISABLED ? true : false; + } + + for (int i = 0; i < CULL_VARIANT_MAX; i++) { + RD::PolygonCullMode cull_mode_rd_table[CULL_VARIANT_MAX][3] = { + { RD::POLYGON_CULL_DISABLED, RD::POLYGON_CULL_FRONT, RD::POLYGON_CULL_BACK }, + { RD::POLYGON_CULL_DISABLED, RD::POLYGON_CULL_BACK, RD::POLYGON_CULL_FRONT }, + { RD::POLYGON_CULL_DISABLED, RD::POLYGON_CULL_DISABLED, RD::POLYGON_CULL_DISABLED } + }; + + RD::PolygonCullMode cull_mode_rd = cull_mode_rd_table[i][cull]; + + for (int j = 0; j < RS::PRIMITIVE_MAX; j++) { + RD::RenderPrimitive primitive_rd_table[RS::PRIMITIVE_MAX] = { + RD::RENDER_PRIMITIVE_POINTS, + RD::RENDER_PRIMITIVE_LINES, + RD::RENDER_PRIMITIVE_LINESTRIPS, + RD::RENDER_PRIMITIVE_TRIANGLES, + RD::RENDER_PRIMITIVE_TRIANGLE_STRIPS, + }; + + RD::RenderPrimitive primitive_rd = uses_point_size ? RD::RENDER_PRIMITIVE_POINTS : primitive_rd_table[j]; + + for (int k = 0; k < SHADER_VERSION_MAX; k++) { + if (!static_cast<SceneShaderForwardMobile *>(singleton)->shader.is_variant_enabled(k)) { + continue; + } + RD::PipelineRasterizationState raster_state; + raster_state.cull_mode = cull_mode_rd; + raster_state.wireframe = wireframe; + + RD::PipelineColorBlendState blend_state; + RD::PipelineDepthStencilState depth_stencil = depth_stencil_state; + RD::PipelineMultisampleState multisample_state; + + if (uses_alpha || uses_blend_alpha) { + // only allow these flags to go through if we have some form of msaa + if (alpha_antialiasing_mode == ALPHA_ANTIALIASING_ALPHA_TO_COVERAGE) { + multisample_state.enable_alpha_to_coverage = true; + } else if (alpha_antialiasing_mode == ALPHA_ANTIALIASING_ALPHA_TO_COVERAGE_AND_TO_ONE) { + multisample_state.enable_alpha_to_coverage = true; + multisample_state.enable_alpha_to_one = true; + } + + if (k == SHADER_VERSION_COLOR_PASS || k == SHADER_VERSION_COLOR_PASS_MULTIVIEW || k == SHADER_VERSION_LIGHTMAP_COLOR_PASS || k == SHADER_VERSION_LIGHTMAP_COLOR_PASS_MULTIVIEW) { + blend_state = blend_state_blend; + if (depth_draw == DEPTH_DRAW_OPAQUE && !uses_alpha_clip) { + depth_stencil.enable_depth_write = false; //alpha does not draw depth + } + } else if (k == SHADER_VERSION_SHADOW_PASS || k == SHADER_VERSION_SHADOW_PASS_MULTIVIEW || k == SHADER_VERSION_SHADOW_PASS_DP) { + //none, blend state contains nothing + } else if (k == SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL) { + blend_state = RD::PipelineColorBlendState::create_disabled(5); //writes to normal and roughness in opaque way + } else { + pipelines[i][j][k].clear(); + continue; // do not use this version (will error if using it is attempted) + } + } else { + if (k == SHADER_VERSION_COLOR_PASS || k == SHADER_VERSION_COLOR_PASS_MULTIVIEW || k == SHADER_VERSION_LIGHTMAP_COLOR_PASS || k == SHADER_VERSION_LIGHTMAP_COLOR_PASS_MULTIVIEW) { + blend_state = blend_state_opaque; + } else if (k == SHADER_VERSION_SHADOW_PASS || k == SHADER_VERSION_SHADOW_PASS_MULTIVIEW || k == SHADER_VERSION_SHADOW_PASS_DP) { + //none, leave empty + } else if (k == SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL) { + blend_state = RD::PipelineColorBlendState::create_disabled(5); //writes to normal and roughness in opaque way + } else { + // ??? + } + } + + RID shader_variant = shader_singleton->shader.version_get_shader(version, k); + pipelines[i][j][k].setup(shader_variant, primitive_rd, raster_state, multisample_state, depth_stencil, blend_state, 0, singleton->default_specialization_constants); + } + } + } + + valid = true; +} + +void SceneShaderForwardMobile::ShaderData::set_default_texture_parameter(const StringName &p_name, RID p_texture, int p_index) { + if (!p_texture.is_valid()) { + if (default_texture_params.has(p_name) && default_texture_params[p_name].has(p_index)) { + default_texture_params[p_name].erase(p_index); + + if (default_texture_params[p_name].is_empty()) { + default_texture_params.erase(p_name); + } + } + } else { + if (!default_texture_params.has(p_name)) { + default_texture_params[p_name] = HashMap<int, RID>(); + } + default_texture_params[p_name][p_index] = p_texture; + } +} + +void SceneShaderForwardMobile::ShaderData::get_shader_uniform_list(List<PropertyInfo> *p_param_list) const { + HashMap<int, StringName> order; + + for (const KeyValue<StringName, ShaderLanguage::ShaderNode::Uniform> &E : uniforms) { + if (E.value.scope != ShaderLanguage::ShaderNode::Uniform::SCOPE_LOCAL || + E.value.hint == ShaderLanguage::ShaderNode::Uniform::HINT_SCREEN_TEXTURE || + E.value.hint == ShaderLanguage::ShaderNode::Uniform::HINT_NORMAL_ROUGHNESS_TEXTURE || + E.value.hint == ShaderLanguage::ShaderNode::Uniform::HINT_DEPTH_TEXTURE) { + continue; + } + + if (E.value.texture_order >= 0) { + order[E.value.texture_order + 100000] = E.key; + } else { + order[E.value.order] = E.key; + } + } + + String last_group; + for (const KeyValue<int, StringName> &E : order) { + String group = uniforms[E.value].group; + if (!uniforms[E.value].subgroup.is_empty()) { + group += "::" + uniforms[E.value].subgroup; + } + + if (group != last_group) { + PropertyInfo pi; + pi.usage = PROPERTY_USAGE_GROUP; + pi.name = group; + p_param_list->push_back(pi); + + last_group = group; + } + + PropertyInfo pi = ShaderLanguage::uniform_to_property_info(uniforms[E.value]); + pi.name = E.value; + p_param_list->push_back(pi); + } +} + +void SceneShaderForwardMobile::ShaderData::get_instance_param_list(List<RendererMaterialStorage::InstanceShaderParam> *p_param_list) const { + for (const KeyValue<StringName, ShaderLanguage::ShaderNode::Uniform> &E : uniforms) { + if (E.value.scope != ShaderLanguage::ShaderNode::Uniform::SCOPE_INSTANCE) { + continue; + } + + RendererMaterialStorage::InstanceShaderParam p; + p.info = ShaderLanguage::uniform_to_property_info(E.value); + p.info.name = E.key; //supply name + p.index = E.value.instance_index; + p.default_value = ShaderLanguage::constant_value_to_variant(E.value.default_value, E.value.type, E.value.array_size, E.value.hint); + p_param_list->push_back(p); + } +} + +bool SceneShaderForwardMobile::ShaderData::is_parameter_texture(const StringName &p_param) const { + if (!uniforms.has(p_param)) { + return false; + } + + return uniforms[p_param].texture_order >= 0; +} + +bool SceneShaderForwardMobile::ShaderData::is_animated() const { + return false; +} + +bool SceneShaderForwardMobile::ShaderData::casts_shadows() const { + return false; +} + +Variant SceneShaderForwardMobile::ShaderData::get_default_parameter(const StringName &p_parameter) const { + if (uniforms.has(p_parameter)) { + ShaderLanguage::ShaderNode::Uniform uniform = uniforms[p_parameter]; + Vector<ShaderLanguage::ConstantNode::Value> default_value = uniform.default_value; + return ShaderLanguage::constant_value_to_variant(default_value, uniform.type, uniform.array_size, uniform.hint); + } + return Variant(); +} + +RS::ShaderNativeSourceCode SceneShaderForwardMobile::ShaderData::get_native_source_code() const { + SceneShaderForwardMobile *shader_singleton = (SceneShaderForwardMobile *)SceneShaderForwardMobile::singleton; + + return shader_singleton->shader.version_get_native_source_code(version); +} + +SceneShaderForwardMobile::ShaderData::ShaderData() : + shader_list_element(this) { +} + +SceneShaderForwardMobile::ShaderData::~ShaderData() { + SceneShaderForwardMobile *shader_singleton = (SceneShaderForwardMobile *)SceneShaderForwardMobile::singleton; + ERR_FAIL_COND(!shader_singleton); + //pipeline variants will clear themselves if shader is gone + if (version.is_valid()) { + shader_singleton->shader.version_free(version); + } +} + +RendererRD::MaterialStorage::ShaderData *SceneShaderForwardMobile::_create_shader_func() { + ShaderData *shader_data = memnew(ShaderData); + singleton->shader_list.add(&shader_data->shader_list_element); + return shader_data; +} + +void SceneShaderForwardMobile::MaterialData::set_render_priority(int p_priority) { + priority = p_priority - RS::MATERIAL_RENDER_PRIORITY_MIN; //8 bits +} + +void SceneShaderForwardMobile::MaterialData::set_next_pass(RID p_pass) { + next_pass = p_pass; +} + +bool SceneShaderForwardMobile::MaterialData::update_parameters(const HashMap<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { + SceneShaderForwardMobile *shader_singleton = (SceneShaderForwardMobile *)SceneShaderForwardMobile::singleton; + + return update_parameters_uniform_set(p_parameters, p_uniform_dirty, p_textures_dirty, shader_data->uniforms, shader_data->ubo_offsets.ptr(), shader_data->texture_uniforms, shader_data->default_texture_params, shader_data->ubo_size, uniform_set, shader_singleton->shader.version_get_shader(shader_data->version, 0), RenderForwardMobile::MATERIAL_UNIFORM_SET, true, RD::BARRIER_MASK_RASTER); +} + +SceneShaderForwardMobile::MaterialData::~MaterialData() { + free_parameters_uniform_set(uniform_set); +} + +RendererRD::MaterialStorage::MaterialData *SceneShaderForwardMobile::_create_material_func(ShaderData *p_shader) { + MaterialData *material_data = memnew(MaterialData); + material_data->shader_data = p_shader; + //update will happen later anyway so do nothing. + return material_data; +} + +/* Scene Shader */ + +SceneShaderForwardMobile *SceneShaderForwardMobile::singleton = nullptr; + +SceneShaderForwardMobile::SceneShaderForwardMobile() { + // there should be only one of these, contained within our RenderForwardMobile singleton. + singleton = this; +} + +void SceneShaderForwardMobile::init(const String p_defines) { + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + + /* SCENE SHADER */ + + { + Vector<String> shader_versions; + shader_versions.push_back(""); // SHADER_VERSION_COLOR_PASS + shader_versions.push_back("\n#define USE_LIGHTMAP\n"); // SHADER_VERSION_LIGHTMAP_COLOR_PASS + shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n"); // SHADER_VERSION_SHADOW_PASS, should probably change this to MODE_RENDER_SHADOW because we don't have a depth pass here... + shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_DUAL_PARABOLOID\n"); // SHADER_VERSION_SHADOW_PASS_DP + shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_RENDER_MATERIAL\n"); // SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL + + // multiview versions of our shaders + shader_versions.push_back("\n#define USE_MULTIVIEW\n"); // SHADER_VERSION_COLOR_PASS_MULTIVIEW + shader_versions.push_back("\n#define USE_MULTIVIEW\n#define USE_LIGHTMAP\n"); // SHADER_VERSION_LIGHTMAP_COLOR_PASS_MULTIVIEW + shader_versions.push_back("\n#define USE_MULTIVIEW\n#define MODE_RENDER_DEPTH\n"); // SHADER_VERSION_SHADOW_PASS_MULTIVIEW + + shader.initialize(shader_versions, p_defines); + + if (!RendererCompositorRD::singleton->is_xr_enabled()) { + shader.set_variant_enabled(SHADER_VERSION_COLOR_PASS_MULTIVIEW, false); + shader.set_variant_enabled(SHADER_VERSION_LIGHTMAP_COLOR_PASS_MULTIVIEW, false); + shader.set_variant_enabled(SHADER_VERSION_SHADOW_PASS_MULTIVIEW, false); + } + } + + material_storage->shader_set_data_request_function(RendererRD::MaterialStorage::SHADER_TYPE_3D, _create_shader_funcs); + material_storage->material_set_data_request_function(RendererRD::MaterialStorage::SHADER_TYPE_3D, _create_material_funcs); + + { + //shader compiler + ShaderCompiler::DefaultIdentifierActions actions; + + actions.renames["MODEL_MATRIX"] = "read_model_matrix"; + actions.renames["MODEL_NORMAL_MATRIX"] = "model_normal_matrix"; + actions.renames["VIEW_MATRIX"] = "scene_data.view_matrix"; + actions.renames["INV_VIEW_MATRIX"] = "inv_view_matrix"; + actions.renames["PROJECTION_MATRIX"] = "projection_matrix"; + actions.renames["INV_PROJECTION_MATRIX"] = "inv_projection_matrix"; + actions.renames["MODELVIEW_MATRIX"] = "modelview"; + actions.renames["MODELVIEW_NORMAL_MATRIX"] = "modelview_normal"; + + actions.renames["VERTEX"] = "vertex"; + actions.renames["NORMAL"] = "normal"; + actions.renames["TANGENT"] = "tangent"; + actions.renames["BINORMAL"] = "binormal"; + actions.renames["POSITION"] = "position"; + actions.renames["UV"] = "uv_interp"; + actions.renames["UV2"] = "uv2_interp"; + actions.renames["COLOR"] = "color_interp"; + actions.renames["POINT_SIZE"] = "gl_PointSize"; + actions.renames["INSTANCE_ID"] = "gl_InstanceIndex"; + actions.renames["VERTEX_ID"] = "gl_VertexIndex"; + + actions.renames["ALPHA_SCISSOR_THRESHOLD"] = "alpha_scissor_threshold"; + actions.renames["ALPHA_HASH_SCALE"] = "alpha_hash_scale"; + actions.renames["ALPHA_ANTIALIASING_EDGE"] = "alpha_antialiasing_edge"; + actions.renames["ALPHA_TEXTURE_COORDINATE"] = "alpha_texture_coordinate"; + + //builtins + + actions.renames["TIME"] = "scene_data_block.data.time"; + actions.renames["PI"] = _MKSTR(Math_PI); + actions.renames["TAU"] = _MKSTR(Math_TAU); + actions.renames["E"] = _MKSTR(Math_E); + actions.renames["VIEWPORT_SIZE"] = "scene_data.viewport_size"; + + actions.renames["FRAGCOORD"] = "gl_FragCoord"; + actions.renames["FRONT_FACING"] = "gl_FrontFacing"; + actions.renames["NORMAL_MAP"] = "normal_map"; + actions.renames["NORMAL_MAP_DEPTH"] = "normal_map_depth"; + actions.renames["ALBEDO"] = "albedo"; + actions.renames["ALPHA"] = "alpha"; + actions.renames["METALLIC"] = "metallic"; + actions.renames["SPECULAR"] = "specular"; + actions.renames["ROUGHNESS"] = "roughness"; + actions.renames["RIM"] = "rim"; + actions.renames["RIM_TINT"] = "rim_tint"; + actions.renames["CLEARCOAT"] = "clearcoat"; + actions.renames["CLEARCOAT_ROUGHNESS"] = "clearcoat_roughness"; + actions.renames["ANISOTROPY"] = "anisotropy"; + actions.renames["ANISOTROPY_FLOW"] = "anisotropy_flow"; + actions.renames["SSS_STRENGTH"] = "sss_strength"; + actions.renames["SSS_TRANSMITTANCE_COLOR"] = "transmittance_color"; + actions.renames["SSS_TRANSMITTANCE_DEPTH"] = "transmittance_depth"; + actions.renames["SSS_TRANSMITTANCE_BOOST"] = "transmittance_boost"; + actions.renames["BACKLIGHT"] = "backlight"; + actions.renames["AO"] = "ao"; + actions.renames["AO_LIGHT_AFFECT"] = "ao_light_affect"; + actions.renames["EMISSION"] = "emission"; + actions.renames["POINT_COORD"] = "gl_PointCoord"; + actions.renames["INSTANCE_CUSTOM"] = "instance_custom"; + actions.renames["SCREEN_UV"] = "screen_uv"; + actions.renames["SCREEN_TEXTURE"] = "color_buffer"; + actions.renames["DEPTH_TEXTURE"] = "depth_buffer"; + actions.renames["NORMAL_ROUGHNESS_TEXTURE"] = "normal_roughness_buffer"; + actions.renames["DEPTH"] = "gl_FragDepth"; + actions.renames["OUTPUT_IS_SRGB"] = "true"; + actions.renames["FOG"] = "fog"; + actions.renames["RADIANCE"] = "custom_radiance"; + actions.renames["IRRADIANCE"] = "custom_irradiance"; + actions.renames["BONE_INDICES"] = "bone_attrib"; + actions.renames["BONE_WEIGHTS"] = "weight_attrib"; + actions.renames["CUSTOM0"] = "custom0_attrib"; + actions.renames["CUSTOM1"] = "custom1_attrib"; + actions.renames["CUSTOM2"] = "custom2_attrib"; + actions.renames["CUSTOM3"] = "custom3_attrib"; + actions.renames["OUTPUT_IS_SRGB"] = "SHADER_IS_SRGB"; + + actions.renames["NODE_POSITION_WORLD"] = "read_model_matrix[3].xyz"; + actions.renames["CAMERA_POSITION_WORLD"] = "scene_data.inv_view_matrix[3].xyz"; + actions.renames["CAMERA_DIRECTION_WORLD"] = "scene_data.view_matrix[3].xyz"; + actions.renames["NODE_POSITION_VIEW"] = "(read_model_matrix * scene_data.view_matrix)[3].xyz"; + + actions.renames["VIEW_INDEX"] = "ViewIndex"; + actions.renames["VIEW_MONO_LEFT"] = "0"; + actions.renames["VIEW_RIGHT"] = "1"; + + //for light + actions.renames["VIEW"] = "view"; + actions.renames["LIGHT_COLOR"] = "light_color"; + actions.renames["LIGHT"] = "light"; + actions.renames["ATTENUATION"] = "attenuation"; + actions.renames["DIFFUSE_LIGHT"] = "diffuse_light"; + actions.renames["SPECULAR_LIGHT"] = "specular_light"; + + actions.usage_defines["NORMAL"] = "#define NORMAL_USED\n"; + actions.usage_defines["TANGENT"] = "#define TANGENT_USED\n"; + actions.usage_defines["BINORMAL"] = "@TANGENT"; + actions.usage_defines["RIM"] = "#define LIGHT_RIM_USED\n"; + actions.usage_defines["RIM_TINT"] = "@RIM"; + actions.usage_defines["CLEARCOAT"] = "#define LIGHT_CLEARCOAT_USED\n"; + actions.usage_defines["CLEARCOAT_ROUGHNESS"] = "@CLEARCOAT"; + actions.usage_defines["ANISOTROPY"] = "#define LIGHT_ANISOTROPY_USED\n"; + actions.usage_defines["ANISOTROPY_FLOW"] = "@ANISOTROPY"; + actions.usage_defines["AO"] = "#define AO_USED\n"; + actions.usage_defines["AO_LIGHT_AFFECT"] = "#define AO_USED\n"; + actions.usage_defines["UV"] = "#define UV_USED\n"; + actions.usage_defines["UV2"] = "#define UV2_USED\n"; + actions.usage_defines["BONE_INDICES"] = "#define BONES_USED\n"; + actions.usage_defines["BONE_WEIGHTS"] = "#define WEIGHTS_USED\n"; + actions.usage_defines["CUSTOM0"] = "#define CUSTOM0_USED\n"; + actions.usage_defines["CUSTOM1"] = "#define CUSTOM1_USED\n"; + actions.usage_defines["CUSTOM2"] = "#define CUSTOM2_USED\n"; + actions.usage_defines["CUSTOM3"] = "#define CUSTOM3_USED\n"; + actions.usage_defines["NORMAL_MAP"] = "#define NORMAL_MAP_USED\n"; + actions.usage_defines["NORMAL_MAP_DEPTH"] = "@NORMAL_MAP"; + actions.usage_defines["COLOR"] = "#define COLOR_USED\n"; + actions.usage_defines["INSTANCE_CUSTOM"] = "#define ENABLE_INSTANCE_CUSTOM\n"; + actions.usage_defines["POSITION"] = "#define OVERRIDE_POSITION\n"; + + actions.usage_defines["ALPHA_SCISSOR_THRESHOLD"] = "#define ALPHA_SCISSOR_USED\n"; + actions.usage_defines["ALPHA_HASH_SCALE"] = "#define ALPHA_HASH_USED\n"; + actions.usage_defines["ALPHA_ANTIALIASING_EDGE"] = "#define ALPHA_ANTIALIASING_EDGE_USED\n"; + actions.usage_defines["ALPHA_TEXTURE_COORDINATE"] = "@ALPHA_ANTIALIASING_EDGE"; + + actions.usage_defines["SSS_STRENGTH"] = "#define ENABLE_SSS\n"; + actions.usage_defines["SSS_TRANSMITTANCE_DEPTH"] = "#define ENABLE_TRANSMITTANCE\n"; + actions.usage_defines["BACKLIGHT"] = "#define LIGHT_BACKLIGHT_USED\n"; + actions.usage_defines["SCREEN_TEXTURE"] = "#define SCREEN_TEXTURE_USED\n"; + actions.usage_defines["SCREEN_UV"] = "#define SCREEN_UV_USED\n"; + + actions.usage_defines["DIFFUSE_LIGHT"] = "#define USE_LIGHT_SHADER_CODE\n"; + actions.usage_defines["SPECULAR_LIGHT"] = "#define USE_LIGHT_SHADER_CODE\n"; + + actions.usage_defines["FOG"] = "#define CUSTOM_FOG_USED\n"; + actions.usage_defines["RADIANCE"] = "#define CUSTOM_RADIANCE_USED\n"; + actions.usage_defines["IRRADIANCE"] = "#define CUSTOM_IRRADIANCE_USED\n"; + + actions.usage_defines["MODEL_MATRIX"] = "#define MODEL_MATRIX_USED\n"; + + actions.render_mode_defines["skip_vertex_transform"] = "#define SKIP_TRANSFORM_USED\n"; + actions.render_mode_defines["world_vertex_coords"] = "#define VERTEX_WORLD_COORDS_USED\n"; + actions.render_mode_defines["ensure_correct_normals"] = "#define ENSURE_CORRECT_NORMALS\n"; + actions.render_mode_defines["cull_front"] = "#define DO_SIDE_CHECK\n"; + actions.render_mode_defines["cull_disabled"] = "#define DO_SIDE_CHECK\n"; + actions.render_mode_defines["particle_trails"] = "#define USE_PARTICLE_TRAILS\n"; + actions.render_mode_defines["depth_draw_opaque"] = "#define USE_OPAQUE_PREPASS\n"; + + bool force_lambert = GLOBAL_GET("rendering/shading/overrides/force_lambert_over_burley"); + if (!force_lambert) { + actions.render_mode_defines["diffuse_burley"] = "#define DIFFUSE_BURLEY\n"; + } + + actions.render_mode_defines["diffuse_lambert_wrap"] = "#define DIFFUSE_LAMBERT_WRAP\n"; + actions.render_mode_defines["diffuse_toon"] = "#define DIFFUSE_TOON\n"; + + actions.render_mode_defines["sss_mode_skin"] = "#define SSS_MODE_SKIN\n"; + + actions.render_mode_defines["specular_schlick_ggx"] = "#define SPECULAR_SCHLICK_GGX\n"; + + actions.render_mode_defines["specular_toon"] = "#define SPECULAR_TOON\n"; + actions.render_mode_defines["specular_disabled"] = "#define SPECULAR_DISABLED\n"; + actions.render_mode_defines["shadows_disabled"] = "#define SHADOWS_DISABLED\n"; + actions.render_mode_defines["ambient_light_disabled"] = "#define AMBIENT_LIGHT_DISABLED\n"; + actions.render_mode_defines["shadow_to_opacity"] = "#define USE_SHADOW_TO_OPACITY\n"; + actions.render_mode_defines["unshaded"] = "#define MODE_UNSHADED\n"; + + actions.sampler_array_name = "material_samplers"; + actions.base_texture_binding_index = 1; + actions.texture_layout_set = RenderForwardMobile::MATERIAL_UNIFORM_SET; + actions.base_uniform_string = "material."; + actions.base_varying_index = 10; + + actions.default_filter = ShaderLanguage::FILTER_LINEAR_MIPMAP; + actions.default_repeat = ShaderLanguage::REPEAT_ENABLE; + actions.global_buffer_array_variable = "global_shader_uniforms.data"; + actions.instance_uniform_index_variable = "draw_call.instance_uniforms_ofs"; + + actions.apply_luminance_multiplier = true; // apply luminance multiplier to screen texture + + compiler.initialize(actions); + } + + { + //default material and shader + default_shader = material_storage->shader_allocate(); + material_storage->shader_initialize(default_shader); + material_storage->shader_set_code(default_shader, R"( +// Default 3D material shader (mobile). + +shader_type spatial; + +void vertex() { + ROUGHNESS = 0.8; +} + +void fragment() { + ALBEDO = vec3(0.6); + ROUGHNESS = 0.8; + METALLIC = 0.2; +} +)"); + default_material = material_storage->material_allocate(); + material_storage->material_initialize(default_material); + material_storage->material_set_shader(default_material, default_shader); + + MaterialData *md = static_cast<MaterialData *>(material_storage->material_get_data(default_material, RendererRD::MaterialStorage::SHADER_TYPE_3D)); + default_shader_rd = shader.version_get_shader(md->shader_data->version, SHADER_VERSION_COLOR_PASS); + + default_material_shader_ptr = md->shader_data; + default_material_uniform_set = md->uniform_set; + } + + { + overdraw_material_shader = material_storage->shader_allocate(); + material_storage->shader_initialize(overdraw_material_shader); + // Use relatively low opacity so that more "layers" of overlapping objects can be distinguished. + material_storage->shader_set_code(overdraw_material_shader, R"( +// 3D editor Overdraw debug draw mode shader (mobile). + +shader_type spatial; + +render_mode blend_add, unshaded; + +void fragment() { + ALBEDO = vec3(0.4, 0.8, 0.8); + ALPHA = 0.1; +} +)"); + overdraw_material = material_storage->material_allocate(); + material_storage->material_initialize(overdraw_material); + material_storage->material_set_shader(overdraw_material, overdraw_material_shader); + + MaterialData *md = static_cast<MaterialData *>(material_storage->material_get_data(overdraw_material, RendererRD::MaterialStorage::SHADER_TYPE_3D)); + overdraw_material_shader_ptr = md->shader_data; + overdraw_material_uniform_set = md->uniform_set; + } + + { + default_vec4_xform_buffer = RD::get_singleton()->storage_buffer_create(256); + Vector<RD::Uniform> uniforms; + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.append_id(default_vec4_xform_buffer); + u.binding = 0; + uniforms.push_back(u); + + default_vec4_xform_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, default_shader_rd, RenderForwardMobile::TRANSFORMS_UNIFORM_SET); + } + { + RD::SamplerState sampler; + sampler.mag_filter = RD::SAMPLER_FILTER_LINEAR; + sampler.min_filter = RD::SAMPLER_FILTER_LINEAR; + sampler.enable_compare = true; + sampler.compare_op = RD::COMPARE_OP_LESS; + shadow_sampler = RD::get_singleton()->sampler_create(sampler); + } +} + +void SceneShaderForwardMobile::set_default_specialization_constants(const Vector<RD::PipelineSpecializationConstant> &p_constants) { + default_specialization_constants = p_constants; + for (SelfList<ShaderData> *E = shader_list.first(); E; E = E->next()) { + for (int i = 0; i < ShaderData::CULL_VARIANT_MAX; i++) { + for (int j = 0; j < RS::PRIMITIVE_MAX; j++) { + for (int k = 0; k < SHADER_VERSION_MAX; k++) { + E->self()->pipelines[i][j][k].update_specialization_constants(default_specialization_constants); + } + } + } + } +} + +SceneShaderForwardMobile::~SceneShaderForwardMobile() { + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + + RD::get_singleton()->free(default_vec4_xform_buffer); + RD::get_singleton()->free(shadow_sampler); + + material_storage->shader_free(overdraw_material_shader); + material_storage->shader_free(default_shader); + + material_storage->material_free(overdraw_material); + material_storage->material_free(default_material); +} diff --git a/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.h b/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.h new file mode 100644 index 0000000000..f67665a02f --- /dev/null +++ b/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.h @@ -0,0 +1,216 @@ +/*************************************************************************/ +/* scene_shader_forward_mobile.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef SCENE_SHADER_FORWARD_MOBILE_H +#define SCENE_SHADER_FORWARD_MOBILE_H + +#include "servers/rendering/renderer_rd/renderer_scene_render_rd.h" +#include "servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile.glsl.gen.h" + +namespace RendererSceneRenderImplementation { + +class SceneShaderForwardMobile { +private: + static SceneShaderForwardMobile *singleton; + +public: + enum ShaderVersion { + SHADER_VERSION_COLOR_PASS, + SHADER_VERSION_LIGHTMAP_COLOR_PASS, + SHADER_VERSION_SHADOW_PASS, + SHADER_VERSION_SHADOW_PASS_DP, + SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL, + + SHADER_VERSION_COLOR_PASS_MULTIVIEW, + SHADER_VERSION_LIGHTMAP_COLOR_PASS_MULTIVIEW, + SHADER_VERSION_SHADOW_PASS_MULTIVIEW, + + SHADER_VERSION_MAX + }; + + struct ShaderData : public RendererRD::MaterialStorage::ShaderData { + enum BlendMode { //used internally + BLEND_MODE_MIX, + BLEND_MODE_ADD, + BLEND_MODE_SUB, + BLEND_MODE_MUL, + BLEND_MODE_ALPHA_TO_COVERAGE + }; + + enum DepthDraw { + DEPTH_DRAW_DISABLED, + DEPTH_DRAW_OPAQUE, + DEPTH_DRAW_ALWAYS + }; + + enum DepthTest { + DEPTH_TEST_DISABLED, + DEPTH_TEST_ENABLED + }; + + enum Cull { + CULL_DISABLED, + CULL_FRONT, + CULL_BACK + }; + + enum CullVariant { + CULL_VARIANT_NORMAL, + CULL_VARIANT_REVERSED, + CULL_VARIANT_DOUBLE_SIDED, + CULL_VARIANT_MAX + + }; + + enum AlphaAntiAliasing { + ALPHA_ANTIALIASING_OFF, + ALPHA_ANTIALIASING_ALPHA_TO_COVERAGE, + ALPHA_ANTIALIASING_ALPHA_TO_COVERAGE_AND_TO_ONE + }; + + bool valid = false; + RID version; + uint32_t vertex_input_mask = 0; + PipelineCacheRD pipelines[CULL_VARIANT_MAX][RS::PRIMITIVE_MAX][SHADER_VERSION_MAX]; + + String path; + + HashMap<StringName, ShaderLanguage::ShaderNode::Uniform> uniforms; + Vector<ShaderCompiler::GeneratedCode::Texture> texture_uniforms; + + Vector<uint32_t> ubo_offsets; + uint32_t ubo_size = 0; + + String code; + HashMap<StringName, HashMap<int, RID>> default_texture_params; + + DepthDraw depth_draw; + DepthTest depth_test; + + bool uses_point_size = false; + bool uses_alpha = false; + bool uses_blend_alpha = false; + bool uses_alpha_clip = false; + bool uses_depth_pre_pass = false; + bool uses_discard = false; + bool uses_roughness = false; + bool uses_normal = false; + bool uses_particle_trails = false; + + bool unshaded = false; + bool uses_vertex = false; + bool uses_sss = false; + bool uses_transmittance = false; + bool uses_screen_texture = false; + bool uses_depth_texture = false; + bool uses_normal_texture = false; + bool uses_time = false; + bool writes_modelview_or_projection = false; + bool uses_world_coordinates = false; + + uint64_t last_pass = 0; + uint32_t index = 0; + + virtual void set_code(const String &p_Code); + virtual void set_path_hint(const String &p_path); + + virtual void set_default_texture_parameter(const StringName &p_name, RID p_texture, int p_index); + virtual void get_shader_uniform_list(List<PropertyInfo> *p_param_list) const; + void get_instance_param_list(List<RendererMaterialStorage::InstanceShaderParam> *p_param_list) const; + + virtual bool is_parameter_texture(const StringName &p_param) const; + virtual bool is_animated() const; + virtual bool casts_shadows() const; + virtual Variant get_default_parameter(const StringName &p_parameter) const; + virtual RS::ShaderNativeSourceCode get_native_source_code() const; + + SelfList<ShaderData> shader_list_element; + + ShaderData(); + virtual ~ShaderData(); + }; + + RendererRD::MaterialStorage::ShaderData *_create_shader_func(); + static RendererRD::MaterialStorage::ShaderData *_create_shader_funcs() { + return static_cast<SceneShaderForwardMobile *>(singleton)->_create_shader_func(); + } + + struct MaterialData : public RendererRD::MaterialStorage::MaterialData { + ShaderData *shader_data = nullptr; + RID uniform_set; + uint64_t last_pass = 0; + uint32_t index = 0; + RID next_pass; + uint8_t priority; + virtual void set_render_priority(int p_priority); + virtual void set_next_pass(RID p_pass); + virtual bool update_parameters(const HashMap<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); + virtual ~MaterialData(); + }; + + SelfList<ShaderData>::List shader_list; + + RendererRD::MaterialStorage::MaterialData *_create_material_func(ShaderData *p_shader); + static RendererRD::MaterialStorage::MaterialData *_create_material_funcs(RendererRD::MaterialStorage::ShaderData *p_shader) { + return static_cast<SceneShaderForwardMobile *>(singleton)->_create_material_func(static_cast<ShaderData *>(p_shader)); + } + + SceneForwardMobileShaderRD shader; + ShaderCompiler compiler; + + RID default_shader; + RID default_material; + RID overdraw_material_shader; + RID overdraw_material; + RID default_shader_rd; + + RID default_vec4_xform_buffer; + RID default_vec4_xform_uniform_set; + + RID shadow_sampler; + + RID default_material_uniform_set; + ShaderData *default_material_shader_ptr = nullptr; + + RID overdraw_material_uniform_set; + ShaderData *overdraw_material_shader_ptr = nullptr; + + SceneShaderForwardMobile(); + ~SceneShaderForwardMobile(); + + Vector<RD::PipelineSpecializationConstant> default_specialization_constants; + + void init(const String p_defines); + void set_default_specialization_constants(const Vector<RD::PipelineSpecializationConstant> &p_constants); +}; + +} // namespace RendererSceneRenderImplementation + +#endif // SCENE_SHADER_FORWARD_MOBILE_H diff --git a/servers/rendering/renderer_rd/framebuffer_cache_rd.cpp b/servers/rendering/renderer_rd/framebuffer_cache_rd.cpp new file mode 100644 index 0000000000..9baa86a32d --- /dev/null +++ b/servers/rendering/renderer_rd/framebuffer_cache_rd.cpp @@ -0,0 +1,64 @@ +/*************************************************************************/ +/* framebuffer_cache_rd.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "framebuffer_cache_rd.h" + +FramebufferCacheRD *FramebufferCacheRD::singleton = nullptr; + +void FramebufferCacheRD::_invalidate(Cache *p_cache) { + if (p_cache->prev) { + p_cache->prev->next = p_cache->next; + } else { + // At beginning of table + uint32_t table_idx = p_cache->hash % HASH_TABLE_SIZE; + hash_table[table_idx] = p_cache->next; + } + + if (p_cache->next) { + p_cache->next->prev = p_cache->prev; + } + + cache_allocator.free(p_cache); + cache_instances_used--; +} +void FramebufferCacheRD::_framebuffer_invalidation_callback(void *p_userdata) { + singleton->_invalidate(reinterpret_cast<Cache *>(p_userdata)); +} + +FramebufferCacheRD::FramebufferCacheRD() { + ERR_FAIL_COND(singleton != nullptr); + singleton = this; +} + +FramebufferCacheRD::~FramebufferCacheRD() { + if (cache_instances_used > 0) { + ERR_PRINT("At exit: " + itos(cache_instances_used) + " framebuffer cache instance(s) still in use."); + } +} diff --git a/servers/rendering/renderer_rd/framebuffer_cache_rd.h b/servers/rendering/renderer_rd/framebuffer_cache_rd.h new file mode 100644 index 0000000000..f50d6baa30 --- /dev/null +++ b/servers/rendering/renderer_rd/framebuffer_cache_rd.h @@ -0,0 +1,310 @@ +/*************************************************************************/ +/* framebuffer_cache_rd.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef FRAMEBUFFER_CACHE_RD_H +#define FRAMEBUFFER_CACHE_RD_H + +#include "core/templates/local_vector.h" +#include "core/templates/paged_allocator.h" +#include "servers/rendering/rendering_device.h" + +class FramebufferCacheRD : public Object { + GDCLASS(FramebufferCacheRD, Object) + + struct Cache { + Cache *prev = nullptr; + Cache *next = nullptr; + uint32_t hash = 0; + RID cache; + LocalVector<RID> textures; + LocalVector<RD::FramebufferPass> passes; + uint32_t views = 0; + }; + + PagedAllocator<Cache> cache_allocator; + + enum { + HASH_TABLE_SIZE = 16381 // Prime + }; + + Cache *hash_table[HASH_TABLE_SIZE] = {}; + + static _FORCE_INLINE_ uint32_t _hash_pass(const RD::FramebufferPass &p, uint32_t h) { + h = hash_murmur3_one_32(p.depth_attachment, h); + h = hash_murmur3_one_32(p.vrs_attachment, h); + + h = hash_murmur3_one_32(p.color_attachments.size(), h); + for (int i = 0; i < p.color_attachments.size(); i++) { + h = hash_murmur3_one_32(p.color_attachments[i], h); + } + + h = hash_murmur3_one_32(p.resolve_attachments.size(), h); + for (int i = 0; i < p.resolve_attachments.size(); i++) { + h = hash_murmur3_one_32(p.resolve_attachments[i], h); + } + + h = hash_murmur3_one_32(p.preserve_attachments.size(), h); + for (int i = 0; i < p.preserve_attachments.size(); i++) { + h = hash_murmur3_one_32(p.preserve_attachments[i], h); + } + + return h; + } + + static _FORCE_INLINE_ bool _compare_pass(const RD::FramebufferPass &a, const RD::FramebufferPass &b) { + if (a.depth_attachment != b.depth_attachment) { + return false; + } + + if (a.vrs_attachment != b.vrs_attachment) { + return false; + } + + if (a.color_attachments.size() != b.color_attachments.size()) { + return false; + } + + for (int i = 0; i < a.color_attachments.size(); i++) { + if (a.color_attachments[i] != b.color_attachments[i]) { + return false; + } + } + + if (a.resolve_attachments.size() != b.resolve_attachments.size()) { + return false; + } + + for (int i = 0; i < a.resolve_attachments.size(); i++) { + if (a.resolve_attachments[i] != b.resolve_attachments[i]) { + return false; + } + } + + if (a.preserve_attachments.size() != b.preserve_attachments.size()) { + return false; + } + + for (int i = 0; i < a.preserve_attachments.size(); i++) { + if (a.preserve_attachments[i] != b.preserve_attachments[i]) { + return false; + } + } + + return true; + } + + _FORCE_INLINE_ uint32_t _hash_rids(uint32_t h, const RID &arg) { + return hash_murmur3_one_64(arg.get_id(), h); + } + + template <typename... Args> + uint32_t _hash_rids(uint32_t h, const RID &arg, Args... args) { + h = hash_murmur3_one_64(arg.get_id(), h); + return _hash_rids(h, args...); + } + + _FORCE_INLINE_ bool _compare_args(uint32_t idx, const LocalVector<RID> &textures, const RID &arg) { + return textures[idx] == arg; + } + + template <typename... Args> + _FORCE_INLINE_ bool _compare_args(uint32_t idx, const LocalVector<RID> &textures, const RID &arg, Args... args) { + if (textures[idx] != arg) { + return false; + } + return _compare_args(idx + 1, textures, args...); + } + + _FORCE_INLINE_ void _create_args(Vector<RID> &textures, const RID &arg) { + textures.push_back(arg); + } + + template <typename... Args> + _FORCE_INLINE_ void _create_args(Vector<RID> &textures, const RID &arg, Args... args) { + textures.push_back(arg); + _create_args(textures, args...); + } + + static FramebufferCacheRD *singleton; + + uint32_t cache_instances_used = 0; + + void _invalidate(Cache *p_cache); + static void _framebuffer_invalidation_callback(void *p_userdata); + + RID _allocate_from_data(uint32_t p_views, uint32_t p_hash, uint32_t p_table_idx, const Vector<RID> &p_textures, const Vector<RD::FramebufferPass> &p_passes) { + RID rid; + if (p_passes.size()) { + rid = RD::get_singleton()->framebuffer_create_multipass(p_textures, p_passes, RD::INVALID_ID, p_views); + } else { + rid = RD::get_singleton()->framebuffer_create(p_textures, RD::INVALID_ID, p_views); + } + + ERR_FAIL_COND_V(rid.is_null(), rid); + + Cache *c = cache_allocator.alloc(); + c->views = p_views; + c->cache = rid; + c->hash = p_hash; + c->textures.resize(p_textures.size()); + for (uint32_t i = 0; i < c->textures.size(); i++) { + c->textures[i] = p_textures[i]; + } + c->passes.resize(p_passes.size()); + for (uint32_t i = 0; i < c->passes.size(); i++) { + c->passes[i] = p_passes[i]; + } + c->prev = nullptr; + c->next = hash_table[p_table_idx]; + if (hash_table[p_table_idx]) { + hash_table[p_table_idx]->prev = c; + } + hash_table[p_table_idx] = c; + + RD::get_singleton()->framebuffer_set_invalidation_callback(rid, _framebuffer_invalidation_callback, c); + + cache_instances_used++; + + return rid; + } + +public: + template <typename... Args> + RID get_cache(Args... args) { + uint32_t h = hash_murmur3_one_32(1); //1 view + h = hash_murmur3_one_32(sizeof...(Args), h); + h = _hash_rids(h, args...); + h = hash_murmur3_one_32(0, h); // 0 passes + h = hash_fmix32(h); + + uint32_t table_idx = h % HASH_TABLE_SIZE; + { + const Cache *c = hash_table[table_idx]; + + while (c) { + if (c->hash == h && c->passes.size() == 0 && c->textures.size() == sizeof...(Args) && c->views == 1 && _compare_args(0, c->textures, args...)) { + return c->cache; + } + c = c->next; + } + } + + // Not in cache, create: + + Vector<RID> textures; + _create_args(textures, args...); + + return _allocate_from_data(1, h, table_idx, textures, Vector<RD::FramebufferPass>()); + } + + template <typename... Args> + RID get_cache_multiview(uint32_t p_views, Args... args) { + uint32_t h = hash_murmur3_one_32(p_views); + h = hash_murmur3_one_32(sizeof...(Args), h); + h = _hash_rids(h, args...); + h = hash_murmur3_one_32(0, h); // 0 passes + h = hash_fmix32(h); + + uint32_t table_idx = h % HASH_TABLE_SIZE; + { + const Cache *c = hash_table[table_idx]; + + while (c) { + if (c->hash == h && c->passes.size() == 0 && c->textures.size() == sizeof...(Args) && c->views == p_views && _compare_args(0, c->textures, args...)) { + return c->cache; + } + c = c->next; + } + } + + // Not in cache, create: + + Vector<RID> textures; + _create_args(textures, args...); + + return _allocate_from_data(p_views, h, table_idx, textures, Vector<RD::FramebufferPass>()); + } + + RID get_cache_multipass(const Vector<RID> &p_textures, const Vector<RD::FramebufferPass> &p_passes, uint32_t p_views = 1) { + uint32_t h = hash_murmur3_one_32(p_views); + h = hash_murmur3_one_32(p_textures.size()); + for (int i = 0; i < p_textures.size(); i++) { + h = hash_murmur3_one_64(p_textures[i].get_id(), h); + } + h = hash_murmur3_one_32(p_passes.size()); + for (int i = 0; i < p_passes.size(); i++) { + h = _hash_pass(p_passes[i], h); + } + + h = hash_fmix32(h); + + uint32_t table_idx = h % HASH_TABLE_SIZE; + { + const Cache *c = hash_table[table_idx]; + + while (c) { + if (c->hash == h && c->views == p_views && c->textures.size() == (uint32_t)p_textures.size() && c->passes.size() == (uint32_t)p_passes.size()) { + bool all_ok = true; + + for (int i = 0; i < p_textures.size(); i++) { + if (p_textures[i] != c->textures[i]) { + all_ok = false; + break; + } + } + + if (all_ok) { + for (int i = 0; i < p_passes.size(); i++) { + if (!_compare_pass(p_passes[i], c->passes[i])) { + all_ok = false; + break; + } + } + } + + if (all_ok) { + return c->cache; + } + } + c = c->next; + } + } + + // Not in cache, create: + return _allocate_from_data(p_views, h, table_idx, p_textures, p_passes); + } + + static FramebufferCacheRD *get_singleton() { return singleton; } + + FramebufferCacheRD(); + ~FramebufferCacheRD(); +}; + +#endif // FRAMEBUFFER_CACHE_RD_H diff --git a/servers/rendering/renderer_rd/pipeline_cache_rd.cpp b/servers/rendering/renderer_rd/pipeline_cache_rd.cpp new file mode 100644 index 0000000000..7a2336970b --- /dev/null +++ b/servers/rendering/renderer_rd/pipeline_cache_rd.cpp @@ -0,0 +1,126 @@ +/*************************************************************************/ +/* pipeline_cache_rd.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "pipeline_cache_rd.h" + +#include "core/os/memory.h" + +RID PipelineCacheRD::_generate_version(RD::VertexFormatID p_vertex_format_id, RD::FramebufferFormatID p_framebuffer_format_id, bool p_wireframe, uint32_t p_render_pass, uint32_t p_bool_specializations) { + RD::PipelineMultisampleState multisample_state_version = multisample_state; + multisample_state_version.sample_count = RD::get_singleton()->framebuffer_format_get_texture_samples(p_framebuffer_format_id, p_render_pass); + + bool wireframe = p_wireframe || rasterization_state.wireframe; + + RD::PipelineRasterizationState raster_state_version = rasterization_state; + raster_state_version.wireframe = wireframe; + + Vector<RD::PipelineSpecializationConstant> specialization_constants = base_specialization_constants; + + uint32_t bool_index = 0; + uint32_t bool_specializations = p_bool_specializations; + while (bool_specializations) { + if (bool_specializations & (1 << bool_index)) { + RD::PipelineSpecializationConstant sc; + sc.bool_value = true; + sc.constant_id = bool_index; + sc.type = RD::PIPELINE_SPECIALIZATION_CONSTANT_TYPE_BOOL; + specialization_constants.push_back(sc); + bool_specializations &= ~(1 << bool_index); + } + bool_index++; + } + + RID pipeline = RD::get_singleton()->render_pipeline_create(shader, p_framebuffer_format_id, p_vertex_format_id, render_primitive, raster_state_version, multisample_state_version, depth_stencil_state, blend_state, dynamic_state_flags, p_render_pass, specialization_constants); + ERR_FAIL_COND_V(pipeline.is_null(), RID()); + versions = static_cast<Version *>(memrealloc(versions, sizeof(Version) * (version_count + 1))); + versions[version_count].framebuffer_id = p_framebuffer_format_id; + versions[version_count].vertex_id = p_vertex_format_id; + versions[version_count].wireframe = wireframe; + versions[version_count].pipeline = pipeline; + versions[version_count].render_pass = p_render_pass; + versions[version_count].bool_specializations = p_bool_specializations; + version_count++; + return pipeline; +} + +void PipelineCacheRD::_clear() { + // TODO: Clear should probably recompile all the variants already compiled instead to avoid stalls? Needs discussion. + if (versions) { + for (uint32_t i = 0; i < version_count; i++) { + //shader may be gone, so this may not be valid + if (RD::get_singleton()->render_pipeline_is_valid(versions[i].pipeline)) { + RD::get_singleton()->free(versions[i].pipeline); + } + } + version_count = 0; + memfree(versions); + versions = nullptr; + } +} + +void PipelineCacheRD::setup(RID p_shader, RD::RenderPrimitive p_primitive, const RD::PipelineRasterizationState &p_rasterization_state, RD::PipelineMultisampleState p_multisample, const RD::PipelineDepthStencilState &p_depth_stencil_state, const RD::PipelineColorBlendState &p_blend_state, int p_dynamic_state_flags, const Vector<RD::PipelineSpecializationConstant> &p_base_specialization_constants) { + ERR_FAIL_COND(p_shader.is_null()); + _clear(); + shader = p_shader; + input_mask = RD::get_singleton()->shader_get_vertex_input_attribute_mask(p_shader); + render_primitive = p_primitive; + rasterization_state = p_rasterization_state; + multisample_state = p_multisample; + depth_stencil_state = p_depth_stencil_state; + blend_state = p_blend_state; + dynamic_state_flags = p_dynamic_state_flags; + base_specialization_constants = p_base_specialization_constants; +} +void PipelineCacheRD::update_specialization_constants(const Vector<RD::PipelineSpecializationConstant> &p_base_specialization_constants) { + base_specialization_constants = p_base_specialization_constants; + _clear(); +} + +void PipelineCacheRD::update_shader(RID p_shader) { + ERR_FAIL_COND(p_shader.is_null()); + _clear(); + setup(p_shader, render_primitive, rasterization_state, multisample_state, depth_stencil_state, blend_state, dynamic_state_flags); +} + +void PipelineCacheRD::clear() { + _clear(); + shader = RID(); //clear shader + input_mask = 0; +} + +PipelineCacheRD::PipelineCacheRD() { + version_count = 0; + versions = nullptr; + input_mask = 0; +} + +PipelineCacheRD::~PipelineCacheRD() { + _clear(); +} diff --git a/servers/rendering/renderer_rd/pipeline_cache_rd.h b/servers/rendering/renderer_rd/pipeline_cache_rd.h new file mode 100644 index 0000000000..882e459bcd --- /dev/null +++ b/servers/rendering/renderer_rd/pipeline_cache_rd.h @@ -0,0 +1,100 @@ +/*************************************************************************/ +/* pipeline_cache_rd.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef PIPELINE_CACHE_RD_H +#define PIPELINE_CACHE_RD_H + +#include "core/os/spin_lock.h" +#include "servers/rendering/rendering_device.h" + +class PipelineCacheRD { + SpinLock spin_lock; + + RID shader; + uint32_t input_mask; + + RD::RenderPrimitive render_primitive; + RD::PipelineRasterizationState rasterization_state; + RD::PipelineMultisampleState multisample_state; + RD::PipelineDepthStencilState depth_stencil_state; + RD::PipelineColorBlendState blend_state; + int dynamic_state_flags = 0; + Vector<RD::PipelineSpecializationConstant> base_specialization_constants; + + struct Version { + RD::VertexFormatID vertex_id; + RD::FramebufferFormatID framebuffer_id; + uint32_t render_pass; + bool wireframe; + uint32_t bool_specializations; + RID pipeline; + }; + + Version *versions = nullptr; + uint32_t version_count; + + RID _generate_version(RD::VertexFormatID p_vertex_format_id, RD::FramebufferFormatID p_framebuffer_format_id, bool p_wireframe, uint32_t p_render_pass, uint32_t p_bool_specializations = 0); + + void _clear(); + +public: + void setup(RID p_shader, RD::RenderPrimitive p_primitive, const RD::PipelineRasterizationState &p_rasterization_state, RD::PipelineMultisampleState p_multisample, const RD::PipelineDepthStencilState &p_depth_stencil_state, const RD::PipelineColorBlendState &p_blend_state, int p_dynamic_state_flags = 0, const Vector<RD::PipelineSpecializationConstant> &p_base_specialization_constants = Vector<RD::PipelineSpecializationConstant>()); + void update_specialization_constants(const Vector<RD::PipelineSpecializationConstant> &p_base_specialization_constants); + void update_shader(RID p_shader); + + _FORCE_INLINE_ RID get_render_pipeline(RD::VertexFormatID p_vertex_format_id, RD::FramebufferFormatID p_framebuffer_format_id, bool p_wireframe = false, uint32_t p_render_pass = 0, uint32_t p_bool_specializations = 0) { +#ifdef DEBUG_ENABLED + ERR_FAIL_COND_V_MSG(shader.is_null(), RID(), + "Attempted to use an unused shader variant (shader is null),"); +#endif + + spin_lock.lock(); + RID result; + for (uint32_t i = 0; i < version_count; i++) { + if (versions[i].vertex_id == p_vertex_format_id && versions[i].framebuffer_id == p_framebuffer_format_id && versions[i].wireframe == p_wireframe && versions[i].render_pass == p_render_pass && versions[i].bool_specializations == p_bool_specializations) { + result = versions[i].pipeline; + spin_lock.unlock(); + return result; + } + } + result = _generate_version(p_vertex_format_id, p_framebuffer_format_id, p_wireframe, p_render_pass, p_bool_specializations); + spin_lock.unlock(); + return result; + } + + _FORCE_INLINE_ uint32_t get_vertex_input_mask() const { + return input_mask; + } + void clear(); + PipelineCacheRD(); + ~PipelineCacheRD(); +}; + +#endif // PIPELINE_CACHE_RD_H diff --git a/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp b/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp new file mode 100644 index 0000000000..dc0214e307 --- /dev/null +++ b/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp @@ -0,0 +1,2866 @@ +/*************************************************************************/ +/* renderer_canvas_render_rd.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "renderer_canvas_render_rd.h" + +#include "core/config/project_settings.h" +#include "core/math/geometry_2d.h" +#include "core/math/math_defs.h" +#include "core/math/math_funcs.h" +#include "renderer_compositor_rd.h" +#include "servers/rendering/renderer_rd/storage_rd/material_storage.h" +#include "servers/rendering/renderer_rd/storage_rd/particles_storage.h" +#include "servers/rendering/renderer_rd/storage_rd/texture_storage.h" +#include "servers/rendering/rendering_server_default.h" + +void RendererCanvasRenderRD::_update_transform_2d_to_mat4(const Transform2D &p_transform, float *p_mat4) { + p_mat4[0] = p_transform.columns[0][0]; + p_mat4[1] = p_transform.columns[0][1]; + p_mat4[2] = 0; + p_mat4[3] = 0; + p_mat4[4] = p_transform.columns[1][0]; + p_mat4[5] = p_transform.columns[1][1]; + p_mat4[6] = 0; + p_mat4[7] = 0; + p_mat4[8] = 0; + p_mat4[9] = 0; + p_mat4[10] = 1; + p_mat4[11] = 0; + p_mat4[12] = p_transform.columns[2][0]; + p_mat4[13] = p_transform.columns[2][1]; + p_mat4[14] = 0; + p_mat4[15] = 1; +} + +void RendererCanvasRenderRD::_update_transform_2d_to_mat2x4(const Transform2D &p_transform, float *p_mat2x4) { + p_mat2x4[0] = p_transform.columns[0][0]; + p_mat2x4[1] = p_transform.columns[1][0]; + p_mat2x4[2] = 0; + p_mat2x4[3] = p_transform.columns[2][0]; + + p_mat2x4[4] = p_transform.columns[0][1]; + p_mat2x4[5] = p_transform.columns[1][1]; + p_mat2x4[6] = 0; + p_mat2x4[7] = p_transform.columns[2][1]; +} + +void RendererCanvasRenderRD::_update_transform_2d_to_mat2x3(const Transform2D &p_transform, float *p_mat2x3) { + p_mat2x3[0] = p_transform.columns[0][0]; + p_mat2x3[1] = p_transform.columns[0][1]; + p_mat2x3[2] = p_transform.columns[1][0]; + p_mat2x3[3] = p_transform.columns[1][1]; + p_mat2x3[4] = p_transform.columns[2][0]; + p_mat2x3[5] = p_transform.columns[2][1]; +} + +void RendererCanvasRenderRD::_update_transform_to_mat4(const Transform3D &p_transform, float *p_mat4) { + p_mat4[0] = p_transform.basis.rows[0][0]; + p_mat4[1] = p_transform.basis.rows[1][0]; + p_mat4[2] = p_transform.basis.rows[2][0]; + p_mat4[3] = 0; + p_mat4[4] = p_transform.basis.rows[0][1]; + p_mat4[5] = p_transform.basis.rows[1][1]; + p_mat4[6] = p_transform.basis.rows[2][1]; + p_mat4[7] = 0; + p_mat4[8] = p_transform.basis.rows[0][2]; + p_mat4[9] = p_transform.basis.rows[1][2]; + p_mat4[10] = p_transform.basis.rows[2][2]; + p_mat4[11] = 0; + p_mat4[12] = p_transform.origin.x; + p_mat4[13] = p_transform.origin.y; + p_mat4[14] = p_transform.origin.z; + p_mat4[15] = 1; +} + +RendererCanvasRender::PolygonID RendererCanvasRenderRD::request_polygon(const Vector<int> &p_indices, const Vector<Point2> &p_points, const Vector<Color> &p_colors, const Vector<Point2> &p_uvs, const Vector<int> &p_bones, const Vector<float> &p_weights) { + // Care must be taken to generate array formats + // in ways where they could be reused, so we will + // put single-occuring elements first, and repeated + // elements later. This way the generated formats are + // the same no matter the length of the arrays. + // This dramatically reduces the amount of pipeline objects + // that need to be created for these formats. + + RendererRD::MeshStorage *mesh_storage = RendererRD::MeshStorage::get_singleton(); + + uint32_t vertex_count = p_points.size(); + uint32_t stride = 2; //vertices always repeat + if ((uint32_t)p_colors.size() == vertex_count || p_colors.size() == 1) { + stride += 4; + } + if ((uint32_t)p_uvs.size() == vertex_count) { + stride += 2; + } + if ((uint32_t)p_bones.size() == vertex_count * 4 && (uint32_t)p_weights.size() == vertex_count * 4) { + stride += 4; + } + + uint32_t buffer_size = stride * p_points.size(); + + Vector<uint8_t> polygon_buffer; + polygon_buffer.resize(buffer_size * sizeof(float)); + Vector<RD::VertexAttribute> descriptions; + descriptions.resize(5); + Vector<RID> buffers; + buffers.resize(5); + + { + uint8_t *r = polygon_buffer.ptrw(); + float *fptr = reinterpret_cast<float *>(r); + uint32_t *uptr = reinterpret_cast<uint32_t *>(r); + uint32_t base_offset = 0; + { //vertices + RD::VertexAttribute vd; + vd.format = RD::DATA_FORMAT_R32G32_SFLOAT; + vd.offset = base_offset * sizeof(float); + vd.location = RS::ARRAY_VERTEX; + vd.stride = stride * sizeof(float); + + descriptions.write[0] = vd; + + const Vector2 *points_ptr = p_points.ptr(); + + for (uint32_t i = 0; i < vertex_count; i++) { + fptr[base_offset + i * stride + 0] = points_ptr[i].x; + fptr[base_offset + i * stride + 1] = points_ptr[i].y; + } + + base_offset += 2; + } + + //colors + if ((uint32_t)p_colors.size() == vertex_count || p_colors.size() == 1) { + RD::VertexAttribute vd; + vd.format = RD::DATA_FORMAT_R32G32B32A32_SFLOAT; + vd.offset = base_offset * sizeof(float); + vd.location = RS::ARRAY_COLOR; + vd.stride = stride * sizeof(float); + + descriptions.write[1] = vd; + + if (p_colors.size() == 1) { + Color color = p_colors[0]; + for (uint32_t i = 0; i < vertex_count; i++) { + fptr[base_offset + i * stride + 0] = color.r; + fptr[base_offset + i * stride + 1] = color.g; + fptr[base_offset + i * stride + 2] = color.b; + fptr[base_offset + i * stride + 3] = color.a; + } + } else { + const Color *color_ptr = p_colors.ptr(); + + for (uint32_t i = 0; i < vertex_count; i++) { + fptr[base_offset + i * stride + 0] = color_ptr[i].r; + fptr[base_offset + i * stride + 1] = color_ptr[i].g; + fptr[base_offset + i * stride + 2] = color_ptr[i].b; + fptr[base_offset + i * stride + 3] = color_ptr[i].a; + } + } + base_offset += 4; + } else { + RD::VertexAttribute vd; + vd.format = RD::DATA_FORMAT_R32G32B32A32_SFLOAT; + vd.offset = 0; + vd.location = RS::ARRAY_COLOR; + vd.stride = 0; + + descriptions.write[1] = vd; + buffers.write[1] = mesh_storage->mesh_get_default_rd_buffer(RendererRD::MeshStorage::DEFAULT_RD_BUFFER_COLOR); + } + + //uvs + if ((uint32_t)p_uvs.size() == vertex_count) { + RD::VertexAttribute vd; + vd.format = RD::DATA_FORMAT_R32G32_SFLOAT; + vd.offset = base_offset * sizeof(float); + vd.location = RS::ARRAY_TEX_UV; + vd.stride = stride * sizeof(float); + + descriptions.write[2] = vd; + + const Vector2 *uv_ptr = p_uvs.ptr(); + + for (uint32_t i = 0; i < vertex_count; i++) { + fptr[base_offset + i * stride + 0] = uv_ptr[i].x; + fptr[base_offset + i * stride + 1] = uv_ptr[i].y; + } + base_offset += 2; + } else { + RD::VertexAttribute vd; + vd.format = RD::DATA_FORMAT_R32G32_SFLOAT; + vd.offset = 0; + vd.location = RS::ARRAY_TEX_UV; + vd.stride = 0; + + descriptions.write[2] = vd; + buffers.write[2] = mesh_storage->mesh_get_default_rd_buffer(RendererRD::MeshStorage::DEFAULT_RD_BUFFER_TEX_UV); + } + + //bones + if ((uint32_t)p_indices.size() == vertex_count * 4 && (uint32_t)p_weights.size() == vertex_count * 4) { + RD::VertexAttribute vd; + vd.format = RD::DATA_FORMAT_R16G16B16A16_UINT; + vd.offset = base_offset * sizeof(float); + vd.location = RS::ARRAY_BONES; + vd.stride = stride * sizeof(float); + + descriptions.write[3] = vd; + + const int *bone_ptr = p_bones.ptr(); + + for (uint32_t i = 0; i < vertex_count; i++) { + uint16_t *bone16w = (uint16_t *)&uptr[base_offset + i * stride]; + + bone16w[0] = bone_ptr[i * 4 + 0]; + bone16w[1] = bone_ptr[i * 4 + 1]; + bone16w[2] = bone_ptr[i * 4 + 2]; + bone16w[3] = bone_ptr[i * 4 + 3]; + } + + base_offset += 2; + } else { + RD::VertexAttribute vd; + vd.format = RD::DATA_FORMAT_R32G32B32A32_UINT; + vd.offset = 0; + vd.location = RS::ARRAY_BONES; + vd.stride = 0; + + descriptions.write[3] = vd; + buffers.write[3] = mesh_storage->mesh_get_default_rd_buffer(RendererRD::MeshStorage::DEFAULT_RD_BUFFER_BONES); + } + + //weights + if ((uint32_t)p_weights.size() == vertex_count * 4) { + RD::VertexAttribute vd; + vd.format = RD::DATA_FORMAT_R16G16B16A16_UNORM; + vd.offset = base_offset * sizeof(float); + vd.location = RS::ARRAY_WEIGHTS; + vd.stride = stride * sizeof(float); + + descriptions.write[4] = vd; + + const float *weight_ptr = p_weights.ptr(); + + for (uint32_t i = 0; i < vertex_count; i++) { + uint16_t *weight16w = (uint16_t *)&uptr[base_offset + i * stride]; + + weight16w[0] = CLAMP(weight_ptr[i * 4 + 0] * 65535, 0, 65535); + weight16w[1] = CLAMP(weight_ptr[i * 4 + 1] * 65535, 0, 65535); + weight16w[2] = CLAMP(weight_ptr[i * 4 + 2] * 65535, 0, 65535); + weight16w[3] = CLAMP(weight_ptr[i * 4 + 3] * 65535, 0, 65535); + } + + base_offset += 2; + } else { + RD::VertexAttribute vd; + vd.format = RD::DATA_FORMAT_R32G32B32A32_SFLOAT; + vd.offset = 0; + vd.location = RS::ARRAY_WEIGHTS; + vd.stride = 0; + + descriptions.write[4] = vd; + buffers.write[4] = mesh_storage->mesh_get_default_rd_buffer(RendererRD::MeshStorage::DEFAULT_RD_BUFFER_WEIGHTS); + } + + //check that everything is as it should be + ERR_FAIL_COND_V(base_offset != stride, 0); //bug + } + + RD::VertexFormatID vertex_id = RD::get_singleton()->vertex_format_create(descriptions); + ERR_FAIL_COND_V(vertex_id == RD::INVALID_ID, 0); + + PolygonBuffers pb; + pb.vertex_buffer = RD::get_singleton()->vertex_buffer_create(polygon_buffer.size(), polygon_buffer); + for (int i = 0; i < descriptions.size(); i++) { + if (buffers[i] == RID()) { //if put in vertex, use as vertex + buffers.write[i] = pb.vertex_buffer; + } + } + + pb.vertex_array = RD::get_singleton()->vertex_array_create(p_points.size(), vertex_id, buffers); + + if (p_indices.size()) { + //create indices, as indices were requested + Vector<uint8_t> index_buffer; + index_buffer.resize(p_indices.size() * sizeof(int32_t)); + { + uint8_t *w = index_buffer.ptrw(); + memcpy(w, p_indices.ptr(), sizeof(int32_t) * p_indices.size()); + } + pb.index_buffer = RD::get_singleton()->index_buffer_create(p_indices.size(), RD::INDEX_BUFFER_FORMAT_UINT32, index_buffer); + pb.indices = RD::get_singleton()->index_array_create(pb.index_buffer, 0, p_indices.size()); + } + + pb.vertex_format_id = vertex_id; + + PolygonID id = polygon_buffers.last_id++; + + polygon_buffers.polygons[id] = pb; + + return id; +} + +void RendererCanvasRenderRD::free_polygon(PolygonID p_polygon) { + PolygonBuffers *pb_ptr = polygon_buffers.polygons.getptr(p_polygon); + ERR_FAIL_COND(!pb_ptr); + + PolygonBuffers &pb = *pb_ptr; + + if (pb.indices.is_valid()) { + RD::get_singleton()->free(pb.indices); + } + if (pb.index_buffer.is_valid()) { + RD::get_singleton()->free(pb.index_buffer); + } + + RD::get_singleton()->free(pb.vertex_array); + RD::get_singleton()->free(pb.vertex_buffer); + + polygon_buffers.polygons.erase(p_polygon); +} + +//////////////////// + +void RendererCanvasRenderRD::_bind_canvas_texture(RD::DrawListID p_draw_list, RID p_texture, RS::CanvasItemTextureFilter p_base_filter, RS::CanvasItemTextureRepeat p_base_repeat, RID &r_last_texture, PushConstant &push_constant, Size2 &r_texpixel_size) { + if (p_texture == RID()) { + p_texture = default_canvas_texture; + } + + if (r_last_texture == p_texture) { + return; //nothing to do, its the same + } + + RID uniform_set; + Color specular_shininess; + Size2i size; + bool use_normal; + bool use_specular; + + bool success = RendererRD::TextureStorage::get_singleton()->canvas_texture_get_uniform_set(p_texture, p_base_filter, p_base_repeat, shader.default_version_rd_shader, CANVAS_TEXTURE_UNIFORM_SET, uniform_set, size, specular_shininess, use_normal, use_specular); + //something odd happened + if (!success) { + _bind_canvas_texture(p_draw_list, default_canvas_texture, p_base_filter, p_base_repeat, r_last_texture, push_constant, r_texpixel_size); + return; + } + + RD::get_singleton()->draw_list_bind_uniform_set(p_draw_list, uniform_set, CANVAS_TEXTURE_UNIFORM_SET); + + if (specular_shininess.a < 0.999) { + push_constant.flags |= FLAGS_DEFAULT_SPECULAR_MAP_USED; + } else { + push_constant.flags &= ~FLAGS_DEFAULT_SPECULAR_MAP_USED; + } + + if (use_normal) { + push_constant.flags |= FLAGS_DEFAULT_NORMAL_MAP_USED; + } else { + push_constant.flags &= ~FLAGS_DEFAULT_NORMAL_MAP_USED; + } + + push_constant.specular_shininess = uint32_t(CLAMP(specular_shininess.a * 255.0, 0, 255)) << 24; + push_constant.specular_shininess |= uint32_t(CLAMP(specular_shininess.b * 255.0, 0, 255)) << 16; + push_constant.specular_shininess |= uint32_t(CLAMP(specular_shininess.g * 255.0, 0, 255)) << 8; + push_constant.specular_shininess |= uint32_t(CLAMP(specular_shininess.r * 255.0, 0, 255)); + + r_texpixel_size.x = 1.0 / float(size.x); + r_texpixel_size.y = 1.0 / float(size.y); + + push_constant.color_texture_pixel_size[0] = r_texpixel_size.x; + push_constant.color_texture_pixel_size[1] = r_texpixel_size.y; + + r_last_texture = p_texture; +} + +void RendererCanvasRenderRD::_render_item(RD::DrawListID p_draw_list, RID p_render_target, const Item *p_item, RD::FramebufferFormatID p_framebuffer_format, const Transform2D &p_canvas_transform_inverse, Item *¤t_clip, Light *p_lights, PipelineVariants *p_pipeline_variants) { + //create an empty push constant + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + RendererRD::MeshStorage *mesh_storage = RendererRD::MeshStorage::get_singleton(); + RendererRD::ParticlesStorage *particles_storage = RendererRD::ParticlesStorage::get_singleton(); + + RS::CanvasItemTextureFilter current_filter = default_filter; + RS::CanvasItemTextureRepeat current_repeat = default_repeat; + + if (p_item->texture_filter != RS::CANVAS_ITEM_TEXTURE_FILTER_DEFAULT) { + current_filter = p_item->texture_filter; + } + + if (p_item->texture_repeat != RS::CANVAS_ITEM_TEXTURE_REPEAT_DEFAULT) { + current_repeat = p_item->texture_repeat; + } + + PushConstant push_constant; + Transform2D base_transform = p_canvas_transform_inverse * p_item->final_transform; + Transform2D draw_transform; + _update_transform_2d_to_mat2x3(base_transform, push_constant.world); + + Color base_color = p_item->final_modulate; + + for (int i = 0; i < 4; i++) { + push_constant.modulation[i] = 0; + push_constant.ninepatch_margins[i] = 0; + push_constant.src_rect[i] = 0; + push_constant.dst_rect[i] = 0; + } + push_constant.flags = 0; + push_constant.color_texture_pixel_size[0] = 0; + push_constant.color_texture_pixel_size[1] = 0; + + push_constant.pad[0] = 0; + push_constant.pad[1] = 0; + + push_constant.lights[0] = 0; + push_constant.lights[1] = 0; + push_constant.lights[2] = 0; + push_constant.lights[3] = 0; + + uint32_t base_flags = 0; + + uint16_t light_count = 0; + PipelineLightMode light_mode; + + { + Light *light = p_lights; + + while (light) { + if (light->render_index_cache >= 0 && p_item->light_mask & light->item_mask && p_item->z_final >= light->z_min && p_item->z_final <= light->z_max && p_item->global_rect_cache.intersects_transformed(light->xform_cache, light->rect_cache)) { + uint32_t light_index = light->render_index_cache; + push_constant.lights[light_count >> 2] |= light_index << ((light_count & 3) * 8); + + light_count++; + + if (light_count == MAX_LIGHTS_PER_ITEM) { + break; + } + } + light = light->next_ptr; + } + + base_flags |= light_count << FLAGS_LIGHT_COUNT_SHIFT; + } + + light_mode = (light_count > 0 || using_directional_lights) ? PIPELINE_LIGHT_MODE_ENABLED : PIPELINE_LIGHT_MODE_DISABLED; + + PipelineVariants *pipeline_variants = p_pipeline_variants; + + bool reclip = false; + + RID last_texture; + Size2 texpixel_size; + + bool skipping = false; + + const Item::Command *c = p_item->commands; + while (c) { + if (skipping && c->type != Item::Command::TYPE_ANIMATION_SLICE) { + c = c->next; + continue; + } + + push_constant.flags = base_flags | (push_constant.flags & (FLAGS_DEFAULT_NORMAL_MAP_USED | FLAGS_DEFAULT_SPECULAR_MAP_USED)); //reset on each command for sanity, keep canvastexture binding config + + switch (c->type) { + case Item::Command::TYPE_RECT: { + const Item::CommandRect *rect = static_cast<const Item::CommandRect *>(c); + + if (rect->flags & CANVAS_RECT_TILE) { + current_repeat = RenderingServer::CanvasItemTextureRepeat::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED; + } + + //bind pipeline + if (rect->flags & CANVAS_RECT_LCD) { + RID pipeline = pipeline_variants->variants[light_mode][PIPELINE_VARIANT_QUAD_LCD_BLEND].get_render_pipeline(RD::INVALID_ID, p_framebuffer_format); + RD::get_singleton()->draw_list_bind_render_pipeline(p_draw_list, pipeline); + RD::get_singleton()->draw_list_set_blend_constants(p_draw_list, rect->modulate); + } else { + RID pipeline = pipeline_variants->variants[light_mode][PIPELINE_VARIANT_QUAD].get_render_pipeline(RD::INVALID_ID, p_framebuffer_format); + RD::get_singleton()->draw_list_bind_render_pipeline(p_draw_list, pipeline); + } + + //bind textures + + _bind_canvas_texture(p_draw_list, rect->texture, current_filter, current_repeat, last_texture, push_constant, texpixel_size); + + Rect2 src_rect; + Rect2 dst_rect; + + if (rect->texture != RID()) { + src_rect = (rect->flags & CANVAS_RECT_REGION) ? Rect2(rect->source.position * texpixel_size, rect->source.size * texpixel_size) : Rect2(0, 0, 1, 1); + dst_rect = Rect2(rect->rect.position, rect->rect.size); + + if (dst_rect.size.width < 0) { + dst_rect.position.x += dst_rect.size.width; + dst_rect.size.width *= -1; + } + if (dst_rect.size.height < 0) { + dst_rect.position.y += dst_rect.size.height; + dst_rect.size.height *= -1; + } + + if (rect->flags & CANVAS_RECT_FLIP_H) { + src_rect.size.x *= -1; + } + + if (rect->flags & CANVAS_RECT_FLIP_V) { + src_rect.size.y *= -1; + } + + if (rect->flags & CANVAS_RECT_TRANSPOSE) { + push_constant.flags |= FLAGS_TRANSPOSE_RECT; + } + + if (rect->flags & CANVAS_RECT_CLIP_UV) { + push_constant.flags |= FLAGS_CLIP_RECT_UV; + } + + } else { + dst_rect = Rect2(rect->rect.position, rect->rect.size); + + if (dst_rect.size.width < 0) { + dst_rect.position.x += dst_rect.size.width; + dst_rect.size.width *= -1; + } + if (dst_rect.size.height < 0) { + dst_rect.position.y += dst_rect.size.height; + dst_rect.size.height *= -1; + } + + src_rect = Rect2(0, 0, 1, 1); + } + + if (rect->flags & CANVAS_RECT_MSDF) { + push_constant.flags |= FLAGS_USE_MSDF; + push_constant.msdf[0] = rect->px_range; // Pixel range. + push_constant.msdf[1] = rect->outline; // Outline size. + push_constant.msdf[2] = 0.f; // Reserved. + push_constant.msdf[3] = 0.f; // Reserved. + } else if (rect->flags & CANVAS_RECT_LCD) { + push_constant.flags |= FLAGS_USE_LCD; + } + + push_constant.modulation[0] = rect->modulate.r * base_color.r; + push_constant.modulation[1] = rect->modulate.g * base_color.g; + push_constant.modulation[2] = rect->modulate.b * base_color.b; + push_constant.modulation[3] = rect->modulate.a * base_color.a; + + push_constant.src_rect[0] = src_rect.position.x; + push_constant.src_rect[1] = src_rect.position.y; + push_constant.src_rect[2] = src_rect.size.width; + push_constant.src_rect[3] = src_rect.size.height; + + push_constant.dst_rect[0] = dst_rect.position.x; + push_constant.dst_rect[1] = dst_rect.position.y; + push_constant.dst_rect[2] = dst_rect.size.width; + push_constant.dst_rect[3] = dst_rect.size.height; + + RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(PushConstant)); + RD::get_singleton()->draw_list_bind_index_array(p_draw_list, shader.quad_index_array); + RD::get_singleton()->draw_list_draw(p_draw_list, true); + + } break; + + case Item::Command::TYPE_NINEPATCH: { + const Item::CommandNinePatch *np = static_cast<const Item::CommandNinePatch *>(c); + + //bind pipeline + { + RID pipeline = pipeline_variants->variants[light_mode][PIPELINE_VARIANT_NINEPATCH].get_render_pipeline(RD::INVALID_ID, p_framebuffer_format); + RD::get_singleton()->draw_list_bind_render_pipeline(p_draw_list, pipeline); + } + + //bind textures + + _bind_canvas_texture(p_draw_list, np->texture, current_filter, current_repeat, last_texture, push_constant, texpixel_size); + + Rect2 src_rect; + Rect2 dst_rect(np->rect.position.x, np->rect.position.y, np->rect.size.x, np->rect.size.y); + + if (np->texture == RID()) { + texpixel_size = Size2(1, 1); + src_rect = Rect2(0, 0, 1, 1); + + } else { + if (np->source != Rect2()) { + src_rect = Rect2(np->source.position.x * texpixel_size.width, np->source.position.y * texpixel_size.height, np->source.size.x * texpixel_size.width, np->source.size.y * texpixel_size.height); + push_constant.color_texture_pixel_size[0] = 1.0 / np->source.size.width; + push_constant.color_texture_pixel_size[1] = 1.0 / np->source.size.height; + + } else { + src_rect = Rect2(0, 0, 1, 1); + } + } + + push_constant.modulation[0] = np->color.r * base_color.r; + push_constant.modulation[1] = np->color.g * base_color.g; + push_constant.modulation[2] = np->color.b * base_color.b; + push_constant.modulation[3] = np->color.a * base_color.a; + + push_constant.src_rect[0] = src_rect.position.x; + push_constant.src_rect[1] = src_rect.position.y; + push_constant.src_rect[2] = src_rect.size.width; + push_constant.src_rect[3] = src_rect.size.height; + + push_constant.dst_rect[0] = dst_rect.position.x; + push_constant.dst_rect[1] = dst_rect.position.y; + push_constant.dst_rect[2] = dst_rect.size.width; + push_constant.dst_rect[3] = dst_rect.size.height; + + push_constant.flags |= int(np->axis_x) << FLAGS_NINEPATCH_H_MODE_SHIFT; + push_constant.flags |= int(np->axis_y) << FLAGS_NINEPATCH_V_MODE_SHIFT; + + if (np->draw_center) { + push_constant.flags |= FLAGS_NINEPACH_DRAW_CENTER; + } + + push_constant.ninepatch_margins[0] = np->margin[SIDE_LEFT]; + push_constant.ninepatch_margins[1] = np->margin[SIDE_TOP]; + push_constant.ninepatch_margins[2] = np->margin[SIDE_RIGHT]; + push_constant.ninepatch_margins[3] = np->margin[SIDE_BOTTOM]; + + RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(PushConstant)); + RD::get_singleton()->draw_list_bind_index_array(p_draw_list, shader.quad_index_array); + RD::get_singleton()->draw_list_draw(p_draw_list, true); + + // Restore if overridden. + push_constant.color_texture_pixel_size[0] = texpixel_size.x; + push_constant.color_texture_pixel_size[1] = texpixel_size.y; + + } break; + case Item::Command::TYPE_POLYGON: { + const Item::CommandPolygon *polygon = static_cast<const Item::CommandPolygon *>(c); + + PolygonBuffers *pb = polygon_buffers.polygons.getptr(polygon->polygon.polygon_id); + ERR_CONTINUE(!pb); + //bind pipeline + { + static const PipelineVariant variant[RS::PRIMITIVE_MAX] = { PIPELINE_VARIANT_ATTRIBUTE_POINTS, PIPELINE_VARIANT_ATTRIBUTE_LINES, PIPELINE_VARIANT_ATTRIBUTE_LINES_STRIP, PIPELINE_VARIANT_ATTRIBUTE_TRIANGLES, PIPELINE_VARIANT_ATTRIBUTE_TRIANGLE_STRIP }; + ERR_CONTINUE(polygon->primitive < 0 || polygon->primitive >= RS::PRIMITIVE_MAX); + RID pipeline = pipeline_variants->variants[light_mode][variant[polygon->primitive]].get_render_pipeline(pb->vertex_format_id, p_framebuffer_format); + RD::get_singleton()->draw_list_bind_render_pipeline(p_draw_list, pipeline); + } + + if (polygon->primitive == RS::PRIMITIVE_LINES) { + //not supported in most hardware, so pointless + //RD::get_singleton()->draw_list_set_line_width(p_draw_list, polygon->line_width); + } + + //bind textures + + _bind_canvas_texture(p_draw_list, polygon->texture, current_filter, current_repeat, last_texture, push_constant, texpixel_size); + + push_constant.modulation[0] = base_color.r; + push_constant.modulation[1] = base_color.g; + push_constant.modulation[2] = base_color.b; + push_constant.modulation[3] = base_color.a; + + for (int j = 0; j < 4; j++) { + push_constant.src_rect[j] = 0; + push_constant.dst_rect[j] = 0; + push_constant.ninepatch_margins[j] = 0; + } + + RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(PushConstant)); + RD::get_singleton()->draw_list_bind_vertex_array(p_draw_list, pb->vertex_array); + if (pb->indices.is_valid()) { + RD::get_singleton()->draw_list_bind_index_array(p_draw_list, pb->indices); + } + RD::get_singleton()->draw_list_draw(p_draw_list, pb->indices.is_valid()); + + } break; + case Item::Command::TYPE_PRIMITIVE: { + const Item::CommandPrimitive *primitive = static_cast<const Item::CommandPrimitive *>(c); + + //bind pipeline + { + static const PipelineVariant variant[4] = { PIPELINE_VARIANT_PRIMITIVE_POINTS, PIPELINE_VARIANT_PRIMITIVE_LINES, PIPELINE_VARIANT_PRIMITIVE_TRIANGLES, PIPELINE_VARIANT_PRIMITIVE_TRIANGLES }; + ERR_CONTINUE(primitive->point_count == 0 || primitive->point_count > 4); + RID pipeline = pipeline_variants->variants[light_mode][variant[primitive->point_count - 1]].get_render_pipeline(RD::INVALID_ID, p_framebuffer_format); + RD::get_singleton()->draw_list_bind_render_pipeline(p_draw_list, pipeline); + } + + //bind textures + + _bind_canvas_texture(p_draw_list, RID(), current_filter, current_repeat, last_texture, push_constant, texpixel_size); + + RD::get_singleton()->draw_list_bind_index_array(p_draw_list, primitive_arrays.index_array[MIN(3u, primitive->point_count) - 1]); + + for (uint32_t j = 0; j < MIN(3u, primitive->point_count); j++) { + push_constant.points[j * 2 + 0] = primitive->points[j].x; + push_constant.points[j * 2 + 1] = primitive->points[j].y; + push_constant.uvs[j * 2 + 0] = primitive->uvs[j].x; + push_constant.uvs[j * 2 + 1] = primitive->uvs[j].y; + Color col = primitive->colors[j] * base_color; + push_constant.colors[j * 2 + 0] = (uint32_t(Math::make_half_float(col.g)) << 16) | Math::make_half_float(col.r); + push_constant.colors[j * 2 + 1] = (uint32_t(Math::make_half_float(col.a)) << 16) | Math::make_half_float(col.b); + } + RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(PushConstant)); + RD::get_singleton()->draw_list_draw(p_draw_list, true); + + if (primitive->point_count == 4) { + for (uint32_t j = 1; j < 3; j++) { + //second half of triangle + push_constant.points[j * 2 + 0] = primitive->points[j + 1].x; + push_constant.points[j * 2 + 1] = primitive->points[j + 1].y; + push_constant.uvs[j * 2 + 0] = primitive->uvs[j + 1].x; + push_constant.uvs[j * 2 + 1] = primitive->uvs[j + 1].y; + Color col = primitive->colors[j + 1] * base_color; + push_constant.colors[j * 2 + 0] = (uint32_t(Math::make_half_float(col.g)) << 16) | Math::make_half_float(col.r); + push_constant.colors[j * 2 + 1] = (uint32_t(Math::make_half_float(col.a)) << 16) | Math::make_half_float(col.b); + } + + RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(PushConstant)); + RD::get_singleton()->draw_list_draw(p_draw_list, true); + } + + } break; + case Item::Command::TYPE_MESH: + case Item::Command::TYPE_MULTIMESH: + case Item::Command::TYPE_PARTICLES: { + RID mesh; + RID mesh_instance; + RID texture; + Color modulate(1, 1, 1, 1); + float world_backup[6]; + int instance_count = 1; + + for (int j = 0; j < 6; j++) { + world_backup[j] = push_constant.world[j]; + } + + if (c->type == Item::Command::TYPE_MESH) { + const Item::CommandMesh *m = static_cast<const Item::CommandMesh *>(c); + mesh = m->mesh; + mesh_instance = m->mesh_instance; + texture = m->texture; + modulate = m->modulate; + _update_transform_2d_to_mat2x3(base_transform * draw_transform * m->transform, push_constant.world); + } else if (c->type == Item::Command::TYPE_MULTIMESH) { + const Item::CommandMultiMesh *mm = static_cast<const Item::CommandMultiMesh *>(c); + RID multimesh = mm->multimesh; + mesh = mesh_storage->multimesh_get_mesh(multimesh); + texture = mm->texture; + + if (mesh_storage->multimesh_get_transform_format(multimesh) != RS::MULTIMESH_TRANSFORM_2D) { + break; + } + + instance_count = mesh_storage->multimesh_get_instances_to_draw(multimesh); + + if (instance_count == 0) { + break; + } + + RID uniform_set = mesh_storage->multimesh_get_2d_uniform_set(multimesh, shader.default_version_rd_shader, TRANSFORMS_UNIFORM_SET); + RD::get_singleton()->draw_list_bind_uniform_set(p_draw_list, uniform_set, TRANSFORMS_UNIFORM_SET); + push_constant.flags |= 1; //multimesh, trails disabled + if (mesh_storage->multimesh_uses_colors(multimesh)) { + push_constant.flags |= FLAGS_INSTANCING_HAS_COLORS; + } + if (mesh_storage->multimesh_uses_custom_data(multimesh)) { + push_constant.flags |= FLAGS_INSTANCING_HAS_CUSTOM_DATA; + } + } else if (c->type == Item::Command::TYPE_PARTICLES) { + const Item::CommandParticles *pt = static_cast<const Item::CommandParticles *>(c); + ERR_BREAK(particles_storage->particles_get_mode(pt->particles) != RS::PARTICLES_MODE_2D); + particles_storage->particles_request_process(pt->particles); + + if (particles_storage->particles_is_inactive(pt->particles)) { + break; + } + + RenderingServerDefault::redraw_request(); // active particles means redraw request + + bool local_coords = true; + int dpc = particles_storage->particles_get_draw_passes(pt->particles); + if (dpc == 0) { + break; //nothing to draw + } + uint32_t divisor = 1; + instance_count = particles_storage->particles_get_amount(pt->particles, divisor); + + RID uniform_set = particles_storage->particles_get_instance_buffer_uniform_set(pt->particles, shader.default_version_rd_shader, TRANSFORMS_UNIFORM_SET); + RD::get_singleton()->draw_list_bind_uniform_set(p_draw_list, uniform_set, TRANSFORMS_UNIFORM_SET); + + push_constant.flags |= divisor; + instance_count /= divisor; + + push_constant.flags |= FLAGS_INSTANCING_HAS_COLORS; + push_constant.flags |= FLAGS_INSTANCING_HAS_CUSTOM_DATA; + + mesh = particles_storage->particles_get_draw_pass_mesh(pt->particles, 0); //higher ones are ignored + texture = pt->texture; + + if (particles_storage->particles_has_collision(pt->particles) && texture_storage->render_target_is_sdf_enabled(p_render_target)) { + //pass collision information + Transform2D xform; + if (local_coords) { + xform = p_item->final_transform; + } else { + xform = p_canvas_transform_inverse; + } + + RID sdf_texture = texture_storage->render_target_get_sdf_texture(p_render_target); + + Rect2 to_screen; + { + Rect2 sdf_rect = texture_storage->render_target_get_sdf_rect(p_render_target); + + to_screen.size = Vector2(1.0 / sdf_rect.size.width, 1.0 / sdf_rect.size.height); + to_screen.position = -sdf_rect.position * to_screen.size; + } + + particles_storage->particles_set_canvas_sdf_collision(pt->particles, true, xform, to_screen, sdf_texture); + } else { + particles_storage->particles_set_canvas_sdf_collision(pt->particles, false, Transform2D(), Rect2(), RID()); + } + } + + if (mesh.is_null()) { + break; + } + + _bind_canvas_texture(p_draw_list, texture, current_filter, current_repeat, last_texture, push_constant, texpixel_size); + + uint32_t surf_count = mesh_storage->mesh_get_surface_count(mesh); + static const PipelineVariant variant[RS::PRIMITIVE_MAX] = { PIPELINE_VARIANT_ATTRIBUTE_POINTS, PIPELINE_VARIANT_ATTRIBUTE_LINES, PIPELINE_VARIANT_ATTRIBUTE_LINES_STRIP, PIPELINE_VARIANT_ATTRIBUTE_TRIANGLES, PIPELINE_VARIANT_ATTRIBUTE_TRIANGLE_STRIP }; + + push_constant.modulation[0] = base_color.r * modulate.r; + push_constant.modulation[1] = base_color.g * modulate.g; + push_constant.modulation[2] = base_color.b * modulate.b; + push_constant.modulation[3] = base_color.a * modulate.a; + + for (int j = 0; j < 4; j++) { + push_constant.src_rect[j] = 0; + push_constant.dst_rect[j] = 0; + push_constant.ninepatch_margins[j] = 0; + } + + for (uint32_t j = 0; j < surf_count; j++) { + void *surface = mesh_storage->mesh_get_surface(mesh, j); + + RS::PrimitiveType primitive = mesh_storage->mesh_surface_get_primitive(surface); + ERR_CONTINUE(primitive < 0 || primitive >= RS::PRIMITIVE_MAX); + + uint32_t input_mask = pipeline_variants->variants[light_mode][variant[primitive]].get_vertex_input_mask(); + + RID vertex_array; + RD::VertexFormatID vertex_format = RD::INVALID_FORMAT_ID; + + if (mesh_instance.is_valid()) { + mesh_storage->mesh_instance_surface_get_vertex_arrays_and_format(mesh_instance, j, input_mask, vertex_array, vertex_format); + } else { + mesh_storage->mesh_surface_get_vertex_arrays_and_format(surface, input_mask, vertex_array, vertex_format); + } + + RID pipeline = pipeline_variants->variants[light_mode][variant[primitive]].get_render_pipeline(vertex_format, p_framebuffer_format); + RD::get_singleton()->draw_list_bind_render_pipeline(p_draw_list, pipeline); + + RID index_array = mesh_storage->mesh_surface_get_index_array(surface, 0); + + if (index_array.is_valid()) { + RD::get_singleton()->draw_list_bind_index_array(p_draw_list, index_array); + } + + RD::get_singleton()->draw_list_bind_vertex_array(p_draw_list, vertex_array); + RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(PushConstant)); + + RD::get_singleton()->draw_list_draw(p_draw_list, index_array.is_valid(), instance_count); + } + + for (int j = 0; j < 6; j++) { + push_constant.world[j] = world_backup[j]; + } + } break; + case Item::Command::TYPE_TRANSFORM: { + const Item::CommandTransform *transform = static_cast<const Item::CommandTransform *>(c); + draw_transform = transform->xform; + _update_transform_2d_to_mat2x3(base_transform * transform->xform, push_constant.world); + + } break; + case Item::Command::TYPE_CLIP_IGNORE: { + const Item::CommandClipIgnore *ci = static_cast<const Item::CommandClipIgnore *>(c); + if (current_clip) { + if (ci->ignore != reclip) { + if (ci->ignore) { + RD::get_singleton()->draw_list_disable_scissor(p_draw_list); + reclip = true; + } else { + RD::get_singleton()->draw_list_enable_scissor(p_draw_list, current_clip->final_clip_rect); + reclip = false; + } + } + } + + } break; + case Item::Command::TYPE_ANIMATION_SLICE: { + const Item::CommandAnimationSlice *as = static_cast<const Item::CommandAnimationSlice *>(c); + double current_time = RendererCompositorRD::singleton->get_total_time(); + double local_time = Math::fposmod(current_time - as->offset, as->animation_length); + skipping = !(local_time >= as->slice_begin && local_time < as->slice_end); + + RenderingServerDefault::redraw_request(); // animation visible means redraw request + } break; + } + + c = c->next; + } + + if (current_clip && reclip) { + //will make it re-enable clipping if needed afterwards + current_clip = nullptr; + } +} + +RID RendererCanvasRenderRD::_create_base_uniform_set(RID p_to_render_target, bool p_backbuffer) { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + + //re create canvas state + Vector<RD::Uniform> uniforms; + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 1; + u.append_id(state.canvas_state_buffer); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 2; + u.append_id(state.lights_uniform_buffer); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 3; + u.append_id(RendererRD::TextureStorage::get_singleton()->decal_atlas_get_texture()); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 4; + u.append_id(state.shadow_texture); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 5; + u.append_id(state.shadow_sampler); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 6; + RID screen; + if (p_backbuffer) { + screen = texture_storage->render_target_get_rd_texture(p_to_render_target); + } else { + screen = texture_storage->render_target_get_rd_backbuffer(p_to_render_target); + if (screen.is_null()) { //unallocated backbuffer + screen = RendererRD::TextureStorage::get_singleton()->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_WHITE); + } + } + u.append_id(screen); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 7; + RID sdf = texture_storage->render_target_get_sdf_texture(p_to_render_target); + u.append_id(sdf); + uniforms.push_back(u); + } + + { + //needs samplers for the material (uses custom textures) create them + Vector<RID> ids; + ids.resize(12); + RID *ids_ptr = ids.ptrw(); + ids_ptr[0] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[1] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[2] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[3] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[4] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[5] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[6] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[7] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[8] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[9] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[10] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[11] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + + RD::Uniform u(RD::UNIFORM_TYPE_SAMPLER, 8, ids); + + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 9; + u.append_id(RendererRD::MaterialStorage::get_singleton()->global_shader_uniforms_get_storage_buffer()); + uniforms.push_back(u); + } + + RID uniform_set = RD::get_singleton()->uniform_set_create(uniforms, shader.default_version_rd_shader, BASE_UNIFORM_SET); + if (p_backbuffer) { + texture_storage->render_target_set_backbuffer_uniform_set(p_to_render_target, uniform_set); + } else { + texture_storage->render_target_set_framebuffer_uniform_set(p_to_render_target, uniform_set); + } + + return uniform_set; +} + +void RendererCanvasRenderRD::_render_items(RID p_to_render_target, int p_item_count, const Transform2D &p_canvas_transform_inverse, Light *p_lights, bool p_to_backbuffer) { + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + + Item *current_clip = nullptr; + + Transform2D canvas_transform_inverse = p_canvas_transform_inverse; + + RID framebuffer; + RID fb_uniform_set; + bool clear = false; + Vector<Color> clear_colors; + + if (p_to_backbuffer) { + framebuffer = texture_storage->render_target_get_rd_backbuffer_framebuffer(p_to_render_target); + fb_uniform_set = texture_storage->render_target_get_backbuffer_uniform_set(p_to_render_target); + } else { + framebuffer = texture_storage->render_target_get_rd_framebuffer(p_to_render_target); + + if (texture_storage->render_target_is_clear_requested(p_to_render_target)) { + clear = true; + clear_colors.push_back(texture_storage->render_target_get_clear_request_color(p_to_render_target)); + texture_storage->render_target_disable_clear_request(p_to_render_target); + } + // TODO: Obtain from framebuffer format eventually when this is implemented. + fb_uniform_set = texture_storage->render_target_get_framebuffer_uniform_set(p_to_render_target); + } + + if (fb_uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(fb_uniform_set)) { + fb_uniform_set = _create_base_uniform_set(p_to_render_target, p_to_backbuffer); + } + + RD::FramebufferFormatID fb_format = RD::get_singleton()->framebuffer_get_format(framebuffer); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(framebuffer, clear ? RD::INITIAL_ACTION_CLEAR : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD, clear_colors); + + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, fb_uniform_set, BASE_UNIFORM_SET); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, state.default_transforms_uniform_set, TRANSFORMS_UNIFORM_SET); + + RID prev_material; + + PipelineVariants *pipeline_variants = &shader.pipeline_variants; + + for (int i = 0; i < p_item_count; i++) { + Item *ci = items[i]; + + if (current_clip != ci->final_clip_owner) { + current_clip = ci->final_clip_owner; + + //setup clip + if (current_clip) { + RD::get_singleton()->draw_list_enable_scissor(draw_list, current_clip->final_clip_rect); + + } else { + RD::get_singleton()->draw_list_disable_scissor(draw_list); + } + } + + RID material = ci->material_owner == nullptr ? ci->material : ci->material_owner->material; + + if (ci->canvas_group != nullptr) { + if (ci->canvas_group->mode == RS::CANVAS_GROUP_MODE_CLIP_AND_DRAW) { + if (!p_to_backbuffer) { + material = default_clip_children_material; + } + } else { + if (material.is_null()) { + if (ci->canvas_group->mode == RS::CANVAS_GROUP_MODE_CLIP_ONLY) { + material = default_clip_children_material; + } else { + material = default_canvas_group_material; + } + } + } + } + + if (material != prev_material) { + CanvasMaterialData *material_data = nullptr; + if (material.is_valid()) { + material_data = static_cast<CanvasMaterialData *>(material_storage->material_get_data(material, RendererRD::MaterialStorage::SHADER_TYPE_2D)); + } + + if (material_data) { + if (material_data->shader_data->version.is_valid() && material_data->shader_data->valid) { + pipeline_variants = &material_data->shader_data->pipeline_variants; + // Update uniform set. + if (material_data->uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(material_data->uniform_set)) { // Material may not have a uniform set. + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, material_data->uniform_set, MATERIAL_UNIFORM_SET); + } + } else { + pipeline_variants = &shader.pipeline_variants; + } + } else { + pipeline_variants = &shader.pipeline_variants; + } + } + + _render_item(draw_list, p_to_render_target, ci, fb_format, canvas_transform_inverse, current_clip, p_lights, pipeline_variants); + + prev_material = material; + } + + RD::get_singleton()->draw_list_end(); +} + +void RendererCanvasRenderRD::canvas_render_items(RID p_to_render_target, Item *p_item_list, const Color &p_modulate, Light *p_light_list, Light *p_directional_light_list, const Transform2D &p_canvas_transform, RenderingServer::CanvasItemTextureFilter p_default_filter, RenderingServer::CanvasItemTextureRepeat p_default_repeat, bool p_snap_2d_vertices_to_pixel, bool &r_sdf_used) { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + RendererRD::MeshStorage *mesh_storage = RendererRD::MeshStorage::get_singleton(); + + r_sdf_used = false; + int item_count = 0; + + //setup canvas state uniforms if needed + + Transform2D canvas_transform_inverse = p_canvas_transform.affine_inverse(); + + //setup directional lights if exist + + uint32_t light_count = 0; + uint32_t directional_light_count = 0; + { + Light *l = p_directional_light_list; + uint32_t index = 0; + + while (l) { + if (index == state.max_lights_per_render) { + l->render_index_cache = -1; + l = l->next_ptr; + continue; + } + + CanvasLight *clight = canvas_light_owner.get_or_null(l->light_internal); + if (!clight) { //unused or invalid texture + l->render_index_cache = -1; + l = l->next_ptr; + ERR_CONTINUE(!clight); + } + + Vector2 canvas_light_dir = l->xform_cache.columns[1].normalized(); + + state.light_uniforms[index].position[0] = -canvas_light_dir.x; + state.light_uniforms[index].position[1] = -canvas_light_dir.y; + + _update_transform_2d_to_mat2x4(clight->shadow.directional_xform, state.light_uniforms[index].shadow_matrix); + + state.light_uniforms[index].height = l->height; //0..1 here + + for (int i = 0; i < 4; i++) { + state.light_uniforms[index].shadow_color[i] = uint8_t(CLAMP(int32_t(l->shadow_color[i] * 255.0), 0, 255)); + state.light_uniforms[index].color[i] = l->color[i]; + } + + state.light_uniforms[index].color[3] = l->energy; //use alpha for energy, so base color can go separate + + if (state.shadow_fb.is_valid()) { + state.light_uniforms[index].shadow_pixel_size = (1.0 / state.shadow_texture_size) * (1.0 + l->shadow_smooth); + state.light_uniforms[index].shadow_z_far_inv = 1.0 / clight->shadow.z_far; + state.light_uniforms[index].shadow_y_ofs = clight->shadow.y_offset; + } else { + state.light_uniforms[index].shadow_pixel_size = 1.0; + state.light_uniforms[index].shadow_z_far_inv = 1.0; + state.light_uniforms[index].shadow_y_ofs = 0; + } + + state.light_uniforms[index].flags = l->blend_mode << LIGHT_FLAGS_BLEND_SHIFT; + state.light_uniforms[index].flags |= l->shadow_filter << LIGHT_FLAGS_FILTER_SHIFT; + if (clight->shadow.enabled) { + state.light_uniforms[index].flags |= LIGHT_FLAGS_HAS_SHADOW; + } + + l->render_index_cache = index; + + index++; + l = l->next_ptr; + } + + light_count = index; + directional_light_count = light_count; + using_directional_lights = directional_light_count > 0; + } + + //setup lights if exist + + { + Light *l = p_light_list; + uint32_t index = light_count; + + while (l) { + if (index == state.max_lights_per_render) { + l->render_index_cache = -1; + l = l->next_ptr; + continue; + } + + CanvasLight *clight = canvas_light_owner.get_or_null(l->light_internal); + if (!clight) { //unused or invalid texture + l->render_index_cache = -1; + l = l->next_ptr; + ERR_CONTINUE(!clight); + } + Transform2D to_light_xform = (p_canvas_transform * l->light_shader_xform).affine_inverse(); + + Vector2 canvas_light_pos = p_canvas_transform.xform(l->xform.get_origin()); //convert light position to canvas coordinates, as all computation is done in canvas coords to avoid precision loss + state.light_uniforms[index].position[0] = canvas_light_pos.x; + state.light_uniforms[index].position[1] = canvas_light_pos.y; + + _update_transform_2d_to_mat2x4(to_light_xform, state.light_uniforms[index].matrix); + _update_transform_2d_to_mat2x4(l->xform_cache.affine_inverse(), state.light_uniforms[index].shadow_matrix); + + state.light_uniforms[index].height = l->height * (p_canvas_transform.columns[0].length() + p_canvas_transform.columns[1].length()) * 0.5; //approximate height conversion to the canvas size, since all calculations are done in canvas coords to avoid precision loss + for (int i = 0; i < 4; i++) { + state.light_uniforms[index].shadow_color[i] = uint8_t(CLAMP(int32_t(l->shadow_color[i] * 255.0), 0, 255)); + state.light_uniforms[index].color[i] = l->color[i]; + } + + state.light_uniforms[index].color[3] = l->energy; //use alpha for energy, so base color can go separate + + if (state.shadow_fb.is_valid()) { + state.light_uniforms[index].shadow_pixel_size = (1.0 / state.shadow_texture_size) * (1.0 + l->shadow_smooth); + state.light_uniforms[index].shadow_z_far_inv = 1.0 / clight->shadow.z_far; + state.light_uniforms[index].shadow_y_ofs = clight->shadow.y_offset; + } else { + state.light_uniforms[index].shadow_pixel_size = 1.0; + state.light_uniforms[index].shadow_z_far_inv = 1.0; + state.light_uniforms[index].shadow_y_ofs = 0; + } + + state.light_uniforms[index].flags = l->blend_mode << LIGHT_FLAGS_BLEND_SHIFT; + state.light_uniforms[index].flags |= l->shadow_filter << LIGHT_FLAGS_FILTER_SHIFT; + if (clight->shadow.enabled) { + state.light_uniforms[index].flags |= LIGHT_FLAGS_HAS_SHADOW; + } + + if (clight->texture.is_valid()) { + Rect2 atlas_rect = RendererRD::TextureStorage::get_singleton()->decal_atlas_get_texture_rect(clight->texture); + state.light_uniforms[index].atlas_rect[0] = atlas_rect.position.x; + state.light_uniforms[index].atlas_rect[1] = atlas_rect.position.y; + state.light_uniforms[index].atlas_rect[2] = atlas_rect.size.width; + state.light_uniforms[index].atlas_rect[3] = atlas_rect.size.height; + + } else { + state.light_uniforms[index].atlas_rect[0] = 0; + state.light_uniforms[index].atlas_rect[1] = 0; + state.light_uniforms[index].atlas_rect[2] = 0; + state.light_uniforms[index].atlas_rect[3] = 0; + } + + l->render_index_cache = index; + + index++; + l = l->next_ptr; + } + + light_count = index; + } + + if (light_count > 0) { + RD::get_singleton()->buffer_update(state.lights_uniform_buffer, 0, sizeof(LightUniform) * light_count, &state.light_uniforms[0]); + } + + { + //update canvas state uniform buffer + State::Buffer state_buffer; + + Size2i ssize = texture_storage->render_target_get_size(p_to_render_target); + + Transform3D screen_transform; + screen_transform.translate_local(-(ssize.width / 2.0f), -(ssize.height / 2.0f), 0.0f); + screen_transform.scale(Vector3(2.0f / ssize.width, 2.0f / ssize.height, 1.0f)); + _update_transform_to_mat4(screen_transform, state_buffer.screen_transform); + _update_transform_2d_to_mat4(p_canvas_transform, state_buffer.canvas_transform); + + Transform2D normal_transform = p_canvas_transform; + normal_transform.columns[0].normalize(); + normal_transform.columns[1].normalize(); + normal_transform.columns[2] = Vector2(); + _update_transform_2d_to_mat4(normal_transform, state_buffer.canvas_normal_transform); + + state_buffer.canvas_modulate[0] = p_modulate.r; + state_buffer.canvas_modulate[1] = p_modulate.g; + state_buffer.canvas_modulate[2] = p_modulate.b; + state_buffer.canvas_modulate[3] = p_modulate.a; + + Size2 render_target_size = texture_storage->render_target_get_size(p_to_render_target); + state_buffer.screen_pixel_size[0] = 1.0 / render_target_size.x; + state_buffer.screen_pixel_size[1] = 1.0 / render_target_size.y; + + state_buffer.time = state.time; + state_buffer.use_pixel_snap = p_snap_2d_vertices_to_pixel; + + state_buffer.directional_light_count = directional_light_count; + + Vector2 canvas_scale = p_canvas_transform.get_scale(); + + state_buffer.sdf_to_screen[0] = render_target_size.width / canvas_scale.x; + state_buffer.sdf_to_screen[1] = render_target_size.height / canvas_scale.y; + + state_buffer.screen_to_sdf[0] = 1.0 / state_buffer.sdf_to_screen[0]; + state_buffer.screen_to_sdf[1] = 1.0 / state_buffer.sdf_to_screen[1]; + + Rect2 sdf_rect = texture_storage->render_target_get_sdf_rect(p_to_render_target); + Rect2 sdf_tex_rect(sdf_rect.position / canvas_scale, sdf_rect.size / canvas_scale); + + state_buffer.sdf_to_tex[0] = 1.0 / sdf_tex_rect.size.width; + state_buffer.sdf_to_tex[1] = 1.0 / sdf_tex_rect.size.height; + state_buffer.sdf_to_tex[2] = -sdf_tex_rect.position.x / sdf_tex_rect.size.width; + state_buffer.sdf_to_tex[3] = -sdf_tex_rect.position.y / sdf_tex_rect.size.height; + + //print_line("w: " + itos(ssize.width) + " s: " + rtos(canvas_scale)); + state_buffer.tex_to_sdf = 1.0 / ((canvas_scale.x + canvas_scale.y) * 0.5); + + RD::get_singleton()->buffer_update(state.canvas_state_buffer, 0, sizeof(State::Buffer), &state_buffer); + } + + { //default filter/repeat + default_filter = p_default_filter; + default_repeat = p_default_repeat; + } + + Item *ci = p_item_list; + + //fill the list until rendering is possible. + bool material_screen_texture_cached = false; + bool material_screen_texture_mipmaps_cached = false; + + Rect2 back_buffer_rect; + bool backbuffer_copy = false; + bool backbuffer_gen_mipmaps = false; + + Item *canvas_group_owner = nullptr; + + bool update_skeletons = false; + bool time_used = false; + + bool backbuffer_cleared = false; + + while (ci) { + if (ci->copy_back_buffer && canvas_group_owner == nullptr) { + backbuffer_copy = true; + + if (ci->copy_back_buffer->full) { + back_buffer_rect = Rect2(); + } else { + back_buffer_rect = ci->copy_back_buffer->rect; + } + } + + RID material = ci->material_owner == nullptr ? ci->material : ci->material_owner->material; + + if (material.is_valid()) { + CanvasMaterialData *md = static_cast<CanvasMaterialData *>(material_storage->material_get_data(material, RendererRD::MaterialStorage::SHADER_TYPE_2D)); + if (md && md->shader_data->valid) { + if (md->shader_data->uses_screen_texture && canvas_group_owner == nullptr) { + if (!material_screen_texture_cached) { + backbuffer_copy = true; + back_buffer_rect = Rect2(); + backbuffer_gen_mipmaps = md->shader_data->uses_screen_texture_mipmaps; + } else if (!material_screen_texture_mipmaps_cached) { + backbuffer_gen_mipmaps = md->shader_data->uses_screen_texture_mipmaps; + } + } + + if (md->shader_data->uses_sdf) { + r_sdf_used = true; + } + if (md->shader_data->uses_time) { + time_used = true; + } + } + } + + if (ci->skeleton.is_valid()) { + const Item::Command *c = ci->commands; + + while (c) { + if (c->type == Item::Command::TYPE_MESH) { + const Item::CommandMesh *cm = static_cast<const Item::CommandMesh *>(c); + if (cm->mesh_instance.is_valid()) { + mesh_storage->mesh_instance_check_for_update(cm->mesh_instance); + update_skeletons = true; + } + } + c = c->next; + } + } + + if (ci->canvas_group_owner != nullptr) { + if (canvas_group_owner == nullptr) { + // Canvas group begins here, render until before this item + if (update_skeletons) { + mesh_storage->update_mesh_instances(); + update_skeletons = false; + } + + _render_items(p_to_render_target, item_count, canvas_transform_inverse, p_light_list); + item_count = 0; + + if (ci->canvas_group_owner->canvas_group->mode != RS::CANVAS_GROUP_MODE_TRANSPARENT) { + Rect2i group_rect = ci->canvas_group_owner->global_rect_cache; + texture_storage->render_target_copy_to_back_buffer(p_to_render_target, group_rect, false); + if (ci->canvas_group_owner->canvas_group->mode == RS::CANVAS_GROUP_MODE_CLIP_AND_DRAW) { + items[item_count++] = ci->canvas_group_owner; + } + } else if (!backbuffer_cleared) { + texture_storage->render_target_clear_back_buffer(p_to_render_target, Rect2i(), Color(0, 0, 0, 0)); + backbuffer_cleared = true; + } + + backbuffer_copy = false; + canvas_group_owner = ci->canvas_group_owner; //continue until owner found + } + + ci->canvas_group_owner = nullptr; //must be cleared + } + + if (!backbuffer_cleared && canvas_group_owner == nullptr && ci->canvas_group != nullptr && !backbuffer_copy) { + texture_storage->render_target_clear_back_buffer(p_to_render_target, Rect2i(), Color(0, 0, 0, 0)); + backbuffer_cleared = true; + } + + if (ci == canvas_group_owner) { + if (update_skeletons) { + mesh_storage->update_mesh_instances(); + update_skeletons = false; + } + + _render_items(p_to_render_target, item_count, canvas_transform_inverse, p_light_list, true); + item_count = 0; + + if (ci->canvas_group->blur_mipmaps) { + texture_storage->render_target_gen_back_buffer_mipmaps(p_to_render_target, ci->global_rect_cache); + } + + canvas_group_owner = nullptr; + // Backbuffer is dirty now and needs to be re-cleared if another CanvasGroup needs it. + backbuffer_cleared = false; + } + + if (backbuffer_copy) { + //render anything pending, including clearing if no items + if (update_skeletons) { + mesh_storage->update_mesh_instances(); + update_skeletons = false; + } + + _render_items(p_to_render_target, item_count, canvas_transform_inverse, p_light_list); + item_count = 0; + + texture_storage->render_target_copy_to_back_buffer(p_to_render_target, back_buffer_rect, backbuffer_gen_mipmaps); + + backbuffer_copy = false; + backbuffer_gen_mipmaps = false; + material_screen_texture_cached = true; // After a backbuffer copy, screen texture makes no further copies. + material_screen_texture_mipmaps_cached = backbuffer_gen_mipmaps; + } + + if (backbuffer_gen_mipmaps) { + texture_storage->render_target_gen_back_buffer_mipmaps(p_to_render_target, back_buffer_rect); + + backbuffer_gen_mipmaps = false; + material_screen_texture_mipmaps_cached = true; + } + + items[item_count++] = ci; + + if (!ci->next || item_count == MAX_RENDER_ITEMS - 1) { + if (update_skeletons) { + mesh_storage->update_mesh_instances(); + update_skeletons = false; + } + + _render_items(p_to_render_target, item_count, canvas_transform_inverse, p_light_list); + //then reset + item_count = 0; + } + + ci = ci->next; + } + + if (time_used) { + RenderingServerDefault::redraw_request(); + } +} + +RID RendererCanvasRenderRD::light_create() { + CanvasLight canvas_light; + return canvas_light_owner.make_rid(canvas_light); +} + +void RendererCanvasRenderRD::light_set_texture(RID p_rid, RID p_texture) { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + + CanvasLight *cl = canvas_light_owner.get_or_null(p_rid); + ERR_FAIL_COND(!cl); + if (cl->texture == p_texture) { + return; + } + if (cl->texture.is_valid()) { + texture_storage->texture_remove_from_decal_atlas(cl->texture); + } + cl->texture = p_texture; + + if (cl->texture.is_valid()) { + texture_storage->texture_add_to_decal_atlas(cl->texture); + } +} + +void RendererCanvasRenderRD::light_set_use_shadow(RID p_rid, bool p_enable) { + CanvasLight *cl = canvas_light_owner.get_or_null(p_rid); + ERR_FAIL_COND(!cl); + + cl->shadow.enabled = p_enable; +} + +void RendererCanvasRenderRD::_update_shadow_atlas() { + if (state.shadow_fb == RID()) { + //ah, we lack the shadow texture.. + RD::get_singleton()->free(state.shadow_texture); //erase placeholder + + Vector<RID> fb_textures; + + { //texture + RD::TextureFormat tf; + tf.texture_type = RD::TEXTURE_TYPE_2D; + tf.width = state.shadow_texture_size; + tf.height = state.max_lights_per_render * 2; + tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; + tf.format = RD::DATA_FORMAT_R32_SFLOAT; + + state.shadow_texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); + fb_textures.push_back(state.shadow_texture); + } + { + RD::TextureFormat tf; + tf.texture_type = RD::TEXTURE_TYPE_2D; + tf.width = state.shadow_texture_size; + tf.height = state.max_lights_per_render * 2; + tf.usage_bits = RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT; + tf.format = RD::DATA_FORMAT_D32_SFLOAT; + //chunks to write + state.shadow_depth_texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); + fb_textures.push_back(state.shadow_depth_texture); + } + + state.shadow_fb = RD::get_singleton()->framebuffer_create(fb_textures); + } +} +void RendererCanvasRenderRD::light_update_shadow(RID p_rid, int p_shadow_index, const Transform2D &p_light_xform, int p_light_mask, float p_near, float p_far, LightOccluderInstance *p_occluders) { + CanvasLight *cl = canvas_light_owner.get_or_null(p_rid); + ERR_FAIL_COND(!cl->shadow.enabled); + + _update_shadow_atlas(); + + cl->shadow.z_far = p_far; + cl->shadow.y_offset = float(p_shadow_index * 2 + 1) / float(state.max_lights_per_render * 2); + Vector<Color> cc; + cc.push_back(Color(p_far, p_far, p_far, 1.0)); + + for (int i = 0; i < 4; i++) { + //make sure it remains orthogonal, makes easy to read angle later + + //light.basis.scale(Vector3(to_light.elements[0].length(),to_light.elements[1].length(),1)); + + Rect2i rect((state.shadow_texture_size / 4) * i, p_shadow_index * 2, (state.shadow_texture_size / 4), 2); + RD::InitialAction initial_action = i == 0 ? RD::INITIAL_ACTION_CLEAR_REGION : RD::INITIAL_ACTION_CLEAR_REGION_CONTINUE; + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(state.shadow_fb, initial_action, i != 3 ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, initial_action, RD::FINAL_ACTION_DISCARD, cc, 1.0, 0, rect); + + Projection projection; + { + real_t fov = 90; + real_t nearp = p_near; + real_t farp = p_far; + real_t aspect = 1.0; + + real_t ymax = nearp * Math::tan(Math::deg_to_rad(fov * 0.5)); + real_t ymin = -ymax; + real_t xmin = ymin * aspect; + real_t xmax = ymax * aspect; + + projection.set_frustum(xmin, xmax, ymin, ymax, nearp, farp); + } + + Vector3 cam_target = Basis::from_euler(Vector3(0, 0, Math_TAU * ((i + 3) / 4.0))).xform(Vector3(0, 1, 0)); + projection = projection * Projection(Transform3D().looking_at(cam_target, Vector3(0, 0, -1)).affine_inverse()); + + ShadowRenderPushConstant push_constant; + for (int y = 0; y < 4; y++) { + for (int x = 0; x < 4; x++) { + push_constant.projection[y * 4 + x] = projection.columns[y][x]; + } + } + static const Vector2 directions[4] = { Vector2(1, 0), Vector2(0, 1), Vector2(-1, 0), Vector2(0, -1) }; + push_constant.direction[0] = directions[i].x; + push_constant.direction[1] = directions[i].y; + push_constant.z_far = p_far; + push_constant.pad = 0; + + LightOccluderInstance *instance = p_occluders; + + while (instance) { + OccluderPolygon *co = occluder_polygon_owner.get_or_null(instance->occluder); + + if (!co || co->index_array.is_null() || !(p_light_mask & instance->light_mask)) { + instance = instance->next; + continue; + } + + _update_transform_2d_to_mat2x4(p_light_xform * instance->xform_cache, push_constant.modelview); + + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, shadow_render.render_pipelines[co->cull_mode]); + RD::get_singleton()->draw_list_bind_vertex_array(draw_list, co->vertex_array); + RD::get_singleton()->draw_list_bind_index_array(draw_list, co->index_array); + RD::get_singleton()->draw_list_set_push_constant(draw_list, &push_constant, sizeof(ShadowRenderPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + + instance = instance->next; + } + + RD::get_singleton()->draw_list_end(); + } +} + +void RendererCanvasRenderRD::light_update_directional_shadow(RID p_rid, int p_shadow_index, const Transform2D &p_light_xform, int p_light_mask, float p_cull_distance, const Rect2 &p_clip_rect, LightOccluderInstance *p_occluders) { + CanvasLight *cl = canvas_light_owner.get_or_null(p_rid); + ERR_FAIL_COND(!cl->shadow.enabled); + + _update_shadow_atlas(); + + Vector2 light_dir = p_light_xform.columns[1].normalized(); + + Vector2 center = p_clip_rect.get_center(); + + float to_edge_distance = ABS(light_dir.dot(p_clip_rect.get_support(light_dir)) - light_dir.dot(center)); + + Vector2 from_pos = center - light_dir * (to_edge_distance + p_cull_distance); + float distance = to_edge_distance * 2.0 + p_cull_distance; + float half_size = p_clip_rect.size.length() * 0.5; //shadow length, must keep this no matter the angle + + cl->shadow.z_far = distance; + cl->shadow.y_offset = float(p_shadow_index * 2 + 1) / float(state.max_lights_per_render * 2); + + Transform2D to_light_xform; + + to_light_xform[2] = from_pos; + to_light_xform[1] = light_dir; + to_light_xform[0] = -light_dir.orthogonal(); + + to_light_xform.invert(); + + Vector<Color> cc; + cc.push_back(Color(1, 1, 1, 1)); + + Rect2i rect(0, p_shadow_index * 2, state.shadow_texture_size, 2); + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(state.shadow_fb, RD::INITIAL_ACTION_CLEAR_REGION, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR_REGION, RD::FINAL_ACTION_DISCARD, cc, 1.0, 0, rect); + + Projection projection; + projection.set_orthogonal(-half_size, half_size, -0.5, 0.5, 0.0, distance); + projection = projection * Projection(Transform3D().looking_at(Vector3(0, 1, 0), Vector3(0, 0, -1)).affine_inverse()); + + ShadowRenderPushConstant push_constant; + for (int y = 0; y < 4; y++) { + for (int x = 0; x < 4; x++) { + push_constant.projection[y * 4 + x] = projection.columns[y][x]; + } + } + + push_constant.direction[0] = 0.0; + push_constant.direction[1] = 1.0; + push_constant.z_far = distance; + push_constant.pad = 0; + + LightOccluderInstance *instance = p_occluders; + + while (instance) { + OccluderPolygon *co = occluder_polygon_owner.get_or_null(instance->occluder); + + if (!co || co->index_array.is_null() || !(p_light_mask & instance->light_mask)) { + instance = instance->next; + continue; + } + + _update_transform_2d_to_mat2x4(to_light_xform * instance->xform_cache, push_constant.modelview); + + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, shadow_render.render_pipelines[co->cull_mode]); + RD::get_singleton()->draw_list_bind_vertex_array(draw_list, co->vertex_array); + RD::get_singleton()->draw_list_bind_index_array(draw_list, co->index_array); + RD::get_singleton()->draw_list_set_push_constant(draw_list, &push_constant, sizeof(ShadowRenderPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + + instance = instance->next; + } + + RD::get_singleton()->draw_list_end(); + + Transform2D to_shadow; + to_shadow.columns[0].x = 1.0 / -(half_size * 2.0); + to_shadow.columns[2].x = 0.5; + + cl->shadow.directional_xform = to_shadow * to_light_xform; +} + +void RendererCanvasRenderRD::render_sdf(RID p_render_target, LightOccluderInstance *p_occluders) { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + + RID fb = texture_storage->render_target_get_sdf_framebuffer(p_render_target); + Rect2i rect = texture_storage->render_target_get_sdf_rect(p_render_target); + + Transform2D to_sdf; + to_sdf.columns[0] *= rect.size.width; + to_sdf.columns[1] *= rect.size.height; + to_sdf.columns[2] = rect.position; + + Transform2D to_clip; + to_clip.columns[0] *= 2.0; + to_clip.columns[1] *= 2.0; + to_clip.columns[2] = -Vector2(1.0, 1.0); + + to_clip = to_clip * to_sdf.affine_inverse(); + + Vector<Color> cc; + cc.push_back(Color(0, 0, 0, 0)); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(fb, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_DISCARD, cc); + + Projection projection; + + ShadowRenderPushConstant push_constant; + for (int y = 0; y < 4; y++) { + for (int x = 0; x < 4; x++) { + push_constant.projection[y * 4 + x] = projection.columns[y][x]; + } + } + + push_constant.direction[0] = 0.0; + push_constant.direction[1] = 0.0; + push_constant.z_far = 0; + push_constant.pad = 0; + + LightOccluderInstance *instance = p_occluders; + + while (instance) { + OccluderPolygon *co = occluder_polygon_owner.get_or_null(instance->occluder); + + if (!co || co->sdf_index_array.is_null()) { + instance = instance->next; + continue; + } + + _update_transform_2d_to_mat2x4(to_clip * instance->xform_cache, push_constant.modelview); + + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, shadow_render.sdf_render_pipelines[co->sdf_is_lines ? SHADOW_RENDER_SDF_LINES : SHADOW_RENDER_SDF_TRIANGLES]); + RD::get_singleton()->draw_list_bind_vertex_array(draw_list, co->sdf_vertex_array); + RD::get_singleton()->draw_list_bind_index_array(draw_list, co->sdf_index_array); + RD::get_singleton()->draw_list_set_push_constant(draw_list, &push_constant, sizeof(ShadowRenderPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + + instance = instance->next; + } + + RD::get_singleton()->draw_list_end(); + + texture_storage->render_target_sdf_process(p_render_target); //done rendering, process it +} + +RID RendererCanvasRenderRD::occluder_polygon_create() { + OccluderPolygon occluder; + occluder.line_point_count = 0; + occluder.sdf_point_count = 0; + occluder.sdf_index_count = 0; + occluder.cull_mode = RS::CANVAS_OCCLUDER_POLYGON_CULL_DISABLED; + return occluder_polygon_owner.make_rid(occluder); +} + +void RendererCanvasRenderRD::occluder_polygon_set_shape(RID p_occluder, const Vector<Vector2> &p_points, bool p_closed) { + OccluderPolygon *oc = occluder_polygon_owner.get_or_null(p_occluder); + ERR_FAIL_COND(!oc); + + Vector<Vector2> lines; + + if (p_points.size()) { + int lc = p_points.size() * 2; + + lines.resize(lc - (p_closed ? 0 : 2)); + { + Vector2 *w = lines.ptrw(); + const Vector2 *r = p_points.ptr(); + + int max = lc / 2; + if (!p_closed) { + max--; + } + for (int i = 0; i < max; i++) { + Vector2 a = r[i]; + Vector2 b = r[(i + 1) % (lc / 2)]; + w[i * 2 + 0] = a; + w[i * 2 + 1] = b; + } + } + } + + if (oc->line_point_count != lines.size() && oc->vertex_array.is_valid()) { + RD::get_singleton()->free(oc->vertex_array); + RD::get_singleton()->free(oc->vertex_buffer); + RD::get_singleton()->free(oc->index_array); + RD::get_singleton()->free(oc->index_buffer); + + oc->vertex_array = RID(); + oc->vertex_buffer = RID(); + oc->index_array = RID(); + oc->index_buffer = RID(); + + oc->line_point_count = lines.size(); + } + + if (lines.size()) { + Vector<uint8_t> geometry; + Vector<uint8_t> indices; + int lc = lines.size(); + + geometry.resize(lc * 6 * sizeof(float)); + indices.resize(lc * 3 * sizeof(uint16_t)); + + { + uint8_t *vw = geometry.ptrw(); + float *vwptr = reinterpret_cast<float *>(vw); + uint8_t *iw = indices.ptrw(); + uint16_t *iwptr = (uint16_t *)iw; + + const Vector2 *lr = lines.ptr(); + + const int POLY_HEIGHT = 16384; + + for (int i = 0; i < lc / 2; i++) { + vwptr[i * 12 + 0] = lr[i * 2 + 0].x; + vwptr[i * 12 + 1] = lr[i * 2 + 0].y; + vwptr[i * 12 + 2] = POLY_HEIGHT; + + vwptr[i * 12 + 3] = lr[i * 2 + 1].x; + vwptr[i * 12 + 4] = lr[i * 2 + 1].y; + vwptr[i * 12 + 5] = POLY_HEIGHT; + + vwptr[i * 12 + 6] = lr[i * 2 + 1].x; + vwptr[i * 12 + 7] = lr[i * 2 + 1].y; + vwptr[i * 12 + 8] = -POLY_HEIGHT; + + vwptr[i * 12 + 9] = lr[i * 2 + 0].x; + vwptr[i * 12 + 10] = lr[i * 2 + 0].y; + vwptr[i * 12 + 11] = -POLY_HEIGHT; + + iwptr[i * 6 + 0] = i * 4 + 0; + iwptr[i * 6 + 1] = i * 4 + 1; + iwptr[i * 6 + 2] = i * 4 + 2; + + iwptr[i * 6 + 3] = i * 4 + 2; + iwptr[i * 6 + 4] = i * 4 + 3; + iwptr[i * 6 + 5] = i * 4 + 0; + } + } + + //if same buffer len is being set, just use BufferSubData to avoid a pipeline flush + + if (oc->vertex_array.is_null()) { + //create from scratch + //vertices + oc->vertex_buffer = RD::get_singleton()->vertex_buffer_create(lc * 6 * sizeof(real_t), geometry); + + Vector<RID> buffer; + buffer.push_back(oc->vertex_buffer); + oc->vertex_array = RD::get_singleton()->vertex_array_create(4 * lc / 2, shadow_render.vertex_format, buffer); + //indices + + oc->index_buffer = RD::get_singleton()->index_buffer_create(3 * lc, RD::INDEX_BUFFER_FORMAT_UINT16, indices); + oc->index_array = RD::get_singleton()->index_array_create(oc->index_buffer, 0, 3 * lc); + + } else { + //update existing + const uint8_t *vr = geometry.ptr(); + RD::get_singleton()->buffer_update(oc->vertex_buffer, 0, geometry.size(), vr); + const uint8_t *ir = indices.ptr(); + RD::get_singleton()->buffer_update(oc->index_buffer, 0, indices.size(), ir); + } + } + + // sdf + + Vector<int> sdf_indices; + + if (p_points.size()) { + if (p_closed) { + sdf_indices = Geometry2D::triangulate_polygon(p_points); + oc->sdf_is_lines = false; + } else { + int max = p_points.size(); + sdf_indices.resize(max * 2); + + int *iw = sdf_indices.ptrw(); + for (int i = 0; i < max; i++) { + iw[i * 2 + 0] = i; + iw[i * 2 + 1] = (i + 1) % max; + } + oc->sdf_is_lines = true; + } + } + + if (oc->sdf_index_count != sdf_indices.size() && oc->sdf_point_count != p_points.size() && oc->sdf_vertex_array.is_valid()) { + RD::get_singleton()->free(oc->sdf_vertex_array); + RD::get_singleton()->free(oc->sdf_vertex_buffer); + RD::get_singleton()->free(oc->sdf_index_array); + RD::get_singleton()->free(oc->sdf_index_buffer); + + oc->sdf_vertex_array = RID(); + oc->sdf_vertex_buffer = RID(); + oc->sdf_index_array = RID(); + oc->sdf_index_buffer = RID(); + + oc->sdf_index_count = sdf_indices.size(); + oc->sdf_point_count = p_points.size(); + + oc->sdf_is_lines = false; + } + + if (sdf_indices.size()) { + if (oc->sdf_vertex_array.is_null()) { + //create from scratch + //vertices + oc->sdf_vertex_buffer = RD::get_singleton()->vertex_buffer_create(p_points.size() * 2 * sizeof(real_t), p_points.to_byte_array()); + oc->sdf_index_buffer = RD::get_singleton()->index_buffer_create(sdf_indices.size(), RD::INDEX_BUFFER_FORMAT_UINT32, sdf_indices.to_byte_array()); + oc->sdf_index_array = RD::get_singleton()->index_array_create(oc->sdf_index_buffer, 0, sdf_indices.size()); + + Vector<RID> buffer; + buffer.push_back(oc->sdf_vertex_buffer); + oc->sdf_vertex_array = RD::get_singleton()->vertex_array_create(p_points.size(), shadow_render.sdf_vertex_format, buffer); + //indices + + } else { + //update existing + RD::get_singleton()->buffer_update(oc->sdf_vertex_buffer, 0, sizeof(real_t) * 2 * p_points.size(), p_points.ptr()); + RD::get_singleton()->buffer_update(oc->sdf_index_buffer, 0, sdf_indices.size() * sizeof(int32_t), sdf_indices.ptr()); + } + } +} + +void RendererCanvasRenderRD::occluder_polygon_set_cull_mode(RID p_occluder, RS::CanvasOccluderPolygonCullMode p_mode) { + OccluderPolygon *oc = occluder_polygon_owner.get_or_null(p_occluder); + ERR_FAIL_COND(!oc); + oc->cull_mode = p_mode; +} + +void RendererCanvasRenderRD::CanvasShaderData::set_path_hint(const String &p_path) { + path = p_path; +} + +void RendererCanvasRenderRD::CanvasShaderData::set_code(const String &p_code) { + //compile + + code = p_code; + valid = false; + ubo_size = 0; + uniforms.clear(); + uses_screen_texture = false; + uses_screen_texture_mipmaps = false; + uses_sdf = false; + uses_time = false; + + if (code.is_empty()) { + return; //just invalid, but no error + } + + ShaderCompiler::GeneratedCode gen_code; + + int blend_mode = BLEND_MODE_MIX; + + ShaderCompiler::IdentifierActions actions; + actions.entry_point_stages["vertex"] = ShaderCompiler::STAGE_VERTEX; + actions.entry_point_stages["fragment"] = ShaderCompiler::STAGE_FRAGMENT; + actions.entry_point_stages["light"] = ShaderCompiler::STAGE_FRAGMENT; + + actions.render_mode_values["blend_add"] = Pair<int *, int>(&blend_mode, BLEND_MODE_ADD); + actions.render_mode_values["blend_mix"] = Pair<int *, int>(&blend_mode, BLEND_MODE_MIX); + actions.render_mode_values["blend_sub"] = Pair<int *, int>(&blend_mode, BLEND_MODE_SUB); + actions.render_mode_values["blend_mul"] = Pair<int *, int>(&blend_mode, BLEND_MODE_MUL); + actions.render_mode_values["blend_premul_alpha"] = Pair<int *, int>(&blend_mode, BLEND_MODE_PMALPHA); + actions.render_mode_values["blend_disabled"] = Pair<int *, int>(&blend_mode, BLEND_MODE_DISABLED); + + actions.usage_flag_pointers["SCREEN_TEXTURE"] = &uses_screen_texture; + actions.usage_flag_pointers["texture_sdf"] = &uses_sdf; + actions.usage_flag_pointers["TIME"] = &uses_time; + + actions.uniforms = &uniforms; + + RendererCanvasRenderRD *canvas_singleton = static_cast<RendererCanvasRenderRD *>(RendererCanvasRender::singleton); + + Error err = canvas_singleton->shader.compiler.compile(RS::SHADER_CANVAS_ITEM, code, &actions, path, gen_code); + ERR_FAIL_COND_MSG(err != OK, "Shader compilation failed."); + + uses_screen_texture_mipmaps = gen_code.uses_screen_texture_mipmaps; + + if (version.is_null()) { + version = canvas_singleton->shader.canvas_shader.version_create(); + } + +#if 0 + print_line("**compiling shader:"); + print_line("**defines:\n"); + for (int i = 0; i < gen_code.defines.size(); i++) { + print_line(gen_code.defines[i]); + } + + HashMap<String, String>::Iterator el = gen_code.code.begin(); + while (el) { + print_line("\n**code " + el->key + ":\n" + el->value); + ++el; + } + + print_line("\n**uniforms:\n" + gen_code.uniforms); + print_line("\n**vertex_globals:\n" + gen_code.stage_globals[ShaderCompiler::STAGE_VERTEX]); + print_line("\n**fragment_globals:\n" + gen_code.stage_globals[ShaderCompiler::STAGE_FRAGMENT]); +#endif + canvas_singleton->shader.canvas_shader.version_set_code(version, gen_code.code, gen_code.uniforms, gen_code.stage_globals[ShaderCompiler::STAGE_VERTEX], gen_code.stage_globals[ShaderCompiler::STAGE_FRAGMENT], gen_code.defines); + ERR_FAIL_COND(!canvas_singleton->shader.canvas_shader.version_is_valid(version)); + + ubo_size = gen_code.uniform_total_size; + ubo_offsets = gen_code.uniform_offsets; + texture_uniforms = gen_code.texture_uniforms; + + //update them pipelines + + RD::PipelineColorBlendState::Attachment attachment; + + switch (blend_mode) { + case BLEND_MODE_DISABLED: { + // nothing to do here, disabled by default + + } break; + case BLEND_MODE_MIX: { + attachment.enable_blend = true; + attachment.color_blend_op = RD::BLEND_OP_ADD; + attachment.src_color_blend_factor = RD::BLEND_FACTOR_SRC_ALPHA; + attachment.dst_color_blend_factor = RD::BLEND_FACTOR_ONE_MINUS_SRC_ALPHA; + + attachment.alpha_blend_op = RD::BLEND_OP_ADD; + attachment.src_alpha_blend_factor = RD::BLEND_FACTOR_ONE; + attachment.dst_alpha_blend_factor = RD::BLEND_FACTOR_ONE_MINUS_SRC_ALPHA; + + } break; + case BLEND_MODE_ADD: { + attachment.enable_blend = true; + attachment.alpha_blend_op = RD::BLEND_OP_ADD; + attachment.color_blend_op = RD::BLEND_OP_ADD; + attachment.src_color_blend_factor = RD::BLEND_FACTOR_SRC_ALPHA; + attachment.dst_color_blend_factor = RD::BLEND_FACTOR_ONE; + attachment.src_alpha_blend_factor = RD::BLEND_FACTOR_SRC_ALPHA; + attachment.dst_alpha_blend_factor = RD::BLEND_FACTOR_ONE; + + } break; + case BLEND_MODE_SUB: { + attachment.enable_blend = true; + attachment.alpha_blend_op = RD::BLEND_OP_SUBTRACT; + attachment.color_blend_op = RD::BLEND_OP_SUBTRACT; + attachment.src_color_blend_factor = RD::BLEND_FACTOR_SRC_ALPHA; + attachment.dst_color_blend_factor = RD::BLEND_FACTOR_ONE; + attachment.src_alpha_blend_factor = RD::BLEND_FACTOR_SRC_ALPHA; + attachment.dst_alpha_blend_factor = RD::BLEND_FACTOR_ONE; + + } break; + case BLEND_MODE_MUL: { + attachment.enable_blend = true; + attachment.alpha_blend_op = RD::BLEND_OP_ADD; + attachment.color_blend_op = RD::BLEND_OP_ADD; + attachment.src_color_blend_factor = RD::BLEND_FACTOR_DST_COLOR; + attachment.dst_color_blend_factor = RD::BLEND_FACTOR_ZERO; + attachment.src_alpha_blend_factor = RD::BLEND_FACTOR_DST_ALPHA; + attachment.dst_alpha_blend_factor = RD::BLEND_FACTOR_ZERO; + + } break; + case BLEND_MODE_PMALPHA: { + attachment.enable_blend = true; + attachment.alpha_blend_op = RD::BLEND_OP_ADD; + attachment.color_blend_op = RD::BLEND_OP_ADD; + attachment.src_color_blend_factor = RD::BLEND_FACTOR_ONE; + attachment.dst_color_blend_factor = RD::BLEND_FACTOR_ONE_MINUS_SRC_ALPHA; + attachment.src_alpha_blend_factor = RD::BLEND_FACTOR_ONE; + attachment.dst_alpha_blend_factor = RD::BLEND_FACTOR_ONE_MINUS_SRC_ALPHA; + + } break; + } + + RD::PipelineColorBlendState blend_state; + blend_state.attachments.push_back(attachment); + + RD::PipelineColorBlendState::Attachment attachment_lcd; + attachment_lcd.enable_blend = true; + attachment_lcd.alpha_blend_op = RD::BLEND_OP_ADD; + attachment_lcd.color_blend_op = RD::BLEND_OP_ADD; + attachment_lcd.src_color_blend_factor = RD::BLEND_FACTOR_CONSTANT_COLOR; + attachment_lcd.dst_color_blend_factor = RD::BLEND_FACTOR_ONE_MINUS_SRC_COLOR; + attachment_lcd.src_alpha_blend_factor = RD::BLEND_FACTOR_ONE; + attachment_lcd.dst_alpha_blend_factor = RD::BLEND_FACTOR_ONE_MINUS_SRC_ALPHA; + + RD::PipelineColorBlendState blend_state_lcd; + blend_state_lcd.attachments.push_back(attachment_lcd); + + //update pipelines + + for (int i = 0; i < PIPELINE_LIGHT_MODE_MAX; i++) { + for (int j = 0; j < PIPELINE_VARIANT_MAX; j++) { + RD::RenderPrimitive primitive[PIPELINE_VARIANT_MAX] = { + RD::RENDER_PRIMITIVE_TRIANGLES, + RD::RENDER_PRIMITIVE_TRIANGLES, + RD::RENDER_PRIMITIVE_TRIANGLES, + RD::RENDER_PRIMITIVE_LINES, + RD::RENDER_PRIMITIVE_POINTS, + RD::RENDER_PRIMITIVE_TRIANGLES, + RD::RENDER_PRIMITIVE_TRIANGLE_STRIPS, + RD::RENDER_PRIMITIVE_LINES, + RD::RENDER_PRIMITIVE_LINESTRIPS, + RD::RENDER_PRIMITIVE_POINTS, + RD::RENDER_PRIMITIVE_TRIANGLES, + }; + + ShaderVariant shader_variants[PIPELINE_LIGHT_MODE_MAX][PIPELINE_VARIANT_MAX] = { + { + //non lit + SHADER_VARIANT_QUAD, + SHADER_VARIANT_NINEPATCH, + SHADER_VARIANT_PRIMITIVE, + SHADER_VARIANT_PRIMITIVE, + SHADER_VARIANT_PRIMITIVE_POINTS, + SHADER_VARIANT_ATTRIBUTES, + SHADER_VARIANT_ATTRIBUTES, + SHADER_VARIANT_ATTRIBUTES, + SHADER_VARIANT_ATTRIBUTES, + SHADER_VARIANT_ATTRIBUTES_POINTS, + SHADER_VARIANT_QUAD, + }, + { + //lit + SHADER_VARIANT_QUAD_LIGHT, + SHADER_VARIANT_NINEPATCH_LIGHT, + SHADER_VARIANT_PRIMITIVE_LIGHT, + SHADER_VARIANT_PRIMITIVE_LIGHT, + SHADER_VARIANT_PRIMITIVE_POINTS_LIGHT, + SHADER_VARIANT_ATTRIBUTES_LIGHT, + SHADER_VARIANT_ATTRIBUTES_LIGHT, + SHADER_VARIANT_ATTRIBUTES_LIGHT, + SHADER_VARIANT_ATTRIBUTES_LIGHT, + SHADER_VARIANT_ATTRIBUTES_POINTS_LIGHT, + SHADER_VARIANT_QUAD_LIGHT, + }, + }; + + RID shader_variant = canvas_singleton->shader.canvas_shader.version_get_shader(version, shader_variants[i][j]); + if (j == PIPELINE_VARIANT_QUAD_LCD_BLEND) { + pipeline_variants.variants[i][j].setup(shader_variant, primitive[j], RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), blend_state_lcd, RD::DYNAMIC_STATE_BLEND_CONSTANTS); + } else { + pipeline_variants.variants[i][j].setup(shader_variant, primitive[j], RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), blend_state, 0); + } + } + } + + valid = true; +} + +void RendererCanvasRenderRD::CanvasShaderData::set_default_texture_parameter(const StringName &p_name, RID p_texture, int p_index) { + if (!p_texture.is_valid()) { + if (default_texture_params.has(p_name) && default_texture_params[p_name].has(p_index)) { + default_texture_params[p_name].erase(p_index); + + if (default_texture_params[p_name].is_empty()) { + default_texture_params.erase(p_name); + } + } + } else { + if (!default_texture_params.has(p_name)) { + default_texture_params[p_name] = HashMap<int, RID>(); + } + default_texture_params[p_name][p_index] = p_texture; + } +} + +void RendererCanvasRenderRD::CanvasShaderData::get_shader_uniform_list(List<PropertyInfo> *p_param_list) const { + HashMap<int, StringName> order; + + for (const KeyValue<StringName, ShaderLanguage::ShaderNode::Uniform> &E : uniforms) { + if (E.value.scope != ShaderLanguage::ShaderNode::Uniform::SCOPE_LOCAL || + E.value.hint == ShaderLanguage::ShaderNode::Uniform::HINT_SCREEN_TEXTURE || + E.value.hint == ShaderLanguage::ShaderNode::Uniform::HINT_NORMAL_ROUGHNESS_TEXTURE || + E.value.hint == ShaderLanguage::ShaderNode::Uniform::HINT_DEPTH_TEXTURE) { + // Don't expose any of these. + continue; + } + if (E.value.texture_order >= 0) { + order[E.value.texture_order + 100000] = E.key; + } else { + order[E.value.order] = E.key; + } + } + + String last_group; + for (const KeyValue<int, StringName> &E : order) { + String group = uniforms[E.value].group; + if (!uniforms[E.value].subgroup.is_empty()) { + group += "::" + uniforms[E.value].subgroup; + } + + if (group != last_group) { + PropertyInfo pi; + pi.usage = PROPERTY_USAGE_GROUP; + pi.name = group; + p_param_list->push_back(pi); + + last_group = group; + } + + PropertyInfo pi = ShaderLanguage::uniform_to_property_info(uniforms[E.value]); + pi.name = E.value; + p_param_list->push_back(pi); + } +} + +void RendererCanvasRenderRD::CanvasShaderData::get_instance_param_list(List<RendererMaterialStorage::InstanceShaderParam> *p_param_list) const { + for (const KeyValue<StringName, ShaderLanguage::ShaderNode::Uniform> &E : uniforms) { + if (E.value.scope != ShaderLanguage::ShaderNode::Uniform::SCOPE_INSTANCE) { + continue; + } + + RendererMaterialStorage::InstanceShaderParam p; + p.info = ShaderLanguage::uniform_to_property_info(E.value); + p.info.name = E.key; //supply name + p.index = E.value.instance_index; + p.default_value = ShaderLanguage::constant_value_to_variant(E.value.default_value, E.value.type, E.value.array_size, E.value.hint); + p_param_list->push_back(p); + } +} + +bool RendererCanvasRenderRD::CanvasShaderData::is_parameter_texture(const StringName &p_param) const { + if (!uniforms.has(p_param)) { + return false; + } + + return uniforms[p_param].texture_order >= 0; +} + +bool RendererCanvasRenderRD::CanvasShaderData::is_animated() const { + return false; +} + +bool RendererCanvasRenderRD::CanvasShaderData::casts_shadows() const { + return false; +} + +Variant RendererCanvasRenderRD::CanvasShaderData::get_default_parameter(const StringName &p_parameter) const { + if (uniforms.has(p_parameter)) { + ShaderLanguage::ShaderNode::Uniform uniform = uniforms[p_parameter]; + Vector<ShaderLanguage::ConstantNode::Value> default_value = uniform.default_value; + return ShaderLanguage::constant_value_to_variant(default_value, uniform.type, uniform.array_size, uniform.hint); + } + return Variant(); +} + +RS::ShaderNativeSourceCode RendererCanvasRenderRD::CanvasShaderData::get_native_source_code() const { + RendererCanvasRenderRD *canvas_singleton = static_cast<RendererCanvasRenderRD *>(RendererCanvasRender::singleton); + return canvas_singleton->shader.canvas_shader.version_get_native_source_code(version); +} + +RendererCanvasRenderRD::CanvasShaderData::~CanvasShaderData() { + RendererCanvasRenderRD *canvas_singleton = static_cast<RendererCanvasRenderRD *>(RendererCanvasRender::singleton); + ERR_FAIL_COND(!canvas_singleton); + //pipeline variants will clear themselves if shader is gone + if (version.is_valid()) { + canvas_singleton->shader.canvas_shader.version_free(version); + } +} + +RendererRD::MaterialStorage::ShaderData *RendererCanvasRenderRD::_create_shader_func() { + CanvasShaderData *shader_data = memnew(CanvasShaderData); + return shader_data; +} + +bool RendererCanvasRenderRD::CanvasMaterialData::update_parameters(const HashMap<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { + RendererCanvasRenderRD *canvas_singleton = static_cast<RendererCanvasRenderRD *>(RendererCanvasRender::singleton); + + return update_parameters_uniform_set(p_parameters, p_uniform_dirty, p_textures_dirty, shader_data->uniforms, shader_data->ubo_offsets.ptr(), shader_data->texture_uniforms, shader_data->default_texture_params, shader_data->ubo_size, uniform_set, canvas_singleton->shader.canvas_shader.version_get_shader(shader_data->version, 0), MATERIAL_UNIFORM_SET, false); +} + +RendererCanvasRenderRD::CanvasMaterialData::~CanvasMaterialData() { + free_parameters_uniform_set(uniform_set); +} + +RendererRD::MaterialStorage::MaterialData *RendererCanvasRenderRD::_create_material_func(CanvasShaderData *p_shader) { + CanvasMaterialData *material_data = memnew(CanvasMaterialData); + material_data->shader_data = p_shader; + //update will happen later anyway so do nothing. + return material_data; +} + +void RendererCanvasRenderRD::set_time(double p_time) { + state.time = p_time; +} + +void RendererCanvasRenderRD::update() { +} + +RendererCanvasRenderRD::RendererCanvasRenderRD() { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + + { //create default samplers + + default_samplers.default_filter = RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR; + default_samplers.default_repeat = RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED; + } + + { //shader variants + + String global_defines; + + uint64_t uniform_max_size = RD::get_singleton()->limit_get(RD::LIMIT_MAX_UNIFORM_BUFFER_SIZE); + if (uniform_max_size < 65536) { + //Yes, you guessed right, ARM again + state.max_lights_per_render = 64; + global_defines += "#define MAX_LIGHTS 64\n"; + } else { + state.max_lights_per_render = DEFAULT_MAX_LIGHTS_PER_RENDER; + global_defines += "#define MAX_LIGHTS " + itos(DEFAULT_MAX_LIGHTS_PER_RENDER) + "\n"; + } + + state.light_uniforms = memnew_arr(LightUniform, state.max_lights_per_render); + Vector<String> variants; + //non light variants + variants.push_back(""); //none by default is first variant + variants.push_back("#define USE_NINEPATCH\n"); //ninepatch is the second variant + variants.push_back("#define USE_PRIMITIVE\n"); //primitive is the third + variants.push_back("#define USE_PRIMITIVE\n#define USE_POINT_SIZE\n"); //points need point size + variants.push_back("#define USE_ATTRIBUTES\n"); // attributes for vertex arrays + variants.push_back("#define USE_ATTRIBUTES\n#define USE_POINT_SIZE\n"); //attributes with point size + //light variants + variants.push_back("#define USE_LIGHTING\n"); //none by default is first variant + variants.push_back("#define USE_LIGHTING\n#define USE_NINEPATCH\n"); //ninepatch is the second variant + variants.push_back("#define USE_LIGHTING\n#define USE_PRIMITIVE\n"); //primitive is the third + variants.push_back("#define USE_LIGHTING\n#define USE_PRIMITIVE\n#define USE_POINT_SIZE\n"); //points need point size + variants.push_back("#define USE_LIGHTING\n#define USE_ATTRIBUTES\n"); // attributes for vertex arrays + variants.push_back("#define USE_LIGHTING\n#define USE_ATTRIBUTES\n#define USE_POINT_SIZE\n"); //attributes with point size + + shader.canvas_shader.initialize(variants, global_defines); + + shader.default_version = shader.canvas_shader.version_create(); + shader.default_version_rd_shader = shader.canvas_shader.version_get_shader(shader.default_version, SHADER_VARIANT_QUAD); + + RD::PipelineColorBlendState blend_state; + RD::PipelineColorBlendState::Attachment blend_attachment; + + blend_attachment.enable_blend = true; + blend_attachment.color_blend_op = RD::BLEND_OP_ADD; + blend_attachment.src_color_blend_factor = RD::BLEND_FACTOR_SRC_ALPHA; + blend_attachment.dst_color_blend_factor = RD::BLEND_FACTOR_ONE_MINUS_SRC_ALPHA; + + blend_attachment.alpha_blend_op = RD::BLEND_OP_ADD; + blend_attachment.src_alpha_blend_factor = RD::BLEND_FACTOR_ONE; + blend_attachment.dst_alpha_blend_factor = RD::BLEND_FACTOR_ONE_MINUS_SRC_ALPHA; + + blend_state.attachments.push_back(blend_attachment); + + RD::PipelineColorBlendState::Attachment attachment_lcd; + attachment_lcd.enable_blend = true; + attachment_lcd.alpha_blend_op = RD::BLEND_OP_ADD; + attachment_lcd.color_blend_op = RD::BLEND_OP_ADD; + attachment_lcd.src_color_blend_factor = RD::BLEND_FACTOR_CONSTANT_COLOR; + attachment_lcd.dst_color_blend_factor = RD::BLEND_FACTOR_ONE_MINUS_SRC_COLOR; + attachment_lcd.src_alpha_blend_factor = RD::BLEND_FACTOR_ONE; + attachment_lcd.dst_alpha_blend_factor = RD::BLEND_FACTOR_ONE_MINUS_SRC_ALPHA; + + RD::PipelineColorBlendState blend_state_lcd; + blend_state_lcd.attachments.push_back(attachment_lcd); + + for (int i = 0; i < PIPELINE_LIGHT_MODE_MAX; i++) { + for (int j = 0; j < PIPELINE_VARIANT_MAX; j++) { + RD::RenderPrimitive primitive[PIPELINE_VARIANT_MAX] = { + RD::RENDER_PRIMITIVE_TRIANGLES, + RD::RENDER_PRIMITIVE_TRIANGLES, + RD::RENDER_PRIMITIVE_TRIANGLES, + RD::RENDER_PRIMITIVE_LINES, + RD::RENDER_PRIMITIVE_POINTS, + RD::RENDER_PRIMITIVE_TRIANGLES, + RD::RENDER_PRIMITIVE_TRIANGLE_STRIPS, + RD::RENDER_PRIMITIVE_LINES, + RD::RENDER_PRIMITIVE_LINESTRIPS, + RD::RENDER_PRIMITIVE_POINTS, + RD::RENDER_PRIMITIVE_TRIANGLES, + }; + + ShaderVariant shader_variants[PIPELINE_LIGHT_MODE_MAX][PIPELINE_VARIANT_MAX] = { + { + //non lit + SHADER_VARIANT_QUAD, + SHADER_VARIANT_NINEPATCH, + SHADER_VARIANT_PRIMITIVE, + SHADER_VARIANT_PRIMITIVE, + SHADER_VARIANT_PRIMITIVE_POINTS, + SHADER_VARIANT_ATTRIBUTES, + SHADER_VARIANT_ATTRIBUTES, + SHADER_VARIANT_ATTRIBUTES, + SHADER_VARIANT_ATTRIBUTES, + SHADER_VARIANT_ATTRIBUTES_POINTS, + SHADER_VARIANT_QUAD, + }, + { + //lit + SHADER_VARIANT_QUAD_LIGHT, + SHADER_VARIANT_NINEPATCH_LIGHT, + SHADER_VARIANT_PRIMITIVE_LIGHT, + SHADER_VARIANT_PRIMITIVE_LIGHT, + SHADER_VARIANT_PRIMITIVE_POINTS_LIGHT, + SHADER_VARIANT_ATTRIBUTES_LIGHT, + SHADER_VARIANT_ATTRIBUTES_LIGHT, + SHADER_VARIANT_ATTRIBUTES_LIGHT, + SHADER_VARIANT_ATTRIBUTES_LIGHT, + SHADER_VARIANT_ATTRIBUTES_POINTS_LIGHT, + SHADER_VARIANT_QUAD_LIGHT, + }, + }; + + RID shader_variant = shader.canvas_shader.version_get_shader(shader.default_version, shader_variants[i][j]); + if (j == PIPELINE_VARIANT_QUAD_LCD_BLEND) { + shader.pipeline_variants.variants[i][j].setup(shader_variant, primitive[j], RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), blend_state_lcd, RD::DYNAMIC_STATE_BLEND_CONSTANTS); + } else { + shader.pipeline_variants.variants[i][j].setup(shader_variant, primitive[j], RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), blend_state, 0); + } + } + } + } + + { + //shader compiler + ShaderCompiler::DefaultIdentifierActions actions; + + actions.renames["VERTEX"] = "vertex"; + actions.renames["LIGHT_VERTEX"] = "light_vertex"; + actions.renames["SHADOW_VERTEX"] = "shadow_vertex"; + actions.renames["UV"] = "uv"; + actions.renames["POINT_SIZE"] = "gl_PointSize"; + + actions.renames["MODEL_MATRIX"] = "model_matrix"; + actions.renames["CANVAS_MATRIX"] = "canvas_data.canvas_transform"; + actions.renames["SCREEN_MATRIX"] = "canvas_data.screen_transform"; + actions.renames["TIME"] = "canvas_data.time"; + actions.renames["PI"] = _MKSTR(Math_PI); + actions.renames["TAU"] = _MKSTR(Math_TAU); + actions.renames["E"] = _MKSTR(Math_E); + actions.renames["AT_LIGHT_PASS"] = "false"; + actions.renames["INSTANCE_CUSTOM"] = "instance_custom"; + + actions.renames["COLOR"] = "color"; + actions.renames["NORMAL"] = "normal"; + actions.renames["NORMAL_MAP"] = "normal_map"; + actions.renames["NORMAL_MAP_DEPTH"] = "normal_map_depth"; + actions.renames["TEXTURE"] = "color_texture"; + actions.renames["TEXTURE_PIXEL_SIZE"] = "draw_data.color_texture_pixel_size"; + actions.renames["NORMAL_TEXTURE"] = "normal_texture"; + actions.renames["SPECULAR_SHININESS_TEXTURE"] = "specular_texture"; + actions.renames["SPECULAR_SHININESS"] = "specular_shininess"; + actions.renames["SCREEN_UV"] = "screen_uv"; + actions.renames["SCREEN_TEXTURE"] = "screen_texture"; + actions.renames["SCREEN_PIXEL_SIZE"] = "canvas_data.screen_pixel_size"; + actions.renames["FRAGCOORD"] = "gl_FragCoord"; + actions.renames["POINT_COORD"] = "gl_PointCoord"; + actions.renames["INSTANCE_ID"] = "gl_InstanceIndex"; + actions.renames["VERTEX_ID"] = "gl_VertexIndex"; + + actions.renames["LIGHT_POSITION"] = "light_position"; + actions.renames["LIGHT_DIRECTION"] = "light_direction"; + actions.renames["LIGHT_IS_DIRECTIONAL"] = "is_directional"; + actions.renames["LIGHT_COLOR"] = "light_color"; + actions.renames["LIGHT_ENERGY"] = "light_energy"; + actions.renames["LIGHT"] = "light"; + actions.renames["SHADOW_MODULATE"] = "shadow_modulate"; + + actions.renames["texture_sdf"] = "texture_sdf"; + actions.renames["texture_sdf_normal"] = "texture_sdf_normal"; + actions.renames["sdf_to_screen_uv"] = "sdf_to_screen_uv"; + actions.renames["screen_uv_to_sdf"] = "screen_uv_to_sdf"; + + actions.usage_defines["COLOR"] = "#define COLOR_USED\n"; + actions.usage_defines["SCREEN_TEXTURE"] = "#define SCREEN_TEXTURE_USED\n"; + actions.usage_defines["SCREEN_UV"] = "#define SCREEN_UV_USED\n"; + actions.usage_defines["SCREEN_PIXEL_SIZE"] = "@SCREEN_UV"; + actions.usage_defines["NORMAL"] = "#define NORMAL_USED\n"; + actions.usage_defines["NORMAL_MAP"] = "#define NORMAL_MAP_USED\n"; + actions.usage_defines["LIGHT"] = "#define LIGHT_SHADER_CODE_USED\n"; + actions.usage_defines["SPECULAR_SHININESS"] = "#define SPECULAR_SHININESS_USED\n"; + + actions.render_mode_defines["skip_vertex_transform"] = "#define SKIP_TRANSFORM_USED\n"; + actions.render_mode_defines["unshaded"] = "#define MODE_UNSHADED\n"; + actions.render_mode_defines["light_only"] = "#define MODE_LIGHT_ONLY\n"; + + actions.custom_samplers["TEXTURE"] = "texture_sampler"; + actions.custom_samplers["NORMAL_TEXTURE"] = "texture_sampler"; + actions.custom_samplers["SPECULAR_SHININESS_TEXTURE"] = "texture_sampler"; + actions.custom_samplers["SCREEN_TEXTURE"] = "material_samplers[3]"; //mipmap and filter for screen texture + actions.sampler_array_name = "material_samplers"; + actions.base_texture_binding_index = 1; + actions.texture_layout_set = MATERIAL_UNIFORM_SET; + actions.base_uniform_string = "material."; + actions.default_filter = ShaderLanguage::FILTER_LINEAR; + actions.default_repeat = ShaderLanguage::REPEAT_DISABLE; + actions.base_varying_index = 4; + + actions.global_buffer_array_variable = "global_shader_uniforms.data"; + + shader.compiler.initialize(actions); + } + + { //shadow rendering + Vector<String> versions; + versions.push_back("\n#define MODE_SHADOW\n"); //shadow + versions.push_back("\n#define MODE_SDF\n"); //sdf + shadow_render.shader.initialize(versions); + + { + Vector<RD::AttachmentFormat> attachments; + + RD::AttachmentFormat af_color; + af_color.format = RD::DATA_FORMAT_R32_SFLOAT; + af_color.usage_flags = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + + attachments.push_back(af_color); + + RD::AttachmentFormat af_depth; + af_depth.format = RD::DATA_FORMAT_D32_SFLOAT; + af_depth.usage_flags = RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT; + + attachments.push_back(af_depth); + + shadow_render.framebuffer_format = RD::get_singleton()->framebuffer_format_create(attachments); + } + + { + Vector<RD::AttachmentFormat> attachments; + + RD::AttachmentFormat af_color; + af_color.format = RD::DATA_FORMAT_R8_UNORM; + af_color.usage_flags = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + + attachments.push_back(af_color); + + shadow_render.sdf_framebuffer_format = RD::get_singleton()->framebuffer_format_create(attachments); + } + + //pipelines + Vector<RD::VertexAttribute> vf; + RD::VertexAttribute vd; + vd.format = sizeof(real_t) == sizeof(float) ? RD::DATA_FORMAT_R32G32B32_SFLOAT : RD::DATA_FORMAT_R64G64B64_SFLOAT; + vd.location = 0; + vd.offset = 0; + vd.stride = sizeof(real_t) * 3; + vf.push_back(vd); + shadow_render.vertex_format = RD::get_singleton()->vertex_format_create(vf); + + vd.format = sizeof(real_t) == sizeof(float) ? RD::DATA_FORMAT_R32G32_SFLOAT : RD::DATA_FORMAT_R64G64_SFLOAT; + vd.stride = sizeof(real_t) * 2; + + vf.write[0] = vd; + shadow_render.sdf_vertex_format = RD::get_singleton()->vertex_format_create(vf); + + shadow_render.shader_version = shadow_render.shader.version_create(); + + for (int i = 0; i < 3; i++) { + RD::PipelineRasterizationState rs; + rs.cull_mode = i == 0 ? RD::POLYGON_CULL_DISABLED : (i == 1 ? RD::POLYGON_CULL_FRONT : RD::POLYGON_CULL_BACK); + RD::PipelineDepthStencilState ds; + ds.enable_depth_write = true; + ds.enable_depth_test = true; + ds.depth_compare_operator = RD::COMPARE_OP_LESS; + shadow_render.render_pipelines[i] = RD::get_singleton()->render_pipeline_create(shadow_render.shader.version_get_shader(shadow_render.shader_version, SHADOW_RENDER_MODE_SHADOW), shadow_render.framebuffer_format, shadow_render.vertex_format, RD::RENDER_PRIMITIVE_TRIANGLES, rs, RD::PipelineMultisampleState(), ds, RD::PipelineColorBlendState::create_disabled(), 0); + } + + for (int i = 0; i < 2; i++) { + shadow_render.sdf_render_pipelines[i] = RD::get_singleton()->render_pipeline_create(shadow_render.shader.version_get_shader(shadow_render.shader_version, SHADOW_RENDER_MODE_SDF), shadow_render.sdf_framebuffer_format, shadow_render.sdf_vertex_format, i == 0 ? RD::RENDER_PRIMITIVE_TRIANGLES : RD::RENDER_PRIMITIVE_LINES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RD::PipelineColorBlendState::create_disabled(), 0); + } + } + + { //bindings + + state.canvas_state_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(State::Buffer)); + state.lights_uniform_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(LightUniform) * state.max_lights_per_render); + + RD::SamplerState shadow_sampler_state; + shadow_sampler_state.mag_filter = RD::SAMPLER_FILTER_LINEAR; + shadow_sampler_state.min_filter = RD::SAMPLER_FILTER_LINEAR; + shadow_sampler_state.repeat_u = RD::SAMPLER_REPEAT_MODE_REPEAT; //shadow wrap around + shadow_sampler_state.compare_op = RD::COMPARE_OP_GREATER; + shadow_sampler_state.enable_compare = true; + state.shadow_sampler = RD::get_singleton()->sampler_create(shadow_sampler_state); + } + + { + //polygon buffers + polygon_buffers.last_id = 1; + } + + { // default index buffer + + Vector<uint8_t> pv; + pv.resize(6 * 4); + { + uint8_t *w = pv.ptrw(); + int *p32 = (int *)w; + p32[0] = 0; + p32[1] = 1; + p32[2] = 2; + p32[3] = 0; + p32[4] = 2; + p32[5] = 3; + } + shader.quad_index_buffer = RD::get_singleton()->index_buffer_create(6, RenderingDevice::INDEX_BUFFER_FORMAT_UINT32, pv); + shader.quad_index_array = RD::get_singleton()->index_array_create(shader.quad_index_buffer, 0, 6); + } + + { //primitive + primitive_arrays.index_array[0] = shader.quad_index_array = RD::get_singleton()->index_array_create(shader.quad_index_buffer, 0, 1); + primitive_arrays.index_array[1] = shader.quad_index_array = RD::get_singleton()->index_array_create(shader.quad_index_buffer, 0, 2); + primitive_arrays.index_array[2] = shader.quad_index_array = RD::get_singleton()->index_array_create(shader.quad_index_buffer, 0, 3); + primitive_arrays.index_array[3] = shader.quad_index_array = RD::get_singleton()->index_array_create(shader.quad_index_buffer, 0, 6); + } + + { //default skeleton buffer + + shader.default_skeleton_uniform_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(SkeletonUniform)); + SkeletonUniform su; + _update_transform_2d_to_mat4(Transform2D(), su.skeleton_inverse); + _update_transform_2d_to_mat4(Transform2D(), su.skeleton_transform); + RD::get_singleton()->buffer_update(shader.default_skeleton_uniform_buffer, 0, sizeof(SkeletonUniform), &su); + + shader.default_skeleton_texture_buffer = RD::get_singleton()->texture_buffer_create(32, RD::DATA_FORMAT_R32G32B32A32_SFLOAT); + } + { + //default shadow texture to keep uniform set happy + RD::TextureFormat tf; + tf.texture_type = RD::TEXTURE_TYPE_2D; + tf.width = 4; + tf.height = 4; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT; + tf.format = RD::DATA_FORMAT_R32_SFLOAT; + + state.shadow_texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); + } + + { + Vector<RD::Uniform> uniforms; + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 0; + u.append_id(RendererRD::MeshStorage::get_singleton()->get_default_rd_storage_buffer()); + uniforms.push_back(u); + } + + state.default_transforms_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, shader.default_version_rd_shader, TRANSFORMS_UNIFORM_SET); + } + + default_canvas_texture = texture_storage->canvas_texture_allocate(); + texture_storage->canvas_texture_initialize(default_canvas_texture); + + state.shadow_texture_size = GLOBAL_GET("rendering/2d/shadow_atlas/size"); + + //create functions for shader and material + material_storage->shader_set_data_request_function(RendererRD::MaterialStorage::SHADER_TYPE_2D, _create_shader_funcs); + material_storage->material_set_data_request_function(RendererRD::MaterialStorage::SHADER_TYPE_2D, _create_material_funcs); + + state.time = 0; + + { + default_canvas_group_shader = material_storage->shader_allocate(); + material_storage->shader_initialize(default_canvas_group_shader); + + material_storage->shader_set_code(default_canvas_group_shader, R"( +// Default CanvasGroup shader. + +shader_type canvas_item; + +void fragment() { + vec4 c = textureLod(SCREEN_TEXTURE, SCREEN_UV, 0.0); + + if (c.a > 0.0001) { + c.rgb /= c.a; + } + + COLOR *= c; +} +)"); + default_canvas_group_material = material_storage->material_allocate(); + material_storage->material_initialize(default_canvas_group_material); + + material_storage->material_set_shader(default_canvas_group_material, default_canvas_group_shader); + } + + { + default_clip_children_shader = material_storage->shader_allocate(); + material_storage->shader_initialize(default_clip_children_shader); + + material_storage->shader_set_code(default_clip_children_shader, R"( +// Default clip children shader. + +shader_type canvas_item; + +void fragment() { + vec4 c = textureLod(SCREEN_TEXTURE, SCREEN_UV, 0.0); + COLOR.rgb = c.rgb; +} +)"); + default_clip_children_material = material_storage->material_allocate(); + material_storage->material_initialize(default_clip_children_material); + + material_storage->material_set_shader(default_clip_children_material, default_clip_children_shader); + } + + static_assert(sizeof(PushConstant) == 128); +} + +bool RendererCanvasRenderRD::free(RID p_rid) { + if (canvas_light_owner.owns(p_rid)) { + CanvasLight *cl = canvas_light_owner.get_or_null(p_rid); + ERR_FAIL_COND_V(!cl, false); + light_set_use_shadow(p_rid, false); + canvas_light_owner.free(p_rid); + } else if (occluder_polygon_owner.owns(p_rid)) { + occluder_polygon_set_shape(p_rid, Vector<Vector2>(), false); + occluder_polygon_owner.free(p_rid); + } else { + return false; + } + + return true; +} + +void RendererCanvasRenderRD::set_shadow_texture_size(int p_size) { + p_size = nearest_power_of_2_templated(p_size); + if (p_size == state.shadow_texture_size) { + return; + } + state.shadow_texture_size = p_size; + if (state.shadow_fb.is_valid()) { + RD::get_singleton()->free(state.shadow_texture); + RD::get_singleton()->free(state.shadow_depth_texture); + state.shadow_fb = RID(); + + { + //create a default shadow texture to keep uniform set happy (and that it gets erased when a new one is created) + RD::TextureFormat tf; + tf.texture_type = RD::TEXTURE_TYPE_2D; + tf.width = 4; + tf.height = 4; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT; + tf.format = RD::DATA_FORMAT_R32_SFLOAT; + + state.shadow_texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); + } + } +} + +RendererCanvasRenderRD::~RendererCanvasRenderRD() { + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + //canvas state + + material_storage->material_free(default_canvas_group_material); + material_storage->shader_free(default_canvas_group_shader); + + material_storage->material_free(default_clip_children_material); + material_storage->shader_free(default_clip_children_shader); + + { + if (state.canvas_state_buffer.is_valid()) { + RD::get_singleton()->free(state.canvas_state_buffer); + } + + memdelete_arr(state.light_uniforms); + RD::get_singleton()->free(state.lights_uniform_buffer); + RD::get_singleton()->free(shader.default_skeleton_uniform_buffer); + RD::get_singleton()->free(shader.default_skeleton_texture_buffer); + } + + //shadow rendering + { + shadow_render.shader.version_free(shadow_render.shader_version); + //this will also automatically clear all pipelines + RD::get_singleton()->free(state.shadow_sampler); + } + //bindings + + //shaders + + shader.canvas_shader.version_free(shader.default_version); + + //buffers + { + RD::get_singleton()->free(shader.quad_index_array); + RD::get_singleton()->free(shader.quad_index_buffer); + //primitives are erase by dependency + } + + if (state.shadow_fb.is_valid()) { + RD::get_singleton()->free(state.shadow_depth_texture); + } + RD::get_singleton()->free(state.shadow_texture); + + RendererRD::TextureStorage::get_singleton()->canvas_texture_free(default_canvas_texture); + //pipelines don't need freeing, they are all gone after shaders are gone +} diff --git a/servers/rendering/renderer_rd/renderer_canvas_render_rd.h b/servers/rendering/renderer_rd/renderer_canvas_render_rd.h new file mode 100644 index 0000000000..4f1f77af5e --- /dev/null +++ b/servers/rendering/renderer_rd/renderer_canvas_render_rd.h @@ -0,0 +1,473 @@ +/*************************************************************************/ +/* renderer_canvas_render_rd.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef RENDERER_CANVAS_RENDER_RD_H +#define RENDERER_CANVAS_RENDER_RD_H + +#include "servers/rendering/renderer_canvas_render.h" +#include "servers/rendering/renderer_compositor.h" +#include "servers/rendering/renderer_rd/pipeline_cache_rd.h" +#include "servers/rendering/renderer_rd/shaders/canvas.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/canvas_occlusion.glsl.gen.h" +#include "servers/rendering/renderer_rd/storage_rd/material_storage.h" +#include "servers/rendering/rendering_device.h" +#include "servers/rendering/shader_compiler.h" + +class RendererCanvasRenderRD : public RendererCanvasRender { + enum { + BASE_UNIFORM_SET = 0, + MATERIAL_UNIFORM_SET = 1, + TRANSFORMS_UNIFORM_SET = 2, + CANVAS_TEXTURE_UNIFORM_SET = 3, + }; + + enum ShaderVariant { + SHADER_VARIANT_QUAD, + SHADER_VARIANT_NINEPATCH, + SHADER_VARIANT_PRIMITIVE, + SHADER_VARIANT_PRIMITIVE_POINTS, + SHADER_VARIANT_ATTRIBUTES, + SHADER_VARIANT_ATTRIBUTES_POINTS, + SHADER_VARIANT_QUAD_LIGHT, + SHADER_VARIANT_NINEPATCH_LIGHT, + SHADER_VARIANT_PRIMITIVE_LIGHT, + SHADER_VARIANT_PRIMITIVE_POINTS_LIGHT, + SHADER_VARIANT_ATTRIBUTES_LIGHT, + SHADER_VARIANT_ATTRIBUTES_POINTS_LIGHT, + SHADER_VARIANT_MAX + }; + + enum { + + FLAGS_INSTANCING_MASK = 0x7F, + FLAGS_INSTANCING_HAS_COLORS = (1 << 7), + FLAGS_INSTANCING_HAS_CUSTOM_DATA = (1 << 8), + + FLAGS_CLIP_RECT_UV = (1 << 9), + FLAGS_TRANSPOSE_RECT = (1 << 10), + + FLAGS_NINEPACH_DRAW_CENTER = (1 << 12), + FLAGS_USING_PARTICLES = (1 << 13), + + FLAGS_USE_SKELETON = (1 << 15), + FLAGS_NINEPATCH_H_MODE_SHIFT = 16, + FLAGS_NINEPATCH_V_MODE_SHIFT = 18, + FLAGS_LIGHT_COUNT_SHIFT = 20, + + FLAGS_DEFAULT_NORMAL_MAP_USED = (1 << 26), + FLAGS_DEFAULT_SPECULAR_MAP_USED = (1 << 27), + + FLAGS_USE_MSDF = (1 << 28), + FLAGS_USE_LCD = (1 << 29), + }; + + enum { + LIGHT_FLAGS_TEXTURE_MASK = 0xFFFF, + LIGHT_FLAGS_BLEND_SHIFT = 16, + LIGHT_FLAGS_BLEND_MASK = (3 << 16), + LIGHT_FLAGS_BLEND_MODE_ADD = (0 << 16), + LIGHT_FLAGS_BLEND_MODE_SUB = (1 << 16), + LIGHT_FLAGS_BLEND_MODE_MIX = (2 << 16), + LIGHT_FLAGS_BLEND_MODE_MASK = (3 << 16), + LIGHT_FLAGS_HAS_SHADOW = (1 << 20), + LIGHT_FLAGS_FILTER_SHIFT = 22 + + }; + + enum { + MAX_RENDER_ITEMS = 256 * 1024, + MAX_LIGHT_TEXTURES = 1024, + MAX_LIGHTS_PER_ITEM = 16, + DEFAULT_MAX_LIGHTS_PER_RENDER = 256 + }; + + /****************/ + /**** SHADER ****/ + /****************/ + + enum PipelineVariant { + PIPELINE_VARIANT_QUAD, + PIPELINE_VARIANT_NINEPATCH, + PIPELINE_VARIANT_PRIMITIVE_TRIANGLES, + PIPELINE_VARIANT_PRIMITIVE_LINES, + PIPELINE_VARIANT_PRIMITIVE_POINTS, + PIPELINE_VARIANT_ATTRIBUTE_TRIANGLES, + PIPELINE_VARIANT_ATTRIBUTE_TRIANGLE_STRIP, + PIPELINE_VARIANT_ATTRIBUTE_LINES, + PIPELINE_VARIANT_ATTRIBUTE_LINES_STRIP, + PIPELINE_VARIANT_ATTRIBUTE_POINTS, + PIPELINE_VARIANT_QUAD_LCD_BLEND, + PIPELINE_VARIANT_MAX + }; + enum PipelineLightMode { + PIPELINE_LIGHT_MODE_DISABLED, + PIPELINE_LIGHT_MODE_ENABLED, + PIPELINE_LIGHT_MODE_MAX + }; + + struct PipelineVariants { + PipelineCacheRD variants[PIPELINE_LIGHT_MODE_MAX][PIPELINE_VARIANT_MAX]; + }; + + struct { + CanvasShaderRD canvas_shader; + RID default_version; + RID default_version_rd_shader; + RID quad_index_buffer; + RID quad_index_array; + PipelineVariants pipeline_variants; + + // default_skeleton uniform set + RID default_skeleton_uniform_buffer; + RID default_skeleton_texture_buffer; + + ShaderCompiler compiler; + } shader; + + struct CanvasShaderData : public RendererRD::MaterialStorage::ShaderData { + enum BlendMode { //used internally + BLEND_MODE_MIX, + BLEND_MODE_ADD, + BLEND_MODE_SUB, + BLEND_MODE_MUL, + BLEND_MODE_PMALPHA, + BLEND_MODE_DISABLED, + }; + + bool valid = false; + RID version; + PipelineVariants pipeline_variants; + String path; + + HashMap<StringName, ShaderLanguage::ShaderNode::Uniform> uniforms; + Vector<ShaderCompiler::GeneratedCode::Texture> texture_uniforms; + + Vector<uint32_t> ubo_offsets; + uint32_t ubo_size = 0; + + String code; + HashMap<StringName, HashMap<int, RID>> default_texture_params; + + bool uses_screen_texture = false; + bool uses_screen_texture_mipmaps = false; + bool uses_sdf = false; + bool uses_time = false; + + virtual void set_code(const String &p_Code); + virtual void set_path_hint(const String &p_path); + virtual void set_default_texture_parameter(const StringName &p_name, RID p_texture, int p_index); + virtual void get_shader_uniform_list(List<PropertyInfo> *p_param_list) const; + virtual void get_instance_param_list(List<RendererMaterialStorage::InstanceShaderParam> *p_param_list) const; + + virtual bool is_parameter_texture(const StringName &p_param) const; + virtual bool is_animated() const; + virtual bool casts_shadows() const; + virtual Variant get_default_parameter(const StringName &p_parameter) const; + virtual RS::ShaderNativeSourceCode get_native_source_code() const; + + CanvasShaderData() {} + virtual ~CanvasShaderData(); + }; + + RendererRD::MaterialStorage::ShaderData *_create_shader_func(); + static RendererRD::MaterialStorage::ShaderData *_create_shader_funcs() { + return static_cast<RendererCanvasRenderRD *>(singleton)->_create_shader_func(); + } + + struct CanvasMaterialData : public RendererRD::MaterialStorage::MaterialData { + CanvasShaderData *shader_data = nullptr; + RID uniform_set; + + virtual void set_render_priority(int p_priority) {} + virtual void set_next_pass(RID p_pass) {} + virtual bool update_parameters(const HashMap<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); + virtual ~CanvasMaterialData(); + }; + + RendererRD::MaterialStorage::MaterialData *_create_material_func(CanvasShaderData *p_shader); + static RendererRD::MaterialStorage::MaterialData *_create_material_funcs(RendererRD::MaterialStorage::ShaderData *p_shader) { + return static_cast<RendererCanvasRenderRD *>(singleton)->_create_material_func(static_cast<CanvasShaderData *>(p_shader)); + } + + /**************************/ + /**** CANVAS TEXTURES *****/ + /**************************/ + + struct { + RS::CanvasItemTextureFilter default_filter; + RS::CanvasItemTextureRepeat default_repeat; + } default_samplers; + + /******************/ + /**** POLYGONS ****/ + /******************/ + + struct PolygonBuffers { + RD::VertexFormatID vertex_format_id; + RID vertex_buffer; + RID vertex_array; + RID index_buffer; + RID indices; + }; + + struct { + HashMap<PolygonID, PolygonBuffers> polygons; + PolygonID last_id; + } polygon_buffers; + + /********************/ + /**** PRIMITIVES ****/ + /********************/ + + struct { + RID index_array[4]; + } primitive_arrays; + + /*******************/ + /**** MATERIALS ****/ + /*******************/ + + /******************/ + /**** LIGHTING ****/ + /******************/ + + struct CanvasLight { + RID texture; + struct { + bool enabled = false; + float z_far; + float y_offset; + Transform2D directional_xform; + } shadow; + }; + + RID_Owner<CanvasLight> canvas_light_owner; + + struct ShadowRenderPushConstant { + float projection[16]; + float modelview[8]; + float direction[2]; + float z_far; + float pad; + }; + + struct OccluderPolygon { + RS::CanvasOccluderPolygonCullMode cull_mode; + int line_point_count; + RID vertex_buffer; + RID vertex_array; + RID index_buffer; + RID index_array; + + int sdf_point_count; + int sdf_index_count; + RID sdf_vertex_buffer; + RID sdf_vertex_array; + RID sdf_index_buffer; + RID sdf_index_array; + bool sdf_is_lines; + }; + + struct LightUniform { + float matrix[8]; //light to texture coordinate matrix + float shadow_matrix[8]; //light to shadow coordinate matrix + float color[4]; + + uint8_t shadow_color[4]; + uint32_t flags; //index to light texture + float shadow_pixel_size; + float height; + + float position[2]; + float shadow_z_far_inv; + float shadow_y_ofs; + + float atlas_rect[4]; + }; + + RID_Owner<OccluderPolygon> occluder_polygon_owner; + + enum ShadowRenderMode { + SHADOW_RENDER_MODE_SHADOW, + SHADOW_RENDER_MODE_SDF, + }; + + enum { + SHADOW_RENDER_SDF_TRIANGLES, + SHADOW_RENDER_SDF_LINES, + }; + + struct { + CanvasOcclusionShaderRD shader; + RID shader_version; + RID render_pipelines[3]; + RID sdf_render_pipelines[2]; + RD::VertexFormatID vertex_format; + RD::VertexFormatID sdf_vertex_format; + RD::FramebufferFormatID framebuffer_format; + RD::FramebufferFormatID sdf_framebuffer_format; + } shadow_render; + + /***************/ + /**** STATE ****/ + /***************/ + + //state that does not vary across rendering all items + + struct State { + //state buffer + struct Buffer { + float canvas_transform[16]; + float screen_transform[16]; + float canvas_normal_transform[16]; + float canvas_modulate[4]; + + float screen_pixel_size[2]; + float time; + uint32_t use_pixel_snap; + + float sdf_to_tex[4]; + float sdf_to_screen[2]; + float screen_to_sdf[2]; + + uint32_t directional_light_count; + float tex_to_sdf; + uint32_t pad1; + uint32_t pad2; + }; + + LightUniform *light_uniforms = nullptr; + + RID lights_uniform_buffer; + RID canvas_state_buffer; + RID shadow_sampler; + RID shadow_texture; + RID shadow_depth_texture; + RID shadow_fb; + int shadow_texture_size = 2048; + + RID default_transforms_uniform_set; + + uint32_t max_lights_per_render; + uint32_t max_lights_per_item; + + double time; + + } state; + + struct PushConstant { + float world[6]; + uint32_t flags; + uint32_t specular_shininess; + union { + //rect + struct { + float modulation[4]; + union { + float msdf[4]; + float ninepatch_margins[4]; + }; + float dst_rect[4]; + float src_rect[4]; + float pad[2]; + }; + //primitive + struct { + float points[6]; // vec2 points[3] + float uvs[6]; // vec2 points[3] + uint32_t colors[6]; // colors encoded as half + }; + }; + float color_texture_pixel_size[2]; + uint32_t lights[4]; + }; + + struct SkeletonUniform { + float skeleton_transform[16]; + float skeleton_inverse[16]; + }; + + Item *items[MAX_RENDER_ITEMS]; + + bool using_directional_lights = false; + RID default_canvas_texture; + + RID default_canvas_group_shader; + RID default_canvas_group_material; + RID default_clip_children_material; + RID default_clip_children_shader; + + RS::CanvasItemTextureFilter default_filter = RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR; + RS::CanvasItemTextureRepeat default_repeat = RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED; + + RID _create_base_uniform_set(RID p_to_render_target, bool p_backbuffer); + + inline void _bind_canvas_texture(RD::DrawListID p_draw_list, RID p_texture, RS::CanvasItemTextureFilter p_base_filter, RS::CanvasItemTextureRepeat p_base_repeat, RID &r_last_texture, PushConstant &push_constant, Size2 &r_texpixel_size); //recursive, so regular inline used instead. + void _render_item(RenderingDevice::DrawListID p_draw_list, RID p_render_target, const Item *p_item, RenderingDevice::FramebufferFormatID p_framebuffer_format, const Transform2D &p_canvas_transform_inverse, Item *¤t_clip, Light *p_lights, PipelineVariants *p_pipeline_variants); + void _render_items(RID p_to_render_target, int p_item_count, const Transform2D &p_canvas_transform_inverse, Light *p_lights, bool p_to_backbuffer = false); + + _FORCE_INLINE_ void _update_transform_2d_to_mat2x4(const Transform2D &p_transform, float *p_mat2x4); + _FORCE_INLINE_ void _update_transform_2d_to_mat2x3(const Transform2D &p_transform, float *p_mat2x3); + + _FORCE_INLINE_ void _update_transform_2d_to_mat4(const Transform2D &p_transform, float *p_mat4); + _FORCE_INLINE_ void _update_transform_to_mat4(const Transform3D &p_transform, float *p_mat4); + + void _update_shadow_atlas(); + +public: + PolygonID request_polygon(const Vector<int> &p_indices, const Vector<Point2> &p_points, const Vector<Color> &p_colors, const Vector<Point2> &p_uvs = Vector<Point2>(), const Vector<int> &p_bones = Vector<int>(), const Vector<float> &p_weights = Vector<float>()); + void free_polygon(PolygonID p_polygon); + + RID light_create(); + void light_set_texture(RID p_rid, RID p_texture); + void light_set_use_shadow(RID p_rid, bool p_enable); + void light_update_shadow(RID p_rid, int p_shadow_index, const Transform2D &p_light_xform, int p_light_mask, float p_near, float p_far, LightOccluderInstance *p_occluders); + void light_update_directional_shadow(RID p_rid, int p_shadow_index, const Transform2D &p_light_xform, int p_light_mask, float p_cull_distance, const Rect2 &p_clip_rect, LightOccluderInstance *p_occluders); + + virtual void render_sdf(RID p_render_target, LightOccluderInstance *p_occluders); + + RID occluder_polygon_create(); + void occluder_polygon_set_shape(RID p_occluder, const Vector<Vector2> &p_points, bool p_closed); + void occluder_polygon_set_cull_mode(RID p_occluder, RS::CanvasOccluderPolygonCullMode p_mode); + + void canvas_render_items(RID p_to_render_target, Item *p_item_list, const Color &p_modulate, Light *p_light_list, Light *p_directional_light_list, const Transform2D &p_canvas_transform, RS::CanvasItemTextureFilter p_default_filter, RS::CanvasItemTextureRepeat p_default_repeat, bool p_snap_2d_vertices_to_pixel, bool &r_sdf_used); + + void canvas_debug_viewport_shadows(Light *p_lights_with_shadow) {} + + virtual void set_shadow_texture_size(int p_size); + + void set_time(double p_time); + void update(); + bool free(RID p_rid); + RendererCanvasRenderRD(); + ~RendererCanvasRenderRD(); +}; + +#endif // RENDERER_CANVAS_RENDER_RD_H diff --git a/servers/rendering/renderer_rd/renderer_compositor_rd.cpp b/servers/rendering/renderer_rd/renderer_compositor_rd.cpp new file mode 100644 index 0000000000..3289bfb0ea --- /dev/null +++ b/servers/rendering/renderer_rd/renderer_compositor_rd.cpp @@ -0,0 +1,325 @@ +/*************************************************************************/ +/* renderer_compositor_rd.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "renderer_compositor_rd.h" + +#include "core/config/project_settings.h" +#include "core/io/dir_access.h" + +void RendererCompositorRD::prepare_for_blitting_render_targets() { + RD::get_singleton()->prepare_screen_for_drawing(); +} + +void RendererCompositorRD::blit_render_targets_to_screen(DisplayServer::WindowID p_screen, const BlitToScreen *p_render_targets, int p_amount) { + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin_for_screen(p_screen); + if (draw_list == RD::INVALID_ID) { + return; // Window is minimized and does not have valid swapchain, skip drawing without printing errors. + } + + for (int i = 0; i < p_amount; i++) { + RID rd_texture = texture_storage->render_target_get_rd_texture(p_render_targets[i].render_target); + ERR_CONTINUE(rd_texture.is_null()); + + if (!render_target_descriptors.has(rd_texture) || !RD::get_singleton()->uniform_set_is_valid(render_target_descriptors[rd_texture])) { + Vector<RD::Uniform> uniforms; + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; + u.binding = 0; + u.append_id(blit.sampler); + u.append_id(rd_texture); + uniforms.push_back(u); + RID uniform_set = RD::get_singleton()->uniform_set_create(uniforms, blit.shader.version_get_shader(blit.shader_version, BLIT_MODE_NORMAL), 0); + + render_target_descriptors[rd_texture] = uniform_set; + } + + Size2 screen_size(RD::get_singleton()->screen_get_width(p_screen), RD::get_singleton()->screen_get_height(p_screen)); + BlitMode mode = p_render_targets[i].lens_distortion.apply ? BLIT_MODE_LENS : (p_render_targets[i].multi_view.use_layer ? BLIT_MODE_USE_LAYER : BLIT_MODE_NORMAL); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, blit.pipelines[mode]); + RD::get_singleton()->draw_list_bind_index_array(draw_list, blit.array); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, render_target_descriptors[rd_texture], 0); + + blit.push_constant.src_rect[0] = p_render_targets[i].src_rect.position.x; + blit.push_constant.src_rect[1] = p_render_targets[i].src_rect.position.y; + blit.push_constant.src_rect[2] = p_render_targets[i].src_rect.size.width; + blit.push_constant.src_rect[3] = p_render_targets[i].src_rect.size.height; + blit.push_constant.dst_rect[0] = p_render_targets[i].dst_rect.position.x / screen_size.width; + blit.push_constant.dst_rect[1] = p_render_targets[i].dst_rect.position.y / screen_size.height; + blit.push_constant.dst_rect[2] = p_render_targets[i].dst_rect.size.width / screen_size.width; + blit.push_constant.dst_rect[3] = p_render_targets[i].dst_rect.size.height / screen_size.height; + blit.push_constant.layer = p_render_targets[i].multi_view.layer; + blit.push_constant.eye_center[0] = p_render_targets[i].lens_distortion.eye_center.x; + blit.push_constant.eye_center[1] = p_render_targets[i].lens_distortion.eye_center.y; + blit.push_constant.k1 = p_render_targets[i].lens_distortion.k1; + blit.push_constant.k2 = p_render_targets[i].lens_distortion.k2; + blit.push_constant.upscale = p_render_targets[i].lens_distortion.upscale; + blit.push_constant.aspect_ratio = p_render_targets[i].lens_distortion.aspect_ratio; + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &blit.push_constant, sizeof(BlitPushConstant)); + RD::get_singleton()->draw_list_draw(draw_list, true); + } + + RD::get_singleton()->draw_list_end(); +} + +void RendererCompositorRD::begin_frame(double frame_step) { + frame++; + delta = frame_step; + time += frame_step; + + double time_roll_over = GLOBAL_GET("rendering/limits/time/time_rollover_secs"); + time = Math::fmod(time, time_roll_over); + + canvas->set_time(time); + scene->set_time(time, frame_step); +} + +void RendererCompositorRD::end_frame(bool p_swap_buffers) { + // TODO: Likely pass a bool to swap buffers to avoid display? + RD::get_singleton()->swap_buffers(); +} + +void RendererCompositorRD::initialize() { + { + // Initialize blit + Vector<String> blit_modes; + blit_modes.push_back("\n"); + blit_modes.push_back("\n#define USE_LAYER\n"); + blit_modes.push_back("\n#define USE_LAYER\n#define APPLY_LENS_DISTORTION\n"); + blit_modes.push_back("\n"); + + blit.shader.initialize(blit_modes); + + blit.shader_version = blit.shader.version_create(); + + for (int i = 0; i < BLIT_MODE_MAX; i++) { + blit.pipelines[i] = RD::get_singleton()->render_pipeline_create(blit.shader.version_get_shader(blit.shader_version, i), RD::get_singleton()->screen_get_framebuffer_format(), RD::INVALID_ID, RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), i == BLIT_MODE_NORMAL_ALPHA ? RenderingDevice::PipelineColorBlendState::create_blend() : RenderingDevice::PipelineColorBlendState::create_disabled(), 0); + } + + //create index array for copy shader + Vector<uint8_t> pv; + pv.resize(6 * 4); + { + uint8_t *w = pv.ptrw(); + int *p32 = (int *)w; + p32[0] = 0; + p32[1] = 1; + p32[2] = 2; + p32[3] = 0; + p32[4] = 2; + p32[5] = 3; + } + blit.index_buffer = RD::get_singleton()->index_buffer_create(6, RenderingDevice::INDEX_BUFFER_FORMAT_UINT32, pv); + blit.array = RD::get_singleton()->index_array_create(blit.index_buffer, 0, 6); + + blit.sampler = RD::get_singleton()->sampler_create(RD::SamplerState()); + } +} + +uint64_t RendererCompositorRD::frame = 1; + +void RendererCompositorRD::finalize() { + memdelete(scene); + memdelete(canvas); + memdelete(effects); + memdelete(fog); + memdelete(particles_storage); + memdelete(light_storage); + memdelete(mesh_storage); + memdelete(material_storage); + memdelete(texture_storage); + memdelete(utilities); + + //only need to erase these, the rest are erased by cascade + blit.shader.version_free(blit.shader_version); + RD::get_singleton()->free(blit.index_buffer); + RD::get_singleton()->free(blit.sampler); +} + +void RendererCompositorRD::set_boot_image(const Ref<Image> &p_image, const Color &p_color, bool p_scale, bool p_use_filter) { + if (p_image.is_null() || p_image->is_empty()) { + return; + } + + RD::get_singleton()->prepare_screen_for_drawing(); + + RID texture = texture_storage->texture_allocate(); + texture_storage->texture_2d_initialize(texture, p_image); + RID rd_texture = texture_storage->texture_get_rd_texture(texture); + + RID uset; + { + Vector<RD::Uniform> uniforms; + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; + u.binding = 0; + u.append_id(blit.sampler); + u.append_id(rd_texture); + uniforms.push_back(u); + uset = RD::get_singleton()->uniform_set_create(uniforms, blit.shader.version_get_shader(blit.shader_version, BLIT_MODE_NORMAL), 0); + } + + Size2 window_size = DisplayServer::get_singleton()->window_get_size(); + + Rect2 imgrect(0, 0, p_image->get_width(), p_image->get_height()); + Rect2 screenrect; + if (p_scale) { + if (window_size.width > window_size.height) { + //scale horizontally + screenrect.size.y = window_size.height; + screenrect.size.x = imgrect.size.x * window_size.height / imgrect.size.y; + screenrect.position.x = (window_size.width - screenrect.size.x) / 2; + + } else { + //scale vertically + screenrect.size.x = window_size.width; + screenrect.size.y = imgrect.size.y * window_size.width / imgrect.size.x; + screenrect.position.y = (window_size.height - screenrect.size.y) / 2; + } + } else { + screenrect = imgrect; + screenrect.position += ((window_size - screenrect.size) / 2.0).floor(); + } + + screenrect.position /= window_size; + screenrect.size /= window_size; + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin_for_screen(DisplayServer::MAIN_WINDOW_ID, p_color); + + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, blit.pipelines[BLIT_MODE_NORMAL_ALPHA]); + RD::get_singleton()->draw_list_bind_index_array(draw_list, blit.array); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uset, 0); + + blit.push_constant.src_rect[0] = 0.0; + blit.push_constant.src_rect[1] = 0.0; + blit.push_constant.src_rect[2] = 1.0; + blit.push_constant.src_rect[3] = 1.0; + blit.push_constant.dst_rect[0] = screenrect.position.x; + blit.push_constant.dst_rect[1] = screenrect.position.y; + blit.push_constant.dst_rect[2] = screenrect.size.width; + blit.push_constant.dst_rect[3] = screenrect.size.height; + blit.push_constant.layer = 0; + blit.push_constant.eye_center[0] = 0; + blit.push_constant.eye_center[1] = 0; + blit.push_constant.k1 = 0; + blit.push_constant.k2 = 0; + blit.push_constant.upscale = 1.0; + blit.push_constant.aspect_ratio = 1.0; + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &blit.push_constant, sizeof(BlitPushConstant)); + RD::get_singleton()->draw_list_draw(draw_list, true); + + RD::get_singleton()->draw_list_end(); + + RD::get_singleton()->swap_buffers(); + + texture_storage->texture_free(texture); +} + +RendererCompositorRD *RendererCompositorRD::singleton = nullptr; + +RendererCompositorRD::RendererCompositorRD() { + uniform_set_cache = memnew(UniformSetCacheRD); + framebuffer_cache = memnew(FramebufferCacheRD); + + { + String shader_cache_dir = Engine::get_singleton()->get_shader_cache_path(); + if (shader_cache_dir.is_empty()) { + shader_cache_dir = "user://"; + } + Ref<DirAccess> da = DirAccess::open(shader_cache_dir); + if (da.is_null()) { + ERR_PRINT("Can't create shader cache folder, no shader caching will happen: " + shader_cache_dir); + } else { + Error err = da->change_dir("shader_cache"); + if (err != OK) { + err = da->make_dir("shader_cache"); + } + if (err != OK) { + ERR_PRINT("Can't create shader cache folder, no shader caching will happen: " + shader_cache_dir); + } else { + shader_cache_dir = shader_cache_dir.path_join("shader_cache"); + + bool shader_cache_enabled = GLOBAL_GET("rendering/shader_compiler/shader_cache/enabled"); + if (!Engine::get_singleton()->is_editor_hint() && !shader_cache_enabled) { + shader_cache_dir = String(); //disable only if not editor + } + + if (!shader_cache_dir.is_empty()) { + bool compress = GLOBAL_GET("rendering/shader_compiler/shader_cache/compress"); + bool use_zstd = GLOBAL_GET("rendering/shader_compiler/shader_cache/use_zstd_compression"); + bool strip_debug = GLOBAL_GET("rendering/shader_compiler/shader_cache/strip_debug"); + + ShaderRD::set_shader_cache_dir(shader_cache_dir); + ShaderRD::set_shader_cache_save_compressed(compress); + ShaderRD::set_shader_cache_save_compressed_zstd(use_zstd); + ShaderRD::set_shader_cache_save_debug(!strip_debug); + } + } + } + } + + singleton = this; + + utilities = memnew(RendererRD::Utilities); + texture_storage = memnew(RendererRD::TextureStorage); + material_storage = memnew(RendererRD::MaterialStorage); + mesh_storage = memnew(RendererRD::MeshStorage); + light_storage = memnew(RendererRD::LightStorage); + particles_storage = memnew(RendererRD::ParticlesStorage); + fog = memnew(RendererRD::Fog); + canvas = memnew(RendererCanvasRenderRD()); + + String rendering_method = GLOBAL_GET("rendering/renderer/rendering_method"); + uint64_t textures_per_stage = RD::get_singleton()->limit_get(RD::LIMIT_MAX_TEXTURES_PER_SHADER_STAGE); + + if (rendering_method == "mobile" || textures_per_stage < 48) { + scene = memnew(RendererSceneRenderImplementation::RenderForwardMobile()); + if (rendering_method == "forward_plus") { + WARN_PRINT_ONCE("Platform supports less than 48 textures per stage which is less than required by the Clustered renderer. Defaulting to Mobile renderer."); + } + } else if (rendering_method == "forward_plus") { + // default to our high end renderer + scene = memnew(RendererSceneRenderImplementation::RenderForwardClustered()); + } else { + ERR_FAIL_MSG("Cannot instantiate RenderingDevice-based renderer with renderer type " + rendering_method); + } + + scene->init(); + + // now we're ready to create our effects, + effects = memnew(EffectsRD(!scene->_render_buffers_can_be_storage())); +} + +RendererCompositorRD::~RendererCompositorRD() { + memdelete(uniform_set_cache); + memdelete(framebuffer_cache); + ShaderRD::set_shader_cache_dir(String()); +} diff --git a/servers/rendering/renderer_rd/renderer_compositor_rd.h b/servers/rendering/renderer_rd/renderer_compositor_rd.h new file mode 100644 index 0000000000..a28335f800 --- /dev/null +++ b/servers/rendering/renderer_rd/renderer_compositor_rd.h @@ -0,0 +1,153 @@ +/*************************************************************************/ +/* renderer_compositor_rd.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef RENDERER_COMPOSITOR_RD_H +#define RENDERER_COMPOSITOR_RD_H + +#include "core/os/os.h" +#include "servers/rendering/renderer_compositor.h" +#include "servers/rendering/renderer_rd/effects_rd.h" +#include "servers/rendering/renderer_rd/environment/fog.h" +#include "servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h" +#include "servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.h" +#include "servers/rendering/renderer_rd/framebuffer_cache_rd.h" +#include "servers/rendering/renderer_rd/renderer_canvas_render_rd.h" +#include "servers/rendering/renderer_rd/shaders/blit.glsl.gen.h" +#include "servers/rendering/renderer_rd/storage_rd/light_storage.h" +#include "servers/rendering/renderer_rd/storage_rd/material_storage.h" +#include "servers/rendering/renderer_rd/storage_rd/mesh_storage.h" +#include "servers/rendering/renderer_rd/storage_rd/particles_storage.h" +#include "servers/rendering/renderer_rd/storage_rd/texture_storage.h" +#include "servers/rendering/renderer_rd/storage_rd/utilities.h" +#include "servers/rendering/renderer_rd/uniform_set_cache_rd.h" + +class RendererCompositorRD : public RendererCompositor { +protected: + UniformSetCacheRD *uniform_set_cache = nullptr; + FramebufferCacheRD *framebuffer_cache = nullptr; + RendererCanvasRenderRD *canvas = nullptr; + RendererRD::Utilities *utilities = nullptr; + RendererRD::LightStorage *light_storage = nullptr; + RendererRD::MaterialStorage *material_storage = nullptr; + RendererRD::MeshStorage *mesh_storage = nullptr; + RendererRD::ParticlesStorage *particles_storage = nullptr; + RendererRD::TextureStorage *texture_storage = nullptr; + RendererRD::Fog *fog = nullptr; + EffectsRD *effects = nullptr; + RendererSceneRenderRD *scene = nullptr; + + enum BlitMode { + BLIT_MODE_NORMAL, + BLIT_MODE_USE_LAYER, + BLIT_MODE_LENS, + BLIT_MODE_NORMAL_ALPHA, + BLIT_MODE_MAX + }; + + struct BlitPushConstant { + float src_rect[4]; + float dst_rect[4]; + + float eye_center[2]; + float k1; + float k2; + + float upscale; + float aspect_ratio; + uint32_t layer; + uint32_t pad1; + }; + + struct Blit { + BlitPushConstant push_constant; + BlitShaderRD shader; + RID shader_version; + RID pipelines[BLIT_MODE_MAX]; + RID index_buffer; + RID array; + RID sampler; + } blit; + + HashMap<RID, RID> render_target_descriptors; + + double time = 0.0; + double delta = 0.0; + + static uint64_t frame; + +public: + RendererUtilities *get_utilities() { return utilities; }; + RendererLightStorage *get_light_storage() { return light_storage; } + RendererMaterialStorage *get_material_storage() { return material_storage; } + RendererMeshStorage *get_mesh_storage() { return mesh_storage; } + RendererParticlesStorage *get_particles_storage() { return particles_storage; } + RendererTextureStorage *get_texture_storage() { return texture_storage; } + RendererGI *get_gi() { + ERR_FAIL_NULL_V(scene, nullptr); + return scene->get_gi(); + } + RendererFog *get_fog() { return fog; } + EffectsRD *get_effects() { return effects; } + RendererCanvasRender *get_canvas() { return canvas; } + RendererSceneRender *get_scene() { return scene; } + + void set_boot_image(const Ref<Image> &p_image, const Color &p_color, bool p_scale, bool p_use_filter); + + void initialize(); + void begin_frame(double frame_step); + void prepare_for_blitting_render_targets(); + void blit_render_targets_to_screen(DisplayServer::WindowID p_screen, const BlitToScreen *p_render_targets, int p_amount); + + void end_frame(bool p_swap_buffers); + void finalize(); + + _ALWAYS_INLINE_ uint64_t get_frame_number() const { return frame; } + _ALWAYS_INLINE_ double get_frame_delta_time() const { return delta; } + _ALWAYS_INLINE_ double get_total_time() const { return time; } + + static Error is_viable() { + return OK; + } + + static RendererCompositor *_create_current() { + return memnew(RendererCompositorRD); + } + + static void make_current() { + _create_func = _create_current; + low_end = false; + } + + static RendererCompositorRD *singleton; + RendererCompositorRD(); + ~RendererCompositorRD(); +}; + +#endif // RENDERER_COMPOSITOR_RD_H diff --git a/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp b/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp new file mode 100644 index 0000000000..75fe84f46b --- /dev/null +++ b/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp @@ -0,0 +1,1470 @@ +/*************************************************************************/ +/* renderer_scene_render_rd.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "renderer_scene_render_rd.h" + +#include "core/config/project_settings.h" +#include "core/os/os.h" +#include "renderer_compositor_rd.h" +#include "servers/rendering/renderer_rd/environment/fog.h" +#include "servers/rendering/renderer_rd/storage_rd/material_storage.h" +#include "servers/rendering/renderer_rd/storage_rd/texture_storage.h" +#include "servers/rendering/rendering_server_default.h" +#include "servers/rendering/storage/camera_attributes_storage.h" + +void get_vogel_disk(float *r_kernel, int p_sample_count) { + const float golden_angle = 2.4; + + for (int i = 0; i < p_sample_count; i++) { + float r = Math::sqrt(float(i) + 0.5) / Math::sqrt(float(p_sample_count)); + float theta = float(i) * golden_angle; + + r_kernel[i * 4] = Math::cos(theta) * r; + r_kernel[i * 4 + 1] = Math::sin(theta) * r; + } +} + +void RendererSceneRenderRD::sdfgi_update(const Ref<RenderSceneBuffers> &p_render_buffers, RID p_environment, const Vector3 &p_world_position) { + Ref<RenderSceneBuffersRD> rb = p_render_buffers; + ERR_FAIL_COND(rb.is_null()); + Ref<RendererRD::GI::SDFGI> sdfgi; + if (rb->has_custom_data(RB_SCOPE_SDFGI)) { + sdfgi = rb->get_custom_data(RB_SCOPE_SDFGI); + } + + bool needs_sdfgi = p_environment.is_valid() && environment_get_sdfgi_enabled(p_environment); + + if (!needs_sdfgi) { + if (sdfgi.is_valid()) { + // delete it + sdfgi.unref(); + rb->set_custom_data(RB_SCOPE_SDFGI, sdfgi); + } + return; + } + + static const uint32_t history_frames_to_converge[RS::ENV_SDFGI_CONVERGE_MAX] = { 5, 10, 15, 20, 25, 30 }; + uint32_t requested_history_size = history_frames_to_converge[gi.sdfgi_frames_to_converge]; + + if (sdfgi.is_valid() && (sdfgi->num_cascades != environment_get_sdfgi_cascades(p_environment) || sdfgi->min_cell_size != environment_get_sdfgi_min_cell_size(p_environment) || requested_history_size != sdfgi->history_size || sdfgi->uses_occlusion != environment_get_sdfgi_use_occlusion(p_environment) || sdfgi->y_scale_mode != environment_get_sdfgi_y_scale(p_environment))) { + //configuration changed, erase + sdfgi.unref(); + rb->set_custom_data(RB_SCOPE_SDFGI, sdfgi); + } + + if (sdfgi.is_null()) { + // re-create + sdfgi = gi.create_sdfgi(p_environment, p_world_position, requested_history_size); + rb->set_custom_data(RB_SCOPE_SDFGI, sdfgi); + } else { + //check for updates + sdfgi->update(p_environment, p_world_position); + } +} + +int RendererSceneRenderRD::sdfgi_get_pending_region_count(const Ref<RenderSceneBuffers> &p_render_buffers) const { + Ref<RenderSceneBuffersRD> rb = p_render_buffers; + ERR_FAIL_COND_V(rb.is_null(), 0); + + if (!rb->has_custom_data(RB_SCOPE_SDFGI)) { + return 0; + } + Ref<RendererRD::GI::SDFGI> sdfgi = rb->get_custom_data(RB_SCOPE_SDFGI); + + int dirty_count = 0; + for (uint32_t i = 0; i < sdfgi->cascades.size(); i++) { + const RendererRD::GI::SDFGI::Cascade &c = sdfgi->cascades[i]; + + if (c.dirty_regions == RendererRD::GI::SDFGI::Cascade::DIRTY_ALL) { + dirty_count++; + } else { + for (int j = 0; j < 3; j++) { + if (c.dirty_regions[j] != 0) { + dirty_count++; + } + } + } + } + + return dirty_count; +} + +AABB RendererSceneRenderRD::sdfgi_get_pending_region_bounds(const Ref<RenderSceneBuffers> &p_render_buffers, int p_region) const { + AABB bounds; + Vector3i from; + Vector3i size; + + Ref<RenderSceneBuffersRD> rb = p_render_buffers; + ERR_FAIL_COND_V(rb.is_null(), AABB()); + Ref<RendererRD::GI::SDFGI> sdfgi = rb->get_custom_data(RB_SCOPE_SDFGI); + ERR_FAIL_COND_V(sdfgi.is_null(), AABB()); + + int c = sdfgi->get_pending_region_data(p_region, from, size, bounds); + ERR_FAIL_COND_V(c == -1, AABB()); + return bounds; +} + +uint32_t RendererSceneRenderRD::sdfgi_get_pending_region_cascade(const Ref<RenderSceneBuffers> &p_render_buffers, int p_region) const { + AABB bounds; + Vector3i from; + Vector3i size; + + Ref<RenderSceneBuffersRD> rb = p_render_buffers; + ERR_FAIL_COND_V(rb.is_null(), -1); + Ref<RendererRD::GI::SDFGI> sdfgi = rb->get_custom_data(RB_SCOPE_SDFGI); + ERR_FAIL_COND_V(sdfgi.is_null(), -1); + + return sdfgi->get_pending_region_data(p_region, from, size, bounds); +} + +RID RendererSceneRenderRD::sky_allocate() { + return sky.allocate_sky_rid(); +} +void RendererSceneRenderRD::sky_initialize(RID p_rid) { + sky.initialize_sky_rid(p_rid); +} + +void RendererSceneRenderRD::sky_set_radiance_size(RID p_sky, int p_radiance_size) { + sky.sky_set_radiance_size(p_sky, p_radiance_size); +} + +void RendererSceneRenderRD::sky_set_mode(RID p_sky, RS::SkyMode p_mode) { + sky.sky_set_mode(p_sky, p_mode); +} + +void RendererSceneRenderRD::sky_set_material(RID p_sky, RID p_material) { + sky.sky_set_material(p_sky, p_material); +} + +Ref<Image> RendererSceneRenderRD::sky_bake_panorama(RID p_sky, float p_energy, bool p_bake_irradiance, const Size2i &p_size) { + return sky.sky_bake_panorama(p_sky, p_energy, p_bake_irradiance, p_size); +} + +void RendererSceneRenderRD::environment_glow_set_use_bicubic_upscale(bool p_enable) { + glow_bicubic_upscale = p_enable; +} + +void RendererSceneRenderRD::environment_glow_set_use_high_quality(bool p_enable) { + glow_high_quality = p_enable; +} + +void RendererSceneRenderRD::environment_set_volumetric_fog_volume_size(int p_size, int p_depth) { + volumetric_fog_size = p_size; + volumetric_fog_depth = p_depth; +} + +void RendererSceneRenderRD::environment_set_volumetric_fog_filter_active(bool p_enable) { + volumetric_fog_filter_active = p_enable; +} + +void RendererSceneRenderRD::environment_set_sdfgi_ray_count(RS::EnvironmentSDFGIRayCount p_ray_count) { + gi.sdfgi_ray_count = p_ray_count; +} + +void RendererSceneRenderRD::environment_set_sdfgi_frames_to_converge(RS::EnvironmentSDFGIFramesToConverge p_frames) { + gi.sdfgi_frames_to_converge = p_frames; +} +void RendererSceneRenderRD::environment_set_sdfgi_frames_to_update_light(RS::EnvironmentSDFGIFramesToUpdateLight p_update) { + gi.sdfgi_frames_to_update_light = p_update; +} + +Ref<Image> RendererSceneRenderRD::environment_bake_panorama(RID p_env, bool p_bake_irradiance, const Size2i &p_size) { + ERR_FAIL_COND_V(p_env.is_null(), Ref<Image>()); + + RS::EnvironmentBG environment_background = environment_get_background(p_env); + + if (environment_background == RS::ENV_BG_CAMERA_FEED || environment_background == RS::ENV_BG_CANVAS || environment_background == RS::ENV_BG_KEEP) { + return Ref<Image>(); //nothing to bake + } + + RS::EnvironmentAmbientSource ambient_source = environment_get_ambient_source(p_env); + + bool use_ambient_light = false; + bool use_cube_map = false; + if (ambient_source == RS::ENV_AMBIENT_SOURCE_BG && (environment_background == RS::ENV_BG_CLEAR_COLOR || environment_background == RS::ENV_BG_COLOR)) { + use_ambient_light = true; + } else { + use_cube_map = (ambient_source == RS::ENV_AMBIENT_SOURCE_BG && environment_background == RS::ENV_BG_SKY) || ambient_source == RS::ENV_AMBIENT_SOURCE_SKY; + use_ambient_light = use_cube_map || ambient_source == RS::ENV_AMBIENT_SOURCE_COLOR; + } + use_cube_map = use_cube_map || (environment_background == RS::ENV_BG_SKY && environment_get_sky(p_env).is_valid()); + + Color ambient_color; + float ambient_color_sky_mix = 0.0; + if (use_ambient_light) { + ambient_color_sky_mix = environment_get_ambient_sky_contribution(p_env); + const float ambient_energy = environment_get_ambient_light_energy(p_env); + ambient_color = environment_get_ambient_light(p_env); + ambient_color = ambient_color.srgb_to_linear(); + ambient_color.r *= ambient_energy; + ambient_color.g *= ambient_energy; + ambient_color.b *= ambient_energy; + } + + if (use_cube_map) { + Ref<Image> panorama = sky_bake_panorama(environment_get_sky(p_env), environment_get_bg_energy_multiplier(p_env), p_bake_irradiance, p_size); + if (use_ambient_light) { + for (int x = 0; x < p_size.width; x++) { + for (int y = 0; y < p_size.height; y++) { + panorama->set_pixel(x, y, ambient_color.lerp(panorama->get_pixel(x, y), ambient_color_sky_mix)); + } + } + } + return panorama; + } else { + const float bg_energy_multiplier = environment_get_bg_energy_multiplier(p_env); + Color panorama_color = ((environment_background == RS::ENV_BG_CLEAR_COLOR) ? RSG::texture_storage->get_default_clear_color() : environment_get_bg_color(p_env)); + panorama_color = panorama_color.srgb_to_linear(); + panorama_color.r *= bg_energy_multiplier; + panorama_color.g *= bg_energy_multiplier; + panorama_color.b *= bg_energy_multiplier; + + if (use_ambient_light) { + panorama_color = ambient_color.lerp(panorama_color, ambient_color_sky_mix); + } + + Ref<Image> panorama = Image::create_empty(p_size.width, p_size.height, false, Image::FORMAT_RGBAF); + panorama->fill(panorama_color); + return panorama; + } +} + +/* REFLECTION PROBE */ + +RID RendererSceneRenderRD::reflection_probe_create_framebuffer(RID p_color, RID p_depth) { + Vector<RID> fb; + fb.push_back(p_color); + fb.push_back(p_depth); + return RD::get_singleton()->framebuffer_create(fb); +} + +/* FOG VOLUME INSTANCE */ + +RID RendererSceneRenderRD::fog_volume_instance_create(RID p_fog_volume) { + return RendererRD::Fog::get_singleton()->fog_volume_instance_create(p_fog_volume); +} + +void RendererSceneRenderRD::fog_volume_instance_set_transform(RID p_fog_volume_instance, const Transform3D &p_transform) { + RendererRD::Fog::get_singleton()->fog_volume_instance_set_transform(p_fog_volume_instance, p_transform); +} + +void RendererSceneRenderRD::fog_volume_instance_set_active(RID p_fog_volume_instance, bool p_active) { + RendererRD::Fog::get_singleton()->fog_volume_instance_set_active(p_fog_volume_instance, p_active); +} + +RID RendererSceneRenderRD::fog_volume_instance_get_volume(RID p_fog_volume_instance) const { + return RendererRD::Fog::get_singleton()->fog_volume_instance_get_volume(p_fog_volume_instance); +} + +Vector3 RendererSceneRenderRD::fog_volume_instance_get_position(RID p_fog_volume_instance) const { + return RendererRD::Fog::get_singleton()->fog_volume_instance_get_position(p_fog_volume_instance); +} + +/* VOXEL GI */ + +RID RendererSceneRenderRD::voxel_gi_instance_create(RID p_base) { + return gi.voxel_gi_instance_create(p_base); +} + +void RendererSceneRenderRD::voxel_gi_instance_set_transform_to_data(RID p_probe, const Transform3D &p_xform) { + if (!is_dynamic_gi_supported()) { + return; + } + + gi.voxel_gi_instance_set_transform_to_data(p_probe, p_xform); +} + +bool RendererSceneRenderRD::voxel_gi_needs_update(RID p_probe) const { + if (!is_dynamic_gi_supported()) { + return false; + } + + return gi.voxel_gi_needs_update(p_probe); +} + +void RendererSceneRenderRD::voxel_gi_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RenderGeometryInstance *> &p_dynamic_objects) { + if (!is_dynamic_gi_supported()) { + return; + } + + gi.voxel_gi_update(p_probe, p_update_light_instances, p_light_instances, p_dynamic_objects); +} + +void RendererSceneRenderRD::_debug_sdfgi_probes(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_framebuffer, const uint32_t p_view_count, const Projection *p_camera_with_transforms, bool p_will_continue_color, bool p_will_continue_depth) { + ERR_FAIL_COND(p_render_buffers.is_null()); + + if (!p_render_buffers->has_custom_data(RB_SCOPE_SDFGI)) { + return; //nothing to debug + } + + Ref<RendererRD::GI::SDFGI> sdfgi = p_render_buffers->get_custom_data(RB_SCOPE_SDFGI); + + sdfgi->debug_probes(p_framebuffer, p_view_count, p_camera_with_transforms, p_will_continue_color, p_will_continue_depth); +} + +//////////////////////////////// +Ref<RenderSceneBuffers> RendererSceneRenderRD::render_buffers_create() { + Ref<RenderSceneBuffersRD> rb; + rb.instantiate(); + + rb->set_can_be_storage(_render_buffers_can_be_storage()); + rb->set_max_cluster_elements(max_cluster_elements); + rb->set_base_data_format(_render_buffers_get_color_format()); + if (vrs) { + rb->set_vrs(vrs); + } + + setup_render_buffer_data(rb); + + return rb; +} + +void RendererSceneRenderRD::_allocate_luminance_textures(Ref<RenderSceneBuffersRD> rb) { + ERR_FAIL_COND(!rb->luminance.current.is_null()); + + Size2i internal_size = rb->get_internal_size(); + int w = internal_size.x; + int h = internal_size.y; + + while (true) { + w = MAX(w / 8, 1); + h = MAX(h / 8, 1); + + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R32_SFLOAT; + tf.width = w; + tf.height = h; + + bool final = w == 1 && h == 1; + + if (_render_buffers_can_be_storage()) { + tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT; + if (final) { + tf.usage_bits |= RD::TEXTURE_USAGE_SAMPLING_BIT; + } + } else { + tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; + } + + RID texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); + + rb->luminance.reduce.push_back(texture); + if (!_render_buffers_can_be_storage()) { + Vector<RID> fb; + fb.push_back(texture); + + rb->luminance.fb.push_back(RD::get_singleton()->framebuffer_create(fb)); + } + + if (final) { + rb->luminance.current = RD::get_singleton()->texture_create(tf, RD::TextureView()); + + if (!_render_buffers_can_be_storage()) { + Vector<RID> fb; + fb.push_back(rb->luminance.current); + + rb->luminance.current_fb = RD::get_singleton()->framebuffer_create(fb); + } + break; + } + } +} + +void RendererSceneRenderRD::_render_buffers_copy_screen_texture(const RenderDataRD *p_render_data) { + Ref<RenderSceneBuffersRD> rb = p_render_data->render_buffers; + ERR_FAIL_COND(rb.is_null()); + + RD::get_singleton()->draw_command_begin_label("Copy screen texture"); + + rb->allocate_blur_textures(); + + bool can_use_storage = _render_buffers_can_be_storage(); + Size2i size = rb->get_internal_size(); + + for (uint32_t v = 0; v < rb->get_view_count(); v++) { + RID texture = rb->get_internal_texture(v); + int mipmaps = int(rb->get_texture_format(RB_SCOPE_BUFFERS, RB_TEX_BLUR_0).mipmaps); + RID dest = rb->get_texture_slice(RB_SCOPE_BUFFERS, RB_TEX_BLUR_0, v, 0); + + if (can_use_storage) { + copy_effects->copy_to_rect(texture, dest, Rect2i(0, 0, size.x, size.y)); + } else { + RID fb = FramebufferCacheRD::get_singleton()->get_cache(dest); + copy_effects->copy_to_fb_rect(texture, fb, Rect2i(0, 0, size.x, size.y)); + } + + for (int i = 1; i < mipmaps; i++) { + RID source = dest; + dest = rb->get_texture_slice(RB_SCOPE_BUFFERS, RB_TEX_BLUR_0, v, i); + Size2i msize = rb->get_texture_slice_size(RB_SCOPE_BUFFERS, RB_TEX_BLUR_0, v, i); + + if (can_use_storage) { + copy_effects->make_mipmap(source, dest, msize); + } else { + copy_effects->make_mipmap_raster(source, dest, msize); + } + } + } + + RD::get_singleton()->draw_command_end_label(); +} + +void RendererSceneRenderRD::_render_buffers_copy_depth_texture(const RenderDataRD *p_render_data) { + Ref<RenderSceneBuffersRD> rb = p_render_data->render_buffers; + ERR_FAIL_COND(rb.is_null()); + + RD::get_singleton()->draw_command_begin_label("Copy depth texture"); + + // note, this only creates our back depth texture if we haven't already created it. + uint32_t usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT; + usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + usage_bits |= RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; // set this as color attachment because we're copying data into it, it's not actually used as a depth buffer + + rb->create_texture(RB_SCOPE_BUFFERS, RB_TEX_BACK_DEPTH, RD::DATA_FORMAT_R32_SFLOAT, usage_bits, RD::TEXTURE_SAMPLES_1); + + bool can_use_storage = _render_buffers_can_be_storage(); + Size2i size = rb->get_internal_size(); + for (uint32_t v = 0; v < p_render_data->scene_data->view_count; v++) { + RID depth_texture = rb->get_depth_texture(v); + RID depth_back_texture = rb->get_texture_slice(RB_SCOPE_BUFFERS, RB_TEX_BACK_DEPTH, v, 0); + + if (can_use_storage) { + copy_effects->copy_to_rect(depth_texture, depth_back_texture, Rect2i(0, 0, size.x, size.y)); + } else { + RID depth_back_fb = FramebufferCacheRD::get_singleton()->get_cache(depth_back_texture); + copy_effects->copy_to_fb_rect(depth_texture, depth_back_fb, Rect2i(0, 0, size.x, size.y)); + } + } + + RD::get_singleton()->draw_command_end_label(); +} + +void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(const RenderDataRD *p_render_data) { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + + Ref<RenderSceneBuffersRD> rb = p_render_data->render_buffers; + ERR_FAIL_COND(rb.is_null()); + + // Glow, auto exposure and DoF (if enabled). + + Size2i internal_size = rb->get_internal_size(); + Size2i target_size = rb->get_target_size(); + + bool can_use_effects = target_size.x >= 8 && target_size.y >= 8; // FIXME I think this should check internal size, we do all our post processing at this size... + bool can_use_storage = _render_buffers_can_be_storage(); + + RID render_target = rb->get_render_target(); + RID internal_texture = rb->get_internal_texture(); + + if (can_use_effects && RSG::camera_attributes->camera_attributes_uses_dof(p_render_data->camera_attributes)) { + RENDER_TIMESTAMP("Depth of Field"); + RD::get_singleton()->draw_command_begin_label("DOF"); + + rb->allocate_blur_textures(); + + RendererRD::BokehDOF::BokehBuffers buffers; + + // Textures we use + buffers.base_texture_size = rb->get_internal_size(); + buffers.secondary_texture = rb->get_texture_slice(RB_SCOPE_BUFFERS, RB_TEX_BLUR_0, 0, 0); + buffers.half_texture[0] = rb->get_texture_slice(RB_SCOPE_BUFFERS, RB_TEX_BLUR_1, 0, 0); + buffers.half_texture[1] = rb->get_texture_slice(RB_SCOPE_BUFFERS, RB_TEX_BLUR_0, 0, 1); + + if (can_use_storage) { + for (uint32_t i = 0; i < rb->get_view_count(); i++) { + buffers.base_texture = rb->get_internal_texture(i); + buffers.depth_texture = rb->get_depth_texture(i); + + // In stereo p_render_data->z_near and p_render_data->z_far can be offset for our combined frustrum + float z_near = p_render_data->scene_data->view_projection[i].get_z_near(); + float z_far = p_render_data->scene_data->view_projection[i].get_z_far(); + bokeh_dof->bokeh_dof_compute(buffers, p_render_data->camera_attributes, z_near, z_far, p_render_data->scene_data->cam_orthogonal); + }; + } else { + // Set framebuffers. + buffers.secondary_fb = rb->weight_buffers[1].fb; + buffers.half_fb[0] = rb->weight_buffers[2].fb; + buffers.half_fb[1] = rb->weight_buffers[3].fb; + buffers.weight_texture[0] = rb->weight_buffers[0].weight; + buffers.weight_texture[1] = rb->weight_buffers[1].weight; + buffers.weight_texture[2] = rb->weight_buffers[2].weight; + buffers.weight_texture[3] = rb->weight_buffers[3].weight; + + // Set weight buffers. + buffers.base_weight_fb = rb->weight_buffers[0].fb; + + for (uint32_t i = 0; i < rb->get_view_count(); i++) { + buffers.base_texture = rb->get_internal_texture(i); + buffers.depth_texture = rb->get_depth_texture(i); + buffers.base_fb = FramebufferCacheRD::get_singleton()->get_cache(buffers.base_texture); // TODO move this into bokeh_dof_raster, we can do this internally + + // In stereo p_render_data->z_near and p_render_data->z_far can be offset for our combined frustrum + float z_near = p_render_data->scene_data->view_projection[i].get_z_near(); + float z_far = p_render_data->scene_data->view_projection[i].get_z_far(); + bokeh_dof->bokeh_dof_raster(buffers, p_render_data->camera_attributes, z_near, z_far, p_render_data->scene_data->cam_orthogonal); + } + } + RD::get_singleton()->draw_command_end_label(); + } + + float auto_exposure_scale = 1.0; + + if (can_use_effects && RSG::camera_attributes->camera_attributes_uses_auto_exposure(p_render_data->camera_attributes)) { + RENDER_TIMESTAMP("Auto exposure"); + + RD::get_singleton()->draw_command_begin_label("Auto exposure"); + if (rb->luminance.current.is_null()) { + _allocate_luminance_textures(rb); + } + uint64_t auto_exposure_version = RSG::camera_attributes->camera_attributes_get_auto_exposure_version(p_render_data->camera_attributes); + bool set_immediate = auto_exposure_version != rb->get_auto_exposure_version(); + rb->set_auto_exposure_version(auto_exposure_version); + + double step = RSG::camera_attributes->camera_attributes_get_auto_exposure_adjust_speed(p_render_data->camera_attributes) * time_step; + float auto_exposure_min_sensitivity = RSG::camera_attributes->camera_attributes_get_auto_exposure_min_sensitivity(p_render_data->camera_attributes); + float auto_exposure_max_sensitivity = RSG::camera_attributes->camera_attributes_get_auto_exposure_max_sensitivity(p_render_data->camera_attributes); + if (can_use_storage) { + RendererCompositorRD::singleton->get_effects()->luminance_reduction(internal_texture, internal_size, rb->luminance.reduce, rb->luminance.current, auto_exposure_min_sensitivity, auto_exposure_max_sensitivity, step, set_immediate); + } else { + RendererCompositorRD::singleton->get_effects()->luminance_reduction_raster(internal_texture, internal_size, rb->luminance.reduce, rb->luminance.fb, rb->luminance.current, auto_exposure_min_sensitivity, auto_exposure_max_sensitivity, step, set_immediate); + } + // Swap final reduce with prev luminance. + SWAP(rb->luminance.current, rb->luminance.reduce.write[rb->luminance.reduce.size() - 1]); + if (!can_use_storage) { + SWAP(rb->luminance.current_fb, rb->luminance.fb.write[rb->luminance.fb.size() - 1]); + } + + auto_exposure_scale = RSG::camera_attributes->camera_attributes_get_auto_exposure_scale(p_render_data->camera_attributes); + + RenderingServerDefault::redraw_request(); // Redraw all the time if auto exposure rendering is on. + RD::get_singleton()->draw_command_end_label(); + } + + int max_glow_level = -1; + + if (can_use_effects && p_render_data->environment.is_valid() && environment_get_glow_enabled(p_render_data->environment)) { + RENDER_TIMESTAMP("Glow"); + RD::get_singleton()->draw_command_begin_label("Gaussian Glow"); + + rb->allocate_blur_textures(); + + for (int i = 0; i < RS::MAX_GLOW_LEVELS; i++) { + if (environment_get_glow_levels(p_render_data->environment)[i] > 0.0) { + int mipmaps = int(rb->get_texture_format(RB_SCOPE_BUFFERS, RB_TEX_BLUR_1).mipmaps); + if (i >= mipmaps) { + max_glow_level = mipmaps - 1; + } else { + max_glow_level = i; + } + } + } + + float luminance_multiplier = _render_buffers_get_luminance_multiplier(); + for (uint32_t l = 0; l < rb->get_view_count(); l++) { + for (int i = 0; i < (max_glow_level + 1); i++) { + Size2i vp_size = rb->get_texture_slice_size(RB_SCOPE_BUFFERS, RB_TEX_BLUR_1, l, i); + + if (i == 0) { + RID luminance_texture; + if (RSG::camera_attributes->camera_attributes_uses_auto_exposure(p_render_data->camera_attributes) && rb->luminance.current.is_valid()) { + luminance_texture = rb->luminance.current; + } + RID source = rb->get_internal_texture(l); + RID dest = rb->get_texture_slice(RB_SCOPE_BUFFERS, RB_TEX_BLUR_1, l, i); + if (can_use_storage) { + copy_effects->gaussian_glow(source, dest, vp_size, environment_get_glow_strength(p_render_data->environment), glow_high_quality, true, environment_get_glow_hdr_luminance_cap(p_render_data->environment), environment_get_exposure(p_render_data->environment), environment_get_glow_bloom(p_render_data->environment), environment_get_glow_hdr_bleed_threshold(p_render_data->environment), environment_get_glow_hdr_bleed_scale(p_render_data->environment), luminance_texture, auto_exposure_scale); + } else { + RID half = rb->get_texture_slice(RB_SCOPE_BUFFERS, RB_TEX_HALF_BLUR, 0, i); // we can reuse this for each view + copy_effects->gaussian_glow_raster(source, half, dest, luminance_multiplier, vp_size, environment_get_glow_strength(p_render_data->environment), glow_high_quality, true, environment_get_glow_hdr_luminance_cap(p_render_data->environment), environment_get_exposure(p_render_data->environment), environment_get_glow_bloom(p_render_data->environment), environment_get_glow_hdr_bleed_threshold(p_render_data->environment), environment_get_glow_hdr_bleed_scale(p_render_data->environment), luminance_texture, auto_exposure_scale); + } + } else { + RID source = rb->get_texture_slice(RB_SCOPE_BUFFERS, RB_TEX_BLUR_1, l, i - 1); + RID dest = rb->get_texture_slice(RB_SCOPE_BUFFERS, RB_TEX_BLUR_1, l, i); + + if (can_use_storage) { + copy_effects->gaussian_glow(source, dest, vp_size, environment_get_glow_strength(p_render_data->environment), glow_high_quality); + } else { + RID half = rb->get_texture_slice(RB_SCOPE_BUFFERS, RB_TEX_HALF_BLUR, 0, i); // we can reuse this for each view + copy_effects->gaussian_glow_raster(source, half, dest, luminance_multiplier, vp_size, environment_get_glow_strength(p_render_data->environment), glow_high_quality); + } + } + } + } + + RD::get_singleton()->draw_command_end_label(); + } + + { + RENDER_TIMESTAMP("Tonemap"); + RD::get_singleton()->draw_command_begin_label("Tonemap"); + + RendererRD::ToneMapper::TonemapSettings tonemap; + + if (can_use_effects && RSG::camera_attributes->camera_attributes_uses_auto_exposure(p_render_data->camera_attributes) && rb->luminance.current.is_valid()) { + tonemap.use_auto_exposure = true; + tonemap.exposure_texture = rb->luminance.current; + tonemap.auto_exposure_scale = auto_exposure_scale; + } else { + tonemap.exposure_texture = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_WHITE); + } + + if (can_use_effects && p_render_data->environment.is_valid() && environment_get_glow_enabled(p_render_data->environment)) { + tonemap.use_glow = true; + tonemap.glow_mode = RendererRD::ToneMapper::TonemapSettings::GlowMode(environment_get_glow_blend_mode(p_render_data->environment)); + tonemap.glow_intensity = environment_get_glow_blend_mode(p_render_data->environment) == RS::ENV_GLOW_BLEND_MODE_MIX ? environment_get_glow_mix(p_render_data->environment) : environment_get_glow_intensity(p_render_data->environment); + for (int i = 0; i < RS::MAX_GLOW_LEVELS; i++) { + tonemap.glow_levels[i] = environment_get_glow_levels(p_render_data->environment)[i]; + } + + Size2i msize = rb->get_texture_slice_size(RB_SCOPE_BUFFERS, RB_TEX_BLUR_1, 0, 0); + tonemap.glow_texture_size.x = msize.width; + tonemap.glow_texture_size.y = msize.height; + tonemap.glow_use_bicubic_upscale = glow_bicubic_upscale; + tonemap.glow_texture = rb->get_texture(RB_SCOPE_BUFFERS, RB_TEX_BLUR_1); + if (environment_get_glow_map(p_render_data->environment).is_valid()) { + tonemap.glow_map_strength = environment_get_glow_map_strength(p_render_data->environment); + tonemap.glow_map = texture_storage->texture_get_rd_texture(environment_get_glow_map(p_render_data->environment)); + } else { + tonemap.glow_map_strength = 0.0f; + tonemap.glow_map = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_WHITE); + } + + } else { + tonemap.glow_texture = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_BLACK); + tonemap.glow_map = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_WHITE); + } + + if (rb->get_screen_space_aa() == RS::VIEWPORT_SCREEN_SPACE_AA_FXAA) { + tonemap.use_fxaa = true; + } + + tonemap.use_debanding = rb->get_use_debanding(); + tonemap.texture_size = Vector2i(rb->get_internal_size().x, rb->get_internal_size().y); + + if (p_render_data->environment.is_valid()) { + tonemap.tonemap_mode = environment_get_tone_mapper(p_render_data->environment); + tonemap.white = environment_get_white(p_render_data->environment); + tonemap.exposure = environment_get_exposure(p_render_data->environment); + } + + tonemap.use_color_correction = false; + tonemap.use_1d_color_correction = false; + tonemap.color_correction_texture = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE); + + if (can_use_effects && p_render_data->environment.is_valid()) { + tonemap.use_bcs = environment_get_adjustments_enabled(p_render_data->environment); + tonemap.brightness = environment_get_adjustments_brightness(p_render_data->environment); + tonemap.contrast = environment_get_adjustments_contrast(p_render_data->environment); + tonemap.saturation = environment_get_adjustments_saturation(p_render_data->environment); + if (environment_get_adjustments_enabled(p_render_data->environment) && environment_get_color_correction(p_render_data->environment).is_valid()) { + tonemap.use_color_correction = true; + tonemap.use_1d_color_correction = environment_get_use_1d_color_correction(p_render_data->environment); + tonemap.color_correction_texture = texture_storage->texture_get_rd_texture(environment_get_color_correction(p_render_data->environment)); + } + } + + tonemap.luminance_multiplier = _render_buffers_get_luminance_multiplier(); + tonemap.view_count = rb->get_view_count(); + + RID dest_fb; + if (fsr && can_use_effects && (internal_size.x != target_size.x || internal_size.y != target_size.y)) { + // If we use FSR to upscale we need to write our result into an intermediate buffer. + // Note that this is cached so we only create the texture the first time. + RID dest_texture = rb->create_texture(SNAME("Tonemapper"), SNAME("destination"), _render_buffers_get_color_format(), RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT); + dest_fb = FramebufferCacheRD::get_singleton()->get_cache(dest_texture); + } else { + // If we do a bilinear upscale we just render into our render target and our shader will upscale automatically. + // Target size in this case is lying as we never get our real target size communicated. + // Bit nasty but... + dest_fb = texture_storage->render_target_get_rd_framebuffer(render_target); + } + + tone_mapper->tonemapper(internal_texture, dest_fb, tonemap); + + RD::get_singleton()->draw_command_end_label(); + } + + if (fsr && can_use_effects && (internal_size.x != target_size.x || internal_size.y != target_size.y)) { + // TODO Investigate? Does this work? We never write into our render target and we've already done so up above in our tonemapper. + // I think FSR should either work before our tonemapper or as an alternative of our tonemapper. + + RD::get_singleton()->draw_command_begin_label("FSR 1.0 Upscale"); + + for (uint32_t v = 0; v < rb->get_view_count(); v++) { + RID source_texture = rb->get_texture_slice(SNAME("Tonemapper"), SNAME("destination"), v, 0); + RID dest_texture = texture_storage->render_target_get_rd_texture_slice(render_target, v); + + fsr->fsr_upscale(rb, source_texture, dest_texture); + } + + RD::get_singleton()->draw_command_end_label(); + } + + texture_storage->render_target_disable_clear_request(render_target); +} + +void RendererSceneRenderRD::_post_process_subpass(RID p_source_texture, RID p_framebuffer, const RenderDataRD *p_render_data) { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + RD::get_singleton()->draw_command_begin_label("Post Process Subpass"); + + Ref<RenderSceneBuffersRD> rb = p_render_data->render_buffers; + ERR_FAIL_COND(rb.is_null()); + + // FIXME: Our input it our internal_texture, shouldn't this be using internal_size ?? + // Seeing we don't support FSR in our mobile renderer right now target_size = internal_size... + Size2i target_size = rb->get_target_size(); + bool can_use_effects = target_size.x >= 8 && target_size.y >= 8; + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_switch_to_next_pass(); + + RendererRD::ToneMapper::TonemapSettings tonemap; + + if (p_render_data->environment.is_valid()) { + tonemap.tonemap_mode = environment_get_tone_mapper(p_render_data->environment); + tonemap.exposure = environment_get_exposure(p_render_data->environment); + tonemap.white = environment_get_white(p_render_data->environment); + } + + // We don't support glow or auto exposure here, if they are needed, don't use subpasses! + // The problem is that we need to use the result so far and process them before we can + // apply this to our results. + if (can_use_effects && p_render_data->environment.is_valid() && environment_get_glow_enabled(p_render_data->environment)) { + ERR_FAIL_MSG("Glow is not supported when using subpasses."); + } + + if (can_use_effects && RSG::camera_attributes->camera_attributes_uses_auto_exposure(p_render_data->camera_attributes)) { + ERR_FAIL_MSG("Auto Exposure is not supported when using subpasses."); + } + + tonemap.use_glow = false; + tonemap.glow_texture = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_BLACK); + tonemap.glow_map = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_WHITE); + tonemap.use_auto_exposure = false; + tonemap.exposure_texture = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_WHITE); + + tonemap.use_color_correction = false; + tonemap.use_1d_color_correction = false; + tonemap.color_correction_texture = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE); + + if (can_use_effects && p_render_data->environment.is_valid()) { + tonemap.use_bcs = environment_get_adjustments_enabled(p_render_data->environment); + tonemap.brightness = environment_get_adjustments_brightness(p_render_data->environment); + tonemap.contrast = environment_get_adjustments_contrast(p_render_data->environment); + tonemap.saturation = environment_get_adjustments_saturation(p_render_data->environment); + if (environment_get_adjustments_enabled(p_render_data->environment) && environment_get_color_correction(p_render_data->environment).is_valid()) { + tonemap.use_color_correction = true; + tonemap.use_1d_color_correction = environment_get_use_1d_color_correction(p_render_data->environment); + tonemap.color_correction_texture = texture_storage->texture_get_rd_texture(environment_get_color_correction(p_render_data->environment)); + } + } + + tonemap.use_debanding = rb->get_use_debanding(); + tonemap.texture_size = Vector2i(target_size.x, target_size.y); + + tonemap.luminance_multiplier = _render_buffers_get_luminance_multiplier(); + tonemap.view_count = rb->get_view_count(); + + tone_mapper->tonemapper(draw_list, p_source_texture, RD::get_singleton()->framebuffer_get_format(p_framebuffer), tonemap); + + RD::get_singleton()->draw_command_end_label(); +} + +void RendererSceneRenderRD::_disable_clear_request(const RenderDataRD *p_render_data) { + ERR_FAIL_COND(p_render_data->render_buffers.is_null()); + + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + texture_storage->render_target_disable_clear_request(p_render_data->render_buffers->get_render_target()); +} + +void RendererSceneRenderRD::_render_buffers_debug_draw(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_shadow_atlas, RID p_occlusion_buffer) { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + + ERR_FAIL_COND(p_render_buffers.is_null()); + + RID render_target = p_render_buffers->get_render_target(); + + if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_SHADOW_ATLAS) { + if (p_shadow_atlas.is_valid()) { + RID shadow_atlas_texture = RendererRD::LightStorage::get_singleton()->shadow_atlas_get_texture(p_shadow_atlas); + + if (shadow_atlas_texture.is_null()) { + shadow_atlas_texture = texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_BLACK); + } + + Size2 rtsize = texture_storage->render_target_get_size(render_target); + copy_effects->copy_to_fb_rect(shadow_atlas_texture, texture_storage->render_target_get_rd_framebuffer(render_target), Rect2i(Vector2(), rtsize / 2), false, true); + } + } + + if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_DIRECTIONAL_SHADOW_ATLAS) { + if (RendererRD::LightStorage::get_singleton()->directional_shadow_get_texture().is_valid()) { + RID shadow_atlas_texture = RendererRD::LightStorage::get_singleton()->directional_shadow_get_texture(); + Size2i rtsize = texture_storage->render_target_get_size(render_target); + + copy_effects->copy_to_fb_rect(shadow_atlas_texture, texture_storage->render_target_get_rd_framebuffer(render_target), Rect2i(Vector2(), rtsize / 2), false, true); + } + } + + if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_DECAL_ATLAS) { + RID decal_atlas = RendererRD::TextureStorage::get_singleton()->decal_atlas_get_texture(); + + if (decal_atlas.is_valid()) { + Size2i rtsize = texture_storage->render_target_get_size(render_target); + + copy_effects->copy_to_fb_rect(decal_atlas, texture_storage->render_target_get_rd_framebuffer(render_target), Rect2i(Vector2(), rtsize / 2), false, false, true); + } + } + + if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_SCENE_LUMINANCE) { + if (p_render_buffers->luminance.current.is_valid()) { + Size2i rtsize = texture_storage->render_target_get_size(render_target); + + copy_effects->copy_to_fb_rect(p_render_buffers->luminance.current, texture_storage->render_target_get_rd_framebuffer(render_target), Rect2(Vector2(), rtsize / 8), false, true); + } + } + + if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_NORMAL_BUFFER && _render_buffers_get_normal_texture(p_render_buffers).is_valid()) { + Size2 rtsize = texture_storage->render_target_get_size(render_target); + copy_effects->copy_to_fb_rect(_render_buffers_get_normal_texture(p_render_buffers), texture_storage->render_target_get_rd_framebuffer(render_target), Rect2(Vector2(), rtsize), false, false); + } + + if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_OCCLUDERS) { + if (p_occlusion_buffer.is_valid()) { + Size2i rtsize = texture_storage->render_target_get_size(render_target); + copy_effects->copy_to_fb_rect(texture_storage->texture_get_rd_texture(p_occlusion_buffer), texture_storage->render_target_get_rd_framebuffer(render_target), Rect2i(Vector2(), rtsize), true, false); + } + } + + if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_MOTION_VECTORS && _render_buffers_get_velocity_texture(p_render_buffers).is_valid()) { + Size2i rtsize = texture_storage->render_target_get_size(render_target); + copy_effects->copy_to_fb_rect(_render_buffers_get_velocity_texture(p_render_buffers), texture_storage->render_target_get_rd_framebuffer(render_target), Rect2(Vector2(), rtsize), false, false); + } +} + +RID RendererSceneRenderRD::render_buffers_get_default_voxel_gi_buffer() { + return gi.default_voxel_gi_buffer; +} + +float RendererSceneRenderRD::_render_buffers_get_luminance_multiplier() { + return 1.0; +} + +RD::DataFormat RendererSceneRenderRD::_render_buffers_get_color_format() { + return RD::DATA_FORMAT_R16G16B16A16_SFLOAT; +} + +bool RendererSceneRenderRD::_render_buffers_can_be_storage() { + return true; +} + +void RendererSceneRenderRD::gi_set_use_half_resolution(bool p_enable) { + gi.half_resolution = p_enable; +} + +void RendererSceneRenderRD::positional_soft_shadow_filter_set_quality(RS::ShadowQuality p_quality) { + ERR_FAIL_INDEX_MSG(p_quality, RS::SHADOW_QUALITY_MAX, "Shadow quality too high, please see RenderingServer's ShadowQuality enum"); + + if (shadows_quality != p_quality) { + shadows_quality = p_quality; + + switch (shadows_quality) { + case RS::SHADOW_QUALITY_HARD: { + penumbra_shadow_samples = 4; + soft_shadow_samples = 0; + shadows_quality_radius = 1.0; + } break; + case RS::SHADOW_QUALITY_SOFT_VERY_LOW: { + penumbra_shadow_samples = 4; + soft_shadow_samples = 1; + shadows_quality_radius = 1.5; + } break; + case RS::SHADOW_QUALITY_SOFT_LOW: { + penumbra_shadow_samples = 8; + soft_shadow_samples = 4; + shadows_quality_radius = 2.0; + } break; + case RS::SHADOW_QUALITY_SOFT_MEDIUM: { + penumbra_shadow_samples = 12; + soft_shadow_samples = 8; + shadows_quality_radius = 2.0; + } break; + case RS::SHADOW_QUALITY_SOFT_HIGH: { + penumbra_shadow_samples = 24; + soft_shadow_samples = 16; + shadows_quality_radius = 3.0; + } break; + case RS::SHADOW_QUALITY_SOFT_ULTRA: { + penumbra_shadow_samples = 32; + soft_shadow_samples = 32; + shadows_quality_radius = 4.0; + } break; + case RS::SHADOW_QUALITY_MAX: + break; + } + get_vogel_disk(penumbra_shadow_kernel, penumbra_shadow_samples); + get_vogel_disk(soft_shadow_kernel, soft_shadow_samples); + } + + _update_shader_quality_settings(); +} + +void RendererSceneRenderRD::directional_soft_shadow_filter_set_quality(RS::ShadowQuality p_quality) { + ERR_FAIL_INDEX_MSG(p_quality, RS::SHADOW_QUALITY_MAX, "Shadow quality too high, please see RenderingServer's ShadowQuality enum"); + + if (directional_shadow_quality != p_quality) { + directional_shadow_quality = p_quality; + + switch (directional_shadow_quality) { + case RS::SHADOW_QUALITY_HARD: { + directional_penumbra_shadow_samples = 4; + directional_soft_shadow_samples = 0; + directional_shadow_quality_radius = 1.0; + } break; + case RS::SHADOW_QUALITY_SOFT_VERY_LOW: { + directional_penumbra_shadow_samples = 4; + directional_soft_shadow_samples = 1; + directional_shadow_quality_radius = 1.5; + } break; + case RS::SHADOW_QUALITY_SOFT_LOW: { + directional_penumbra_shadow_samples = 8; + directional_soft_shadow_samples = 4; + directional_shadow_quality_radius = 2.0; + } break; + case RS::SHADOW_QUALITY_SOFT_MEDIUM: { + directional_penumbra_shadow_samples = 12; + directional_soft_shadow_samples = 8; + directional_shadow_quality_radius = 2.0; + } break; + case RS::SHADOW_QUALITY_SOFT_HIGH: { + directional_penumbra_shadow_samples = 24; + directional_soft_shadow_samples = 16; + directional_shadow_quality_radius = 3.0; + } break; + case RS::SHADOW_QUALITY_SOFT_ULTRA: { + directional_penumbra_shadow_samples = 32; + directional_soft_shadow_samples = 32; + directional_shadow_quality_radius = 4.0; + } break; + case RS::SHADOW_QUALITY_MAX: + break; + } + get_vogel_disk(directional_penumbra_shadow_kernel, directional_penumbra_shadow_samples); + get_vogel_disk(directional_soft_shadow_kernel, directional_soft_shadow_samples); + } + + _update_shader_quality_settings(); +} + +void RendererSceneRenderRD::decals_set_filter(RenderingServer::DecalFilter p_filter) { + if (decals_filter == p_filter) { + return; + } + decals_filter = p_filter; + _update_shader_quality_settings(); +} +void RendererSceneRenderRD::light_projectors_set_filter(RenderingServer::LightProjectorFilter p_filter) { + if (light_projectors_filter == p_filter) { + return; + } + light_projectors_filter = p_filter; + _update_shader_quality_settings(); +} + +int RendererSceneRenderRD::get_roughness_layers() const { + return sky.roughness_layers; +} + +bool RendererSceneRenderRD::is_using_radiance_cubemap_array() const { + return sky.sky_use_cubemap_array; +} + +void RendererSceneRenderRD::_update_vrs(Ref<RenderSceneBuffersRD> p_render_buffers) { + if (p_render_buffers.is_valid() && vrs) { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + + RS::ViewportVRSMode vrs_mode = texture_storage->render_target_get_vrs_mode(p_render_buffers->get_render_target()); + if (vrs_mode != RS::VIEWPORT_VRS_DISABLED) { + RID vrs_texture = p_render_buffers->get_texture(RB_SCOPE_VRS, RB_TEXTURE); + + // We use get_cache_multipass instead of get_cache_multiview because the default behavior is for + // our vrs_texture to be used as the VRS attachment. In this particular case we're writing to it + // so it needs to be set as our color attachment + + Vector<RID> textures; + textures.push_back(vrs_texture); + + Vector<RD::FramebufferPass> passes; + RD::FramebufferPass pass; + pass.color_attachments.push_back(0); + passes.push_back(pass); + + RID vrs_fb = FramebufferCacheRD::get_singleton()->get_cache_multipass(textures, passes, p_render_buffers->get_view_count()); + + vrs->update_vrs_texture(vrs_fb, p_render_buffers->get_render_target()); + } + } +} + +bool RendererSceneRenderRD::_needs_post_prepass_render(RenderDataRD *p_render_data, bool p_use_gi) { + if (p_render_data->render_buffers.is_valid()) { + if (p_render_data->render_buffers->has_custom_data(RB_SCOPE_SDFGI)) { + return true; + } + } + return false; +} + +void RendererSceneRenderRD::_post_prepass_render(RenderDataRD *p_render_data, bool p_use_gi) { + if (p_render_data->render_buffers.is_valid() && p_use_gi) { + if (!p_render_data->render_buffers->has_custom_data(RB_SCOPE_SDFGI)) { + return; + } + + Ref<RendererRD::GI::SDFGI> sdfgi = p_render_data->render_buffers->get_custom_data(RB_SCOPE_SDFGI); + sdfgi->update_probes(p_render_data->environment, sky.sky_owner.get_or_null(environment_get_sky(p_render_data->environment))); + } +} + +void RendererSceneRenderRD::_pre_resolve_render(RenderDataRD *p_render_data, bool p_use_gi) { + if (p_render_data->render_buffers.is_valid()) { + if (p_use_gi) { + RD::get_singleton()->compute_list_end(); + } + } +} + +void RendererSceneRenderRD::render_scene(const Ref<RenderSceneBuffers> &p_render_buffers, const CameraData *p_camera_data, const CameraData *p_prev_camera_data, const PagedArray<RenderGeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_voxel_gi_instances, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, const PagedArray<RID> &p_fog_volumes, RID p_environment, RID p_camera_attributes, RID p_shadow_atlas, RID p_occluder_debug_tex, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_mesh_lod_threshold, const RenderShadowData *p_render_shadows, int p_render_shadow_count, const RenderSDFGIData *p_render_sdfgi_regions, int p_render_sdfgi_region_count, const RenderSDFGIUpdateData *p_sdfgi_update_data, RenderingMethod::RenderInfo *r_render_info) { + RendererRD::LightStorage *light_storage = RendererRD::LightStorage::get_singleton(); + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + + // getting this here now so we can direct call a bunch of things more easily + Ref<RenderSceneBuffersRD> rb; + if (p_render_buffers.is_valid()) { + rb = p_render_buffers; // cast it... + ERR_FAIL_COND(rb.is_null()); + } + + // setup scene data + RenderSceneDataRD scene_data; + { + // Our first camera is used by default + scene_data.cam_transform = p_camera_data->main_transform; + scene_data.cam_projection = p_camera_data->main_projection; + scene_data.cam_orthogonal = p_camera_data->is_orthogonal; + scene_data.taa_jitter = p_camera_data->taa_jitter; + + scene_data.view_count = p_camera_data->view_count; + for (uint32_t v = 0; v < p_camera_data->view_count; v++) { + scene_data.view_eye_offset[v] = p_camera_data->view_offset[v].origin; + scene_data.view_projection[v] = p_camera_data->view_projection[v]; + } + + scene_data.prev_cam_transform = p_prev_camera_data->main_transform; + scene_data.prev_cam_projection = p_prev_camera_data->main_projection; + scene_data.prev_taa_jitter = p_prev_camera_data->taa_jitter; + + for (uint32_t v = 0; v < p_camera_data->view_count; v++) { + scene_data.prev_view_projection[v] = p_prev_camera_data->view_projection[v]; + } + + scene_data.z_near = p_camera_data->main_projection.get_z_near(); + scene_data.z_far = p_camera_data->main_projection.get_z_far(); + + // this should be the same for all cameras.. + scene_data.lod_distance_multiplier = p_camera_data->main_projection.get_lod_multiplier(); + + if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_DISABLE_LOD) { + scene_data.screen_mesh_lod_threshold = 0.0; + } else { + scene_data.screen_mesh_lod_threshold = p_screen_mesh_lod_threshold; + } + + if (p_shadow_atlas.is_valid()) { + int shadow_atlas_size = light_storage->shadow_atlas_get_size(p_shadow_atlas); + scene_data.shadow_atlas_pixel_size.x = 1.0 / shadow_atlas_size; + scene_data.shadow_atlas_pixel_size.y = 1.0 / shadow_atlas_size; + } + { + int directional_shadow_size = light_storage->directional_shadow_get_size(); + scene_data.directional_shadow_pixel_size.x = 1.0 / directional_shadow_size; + scene_data.directional_shadow_pixel_size.y = 1.0 / directional_shadow_size; + } + + scene_data.time = time; + scene_data.time_step = time_step; + } + + //assign render data + RenderDataRD render_data; + { + render_data.render_buffers = rb; + render_data.scene_data = &scene_data; + + render_data.instances = &p_instances; + render_data.lights = &p_lights; + render_data.reflection_probes = &p_reflection_probes; + render_data.voxel_gi_instances = &p_voxel_gi_instances; + render_data.decals = &p_decals; + render_data.lightmaps = &p_lightmaps; + render_data.fog_volumes = &p_fog_volumes; + render_data.environment = p_environment; + render_data.camera_attributes = p_camera_attributes; + render_data.shadow_atlas = p_shadow_atlas; + render_data.occluder_debug_tex = p_occluder_debug_tex; + render_data.reflection_atlas = p_reflection_atlas; + render_data.reflection_probe = p_reflection_probe; + render_data.reflection_probe_pass = p_reflection_probe_pass; + + render_data.render_shadows = p_render_shadows; + render_data.render_shadow_count = p_render_shadow_count; + render_data.render_sdfgi_regions = p_render_sdfgi_regions; + render_data.render_sdfgi_region_count = p_render_sdfgi_region_count; + render_data.sdfgi_update_data = p_sdfgi_update_data; + + render_data.render_info = r_render_info; + } + + PagedArray<RID> empty; + + if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_UNSHADED) { + render_data.lights = ∅ + render_data.reflection_probes = ∅ + render_data.voxel_gi_instances = ∅ + } + + Color clear_color; + if (p_render_buffers.is_valid()) { + clear_color = texture_storage->render_target_get_clear_request_color(rb->get_render_target()); + } else { + clear_color = RSG::texture_storage->get_default_clear_color(); + } + + //calls _pre_opaque_render between depth pre-pass and opaque pass + _render_scene(&render_data, clear_color); +} + +void RendererSceneRenderRD::render_material(const Transform3D &p_cam_transform, const Projection &p_cam_projection, bool p_cam_orthogonal, const PagedArray<RenderGeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) { + _render_material(p_cam_transform, p_cam_projection, p_cam_orthogonal, p_instances, p_framebuffer, p_region, 1.0); +} + +void RendererSceneRenderRD::render_particle_collider_heightfield(RID p_collider, const Transform3D &p_transform, const PagedArray<RenderGeometryInstance *> &p_instances) { + RendererRD::ParticlesStorage *particles_storage = RendererRD::ParticlesStorage::get_singleton(); + + ERR_FAIL_COND(!particles_storage->particles_collision_is_heightfield(p_collider)); + Vector3 extents = particles_storage->particles_collision_get_extents(p_collider) * p_transform.basis.get_scale(); + Projection cm; + cm.set_orthogonal(-extents.x, extents.x, -extents.z, extents.z, 0, extents.y * 2.0); + + Vector3 cam_pos = p_transform.origin; + cam_pos.y += extents.y; + + Transform3D cam_xform; + cam_xform.set_look_at(cam_pos, cam_pos - p_transform.basis.get_column(Vector3::AXIS_Y), -p_transform.basis.get_column(Vector3::AXIS_Z).normalized()); + + RID fb = particles_storage->particles_collision_get_heightfield_framebuffer(p_collider); + + _render_particle_collider_heightfield(fb, cam_xform, cm, p_instances); +} + +bool RendererSceneRenderRD::free(RID p_rid) { + if (is_environment(p_rid)) { + environment_free(p_rid); + } else if (RSG::camera_attributes->owns_camera_attributes(p_rid)) { + RSG::camera_attributes->camera_attributes_free(p_rid); + } else if (gi.voxel_gi_instance_owns(p_rid)) { + gi.voxel_gi_instance_free(p_rid); + } else if (sky.sky_owner.owns(p_rid)) { + sky.update_dirty_skys(); + sky.free_sky(p_rid); + } else if (RendererRD::Fog::get_singleton()->owns_fog_volume_instance(p_rid)) { + RendererRD::Fog::get_singleton()->fog_instance_free(p_rid); + } else { + return false; + } + + return true; +} + +void RendererSceneRenderRD::set_debug_draw_mode(RS::ViewportDebugDraw p_debug_draw) { + debug_draw = p_debug_draw; +} + +void RendererSceneRenderRD::update() { + sky.update_dirty_skys(); +} + +void RendererSceneRenderRD::set_time(double p_time, double p_step) { + time = p_time; + time_step = p_step; +} + +void RendererSceneRenderRD::screen_space_roughness_limiter_set_active(bool p_enable, float p_amount, float p_limit) { + screen_space_roughness_limiter = p_enable; + screen_space_roughness_limiter_amount = p_amount; + screen_space_roughness_limiter_limit = p_limit; +} + +bool RendererSceneRenderRD::screen_space_roughness_limiter_is_active() const { + return screen_space_roughness_limiter; +} + +float RendererSceneRenderRD::screen_space_roughness_limiter_get_amount() const { + return screen_space_roughness_limiter_amount; +} + +float RendererSceneRenderRD::screen_space_roughness_limiter_get_limit() const { + return screen_space_roughness_limiter_limit; +} + +TypedArray<Image> RendererSceneRenderRD::bake_render_uv2(RID p_base, const TypedArray<RID> &p_material_overrides, const Size2i &p_image_size) { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + tf.width = p_image_size.width; // Always 64x64 + tf.height = p_image_size.height; + tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT; + + RID albedo_alpha_tex = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RID normal_tex = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RID orm_tex = RD::get_singleton()->texture_create(tf, RD::TextureView()); + + tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + RID emission_tex = RD::get_singleton()->texture_create(tf, RD::TextureView()); + + tf.format = RD::DATA_FORMAT_R32_SFLOAT; + RID depth_write_tex = RD::get_singleton()->texture_create(tf, RD::TextureView()); + + tf.usage_bits = RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT; + tf.format = RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_D32_SFLOAT, RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) ? RD::DATA_FORMAT_D32_SFLOAT : RD::DATA_FORMAT_X8_D24_UNORM_PACK32; + RID depth_tex = RD::get_singleton()->texture_create(tf, RD::TextureView()); + + Vector<RID> fb_tex; + fb_tex.push_back(albedo_alpha_tex); + fb_tex.push_back(normal_tex); + fb_tex.push_back(orm_tex); + fb_tex.push_back(emission_tex); + fb_tex.push_back(depth_write_tex); + fb_tex.push_back(depth_tex); + + RID fb = RD::get_singleton()->framebuffer_create(fb_tex); + + //RID sampled_light; + + RenderGeometryInstance *gi_inst = geometry_instance_create(p_base); + ERR_FAIL_NULL_V(gi_inst, TypedArray<Image>()); + + uint32_t sc = RSG::mesh_storage->mesh_get_surface_count(p_base); + Vector<RID> materials; + materials.resize(sc); + + for (uint32_t i = 0; i < sc; i++) { + if (i < (uint32_t)p_material_overrides.size()) { + materials.write[i] = p_material_overrides[i]; + } + } + + gi_inst->set_surface_materials(materials); + + if (cull_argument.size() == 0) { + cull_argument.push_back(nullptr); + } + cull_argument[0] = gi_inst; + _render_uv2(cull_argument, fb, Rect2i(0, 0, p_image_size.width, p_image_size.height)); + + geometry_instance_free(gi_inst); + + TypedArray<Image> ret; + + { + PackedByteArray data = RD::get_singleton()->texture_get_data(albedo_alpha_tex, 0); + Ref<Image> img = Image::create_from_data(p_image_size.width, p_image_size.height, false, Image::FORMAT_RGBA8, data); + RD::get_singleton()->free(albedo_alpha_tex); + ret.push_back(img); + } + + { + PackedByteArray data = RD::get_singleton()->texture_get_data(normal_tex, 0); + Ref<Image> img = Image::create_from_data(p_image_size.width, p_image_size.height, false, Image::FORMAT_RGBA8, data); + RD::get_singleton()->free(normal_tex); + ret.push_back(img); + } + + { + PackedByteArray data = RD::get_singleton()->texture_get_data(orm_tex, 0); + Ref<Image> img = Image::create_from_data(p_image_size.width, p_image_size.height, false, Image::FORMAT_RGBA8, data); + RD::get_singleton()->free(orm_tex); + ret.push_back(img); + } + + { + PackedByteArray data = RD::get_singleton()->texture_get_data(emission_tex, 0); + Ref<Image> img = Image::create_from_data(p_image_size.width, p_image_size.height, false, Image::FORMAT_RGBAH, data); + RD::get_singleton()->free(emission_tex); + ret.push_back(img); + } + + RD::get_singleton()->free(depth_write_tex); + RD::get_singleton()->free(depth_tex); + + return ret; +} + +void RendererSceneRenderRD::sdfgi_set_debug_probe_select(const Vector3 &p_position, const Vector3 &p_dir) { + gi.sdfgi_debug_probe_pos = p_position; + gi.sdfgi_debug_probe_dir = p_dir; +} + +RendererSceneRenderRD *RendererSceneRenderRD::singleton = nullptr; + +bool RendererSceneRenderRD::is_vrs_supported() const { + return RD::get_singleton()->has_feature(RD::SUPPORTS_ATTACHMENT_VRS); +} + +bool RendererSceneRenderRD::is_dynamic_gi_supported() const { + // usable by default (unless low end = true) + return true; +} + +bool RendererSceneRenderRD::is_volumetric_supported() const { + // usable by default (unless low end = true) + return true; +} + +uint32_t RendererSceneRenderRD::get_max_elements() const { + return GLOBAL_GET("rendering/limits/cluster_builder/max_clustered_elements"); +} + +RendererSceneRenderRD::RendererSceneRenderRD() { + singleton = this; +} + +void RendererSceneRenderRD::init() { + max_cluster_elements = get_max_elements(); + RendererRD::LightStorage::get_singleton()->set_max_cluster_elements(max_cluster_elements); + + /* Forward ID */ + forward_id_storage = create_forward_id_storage(); + + /* SKY SHADER */ + + sky.init(); + + /* GI */ + + if (is_dynamic_gi_supported()) { + gi.init(&sky); + } + + { //decals + RendererRD::TextureStorage::get_singleton()->set_max_decals(max_cluster_elements); + } + + { //lights + } + + if (is_volumetric_supported()) { + RendererRD::Fog::get_singleton()->init_fog_shader(RendererRD::LightStorage::get_singleton()->get_max_directional_lights(), get_roughness_layers(), is_using_radiance_cubemap_array()); + } + + RSG::camera_attributes->camera_attributes_set_dof_blur_bokeh_shape(RS::DOFBokehShape(int(GLOBAL_GET("rendering/camera/depth_of_field/depth_of_field_bokeh_shape")))); + RSG::camera_attributes->camera_attributes_set_dof_blur_quality(RS::DOFBlurQuality(int(GLOBAL_GET("rendering/camera/depth_of_field/depth_of_field_bokeh_quality"))), GLOBAL_GET("rendering/camera/depth_of_field/depth_of_field_use_jitter")); + use_physical_light_units = GLOBAL_GET("rendering/lights_and_shadows/use_physical_light_units"); + + screen_space_roughness_limiter = GLOBAL_GET("rendering/anti_aliasing/screen_space_roughness_limiter/enabled"); + screen_space_roughness_limiter_amount = GLOBAL_GET("rendering/anti_aliasing/screen_space_roughness_limiter/amount"); + screen_space_roughness_limiter_limit = GLOBAL_GET("rendering/anti_aliasing/screen_space_roughness_limiter/limit"); + glow_bicubic_upscale = int(GLOBAL_GET("rendering/environment/glow/upscale_mode")) > 0; + glow_high_quality = GLOBAL_GET("rendering/environment/glow/use_high_quality"); + + directional_penumbra_shadow_kernel = memnew_arr(float, 128); + directional_soft_shadow_kernel = memnew_arr(float, 128); + penumbra_shadow_kernel = memnew_arr(float, 128); + soft_shadow_kernel = memnew_arr(float, 128); + positional_soft_shadow_filter_set_quality(RS::ShadowQuality(int(GLOBAL_GET("rendering/lights_and_shadows/positional_shadow/soft_shadow_filter_quality")))); + directional_soft_shadow_filter_set_quality(RS::ShadowQuality(int(GLOBAL_GET("rendering/lights_and_shadows/directional_shadow/soft_shadow_filter_quality")))); + + environment_set_volumetric_fog_volume_size(GLOBAL_GET("rendering/environment/volumetric_fog/volume_size"), GLOBAL_GET("rendering/environment/volumetric_fog/volume_depth")); + environment_set_volumetric_fog_filter_active(GLOBAL_GET("rendering/environment/volumetric_fog/use_filter")); + + decals_set_filter(RS::DecalFilter(int(GLOBAL_GET("rendering/textures/decals/filter")))); + light_projectors_set_filter(RS::LightProjectorFilter(int(GLOBAL_GET("rendering/textures/light_projectors/filter")))); + + cull_argument.set_page_pool(&cull_argument_pool); + + bool can_use_storage = _render_buffers_can_be_storage(); + bokeh_dof = memnew(RendererRD::BokehDOF(!can_use_storage)); + copy_effects = memnew(RendererRD::CopyEffects(!can_use_storage)); + tone_mapper = memnew(RendererRD::ToneMapper); + vrs = memnew(RendererRD::VRS); + if (can_use_storage) { + fsr = memnew(RendererRD::FSR); + } +} + +RendererSceneRenderRD::~RendererSceneRenderRD() { + if (forward_id_storage) { + memdelete(forward_id_storage); + } + + if (bokeh_dof) { + memdelete(bokeh_dof); + } + if (copy_effects) { + memdelete(copy_effects); + } + if (tone_mapper) { + memdelete(tone_mapper); + } + if (vrs) { + memdelete(vrs); + } + if (fsr) { + memdelete(fsr); + } + + if (sky.sky_scene_state.uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(sky.sky_scene_state.uniform_set)) { + RD::get_singleton()->free(sky.sky_scene_state.uniform_set); + } + + if (is_dynamic_gi_supported()) { + gi.free(); + } + + if (is_volumetric_supported()) { + RendererRD::Fog::get_singleton()->free_fog_shader(); + } + + memdelete_arr(directional_penumbra_shadow_kernel); + memdelete_arr(directional_soft_shadow_kernel); + memdelete_arr(penumbra_shadow_kernel); + memdelete_arr(soft_shadow_kernel); + + RSG::light_storage->directional_shadow_atlas_set_size(0); + cull_argument.reset(); //avoid exit error +} diff --git a/servers/rendering/renderer_rd/renderer_scene_render_rd.h b/servers/rendering/renderer_rd/renderer_scene_render_rd.h new file mode 100644 index 0000000000..2312603829 --- /dev/null +++ b/servers/rendering/renderer_rd/renderer_scene_render_rd.h @@ -0,0 +1,393 @@ +/*************************************************************************/ +/* renderer_scene_render_rd.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef RENDERER_SCENE_RENDER_RD_H +#define RENDERER_SCENE_RENDER_RD_H + +#include "core/templates/local_vector.h" +#include "core/templates/rid_owner.h" +#include "servers/rendering/renderer_compositor.h" +#include "servers/rendering/renderer_rd/cluster_builder_rd.h" +#include "servers/rendering/renderer_rd/effects/bokeh_dof.h" +#include "servers/rendering/renderer_rd/effects/copy_effects.h" +#include "servers/rendering/renderer_rd/effects/fsr.h" +#include "servers/rendering/renderer_rd/effects/tone_mapper.h" +#include "servers/rendering/renderer_rd/effects/vrs.h" +#include "servers/rendering/renderer_rd/environment/fog.h" +#include "servers/rendering/renderer_rd/environment/gi.h" +#include "servers/rendering/renderer_rd/environment/sky.h" +#include "servers/rendering/renderer_rd/framebuffer_cache_rd.h" +#include "servers/rendering/renderer_rd/storage_rd/light_storage.h" +#include "servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.h" +#include "servers/rendering/renderer_rd/storage_rd/render_scene_data_rd.h" +#include "servers/rendering/renderer_scene_render.h" +#include "servers/rendering/rendering_device.h" +#include "servers/rendering/rendering_method.h" + +// For RenderDataRD, possibly inherited from RefCounted and add proper getters for our implementation classes + +struct RenderDataRD { + Ref<RenderSceneBuffersRD> render_buffers; + RenderSceneDataRD *scene_data; + + const PagedArray<RenderGeometryInstance *> *instances = nullptr; + const PagedArray<RID> *lights = nullptr; + const PagedArray<RID> *reflection_probes = nullptr; + const PagedArray<RID> *voxel_gi_instances = nullptr; + const PagedArray<RID> *decals = nullptr; + const PagedArray<RID> *lightmaps = nullptr; + const PagedArray<RID> *fog_volumes = nullptr; + RID environment; + RID camera_attributes; + RID shadow_atlas; + RID occluder_debug_tex; + RID reflection_atlas; + RID reflection_probe; + int reflection_probe_pass = 0; + + RID cluster_buffer; + uint32_t cluster_size = 0; + uint32_t cluster_max_elements = 0; + + uint32_t directional_light_count = 0; + bool directional_light_soft_shadows = false; + + RenderingMethod::RenderInfo *render_info = nullptr; + + /* Shadow data */ + const RendererSceneRender::RenderShadowData *render_shadows = nullptr; + int render_shadow_count = 0; + + LocalVector<int> cube_shadows; + LocalVector<int> shadows; + LocalVector<int> directional_shadows; + + /* GI info */ + const RendererSceneRender::RenderSDFGIData *render_sdfgi_regions = nullptr; + int render_sdfgi_region_count = 0; + const RendererSceneRender::RenderSDFGIUpdateData *sdfgi_update_data = nullptr; + + uint32_t voxel_gi_count = 0; +}; + +class RendererSceneRenderRD : public RendererSceneRender { + friend RendererRD::SkyRD; + friend RendererRD::GI; + +protected: + RendererRD::ForwardIDStorage *forward_id_storage = nullptr; + RendererRD::BokehDOF *bokeh_dof = nullptr; + RendererRD::CopyEffects *copy_effects = nullptr; + RendererRD::ToneMapper *tone_mapper = nullptr; + RendererRD::FSR *fsr = nullptr; + RendererRD::VRS *vrs = nullptr; + double time = 0.0; + double time_step = 0.0; + + /* ENVIRONMENT */ + + bool glow_bicubic_upscale = false; + bool glow_high_quality = false; + + bool use_physical_light_units = false; + + //////////////////////////////// + + virtual RendererRD::ForwardIDStorage *create_forward_id_storage() { return memnew(RendererRD::ForwardIDStorage); }; + + void _update_vrs(Ref<RenderSceneBuffersRD> p_render_buffers); + + virtual void setup_render_buffer_data(Ref<RenderSceneBuffersRD> p_render_buffers) = 0; + + virtual void _render_scene(RenderDataRD *p_render_data, const Color &p_default_color) = 0; + virtual void _render_buffers_debug_draw(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_shadow_atlas, RID p_occlusion_buffer); + + virtual void _render_material(const Transform3D &p_cam_transform, const Projection &p_cam_projection, bool p_cam_orthogonal, const PagedArray<RenderGeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region, float p_exposure_normalization) = 0; + virtual void _render_uv2(const PagedArray<RenderGeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) = 0; + virtual void _render_sdfgi(Ref<RenderSceneBuffersRD> p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<RenderGeometryInstance *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture, float p_exposure_normalization) = 0; + virtual void _render_particle_collider_heightfield(RID p_fb, const Transform3D &p_cam_transform, const Projection &p_cam_projection, const PagedArray<RenderGeometryInstance *> &p_instances) = 0; + + void _debug_sdfgi_probes(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_framebuffer, uint32_t p_view_count, const Projection *p_camera_with_transforms, bool p_will_continue_color, bool p_will_continue_depth); + + virtual RID _render_buffers_get_normal_texture(Ref<RenderSceneBuffersRD> p_render_buffers) = 0; + virtual RID _render_buffers_get_velocity_texture(Ref<RenderSceneBuffersRD> p_render_buffers) = 0; + + bool _needs_post_prepass_render(RenderDataRD *p_render_data, bool p_use_gi); + void _post_prepass_render(RenderDataRD *p_render_data, bool p_use_gi); + void _pre_resolve_render(RenderDataRD *p_render_data, bool p_use_gi); + + void _render_buffers_copy_screen_texture(const RenderDataRD *p_render_data); + void _render_buffers_copy_depth_texture(const RenderDataRD *p_render_data); + void _render_buffers_post_process_and_tonemap(const RenderDataRD *p_render_data); + void _post_process_subpass(RID p_source_texture, RID p_framebuffer, const RenderDataRD *p_render_data); + void _disable_clear_request(const RenderDataRD *p_render_data); + + // needed for a single argument calls (material and uv2) + PagedArrayPool<RenderGeometryInstance *> cull_argument_pool; + PagedArray<RenderGeometryInstance *> cull_argument; //need this to exist + + RendererRD::SkyRD sky; + RendererRD::GI gi; + + virtual void _update_shader_quality_settings() {} + +private: + RS::ViewportDebugDraw debug_draw = RS::VIEWPORT_DEBUG_DRAW_DISABLED; + static RendererSceneRenderRD *singleton; + + /* Shadow atlas */ + RS::ShadowQuality shadows_quality = RS::SHADOW_QUALITY_MAX; //So it always updates when first set + RS::ShadowQuality directional_shadow_quality = RS::SHADOW_QUALITY_MAX; + float shadows_quality_radius = 1.0; + float directional_shadow_quality_radius = 1.0; + + float *directional_penumbra_shadow_kernel = nullptr; + float *directional_soft_shadow_kernel = nullptr; + float *penumbra_shadow_kernel = nullptr; + float *soft_shadow_kernel = nullptr; + int directional_penumbra_shadow_samples = 0; + int directional_soft_shadow_samples = 0; + int penumbra_shadow_samples = 0; + int soft_shadow_samples = 0; + RS::DecalFilter decals_filter = RS::DECAL_FILTER_LINEAR_MIPMAPS; + RS::LightProjectorFilter light_projectors_filter = RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS; + + /* RENDER BUFFERS */ + + // TODO move into effects/luminance.h/cpp + void _allocate_luminance_textures(Ref<RenderSceneBuffersRD> rb); + + /* GI */ + bool screen_space_roughness_limiter = false; + float screen_space_roughness_limiter_amount = 0.25; + float screen_space_roughness_limiter_limit = 0.18; + + /* Light data */ + + uint64_t scene_pass = 0; + + uint32_t max_cluster_elements = 512; + + /* Volumetric Fog */ + + uint32_t volumetric_fog_size = 128; + uint32_t volumetric_fog_depth = 128; + bool volumetric_fog_filter_active = true; + +public: + static RendererSceneRenderRD *get_singleton() { return singleton; } + + /* LIGHTING */ + + virtual void setup_added_reflection_probe(const Transform3D &p_transform, const Vector3 &p_half_extents){}; + virtual void setup_added_light(const RS::LightType p_type, const Transform3D &p_transform, float p_radius, float p_spot_aperture){}; + virtual void setup_added_decal(const Transform3D &p_transform, const Vector3 &p_half_extents){}; + + /* GI */ + + RendererRD::GI *get_gi() { return &gi; } + + /* SKY */ + + RendererRD::SkyRD *get_sky() { return &sky; } + + /* SDFGI UPDATE */ + + virtual void sdfgi_update(const Ref<RenderSceneBuffers> &p_render_buffers, RID p_environment, const Vector3 &p_world_position) override; + virtual int sdfgi_get_pending_region_count(const Ref<RenderSceneBuffers> &p_render_buffers) const override; + virtual AABB sdfgi_get_pending_region_bounds(const Ref<RenderSceneBuffers> &p_render_buffers, int p_region) const override; + virtual uint32_t sdfgi_get_pending_region_cascade(const Ref<RenderSceneBuffers> &p_render_buffers, int p_region) const override; + RID sdfgi_get_ubo() const { return gi.sdfgi_ubo; } + + /* SKY API */ + + virtual RID sky_allocate() override; + virtual void sky_initialize(RID p_rid) override; + + virtual void sky_set_radiance_size(RID p_sky, int p_radiance_size) override; + virtual void sky_set_mode(RID p_sky, RS::SkyMode p_mode) override; + virtual void sky_set_material(RID p_sky, RID p_material) override; + virtual Ref<Image> sky_bake_panorama(RID p_sky, float p_energy, bool p_bake_irradiance, const Size2i &p_size) override; + + /* ENVIRONMENT API */ + + virtual void environment_glow_set_use_bicubic_upscale(bool p_enable) override; + virtual void environment_glow_set_use_high_quality(bool p_enable) override; + + virtual void environment_set_volumetric_fog_volume_size(int p_size, int p_depth) override; + virtual void environment_set_volumetric_fog_filter_active(bool p_enable) override; + + virtual void environment_set_sdfgi_ray_count(RS::EnvironmentSDFGIRayCount p_ray_count) override; + virtual void environment_set_sdfgi_frames_to_converge(RS::EnvironmentSDFGIFramesToConverge p_frames) override; + virtual void environment_set_sdfgi_frames_to_update_light(RS::EnvironmentSDFGIFramesToUpdateLight p_update) override; + + virtual Ref<Image> environment_bake_panorama(RID p_env, bool p_bake_irradiance, const Size2i &p_size) override; + + _FORCE_INLINE_ bool is_using_physical_light_units() { + return use_physical_light_units; + } + + /* REFLECTION PROBE */ + + virtual RID reflection_probe_create_framebuffer(RID p_color, RID p_depth); + + /* FOG VOLUMES */ + + uint32_t get_volumetric_fog_size() const { return volumetric_fog_size; } + uint32_t get_volumetric_fog_depth() const { return volumetric_fog_depth; } + bool get_volumetric_fog_filter_active() const { return volumetric_fog_filter_active; } + + virtual RID fog_volume_instance_create(RID p_fog_volume) override; + virtual void fog_volume_instance_set_transform(RID p_fog_volume_instance, const Transform3D &p_transform) override; + virtual void fog_volume_instance_set_active(RID p_fog_volume_instance, bool p_active) override; + virtual RID fog_volume_instance_get_volume(RID p_fog_volume_instance) const override; + virtual Vector3 fog_volume_instance_get_position(RID p_fog_volume_instance) const override; + + /* gi light probes */ + + virtual RID voxel_gi_instance_create(RID p_base) override; + virtual void voxel_gi_instance_set_transform_to_data(RID p_probe, const Transform3D &p_xform) override; + virtual bool voxel_gi_needs_update(RID p_probe) const override; + virtual void voxel_gi_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RenderGeometryInstance *> &p_dynamic_objects) override; + virtual void voxel_gi_set_quality(RS::VoxelGIQuality p_quality) override { gi.voxel_gi_quality = p_quality; } + + /* render buffers */ + + virtual float _render_buffers_get_luminance_multiplier(); + virtual RD::DataFormat _render_buffers_get_color_format(); + virtual bool _render_buffers_can_be_storage(); + virtual Ref<RenderSceneBuffers> render_buffers_create() override; + virtual void gi_set_use_half_resolution(bool p_enable) override; + + RID render_buffers_get_default_voxel_gi_buffer(); + + virtual void base_uniforms_changed() = 0; + virtual void update_uniform_sets(){}; + + virtual void render_scene(const Ref<RenderSceneBuffers> &p_render_buffers, const CameraData *p_camera_data, const CameraData *p_prev_camera_data, const PagedArray<RenderGeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_voxel_gi_instances, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, const PagedArray<RID> &p_fog_volumes, RID p_environment, RID p_camera_attributes, RID p_shadow_atlas, RID p_occluder_debug_tex, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_mesh_lod_threshold, const RenderShadowData *p_render_shadows, int p_render_shadow_count, const RenderSDFGIData *p_render_sdfgi_regions, int p_render_sdfgi_region_count, const RenderSDFGIUpdateData *p_sdfgi_update_data = nullptr, RenderingMethod::RenderInfo *r_render_info = nullptr) override; + + virtual void render_material(const Transform3D &p_cam_transform, const Projection &p_cam_projection, bool p_cam_orthogonal, const PagedArray<RenderGeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) override; + + virtual void render_particle_collider_heightfield(RID p_collider, const Transform3D &p_transform, const PagedArray<RenderGeometryInstance *> &p_instances) override; + + virtual void set_scene_pass(uint64_t p_pass) override { + scene_pass = p_pass; + } + _FORCE_INLINE_ uint64_t get_scene_pass() { + return scene_pass; + } + + virtual void screen_space_roughness_limiter_set_active(bool p_enable, float p_amount, float p_limit) override; + virtual bool screen_space_roughness_limiter_is_active() const override; + virtual float screen_space_roughness_limiter_get_amount() const; + virtual float screen_space_roughness_limiter_get_limit() const; + + virtual void positional_soft_shadow_filter_set_quality(RS::ShadowQuality p_quality) override; + virtual void directional_soft_shadow_filter_set_quality(RS::ShadowQuality p_quality) override; + + virtual void decals_set_filter(RS::DecalFilter p_filter) override; + virtual void light_projectors_set_filter(RS::LightProjectorFilter p_filter) override; + + _FORCE_INLINE_ RS::ShadowQuality shadows_quality_get() const { + return shadows_quality; + } + _FORCE_INLINE_ RS::ShadowQuality directional_shadow_quality_get() const { + return directional_shadow_quality; + } + _FORCE_INLINE_ float shadows_quality_radius_get() const { + return shadows_quality_radius; + } + _FORCE_INLINE_ float directional_shadow_quality_radius_get() const { + return directional_shadow_quality_radius; + } + + _FORCE_INLINE_ float *directional_penumbra_shadow_kernel_get() { + return directional_penumbra_shadow_kernel; + } + _FORCE_INLINE_ float *directional_soft_shadow_kernel_get() { + return directional_soft_shadow_kernel; + } + _FORCE_INLINE_ float *penumbra_shadow_kernel_get() { + return penumbra_shadow_kernel; + } + _FORCE_INLINE_ float *soft_shadow_kernel_get() { + return soft_shadow_kernel; + } + + _FORCE_INLINE_ int directional_penumbra_shadow_samples_get() const { + return directional_penumbra_shadow_samples; + } + _FORCE_INLINE_ int directional_soft_shadow_samples_get() const { + return directional_soft_shadow_samples; + } + _FORCE_INLINE_ int penumbra_shadow_samples_get() const { + return penumbra_shadow_samples; + } + _FORCE_INLINE_ int soft_shadow_samples_get() const { + return soft_shadow_samples; + } + + _FORCE_INLINE_ RS::LightProjectorFilter light_projectors_get_filter() const { + return light_projectors_filter; + } + _FORCE_INLINE_ RS::DecalFilter decals_get_filter() const { + return decals_filter; + } + + int get_roughness_layers() const; + bool is_using_radiance_cubemap_array() const; + + virtual TypedArray<Image> bake_render_uv2(RID p_base, const TypedArray<RID> &p_material_overrides, const Size2i &p_image_size) override; + + virtual bool free(RID p_rid) override; + + virtual void update() override; + + virtual void set_debug_draw_mode(RS::ViewportDebugDraw p_debug_draw) override; + _FORCE_INLINE_ RS::ViewportDebugDraw get_debug_draw_mode() const { + return debug_draw; + } + + virtual void set_time(double p_time, double p_step) override; + + virtual void sdfgi_set_debug_probe_select(const Vector3 &p_position, const Vector3 &p_dir) override; + + virtual bool is_vrs_supported() const; + virtual bool is_dynamic_gi_supported() const; + virtual bool is_volumetric_supported() const; + virtual uint32_t get_max_elements() const; + + void init(); + + RendererSceneRenderRD(); + ~RendererSceneRenderRD(); +}; + +#endif // RENDERER_SCENE_RENDER_RD_H diff --git a/servers/rendering/renderer_rd/shader_rd.cpp b/servers/rendering/renderer_rd/shader_rd.cpp new file mode 100644 index 0000000000..5e9eadadd9 --- /dev/null +++ b/servers/rendering/renderer_rd/shader_rd.cpp @@ -0,0 +1,709 @@ +/*************************************************************************/ +/* shader_rd.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "shader_rd.h" + +#include "core/io/compression.h" +#include "core/io/dir_access.h" +#include "core/io/file_access.h" +#include "renderer_compositor_rd.h" +#include "servers/rendering/rendering_device.h" +#include "thirdparty/misc/smolv.h" + +void ShaderRD::_add_stage(const char *p_code, StageType p_stage_type) { + Vector<String> lines = String(p_code).split("\n"); + + String text; + + for (int i = 0; i < lines.size(); i++) { + String l = lines[i]; + bool push_chunk = false; + + StageTemplate::Chunk chunk; + + if (l.begins_with("#VERSION_DEFINES")) { + chunk.type = StageTemplate::Chunk::TYPE_VERSION_DEFINES; + push_chunk = true; + } else if (l.begins_with("#GLOBALS")) { + switch (p_stage_type) { + case STAGE_TYPE_VERTEX: + chunk.type = StageTemplate::Chunk::TYPE_VERTEX_GLOBALS; + break; + case STAGE_TYPE_FRAGMENT: + chunk.type = StageTemplate::Chunk::TYPE_FRAGMENT_GLOBALS; + break; + case STAGE_TYPE_COMPUTE: + chunk.type = StageTemplate::Chunk::TYPE_COMPUTE_GLOBALS; + break; + default: { + } + } + + push_chunk = true; + } else if (l.begins_with("#MATERIAL_UNIFORMS")) { + chunk.type = StageTemplate::Chunk::TYPE_MATERIAL_UNIFORMS; + push_chunk = true; + } else if (l.begins_with("#CODE")) { + chunk.type = StageTemplate::Chunk::TYPE_CODE; + push_chunk = true; + chunk.code = l.replace_first("#CODE", String()).replace(":", "").strip_edges().to_upper(); + } else { + text += l + "\n"; + } + + if (push_chunk) { + if (!text.is_empty()) { + StageTemplate::Chunk text_chunk; + text_chunk.type = StageTemplate::Chunk::TYPE_TEXT; + text_chunk.text = text.utf8(); + stage_templates[p_stage_type].chunks.push_back(text_chunk); + text = String(); + } + stage_templates[p_stage_type].chunks.push_back(chunk); + } + } + + if (!text.is_empty()) { + StageTemplate::Chunk text_chunk; + text_chunk.type = StageTemplate::Chunk::TYPE_TEXT; + text_chunk.text = text.utf8(); + stage_templates[p_stage_type].chunks.push_back(text_chunk); + text = String(); + } +} + +void ShaderRD::setup(const char *p_vertex_code, const char *p_fragment_code, const char *p_compute_code, const char *p_name) { + name = p_name; + + if (p_compute_code) { + _add_stage(p_compute_code, STAGE_TYPE_COMPUTE); + is_compute = true; + } else { + is_compute = false; + if (p_vertex_code) { + _add_stage(p_vertex_code, STAGE_TYPE_VERTEX); + } + if (p_fragment_code) { + _add_stage(p_fragment_code, STAGE_TYPE_FRAGMENT); + } + } + + StringBuilder tohash; + tohash.append("[SpirvCacheKey]"); + tohash.append(RenderingDevice::get_singleton()->shader_get_spirv_cache_key()); + tohash.append("[BinaryCacheKey]"); + tohash.append(RenderingDevice::get_singleton()->shader_get_binary_cache_key()); + tohash.append("[Vertex]"); + tohash.append(p_vertex_code ? p_vertex_code : ""); + tohash.append("[Fragment]"); + tohash.append(p_fragment_code ? p_fragment_code : ""); + tohash.append("[Compute]"); + tohash.append(p_compute_code ? p_compute_code : ""); + + base_sha256 = tohash.as_string().sha256_text(); +} + +RID ShaderRD::version_create() { + //initialize() was never called + ERR_FAIL_COND_V(variant_defines.size() == 0, RID()); + + Version version; + version.dirty = true; + version.valid = false; + version.initialize_needed = true; + version.variants = nullptr; + return version_owner.make_rid(version); +} + +void ShaderRD::_clear_version(Version *p_version) { + //clear versions if they exist + if (p_version->variants) { + for (int i = 0; i < variant_defines.size(); i++) { + if (variants_enabled[i]) { + RD::get_singleton()->free(p_version->variants[i]); + } + } + + memdelete_arr(p_version->variants); + if (p_version->variant_data) { + memdelete_arr(p_version->variant_data); + } + p_version->variants = nullptr; + } +} + +void ShaderRD::_build_variant_code(StringBuilder &builder, uint32_t p_variant, const Version *p_version, const StageTemplate &p_template) { + for (uint32_t i = 0; i < p_template.chunks.size(); i++) { + const StageTemplate::Chunk &chunk = p_template.chunks[i]; + switch (chunk.type) { + case StageTemplate::Chunk::TYPE_VERSION_DEFINES: { + builder.append("\n"); //make sure defines begin at newline + builder.append(general_defines.get_data()); + builder.append(variant_defines[p_variant].get_data()); + for (int j = 0; j < p_version->custom_defines.size(); j++) { + builder.append(p_version->custom_defines[j].get_data()); + } + builder.append("\n"); //make sure defines begin at newline + if (p_version->uniforms.size()) { + builder.append("#define MATERIAL_UNIFORMS_USED\n"); + } + for (const KeyValue<StringName, CharString> &E : p_version->code_sections) { + builder.append(String("#define ") + String(E.key) + "_CODE_USED\n"); + } +#if defined(MACOS_ENABLED) || defined(IOS_ENABLED) + builder.append("#define MOLTENVK_USED\n"); +#endif + builder.append(String("#define RENDER_DRIVER_") + OS::get_singleton()->get_current_rendering_driver_name().to_upper() + "\n"); + } break; + case StageTemplate::Chunk::TYPE_MATERIAL_UNIFORMS: { + builder.append(p_version->uniforms.get_data()); //uniforms (same for vertex and fragment) + } break; + case StageTemplate::Chunk::TYPE_VERTEX_GLOBALS: { + builder.append(p_version->vertex_globals.get_data()); // vertex globals + } break; + case StageTemplate::Chunk::TYPE_FRAGMENT_GLOBALS: { + builder.append(p_version->fragment_globals.get_data()); // fragment globals + } break; + case StageTemplate::Chunk::TYPE_COMPUTE_GLOBALS: { + builder.append(p_version->compute_globals.get_data()); // compute globals + } break; + case StageTemplate::Chunk::TYPE_CODE: { + if (p_version->code_sections.has(chunk.code)) { + builder.append(p_version->code_sections[chunk.code].get_data()); + } + } break; + case StageTemplate::Chunk::TYPE_TEXT: { + builder.append(chunk.text.get_data()); + } break; + } + } +} + +void ShaderRD::_compile_variant(uint32_t p_variant, Version *p_version) { + if (!variants_enabled[p_variant]) { + return; //variant is disabled, return + } + + Vector<RD::ShaderStageSPIRVData> stages; + + String error; + String current_source; + RD::ShaderStage current_stage = RD::SHADER_STAGE_VERTEX; + bool build_ok = true; + + if (!is_compute) { + //vertex stage + + StringBuilder builder; + _build_variant_code(builder, p_variant, p_version, stage_templates[STAGE_TYPE_VERTEX]); + + current_source = builder.as_string(); + RD::ShaderStageSPIRVData stage; + stage.spir_v = RD::get_singleton()->shader_compile_spirv_from_source(RD::SHADER_STAGE_VERTEX, current_source, RD::SHADER_LANGUAGE_GLSL, &error); + if (stage.spir_v.size() == 0) { + build_ok = false; + } else { + stage.shader_stage = RD::SHADER_STAGE_VERTEX; + stages.push_back(stage); + } + } + + if (!is_compute && build_ok) { + //fragment stage + current_stage = RD::SHADER_STAGE_FRAGMENT; + + StringBuilder builder; + _build_variant_code(builder, p_variant, p_version, stage_templates[STAGE_TYPE_FRAGMENT]); + + current_source = builder.as_string(); + RD::ShaderStageSPIRVData stage; + stage.spir_v = RD::get_singleton()->shader_compile_spirv_from_source(RD::SHADER_STAGE_FRAGMENT, current_source, RD::SHADER_LANGUAGE_GLSL, &error); + if (stage.spir_v.size() == 0) { + build_ok = false; + } else { + stage.shader_stage = RD::SHADER_STAGE_FRAGMENT; + stages.push_back(stage); + } + } + + if (is_compute) { + //compute stage + current_stage = RD::SHADER_STAGE_COMPUTE; + + StringBuilder builder; + _build_variant_code(builder, p_variant, p_version, stage_templates[STAGE_TYPE_COMPUTE]); + + current_source = builder.as_string(); + + RD::ShaderStageSPIRVData stage; + stage.spir_v = RD::get_singleton()->shader_compile_spirv_from_source(RD::SHADER_STAGE_COMPUTE, current_source, RD::SHADER_LANGUAGE_GLSL, &error); + if (stage.spir_v.size() == 0) { + build_ok = false; + } else { + stage.shader_stage = RD::SHADER_STAGE_COMPUTE; + stages.push_back(stage); + } + } + + if (!build_ok) { + MutexLock lock(variant_set_mutex); //properly print the errors + ERR_PRINT("Error compiling " + String(current_stage == RD::SHADER_STAGE_COMPUTE ? "Compute " : (current_stage == RD::SHADER_STAGE_VERTEX ? "Vertex" : "Fragment")) + " shader, variant #" + itos(p_variant) + " (" + variant_defines[p_variant].get_data() + ")."); + ERR_PRINT(error); + +#ifdef DEBUG_ENABLED + ERR_PRINT("code:\n" + current_source.get_with_code_lines()); +#endif + return; + } + + Vector<uint8_t> shader_data = RD::get_singleton()->shader_compile_binary_from_spirv(stages, name + ":" + itos(p_variant)); + + ERR_FAIL_COND(shader_data.size() == 0); + + RID shader = RD::get_singleton()->shader_create_from_bytecode(shader_data); + { + MutexLock lock(variant_set_mutex); + p_version->variants[p_variant] = shader; + p_version->variant_data[p_variant] = shader_data; + } +} + +RS::ShaderNativeSourceCode ShaderRD::version_get_native_source_code(RID p_version) { + Version *version = version_owner.get_or_null(p_version); + RS::ShaderNativeSourceCode source_code; + ERR_FAIL_COND_V(!version, source_code); + + source_code.versions.resize(variant_defines.size()); + + for (int i = 0; i < source_code.versions.size(); i++) { + if (!is_compute) { + //vertex stage + + StringBuilder builder; + _build_variant_code(builder, i, version, stage_templates[STAGE_TYPE_VERTEX]); + + RS::ShaderNativeSourceCode::Version::Stage stage; + stage.name = "vertex"; + stage.code = builder.as_string(); + + source_code.versions.write[i].stages.push_back(stage); + } + + if (!is_compute) { + //fragment stage + + StringBuilder builder; + _build_variant_code(builder, i, version, stage_templates[STAGE_TYPE_FRAGMENT]); + + RS::ShaderNativeSourceCode::Version::Stage stage; + stage.name = "fragment"; + stage.code = builder.as_string(); + + source_code.versions.write[i].stages.push_back(stage); + } + + if (is_compute) { + //compute stage + + StringBuilder builder; + _build_variant_code(builder, i, version, stage_templates[STAGE_TYPE_COMPUTE]); + + RS::ShaderNativeSourceCode::Version::Stage stage; + stage.name = "compute"; + stage.code = builder.as_string(); + + source_code.versions.write[i].stages.push_back(stage); + } + } + + return source_code; +} + +String ShaderRD::_version_get_sha1(Version *p_version) const { + StringBuilder hash_build; + + hash_build.append("[uniforms]"); + hash_build.append(p_version->uniforms.get_data()); + hash_build.append("[vertex_globals]"); + hash_build.append(p_version->vertex_globals.get_data()); + hash_build.append("[fragment_globals]"); + hash_build.append(p_version->fragment_globals.get_data()); + hash_build.append("[compute_globals]"); + hash_build.append(p_version->compute_globals.get_data()); + + Vector<StringName> code_sections; + for (const KeyValue<StringName, CharString> &E : p_version->code_sections) { + code_sections.push_back(E.key); + } + code_sections.sort_custom<StringName::AlphCompare>(); + + for (int i = 0; i < code_sections.size(); i++) { + hash_build.append(String("[code:") + String(code_sections[i]) + "]"); + hash_build.append(p_version->code_sections[code_sections[i]].get_data()); + } + for (int i = 0; i < p_version->custom_defines.size(); i++) { + hash_build.append("[custom_defines:" + itos(i) + "]"); + hash_build.append(p_version->custom_defines[i].get_data()); + } + + return hash_build.as_string().sha1_text(); +} + +static const char *shader_file_header = "GDSC"; +static const uint32_t cache_file_version = 2; + +bool ShaderRD::_load_from_cache(Version *p_version) { + String sha1 = _version_get_sha1(p_version); + String path = shader_cache_dir.path_join(name).path_join(base_sha256).path_join(sha1) + ".cache"; + + Ref<FileAccess> f = FileAccess::open(path, FileAccess::READ); + if (f.is_null()) { + return false; + } + + char header[5] = { 0, 0, 0, 0, 0 }; + f->get_buffer((uint8_t *)header, 4); + ERR_FAIL_COND_V(header != String(shader_file_header), false); + + uint32_t file_version = f->get_32(); + if (file_version != cache_file_version) { + return false; // wrong version + } + + uint32_t variant_count = f->get_32(); + + ERR_FAIL_COND_V(variant_count != (uint32_t)variant_defines.size(), false); //should not happen but check + + for (uint32_t i = 0; i < variant_count; i++) { + uint32_t variant_size = f->get_32(); + ERR_FAIL_COND_V(variant_size == 0 && variants_enabled[i], false); + if (!variants_enabled[i]) { + continue; + } + Vector<uint8_t> variant_bytes; + variant_bytes.resize(variant_size); + + uint32_t br = f->get_buffer(variant_bytes.ptrw(), variant_size); + + ERR_FAIL_COND_V(br != variant_size, false); + + p_version->variant_data[i] = variant_bytes; + } + + for (uint32_t i = 0; i < variant_count; i++) { + if (!variants_enabled[i]) { + MutexLock lock(variant_set_mutex); + p_version->variants[i] = RID(); + continue; + } + RID shader = RD::get_singleton()->shader_create_from_bytecode(p_version->variant_data[i]); + if (shader.is_null()) { + for (uint32_t j = 0; j < i; j++) { + RD::get_singleton()->free(p_version->variants[i]); + } + ERR_FAIL_COND_V(shader.is_null(), false); + } + { + MutexLock lock(variant_set_mutex); + p_version->variants[i] = shader; + } + } + + memdelete_arr(p_version->variant_data); //clear stages + p_version->variant_data = nullptr; + p_version->valid = true; + return true; +} + +void ShaderRD::_save_to_cache(Version *p_version) { + String sha1 = _version_get_sha1(p_version); + String path = shader_cache_dir.path_join(name).path_join(base_sha256).path_join(sha1) + ".cache"; + + Ref<FileAccess> f = FileAccess::open(path, FileAccess::WRITE); + ERR_FAIL_COND(f.is_null()); + f->store_buffer((const uint8_t *)shader_file_header, 4); + f->store_32(cache_file_version); //file version + uint32_t variant_count = variant_defines.size(); + f->store_32(variant_count); //variant count + + for (uint32_t i = 0; i < variant_count; i++) { + f->store_32(p_version->variant_data[i].size()); //stage count + f->store_buffer(p_version->variant_data[i].ptr(), p_version->variant_data[i].size()); + } +} + +void ShaderRD::_compile_version(Version *p_version) { + _clear_version(p_version); + + p_version->valid = false; + p_version->dirty = false; + + p_version->variants = memnew_arr(RID, variant_defines.size()); + typedef Vector<uint8_t> ShaderStageData; + p_version->variant_data = memnew_arr(ShaderStageData, variant_defines.size()); + + if (shader_cache_dir_valid) { + if (_load_from_cache(p_version)) { + return; + } + } + +#if 1 + + WorkerThreadPool::GroupID group_task = WorkerThreadPool::get_singleton()->add_template_group_task(this, &ShaderRD::_compile_variant, p_version, variant_defines.size(), -1, true, SNAME("ShaderCompilation")); + WorkerThreadPool::get_singleton()->wait_for_group_task_completion(group_task); + +#else + for (int i = 0; i < variant_defines.size(); i++) { + _compile_variant(i, p_version); + } +#endif + + bool all_valid = true; + for (int i = 0; i < variant_defines.size(); i++) { + if (!variants_enabled[i]) { + continue; //disabled + } + if (p_version->variants[i].is_null()) { + all_valid = false; + break; + } + } + + if (!all_valid) { + //clear versions if they exist + for (int i = 0; i < variant_defines.size(); i++) { + if (!variants_enabled[i]) { + continue; //disabled + } + if (!p_version->variants[i].is_null()) { + RD::get_singleton()->free(p_version->variants[i]); + } + } + memdelete_arr(p_version->variants); + if (p_version->variant_data) { + memdelete_arr(p_version->variant_data); + } + p_version->variants = nullptr; + p_version->variant_data = nullptr; + return; + } else if (shader_cache_dir_valid) { + //save shader cache + _save_to_cache(p_version); + } + + memdelete_arr(p_version->variant_data); //clear stages + p_version->variant_data = nullptr; + + p_version->valid = true; +} + +void ShaderRD::version_set_code(RID p_version, const HashMap<String, String> &p_code, const String &p_uniforms, const String &p_vertex_globals, const String &p_fragment_globals, const Vector<String> &p_custom_defines) { + ERR_FAIL_COND(is_compute); + + Version *version = version_owner.get_or_null(p_version); + ERR_FAIL_COND(!version); + version->vertex_globals = p_vertex_globals.utf8(); + version->fragment_globals = p_fragment_globals.utf8(); + version->uniforms = p_uniforms.utf8(); + version->code_sections.clear(); + for (const KeyValue<String, String> &E : p_code) { + version->code_sections[StringName(E.key.to_upper())] = E.value.utf8(); + } + + version->custom_defines.clear(); + for (int i = 0; i < p_custom_defines.size(); i++) { + version->custom_defines.push_back(p_custom_defines[i].utf8()); + } + + version->dirty = true; + if (version->initialize_needed) { + _compile_version(version); + version->initialize_needed = false; + } +} + +void ShaderRD::version_set_compute_code(RID p_version, const HashMap<String, String> &p_code, const String &p_uniforms, const String &p_compute_globals, const Vector<String> &p_custom_defines) { + ERR_FAIL_COND(!is_compute); + + Version *version = version_owner.get_or_null(p_version); + ERR_FAIL_COND(!version); + + version->compute_globals = p_compute_globals.utf8(); + version->uniforms = p_uniforms.utf8(); + + version->code_sections.clear(); + for (const KeyValue<String, String> &E : p_code) { + version->code_sections[StringName(E.key.to_upper())] = E.value.utf8(); + } + + version->custom_defines.clear(); + for (int i = 0; i < p_custom_defines.size(); i++) { + version->custom_defines.push_back(p_custom_defines[i].utf8()); + } + + version->dirty = true; + if (version->initialize_needed) { + _compile_version(version); + version->initialize_needed = false; + } +} + +bool ShaderRD::version_is_valid(RID p_version) { + Version *version = version_owner.get_or_null(p_version); + ERR_FAIL_COND_V(!version, false); + + if (version->dirty) { + _compile_version(version); + } + + return version->valid; +} + +bool ShaderRD::version_free(RID p_version) { + if (version_owner.owns(p_version)) { + Version *version = version_owner.get_or_null(p_version); + _clear_version(version); + version_owner.free(p_version); + } else { + return false; + } + + return true; +} + +void ShaderRD::set_variant_enabled(int p_variant, bool p_enabled) { + ERR_FAIL_COND(version_owner.get_rid_count() > 0); //versions exist + ERR_FAIL_INDEX(p_variant, variants_enabled.size()); + variants_enabled.write[p_variant] = p_enabled; +} + +bool ShaderRD::is_variant_enabled(int p_variant) const { + ERR_FAIL_INDEX_V(p_variant, variants_enabled.size(), false); + return variants_enabled[p_variant]; +} + +bool ShaderRD::shader_cache_cleanup_on_start = false; + +ShaderRD::ShaderRD() { + // Do not feel forced to use this, in most cases it makes little to no difference. + bool use_32_threads = false; + if (RD::get_singleton()->get_device_vendor_name() == "NVIDIA") { + use_32_threads = true; + } + String base_compute_define_text; + if (use_32_threads) { + base_compute_define_text = "\n#define NATIVE_LOCAL_GROUP_SIZE 32\n#define NATIVE_LOCAL_SIZE_2D_X 8\n#define NATIVE_LOCAL_SIZE_2D_Y 4\n"; + } else { + base_compute_define_text = "\n#define NATIVE_LOCAL_GROUP_SIZE 64\n#define NATIVE_LOCAL_SIZE_2D_X 8\n#define NATIVE_LOCAL_SIZE_2D_Y 8\n"; + } + + base_compute_defines = base_compute_define_text.ascii(); +} + +void ShaderRD::initialize(const Vector<String> &p_variant_defines, const String &p_general_defines) { + ERR_FAIL_COND(variant_defines.size()); + ERR_FAIL_COND(p_variant_defines.size() == 0); + + general_defines = p_general_defines.utf8(); + + for (int i = 0; i < p_variant_defines.size(); i++) { + variant_defines.push_back(p_variant_defines[i].utf8()); + variants_enabled.push_back(true); + } + + if (!shader_cache_dir.is_empty()) { + StringBuilder hash_build; + + hash_build.append("[base_hash]"); + hash_build.append(base_sha256); + hash_build.append("[general_defines]"); + hash_build.append(general_defines.get_data()); + for (int i = 0; i < variant_defines.size(); i++) { + hash_build.append("[variant_defines:" + itos(i) + "]"); + hash_build.append(variant_defines[i].get_data()); + } + + base_sha256 = hash_build.as_string().sha256_text(); + + Ref<DirAccess> d = DirAccess::open(shader_cache_dir); + ERR_FAIL_COND(d.is_null()); + if (d->change_dir(name) != OK) { + Error err = d->make_dir(name); + ERR_FAIL_COND(err != OK); + d->change_dir(name); + } + + //erase other versions? + if (shader_cache_cleanup_on_start) { + } + // + if (d->change_dir(base_sha256) != OK) { + Error err = d->make_dir(base_sha256); + ERR_FAIL_COND(err != OK); + } + shader_cache_dir_valid = true; + + print_verbose("Shader '" + name + "' SHA256: " + base_sha256); + } +} + +void ShaderRD::set_shader_cache_dir(const String &p_dir) { + shader_cache_dir = p_dir; +} + +void ShaderRD::set_shader_cache_save_compressed(bool p_enable) { + shader_cache_save_compressed = p_enable; +} + +void ShaderRD::set_shader_cache_save_compressed_zstd(bool p_enable) { + shader_cache_save_compressed_zstd = p_enable; +} + +void ShaderRD::set_shader_cache_save_debug(bool p_enable) { + shader_cache_save_debug = p_enable; +} + +String ShaderRD::shader_cache_dir; +bool ShaderRD::shader_cache_save_compressed = true; +bool ShaderRD::shader_cache_save_compressed_zstd = true; +bool ShaderRD::shader_cache_save_debug = true; + +ShaderRD::~ShaderRD() { + List<RID> remaining; + version_owner.get_owned_list(&remaining); + if (remaining.size()) { + ERR_PRINT(itos(remaining.size()) + " shaders of type " + name + " were never freed"); + while (remaining.size()) { + version_free(remaining.front()->get()); + remaining.pop_front(); + } + } +} diff --git a/servers/rendering/renderer_rd/shader_rd.h b/servers/rendering/renderer_rd/shader_rd.h new file mode 100644 index 0000000000..40b10808c2 --- /dev/null +++ b/servers/rendering/renderer_rd/shader_rd.h @@ -0,0 +1,171 @@ +/*************************************************************************/ +/* shader_rd.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef SHADER_RD_H +#define SHADER_RD_H + +#include "core/os/mutex.h" +#include "core/string/string_builder.h" +#include "core/templates/hash_map.h" +#include "core/templates/local_vector.h" +#include "core/templates/rb_map.h" +#include "core/templates/rid_owner.h" +#include "core/variant/variant.h" +#include "servers/rendering_server.h" + +class ShaderRD { + //versions + CharString general_defines; + Vector<CharString> variant_defines; + Vector<bool> variants_enabled; + + struct Version { + CharString uniforms; + CharString vertex_globals; + CharString compute_globals; + CharString fragment_globals; + HashMap<StringName, CharString> code_sections; + Vector<CharString> custom_defines; + + Vector<uint8_t> *variant_data = nullptr; + RID *variants = nullptr; //same size as version defines + + bool valid; + bool dirty; + bool initialize_needed; + }; + + Mutex variant_set_mutex; + + void _compile_variant(uint32_t p_variant, Version *p_version); + + void _clear_version(Version *p_version); + void _compile_version(Version *p_version); + + RID_Owner<Version> version_owner; + + struct StageTemplate { + struct Chunk { + enum Type { + TYPE_VERSION_DEFINES, + TYPE_MATERIAL_UNIFORMS, + TYPE_VERTEX_GLOBALS, + TYPE_FRAGMENT_GLOBALS, + TYPE_COMPUTE_GLOBALS, + TYPE_CODE, + TYPE_TEXT + }; + + Type type; + StringName code; + CharString text; + }; + LocalVector<Chunk> chunks; + }; + + bool is_compute = false; + + String name; + + CharString base_compute_defines; + + String base_sha256; + + static String shader_cache_dir; + static bool shader_cache_cleanup_on_start; + static bool shader_cache_save_compressed; + static bool shader_cache_save_compressed_zstd; + static bool shader_cache_save_debug; + bool shader_cache_dir_valid = false; + + enum StageType { + STAGE_TYPE_VERTEX, + STAGE_TYPE_FRAGMENT, + STAGE_TYPE_COMPUTE, + STAGE_TYPE_MAX, + }; + + StageTemplate stage_templates[STAGE_TYPE_MAX]; + + void _build_variant_code(StringBuilder &p_builder, uint32_t p_variant, const Version *p_version, const StageTemplate &p_template); + + void _add_stage(const char *p_code, StageType p_stage_type); + + String _version_get_sha1(Version *p_version) const; + bool _load_from_cache(Version *p_version); + void _save_to_cache(Version *p_version); + +protected: + ShaderRD(); + void setup(const char *p_vertex_code, const char *p_fragment_code, const char *p_compute_code, const char *p_name); + +public: + RID version_create(); + + void version_set_code(RID p_version, const HashMap<String, String> &p_code, const String &p_uniforms, const String &p_vertex_globals, const String &p_fragment_globals, const Vector<String> &p_custom_defines); + void version_set_compute_code(RID p_version, const HashMap<String, String> &p_code, const String &p_uniforms, const String &p_compute_globals, const Vector<String> &p_custom_defines); + + _FORCE_INLINE_ RID version_get_shader(RID p_version, int p_variant) { + ERR_FAIL_INDEX_V(p_variant, variant_defines.size(), RID()); + ERR_FAIL_COND_V(!variants_enabled[p_variant], RID()); + + Version *version = version_owner.get_or_null(p_version); + ERR_FAIL_COND_V(!version, RID()); + + if (version->dirty) { + _compile_version(version); + } + + if (!version->valid) { + return RID(); + } + + return version->variants[p_variant]; + } + + bool version_is_valid(RID p_version); + + bool version_free(RID p_version); + + void set_variant_enabled(int p_variant, bool p_enabled); + bool is_variant_enabled(int p_variant) const; + + static void set_shader_cache_dir(const String &p_dir); + static void set_shader_cache_save_compressed(bool p_enable); + static void set_shader_cache_save_compressed_zstd(bool p_enable); + static void set_shader_cache_save_debug(bool p_enable); + + RS::ShaderNativeSourceCode version_get_native_source_code(RID p_version); + + void initialize(const Vector<String> &p_variant_defines, const String &p_general_defines = ""); + virtual ~ShaderRD(); +}; + +#endif // SHADER_RD_H diff --git a/servers/rendering/renderer_rd/shaders/SCsub b/servers/rendering/renderer_rd/shaders/SCsub new file mode 100644 index 0000000000..5405985741 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/SCsub @@ -0,0 +1,22 @@ +#!/usr/bin/env python + +Import("env") + +if "RD_GLSL" in env["BUILDERS"]: + # find all include files + gl_include_files = [str(f) for f in Glob("*_inc.glsl")] + + # find all shader code(all glsl files excluding our include files) + glsl_files = [str(f) for f in Glob("*.glsl") if str(f) not in gl_include_files] + + # make sure we recompile shaders if include files change + env.Depends([f + ".gen.h" for f in glsl_files], gl_include_files + ["#glsl_builders.py"]) + + # compile shaders + for glsl_file in glsl_files: + env.RD_GLSL(glsl_file) + +SConscript("effects/SCsub") +SConscript("environment/SCsub") +SConscript("forward_clustered/SCsub") +SConscript("forward_mobile/SCsub") diff --git a/servers/rendering/renderer_rd/shaders/blit.glsl b/servers/rendering/renderer_rd/shaders/blit.glsl new file mode 100644 index 0000000000..14f190a49f --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/blit.glsl @@ -0,0 +1,97 @@ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +layout(push_constant, std140) uniform Pos { + vec4 src_rect; + vec4 dst_rect; + + vec2 eye_center; + float k1; + float k2; + + float upscale; + float aspect_ratio; + uint layer; + uint pad1; +} +data; + +layout(location = 0) out vec2 uv; + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv = data.src_rect.xy + base_arr[gl_VertexIndex] * data.src_rect.zw; + vec2 vtx = data.dst_rect.xy + base_arr[gl_VertexIndex] * data.dst_rect.zw; + gl_Position = vec4(vtx * 2.0 - 1.0, 0.0, 1.0); +} + +#[fragment] + +#version 450 + +#VERSION_DEFINES + +layout(push_constant, std140) uniform Pos { + vec4 src_rect; + vec4 dst_rect; + + vec2 eye_center; + float k1; + float k2; + + float upscale; + float aspect_ratio; + uint layer; + uint pad1; +} +data; + +layout(location = 0) in vec2 uv; + +layout(location = 0) out vec4 color; + +#ifdef USE_LAYER +layout(binding = 0) uniform sampler2DArray src_rt; +#else +layout(binding = 0) uniform sampler2D src_rt; +#endif + +void main() { +#ifdef APPLY_LENS_DISTORTION + vec2 coords = uv * 2.0 - 1.0; + vec2 offset = coords - data.eye_center; + + // take aspect ratio into account + offset.y /= data.aspect_ratio; + + // distort + vec2 offset_sq = offset * offset; + float radius_sq = offset_sq.x + offset_sq.y; + float radius_s4 = radius_sq * radius_sq; + float distortion_scale = 1.0 + (data.k1 * radius_sq) + (data.k2 * radius_s4); + offset *= distortion_scale; + + // reapply aspect ratio + offset.y *= data.aspect_ratio; + + // add our eye center back in + coords = offset + data.eye_center; + coords /= data.upscale; + + // and check our color + if (coords.x < -1.0 || coords.y < -1.0 || coords.x > 1.0 || coords.y > 1.0) { + color = vec4(0.0, 0.0, 0.0, 1.0); + } else { + // layer is always used here + coords = (coords + vec2(1.0)) / vec2(2.0); + color = texture(src_rt, vec3(coords, data.layer)); + } +#elif defined(USE_LAYER) + color = texture(src_rt, vec3(uv, data.layer)); +#else + color = texture(src_rt, uv); +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/canvas.glsl b/servers/rendering/renderer_rd/shaders/canvas.glsl new file mode 100644 index 0000000000..8593e6b265 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/canvas.glsl @@ -0,0 +1,738 @@ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#ifdef USE_ATTRIBUTES +layout(location = 0) in vec2 vertex_attrib; +layout(location = 3) in vec4 color_attrib; +layout(location = 4) in vec2 uv_attrib; + +layout(location = 10) in uvec4 bone_attrib; +layout(location = 11) in vec4 weight_attrib; + +#endif + +#include "canvas_uniforms_inc.glsl" + +layout(location = 0) out vec2 uv_interp; +layout(location = 1) out vec4 color_interp; +layout(location = 2) out vec2 vertex_interp; + +#ifdef USE_NINEPATCH + +layout(location = 3) out vec2 pixel_size_interp; + +#endif + +#ifdef MATERIAL_UNIFORMS_USED +layout(set = 1, binding = 0, std140) uniform MaterialUniforms{ + +#MATERIAL_UNIFORMS + +} material; +#endif + +#GLOBALS + +void main() { + vec4 instance_custom = vec4(0.0); +#ifdef USE_PRIMITIVE + + //weird bug, + //this works + vec2 vertex; + vec2 uv; + vec4 color; + + if (gl_VertexIndex == 0) { + vertex = draw_data.points[0]; + uv = draw_data.uvs[0]; + color = vec4(unpackHalf2x16(draw_data.colors[0]), unpackHalf2x16(draw_data.colors[1])); + } else if (gl_VertexIndex == 1) { + vertex = draw_data.points[1]; + uv = draw_data.uvs[1]; + color = vec4(unpackHalf2x16(draw_data.colors[2]), unpackHalf2x16(draw_data.colors[3])); + } else { + vertex = draw_data.points[2]; + uv = draw_data.uvs[2]; + color = vec4(unpackHalf2x16(draw_data.colors[4]), unpackHalf2x16(draw_data.colors[5])); + } + uvec4 bones = uvec4(0, 0, 0, 0); + vec4 bone_weights = vec4(0.0); + +#elif defined(USE_ATTRIBUTES) + + vec2 vertex = vertex_attrib; + vec4 color = color_attrib * draw_data.modulation; + vec2 uv = uv_attrib; + + uvec4 bones = bone_attrib; + vec4 bone_weights = weight_attrib; +#else + + vec2 vertex_base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + vec2 vertex_base = vertex_base_arr[gl_VertexIndex]; + + vec2 uv = draw_data.src_rect.xy + abs(draw_data.src_rect.zw) * ((draw_data.flags & FLAGS_TRANSPOSE_RECT) != 0 ? vertex_base.yx : vertex_base.xy); + vec4 color = draw_data.modulation; + vec2 vertex = draw_data.dst_rect.xy + abs(draw_data.dst_rect.zw) * mix(vertex_base, vec2(1.0, 1.0) - vertex_base, lessThan(draw_data.src_rect.zw, vec2(0.0, 0.0))); + uvec4 bones = uvec4(0, 0, 0, 0); + +#endif + + mat4 model_matrix = mat4(vec4(draw_data.world_x, 0.0, 0.0), vec4(draw_data.world_y, 0.0, 0.0), vec4(0.0, 0.0, 1.0, 0.0), vec4(draw_data.world_ofs, 0.0, 1.0)); + +#define FLAGS_INSTANCING_MASK 0x7F +#define FLAGS_INSTANCING_HAS_COLORS (1 << 7) +#define FLAGS_INSTANCING_HAS_CUSTOM_DATA (1 << 8) + + uint instancing = draw_data.flags & FLAGS_INSTANCING_MASK; + +#ifdef USE_ATTRIBUTES + if (instancing > 1) { + // trails + + uint stride = 2 + 1 + 1; //particles always uses this format + + uint trail_size = instancing; + + uint offset = trail_size * stride * gl_InstanceIndex; + + vec4 pcolor; + vec2 new_vertex; + { + uint boffset = offset + bone_attrib.x * stride; + new_vertex = (vec4(vertex, 0.0, 1.0) * mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0))).xy * weight_attrib.x; + pcolor = transforms.data[boffset + 2] * weight_attrib.x; + } + if (weight_attrib.y > 0.001) { + uint boffset = offset + bone_attrib.y * stride; + new_vertex += (vec4(vertex, 0.0, 1.0) * mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0))).xy * weight_attrib.y; + pcolor += transforms.data[boffset + 2] * weight_attrib.y; + } + if (weight_attrib.z > 0.001) { + uint boffset = offset + bone_attrib.z * stride; + new_vertex += (vec4(vertex, 0.0, 1.0) * mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0))).xy * weight_attrib.z; + pcolor += transforms.data[boffset + 2] * weight_attrib.z; + } + if (weight_attrib.w > 0.001) { + uint boffset = offset + bone_attrib.w * stride; + new_vertex += (vec4(vertex, 0.0, 1.0) * mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0))).xy * weight_attrib.w; + pcolor += transforms.data[boffset + 2] * weight_attrib.w; + } + + instance_custom = transforms.data[offset + 3]; + + vertex = new_vertex; + color *= pcolor; + } else +#endif // USE_ATTRIBUTES + { + if (instancing == 1) { + uint stride = 2; + { + if (bool(draw_data.flags & FLAGS_INSTANCING_HAS_COLORS)) { + stride += 1; + } + if (bool(draw_data.flags & FLAGS_INSTANCING_HAS_CUSTOM_DATA)) { + stride += 1; + } + } + + uint offset = stride * gl_InstanceIndex; + + mat4 matrix = mat4(transforms.data[offset + 0], transforms.data[offset + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0)); + offset += 2; + + if (bool(draw_data.flags & FLAGS_INSTANCING_HAS_COLORS)) { + color *= transforms.data[offset]; + offset += 1; + } + + if (bool(draw_data.flags & FLAGS_INSTANCING_HAS_CUSTOM_DATA)) { + instance_custom = transforms.data[offset]; + } + + matrix = transpose(matrix); + model_matrix = model_matrix * matrix; + } + } + +#if !defined(USE_ATTRIBUTES) && !defined(USE_PRIMITIVE) + if (bool(draw_data.flags & FLAGS_USING_PARTICLES)) { + //scale by texture size + vertex /= draw_data.color_texture_pixel_size; + } +#endif + +#ifdef USE_POINT_SIZE + float point_size = 1.0; +#endif + { +#CODE : VERTEX + } + +#ifdef USE_NINEPATCH + pixel_size_interp = abs(draw_data.dst_rect.zw) * vertex_base; +#endif + +#if !defined(SKIP_TRANSFORM_USED) + vertex = (model_matrix * vec4(vertex, 0.0, 1.0)).xy; +#endif + + color_interp = color; + + if (canvas_data.use_pixel_snap) { + vertex = floor(vertex + 0.5); + // precision issue on some hardware creates artifacts within texture + // offset uv by a small amount to avoid + uv += 1e-5; + } + +#ifdef USE_ATTRIBUTES +#if 0 + if (bool(draw_data.flags & FLAGS_USE_SKELETON) && bone_weights != vec4(0.0)) { //must be a valid bone + //skeleton transform + ivec4 bone_indicesi = ivec4(bone_indices); + + uvec2 tex_ofs = bone_indicesi.x * 2; + + mat2x4 m; + m = mat2x4( + texelFetch(skeleton_buffer, tex_ofs + 0), + texelFetch(skeleton_buffer, tex_ofs + 1)) * + bone_weights.x; + + tex_ofs = bone_indicesi.y * 2; + + m += mat2x4( + texelFetch(skeleton_buffer, tex_ofs + 0), + texelFetch(skeleton_buffer, tex_ofs + 1)) * + bone_weights.y; + + tex_ofs = bone_indicesi.z * 2; + + m += mat2x4( + texelFetch(skeleton_buffer, tex_ofs + 0), + texelFetch(skeleton_buffer, tex_ofs + 1)) * + bone_weights.z; + + tex_ofs = bone_indicesi.w * 2; + + m += mat2x4( + texelFetch(skeleton_buffer, tex_ofs + 0), + texelFetch(skeleton_buffer, tex_ofs + 1)) * + bone_weights.w; + + mat4 bone_matrix = skeleton_data.skeleton_transform * transpose(mat4(m[0], m[1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0))) * skeleton_data.skeleton_transform_inverse; + + //outvec = bone_matrix * outvec; + } +#endif +#endif + + vertex = (canvas_data.canvas_transform * vec4(vertex, 0.0, 1.0)).xy; + + vertex_interp = vertex; + uv_interp = uv; + + gl_Position = canvas_data.screen_transform * vec4(vertex, 0.0, 1.0); + +#ifdef USE_POINT_SIZE + gl_PointSize = point_size; +#endif +} + +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#include "canvas_uniforms_inc.glsl" + +layout(location = 0) in vec2 uv_interp; +layout(location = 1) in vec4 color_interp; +layout(location = 2) in vec2 vertex_interp; + +#ifdef USE_NINEPATCH + +layout(location = 3) in vec2 pixel_size_interp; + +#endif + +layout(location = 0) out vec4 frag_color; + +#ifdef MATERIAL_UNIFORMS_USED +layout(set = 1, binding = 0, std140) uniform MaterialUniforms{ + +#MATERIAL_UNIFORMS + +} material; +#endif + +vec2 screen_uv_to_sdf(vec2 p_uv) { + return canvas_data.screen_to_sdf * p_uv; +} + +float texture_sdf(vec2 p_sdf) { + vec2 uv = p_sdf * canvas_data.sdf_to_tex.xy + canvas_data.sdf_to_tex.zw; + float d = texture(sampler2D(sdf_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uv).r; + d *= SDF_MAX_LENGTH; + return d * canvas_data.tex_to_sdf; +} + +vec2 texture_sdf_normal(vec2 p_sdf) { + vec2 uv = p_sdf * canvas_data.sdf_to_tex.xy + canvas_data.sdf_to_tex.zw; + + const float EPSILON = 0.001; + return normalize(vec2( + texture(sampler2D(sdf_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uv + vec2(EPSILON, 0.0)).r - texture(sampler2D(sdf_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uv - vec2(EPSILON, 0.0)).r, + texture(sampler2D(sdf_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uv + vec2(0.0, EPSILON)).r - texture(sampler2D(sdf_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uv - vec2(0.0, EPSILON)).r)); +} + +vec2 sdf_to_screen_uv(vec2 p_sdf) { + return p_sdf * canvas_data.sdf_to_screen; +} + +#GLOBALS + +#ifdef LIGHT_CODE_USED + +vec4 light_compute( + vec3 light_vertex, + vec3 light_position, + vec3 normal, + vec4 light_color, + float light_energy, + vec4 specular_shininess, + inout vec4 shadow_modulate, + vec2 screen_uv, + vec2 uv, + vec4 color, bool is_directional) { + vec4 light = vec4(0.0); + vec3 light_direction = vec3(0.0); + + if (is_directional) { + light_direction = normalize(mix(vec3(light_position.xy, 0.0), vec3(0, 0, 1), light_position.z)); + light_position = vec3(0.0); + } else { + light_direction = normalize(light_position - light_vertex); + } + +#CODE : LIGHT + + return light; +} + +#endif + +#ifdef USE_NINEPATCH + +float map_ninepatch_axis(float pixel, float draw_size, float tex_pixel_size, float margin_begin, float margin_end, int np_repeat, inout int draw_center) { + float tex_size = 1.0 / tex_pixel_size; + + if (pixel < margin_begin) { + return pixel * tex_pixel_size; + } else if (pixel >= draw_size - margin_end) { + return (tex_size - (draw_size - pixel)) * tex_pixel_size; + } else { + if (!bool(draw_data.flags & FLAGS_NINEPACH_DRAW_CENTER)) { + draw_center--; + } + + // np_repeat is passed as uniform using NinePatchRect::AxisStretchMode enum. + if (np_repeat == 0) { // Stretch. + // Convert to ratio. + float ratio = (pixel - margin_begin) / (draw_size - margin_begin - margin_end); + // Scale to source texture. + return (margin_begin + ratio * (tex_size - margin_begin - margin_end)) * tex_pixel_size; + } else if (np_repeat == 1) { // Tile. + // Convert to offset. + float ofs = mod((pixel - margin_begin), tex_size - margin_begin - margin_end); + // Scale to source texture. + return (margin_begin + ofs) * tex_pixel_size; + } else if (np_repeat == 2) { // Tile Fit. + // Calculate scale. + float src_area = draw_size - margin_begin - margin_end; + float dst_area = tex_size - margin_begin - margin_end; + float scale = max(1.0, floor(src_area / max(dst_area, 0.0000001) + 0.5)); + // Convert to ratio. + float ratio = (pixel - margin_begin) / src_area; + ratio = mod(ratio * scale, 1.0); + // Scale to source texture. + return (margin_begin + ratio * dst_area) * tex_pixel_size; + } else { // Shouldn't happen, but silences compiler warning. + return 0.0; + } + } +} + +#endif + +#ifdef USE_LIGHTING + +vec3 light_normal_compute(vec3 light_vec, vec3 normal, vec3 base_color, vec3 light_color, vec4 specular_shininess, bool specular_shininess_used) { + float cNdotL = max(0.0, dot(normal, light_vec)); + + if (specular_shininess_used) { + //blinn + vec3 view = vec3(0.0, 0.0, 1.0); // not great but good enough + vec3 half_vec = normalize(view + light_vec); + + float cNdotV = max(dot(normal, view), 0.0); + float cNdotH = max(dot(normal, half_vec), 0.0); + float cVdotH = max(dot(view, half_vec), 0.0); + float cLdotH = max(dot(light_vec, half_vec), 0.0); + float shininess = exp2(15.0 * specular_shininess.a + 1.0) * 0.25; + float blinn = pow(cNdotH, shininess); + blinn *= (shininess + 8.0) * (1.0 / (8.0 * M_PI)); + float s = (blinn) / max(4.0 * cNdotV * cNdotL, 0.75); + + return specular_shininess.rgb * light_color * s + light_color * base_color * cNdotL; + } else { + return light_color * base_color * cNdotL; + } +} + +//float distance = length(shadow_pos); +vec4 light_shadow_compute(uint light_base, vec4 light_color, vec4 shadow_uv +#ifdef LIGHT_CODE_USED + , + vec3 shadow_modulate +#endif +) { + float shadow; + uint shadow_mode = light_array.data[light_base].flags & LIGHT_FLAGS_FILTER_MASK; + + if (shadow_mode == LIGHT_FLAGS_SHADOW_NEAREST) { + shadow = textureProjLod(sampler2DShadow(shadow_atlas_texture, shadow_sampler), shadow_uv, 0.0).x; + } else if (shadow_mode == LIGHT_FLAGS_SHADOW_PCF5) { + vec4 shadow_pixel_size = vec4(light_array.data[light_base].shadow_pixel_size, 0.0, 0.0, 0.0); + shadow = 0.0; + shadow += textureProjLod(sampler2DShadow(shadow_atlas_texture, shadow_sampler), shadow_uv - shadow_pixel_size * 2.0, 0.0).x; + shadow += textureProjLod(sampler2DShadow(shadow_atlas_texture, shadow_sampler), shadow_uv - shadow_pixel_size, 0.0).x; + shadow += textureProjLod(sampler2DShadow(shadow_atlas_texture, shadow_sampler), shadow_uv, 0.0).x; + shadow += textureProjLod(sampler2DShadow(shadow_atlas_texture, shadow_sampler), shadow_uv + shadow_pixel_size, 0.0).x; + shadow += textureProjLod(sampler2DShadow(shadow_atlas_texture, shadow_sampler), shadow_uv + shadow_pixel_size * 2.0, 0.0).x; + shadow /= 5.0; + } else { //PCF13 + vec4 shadow_pixel_size = vec4(light_array.data[light_base].shadow_pixel_size, 0.0, 0.0, 0.0); + shadow = 0.0; + shadow += textureProjLod(sampler2DShadow(shadow_atlas_texture, shadow_sampler), shadow_uv - shadow_pixel_size * 6.0, 0.0).x; + shadow += textureProjLod(sampler2DShadow(shadow_atlas_texture, shadow_sampler), shadow_uv - shadow_pixel_size * 5.0, 0.0).x; + shadow += textureProjLod(sampler2DShadow(shadow_atlas_texture, shadow_sampler), shadow_uv - shadow_pixel_size * 4.0, 0.0).x; + shadow += textureProjLod(sampler2DShadow(shadow_atlas_texture, shadow_sampler), shadow_uv - shadow_pixel_size * 3.0, 0.0).x; + shadow += textureProjLod(sampler2DShadow(shadow_atlas_texture, shadow_sampler), shadow_uv - shadow_pixel_size * 2.0, 0.0).x; + shadow += textureProjLod(sampler2DShadow(shadow_atlas_texture, shadow_sampler), shadow_uv - shadow_pixel_size, 0.0).x; + shadow += textureProjLod(sampler2DShadow(shadow_atlas_texture, shadow_sampler), shadow_uv, 0.0).x; + shadow += textureProjLod(sampler2DShadow(shadow_atlas_texture, shadow_sampler), shadow_uv + shadow_pixel_size, 0.0).x; + shadow += textureProjLod(sampler2DShadow(shadow_atlas_texture, shadow_sampler), shadow_uv + shadow_pixel_size * 2.0, 0.0).x; + shadow += textureProjLod(sampler2DShadow(shadow_atlas_texture, shadow_sampler), shadow_uv + shadow_pixel_size * 3.0, 0.0).x; + shadow += textureProjLod(sampler2DShadow(shadow_atlas_texture, shadow_sampler), shadow_uv + shadow_pixel_size * 4.0, 0.0).x; + shadow += textureProjLod(sampler2DShadow(shadow_atlas_texture, shadow_sampler), shadow_uv + shadow_pixel_size * 5.0, 0.0).x; + shadow += textureProjLod(sampler2DShadow(shadow_atlas_texture, shadow_sampler), shadow_uv + shadow_pixel_size * 6.0, 0.0).x; + shadow /= 13.0; + } + + vec4 shadow_color = unpackUnorm4x8(light_array.data[light_base].shadow_color); +#ifdef LIGHT_CODE_USED + shadow_color.rgb *= shadow_modulate; +#endif + + shadow_color.a *= light_color.a; //respect light alpha + + return mix(light_color, shadow_color, shadow); +} + +void light_blend_compute(uint light_base, vec4 light_color, inout vec3 color) { + uint blend_mode = light_array.data[light_base].flags & LIGHT_FLAGS_BLEND_MASK; + + switch (blend_mode) { + case LIGHT_FLAGS_BLEND_MODE_ADD: { + color.rgb += light_color.rgb * light_color.a; + } break; + case LIGHT_FLAGS_BLEND_MODE_SUB: { + color.rgb -= light_color.rgb * light_color.a; + } break; + case LIGHT_FLAGS_BLEND_MODE_MIX: { + color.rgb = mix(color.rgb, light_color.rgb, light_color.a); + } break; + } +} + +#endif + +float msdf_median(float r, float g, float b, float a) { + return min(max(min(r, g), min(max(r, g), b)), a); +} + +void main() { + vec4 color = color_interp; + vec2 uv = uv_interp; + vec2 vertex = vertex_interp; + +#if !defined(USE_ATTRIBUTES) && !defined(USE_PRIMITIVE) + +#ifdef USE_NINEPATCH + + int draw_center = 2; + uv = vec2( + map_ninepatch_axis(pixel_size_interp.x, abs(draw_data.dst_rect.z), draw_data.color_texture_pixel_size.x, draw_data.ninepatch_margins.x, draw_data.ninepatch_margins.z, int(draw_data.flags >> FLAGS_NINEPATCH_H_MODE_SHIFT) & 0x3, draw_center), + map_ninepatch_axis(pixel_size_interp.y, abs(draw_data.dst_rect.w), draw_data.color_texture_pixel_size.y, draw_data.ninepatch_margins.y, draw_data.ninepatch_margins.w, int(draw_data.flags >> FLAGS_NINEPATCH_V_MODE_SHIFT) & 0x3, draw_center)); + + if (draw_center == 0) { + color.a = 0.0; + } + + uv = uv * draw_data.src_rect.zw + draw_data.src_rect.xy; //apply region if needed + +#endif + if (bool(draw_data.flags & FLAGS_CLIP_RECT_UV)) { + uv = clamp(uv, draw_data.src_rect.xy, draw_data.src_rect.xy + abs(draw_data.src_rect.zw)); + } + +#endif + +#ifndef USE_PRIMITIVE + if (bool(draw_data.flags & FLAGS_USE_MSDF)) { + float px_range = draw_data.ninepatch_margins.x; + float outline_thickness = draw_data.ninepatch_margins.y; + //float reserved1 = draw_data.ninepatch_margins.z; + //float reserved2 = draw_data.ninepatch_margins.w; + + vec4 msdf_sample = texture(sampler2D(color_texture, texture_sampler), uv); + vec2 msdf_size = vec2(textureSize(sampler2D(color_texture, texture_sampler), 0)); + vec2 dest_size = vec2(1.0) / fwidth(uv); + float px_size = max(0.5 * dot((vec2(px_range) / msdf_size), dest_size), 1.0); + float d = msdf_median(msdf_sample.r, msdf_sample.g, msdf_sample.b, msdf_sample.a) - 0.5; + + if (outline_thickness > 0) { + float cr = clamp(outline_thickness, 0.0, px_range / 2) / px_range; + float a = clamp((d + cr) * px_size, 0.0, 1.0); + color.a = a * color.a; + } else { + float a = clamp(d * px_size + 0.5, 0.0, 1.0); + color.a = a * color.a; + } + } else if (bool(draw_data.flags & FLAGS_USE_LCD)) { + vec4 lcd_sample = texture(sampler2D(color_texture, texture_sampler), uv); + if (lcd_sample.a == 1.0) { + color.rgb = lcd_sample.rgb * color.a; + } else { + color = vec4(0.0, 0.0, 0.0, 0.0); + } + } else { +#else + { +#endif + color *= texture(sampler2D(color_texture, texture_sampler), uv); + } + + uint light_count = (draw_data.flags >> FLAGS_LIGHT_COUNT_SHIFT) & 0xF; //max 16 lights + bool using_light = light_count > 0 || canvas_data.directional_light_count > 0; + + vec3 normal; + +#if defined(NORMAL_USED) + bool normal_used = true; +#else + bool normal_used = false; +#endif + + if (normal_used || (using_light && bool(draw_data.flags & FLAGS_DEFAULT_NORMAL_MAP_USED))) { + normal.xy = texture(sampler2D(normal_texture, texture_sampler), uv).xy * vec2(2.0, -2.0) - vec2(1.0, -1.0); + normal.z = sqrt(1.0 - dot(normal.xy, normal.xy)); + normal_used = true; + } else { + normal = vec3(0.0, 0.0, 1.0); + } + + vec4 specular_shininess; + +#if defined(SPECULAR_SHININESS_USED) + + bool specular_shininess_used = true; +#else + bool specular_shininess_used = false; +#endif + + if (specular_shininess_used || (using_light && normal_used && bool(draw_data.flags & FLAGS_DEFAULT_SPECULAR_MAP_USED))) { + specular_shininess = texture(sampler2D(specular_texture, texture_sampler), uv); + specular_shininess *= unpackUnorm4x8(draw_data.specular_shininess); + specular_shininess_used = true; + } else { + specular_shininess = vec4(1.0); + } + +#if defined(SCREEN_UV_USED) + vec2 screen_uv = gl_FragCoord.xy * canvas_data.screen_pixel_size; +#else + vec2 screen_uv = vec2(0.0); +#endif + + vec3 light_vertex = vec3(vertex, 0.0); + vec2 shadow_vertex = vertex; + + { + float normal_map_depth = 1.0; + +#if defined(NORMAL_MAP_USED) + vec3 normal_map = vec3(0.0, 0.0, 1.0); + normal_used = true; +#endif + +#CODE : FRAGMENT + +#if defined(NORMAL_MAP_USED) + normal = mix(vec3(0.0, 0.0, 1.0), normal_map * vec3(2.0, -2.0, 1.0) - vec3(1.0, -1.0, 0.0), normal_map_depth); +#endif + } + + if (normal_used) { + //convert by item transform + normal.xy = mat2(normalize(draw_data.world_x), normalize(draw_data.world_y)) * normal.xy; + //convert by canvas transform + normal = normalize((canvas_data.canvas_normal_transform * vec4(normal, 0.0)).xyz); + } + + vec4 base_color = color; + if (bool(draw_data.flags & FLAGS_USING_LIGHT_MASK)) { + color = vec4(0.0); //invisible by default due to using light mask + } + +#ifdef MODE_LIGHT_ONLY + color = vec4(0.0); +#elif !defined(MODE_UNSHADED) + color *= canvas_data.canvas_modulation; +#endif + +#if defined(USE_LIGHTING) && !defined(MODE_UNSHADED) + + // Directional Lights + + for (uint i = 0; i < canvas_data.directional_light_count; i++) { + uint light_base = i; + + vec2 direction = light_array.data[light_base].position; + vec4 light_color = light_array.data[light_base].color; + +#ifdef LIGHT_CODE_USED + + vec4 shadow_modulate = vec4(1.0); + light_color = light_compute(light_vertex, vec3(direction, light_array.data[light_base].height), normal, light_color, light_color.a, specular_shininess, shadow_modulate, screen_uv, uv, base_color, true); +#else + + if (normal_used) { + vec3 light_vec = normalize(mix(vec3(direction, 0.0), vec3(0, 0, 1), light_array.data[light_base].height)); + light_color.rgb = light_normal_compute(light_vec, normal, base_color.rgb, light_color.rgb, specular_shininess, specular_shininess_used); + } else { + light_color.rgb *= base_color.rgb; + } +#endif + + if (bool(light_array.data[light_base].flags & LIGHT_FLAGS_HAS_SHADOW)) { + vec2 shadow_pos = (vec4(shadow_vertex, 0.0, 1.0) * mat4(light_array.data[light_base].shadow_matrix[0], light_array.data[light_base].shadow_matrix[1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0))).xy; //multiply inverse given its transposed. Optimizer removes useless operations. + + vec4 shadow_uv = vec4(shadow_pos.x, light_array.data[light_base].shadow_y_ofs, shadow_pos.y * light_array.data[light_base].shadow_zfar_inv, 1.0); + + light_color = light_shadow_compute(light_base, light_color, shadow_uv +#ifdef LIGHT_CODE_USED + , + shadow_modulate.rgb +#endif + ); + } + + light_blend_compute(light_base, light_color, color.rgb); + } + + // Positional Lights + + for (uint i = 0; i < MAX_LIGHTS_PER_ITEM; i++) { + if (i >= light_count) { + break; + } + uint light_base = draw_data.lights[i >> 2]; + light_base >>= (i & 3) * 8; + light_base &= 0xFF; + + vec2 tex_uv = (vec4(vertex, 0.0, 1.0) * mat4(light_array.data[light_base].texture_matrix[0], light_array.data[light_base].texture_matrix[1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0))).xy; //multiply inverse given its transposed. Optimizer removes useless operations. + vec2 tex_uv_atlas = tex_uv * light_array.data[light_base].atlas_rect.zw + light_array.data[light_base].atlas_rect.xy; + vec4 light_color = textureLod(sampler2D(atlas_texture, texture_sampler), tex_uv_atlas, 0.0); + vec4 light_base_color = light_array.data[light_base].color; + +#ifdef LIGHT_CODE_USED + + vec4 shadow_modulate = vec4(1.0); + vec3 light_position = vec3(light_array.data[light_base].position, light_array.data[light_base].height); + + light_color.rgb *= light_base_color.rgb; + light_color = light_compute(light_vertex, light_position, normal, light_color, light_base_color.a, specular_shininess, shadow_modulate, screen_uv, uv, base_color, false); +#else + + light_color.rgb *= light_base_color.rgb * light_base_color.a; + + if (normal_used) { + vec3 light_pos = vec3(light_array.data[light_base].position, light_array.data[light_base].height); + vec3 pos = light_vertex; + vec3 light_vec = normalize(light_pos - pos); + + light_color.rgb = light_normal_compute(light_vec, normal, base_color.rgb, light_color.rgb, specular_shininess, specular_shininess_used); + } else { + light_color.rgb *= base_color.rgb; + } +#endif + if (any(lessThan(tex_uv, vec2(0.0, 0.0))) || any(greaterThanEqual(tex_uv, vec2(1.0, 1.0)))) { + //if outside the light texture, light color is zero + light_color.a = 0.0; + } + + if (bool(light_array.data[light_base].flags & LIGHT_FLAGS_HAS_SHADOW)) { + vec2 shadow_pos = (vec4(shadow_vertex, 0.0, 1.0) * mat4(light_array.data[light_base].shadow_matrix[0], light_array.data[light_base].shadow_matrix[1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0))).xy; //multiply inverse given its transposed. Optimizer removes useless operations. + + vec2 pos_norm = normalize(shadow_pos); + vec2 pos_abs = abs(pos_norm); + vec2 pos_box = pos_norm / max(pos_abs.x, pos_abs.y); + vec2 pos_rot = pos_norm * mat2(vec2(0.7071067811865476, -0.7071067811865476), vec2(0.7071067811865476, 0.7071067811865476)); //is there a faster way to 45 degrees rot? + float tex_ofs; + float distance; + if (pos_rot.y > 0) { + if (pos_rot.x > 0) { + tex_ofs = pos_box.y * 0.125 + 0.125; + distance = shadow_pos.x; + } else { + tex_ofs = pos_box.x * -0.125 + (0.25 + 0.125); + distance = shadow_pos.y; + } + } else { + if (pos_rot.x < 0) { + tex_ofs = pos_box.y * -0.125 + (0.5 + 0.125); + distance = -shadow_pos.x; + } else { + tex_ofs = pos_box.x * 0.125 + (0.75 + 0.125); + distance = -shadow_pos.y; + } + } + + distance *= light_array.data[light_base].shadow_zfar_inv; + + //float distance = length(shadow_pos); + vec4 shadow_uv = vec4(tex_ofs, light_array.data[light_base].shadow_y_ofs, distance, 1.0); + + light_color = light_shadow_compute(light_base, light_color, shadow_uv +#ifdef LIGHT_CODE_USED + , + shadow_modulate.rgb +#endif + ); + } + + light_blend_compute(light_base, light_color, color.rgb); + } +#endif + + frag_color = color; +} diff --git a/servers/rendering/renderer_rd/shaders/canvas_occlusion.glsl b/servers/rendering/renderer_rd/shaders/canvas_occlusion.glsl new file mode 100644 index 0000000000..930cf792cb --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/canvas_occlusion.glsl @@ -0,0 +1,59 @@ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +layout(location = 0) in highp vec3 vertex; + +layout(push_constant, std430) uniform Constants { + mat4 projection; + mat2x4 modelview; + vec2 direction; + float z_far; + float pad; +} +constants; + +#ifdef MODE_SHADOW +layout(location = 0) out highp float depth; +#endif + +void main() { + highp vec4 vtx = vec4(vertex, 1.0) * mat4(constants.modelview[0], constants.modelview[1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0)); + +#ifdef MODE_SHADOW + depth = dot(constants.direction, vtx.xy); +#endif + gl_Position = constants.projection * vtx; +} + +#[fragment] + +#version 450 + +#VERSION_DEFINES + +layout(push_constant, std430) uniform Constants { + mat4 projection; + mat2x4 modelview; + vec2 direction; + float z_far; + float pad; +} +constants; + +#ifdef MODE_SHADOW +layout(location = 0) in highp float depth; +layout(location = 0) out highp float distance_buf; +#else +layout(location = 0) out highp float sdf_buf; +#endif + +void main() { +#ifdef MODE_SHADOW + distance_buf = depth / constants.z_far; +#else + sdf_buf = 1.0; +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/canvas_sdf.glsl b/servers/rendering/renderer_rd/shaders/canvas_sdf.glsl new file mode 100644 index 0000000000..0fafc7d486 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/canvas_sdf.glsl @@ -0,0 +1,179 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +layout(r8, set = 0, binding = 1) uniform restrict readonly image2D src_pixels; +layout(r16_snorm, set = 0, binding = 2) uniform restrict writeonly image2D dst_sdf; + +layout(rg16i, set = 0, binding = 3) uniform restrict readonly iimage2D src_process; +layout(rg16i, set = 0, binding = 4) uniform restrict writeonly iimage2D dst_process; + +layout(push_constant, std430) uniform Params { + ivec2 size; + int stride; + int shift; + ivec2 base_size; + uvec2 pad; +} +params; + +#define SDF_MAX_LENGTH 16384.0 + +void main() { + ivec2 pos = ivec2(gl_GlobalInvocationID.xy); + if (any(greaterThanEqual(pos, params.size))) { //too large, do nothing + return; + } + +#ifdef MODE_LOAD + + bool solid = imageLoad(src_pixels, pos).r > 0.5; + imageStore(dst_process, pos, solid ? ivec4(ivec2(-32767), 0, 0) : ivec4(ivec2(32767), 0, 0)); +#endif + +#ifdef MODE_LOAD_SHRINK + + int s = 1 << params.shift; + ivec2 base = pos << params.shift; + ivec2 center = base + ivec2(params.shift); + + ivec2 rel = ivec2(32767); + float d = 1e20; + int found = 0; + int solid_found = 0; + for (int i = 0; i < s; i++) { + for (int j = 0; j < s; j++) { + ivec2 src_pos = base + ivec2(i, j); + if (any(greaterThanEqual(src_pos, params.base_size))) { + continue; + } + bool solid = imageLoad(src_pixels, src_pos).r > 0.5; + if (solid) { + float dist = length(vec2(src_pos - center)); + if (dist < d) { + d = dist; + rel = src_pos; + } + solid_found++; + } + found++; + } + } + + if (solid_found == found) { + //mark solid only if all are solid + rel = ivec2(-32767); + } + + imageStore(dst_process, pos, ivec4(rel, 0, 0)); +#endif + +#ifdef MODE_PROCESS + + ivec2 base = pos << params.shift; + ivec2 center = base + ivec2(params.shift); + + ivec2 rel = imageLoad(src_process, pos).xy; + + bool solid = rel.x < 0; + + if (solid) { + rel = -rel - ivec2(1); + } + + if (center != rel) { + //only process if it does not point to itself + const int ofs_table_size = 8; + const ivec2 ofs_table[ofs_table_size] = ivec2[]( + ivec2(-1, -1), + ivec2(0, -1), + ivec2(+1, -1), + + ivec2(-1, 0), + ivec2(+1, 0), + + ivec2(-1, +1), + ivec2(0, +1), + ivec2(+1, +1)); + + float dist = length(vec2(rel - center)); + for (int i = 0; i < ofs_table_size; i++) { + ivec2 src_pos = pos + ofs_table[i] * params.stride; + if (any(lessThan(src_pos, ivec2(0))) || any(greaterThanEqual(src_pos, params.size))) { + continue; + } + ivec2 src_rel = imageLoad(src_process, src_pos).xy; + bool src_solid = src_rel.x < 0; + if (src_solid) { + src_rel = -src_rel - ivec2(1); + } + + if (src_solid != solid) { + src_rel = ivec2(src_pos << params.shift); //point to itself if of different type + } + + float src_dist = length(vec2(src_rel - center)); + if (src_dist < dist) { + dist = src_dist; + rel = src_rel; + } + } + } + + if (solid) { + rel = -rel - ivec2(1); + } + + imageStore(dst_process, pos, ivec4(rel, 0, 0)); +#endif + +#ifdef MODE_STORE + + ivec2 rel = imageLoad(src_process, pos).xy; + + bool solid = rel.x < 0; + + if (solid) { + rel = -rel - ivec2(1); + } + + float d = length(vec2(rel - pos)); + + if (solid) { + d = -d; + } + + d /= SDF_MAX_LENGTH; + d = clamp(d, -1.0, 1.0); + imageStore(dst_sdf, pos, vec4(d)); + +#endif + +#ifdef MODE_STORE_SHRINK + + ivec2 base = pos << params.shift; + ivec2 center = base + ivec2(params.shift); + + ivec2 rel = imageLoad(src_process, pos).xy; + + bool solid = rel.x < 0; + + if (solid) { + rel = -rel - ivec2(1); + } + + float d = length(vec2(rel - center)); + + if (solid) { + d = -d; + } + d /= SDF_MAX_LENGTH; + d = clamp(d, -1.0, 1.0); + imageStore(dst_sdf, pos, vec4(d)); + +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl b/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl new file mode 100644 index 0000000000..1b627a3e81 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl @@ -0,0 +1,163 @@ + +#define MAX_LIGHTS_PER_ITEM 16 + +#define M_PI 3.14159265359 + +#define SDF_MAX_LENGTH 16384.0 + +//1 means enabled, 2+ means trails in use +#define FLAGS_INSTANCING_MASK 0x7F +#define FLAGS_INSTANCING_HAS_COLORS (1 << 7) +#define FLAGS_INSTANCING_HAS_CUSTOM_DATA (1 << 8) + +#define FLAGS_CLIP_RECT_UV (1 << 9) +#define FLAGS_TRANSPOSE_RECT (1 << 10) +#define FLAGS_USING_LIGHT_MASK (1 << 11) +#define FLAGS_NINEPACH_DRAW_CENTER (1 << 12) +#define FLAGS_USING_PARTICLES (1 << 13) + +#define FLAGS_NINEPATCH_H_MODE_SHIFT 16 +#define FLAGS_NINEPATCH_V_MODE_SHIFT 18 + +#define FLAGS_LIGHT_COUNT_SHIFT 20 + +#define FLAGS_DEFAULT_NORMAL_MAP_USED (1 << 26) +#define FLAGS_DEFAULT_SPECULAR_MAP_USED (1 << 27) + +#define FLAGS_USE_MSDF (1 << 28) +#define FLAGS_USE_LCD (1 << 29) + +#define SAMPLER_NEAREST_CLAMP 0 +#define SAMPLER_LINEAR_CLAMP 1 +#define SAMPLER_NEAREST_WITH_MIPMAPS_CLAMP 2 +#define SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP 3 +#define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_CLAMP 4 +#define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_CLAMP 5 +#define SAMPLER_NEAREST_REPEAT 6 +#define SAMPLER_LINEAR_REPEAT 7 +#define SAMPLER_NEAREST_WITH_MIPMAPS_REPEAT 8 +#define SAMPLER_LINEAR_WITH_MIPMAPS_REPEAT 9 +#define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_REPEAT 10 +#define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_REPEAT 11 + +// Push Constant + +layout(push_constant, std430) uniform DrawData { + vec2 world_x; + vec2 world_y; + vec2 world_ofs; + uint flags; + uint specular_shininess; +#ifdef USE_PRIMITIVE + vec2 points[3]; + vec2 uvs[3]; + uint colors[6]; +#else + vec4 modulation; + vec4 ninepatch_margins; + vec4 dst_rect; //for built-in rect and UV + vec4 src_rect; + vec2 pad; + +#endif + vec2 color_texture_pixel_size; + uint lights[4]; +} +draw_data; + +// In vulkan, sets should always be ordered using the following logic: +// Lower Sets: Sets that change format and layout less often +// Higher sets: Sets that change format and layout very often +// This is because changing a set for another with a different layout or format, +// invalidates all the upper ones (as likely internal base offset changes) + +/* SET0: Globals */ + +// The values passed per draw primitives are cached within it + +layout(set = 0, binding = 1, std140) uniform CanvasData { + mat4 canvas_transform; + mat4 screen_transform; + mat4 canvas_normal_transform; + vec4 canvas_modulation; + vec2 screen_pixel_size; + float time; + bool use_pixel_snap; + + vec4 sdf_to_tex; + vec2 screen_to_sdf; + vec2 sdf_to_screen; + + uint directional_light_count; + float tex_to_sdf; + uint pad1; + uint pad2; +} +canvas_data; + +#define LIGHT_FLAGS_BLEND_MASK (3 << 16) +#define LIGHT_FLAGS_BLEND_MODE_ADD (0 << 16) +#define LIGHT_FLAGS_BLEND_MODE_SUB (1 << 16) +#define LIGHT_FLAGS_BLEND_MODE_MIX (2 << 16) +#define LIGHT_FLAGS_BLEND_MODE_MASK (3 << 16) +#define LIGHT_FLAGS_HAS_SHADOW (1 << 20) +#define LIGHT_FLAGS_FILTER_SHIFT 22 +#define LIGHT_FLAGS_FILTER_MASK (3 << 22) +#define LIGHT_FLAGS_SHADOW_NEAREST (0 << 22) +#define LIGHT_FLAGS_SHADOW_PCF5 (1 << 22) +#define LIGHT_FLAGS_SHADOW_PCF13 (2 << 22) + +struct Light { + mat2x4 texture_matrix; //light to texture coordinate matrix (transposed) + mat2x4 shadow_matrix; //light to shadow coordinate matrix (transposed) + vec4 color; + + uint shadow_color; // packed + uint flags; //index to light texture + float shadow_pixel_size; + float height; + + vec2 position; + float shadow_zfar_inv; + float shadow_y_ofs; + + vec4 atlas_rect; +}; + +layout(set = 0, binding = 2, std140) uniform LightData { + Light data[MAX_LIGHTS]; +} +light_array; + +layout(set = 0, binding = 3) uniform texture2D atlas_texture; +layout(set = 0, binding = 4) uniform texture2D shadow_atlas_texture; + +layout(set = 0, binding = 5) uniform sampler shadow_sampler; + +layout(set = 0, binding = 6) uniform texture2D screen_texture; +layout(set = 0, binding = 7) uniform texture2D sdf_texture; + +layout(set = 0, binding = 8) uniform sampler material_samplers[12]; + +layout(set = 0, binding = 9, std430) restrict readonly buffer GlobalShaderUniformData { + vec4 data[]; +} +global_shader_uniforms; + +/* SET1: Is reserved for the material */ + +// + +/* SET2: Instancing and Skeleton */ + +layout(set = 2, binding = 0, std430) restrict readonly buffer Transforms { + vec4 data[]; +} +transforms; + +/* SET3: Texture */ + +layout(set = 3, binding = 0) uniform texture2D color_texture; +layout(set = 3, binding = 1) uniform texture2D normal_texture; +layout(set = 3, binding = 2) uniform texture2D specular_texture; +layout(set = 3, binding = 3) uniform sampler texture_sampler; diff --git a/servers/rendering/renderer_rd/shaders/cluster_data_inc.glsl b/servers/rendering/renderer_rd/shaders/cluster_data_inc.glsl new file mode 100644 index 0000000000..8e616ebe1f --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cluster_data_inc.glsl @@ -0,0 +1,3 @@ +#define CLUSTER_COUNTER_SHIFT 20 +#define CLUSTER_POINTER_MASK ((1 << CLUSTER_COUNTER_SHIFT) - 1) +#define CLUSTER_COUNTER_MASK 0xfff diff --git a/servers/rendering/renderer_rd/shaders/cluster_debug.glsl b/servers/rendering/renderer_rd/shaders/cluster_debug.glsl new file mode 100644 index 0000000000..0034de8c91 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cluster_debug.glsl @@ -0,0 +1,115 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +const vec3 usage_gradient[33] = vec3[]( // 1 (none) + 32 + vec3(0.14, 0.17, 0.23), + vec3(0.24, 0.44, 0.83), + vec3(0.23, 0.57, 0.84), + vec3(0.22, 0.71, 0.84), + vec3(0.22, 0.85, 0.83), + vec3(0.21, 0.85, 0.72), + vec3(0.21, 0.85, 0.57), + vec3(0.20, 0.85, 0.42), + vec3(0.20, 0.85, 0.27), + vec3(0.27, 0.86, 0.19), + vec3(0.51, 0.85, 0.19), + vec3(0.57, 0.86, 0.19), + vec3(0.62, 0.85, 0.19), + vec3(0.67, 0.86, 0.20), + vec3(0.73, 0.85, 0.20), + vec3(0.78, 0.85, 0.20), + vec3(0.83, 0.85, 0.20), + vec3(0.85, 0.82, 0.20), + vec3(0.85, 0.76, 0.20), + vec3(0.85, 0.81, 0.20), + vec3(0.85, 0.65, 0.20), + vec3(0.84, 0.60, 0.21), + vec3(0.84, 0.56, 0.21), + vec3(0.84, 0.51, 0.21), + vec3(0.84, 0.46, 0.21), + vec3(0.84, 0.41, 0.21), + vec3(0.84, 0.36, 0.21), + vec3(0.84, 0.31, 0.21), + vec3(0.84, 0.27, 0.21), + vec3(0.83, 0.22, 0.22), + vec3(0.83, 0.22, 0.27), + vec3(0.83, 0.22, 0.32), + vec3(1.00, 0.63, 0.70)); +layout(push_constant, std430) uniform Params { + uvec2 screen_size; + uvec2 cluster_screen_size; + + uint cluster_shift; + uint cluster_type; + float z_near; + float z_far; + + bool orthogonal; + uint max_cluster_element_count_div_32; + uint pad1; + uint pad2; +} +params; + +layout(set = 0, binding = 1, std430) buffer restrict readonly ClusterData { + uint data[]; +} +cluster_data; + +layout(rgba16f, set = 0, binding = 2) uniform restrict writeonly image2D screen_buffer; +layout(set = 0, binding = 3) uniform texture2D depth_buffer; +layout(set = 0, binding = 4) uniform sampler depth_buffer_sampler; + +void main() { + uvec2 screen_pos = gl_GlobalInvocationID.xy; + if (any(greaterThanEqual(screen_pos, params.screen_size))) { + return; + } + + uvec2 cluster_pos = screen_pos >> params.cluster_shift; + + uint offset = cluster_pos.y * params.cluster_screen_size.x + cluster_pos.x; + offset += params.cluster_screen_size.x * params.cluster_screen_size.y * params.cluster_type; + offset *= (params.max_cluster_element_count_div_32 + 32); + + //depth buffers generally can't be accessed via image API + float depth = texelFetch(sampler2D(depth_buffer, depth_buffer_sampler), ivec2(screen_pos), 0).r * 2.0 - 1.0; + + if (params.orthogonal) { + depth = ((depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0; + } else { + depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - depth * (params.z_far - params.z_near)); + } + depth /= params.z_far; + + uint slice = uint(clamp(floor(depth * 32.0), 0.0, 31.0)); + uint slice_minmax = cluster_data.data[offset + params.max_cluster_element_count_div_32 + slice]; + uint item_min = slice_minmax & 0xFFFF; + uint item_max = slice_minmax >> 16; + + uint item_count = 0; + for (uint i = 0; i < params.max_cluster_element_count_div_32; i++) { + uint slice_bits = cluster_data.data[offset + i]; + while (slice_bits != 0) { + uint bit = findLSB(slice_bits); + uint item = i * 32 + bit; + if ((item >= item_min && item < item_max)) { + item_count++; + } + slice_bits &= ~(1 << bit); + } + } + + item_count = min(item_count, 32); + + vec3 color = usage_gradient[item_count]; + + color = mix(color * 1.2, color * 0.3, float(slice) / 31.0); + + imageStore(screen_buffer, ivec2(screen_pos), vec4(color, 1.0)); +} diff --git a/servers/rendering/renderer_rd/shaders/cluster_render.glsl b/servers/rendering/renderer_rd/shaders/cluster_render.glsl new file mode 100644 index 0000000000..932312de82 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cluster_render.glsl @@ -0,0 +1,169 @@ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +layout(location = 0) in vec3 vertex_attrib; + +layout(location = 0) out float depth_interp; +layout(location = 1) out flat uint element_index; + +layout(push_constant, std430) uniform Params { + uint base_index; + uint pad0; + uint pad1; + uint pad2; +} +params; + +layout(set = 0, binding = 1, std140) uniform State { + mat4 projection; + + float inv_z_far; + uint screen_to_clusters_shift; // shift to obtain coordinates in block indices + uint cluster_screen_width; // + uint cluster_data_size; // how much data for a single cluster takes + + uint cluster_depth_offset; + uint pad0; + uint pad1; + uint pad2; +} +state; + +struct RenderElement { + uint type; //0-4 + bool touches_near; + bool touches_far; + uint original_index; + mat3x4 transform_inv; + vec3 scale; + uint pad; +}; + +layout(set = 0, binding = 2, std430) buffer restrict readonly RenderElements { + RenderElement data[]; +} +render_elements; + +void main() { + element_index = params.base_index + gl_InstanceIndex; + + vec3 vertex = vertex_attrib; + vertex *= render_elements.data[element_index].scale; + + vertex = vec4(vertex, 1.0) * render_elements.data[element_index].transform_inv; + depth_interp = -vertex.z; + + gl_Position = state.projection * vec4(vertex, 1.0); +} + +#[fragment] + +#version 450 + +#VERSION_DEFINES +#ifndef MOLTENVK_USED // Metal will corrupt GPU state otherwise +#if defined(has_GL_KHR_shader_subgroup_ballot) && defined(has_GL_KHR_shader_subgroup_arithmetic) && defined(has_GL_KHR_shader_subgroup_vote) + +#extension GL_KHR_shader_subgroup_ballot : enable +#extension GL_KHR_shader_subgroup_arithmetic : enable +#extension GL_KHR_shader_subgroup_vote : enable + +#define USE_SUBGROUPS +#endif +#endif + +layout(location = 0) in float depth_interp; +layout(location = 1) in flat uint element_index; + +layout(set = 0, binding = 1, std140) uniform State { + mat4 projection; + float inv_z_far; + uint screen_to_clusters_shift; // shift to obtain coordinates in block indices + uint cluster_screen_width; // + uint cluster_data_size; // how much data for a single cluster takes + uint cluster_depth_offset; + uint pad0; + uint pad1; + uint pad2; +} +state; + +//cluster data is layout linearly, each cell contains the follow information: +// - list of bits for every element to mark as used, so (max_elem_count/32)*4 uints +// - a uint for each element to mark the depth bits used when rendering (0-31) + +layout(set = 0, binding = 3, std430) buffer restrict ClusterRender { + uint data[]; +} +cluster_render; + +void main() { + //convert from screen to cluster + uvec2 cluster = uvec2(gl_FragCoord.xy) >> state.screen_to_clusters_shift; + + //get linear cluster offset from screen poss + uint cluster_offset = cluster.x + state.cluster_screen_width * cluster.y; + //multiply by data size to position at the beginning of the element list for this cluster + cluster_offset *= state.cluster_data_size; + + //find the current element in the list and plot the bit to mark it as used + uint usage_write_offset = cluster_offset + (element_index >> 5); + uint usage_write_bit = 1 << (element_index & 0x1F); + +#ifdef USE_SUBGROUPS + + uint cluster_thread_group_index; + + if (!gl_HelperInvocation) { + //https://advances.realtimerendering.com/s2017/2017_Sig_Improved_Culling_final.pdf + + uvec4 mask; + + while (true) { + // find the cluster offset of the first active thread + // threads that did break; go inactive and no longer count + uint first = subgroupBroadcastFirst(cluster_offset); + // update the mask for thread that match this cluster + mask = subgroupBallot(first == cluster_offset); + if (first == cluster_offset) { + // This thread belongs to the group of threads that match this offset, + // so exit the loop. + break; + } + } + + cluster_thread_group_index = subgroupBallotExclusiveBitCount(mask); + + if (cluster_thread_group_index == 0) { + atomicOr(cluster_render.data[usage_write_offset], usage_write_bit); + } + } +#else + if (!gl_HelperInvocation) { + atomicOr(cluster_render.data[usage_write_offset], usage_write_bit); + } +#endif + //find the current element in the depth usage list and mark the current depth as used + float unit_depth = depth_interp * state.inv_z_far; + + uint z_bit = clamp(uint(floor(unit_depth * 32.0)), 0, 31); + + uint z_write_offset = cluster_offset + state.cluster_depth_offset + element_index; + uint z_write_bit = 1 << z_bit; + +#ifdef USE_SUBGROUPS + if (!gl_HelperInvocation) { + z_write_bit = subgroupOr(z_write_bit); //merge all Zs + if (cluster_thread_group_index == 0) { + atomicOr(cluster_render.data[z_write_offset], z_write_bit); + } + } +#else + if (!gl_HelperInvocation) { + atomicOr(cluster_render.data[z_write_offset], z_write_bit); + } +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/cluster_store.glsl b/servers/rendering/renderer_rd/shaders/cluster_store.glsl new file mode 100644 index 0000000000..64a145f3c6 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cluster_store.glsl @@ -0,0 +1,119 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +layout(push_constant, std430) uniform Params { + uint cluster_render_data_size; // how much data for a single cluster takes + uint max_render_element_count_div_32; //divided by 32 + uvec2 cluster_screen_size; + uint render_element_count_div_32; //divided by 32 + + uint max_cluster_element_count_div_32; //divided by 32 + uint pad1; + uint pad2; +} +params; + +layout(set = 0, binding = 1, std430) buffer restrict readonly ClusterRender { + uint data[]; +} +cluster_render; + +layout(set = 0, binding = 2, std430) buffer restrict ClusterStore { + uint data[]; +} +cluster_store; + +struct RenderElement { + uint type; //0-4 + bool touches_near; + bool touches_far; + uint original_index; + mat3x4 transform_inv; + vec3 scale; + uint pad; +}; + +layout(set = 0, binding = 3, std430) buffer restrict readonly RenderElements { + RenderElement data[]; +} +render_elements; + +void main() { + uvec2 pos = gl_GlobalInvocationID.xy; + if (any(greaterThanEqual(pos, params.cluster_screen_size))) { + return; + } + + //counter for each type of render_element + + //base offset for this cluster + uint base_offset = (pos.x + params.cluster_screen_size.x * pos.y); + uint src_offset = base_offset * params.cluster_render_data_size; + + uint render_element_offset = 0; + + //check all render_elements and see which one was written to + while (render_element_offset < params.render_element_count_div_32) { + uint bits = cluster_render.data[src_offset + render_element_offset]; + while (bits != 0) { + //if bits exist, check the render_element + uint index_bit = findLSB(bits); + uint index = render_element_offset * 32 + index_bit; + uint type = render_elements.data[index].type; + + uint z_range_offset = src_offset + params.max_render_element_count_div_32 + index; + uint z_range = cluster_render.data[z_range_offset]; + + //if object was written, z was written, but check just in case + if (z_range != 0) { //should always be > 0 + + uint from_z = findLSB(z_range); + uint to_z = findMSB(z_range) + 1; + + if (render_elements.data[index].touches_near) { + from_z = 0; + } + + if (render_elements.data[index].touches_far) { + to_z = 32; + } + + // find cluster offset in the buffer used for indexing in the renderer + uint dst_offset = (base_offset + type * (params.cluster_screen_size.x * params.cluster_screen_size.y)) * (params.max_cluster_element_count_div_32 + 32); + + uint orig_index = render_elements.data[index].original_index; + //store this index in the Z slices by setting the relevant bit + for (uint i = from_z; i < to_z; i++) { + uint slice_ofs = dst_offset + params.max_cluster_element_count_div_32 + i; + + uint minmax = cluster_store.data[slice_ofs]; + + if (minmax == 0) { + minmax = 0xFFFF; //min 0, max 0xFFFF + } + + uint elem_min = min(orig_index, minmax & 0xFFFF); + uint elem_max = max(orig_index + 1, minmax >> 16); //always store plus one, so zero means range is empty when not written to + + minmax = elem_min | (elem_max << 16); + cluster_store.data[slice_ofs] = minmax; + } + + uint store_word = orig_index >> 5; + uint store_bit = orig_index & 0x1F; + + //store the actual render_element index at the end, so the rendering code can reference it + cluster_store.data[dst_offset + store_word] |= 1 << store_bit; + } + + bits &= ~(1 << index_bit); //clear the bit to continue iterating + } + + render_element_offset++; + } +} diff --git a/servers/rendering/renderer_rd/shaders/decal_data_inc.glsl b/servers/rendering/renderer_rd/shaders/decal_data_inc.glsl new file mode 100644 index 0000000000..158096d3c7 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/decal_data_inc.glsl @@ -0,0 +1,18 @@ + +struct DecalData { + highp mat4 xform; //to decal transform + highp vec3 inv_extents; + mediump float albedo_mix; + highp vec4 albedo_rect; + highp vec4 normal_rect; + highp vec4 orm_rect; + highp vec4 emission_rect; + highp vec4 modulate; + mediump float emission_energy; + uint mask; + mediump float upper_fade; + mediump float lower_fade; + mediump mat3x4 normal_xform; + mediump vec3 normal; + mediump float normal_fade; +}; diff --git a/servers/rendering/renderer_rd/shaders/effects/SCsub b/servers/rendering/renderer_rd/shaders/effects/SCsub new file mode 100644 index 0000000000..f06a2d86e2 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/SCsub @@ -0,0 +1,17 @@ +#!/usr/bin/env python + +Import("env") + +if "RD_GLSL" in env["BUILDERS"]: + # find all include files + gl_include_files = [str(f) for f in Glob("*_inc.glsl")] + [str(f) for f in Glob("../*_inc.glsl")] + + # find all shader code(all glsl files excluding our include files) + glsl_files = [str(f) for f in Glob("*.glsl") if str(f) not in gl_include_files] + + # make sure we recompile shaders if include files change + env.Depends([f + ".gen.h" for f in glsl_files], gl_include_files + ["#glsl_builders.py"]) + + # compile shaders + for glsl_file in glsl_files: + env.RD_GLSL(glsl_file) diff --git a/servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl b/servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl new file mode 100644 index 0000000000..cb06250cf2 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl @@ -0,0 +1,148 @@ +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#include "blur_raster_inc.glsl" + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + + gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#include "blur_raster_inc.glsl" + +layout(location = 0) in vec2 uv_interp; +/* clang-format on */ + +layout(set = 0, binding = 0) uniform sampler2D source_color; + +#ifdef GLOW_USE_AUTO_EXPOSURE +layout(set = 1, binding = 0) uniform sampler2D source_auto_exposure; +#endif + +layout(location = 0) out vec4 frag_color; + +void main() { + // We do not apply our color scale for our mobile renderer here, we'll leave our colors at half brightness and apply scale in the tonemap raster. + +#ifdef MODE_MIPMAP + + vec2 pix_size = blur.pixel_size; + vec4 color = texture(source_color, uv_interp + vec2(-0.5, -0.5) * pix_size); + color += texture(source_color, uv_interp + vec2(0.5, -0.5) * pix_size); + color += texture(source_color, uv_interp + vec2(0.5, 0.5) * pix_size); + color += texture(source_color, uv_interp + vec2(-0.5, 0.5) * pix_size); + frag_color = color / 4.0; + +#endif + +#ifdef MODE_GAUSSIAN_BLUR + + // Simpler blur uses SIGMA2 for the gaussian kernel for a stronger effect + + // note, for blur blur.luminance_multiplier is irrelavant, we would be multiplying and then dividing by this amount. + + if (bool(blur.flags & FLAG_HORIZONTAL)) { + vec2 pix_size = blur.pixel_size; + pix_size *= 0.5; //reading from larger buffer, so use more samples + vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.214607; + color += texture(source_color, uv_interp + vec2(1.0, 0.0) * pix_size) * 0.189879; + color += texture(source_color, uv_interp + vec2(2.0, 0.0) * pix_size) * 0.131514; + color += texture(source_color, uv_interp + vec2(3.0, 0.0) * pix_size) * 0.071303; + color += texture(source_color, uv_interp + vec2(-1.0, 0.0) * pix_size) * 0.189879; + color += texture(source_color, uv_interp + vec2(-2.0, 0.0) * pix_size) * 0.131514; + color += texture(source_color, uv_interp + vec2(-3.0, 0.0) * pix_size) * 0.071303; + frag_color = color; + } else { + vec2 pix_size = blur.pixel_size; + vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.38774; + color += texture(source_color, uv_interp + vec2(0.0, 1.0) * pix_size) * 0.24477; + color += texture(source_color, uv_interp + vec2(0.0, 2.0) * pix_size) * 0.06136; + color += texture(source_color, uv_interp + vec2(0.0, -1.0) * pix_size) * 0.24477; + color += texture(source_color, uv_interp + vec2(0.0, -2.0) * pix_size) * 0.06136; + frag_color = color; + } +#endif + +#ifdef MODE_GAUSSIAN_GLOW + + //Glow uses larger sigma 1 for a more rounded blur effect + +#define GLOW_ADD(m_ofs, m_mult) \ + { \ + vec2 ofs = uv_interp + m_ofs * pix_size; \ + vec4 c = texture(source_color, ofs) * m_mult; \ + if (any(lessThan(ofs, vec2(0.0))) || any(greaterThan(ofs, vec2(1.0)))) { \ + c *= 0.0; \ + } \ + color += c; \ + } + + if (bool(blur.flags & FLAG_HORIZONTAL)) { + vec2 pix_size = blur.pixel_size; + pix_size *= 0.5; //reading from larger buffer, so use more samples + + vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.174938; + GLOW_ADD(vec2(1.0, 0.0), 0.165569); + GLOW_ADD(vec2(2.0, 0.0), 0.140367); + GLOW_ADD(vec2(3.0, 0.0), 0.106595); + GLOW_ADD(vec2(-1.0, 0.0), 0.165569); + GLOW_ADD(vec2(-2.0, 0.0), 0.140367); + GLOW_ADD(vec2(-3.0, 0.0), 0.106595); + + // only do this in the horizontal pass, if we also do this in the vertical pass we're doubling up. + color *= blur.glow_strength; + + frag_color = color; + } else { + vec2 pix_size = blur.pixel_size; + vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.288713; + GLOW_ADD(vec2(0.0, 1.0), 0.233062); + GLOW_ADD(vec2(0.0, 2.0), 0.122581); + GLOW_ADD(vec2(0.0, -1.0), 0.233062); + GLOW_ADD(vec2(0.0, -2.0), 0.122581); + + frag_color = color; + } + +#undef GLOW_ADD + + if (bool(blur.flags & FLAG_GLOW_FIRST_PASS)) { + // In the first pass bring back to correct color range else we're applying the wrong threshold + // in subsequent passes we can use it as is as we'd just be undoing it right after. + frag_color *= blur.luminance_multiplier; + +#ifdef GLOW_USE_AUTO_EXPOSURE + + frag_color /= texelFetch(source_auto_exposure, ivec2(0, 0), 0).r / blur.glow_auto_exposure_scale; +#endif + frag_color *= blur.glow_exposure; + + float luminance = max(frag_color.r, max(frag_color.g, frag_color.b)); + float feedback = max(smoothstep(blur.glow_hdr_threshold, blur.glow_hdr_threshold + blur.glow_hdr_scale, luminance), blur.glow_bloom); + + frag_color = min(frag_color * feedback, vec4(blur.glow_luminance_cap)) / blur.luminance_multiplier; + } + +#endif // MODE_GAUSSIAN_GLOW + +#ifdef MODE_COPY + vec4 color = textureLod(source_color, uv_interp, 0.0); + frag_color = color; +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/effects/blur_raster_inc.glsl b/servers/rendering/renderer_rd/shaders/effects/blur_raster_inc.glsl new file mode 100644 index 0000000000..06ca198f37 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/blur_raster_inc.glsl @@ -0,0 +1,26 @@ +#define FLAG_HORIZONTAL (1 << 0) +#define FLAG_USE_ORTHOGONAL_PROJECTION (1 << 1) +#define FLAG_GLOW_FIRST_PASS (1 << 2) + +layout(push_constant, std430) uniform Blur { + vec2 pixel_size; // 08 - 08 + uint flags; // 04 - 12 + uint pad; // 04 - 16 + + // Glow. + float glow_strength; // 04 - 20 + float glow_bloom; // 04 - 24 + float glow_hdr_threshold; // 04 - 28 + float glow_hdr_scale; // 04 - 32 + + float glow_exposure; // 04 - 36 + float glow_white; // 04 - 40 + float glow_luminance_cap; // 04 - 44 + float glow_auto_exposure_scale; // 04 - 48 + + float luminance_multiplier; // 04 - 52 + float res1; // 04 - 56 + float res2; // 04 - 60 + float res3; // 04 - 64 +} +blur; diff --git a/servers/rendering/renderer_rd/shaders/effects/bokeh_dof.glsl b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof.glsl new file mode 100644 index 0000000000..fe770ac065 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof.glsl @@ -0,0 +1,230 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +#define BLOCK_SIZE 8 + +layout(local_size_x = BLOCK_SIZE, local_size_y = BLOCK_SIZE, local_size_z = 1) in; + +#ifdef MODE_GEN_BLUR_SIZE +layout(rgba16f, set = 0, binding = 0) uniform restrict image2D color_image; +layout(set = 1, binding = 0) uniform sampler2D source_depth; +#endif + +#if defined(MODE_BOKEH_BOX) || defined(MODE_BOKEH_HEXAGONAL) || defined(MODE_BOKEH_CIRCULAR) +layout(set = 1, binding = 0) uniform sampler2D color_texture; +layout(rgba16f, set = 0, binding = 0) uniform restrict writeonly image2D bokeh_image; +#endif + +#ifdef MODE_COMPOSITE_BOKEH +layout(rgba16f, set = 0, binding = 0) uniform restrict image2D color_image; +layout(set = 1, binding = 0) uniform sampler2D source_bokeh; +#endif + +// based on https://www.shadertoy.com/view/Xd3GDl + +#include "bokeh_dof_inc.glsl" + +#ifdef MODE_GEN_BLUR_SIZE + +float get_depth_at_pos(vec2 uv) { + float depth = textureLod(source_depth, uv, 0.0).x * 2.0 - 1.0; + if (params.orthogonal) { + depth = ((depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0; + } else { + depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - depth * (params.z_far - params.z_near)); + } + return depth; +} + +float get_blur_size(float depth) { + if (params.blur_near_active && depth < params.blur_near_begin) { + if (params.use_physical_near) { + // Physically-based. + float d = abs(params.blur_near_begin - depth); + return -(d / (params.blur_near_begin - d)) * params.blur_size_near - DEPTH_GAP; // Near blur is negative. + } else { + // Non-physically-based. + return -(1.0 - smoothstep(params.blur_near_end, params.blur_near_begin, depth)) * params.blur_size - DEPTH_GAP; // Near blur is negative. + } + } + + if (params.blur_far_active && depth > params.blur_far_begin) { + if (params.use_physical_far) { + // Physically-based. + float d = abs(params.blur_far_begin - depth); + return (d / (params.blur_far_begin + d)) * params.blur_size_far + DEPTH_GAP; + } else { + // Non-physically-based. + return smoothstep(params.blur_far_begin, params.blur_far_end, depth) * params.blur_size + DEPTH_GAP; + } + } + + return 0.0; +} + +#endif + +#if defined(MODE_BOKEH_BOX) || defined(MODE_BOKEH_HEXAGONAL) + +vec4 weighted_filter_dir(vec2 dir, vec2 uv, vec2 pixel_size) { + dir *= pixel_size; + vec4 color = texture(color_texture, uv); + + vec4 accum = color; + float total = 1.0; + + float blur_scale = params.blur_size / float(params.blur_steps); + + if (params.use_jitter) { + uv += dir * (hash12n(uv + params.jitter_seed) - 0.5); + } + + for (int i = -params.blur_steps; i <= params.blur_steps; i++) { + if (i == 0) { + continue; + } + float radius = float(i) * blur_scale; + vec2 suv = uv + dir * radius; + radius = abs(radius); + + vec4 sample_color = texture(color_texture, suv); + float limit; + + if (sample_color.a < color.a) { + limit = abs(sample_color.a); + } else { + limit = abs(color.a); + } + + limit -= DEPTH_GAP; + + float m = smoothstep(radius - 0.5, radius + 0.5, limit); + + accum += mix(color, sample_color, m); + + total += 1.0; + } + + return accum / total; +} + +#endif + +void main() { + ivec2 pos = ivec2(gl_GlobalInvocationID.xy); + + if (any(greaterThan(pos, params.size))) { //too large, do nothing + return; + } + + vec2 pixel_size = 1.0 / vec2(params.size); + vec2 uv = vec2(pos) / vec2(params.size); + +#ifdef MODE_GEN_BLUR_SIZE + uv += pixel_size * 0.5; + //precompute size in alpha channel + float depth = get_depth_at_pos(uv); + float size = get_blur_size(depth); + + vec4 color = imageLoad(color_image, pos); + color.a = size; + imageStore(color_image, pos, color); +#endif + +#ifdef MODE_BOKEH_BOX + + //pixel_size*=0.5; //resolution is doubled + if (params.second_pass || !params.half_size) { + uv += pixel_size * 0.5; //half pixel to read centers + } else { + uv += pixel_size * 0.25; //half pixel to read centers from full res + } + + vec2 dir = (params.second_pass ? vec2(0.0, 1.0) : vec2(1.0, 0.0)); + + vec4 color = weighted_filter_dir(dir, uv, pixel_size); + + imageStore(bokeh_image, pos, color); + +#endif + +#ifdef MODE_BOKEH_HEXAGONAL + + //pixel_size*=0.5; //resolution is doubled + if (params.second_pass || !params.half_size) { + uv += pixel_size * 0.5; //half pixel to read centers + } else { + uv += pixel_size * 0.25; //half pixel to read centers from full res + } + + vec2 dir = (params.second_pass ? normalize(vec2(1.0, 0.577350269189626)) : vec2(0.0, 1.0)); + + vec4 color = weighted_filter_dir(dir, uv, pixel_size); + + if (params.second_pass) { + dir = normalize(vec2(-1.0, 0.577350269189626)); + + vec4 color2 = weighted_filter_dir(dir, uv, pixel_size); + + color.rgb = min(color.rgb, color2.rgb); + color.a = (color.a + color2.a) * 0.5; + } + + imageStore(bokeh_image, pos, color); + +#endif + +#ifdef MODE_BOKEH_CIRCULAR + + if (params.half_size) { + pixel_size *= 0.5; //resolution is doubled + } + + uv += pixel_size * 0.5; //half pixel to read centers + + vec4 color = texture(color_texture, uv); + float initial_blur = color.a; + float accum = 1.0; + float radius = params.blur_scale; + + for (float ang = 0.0; radius < params.blur_size; ang += GOLDEN_ANGLE) { + vec2 suv = uv + vec2(cos(ang), sin(ang)) * pixel_size * radius; + vec4 sample_color = texture(color_texture, suv); + float sample_size = abs(sample_color.a); + if (sample_color.a > initial_blur) { + sample_size = clamp(sample_size, 0.0, abs(initial_blur) * 2.0); + } + + float m = smoothstep(radius - 0.5, radius + 0.5, sample_size); + color += mix(color / accum, sample_color, m); + accum += 1.0; + radius += params.blur_scale / radius; + } + + color /= accum; + + imageStore(bokeh_image, pos, color); +#endif + +#ifdef MODE_COMPOSITE_BOKEH + + uv += pixel_size * 0.5; + vec4 color = imageLoad(color_image, pos); + vec4 bokeh = texture(source_bokeh, uv); + + float mix_amount; + if (bokeh.a < color.a) { + mix_amount = min(1.0, max(0.0, max(abs(color.a), abs(bokeh.a)) - DEPTH_GAP)); + } else { + mix_amount = min(1.0, max(0.0, abs(color.a) - DEPTH_GAP)); + } + + color.rgb = mix(color.rgb, bokeh.rgb, mix_amount); //blend between hires and lowres + + color.a = 0; //reset alpha + imageStore(color_image, pos, color); +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_inc.glsl b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_inc.glsl new file mode 100644 index 0000000000..4a2b0edc18 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_inc.glsl @@ -0,0 +1,42 @@ +layout(push_constant, std430) uniform Params { + ivec2 size; + float z_far; + float z_near; + + bool orthogonal; + float blur_size; + float blur_scale; + int blur_steps; + + bool blur_near_active; + float blur_near_begin; + float blur_near_end; + bool blur_far_active; + + float blur_far_begin; + float blur_far_end; + bool second_pass; + bool half_size; + + bool use_jitter; + float jitter_seed; + bool use_physical_near; + bool use_physical_far; + + float blur_size_near; + float blur_size_far; + uint pad[2]; +} +params; + +//used to work around downsampling filter +#define DEPTH_GAP 0.0 + +const float GOLDEN_ANGLE = 2.39996323; + +//note: uniform pdf rand [0;1[ +float hash12n(vec2 p) { + p = fract(p * vec2(5.3987, 5.4421)); + p += dot(p.yx, p.xy + vec2(21.5351, 14.3137)); + return fract(p.x * p.y * 95.4307); +} diff --git a/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_raster.glsl b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_raster.glsl new file mode 100644 index 0000000000..1b487835d2 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_raster.glsl @@ -0,0 +1,264 @@ +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#include "bokeh_dof_inc.glsl" + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + + gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#include "bokeh_dof_inc.glsl" + +layout(location = 0) in vec2 uv_interp; +/* clang-format on */ + +#ifdef MODE_GEN_BLUR_SIZE +layout(location = 0) out float weight; + +layout(set = 0, binding = 0) uniform sampler2D source_depth; +#else +layout(location = 0) out vec4 frag_color; +#ifdef OUTPUT_WEIGHT +layout(location = 1) out float weight; +#endif + +layout(set = 0, binding = 0) uniform sampler2D source_color; +layout(set = 1, binding = 0) uniform sampler2D source_weight; +#ifdef MODE_COMPOSITE_BOKEH +layout(set = 2, binding = 0) uniform sampler2D original_weight; +#endif +#endif + +//DOF +// Bokeh single pass implementation based on https://tuxedolabs.blogspot.com/2018/05/bokeh-depth-of-field-in-single-pass.html + +#ifdef MODE_GEN_BLUR_SIZE + +float get_depth_at_pos(vec2 uv) { + float depth = textureLod(source_depth, uv, 0.0).x * 2.0 - 1.0; + if (params.orthogonal) { + depth = ((depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0; + } else { + depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - depth * (params.z_far - params.z_near)); + } + return depth; +} + +float get_blur_size(float depth) { + if (params.blur_near_active && depth < params.blur_near_begin) { + if (params.use_physical_near) { + // Physically-based. + float d = abs(params.blur_near_begin - depth); + return -(d / (params.blur_near_begin - d)) * params.blur_size_near - DEPTH_GAP; // Near blur is negative. + } else { + // Non-physically-based. + return -(1.0 - smoothstep(params.blur_near_end, params.blur_near_begin, depth)) * params.blur_size - DEPTH_GAP; // Near blur is negative. + } + } + + if (params.blur_far_active && depth > params.blur_far_begin) { + if (params.use_physical_far) { + // Physically-based. + float d = abs(params.blur_far_begin - depth); + return (d / (params.blur_far_begin + d)) * params.blur_size_far + DEPTH_GAP; + } else { + // Non-physically-based. + return smoothstep(params.blur_far_begin, params.blur_far_end, depth) * params.blur_size + DEPTH_GAP; + } + } + + return 0.0; +} + +#endif + +#if defined(MODE_BOKEH_BOX) || defined(MODE_BOKEH_HEXAGONAL) + +vec4 weighted_filter_dir(vec2 dir, vec2 uv, vec2 pixel_size) { + dir *= pixel_size; + vec4 color = texture(source_color, uv); + color.a = texture(source_weight, uv).r; + + vec4 accum = color; + float total = 1.0; + + float blur_scale = params.blur_size / float(params.blur_steps); + + if (params.use_jitter) { + uv += dir * (hash12n(uv + params.jitter_seed) - 0.5); + } + + for (int i = -params.blur_steps; i <= params.blur_steps; i++) { + if (i == 0) { + continue; + } + float radius = float(i) * blur_scale; + vec2 suv = uv + dir * radius; + radius = abs(radius); + + vec4 sample_color = texture(source_color, suv); + sample_color.a = texture(source_weight, suv).r; + float limit; + + if (sample_color.a < color.a) { + limit = abs(sample_color.a); + } else { + limit = abs(color.a); + } + + limit -= DEPTH_GAP; + + float m = smoothstep(radius - 0.5, radius + 0.5, limit); + + accum += mix(color, sample_color, m); + + total += 1.0; + } + + return accum / total; +} + +#endif + +void main() { + vec2 pixel_size = 1.0 / vec2(params.size); + vec2 uv = uv_interp; + +#ifdef MODE_GEN_BLUR_SIZE + uv += pixel_size * 0.5; + float center_depth = get_depth_at_pos(uv); + weight = get_blur_size(center_depth); +#endif + +#ifdef MODE_BOKEH_BOX + //pixel_size*=0.5; //resolution is doubled + if (params.second_pass || !params.half_size) { + uv += pixel_size * 0.5; //half pixel to read centers + } else { + uv += pixel_size * 0.25; //half pixel to read centers from full res + } + + float alpha = texture(source_color, uv).a; // retain this + vec2 dir = (params.second_pass ? vec2(0.0, 1.0) : vec2(1.0, 0.0)); + + vec4 color = weighted_filter_dir(dir, uv, pixel_size); + + frag_color = color; + frag_color.a = alpha; // attempt to retain this in case we have a transparent background, ignored if half_size +#ifdef OUTPUT_WEIGHT + weight = color.a; +#endif + +#endif + +#ifdef MODE_BOKEH_HEXAGONAL + + //pixel_size*=0.5; //resolution is doubled + if (params.second_pass || !params.half_size) { + uv += pixel_size * 0.5; //half pixel to read centers + } else { + uv += pixel_size * 0.25; //half pixel to read centers from full res + } + + float alpha = texture(source_color, uv).a; // retain this + + vec2 dir = (params.second_pass ? normalize(vec2(1.0, 0.577350269189626)) : vec2(0.0, 1.0)); + + vec4 color = weighted_filter_dir(dir, uv, pixel_size); + + if (params.second_pass) { + dir = normalize(vec2(-1.0, 0.577350269189626)); + + vec4 color2 = weighted_filter_dir(dir, uv, pixel_size); + + color.rgb = min(color.rgb, color2.rgb); + color.a = (color.a + color2.a) * 0.5; + } + + frag_color = color; + frag_color.a = alpha; // attempt to retain this in case we have a transparent background, ignored if half_size +#ifdef OUTPUT_WEIGHT + weight = color.a; +#endif + +#endif + +#ifdef MODE_BOKEH_CIRCULAR + if (params.half_size) { + pixel_size *= 0.5; //resolution is doubled + } + + uv += pixel_size * 0.5; //half pixel to read centers + + vec4 color = texture(source_color, uv); + float alpha = color.a; // retain this + color.a = texture(source_weight, uv).r; + + vec4 color_accum = color; + float accum = 1.0; + + float radius = params.blur_scale; + for (float ang = 0.0; radius < params.blur_size; ang += GOLDEN_ANGLE) { + vec2 uv_adj = uv + vec2(cos(ang), sin(ang)) * pixel_size * radius; + + vec4 sample_color = texture(source_color, uv_adj); + sample_color.a = texture(source_weight, uv_adj).r; + + float limit = abs(sample_color.a); + if (sample_color.a > color.a) { + limit = clamp(limit, 0.0, abs(color.a) * 2.0); + } + + limit -= DEPTH_GAP; + + float m = smoothstep(radius - 0.5, radius + 0.5, limit); + color_accum += mix(color_accum / accum, sample_color, m); + accum += 1.0; + + radius += params.blur_scale / radius; + } + + color_accum = color_accum / accum; + + frag_color.rgb = color_accum.rgb; + frag_color.a = alpha; // attempt to retain this in case we have a transparent background, ignored if half_size +#ifdef OUTPUT_WEIGHT + weight = color_accum.a; +#endif + +#endif + +#ifdef MODE_COMPOSITE_BOKEH + frag_color.rgb = texture(source_color, uv).rgb; + + float center_weigth = texture(source_weight, uv).r; + float sample_weight = texture(original_weight, uv).r; + + float mix_amount; + if (sample_weight < center_weigth) { + mix_amount = min(1.0, max(0.0, max(abs(center_weigth), abs(sample_weight)) - DEPTH_GAP)); + } else { + mix_amount = min(1.0, max(0.0, abs(center_weigth) - DEPTH_GAP)); + } + + // let alpha blending take care of mixing + frag_color.a = mix_amount; +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/effects/copy.glsl b/servers/rendering/renderer_rd/shaders/effects/copy.glsl new file mode 100644 index 0000000000..bfe329b8ec --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/copy.glsl @@ -0,0 +1,284 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +#define FLAG_HORIZONTAL (1 << 0) +#define FLAG_USE_BLUR_SECTION (1 << 1) +#define FLAG_USE_ORTHOGONAL_PROJECTION (1 << 2) +#define FLAG_DOF_NEAR_FIRST_TAP (1 << 3) +#define FLAG_GLOW_FIRST_PASS (1 << 4) +#define FLAG_FLIP_Y (1 << 5) +#define FLAG_FORCE_LUMINANCE (1 << 6) +#define FLAG_COPY_ALL_SOURCE (1 << 7) +#define FLAG_HIGH_QUALITY_GLOW (1 << 8) +#define FLAG_ALPHA_TO_ONE (1 << 9) + +layout(push_constant, std430) uniform Params { + ivec4 section; + ivec2 target; + uint flags; + uint pad; + // Glow. + float glow_strength; + float glow_bloom; + float glow_hdr_threshold; + float glow_hdr_scale; + + float glow_exposure; + float glow_white; + float glow_luminance_cap; + float glow_auto_exposure_scale; + // DOF. + float camera_z_far; + float camera_z_near; + uint pad2[2]; + + vec4 set_color; +} +params; + +#ifdef MODE_CUBEMAP_ARRAY_TO_PANORAMA +layout(set = 0, binding = 0) uniform samplerCubeArray source_color; +#elif defined(MODE_CUBEMAP_TO_PANORAMA) +layout(set = 0, binding = 0) uniform samplerCube source_color; +#elif !defined(MODE_SET_COLOR) +layout(set = 0, binding = 0) uniform sampler2D source_color; +#endif + +#ifdef GLOW_USE_AUTO_EXPOSURE +layout(set = 1, binding = 0) uniform sampler2D source_auto_exposure; +#endif + +#if defined(MODE_LINEARIZE_DEPTH_COPY) || defined(MODE_SIMPLE_COPY_DEPTH) +layout(r32f, set = 3, binding = 0) uniform restrict writeonly image2D dest_buffer; +#elif defined(DST_IMAGE_8BIT) +layout(rgba8, set = 3, binding = 0) uniform restrict writeonly image2D dest_buffer; +#else +layout(rgba32f, set = 3, binding = 0) uniform restrict writeonly image2D dest_buffer; +#endif + +#ifdef MODE_GAUSSIAN_BLUR +shared vec4 local_cache[256]; +shared vec4 temp_cache[128]; +#endif + +void main() { + // Pixel being shaded + ivec2 pos = ivec2(gl_GlobalInvocationID.xy); + +#ifndef MODE_GAUSSIAN_BLUR // Gaussian blur needs the extra threads + if (any(greaterThanEqual(pos, params.section.zw))) { //too large, do nothing + return; + } +#endif + +#ifdef MODE_MIPMAP + + ivec2 base_pos = (pos + params.section.xy) << 1; + vec4 color = texelFetch(source_color, base_pos, 0); + color += texelFetch(source_color, base_pos + ivec2(0, 1), 0); + color += texelFetch(source_color, base_pos + ivec2(1, 0), 0); + color += texelFetch(source_color, base_pos + ivec2(1, 1), 0); + color /= 4.0; + color = mix(color, vec4(100.0, 100.0, 100.0, 1.0), isinf(color)); + color = mix(color, vec4(100.0, 100.0, 100.0, 1.0), isnan(color)); + + imageStore(dest_buffer, pos + params.target, color); +#endif + +#ifdef MODE_GAUSSIAN_BLUR + + // First pass copy texture into 16x16 local memory for every 8x8 thread block + vec2 quad_center_uv = clamp(vec2(gl_GlobalInvocationID.xy + gl_LocalInvocationID.xy - 3.5) / params.section.zw, vec2(0.5 / params.section.zw), vec2(1.0 - 1.5 / params.section.zw)); + uint dest_index = gl_LocalInvocationID.x * 2 + gl_LocalInvocationID.y * 2 * 16; + +#ifdef MODE_GLOW + if (bool(params.flags & FLAG_HIGH_QUALITY_GLOW)) { + vec2 quad_offset_uv = clamp((vec2(gl_GlobalInvocationID.xy + gl_LocalInvocationID.xy - 3.0)) / params.section.zw, vec2(0.5 / params.section.zw), vec2(1.0 - 1.5 / params.section.zw)); + + local_cache[dest_index] = (textureLod(source_color, quad_center_uv, 0) + textureLod(source_color, quad_offset_uv, 0)) * 0.5; + local_cache[dest_index + 1] = (textureLod(source_color, quad_center_uv + vec2(1.0 / params.section.z, 0.0), 0) + textureLod(source_color, quad_offset_uv + vec2(1.0 / params.section.z, 0.0), 0)) * 0.5; + local_cache[dest_index + 16] = (textureLod(source_color, quad_center_uv + vec2(0.0, 1.0 / params.section.w), 0) + textureLod(source_color, quad_offset_uv + vec2(0.0, 1.0 / params.section.w), 0)) * 0.5; + local_cache[dest_index + 16 + 1] = (textureLod(source_color, quad_center_uv + vec2(1.0 / params.section.zw), 0) + textureLod(source_color, quad_offset_uv + vec2(1.0 / params.section.zw), 0)) * 0.5; + } else +#endif + { + local_cache[dest_index] = textureLod(source_color, quad_center_uv, 0); + local_cache[dest_index + 1] = textureLod(source_color, quad_center_uv + vec2(1.0 / params.section.z, 0.0), 0); + local_cache[dest_index + 16] = textureLod(source_color, quad_center_uv + vec2(0.0, 1.0 / params.section.w), 0); + local_cache[dest_index + 16 + 1] = textureLod(source_color, quad_center_uv + vec2(1.0 / params.section.zw), 0); + } +#ifdef MODE_GLOW + if (bool(params.flags & FLAG_GLOW_FIRST_PASS)) { + // Tonemap initial samples to reduce weight of fireflies: https://graphicrants.blogspot.com/2013/12/tone-mapping.html + local_cache[dest_index] /= 1.0 + dot(local_cache[dest_index].rgb, vec3(0.299, 0.587, 0.114)); + local_cache[dest_index + 1] /= 1.0 + dot(local_cache[dest_index + 1].rgb, vec3(0.299, 0.587, 0.114)); + local_cache[dest_index + 16] /= 1.0 + dot(local_cache[dest_index + 16].rgb, vec3(0.299, 0.587, 0.114)); + local_cache[dest_index + 16 + 1] /= 1.0 + dot(local_cache[dest_index + 16 + 1].rgb, vec3(0.299, 0.587, 0.114)); + } + const float kernel[4] = { 0.174938, 0.165569, 0.140367, 0.106595 }; +#else + // Simpler blur uses SIGMA2 for the gaussian kernel for a stronger effect. + const float kernel[4] = { 0.214607, 0.189879, 0.131514, 0.071303 }; +#endif + memoryBarrierShared(); + barrier(); + + // Horizontal pass. Needs to copy into 8x16 chunk of local memory so vertical pass has full resolution + uint read_index = gl_LocalInvocationID.x + gl_LocalInvocationID.y * 32 + 4; + vec4 color_top = vec4(0.0); + color_top += local_cache[read_index] * kernel[0]; + color_top += local_cache[read_index + 1] * kernel[1]; + color_top += local_cache[read_index + 2] * kernel[2]; + color_top += local_cache[read_index + 3] * kernel[3]; + color_top += local_cache[read_index - 1] * kernel[1]; + color_top += local_cache[read_index - 2] * kernel[2]; + color_top += local_cache[read_index - 3] * kernel[3]; + + vec4 color_bottom = vec4(0.0); + color_bottom += local_cache[read_index + 16] * kernel[0]; + color_bottom += local_cache[read_index + 1 + 16] * kernel[1]; + color_bottom += local_cache[read_index + 2 + 16] * kernel[2]; + color_bottom += local_cache[read_index + 3 + 16] * kernel[3]; + color_bottom += local_cache[read_index - 1 + 16] * kernel[1]; + color_bottom += local_cache[read_index - 2 + 16] * kernel[2]; + color_bottom += local_cache[read_index - 3 + 16] * kernel[3]; + + // rotate samples to take advantage of cache coherency + uint write_index = gl_LocalInvocationID.y * 2 + gl_LocalInvocationID.x * 16; + + temp_cache[write_index] = color_top; + temp_cache[write_index + 1] = color_bottom; + + memoryBarrierShared(); + barrier(); + + // If destination outside of texture, can stop doing work now + if (any(greaterThanEqual(pos, params.section.zw))) { + return; + } + + // Vertical pass + uint index = gl_LocalInvocationID.y + gl_LocalInvocationID.x * 16 + 4; + vec4 color = vec4(0.0); + + color += temp_cache[index] * kernel[0]; + color += temp_cache[index + 1] * kernel[1]; + color += temp_cache[index + 2] * kernel[2]; + color += temp_cache[index + 3] * kernel[3]; + color += temp_cache[index - 1] * kernel[1]; + color += temp_cache[index - 2] * kernel[2]; + color += temp_cache[index - 3] * kernel[3]; + +#ifdef MODE_GLOW + if (bool(params.flags & FLAG_GLOW_FIRST_PASS)) { + // Undo tonemap to restore range: https://graphicrants.blogspot.com/2013/12/tone-mapping.html + color /= 1.0 - dot(color.rgb, vec3(0.299, 0.587, 0.114)); + } + + color *= params.glow_strength; + + if (bool(params.flags & FLAG_GLOW_FIRST_PASS)) { +#ifdef GLOW_USE_AUTO_EXPOSURE + + color /= texelFetch(source_auto_exposure, ivec2(0, 0), 0).r / params.glow_auto_exposure_scale; +#endif + color *= params.glow_exposure; + + float luminance = max(color.r, max(color.g, color.b)); + float feedback = max(smoothstep(params.glow_hdr_threshold, params.glow_hdr_threshold + params.glow_hdr_scale, luminance), params.glow_bloom); + + color = min(color * feedback, vec4(params.glow_luminance_cap)); + } +#endif + imageStore(dest_buffer, pos + params.target, color); + +#endif + +#ifdef MODE_SIMPLE_COPY + + vec4 color; + if (bool(params.flags & FLAG_COPY_ALL_SOURCE)) { + vec2 uv = vec2(pos) / vec2(params.section.zw); + if (bool(params.flags & FLAG_FLIP_Y)) { + uv.y = 1.0 - uv.y; + } + color = textureLod(source_color, uv, 0.0); + + } else { + color = texelFetch(source_color, pos + params.section.xy, 0); + + if (bool(params.flags & FLAG_FLIP_Y)) { + pos.y = params.section.w - pos.y - 1; + } + } + + if (bool(params.flags & FLAG_FORCE_LUMINANCE)) { + color.rgb = vec3(max(max(color.r, color.g), color.b)); + } + + if (bool(params.flags & FLAG_ALPHA_TO_ONE)) { + color.a = 1.0; + } + + imageStore(dest_buffer, pos + params.target, color); + +#endif + +#ifdef MODE_SIMPLE_COPY_DEPTH + + vec4 color = texelFetch(source_color, pos + params.section.xy, 0); + + if (bool(params.flags & FLAG_FLIP_Y)) { + pos.y = params.section.w - pos.y - 1; + } + + imageStore(dest_buffer, pos + params.target, vec4(color.r)); + +#endif + +#ifdef MODE_LINEARIZE_DEPTH_COPY + + float depth = texelFetch(source_color, pos + params.section.xy, 0).r; + depth = depth * 2.0 - 1.0; + depth = 2.0 * params.camera_z_near * params.camera_z_far / (params.camera_z_far + params.camera_z_near - depth * (params.camera_z_far - params.camera_z_near)); + vec4 color = vec4(depth / params.camera_z_far); + + if (bool(params.flags & FLAG_FLIP_Y)) { + pos.y = params.section.w - pos.y - 1; + } + + imageStore(dest_buffer, pos + params.target, color); +#endif + +#if defined(MODE_CUBEMAP_TO_PANORAMA) || defined(MODE_CUBEMAP_ARRAY_TO_PANORAMA) + + const float PI = 3.14159265359; + vec2 uv = vec2(pos) / vec2(params.section.zw); + if (bool(params.flags & FLAG_FLIP_Y)) { + uv.y = 1.0 - uv.y; + } + float phi = uv.x * 2.0 * PI; + float theta = uv.y * PI; + + vec3 normal; + normal.x = sin(phi) * sin(theta) * -1.0; + normal.y = cos(theta); + normal.z = cos(phi) * sin(theta) * -1.0; + +#ifdef MODE_CUBEMAP_TO_PANORAMA + vec4 color = textureLod(source_color, normal, params.camera_z_far); //the biggest the lod the least the acne +#else + vec4 color = textureLod(source_color, vec4(normal, params.camera_z_far), 0.0); //the biggest the lod the least the acne +#endif + imageStore(dest_buffer, pos + params.target, color); +#endif + +#ifdef MODE_SET_COLOR + imageStore(dest_buffer, pos + params.target, params.set_color); +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/effects/copy_to_fb.glsl b/servers/rendering/renderer_rd/shaders/effects/copy_to_fb.glsl new file mode 100644 index 0000000000..1c17eabb56 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/copy_to_fb.glsl @@ -0,0 +1,169 @@ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#ifdef MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#endif //MULTIVIEW + +#ifdef MULTIVIEW +layout(location = 0) out vec3 uv_interp; +#else +layout(location = 0) out vec2 uv_interp; +#endif + +layout(push_constant, std430) uniform Params { + vec4 section; + vec2 pixel_size; + bool flip_y; + bool use_section; + + bool force_luminance; + uint pad[3]; +} +params; + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp.xy = base_arr[gl_VertexIndex]; +#ifdef MULTIVIEW + uv_interp.z = ViewIndex; +#endif + vec2 vpos = uv_interp.xy; + if (params.use_section) { + vpos = params.section.xy + vpos * params.section.zw; + } + + gl_Position = vec4(vpos * 2.0 - 1.0, 0.0, 1.0); + + if (params.flip_y) { + uv_interp.y = 1.0 - uv_interp.y; + } +} + +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#ifdef MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#endif //MULTIVIEW + +layout(push_constant, std430) uniform Params { + vec4 section; + vec2 pixel_size; + bool flip_y; + bool use_section; + + bool force_luminance; + bool alpha_to_zero; + bool srgb; + uint pad; +} +params; + +#ifdef MULTIVIEW +layout(location = 0) in vec3 uv_interp; +#else +layout(location = 0) in vec2 uv_interp; +#endif + +#ifdef MULTIVIEW +layout(set = 0, binding = 0) uniform sampler2DArray source_color; +#ifdef MODE_TWO_SOURCES +layout(set = 1, binding = 0) uniform sampler2DArray source_depth; +layout(location = 1) out float depth; +#endif /* MODE_TWO_SOURCES */ +#else /* MULTIVIEW */ +layout(set = 0, binding = 0) uniform sampler2D source_color; +#ifdef MODE_TWO_SOURCES +layout(set = 1, binding = 0) uniform sampler2D source_color2; +#endif /* MODE_TWO_SOURCES */ +#endif /* MULTIVIEW */ + +layout(location = 0) out vec4 frag_color; + +vec3 linear_to_srgb(vec3 color) { + //if going to srgb, clamp from 0 to 1. + color = clamp(color, vec3(0.0), vec3(1.0)); + const vec3 a = vec3(0.055f); + return mix((vec3(1.0f) + a) * pow(color.rgb, vec3(1.0f / 2.4f)) - a, 12.92f * color.rgb, lessThan(color.rgb, vec3(0.0031308f))); +} + +void main() { +#ifdef MULTIVIEW + vec3 uv = uv_interp; +#else + vec2 uv = uv_interp; +#endif + +#ifdef MODE_PANORAMA_TO_DP + // Note, multiview and panorama should not be mixed at this time + + //obtain normal from dual paraboloid uv +#define M_PI 3.14159265359 + + float side; + uv.y = modf(uv.y * 2.0, side); + side = side * 2.0 - 1.0; + vec3 normal = vec3(uv * 2.0 - 1.0, 0.0); + normal.z = 0.5 - 0.5 * ((normal.x * normal.x) + (normal.y * normal.y)); + normal *= -side; + normal = normalize(normal); + + //now convert normal to panorama uv + + vec2 st = vec2(atan(normal.x, normal.z), acos(normal.y)); + + if (st.x < 0.0) { + st.x += M_PI * 2.0; + } + + uv = st / vec2(M_PI * 2.0, M_PI); + + if (side < 0.0) { + //uv.y = 1.0 - uv.y; + uv = 1.0 - uv; + } +#endif /* MODE_PANORAMA_TO_DP */ + +#ifdef MULTIVIEW + vec4 color = textureLod(source_color, uv, 0.0); +#ifdef MODE_TWO_SOURCES + // In multiview our 2nd input will be our depth map + depth = textureLod(source_depth, uv, 0.0).r; +#endif /* MODE_TWO_SOURCES */ + +#else /* MULTIVIEW */ + vec4 color = textureLod(source_color, uv, 0.0); +#ifdef MODE_TWO_SOURCES + color += textureLod(source_color2, uv, 0.0); +#endif /* MODE_TWO_SOURCES */ +#endif /* MULTIVIEW */ + + if (params.force_luminance) { + color.rgb = vec3(max(max(color.r, color.g), color.b)); + } + if (params.alpha_to_zero) { + color.rgb *= color.a; + } + if (params.srgb) { + color.rgb = linear_to_srgb(color.rgb); + } + + frag_color = color; +} diff --git a/servers/rendering/renderer_rd/shaders/effects/cube_to_dp.glsl b/servers/rendering/renderer_rd/shaders/effects/cube_to_dp.glsl new file mode 100644 index 0000000000..e77d0de719 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/cube_to_dp.glsl @@ -0,0 +1,84 @@ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +layout(push_constant, std430) uniform Params { + float z_far; + float z_near; + vec2 texel_size; + vec4 screen_rect; +} +params; + +layout(location = 0) out vec2 uv_interp; + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + vec2 screen_pos = uv_interp * params.screen_rect.zw + params.screen_rect.xy; + gl_Position = vec4(screen_pos * 2.0 - 1.0, 0.0, 1.0); +} + +#[fragment] + +#version 450 + +#VERSION_DEFINES + +layout(location = 0) in vec2 uv_interp; + +layout(set = 0, binding = 0) uniform samplerCube source_cube; + +layout(push_constant, std430) uniform Params { + float z_far; + float z_near; + vec2 texel_size; + vec4 screen_rect; +} +params; + +void main() { + vec2 uv = uv_interp; + vec2 texel_size = abs(params.texel_size); + + uv = clamp(uv * (1.0 + 2.0 * texel_size) - texel_size, vec2(0.0), vec2(1.0)); + + vec3 normal = vec3(uv * 2.0 - 1.0, 0.0); + normal.z = 0.5 * (1.0 - dot(normal.xy, normal.xy)); // z = 1/2 - 1/2 * (x^2 + y^2) + normal = normalize(normal); + + normal.y = -normal.y; //needs to be flipped to match projection matrix + if (params.texel_size.x >= 0.0) { // Sign is used to encode Z flip + normal.z = -normal.z; + } + + float depth = texture(source_cube, normal).r; + + // absolute values for direction cosines, bigger value equals closer to basis axis + vec3 unorm = abs(normal); + + if ((unorm.x >= unorm.y) && (unorm.x >= unorm.z)) { + // x code + unorm = normal.x > 0.0 ? vec3(1.0, 0.0, 0.0) : vec3(-1.0, 0.0, 0.0); + } else if ((unorm.y > unorm.x) && (unorm.y >= unorm.z)) { + // y code + unorm = normal.y > 0.0 ? vec3(0.0, 1.0, 0.0) : vec3(0.0, -1.0, 0.0); + } else if ((unorm.z > unorm.x) && (unorm.z > unorm.y)) { + // z code + unorm = normal.z > 0.0 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 0.0, -1.0); + } else { + // oh-no we messed up code + // has to be + unorm = vec3(1.0, 0.0, 0.0); + } + + float depth_fix = 1.0 / dot(normal, unorm); + + depth = 2.0 * depth - 1.0; + float linear_depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - depth * (params.z_far - params.z_near)); + depth = (linear_depth * depth_fix) / params.z_far; + + gl_FragDepth = depth; +} diff --git a/servers/rendering/renderer_rd/shaders/effects/cubemap_downsampler.glsl b/servers/rendering/renderer_rd/shaders/effects/cubemap_downsampler.glsl new file mode 100644 index 0000000000..63f0ce690e --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/cubemap_downsampler.glsl @@ -0,0 +1,145 @@ +// Copyright 2016 Activision Publishing, Inc. +// +// Permission is hereby granted, free of charge, to any person obtaining +// a copy of this software and associated documentation files (the "Software"), +// to deal in the Software without restriction, including without limitation +// the rights to use, copy, modify, merge, publish, distribute, sublicense, +// and/or sell copies of the Software, and to permit persons to whom the Software +// is furnished to do so, subject to the following conditions: +// +// The above copyright notice and this permission notice shall be included in all +// copies or substantial portions of the Software. +// +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +// SOFTWARE. + +#[compute] + +#version 450 + +#VERSION_DEFINES + +#define BLOCK_SIZE 8 + +layout(local_size_x = BLOCK_SIZE, local_size_y = BLOCK_SIZE, local_size_z = 1) in; + +layout(set = 0, binding = 0) uniform samplerCube source_cubemap; + +layout(rgba16f, set = 1, binding = 0) uniform restrict writeonly imageCube dest_cubemap; + +#include "cubemap_downsampler_inc.glsl" + +void main() { + uvec3 id = gl_GlobalInvocationID; + uint face_size = params.face_size; + + if (id.x < face_size && id.y < face_size) { + float inv_face_size = 1.0 / float(face_size); + + float u0 = (float(id.x) * 2.0 + 1.0 - 0.75) * inv_face_size - 1.0; + float u1 = (float(id.x) * 2.0 + 1.0 + 0.75) * inv_face_size - 1.0; + + float v0 = (float(id.y) * 2.0 + 1.0 - 0.75) * -inv_face_size + 1.0; + float v1 = (float(id.y) * 2.0 + 1.0 + 0.75) * -inv_face_size + 1.0; + + float weights[4]; + weights[0] = calcWeight(u0, v0); + weights[1] = calcWeight(u1, v0); + weights[2] = calcWeight(u0, v1); + weights[3] = calcWeight(u1, v1); + + const float wsum = 0.5 / (weights[0] + weights[1] + weights[2] + weights[3]); + for (int i = 0; i < 4; i++) { + weights[i] = weights[i] * wsum + .125; + } + + vec3 dir; + vec4 color; + switch (id.z) { + case 0: + get_dir_0(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_0(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_0(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_0(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + case 1: + get_dir_1(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_1(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_1(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_1(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + case 2: + get_dir_2(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_2(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_2(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_2(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + case 3: + get_dir_3(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_3(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_3(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_3(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + case 4: + get_dir_4(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_4(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_4(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_4(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + default: + get_dir_5(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_5(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_5(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_5(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + } + imageStore(dest_cubemap, ivec3(id), color); + } +} diff --git a/servers/rendering/renderer_rd/shaders/effects/cubemap_downsampler_inc.glsl b/servers/rendering/renderer_rd/shaders/effects/cubemap_downsampler_inc.glsl new file mode 100644 index 0000000000..641e0906f5 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/cubemap_downsampler_inc.glsl @@ -0,0 +1,48 @@ +layout(push_constant, std430) uniform Params { + uint face_size; + uint face_id; // only used in raster shader +} +params; + +#define M_PI 3.14159265359 + +void get_dir_0(out vec3 dir, in float u, in float v) { + dir[0] = 1.0; + dir[1] = v; + dir[2] = -u; +} + +void get_dir_1(out vec3 dir, in float u, in float v) { + dir[0] = -1.0; + dir[1] = v; + dir[2] = u; +} + +void get_dir_2(out vec3 dir, in float u, in float v) { + dir[0] = u; + dir[1] = 1.0; + dir[2] = -v; +} + +void get_dir_3(out vec3 dir, in float u, in float v) { + dir[0] = u; + dir[1] = -1.0; + dir[2] = v; +} + +void get_dir_4(out vec3 dir, in float u, in float v) { + dir[0] = u; + dir[1] = v; + dir[2] = 1.0; +} + +void get_dir_5(out vec3 dir, in float u, in float v) { + dir[0] = -u; + dir[1] = v; + dir[2] = -1.0; +} + +float calcWeight(float u, float v) { + float val = u * u + v * v + 1.0; + return val * sqrt(val); +} diff --git a/servers/rendering/renderer_rd/shaders/effects/cubemap_downsampler_raster.glsl b/servers/rendering/renderer_rd/shaders/effects/cubemap_downsampler_raster.glsl new file mode 100644 index 0000000000..0828ffd921 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/cubemap_downsampler_raster.glsl @@ -0,0 +1,163 @@ +// Copyright 2016 Activision Publishing, Inc. +// +// Permission is hereby granted, free of charge, to any person obtaining +// a copy of this software and associated documentation files (the "Software"), +// to deal in the Software without restriction, including without limitation +// the rights to use, copy, modify, merge, publish, distribute, sublicense, +// and/or sell copies of the Software, and to permit persons to whom the Software +// is furnished to do so, subject to the following conditions: +// +// The above copyright notice and this permission notice shall be included in all +// copies or substantial portions of the Software. +// +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +// SOFTWARE. + +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#include "cubemap_downsampler_inc.glsl" + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex] * float(params.face_size); + gl_Position = vec4(base_arr[gl_VertexIndex] * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#include "cubemap_downsampler_inc.glsl" + +layout(set = 0, binding = 0) uniform samplerCube source_cubemap; + +layout(location = 0) in vec2 uv_interp; +layout(location = 0) out vec4 frag_color; +/* clang-format on */ + +void main() { + // Converted from compute shader which uses absolute coordinates. + // Could possibly simplify this + float face_size = float(params.face_size); + + if (uv_interp.x < face_size && uv_interp.y < face_size) { + float inv_face_size = 1.0 / face_size; + + float u0 = (uv_interp.x * 2.0 + 1.0 - 0.75) * inv_face_size - 1.0; + float u1 = (uv_interp.x * 2.0 + 1.0 + 0.75) * inv_face_size - 1.0; + + float v0 = (uv_interp.y * 2.0 + 1.0 - 0.75) * -inv_face_size + 1.0; + float v1 = (uv_interp.y * 2.0 + 1.0 + 0.75) * -inv_face_size + 1.0; + + float weights[4]; + weights[0] = calcWeight(u0, v0); + weights[1] = calcWeight(u1, v0); + weights[2] = calcWeight(u0, v1); + weights[3] = calcWeight(u1, v1); + + const float wsum = 0.5 / (weights[0] + weights[1] + weights[2] + weights[3]); + for (int i = 0; i < 4; i++) { + weights[i] = weights[i] * wsum + .125; + } + + vec3 dir; + vec4 color; + switch (params.face_id) { + case 0: + get_dir_0(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_0(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_0(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_0(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + case 1: + get_dir_1(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_1(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_1(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_1(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + case 2: + get_dir_2(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_2(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_2(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_2(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + case 3: + get_dir_3(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_3(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_3(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_3(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + case 4: + get_dir_4(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_4(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_4(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_4(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + default: + get_dir_5(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_5(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_5(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_5(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + } + frag_color = color; + } +} diff --git a/servers/rendering/renderer_rd/shaders/effects/cubemap_filter.glsl b/servers/rendering/renderer_rd/shaders/effects/cubemap_filter.glsl new file mode 100644 index 0000000000..2a774b0eb4 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/cubemap_filter.glsl @@ -0,0 +1,326 @@ +// Copyright 2016 Activision Publishing, Inc. +// +// Permission is hereby granted, free of charge, to any person obtaining +// a copy of this software and associated documentation files (the "Software"), +// to deal in the Software without restriction, including without limitation +// the rights to use, copy, modify, merge, publish, distribute, sublicense, +// and/or sell copies of the Software, and to permit persons to whom the Software +// is furnished to do so, subject to the following conditions: +// +// The above copyright notice and this permission notice shall be included in all +// copies or substantial portions of the Software. +// +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +// SOFTWARE. + +#[compute] + +#version 450 + +#VERSION_DEFINES + +#define GROUP_SIZE 64 + +layout(local_size_x = GROUP_SIZE, local_size_y = 1, local_size_z = 1) in; + +layout(set = 0, binding = 0) uniform samplerCube source_cubemap; +layout(rgba16f, set = 2, binding = 0) uniform restrict writeonly imageCube dest_cubemap0; +layout(rgba16f, set = 2, binding = 1) uniform restrict writeonly imageCube dest_cubemap1; +layout(rgba16f, set = 2, binding = 2) uniform restrict writeonly imageCube dest_cubemap2; +layout(rgba16f, set = 2, binding = 3) uniform restrict writeonly imageCube dest_cubemap3; +layout(rgba16f, set = 2, binding = 4) uniform restrict writeonly imageCube dest_cubemap4; +layout(rgba16f, set = 2, binding = 5) uniform restrict writeonly imageCube dest_cubemap5; +layout(rgba16f, set = 2, binding = 6) uniform restrict writeonly imageCube dest_cubemap6; + +#ifdef USE_HIGH_QUALITY +#define NUM_TAPS 32 +#else +#define NUM_TAPS 8 +#endif + +#define BASE_RESOLUTION 128 + +#ifdef USE_HIGH_QUALITY +layout(set = 1, binding = 0, std430) buffer restrict readonly Data { + vec4[7][5][3][24] coeffs; +} +data; +#else +layout(set = 1, binding = 0, std430) buffer restrict readonly Data { + vec4[7][5][6] coeffs; +} +data; +#endif + +void get_dir(out vec3 dir, in vec2 uv, in uint face) { + switch (face) { + case 0: + dir = vec3(1.0, uv[1], -uv[0]); + break; + case 1: + dir = vec3(-1.0, uv[1], uv[0]); + break; + case 2: + dir = vec3(uv[0], 1.0, -uv[1]); + break; + case 3: + dir = vec3(uv[0], -1.0, uv[1]); + break; + case 4: + dir = vec3(uv[0], uv[1], 1.0); + break; + default: + dir = vec3(-uv[0], uv[1], -1.0); + break; + } +} + +void main() { + // INPUT: + // id.x = the linear address of the texel (ignoring face) + // id.y = the face + // -> use to index output texture + // id.x = texel x + // id.y = texel y + // id.z = face + uvec3 id = gl_GlobalInvocationID; + + // determine which texel this is +#ifndef USE_TEXTURE_ARRAY + // NOTE (macOS/MoltenVK): Do not rename, "level" variable name conflicts with the Metal "level(float lod)" mipmap sampling function name. + int mip_level = 0; + if (id.x < (128 * 128)) { + mip_level = 0; + } else if (id.x < (128 * 128 + 64 * 64)) { + mip_level = 1; + id.x -= (128 * 128); + } else if (id.x < (128 * 128 + 64 * 64 + 32 * 32)) { + mip_level = 2; + id.x -= (128 * 128 + 64 * 64); + } else if (id.x < (128 * 128 + 64 * 64 + 32 * 32 + 16 * 16)) { + mip_level = 3; + id.x -= (128 * 128 + 64 * 64 + 32 * 32); + } else if (id.x < (128 * 128 + 64 * 64 + 32 * 32 + 16 * 16 + 8 * 8)) { + mip_level = 4; + id.x -= (128 * 128 + 64 * 64 + 32 * 32 + 16 * 16); + } else if (id.x < (128 * 128 + 64 * 64 + 32 * 32 + 16 * 16 + 8 * 8 + 4 * 4)) { + mip_level = 5; + id.x -= (128 * 128 + 64 * 64 + 32 * 32 + 16 * 16 + 8 * 8); + } else if (id.x < (128 * 128 + 64 * 64 + 32 * 32 + 16 * 16 + 8 * 8 + 4 * 4 + 2 * 2)) { + mip_level = 6; + id.x -= (128 * 128 + 64 * 64 + 32 * 32 + 16 * 16 + 8 * 8 + 4 * 4); + } else { + return; + } + int res = BASE_RESOLUTION >> mip_level; +#else // Using Texture Arrays so all levels are the same resolution + int res = BASE_RESOLUTION; + int mip_level = int(id.x / (BASE_RESOLUTION * BASE_RESOLUTION)); + id.x -= mip_level * BASE_RESOLUTION * BASE_RESOLUTION; +#endif + + // determine dir / pos for the texel + vec3 dir, adir, frameZ; + { + id.z = id.y; + id.y = id.x / res; + id.x -= id.y * res; + + vec2 uv; + uv.x = (float(id.x) * 2.0 + 1.0) / float(res) - 1.0; + uv.y = -(float(id.y) * 2.0 + 1.0) / float(res) + 1.0; + + get_dir(dir, uv, id.z); + frameZ = normalize(dir); + + adir = abs(dir); + } + + // GGX gather colors + vec4 color = vec4(0.0); + for (int axis = 0; axis < 3; axis++) { + const int otherAxis0 = 1 - (axis & 1) - (axis >> 1); + const int otherAxis1 = 2 - (axis >> 1); + + float frameweight = (max(adir[otherAxis0], adir[otherAxis1]) - .75) / .25; + if (frameweight > 0.0) { + // determine frame + vec3 UpVector; + switch (axis) { + case 0: + UpVector = vec3(1, 0, 0); + break; + case 1: + UpVector = vec3(0, 1, 0); + break; + default: + UpVector = vec3(0, 0, 1); + break; + } + + vec3 frameX = normalize(cross(UpVector, frameZ)); + vec3 frameY = cross(frameZ, frameX); + + // calculate parametrization for polynomial + float Nx = dir[otherAxis0]; + float Ny = dir[otherAxis1]; + float Nz = adir[axis]; + + float NmaxXY = max(abs(Ny), abs(Nx)); + Nx /= NmaxXY; + Ny /= NmaxXY; + + float theta; + if (Ny < Nx) { + if (Ny <= -0.999) + theta = Nx; + else + theta = Ny; + } else { + if (Ny >= 0.999) + theta = -Nx; + else + theta = -Ny; + } + + float phi; + if (Nz <= -0.999) + phi = -NmaxXY; + else if (Nz >= 0.999) + phi = NmaxXY; + else + phi = Nz; + + float theta2 = theta * theta; + float phi2 = phi * phi; + + // sample + for (int iSuperTap = 0; iSuperTap < NUM_TAPS / 4; iSuperTap++) { + const int index = (NUM_TAPS / 4) * axis + iSuperTap; + +#ifdef USE_HIGH_QUALITY + vec4 coeffsDir0[3]; + vec4 coeffsDir1[3]; + vec4 coeffsDir2[3]; + vec4 coeffsLevel[3]; + vec4 coeffsWeight[3]; + + for (int iCoeff = 0; iCoeff < 3; iCoeff++) { + coeffsDir0[iCoeff] = data.coeffs[mip_level][0][iCoeff][index]; + coeffsDir1[iCoeff] = data.coeffs[mip_level][1][iCoeff][index]; + coeffsDir2[iCoeff] = data.coeffs[mip_level][2][iCoeff][index]; + coeffsLevel[iCoeff] = data.coeffs[mip_level][3][iCoeff][index]; + coeffsWeight[iCoeff] = data.coeffs[mip_level][4][iCoeff][index]; + } + + for (int iSubTap = 0; iSubTap < 4; iSubTap++) { + // determine sample attributes (dir, weight, mip_level) + vec3 sample_dir = frameX * (coeffsDir0[0][iSubTap] + coeffsDir0[1][iSubTap] * theta2 + coeffsDir0[2][iSubTap] * phi2) + frameY * (coeffsDir1[0][iSubTap] + coeffsDir1[1][iSubTap] * theta2 + coeffsDir1[2][iSubTap] * phi2) + frameZ * (coeffsDir2[0][iSubTap] + coeffsDir2[1][iSubTap] * theta2 + coeffsDir2[2][iSubTap] * phi2); + + float sample_level = coeffsLevel[0][iSubTap] + coeffsLevel[1][iSubTap] * theta2 + coeffsLevel[2][iSubTap] * phi2; + + float sample_weight = coeffsWeight[0][iSubTap] + coeffsWeight[1][iSubTap] * theta2 + coeffsWeight[2][iSubTap] * phi2; +#else + vec4 coeffsDir0 = data.coeffs[mip_level][0][index]; + vec4 coeffsDir1 = data.coeffs[mip_level][1][index]; + vec4 coeffsDir2 = data.coeffs[mip_level][2][index]; + vec4 coeffsLevel = data.coeffs[mip_level][3][index]; + vec4 coeffsWeight = data.coeffs[mip_level][4][index]; + + for (int iSubTap = 0; iSubTap < 4; iSubTap++) { + // determine sample attributes (dir, weight, mip_level) + vec3 sample_dir = frameX * coeffsDir0[iSubTap] + frameY * coeffsDir1[iSubTap] + frameZ * coeffsDir2[iSubTap]; + + float sample_level = coeffsLevel[iSubTap]; + + float sample_weight = coeffsWeight[iSubTap]; +#endif + + sample_weight *= frameweight; + + // adjust for jacobian + sample_dir /= max(abs(sample_dir[0]), max(abs(sample_dir[1]), abs(sample_dir[2]))); + sample_level += 0.75 * log2(dot(sample_dir, sample_dir)); +#ifndef USE_TEXTURE_ARRAY + sample_level += float(mip_level) / 6.0; // Hack to increase the perceived roughness and reduce upscaling artifacts +#endif + // sample cubemap + color.xyz += textureLod(source_cubemap, normalize(sample_dir), sample_level).xyz * sample_weight; + color.w += sample_weight; + } + } + } + } + color /= color.w; + + // write color + color.xyz = max(vec3(0.0), color.xyz); + color.w = 1.0; +#ifdef USE_TEXTURE_ARRAY + id.xy *= uvec2(2, 2); +#endif + + switch (mip_level) { + case 0: + imageStore(dest_cubemap0, ivec3(id), color); +#ifdef USE_TEXTURE_ARRAY + imageStore(dest_cubemap0, ivec3(id) + ivec3(1.0, 0.0, 0.0), color); + imageStore(dest_cubemap0, ivec3(id) + ivec3(0.0, 1.0, 0.0), color); + imageStore(dest_cubemap0, ivec3(id) + ivec3(1.0, 1.0, 0.0), color); +#endif + break; + case 1: + imageStore(dest_cubemap1, ivec3(id), color); +#ifdef USE_TEXTURE_ARRAY + imageStore(dest_cubemap1, ivec3(id) + ivec3(1.0, 0.0, 0.0), color); + imageStore(dest_cubemap1, ivec3(id) + ivec3(0.0, 1.0, 0.0), color); + imageStore(dest_cubemap1, ivec3(id) + ivec3(1.0, 1.0, 0.0), color); +#endif + break; + case 2: + imageStore(dest_cubemap2, ivec3(id), color); +#ifdef USE_TEXTURE_ARRAY + imageStore(dest_cubemap2, ivec3(id) + ivec3(1.0, 0.0, 0.0), color); + imageStore(dest_cubemap2, ivec3(id) + ivec3(0.0, 1.0, 0.0), color); + imageStore(dest_cubemap2, ivec3(id) + ivec3(1.0, 1.0, 0.0), color); +#endif + break; + case 3: + imageStore(dest_cubemap3, ivec3(id), color); +#ifdef USE_TEXTURE_ARRAY + imageStore(dest_cubemap3, ivec3(id) + ivec3(1.0, 0.0, 0.0), color); + imageStore(dest_cubemap3, ivec3(id) + ivec3(0.0, 1.0, 0.0), color); + imageStore(dest_cubemap3, ivec3(id) + ivec3(1.0, 1.0, 0.0), color); +#endif + break; + case 4: + imageStore(dest_cubemap4, ivec3(id), color); +#ifdef USE_TEXTURE_ARRAY + imageStore(dest_cubemap4, ivec3(id) + ivec3(1.0, 0.0, 0.0), color); + imageStore(dest_cubemap4, ivec3(id) + ivec3(0.0, 1.0, 0.0), color); + imageStore(dest_cubemap4, ivec3(id) + ivec3(1.0, 1.0, 0.0), color); +#endif + break; + case 5: + imageStore(dest_cubemap5, ivec3(id), color); +#ifdef USE_TEXTURE_ARRAY + imageStore(dest_cubemap5, ivec3(id) + ivec3(1.0, 0.0, 0.0), color); + imageStore(dest_cubemap5, ivec3(id) + ivec3(0.0, 1.0, 0.0), color); + imageStore(dest_cubemap5, ivec3(id) + ivec3(1.0, 1.0, 0.0), color); +#endif + break; + default: + imageStore(dest_cubemap6, ivec3(id), color); +#ifdef USE_TEXTURE_ARRAY + imageStore(dest_cubemap6, ivec3(id) + ivec3(1.0, 0.0, 0.0), color); + imageStore(dest_cubemap6, ivec3(id) + ivec3(0.0, 1.0, 0.0), color); + imageStore(dest_cubemap6, ivec3(id) + ivec3(1.0, 1.0, 0.0), color); +#endif + break; + } +} diff --git a/servers/rendering/renderer_rd/shaders/effects/cubemap_filter_raster.glsl b/servers/rendering/renderer_rd/shaders/effects/cubemap_filter_raster.glsl new file mode 100644 index 0000000000..0990dc7c2f --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/cubemap_filter_raster.glsl @@ -0,0 +1,256 @@ +// Copyright 2016 Activision Publishing, Inc. +// +// Permission is hereby granted, free of charge, to any person obtaining +// a copy of this software and associated documentation files (the "Software"), +// to deal in the Software without restriction, including without limitation +// the rights to use, copy, modify, merge, publish, distribute, sublicense, +// and/or sell copies of the Software, and to permit persons to whom the Software +// is furnished to do so, subject to the following conditions: +// +// The above copyright notice and this permission notice shall be included in all +// copies or substantial portions of the Software. +// +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +// SOFTWARE. + +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +layout(push_constant, std430) uniform Params { + int mip_level; + uint face_id; +} +params; + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + gl_Position = vec4(base_arr[gl_VertexIndex] * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +layout(push_constant, std430) uniform Params { + int mip_level; + uint face_id; +} +params; + +layout(set = 0, binding = 0) uniform samplerCube source_cubemap; + +layout(location = 0) in vec2 uv_interp; +layout(location = 0) out vec4 frag_color; + +/* clang-format on */ + +#ifdef USE_HIGH_QUALITY +#define NUM_TAPS 32 +#else +#define NUM_TAPS 8 +#endif + +#define BASE_RESOLUTION 128 + +#ifdef USE_HIGH_QUALITY +layout(set = 1, binding = 0, std430) buffer restrict readonly Data { + vec4[7][5][3][24] coeffs; +} +data; +#else +layout(set = 1, binding = 0, std430) buffer restrict readonly Data { + vec4[7][5][6] coeffs; +} +data; +#endif + +void get_dir(out vec3 dir, in vec2 uv, in uint face) { + switch (face) { + case 0: + dir = vec3(1.0, uv[1], -uv[0]); + break; + case 1: + dir = vec3(-1.0, uv[1], uv[0]); + break; + case 2: + dir = vec3(uv[0], 1.0, -uv[1]); + break; + case 3: + dir = vec3(uv[0], -1.0, uv[1]); + break; + case 4: + dir = vec3(uv[0], uv[1], 1.0); + break; + default: + dir = vec3(-uv[0], uv[1], -1.0); + break; + } +} + +void main() { + // determine dir / pos for the texel + vec3 dir, adir, frameZ; + { + vec2 uv; + uv.x = uv_interp.x; + uv.y = 1.0 - uv_interp.y; + uv = uv * 2.0 - 1.0; + + get_dir(dir, uv, params.face_id); + frameZ = normalize(dir); + + adir = abs(dir); + } + + // determine which texel this is + // NOTE (macOS/MoltenVK): Do not rename, "level" variable name conflicts with the Metal "level(float lod)" mipmap sampling function name. + int mip_level = 0; + + if (params.mip_level < 0) { + // return as is + frag_color.rgb = textureLod(source_cubemap, frameZ, 0.0).rgb; + frag_color.a = 1.0; + return; + } else if (params.mip_level > 6) { + // maximum level + mip_level = 6; + } else { + mip_level = params.mip_level; + } + + // GGX gather colors + vec4 color = vec4(0.0); + for (int axis = 0; axis < 3; axis++) { + const int otherAxis0 = 1 - (axis & 1) - (axis >> 1); + const int otherAxis1 = 2 - (axis >> 1); + + float frameweight = (max(adir[otherAxis0], adir[otherAxis1]) - .75) / .25; + if (frameweight > 0.0) { + // determine frame + vec3 UpVector; + switch (axis) { + case 0: + UpVector = vec3(1, 0, 0); + break; + case 1: + UpVector = vec3(0, 1, 0); + break; + default: + UpVector = vec3(0, 0, 1); + break; + } + + vec3 frameX = normalize(cross(UpVector, frameZ)); + vec3 frameY = cross(frameZ, frameX); + + // calculate parametrization for polynomial + float Nx = dir[otherAxis0]; + float Ny = dir[otherAxis1]; + float Nz = adir[axis]; + + float NmaxXY = max(abs(Ny), abs(Nx)); + Nx /= NmaxXY; + Ny /= NmaxXY; + + float theta; + if (Ny < Nx) { + if (Ny <= -0.999) + theta = Nx; + else + theta = Ny; + } else { + if (Ny >= 0.999) + theta = -Nx; + else + theta = -Ny; + } + + float phi; + if (Nz <= -0.999) + phi = -NmaxXY; + else if (Nz >= 0.999) + phi = NmaxXY; + else + phi = Nz; + + float theta2 = theta * theta; + float phi2 = phi * phi; + + // sample + for (int iSuperTap = 0; iSuperTap < NUM_TAPS / 4; iSuperTap++) { + const int index = (NUM_TAPS / 4) * axis + iSuperTap; + +#ifdef USE_HIGH_QUALITY + vec4 coeffsDir0[3]; + vec4 coeffsDir1[3]; + vec4 coeffsDir2[3]; + vec4 coeffsLevel[3]; + vec4 coeffsWeight[3]; + + for (int iCoeff = 0; iCoeff < 3; iCoeff++) { + coeffsDir0[iCoeff] = data.coeffs[mip_level][0][iCoeff][index]; + coeffsDir1[iCoeff] = data.coeffs[mip_level][1][iCoeff][index]; + coeffsDir2[iCoeff] = data.coeffs[mip_level][2][iCoeff][index]; + coeffsLevel[iCoeff] = data.coeffs[mip_level][3][iCoeff][index]; + coeffsWeight[iCoeff] = data.coeffs[mip_level][4][iCoeff][index]; + } + + for (int iSubTap = 0; iSubTap < 4; iSubTap++) { + // determine sample attributes (dir, weight, mip_level) + vec3 sample_dir = frameX * (coeffsDir0[0][iSubTap] + coeffsDir0[1][iSubTap] * theta2 + coeffsDir0[2][iSubTap] * phi2) + frameY * (coeffsDir1[0][iSubTap] + coeffsDir1[1][iSubTap] * theta2 + coeffsDir1[2][iSubTap] * phi2) + frameZ * (coeffsDir2[0][iSubTap] + coeffsDir2[1][iSubTap] * theta2 + coeffsDir2[2][iSubTap] * phi2); + + float sample_level = coeffsLevel[0][iSubTap] + coeffsLevel[1][iSubTap] * theta2 + coeffsLevel[2][iSubTap] * phi2; + + float sample_weight = coeffsWeight[0][iSubTap] + coeffsWeight[1][iSubTap] * theta2 + coeffsWeight[2][iSubTap] * phi2; +#else + vec4 coeffsDir0 = data.coeffs[mip_level][0][index]; + vec4 coeffsDir1 = data.coeffs[mip_level][1][index]; + vec4 coeffsDir2 = data.coeffs[mip_level][2][index]; + vec4 coeffsLevel = data.coeffs[mip_level][3][index]; + vec4 coeffsWeight = data.coeffs[mip_level][4][index]; + + for (int iSubTap = 0; iSubTap < 4; iSubTap++) { + // determine sample attributes (dir, weight, mip_level) + vec3 sample_dir = frameX * coeffsDir0[iSubTap] + frameY * coeffsDir1[iSubTap] + frameZ * coeffsDir2[iSubTap]; + + float sample_level = coeffsLevel[iSubTap]; + + float sample_weight = coeffsWeight[iSubTap]; +#endif + + sample_weight *= frameweight; + + // adjust for jacobian + sample_dir /= max(abs(sample_dir[0]), max(abs(sample_dir[1]), abs(sample_dir[2]))); + sample_level += 0.75 * log2(dot(sample_dir, sample_dir)); + // sample cubemap + color.xyz += textureLod(source_cubemap, normalize(sample_dir), sample_level).xyz * sample_weight; + color.w += sample_weight; + } + } + } + } + color /= color.w; + + // write color + color.xyz = max(vec3(0.0), color.xyz); + color.w = 1.0; + + frag_color = color; +} diff --git a/servers/rendering/renderer_rd/shaders/effects/cubemap_roughness.glsl b/servers/rendering/renderer_rd/shaders/effects/cubemap_roughness.glsl new file mode 100644 index 0000000000..1d46f59408 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/cubemap_roughness.glsl @@ -0,0 +1,63 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +#define GROUP_SIZE 8 + +layout(local_size_x = GROUP_SIZE, local_size_y = GROUP_SIZE, local_size_z = 1) in; + +layout(set = 0, binding = 0) uniform samplerCube source_cube; + +layout(rgba16f, set = 1, binding = 0) uniform restrict writeonly imageCube dest_cubemap; + +#include "cubemap_roughness_inc.glsl" + +void main() { + uvec3 id = gl_GlobalInvocationID; + id.z += params.face_id; + + vec2 uv = ((vec2(id.xy) * 2.0 + 1.0) / (params.face_size) - 1.0); + vec3 N = texelCoordToVec(uv, id.z); + + if (params.use_direct_write) { + imageStore(dest_cubemap, ivec3(id), vec4(texture(source_cube, N).rgb, 1.0)); + } else { + vec4 sum = vec4(0.0, 0.0, 0.0, 0.0); + + float solid_angle_texel = 4.0 * M_PI / (6.0 * params.face_size * params.face_size); + float roughness2 = params.roughness * params.roughness; + float roughness4 = roughness2 * roughness2; + vec3 UpVector = abs(N.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(1.0, 0.0, 0.0); + mat3 T; + T[0] = normalize(cross(UpVector, N)); + T[1] = cross(N, T[0]); + T[2] = N; + + for (uint sampleNum = 0u; sampleNum < params.sample_count; sampleNum++) { + vec2 xi = Hammersley(sampleNum, params.sample_count); + + vec3 H = T * ImportanceSampleGGX(xi, roughness4); + float NdotH = dot(N, H); + vec3 L = (2.0 * NdotH * H - N); + + float ndotl = clamp(dot(N, L), 0.0, 1.0); + + if (ndotl > 0.0) { + float D = DistributionGGX(NdotH, roughness4); + float pdf = D * NdotH / (4.0 * NdotH) + 0.0001; + + float solid_angle_sample = 1.0 / (float(params.sample_count) * pdf + 0.0001); + + float mipLevel = params.roughness == 0.0 ? 0.0 : 0.5 * log2(solid_angle_sample / solid_angle_texel); + + sum.rgb += textureLod(source_cube, L, mipLevel).rgb * ndotl; + sum.a += ndotl; + } + } + sum /= sum.a; + + imageStore(dest_cubemap, ivec3(id), vec4(sum.rgb, 1.0)); + } +} diff --git a/servers/rendering/renderer_rd/shaders/effects/cubemap_roughness_inc.glsl b/servers/rendering/renderer_rd/shaders/effects/cubemap_roughness_inc.glsl new file mode 100644 index 0000000000..1bee428a6f --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/cubemap_roughness_inc.glsl @@ -0,0 +1,95 @@ +#define M_PI 3.14159265359 + +layout(push_constant, std430) uniform Params { + uint face_id; + uint sample_count; + float roughness; + bool use_direct_write; + float face_size; +} +params; + +vec3 texelCoordToVec(vec2 uv, uint faceID) { + mat3 faceUvVectors[6]; + + // -x + faceUvVectors[1][0] = vec3(0.0, 0.0, 1.0); // u -> +z + faceUvVectors[1][1] = vec3(0.0, -1.0, 0.0); // v -> -y + faceUvVectors[1][2] = vec3(-1.0, 0.0, 0.0); // -x face + + // +x + faceUvVectors[0][0] = vec3(0.0, 0.0, -1.0); // u -> -z + faceUvVectors[0][1] = vec3(0.0, -1.0, 0.0); // v -> -y + faceUvVectors[0][2] = vec3(1.0, 0.0, 0.0); // +x face + + // -y + faceUvVectors[3][0] = vec3(1.0, 0.0, 0.0); // u -> +x + faceUvVectors[3][1] = vec3(0.0, 0.0, -1.0); // v -> -z + faceUvVectors[3][2] = vec3(0.0, -1.0, 0.0); // -y face + + // +y + faceUvVectors[2][0] = vec3(1.0, 0.0, 0.0); // u -> +x + faceUvVectors[2][1] = vec3(0.0, 0.0, 1.0); // v -> +z + faceUvVectors[2][2] = vec3(0.0, 1.0, 0.0); // +y face + + // -z + faceUvVectors[5][0] = vec3(-1.0, 0.0, 0.0); // u -> -x + faceUvVectors[5][1] = vec3(0.0, -1.0, 0.0); // v -> -y + faceUvVectors[5][2] = vec3(0.0, 0.0, -1.0); // -z face + + // +z + faceUvVectors[4][0] = vec3(1.0, 0.0, 0.0); // u -> +x + faceUvVectors[4][1] = vec3(0.0, -1.0, 0.0); // v -> -y + faceUvVectors[4][2] = vec3(0.0, 0.0, 1.0); // +z face + + // out = u * s_faceUv[0] + v * s_faceUv[1] + s_faceUv[2]. + vec3 result = (faceUvVectors[faceID][0] * uv.x) + (faceUvVectors[faceID][1] * uv.y) + faceUvVectors[faceID][2]; + return normalize(result); +} + +vec3 ImportanceSampleGGX(vec2 xi, float roughness4) { + // Compute distribution direction + float Phi = 2.0 * M_PI * xi.x; + float CosTheta = sqrt((1.0 - xi.y) / (1.0 + (roughness4 - 1.0) * xi.y)); + float SinTheta = sqrt(1.0 - CosTheta * CosTheta); + + // Convert to spherical direction + vec3 H; + H.x = SinTheta * cos(Phi); + H.y = SinTheta * sin(Phi); + H.z = CosTheta; + + return H; +} + +float DistributionGGX(float NdotH, float roughness4) { + float NdotH2 = NdotH * NdotH; + float denom = (NdotH2 * (roughness4 - 1.0) + 1.0); + denom = M_PI * denom * denom; + + return roughness4 / denom; +} + +// https://graphicrants.blogspot.com.au/2013/08/specular-brdf-reference.html +float GGX(float NdotV, float a) { + float k = a / 2.0; + return NdotV / (NdotV * (1.0 - k) + k); +} + +// https://graphicrants.blogspot.com.au/2013/08/specular-brdf-reference.html +float G_Smith(float a, float nDotV, float nDotL) { + return GGX(nDotL, a * a) * GGX(nDotV, a * a); +} + +float radicalInverse_VdC(uint bits) { + bits = (bits << 16u) | (bits >> 16u); + bits = ((bits & 0x55555555u) << 1u) | ((bits & 0xAAAAAAAAu) >> 1u); + bits = ((bits & 0x33333333u) << 2u) | ((bits & 0xCCCCCCCCu) >> 2u); + bits = ((bits & 0x0F0F0F0Fu) << 4u) | ((bits & 0xF0F0F0F0u) >> 4u); + bits = ((bits & 0x00FF00FFu) << 8u) | ((bits & 0xFF00FF00u) >> 8u); + return float(bits) * 2.3283064365386963e-10; // / 0x100000000 +} + +vec2 Hammersley(uint i, uint N) { + return vec2(float(i) / float(N), radicalInverse_VdC(i)); +} diff --git a/servers/rendering/renderer_rd/shaders/effects/cubemap_roughness_raster.glsl b/servers/rendering/renderer_rd/shaders/effects/cubemap_roughness_raster.glsl new file mode 100644 index 0000000000..c29accd8a7 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/cubemap_roughness_raster.glsl @@ -0,0 +1,79 @@ +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#include "cubemap_roughness_inc.glsl" + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#include "cubemap_roughness_inc.glsl" + +layout(location = 0) in vec2 uv_interp; + +layout(set = 0, binding = 0) uniform samplerCube source_cube; + +layout(location = 0) out vec4 frag_color; +/* clang-format on */ + +void main() { + vec3 N = texelCoordToVec(uv_interp * 2.0 - 1.0, params.face_id); + + //vec4 color = color_interp; + + if (params.use_direct_write) { + frag_color = vec4(texture(source_cube, N).rgb, 1.0); + } else { + vec4 sum = vec4(0.0, 0.0, 0.0, 0.0); + + float solid_angle_texel = 4.0 * M_PI / (6.0 * params.face_size * params.face_size); + float roughness2 = params.roughness * params.roughness; + float roughness4 = roughness2 * roughness2; + vec3 UpVector = abs(N.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(1.0, 0.0, 0.0); + mat3 T; + T[0] = normalize(cross(UpVector, N)); + T[1] = cross(N, T[0]); + T[2] = N; + + for (uint sampleNum = 0u; sampleNum < params.sample_count; sampleNum++) { + vec2 xi = Hammersley(sampleNum, params.sample_count); + + vec3 H = T * ImportanceSampleGGX(xi, roughness4); + float NdotH = dot(N, H); + vec3 L = (2.0 * NdotH * H - N); + + float ndotl = clamp(dot(N, L), 0.0, 1.0); + + if (ndotl > 0.0) { + float D = DistributionGGX(NdotH, roughness4); + float pdf = D * NdotH / (4.0 * NdotH) + 0.0001; + + float solid_angle_sample = 1.0 / (float(params.sample_count) * pdf + 0.0001); + + float mipLevel = params.roughness == 0.0 ? 0.0 : 0.5 * log2(solid_angle_sample / solid_angle_texel); + + sum.rgb += textureLod(source_cube, L, mipLevel).rgb * ndotl; + sum.a += ndotl; + } + } + sum /= sum.a; + + frag_color = vec4(sum.rgb, 1.0); + } +} diff --git a/servers/rendering/renderer_rd/shaders/effects/fsr_upscale.glsl b/servers/rendering/renderer_rd/shaders/effects/fsr_upscale.glsl new file mode 100644 index 0000000000..c8eb78a2f0 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/fsr_upscale.glsl @@ -0,0 +1,173 @@ +/*************************************************************************/ +/* fsr_upscale.glsl */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#[compute] + +#version 450 + +#VERSION_DEFINES + +#define A_GPU +#define A_GLSL + +#ifdef MODE_FSR_UPSCALE_NORMAL + +#define A_HALF + +#endif + +#include "thirdparty/amd-fsr/ffx_a.h" + +layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in; + +layout(rgba16f, set = 1, binding = 0) uniform restrict writeonly image2D fsr_image; +layout(set = 0, binding = 0) uniform sampler2D source_image; + +#define FSR_UPSCALE_PASS_TYPE_EASU 0 +#define FSR_UPSCALE_PASS_TYPE_RCAS 1 + +layout(push_constant, std430) uniform Params { + float resolution_width; + float resolution_height; + float upscaled_width; + float upscaled_height; + float sharpness; + int pass; +} +params; + +AU4 Const0, Const1, Const2, Const3; + +#ifdef MODE_FSR_UPSCALE_FALLBACK + +#define FSR_EASU_F +AF4 FsrEasuRF(AF2 p) { + AF4 res = textureGather(source_image, p, 0); + return res; +} +AF4 FsrEasuGF(AF2 p) { + AF4 res = textureGather(source_image, p, 1); + return res; +} +AF4 FsrEasuBF(AF2 p) { + AF4 res = textureGather(source_image, p, 2); + return res; +} + +#define FSR_RCAS_F +AF4 FsrRcasLoadF(ASU2 p) { + return AF4(texelFetch(source_image, ASU2(p), 0)); +} +void FsrRcasInputF(inout AF1 r, inout AF1 g, inout AF1 b) {} + +#else + +#define FSR_EASU_H +AH4 FsrEasuRH(AF2 p) { + AH4 res = AH4(textureGather(source_image, p, 0)); + return res; +} +AH4 FsrEasuGH(AF2 p) { + AH4 res = AH4(textureGather(source_image, p, 1)); + return res; +} +AH4 FsrEasuBH(AF2 p) { + AH4 res = AH4(textureGather(source_image, p, 2)); + return res; +} + +#define FSR_RCAS_H +AH4 FsrRcasLoadH(ASW2 p) { + return AH4(texelFetch(source_image, ASU2(p), 0)); +} +void FsrRcasInputH(inout AH1 r, inout AH1 g, inout AH1 b) {} + +#endif + +#include "thirdparty/amd-fsr/ffx_fsr1.h" + +void fsr_easu_pass(AU2 pos) { +#ifdef MODE_FSR_UPSCALE_NORMAL + + AH3 Gamma2Color = AH3(0, 0, 0); + FsrEasuH(Gamma2Color, pos, Const0, Const1, Const2, Const3); + imageStore(fsr_image, ASU2(pos), AH4(Gamma2Color, 1)); + +#else + + AF3 Gamma2Color = AF3(0, 0, 0); + FsrEasuF(Gamma2Color, pos, Const0, Const1, Const2, Const3); + imageStore(fsr_image, ASU2(pos), AF4(Gamma2Color, 1)); + +#endif +} + +void fsr_rcas_pass(AU2 pos) { +#ifdef MODE_FSR_UPSCALE_NORMAL + + AH3 Gamma2Color = AH3(0, 0, 0); + FsrRcasH(Gamma2Color.r, Gamma2Color.g, Gamma2Color.b, pos, Const0); + imageStore(fsr_image, ASU2(pos), AH4(Gamma2Color, 1)); + +#else + + AF3 Gamma2Color = AF3(0, 0, 0); + FsrRcasF(Gamma2Color.r, Gamma2Color.g, Gamma2Color.b, pos, Const0); + imageStore(fsr_image, ASU2(pos), AF4(Gamma2Color, 1)); + +#endif +} + +void fsr_pass(AU2 pos) { + if (params.pass == FSR_UPSCALE_PASS_TYPE_EASU) { + fsr_easu_pass(pos); + } else if (params.pass == FSR_UPSCALE_PASS_TYPE_RCAS) { + fsr_rcas_pass(pos); + } +} + +void main() { + // Clang does not like unused functions. If ffx_a.h is included in the binary, clang will throw a fit and not compile so we must configure FSR in this shader + if (params.pass == FSR_UPSCALE_PASS_TYPE_EASU) { + FsrEasuCon(Const0, Const1, Const2, Const3, params.resolution_width, params.resolution_height, params.resolution_width, params.resolution_height, params.upscaled_width, params.upscaled_height); + } else if (params.pass == FSR_UPSCALE_PASS_TYPE_RCAS) { + FsrRcasCon(Const0, params.sharpness); + } + + AU2 gxy = ARmp8x8(gl_LocalInvocationID.x) + AU2(gl_WorkGroupID.x << 4u, gl_WorkGroupID.y << 4u); + + fsr_pass(gxy); + gxy.x += 8u; + fsr_pass(gxy); + gxy.y += 8u; + fsr_pass(gxy); + gxy.x -= 8u; + fsr_pass(gxy); +} diff --git a/servers/rendering/renderer_rd/shaders/effects/resolve.glsl b/servers/rendering/renderer_rd/shaders/effects/resolve.glsl new file mode 100644 index 0000000000..0e086331c0 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/resolve.glsl @@ -0,0 +1,236 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +#ifdef MODE_RESOLVE_DEPTH +layout(set = 0, binding = 0) uniform sampler2DMS source_depth; +layout(r32f, set = 1, binding = 0) uniform restrict writeonly image2D dest_depth; +#endif + +#ifdef MODE_RESOLVE_GI +layout(set = 0, binding = 0) uniform sampler2DMS source_depth; +layout(set = 0, binding = 1) uniform sampler2DMS source_normal_roughness; + +layout(r32f, set = 1, binding = 0) uniform restrict writeonly image2D dest_depth; +layout(rgba8, set = 1, binding = 1) uniform restrict writeonly image2D dest_normal_roughness; + +#ifdef VOXEL_GI_RESOLVE +layout(set = 2, binding = 0) uniform usampler2DMS source_voxel_gi; +layout(rg8ui, set = 3, binding = 0) uniform restrict writeonly uimage2D dest_voxel_gi; +#endif + +#endif + +layout(push_constant, std430) uniform Params { + ivec2 screen_size; + int sample_count; + uint pad; +} +params; + +void main() { + // Pixel being shaded + ivec2 pos = ivec2(gl_GlobalInvocationID.xy); + if (any(greaterThanEqual(pos, params.screen_size))) { //too large, do nothing + return; + } + +#ifdef MODE_RESOLVE_DEPTH + + float depth_avg = 0.0; + for (int i = 0; i < params.sample_count; i++) { + depth_avg += texelFetch(source_depth, pos, i).r; + } + depth_avg /= float(params.sample_count); + imageStore(dest_depth, pos, vec4(depth_avg)); + +#endif + +#ifdef MODE_RESOLVE_GI + + float best_depth = 1e20; + vec4 best_normal_roughness = vec4(0.0); +#ifdef VOXEL_GI_RESOLVE + uvec2 best_voxel_gi; +#endif + +#if 0 + + for(int i=0;i<params.sample_count;i++) { + float depth = texelFetch(source_depth,pos,i).r; + if (depth < best_depth) { //use the depth closest to camera + best_depth = depth; + best_normal_roughness = texelFetch(source_normal_roughness,pos,i); + +#ifdef VOXEL_GI_RESOLVE + best_voxel_gi = texelFetch(source_voxel_gi,pos,i).rg; +#endif + } + } + +#else + +#if 1 + + vec4 group1; + vec4 group2; + vec4 group3; + vec4 group4; + int best_index = 0; + + //2X + group1.x = texelFetch(source_depth, pos, 0).r; + group1.y = texelFetch(source_depth, pos, 1).r; + + //4X + if (params.sample_count >= 4) { + group1.z = texelFetch(source_depth, pos, 2).r; + group1.w = texelFetch(source_depth, pos, 3).r; + } + //8X + if (params.sample_count >= 8) { + group2.x = texelFetch(source_depth, pos, 4).r; + group2.y = texelFetch(source_depth, pos, 5).r; + group2.z = texelFetch(source_depth, pos, 6).r; + group2.w = texelFetch(source_depth, pos, 7).r; + } + //16X + if (params.sample_count >= 16) { + group3.x = texelFetch(source_depth, pos, 8).r; + group3.y = texelFetch(source_depth, pos, 9).r; + group3.z = texelFetch(source_depth, pos, 10).r; + group3.w = texelFetch(source_depth, pos, 11).r; + + group4.x = texelFetch(source_depth, pos, 12).r; + group4.y = texelFetch(source_depth, pos, 13).r; + group4.z = texelFetch(source_depth, pos, 14).r; + group4.w = texelFetch(source_depth, pos, 15).r; + } + + if (params.sample_count == 2) { + best_index = (pos.x & 1) ^ ((pos.y >> 1) & 1); //not much can be done here + } else if (params.sample_count == 4) { + vec4 freq = vec4(equal(group1, vec4(group1.x))); + freq += vec4(equal(group1, vec4(group1.y))); + freq += vec4(equal(group1, vec4(group1.z))); + freq += vec4(equal(group1, vec4(group1.w))); + + float min_f = freq.x; + best_index = 0; + if (freq.y < min_f) { + best_index = 1; + min_f = freq.y; + } + if (freq.z < min_f) { + best_index = 2; + min_f = freq.z; + } + if (freq.w < min_f) { + best_index = 3; + } + } else if (params.sample_count == 8) { + vec4 freq0 = vec4(equal(group1, vec4(group1.x))); + vec4 freq1 = vec4(equal(group2, vec4(group1.x))); + freq0 += vec4(equal(group1, vec4(group1.y))); + freq1 += vec4(equal(group2, vec4(group1.y))); + freq0 += vec4(equal(group1, vec4(group1.z))); + freq1 += vec4(equal(group2, vec4(group1.z))); + freq0 += vec4(equal(group1, vec4(group1.w))); + freq1 += vec4(equal(group2, vec4(group1.w))); + freq0 += vec4(equal(group1, vec4(group2.x))); + freq1 += vec4(equal(group2, vec4(group2.x))); + freq0 += vec4(equal(group1, vec4(group2.y))); + freq1 += vec4(equal(group2, vec4(group2.y))); + freq0 += vec4(equal(group1, vec4(group2.z))); + freq1 += vec4(equal(group2, vec4(group2.z))); + freq0 += vec4(equal(group1, vec4(group2.w))); + freq1 += vec4(equal(group2, vec4(group2.w))); + + float min_f0 = freq0.x; + int best_index0 = 0; + if (freq0.y < min_f0) { + best_index0 = 1; + min_f0 = freq0.y; + } + if (freq0.z < min_f0) { + best_index0 = 2; + min_f0 = freq0.z; + } + if (freq0.w < min_f0) { + best_index0 = 3; + min_f0 = freq0.w; + } + + float min_f1 = freq1.x; + int best_index1 = 4; + if (freq1.y < min_f1) { + best_index1 = 5; + min_f1 = freq1.y; + } + if (freq1.z < min_f1) { + best_index1 = 6; + min_f1 = freq1.z; + } + if (freq1.w < min_f1) { + best_index1 = 7; + min_f1 = freq1.w; + } + + best_index = mix(best_index0, best_index1, min_f0 < min_f1); + } + +#else + float depths[16]; + int depth_indices[16]; + int depth_amount[16]; + int depth_count = 0; + + for (int i = 0; i < params.sample_count; i++) { + float depth = texelFetch(source_depth, pos, i).r; + int depth_index = -1; + for (int j = 0; j < depth_count; j++) { + if (abs(depths[j] - depth) < 0.000001) { + depth_index = j; + break; + } + } + + if (depth_index == -1) { + depths[depth_count] = depth; + depth_indices[depth_count] = i; + depth_amount[depth_count] = 1; + depth_count += 1; + } else { + depth_amount[depth_index] += 1; + } + } + + int depth_least = 0xFFFF; + int best_index = 0; + for (int j = 0; j < depth_count; j++) { + if (depth_amount[j] < depth_least) { + best_index = depth_indices[j]; + depth_least = depth_amount[j]; + } + } +#endif + best_depth = texelFetch(source_depth, pos, best_index).r; + best_normal_roughness = texelFetch(source_normal_roughness, pos, best_index); +#ifdef VOXEL_GI_RESOLVE + best_voxel_gi = texelFetch(source_voxel_gi, pos, best_index).rg; +#endif + +#endif + + imageStore(dest_depth, pos, vec4(best_depth)); + imageStore(dest_normal_roughness, pos, vec4(best_normal_roughness)); +#ifdef VOXEL_GI_RESOLVE + imageStore(dest_voxel_gi, pos, uvec4(best_voxel_gi, 0, 0)); +#endif + +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection.glsl b/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection.glsl new file mode 100644 index 0000000000..9d8294a7da --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection.glsl @@ -0,0 +1,268 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +layout(rgba16f, set = 0, binding = 0) uniform restrict readonly image2D source_diffuse; +layout(r32f, set = 0, binding = 1) uniform restrict readonly image2D source_depth; +layout(rgba16f, set = 1, binding = 0) uniform restrict writeonly image2D ssr_image; +#ifdef MODE_ROUGH +layout(r8, set = 1, binding = 1) uniform restrict writeonly image2D blur_radius_image; +#endif +layout(rgba8, set = 2, binding = 0) uniform restrict readonly image2D source_normal_roughness; +layout(set = 3, binding = 0) uniform sampler2D source_metallic; + +layout(push_constant, std430) uniform Params { + vec4 proj_info; + + ivec2 screen_size; + float camera_z_near; + float camera_z_far; + + int num_steps; + float depth_tolerance; + float distance_fade; + float curve_fade_in; + + bool orthogonal; + float filter_mipmap_levels; + bool use_half_res; + uint view_index; +} +params; + +#include "screen_space_reflection_inc.glsl" + +vec2 view_to_screen(vec3 view_pos, out float w) { + vec4 projected = scene_data.projection[params.view_index] * vec4(view_pos, 1.0); + projected.xyz /= projected.w; + projected.xy = projected.xy * 0.5 + 0.5; + w = projected.w; + return projected.xy; +} + +#define M_PI 3.14159265359 + +void main() { + // Pixel being shaded + ivec2 ssC = ivec2(gl_GlobalInvocationID.xy); + + if (any(greaterThanEqual(ssC.xy, params.screen_size))) { //too large, do nothing + return; + } + + vec2 pixel_size = 1.0 / vec2(params.screen_size); + vec2 uv = vec2(ssC.xy) * pixel_size; + + uv += pixel_size * 0.5; + + float base_depth = imageLoad(source_depth, ssC).r; + + // World space point being shaded + vec3 vertex = reconstructCSPosition(uv * vec2(params.screen_size), base_depth); + + vec4 normal_roughness = imageLoad(source_normal_roughness, ssC); + vec3 normal = normal_roughness.xyz * 2.0 - 1.0; + normal = normalize(normal); + normal.y = -normal.y; //because this code reads flipped + + vec3 view_dir; + if (sc_multiview) { + view_dir = normalize(vertex + scene_data.eye_offset[params.view_index].xyz); + } else { + view_dir = normalize(vertex); + } + vec3 ray_dir = normalize(reflect(view_dir, normal)); + + if (dot(ray_dir, normal) < 0.001) { + imageStore(ssr_image, ssC, vec4(0.0)); + return; + } + //ray_dir = normalize(view_dir - normal * dot(normal,view_dir) * 2.0); + //ray_dir = normalize(vec3(1.0, 1.0, -1.0)); + + //////////////// + + // make ray length and clip it against the near plane (don't want to trace beyond visible) + float ray_len = (vertex.z + ray_dir.z * params.camera_z_far) > -params.camera_z_near ? (-params.camera_z_near - vertex.z) / ray_dir.z : params.camera_z_far; + vec3 ray_end = vertex + ray_dir * ray_len; + + float w_begin; + vec2 vp_line_begin = view_to_screen(vertex, w_begin); + float w_end; + vec2 vp_line_end = view_to_screen(ray_end, w_end); + vec2 vp_line_dir = vp_line_end - vp_line_begin; + + // we need to interpolate w along the ray, to generate perspective correct reflections + w_begin = 1.0 / w_begin; + w_end = 1.0 / w_end; + + float z_begin = vertex.z * w_begin; + float z_end = ray_end.z * w_end; + + vec2 line_begin = vp_line_begin / pixel_size; + vec2 line_dir = vp_line_dir / pixel_size; + float z_dir = z_end - z_begin; + float w_dir = w_end - w_begin; + + // clip the line to the viewport edges + + float scale_max_x = min(1.0, 0.99 * (1.0 - vp_line_begin.x) / max(1e-5, vp_line_dir.x)); + float scale_max_y = min(1.0, 0.99 * (1.0 - vp_line_begin.y) / max(1e-5, vp_line_dir.y)); + float scale_min_x = min(1.0, 0.99 * vp_line_begin.x / max(1e-5, -vp_line_dir.x)); + float scale_min_y = min(1.0, 0.99 * vp_line_begin.y / max(1e-5, -vp_line_dir.y)); + float line_clip = min(scale_max_x, scale_max_y) * min(scale_min_x, scale_min_y); + line_dir *= line_clip; + z_dir *= line_clip; + w_dir *= line_clip; + + // clip z and w advance to line advance + vec2 line_advance = normalize(line_dir); // down to pixel + float step_size = 1.0 / length(line_dir); + float z_advance = z_dir * step_size; // adapt z advance to line advance + float w_advance = w_dir * step_size; // adapt w advance to line advance + + // make line advance faster if direction is closer to pixel edges (this avoids sampling the same pixel twice) + float advance_angle_adj = 1.0 / max(abs(line_advance.x), abs(line_advance.y)); + line_advance *= advance_angle_adj; // adapt z advance to line advance + z_advance *= advance_angle_adj; + w_advance *= advance_angle_adj; + + vec2 pos = line_begin; + float z = z_begin; + float w = w_begin; + float z_from = z / w; + float z_to = z_from; + float depth; + vec2 prev_pos = pos; + + if (ivec2(pos + line_advance - 0.5) == ssC) { + // It is possible for rounding to cause our first pixel to check to be the pixel we're reflecting. + // Make sure we skip it + pos += line_advance; + z += z_advance; + w += w_advance; + } + + bool found = false; + + float steps_taken = 0.0; + + for (int i = 0; i < params.num_steps; i++) { + pos += line_advance; + z += z_advance; + w += w_advance; + + // convert to linear depth + ivec2 test_pos = ivec2(pos - 0.5); + depth = imageLoad(source_depth, test_pos).r; + if (sc_multiview) { + depth = depth * 2.0 - 1.0; + depth = 2.0 * params.camera_z_near * params.camera_z_far / (params.camera_z_far + params.camera_z_near - depth * (params.camera_z_far - params.camera_z_near)); + depth = -depth; + } + + z_from = z_to; + z_to = z / w; + + if (depth > z_to) { + // Test if our ray is hitting the "right" side of the surface, if not we're likely self reflecting and should skip. + vec4 test_normal_roughness = imageLoad(source_normal_roughness, test_pos); + vec3 test_normal = test_normal_roughness.xyz * 2.0 - 1.0; + test_normal = normalize(test_normal); + test_normal.y = -test_normal.y; //because this code reads flipped + + if (dot(ray_dir, test_normal) < 0.001) { + // if depth was surpassed + if (depth <= max(z_to, z_from) + params.depth_tolerance && -depth < params.camera_z_far * 0.95) { + // check the depth tolerance and far clip + // check that normal is valid + found = true; + } + break; + } + } + + steps_taken += 1.0; + prev_pos = pos; + } + + if (found) { + float margin_blend = 1.0; + + vec2 margin = vec2((params.screen_size.x + params.screen_size.y) * 0.05); // make a uniform margin + if (any(bvec4(lessThan(pos, vec2(0.0, 0.0)), greaterThan(pos, params.screen_size)))) { + // clip at the screen edges + imageStore(ssr_image, ssC, vec4(0.0)); + return; + } + + { + //blend fading out towards inner margin + // 0.5 = midpoint of reflection + vec2 margin_grad = mix(params.screen_size - pos, pos, lessThan(pos, params.screen_size * 0.5)); + margin_blend = smoothstep(0.0, margin.x * margin.y, margin_grad.x * margin_grad.y); + //margin_blend = 1.0; + } + + vec2 final_pos; + float grad = (steps_taken + 1.0) / float(params.num_steps); + float initial_fade = params.curve_fade_in == 0.0 ? 1.0 : pow(clamp(grad, 0.0, 1.0), params.curve_fade_in); + float fade = pow(clamp(1.0 - grad, 0.0, 1.0), params.distance_fade) * initial_fade; + final_pos = pos; + + vec4 final_color; + +#ifdef MODE_ROUGH + + // if roughness is enabled, do screen space cone tracing + float blur_radius = 0.0; + float roughness = normal_roughness.w; + + if (roughness > 0.001) { + float cone_angle = min(roughness, 0.999) * M_PI * 0.5; + float cone_len = length(final_pos - line_begin); + float op_len = 2.0 * tan(cone_angle) * cone_len; // opposite side of iso triangle + { + // fit to sphere inside cone (sphere ends at end of cone), something like this: + // ___ + // \O/ + // V + // + // as it avoids bleeding from beyond the reflection as much as possible. As a plus + // it also makes the rough reflection more elongated. + float a = op_len; + float h = cone_len; + float a2 = a * a; + float fh2 = 4.0f * h * h; + blur_radius = (a * (sqrt(a2 + fh2) - a)) / (4.0f * h); + } + } + + imageStore(blur_radius_image, ssC, vec4(blur_radius / 255.0)); //stored in r8 + +#endif // MODE_ROUGH + + final_color = vec4(imageLoad(source_diffuse, ivec2(final_pos - 0.5)).rgb, fade * margin_blend); + + // Schlick term. + float metallic = texelFetch(source_metallic, ssC << 1, 0).w; + float f0 = mix(0.04, 1.0, metallic); // Assume a "specular" amount of 0.5 + normal.y = -normal.y; + float m = clamp(1.0 - dot(normalize(normal), -view_dir), 0.0, 1.0); + float m2 = m * m; + m = m2 * m2 * m; // pow(m,5) + final_color.a *= f0 + (1.0 - f0) * m; // Fresnel Schlick term. + + imageStore(ssr_image, ssC, final_color); + + } else { +#ifdef MODE_ROUGH + imageStore(blur_radius_image, ssC, vec4(0.0)); +#endif + imageStore(ssr_image, ssC, vec4(0.0)); + } +} diff --git a/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection_filter.glsl b/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection_filter.glsl new file mode 100644 index 0000000000..a63d60e0b2 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection_filter.glsl @@ -0,0 +1,148 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +layout(rgba16f, set = 0, binding = 0) uniform restrict readonly image2D source_ssr; +layout(r8, set = 0, binding = 1) uniform restrict readonly image2D source_radius; +layout(rgba8, set = 1, binding = 0) uniform restrict readonly image2D source_normal; + +layout(rgba16f, set = 2, binding = 0) uniform restrict writeonly image2D dest_ssr; +#ifndef VERTICAL_PASS +layout(r8, set = 2, binding = 1) uniform restrict writeonly image2D dest_radius; +#endif +layout(r32f, set = 3, binding = 0) uniform restrict readonly image2D source_depth; + +layout(push_constant, std430) uniform Params { + vec4 proj_info; + + bool orthogonal; + float edge_tolerance; + int increment; + uint view_index; + + ivec2 screen_size; + bool vertical; + uint steps; +} +params; + +#include "screen_space_reflection_inc.glsl" + +#define GAUSS_TABLE_SIZE 15 + +const float gauss_table[GAUSS_TABLE_SIZE + 1] = float[]( + 0.1847392078702266, + 0.16595854345772326, + 0.12031364177766891, + 0.07038755277896766, + 0.03322925565155569, + 0.012657819729901945, + 0.0038903040680094217, + 0.0009646503390864025, + 0.00019297087402915717, + 0.000031139936308099136, + 0.000004053309048174758, + 4.255228059965837e-7, + 3.602517634249573e-8, + 2.4592560765896795e-9, + 1.3534945386863618e-10, + 0.0 //one more for interpolation +); + +float gauss_weight(float p_val) { + float idxf; + float c = modf(max(0.0, p_val * float(GAUSS_TABLE_SIZE)), idxf); + int idx = int(idxf); + if (idx >= GAUSS_TABLE_SIZE + 1) { + return 0.0; + } + + return mix(gauss_table[idx], gauss_table[idx + 1], c); +} + +#define M_PI 3.14159265359 + +void do_filter(inout vec4 accum, inout float accum_radius, inout float divisor, ivec2 texcoord, ivec2 increment, vec3 p_pos, vec3 normal, float p_limit_radius) { + for (int i = 1; i < params.steps; i++) { + float d = float(i * params.increment); + ivec2 tc = texcoord + increment * i; + float depth = imageLoad(source_depth, tc).r; + vec3 view_pos = reconstructCSPosition(vec2(tc) + 0.5, depth); + vec3 view_normal = normalize(imageLoad(source_normal, tc).rgb * 2.0 - 1.0); + view_normal.y = -view_normal.y; + + float r = imageLoad(source_radius, tc).r; + float radius = round(r * 255.0); + + float angle_n = 1.0 - abs(dot(normal, view_normal)); + if (angle_n > params.edge_tolerance) { + break; + } + + float angle = abs(dot(normal, normalize(view_pos - p_pos))); + + if (angle > params.edge_tolerance) { + break; + } + + if (d < radius) { + float w = gauss_weight(d / radius); + accum += imageLoad(source_ssr, tc) * w; +#ifndef VERTICAL_PASS + accum_radius += r * w; +#endif + divisor += w; + } + } +} + +void main() { + // Pixel being shaded + ivec2 ssC = ivec2(gl_GlobalInvocationID.xy); + + if (any(greaterThanEqual(ssC.xy, params.screen_size))) { //too large, do nothing + return; + } + + float base_contrib = gauss_table[0]; + + vec4 accum = imageLoad(source_ssr, ssC); + + float accum_radius = imageLoad(source_radius, ssC).r; + float radius = accum_radius * 255.0; + + float divisor = gauss_table[0]; + accum *= divisor; + accum_radius *= divisor; +#ifdef VERTICAL_PASS + ivec2 direction = ivec2(0, params.increment); +#else + ivec2 direction = ivec2(params.increment, 0); +#endif + float depth = imageLoad(source_depth, ssC).r; + vec3 pos = reconstructCSPosition(vec2(ssC.xy) + 0.5, depth); + vec3 normal = imageLoad(source_normal, ssC).xyz * 2.0 - 1.0; + normal = normalize(normal); + normal.y = -normal.y; + + do_filter(accum, accum_radius, divisor, ssC.xy, direction, pos, normal, radius); + do_filter(accum, accum_radius, divisor, ssC.xy, -direction, pos, normal, radius); + + if (divisor > 0.0) { + accum /= divisor; + accum_radius /= divisor; + } else { + accum = vec4(0.0); + accum_radius = 0.0; + } + + imageStore(dest_ssr, ssC, accum); + +#ifndef VERTICAL_PASS + imageStore(dest_radius, ssC, vec4(accum_radius)); +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection_inc.glsl b/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection_inc.glsl new file mode 100644 index 0000000000..26405ab040 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection_inc.glsl @@ -0,0 +1,28 @@ +layout(constant_id = 0) const bool sc_multiview = false; + +layout(set = 4, binding = 0, std140) uniform SceneData { + mat4x4 projection[2]; + mat4x4 inv_projection[2]; + vec4 eye_offset[2]; +} +scene_data; + +vec3 reconstructCSPosition(vec2 screen_pos, float z) { + if (sc_multiview) { + vec4 pos; + pos.xy = (2.0 * vec2(screen_pos) / vec2(params.screen_size)) - 1.0; + pos.z = z * 2.0 - 1.0; + pos.w = 1.0; + + pos = scene_data.inv_projection[params.view_index] * pos; + pos.xyz /= pos.w; + + return pos.xyz; + } else { + if (params.orthogonal) { + return vec3((screen_pos.xy * params.proj_info.xy + params.proj_info.zw), z); + } else { + return vec3((screen_pos.xy * params.proj_info.xy + params.proj_info.zw) * z, z); + } + } +} diff --git a/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection_scale.glsl b/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection_scale.glsl new file mode 100644 index 0000000000..a7da0812df --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection_scale.glsl @@ -0,0 +1,106 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +/* Specialization Constants (Toggles) */ + +layout(constant_id = 0) const bool sc_multiview = false; + +/* inputs */ +layout(set = 0, binding = 0) uniform sampler2D source_ssr; +layout(set = 1, binding = 0) uniform sampler2D source_depth; +layout(set = 1, binding = 1) uniform sampler2D source_normal; +layout(rgba16f, set = 2, binding = 0) uniform restrict writeonly image2D dest_ssr; +layout(r32f, set = 3, binding = 0) uniform restrict writeonly image2D dest_depth; +layout(rgba8, set = 3, binding = 1) uniform restrict writeonly image2D dest_normal; + +layout(push_constant, std430) uniform Params { + ivec2 screen_size; + float camera_z_near; + float camera_z_far; + + bool orthogonal; + bool filtered; + uint pad[2]; +} +params; + +void main() { + // Pixel being shaded + ivec2 ssC = ivec2(gl_GlobalInvocationID.xy); + + if (any(greaterThanEqual(ssC.xy, params.screen_size))) { //too large, do nothing + return; + } + //do not filter, SSR will generate arctifacts if this is done + + float divisor = 0.0; + vec4 color; + float depth; + vec4 normal; + + if (params.filtered) { + color = vec4(0.0); + depth = 0.0; + normal = vec4(0.0); + + for (int i = 0; i < 4; i++) { + ivec2 ofs = ssC << 1; + if (bool(i & 1)) { + ofs.x += 1; + } + if (bool(i & 2)) { + ofs.y += 1; + } + color += texelFetch(source_ssr, ofs, 0); + float d = texelFetch(source_depth, ofs, 0).r; + vec4 nr = texelFetch(source_normal, ofs, 0); + normal.xyz += nr.xyz * 2.0 - 1.0; + normal.w += nr.w; + + if (sc_multiview) { + // we're doing a full unproject so we need the value as is. + depth += d; + } else { + // unproject our Z value so we can use it directly. + d = d * 2.0 - 1.0; + if (params.orthogonal) { + d = ((d + (params.camera_z_far + params.camera_z_near) / (params.camera_z_far - params.camera_z_near)) * (params.camera_z_far - params.camera_z_near)) / 2.0; + } else { + d = 2.0 * params.camera_z_near * params.camera_z_far / (params.camera_z_far + params.camera_z_near - d * (params.camera_z_far - params.camera_z_near)); + } + depth += -d; + } + } + + color /= 4.0; + depth /= 4.0; + normal.xyz = normalize(normal.xyz / 4.0) * 0.5 + 0.5; + normal.w /= 4.0; + } else { + ivec2 ofs = ssC << 1; + + color = texelFetch(source_ssr, ofs, 0); + depth = texelFetch(source_depth, ofs, 0).r; + normal = texelFetch(source_normal, ofs, 0); + + if (!sc_multiview) { + // unproject our Z value so we can use it directly. + depth = depth * 2.0 - 1.0; + if (params.orthogonal) { + depth = ((depth + (params.camera_z_far + params.camera_z_near) / (params.camera_z_far - params.camera_z_near)) * (params.camera_z_far - params.camera_z_near)) / 2.0; + } else { + depth = 2.0 * params.camera_z_near * params.camera_z_far / (params.camera_z_far + params.camera_z_near - depth * (params.camera_z_far - params.camera_z_near)); + } + depth = -depth; + } + } + + imageStore(dest_ssr, ssC, color); + imageStore(dest_depth, ssC, vec4(depth)); + imageStore(dest_normal, ssC, normal); +} diff --git a/servers/rendering/renderer_rd/shaders/effects/specular_merge.glsl b/servers/rendering/renderer_rd/shaders/effects/specular_merge.glsl new file mode 100644 index 0000000000..c62144fdaf --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/specular_merge.glsl @@ -0,0 +1,112 @@ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#if defined(USE_MULTIVIEW) && defined(has_VK_KHR_multiview) +#extension GL_EXT_multiview : enable +#endif + +#ifdef USE_MULTIVIEW +#ifdef has_VK_KHR_multiview +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +// !BAS! This needs to become an input once we implement our fallback! +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#else // USE_MULTIVIEW +// Set to zero, not supported in non stereo +#define ViewIndex 0 +#endif //USE_MULTIVIEW + +#ifdef USE_MULTIVIEW +layout(location = 0) out vec3 uv_interp; +#else // USE_MULTIVIEW +layout(location = 0) out vec2 uv_interp; +#endif //USE_MULTIVIEW + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + +#ifdef USE_MULTIVIEW + uv_interp = vec3(base_arr[gl_VertexIndex], ViewIndex); + + gl_Position = vec4(uv_interp.xy * 2.0 - 1.0, 0.0, 1.0); +#else + uv_interp = base_arr[gl_VertexIndex]; + + gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0); +#endif +} + +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#if defined(USE_MULTIVIEW) && defined(has_VK_KHR_multiview) +#extension GL_EXT_multiview : enable +#endif + +#ifdef USE_MULTIVIEW +#ifdef has_VK_KHR_multiview +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +// !BAS! This needs to become an input once we implement our fallback! +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#else // USE_MULTIVIEW +// Set to zero, not supported in non stereo +#define ViewIndex 0 +#endif //USE_MULTIVIEW + +#ifdef USE_MULTIVIEW +layout(location = 0) in vec3 uv_interp; +#else // USE_MULTIVIEW +layout(location = 0) in vec2 uv_interp; +#endif //USE_MULTIVIEW + +#ifdef USE_MULTIVIEW +layout(set = 0, binding = 0) uniform sampler2DArray specular; +#else // USE_MULTIVIEW +layout(set = 0, binding = 0) uniform sampler2D specular; +#endif //USE_MULTIVIEW + +#ifdef MODE_SSR + +#ifdef USE_MULTIVIEW +layout(set = 1, binding = 0) uniform sampler2DArray ssr; +#else // USE_MULTIVIEW +layout(set = 1, binding = 0) uniform sampler2D ssr; +#endif //USE_MULTIVIEW + +#endif + +#ifdef MODE_MERGE + +#ifdef USE_MULTIVIEW +layout(set = 2, binding = 0) uniform sampler2DArray diffuse; +#else // USE_MULTIVIEW +layout(set = 2, binding = 0) uniform sampler2D diffuse; +#endif //USE_MULTIVIEW + +#endif + +layout(location = 0) out vec4 frag_color; + +void main() { + frag_color.rgb = texture(specular, uv_interp).rgb; + frag_color.a = 0.0; +#ifdef MODE_SSR + + vec4 ssr_color = texture(ssr, uv_interp); + frag_color.rgb = mix(frag_color.rgb, ssr_color.rgb, ssr_color.a); +#endif + +#ifdef MODE_MERGE + frag_color += texture(diffuse, uv_interp); +#endif + //added using additive blend +} diff --git a/servers/rendering/renderer_rd/shaders/effects/ss_effects_downsample.glsl b/servers/rendering/renderer_rd/shaders/effects/ss_effects_downsample.glsl new file mode 100644 index 0000000000..134aae5ce7 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/ss_effects_downsample.glsl @@ -0,0 +1,229 @@ +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// Copyright (c) 2016, Intel Corporation +// Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated +// documentation files (the "Software"), to deal in the Software without restriction, including without limitation +// the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to +// permit persons to whom the Software is furnished to do so, subject to the following conditions: +// The above copyright notice and this permission notice shall be included in all copies or substantial portions of +// the Software. +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO +// THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +// TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +// SOFTWARE. +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// File changes (yyyy-mm-dd) +// 2016-09-07: filip.strugar@intel.com: first commit +// 2020-12-05: clayjohn: convert to Vulkan and Godot +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// + +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +layout(push_constant, std430) uniform Params { + vec2 pixel_size; + float z_far; + float z_near; + bool orthogonal; + float radius_sq; + uvec2 pad; +} +params; + +layout(set = 0, binding = 0) uniform sampler2D source_depth; + +layout(r16f, set = 1, binding = 0) uniform restrict writeonly image2DArray dest_image0; //rename +#ifdef GENERATE_MIPS +layout(r16f, set = 2, binding = 0) uniform restrict writeonly image2DArray dest_image1; +layout(r16f, set = 2, binding = 1) uniform restrict writeonly image2DArray dest_image2; +layout(r16f, set = 2, binding = 2) uniform restrict writeonly image2DArray dest_image3; +#ifdef GENERATE_FULL_MIPS +layout(r16f, set = 2, binding = 3) uniform restrict writeonly image2DArray dest_image4; +#endif +#endif + +vec4 screen_space_to_view_space_depth(vec4 p_depth) { + if (params.orthogonal) { + vec4 depth = p_depth * 2.0 - 1.0; + return ((depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0; + } + + float depth_linearize_mul = params.z_near; + float depth_linearize_add = params.z_far; + + // Optimised version of "-cameraClipNear / (cameraClipFar - projDepth * (cameraClipFar - cameraClipNear)) * cameraClipFar" + + // Set your depth_linearize_mul and depth_linearize_add to: + // depth_linearize_mul = ( cameraClipFar * cameraClipNear) / ( cameraClipFar - cameraClipNear ); + // depth_linearize_add = cameraClipFar / ( cameraClipFar - cameraClipNear ); + + return depth_linearize_mul / (depth_linearize_add - p_depth); +} + +float screen_space_to_view_space_depth(float p_depth) { + if (params.orthogonal) { + float depth = p_depth * 2.0 - 1.0; + return ((depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / (2.0 * params.z_far); + } + + float depth_linearize_mul = params.z_near; + float depth_linearize_add = params.z_far; + + return depth_linearize_mul / (depth_linearize_add - p_depth); +} + +#ifdef GENERATE_MIPS + +shared float depth_buffer[4][8][8]; + +float mip_smart_average(vec4 p_depths) { + float closest = min(min(p_depths.x, p_depths.y), min(p_depths.z, p_depths.w)); + float fallof_sq = -1.0f / params.radius_sq; + vec4 dists = p_depths - closest.xxxx; + vec4 weights = clamp(dists * dists * fallof_sq + 1.0, 0.0, 1.0); + return dot(weights, p_depths) / dot(weights, vec4(1.0, 1.0, 1.0, 1.0)); +} + +void prepare_depths_and_mips(vec4 p_samples, uvec2 p_output_coord, uvec2 p_gtid) { + p_samples = screen_space_to_view_space_depth(p_samples); + + depth_buffer[0][p_gtid.x][p_gtid.y] = p_samples.w; + depth_buffer[1][p_gtid.x][p_gtid.y] = p_samples.z; + depth_buffer[2][p_gtid.x][p_gtid.y] = p_samples.x; + depth_buffer[3][p_gtid.x][p_gtid.y] = p_samples.y; + + imageStore(dest_image0, ivec3(p_output_coord.x, p_output_coord.y, 0), vec4(p_samples.w)); + imageStore(dest_image0, ivec3(p_output_coord.x, p_output_coord.y, 1), vec4(p_samples.z)); + imageStore(dest_image0, ivec3(p_output_coord.x, p_output_coord.y, 2), vec4(p_samples.x)); + imageStore(dest_image0, ivec3(p_output_coord.x, p_output_coord.y, 3), vec4(p_samples.y)); + + uint depth_array_index = 2 * (p_gtid.y % 2) + (p_gtid.x % 2); + uvec2 depth_array_offset = ivec2(p_gtid.x % 2, p_gtid.y % 2); + ivec2 buffer_coord = ivec2(p_gtid) - ivec2(depth_array_offset); + + p_output_coord /= 2; + groupMemoryBarrier(); + barrier(); + + // if (still_alive) <-- all threads alive here + { + float sample_00 = depth_buffer[depth_array_index][buffer_coord.x + 0][buffer_coord.y + 0]; + float sample_01 = depth_buffer[depth_array_index][buffer_coord.x + 0][buffer_coord.y + 1]; + float sample_10 = depth_buffer[depth_array_index][buffer_coord.x + 1][buffer_coord.y + 0]; + float sample_11 = depth_buffer[depth_array_index][buffer_coord.x + 1][buffer_coord.y + 1]; + + float avg = mip_smart_average(vec4(sample_00, sample_01, sample_10, sample_11)); + imageStore(dest_image1, ivec3(p_output_coord.x, p_output_coord.y, depth_array_index), vec4(avg)); + depth_buffer[depth_array_index][buffer_coord.x][buffer_coord.y] = avg; + } + + bool still_alive = p_gtid.x % 4 == depth_array_offset.x && p_gtid.y % 4 == depth_array_offset.y; + + p_output_coord /= 2; + groupMemoryBarrier(); + barrier(); + + if (still_alive) { + float sample_00 = depth_buffer[depth_array_index][buffer_coord.x + 0][buffer_coord.y + 0]; + float sample_01 = depth_buffer[depth_array_index][buffer_coord.x + 0][buffer_coord.y + 2]; + float sample_10 = depth_buffer[depth_array_index][buffer_coord.x + 2][buffer_coord.y + 0]; + float sample_11 = depth_buffer[depth_array_index][buffer_coord.x + 2][buffer_coord.y + 2]; + + float avg = mip_smart_average(vec4(sample_00, sample_01, sample_10, sample_11)); + imageStore(dest_image2, ivec3(p_output_coord.x, p_output_coord.y, depth_array_index), vec4(avg)); + depth_buffer[depth_array_index][buffer_coord.x][buffer_coord.y] = avg; + } + + still_alive = p_gtid.x % 8 == depth_array_offset.x && depth_array_offset.y % 8 == depth_array_offset.y; + + p_output_coord /= 2; + groupMemoryBarrier(); + barrier(); + + if (still_alive) { + float sample_00 = depth_buffer[depth_array_index][buffer_coord.x + 0][buffer_coord.y + 0]; + float sample_01 = depth_buffer[depth_array_index][buffer_coord.x + 0][buffer_coord.y + 4]; + float sample_10 = depth_buffer[depth_array_index][buffer_coord.x + 4][buffer_coord.y + 0]; + float sample_11 = depth_buffer[depth_array_index][buffer_coord.x + 4][buffer_coord.y + 4]; + + float avg = mip_smart_average(vec4(sample_00, sample_01, sample_10, sample_11)); + imageStore(dest_image3, ivec3(p_output_coord.x, p_output_coord.y, depth_array_index), vec4(avg)); +#ifndef GENERATE_FULL_MIPS + } +#else + depth_buffer[depth_array_index][buffer_coord.x][buffer_coord.y] = avg; + } + still_alive = p_gtid.x % 16 == depth_array_offset.x && depth_array_offset.y % 16 == depth_array_offset.y; + + p_output_coord /= 2; + groupMemoryBarrier(); + barrier(); + + if (still_alive) { + float sample_00 = depth_buffer[depth_array_index][buffer_coord.x + 0][buffer_coord.y + 0]; + float sample_01 = depth_buffer[depth_array_index][buffer_coord.x + 0][buffer_coord.y + 8]; + float sample_10 = depth_buffer[depth_array_index][buffer_coord.x + 8][buffer_coord.y + 0]; + float sample_11 = depth_buffer[depth_array_index][buffer_coord.x + 8][buffer_coord.y + 8]; + + float avg = mip_smart_average(vec4(sample_00, sample_01, sample_10, sample_11)); + imageStore(dest_image4, ivec3(p_output_coord.x, p_output_coord.y, depth_array_index), vec4(avg)); + } +#endif +} +#else +#ifndef USE_HALF_BUFFERS +void prepare_depths(vec4 p_samples, uvec2 p_tid) { + p_samples = screen_space_to_view_space_depth(p_samples); + + imageStore(dest_image0, ivec3(p_tid, 0), vec4(p_samples.w)); + imageStore(dest_image0, ivec3(p_tid, 1), vec4(p_samples.z)); + imageStore(dest_image0, ivec3(p_tid, 2), vec4(p_samples.x)); + imageStore(dest_image0, ivec3(p_tid, 3), vec4(p_samples.y)); +} +#endif +#endif + +void main() { +#ifdef USE_HALF_BUFFERS +#ifdef USE_HALF_SIZE + float sample_00 = texelFetch(source_depth, ivec2(4 * gl_GlobalInvocationID.x + 0, 4 * gl_GlobalInvocationID.y + 0), 0).x; + float sample_11 = texelFetch(source_depth, ivec2(4 * gl_GlobalInvocationID.x + 2, 4 * gl_GlobalInvocationID.y + 2), 0).x; +#else + float sample_00 = texelFetch(source_depth, ivec2(2 * gl_GlobalInvocationID.x + 0, 2 * gl_GlobalInvocationID.y + 0), 0).x; + float sample_11 = texelFetch(source_depth, ivec2(2 * gl_GlobalInvocationID.x + 1, 2 * gl_GlobalInvocationID.y + 1), 0).x; +#endif + sample_00 = screen_space_to_view_space_depth(sample_00); + sample_11 = screen_space_to_view_space_depth(sample_11); + + imageStore(dest_image0, ivec3(gl_GlobalInvocationID.xy, 0), vec4(sample_00)); + imageStore(dest_image0, ivec3(gl_GlobalInvocationID.xy, 3), vec4(sample_11)); +#else //!USE_HALF_BUFFERS +#ifdef USE_HALF_SIZE + ivec2 depth_buffer_coord = 4 * ivec2(gl_GlobalInvocationID.xy); + ivec2 output_coord = ivec2(gl_GlobalInvocationID); + + vec2 uv = (vec2(depth_buffer_coord) + 0.5f) * params.pixel_size; + vec4 samples; + samples.x = textureLodOffset(source_depth, uv, 0, ivec2(0, 2)).x; + samples.y = textureLodOffset(source_depth, uv, 0, ivec2(2, 2)).x; + samples.z = textureLodOffset(source_depth, uv, 0, ivec2(2, 0)).x; + samples.w = textureLodOffset(source_depth, uv, 0, ivec2(0, 0)).x; +#else + ivec2 depth_buffer_coord = 2 * ivec2(gl_GlobalInvocationID.xy); + ivec2 output_coord = ivec2(gl_GlobalInvocationID); + + vec2 uv = (vec2(depth_buffer_coord) + 0.5f) * params.pixel_size; + vec4 samples = textureGather(source_depth, uv); +#endif +#ifdef GENERATE_MIPS + prepare_depths_and_mips(samples, output_coord, gl_LocalInvocationID.xy); +#else + prepare_depths(samples, gl_GlobalInvocationID.xy); +#endif +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/effects/ssao.glsl b/servers/rendering/renderer_rd/shaders/effects/ssao.glsl new file mode 100644 index 0000000000..2a87e273bc --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/ssao.glsl @@ -0,0 +1,483 @@ +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// Copyright (c) 2016, Intel Corporation +// Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated +// documentation files (the "Software"), to deal in the Software without restriction, including without limitation +// the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to +// permit persons to whom the Software is furnished to do so, subject to the following conditions: +// The above copyright notice and this permission notice shall be included in all copies or substantial portions of +// the Software. +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO +// THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +// TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +// SOFTWARE. +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// File changes (yyyy-mm-dd) +// 2016-09-07: filip.strugar@intel.com: first commit +// 2020-12-05: clayjohn: convert to Vulkan and Godot +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// + +#[compute] + +#version 450 + +#VERSION_DEFINES + +#define INTELSSAO_MAIN_DISK_SAMPLE_COUNT (32) +const vec4 sample_pattern[INTELSSAO_MAIN_DISK_SAMPLE_COUNT] = { + vec4(0.78488064, 0.56661671, 1.500000, -0.126083), vec4(0.26022232, -0.29575172, 1.500000, -1.064030), vec4(0.10459357, 0.08372527, 1.110000, -2.730563), vec4(-0.68286800, 0.04963045, 1.090000, -0.498827), + vec4(-0.13570161, -0.64190155, 1.250000, -0.532765), vec4(-0.26193795, -0.08205118, 0.670000, -1.783245), vec4(-0.61177456, 0.66664219, 0.710000, -0.044234), vec4(0.43675563, 0.25119025, 0.610000, -1.167283), + vec4(0.07884444, 0.86618668, 0.640000, -0.459002), vec4(-0.12790935, -0.29869005, 0.600000, -1.729424), vec4(-0.04031125, 0.02413622, 0.600000, -4.792042), vec4(0.16201244, -0.52851415, 0.790000, -1.067055), + vec4(-0.70991218, 0.47301072, 0.640000, -0.335236), vec4(0.03277707, -0.22349690, 0.600000, -1.982384), vec4(0.68921727, 0.36800742, 0.630000, -0.266718), vec4(0.29251814, 0.37775412, 0.610000, -1.422520), + vec4(-0.12224089, 0.96582592, 0.600000, -0.426142), vec4(0.11071457, -0.16131058, 0.600000, -2.165947), vec4(0.46562141, -0.59747696, 0.600000, -0.189760), vec4(-0.51548797, 0.11804193, 0.600000, -1.246800), + vec4(0.89141309, -0.42090443, 0.600000, 0.028192), vec4(-0.32402530, -0.01591529, 0.600000, -1.543018), vec4(0.60771245, 0.41635221, 0.600000, -0.605411), vec4(0.02379565, -0.08239821, 0.600000, -3.809046), + vec4(0.48951152, -0.23657045, 0.600000, -1.189011), vec4(-0.17611565, -0.81696892, 0.600000, -0.513724), vec4(-0.33930185, -0.20732205, 0.600000, -1.698047), vec4(-0.91974425, 0.05403209, 0.600000, 0.062246), + vec4(-0.15064627, -0.14949332, 0.600000, -1.896062), vec4(0.53180975, -0.35210401, 0.600000, -0.758838), vec4(0.41487166, 0.81442589, 0.600000, -0.505648), vec4(-0.24106961, -0.32721516, 0.600000, -1.665244) +}; + +// these values can be changed (up to SSAO_MAX_TAPS) with no changes required elsewhere; values for 4th and 5th preset are ignored but array needed to avoid compilation errors +// the actual number of texture samples is two times this value (each "tap" has two symmetrical depth texture samples) +const int num_taps[5] = { 3, 5, 12, 0, 0 }; + +#define SSAO_TILT_SAMPLES_ENABLE_AT_QUALITY_PRESET (99) // to disable simply set to 99 or similar +#define SSAO_TILT_SAMPLES_AMOUNT (0.4) +// +#define SSAO_HALOING_REDUCTION_ENABLE_AT_QUALITY_PRESET (1) // to disable simply set to 99 or similar +#define SSAO_HALOING_REDUCTION_AMOUNT (0.6) // values from 0.0 - 1.0, 1.0 means max weighting (will cause artifacts, 0.8 is more reasonable) +// +#define SSAO_NORMAL_BASED_EDGES_ENABLE_AT_QUALITY_PRESET (2) // to disable simply set to 99 or similar +#define SSAO_NORMAL_BASED_EDGES_DOT_THRESHOLD (0.5) // use 0-0.1 for super-sharp normal-based edges +// +#define SSAO_DETAIL_AO_ENABLE_AT_QUALITY_PRESET (1) // whether to use detail; to disable simply set to 99 or similar +// +#define SSAO_DEPTH_MIPS_ENABLE_AT_QUALITY_PRESET (2) // !!warning!! the MIP generation on the C++ side will be enabled on quality preset 2 regardless of this value, so if changing here, change the C++ side too +#define SSAO_DEPTH_MIPS_GLOBAL_OFFSET (-4.3) // best noise/quality/performance tradeoff, found empirically +// +// !!warning!! the edge handling is hard-coded to 'disabled' on quality level 0, and enabled above, on the C++ side; while toggling it here will work for +// testing purposes, it will not yield performance gains (or correct results) +#define SSAO_DEPTH_BASED_EDGES_ENABLE_AT_QUALITY_PRESET (1) +// +#define SSAO_REDUCE_RADIUS_NEAR_SCREEN_BORDER_ENABLE_AT_QUALITY_PRESET (1) + +#define SSAO_MAX_TAPS 32 +#define SSAO_ADAPTIVE_TAP_BASE_COUNT 5 +#define SSAO_ADAPTIVE_TAP_FLEXIBLE_COUNT (SSAO_MAX_TAPS - SSAO_ADAPTIVE_TAP_BASE_COUNT) +#define SSAO_DEPTH_MIP_LEVELS 4 + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +layout(set = 0, binding = 0) uniform sampler2DArray source_depth_mipmaps; +layout(rgba8, set = 0, binding = 1) uniform restrict readonly image2D source_normal; +layout(set = 0, binding = 2) uniform Constants { //get into a lower set + vec4 rotation_matrices[20]; +} +constants; + +#ifdef ADAPTIVE +layout(rg8, set = 1, binding = 0) uniform restrict readonly image2DArray source_ssao; +layout(set = 1, binding = 1) uniform sampler2D source_importance; +layout(set = 1, binding = 2, std430) buffer Counter { + uint sum; +} +counter; +#endif + +layout(rg8, set = 2, binding = 0) uniform restrict writeonly image2D dest_image; + +// This push_constant is full - 128 bytes - if you need to add more data, consider adding to the uniform buffer instead +layout(push_constant, std430) uniform Params { + ivec2 screen_size; + int pass; + int quality; + + vec2 half_screen_pixel_size; + int size_multiplier; + float detail_intensity; + + vec2 NDC_to_view_mul; + vec2 NDC_to_view_add; + + vec2 pad2; + vec2 half_screen_pixel_size_x025; + + float radius; + float intensity; + float shadow_power; + float shadow_clamp; + + float fade_out_mul; + float fade_out_add; + float horizon_angle_threshold; + float inv_radius_near_limit; + + bool is_orthogonal; + float neg_inv_radius; + float load_counter_avg_div; + float adaptive_sample_limit; + + ivec2 pass_coord_offset; + vec2 pass_uv_offset; +} +params; + +// packing/unpacking for edges; 2 bits per edge mean 4 gradient values (0, 0.33, 0.66, 1) for smoother transitions! +float pack_edges(vec4 p_edgesLRTB) { + p_edgesLRTB = round(clamp(p_edgesLRTB, 0.0, 1.0) * 3.05); + return dot(p_edgesLRTB, vec4(64.0 / 255.0, 16.0 / 255.0, 4.0 / 255.0, 1.0 / 255.0)); +} + +vec3 NDC_to_view_space(vec2 p_pos, float p_viewspace_depth) { + if (params.is_orthogonal) { + return vec3((params.NDC_to_view_mul * p_pos.xy + params.NDC_to_view_add), p_viewspace_depth); + } else { + return vec3((params.NDC_to_view_mul * p_pos.xy + params.NDC_to_view_add) * p_viewspace_depth, p_viewspace_depth); + } +} + +// calculate effect radius and fit our screen sampling pattern inside it +void calculate_radius_parameters(const float p_pix_center_length, const vec2 p_pixel_size_at_center, out float r_lookup_radius, out float r_radius, out float r_fallof_sq) { + r_radius = params.radius; + + // when too close, on-screen sampling disk will grow beyond screen size; limit this to avoid closeup temporal artifacts + const float too_close_limit = clamp(p_pix_center_length * params.inv_radius_near_limit, 0.0, 1.0) * 0.8 + 0.2; + + r_radius *= too_close_limit; + + // 0.85 is to reduce the radius to allow for more samples on a slope to still stay within influence + r_lookup_radius = (0.85 * r_radius) / p_pixel_size_at_center.x; + + // used to calculate falloff (both for AO samples and per-sample weights) + r_fallof_sq = -1.0 / (r_radius * r_radius); +} + +vec4 calculate_edges(const float p_center_z, const float p_left_z, const float p_right_z, const float p_top_z, const float p_bottom_z) { + // slope-sensitive depth-based edge detection + vec4 edgesLRTB = vec4(p_left_z, p_right_z, p_top_z, p_bottom_z) - p_center_z; + vec4 edgesLRTB_slope_adjusted = edgesLRTB + edgesLRTB.yxwz; + edgesLRTB = min(abs(edgesLRTB), abs(edgesLRTB_slope_adjusted)); + return clamp((1.3 - edgesLRTB / (p_center_z * 0.040)), 0.0, 1.0); +} + +vec3 decode_normal(vec3 p_encoded_normal) { + vec3 normal = p_encoded_normal * 2.0 - 1.0; + return normal; +} + +vec3 load_normal(ivec2 p_pos) { + vec3 encoded_normal = imageLoad(source_normal, p_pos).xyz; + encoded_normal.z = 1.0 - encoded_normal.z; + return decode_normal(encoded_normal); +} + +vec3 load_normal(ivec2 p_pos, ivec2 p_offset) { + vec3 encoded_normal = imageLoad(source_normal, p_pos + p_offset).xyz; + encoded_normal.z = 1.0 - encoded_normal.z; + return decode_normal(encoded_normal); +} + +// all vectors in viewspace +float calculate_pixel_obscurance(vec3 p_pixel_normal, vec3 p_hit_delta, float p_fallof_sq) { + float length_sq = dot(p_hit_delta, p_hit_delta); + float NdotD = dot(p_pixel_normal, p_hit_delta) / sqrt(length_sq); + + float falloff_mult = max(0.0, length_sq * p_fallof_sq + 1.0); + + return max(0, NdotD - params.horizon_angle_threshold) * falloff_mult; +} + +void SSAO_tap_inner(const int p_quality_level, inout float r_obscurance_sum, inout float r_weight_sum, const vec2 p_sampling_uv, const float p_mip_level, const vec3 p_pix_center_pos, vec3 p_pixel_normal, const float p_fallof_sq, const float p_weight_mod) { + // get depth at sample + float viewspace_sample_z = textureLod(source_depth_mipmaps, vec3(p_sampling_uv, params.pass), p_mip_level).x; + + // convert to viewspace + vec3 hit_pos = NDC_to_view_space(p_sampling_uv.xy, viewspace_sample_z).xyz; + vec3 hit_delta = hit_pos - p_pix_center_pos; + + float obscurance = calculate_pixel_obscurance(p_pixel_normal, hit_delta, p_fallof_sq); + float weight = 1.0; + + if (p_quality_level >= SSAO_HALOING_REDUCTION_ENABLE_AT_QUALITY_PRESET) { + float reduct = max(0, -hit_delta.z); + reduct = clamp(reduct * params.neg_inv_radius + 2.0, 0.0, 1.0); + weight = SSAO_HALOING_REDUCTION_AMOUNT * reduct + (1.0 - SSAO_HALOING_REDUCTION_AMOUNT); + } + weight *= p_weight_mod; + r_obscurance_sum += obscurance * weight; + r_weight_sum += weight; +} + +void SSAOTap(const int p_quality_level, inout float r_obscurance_sum, inout float r_weight_sum, const int p_tap_index, const mat2 p_rot_scale, const vec3 p_pix_center_pos, vec3 p_pixel_normal, const vec2 p_normalized_screen_pos, const float p_mip_offset, const float p_fallof_sq, float p_weight_mod, vec2 p_norm_xy, float p_norm_xy_length) { + vec2 sample_offset; + float sample_pow_2_len; + + // patterns + { + vec4 new_sample = sample_pattern[p_tap_index]; + sample_offset = new_sample.xy * p_rot_scale; + sample_pow_2_len = new_sample.w; // precalculated, same as: sample_pow_2_len = log2( length( new_sample.xy ) ); + p_weight_mod *= new_sample.z; + } + + // snap to pixel center (more correct obscurance math, avoids artifacts) + sample_offset = round(sample_offset); + + // calculate MIP based on the sample distance from the centre, similar to as described + // in http://graphics.cs.williams.edu/papers/SAOHPG12/. + float mip_level = (p_quality_level < SSAO_DEPTH_MIPS_ENABLE_AT_QUALITY_PRESET) ? (0) : (sample_pow_2_len + p_mip_offset); + + vec2 sampling_uv = sample_offset * params.half_screen_pixel_size + p_normalized_screen_pos; + + SSAO_tap_inner(p_quality_level, r_obscurance_sum, r_weight_sum, sampling_uv, mip_level, p_pix_center_pos, p_pixel_normal, p_fallof_sq, p_weight_mod); + + // for the second tap, just use the mirrored offset + vec2 sample_offset_mirrored_uv = -sample_offset; + + // tilt the second set of samples so that the disk is effectively rotated by the normal + // effective at removing one set of artifacts, but too expensive for lower quality settings + if (p_quality_level >= SSAO_TILT_SAMPLES_ENABLE_AT_QUALITY_PRESET) { + float dot_norm = dot(sample_offset_mirrored_uv, p_norm_xy); + sample_offset_mirrored_uv -= dot_norm * p_norm_xy_length * p_norm_xy; + sample_offset_mirrored_uv = round(sample_offset_mirrored_uv); + } + + // snap to pixel center (more correct obscurance math, avoids artifacts) + vec2 sampling_mirrored_uv = sample_offset_mirrored_uv * params.half_screen_pixel_size + p_normalized_screen_pos; + + SSAO_tap_inner(p_quality_level, r_obscurance_sum, r_weight_sum, sampling_mirrored_uv, mip_level, p_pix_center_pos, p_pixel_normal, p_fallof_sq, p_weight_mod); +} + +void generate_SSAO_shadows_internal(out float r_shadow_term, out vec4 r_edges, out float r_weight, const vec2 p_pos, int p_quality_level, bool p_adaptive_base) { + vec2 pos_rounded = trunc(p_pos); + uvec2 upos = uvec2(pos_rounded); + + const int number_of_taps = (p_adaptive_base) ? (SSAO_ADAPTIVE_TAP_BASE_COUNT) : (num_taps[p_quality_level]); + float pix_z, pix_left_z, pix_top_z, pix_right_z, pix_bottom_z; + + vec4 valuesUL = textureGather(source_depth_mipmaps, vec3(pos_rounded * params.half_screen_pixel_size, params.pass)); + vec4 valuesBR = textureGather(source_depth_mipmaps, vec3((pos_rounded + vec2(1.0)) * params.half_screen_pixel_size, params.pass)); + + // get this pixel's viewspace depth + pix_z = valuesUL.y; + + // get left right top bottom neighbouring pixels for edge detection (gets compiled out on quality_level == 0) + pix_left_z = valuesUL.x; + pix_top_z = valuesUL.z; + pix_right_z = valuesBR.z; + pix_bottom_z = valuesBR.x; + + vec2 normalized_screen_pos = pos_rounded * params.half_screen_pixel_size + params.half_screen_pixel_size_x025; + vec3 pix_center_pos = NDC_to_view_space(normalized_screen_pos, pix_z); + + // Load this pixel's viewspace normal + uvec2 full_res_coord = upos * 2 * params.size_multiplier + params.pass_coord_offset.xy; + vec3 pixel_normal = load_normal(ivec2(full_res_coord)); + + const vec2 pixel_size_at_center = NDC_to_view_space(normalized_screen_pos.xy + params.half_screen_pixel_size, pix_center_pos.z).xy - pix_center_pos.xy; + + float pixel_lookup_radius; + float fallof_sq; + + // calculate effect radius and fit our screen sampling pattern inside it + float viewspace_radius; + calculate_radius_parameters(length(pix_center_pos), pixel_size_at_center, pixel_lookup_radius, viewspace_radius, fallof_sq); + + // calculate samples rotation/scaling + mat2 rot_scale_matrix; + uint pseudo_random_index; + + { + vec4 rotation_scale; + // reduce effect radius near the screen edges slightly; ideally, one would render a larger depth buffer (5% on each side) instead + if (!p_adaptive_base && (p_quality_level >= SSAO_REDUCE_RADIUS_NEAR_SCREEN_BORDER_ENABLE_AT_QUALITY_PRESET)) { + float near_screen_border = min(min(normalized_screen_pos.x, 1.0 - normalized_screen_pos.x), min(normalized_screen_pos.y, 1.0 - normalized_screen_pos.y)); + near_screen_border = clamp(10.0 * near_screen_border + 0.6, 0.0, 1.0); + pixel_lookup_radius *= near_screen_border; + } + + // load & update pseudo-random rotation matrix + pseudo_random_index = uint(pos_rounded.y * 2 + pos_rounded.x) % 5; + rotation_scale = constants.rotation_matrices[params.pass * 5 + pseudo_random_index]; + rot_scale_matrix = mat2(rotation_scale.x * pixel_lookup_radius, rotation_scale.y * pixel_lookup_radius, rotation_scale.z * pixel_lookup_radius, rotation_scale.w * pixel_lookup_radius); + } + + // the main obscurance & sample weight storage + float obscurance_sum = 0.0; + float weight_sum = 0.0; + + // edge mask for between this and left/right/top/bottom neighbour pixels - not used in quality level 0 so initialize to "no edge" (1 is no edge, 0 is edge) + vec4 edgesLRTB = vec4(1.0, 1.0, 1.0, 1.0); + + // Move center pixel slightly towards camera to avoid imprecision artifacts due to using of 16bit depth buffer; a lot smaller offsets needed when using 32bit floats + pix_center_pos *= 0.9992; + + if (!p_adaptive_base && (p_quality_level >= SSAO_DEPTH_BASED_EDGES_ENABLE_AT_QUALITY_PRESET)) { + edgesLRTB = calculate_edges(pix_z, pix_left_z, pix_right_z, pix_top_z, pix_bottom_z); + } + + // adds a more high definition sharp effect, which gets blurred out (reuses left/right/top/bottom samples that we used for edge detection) + if (!p_adaptive_base && (p_quality_level >= SSAO_DETAIL_AO_ENABLE_AT_QUALITY_PRESET)) { + // disable in case of quality level 4 (reference) + if (p_quality_level != 4) { + //approximate neighbouring pixels positions (actually just deltas or "positions - pix_center_pos" ) + vec3 normalized_viewspace_dir = vec3(pix_center_pos.xy / pix_center_pos.zz, 1.0); + vec3 pixel_left_delta = vec3(-pixel_size_at_center.x, 0.0, 0.0) + normalized_viewspace_dir * (pix_left_z - pix_center_pos.z); + vec3 pixel_right_delta = vec3(+pixel_size_at_center.x, 0.0, 0.0) + normalized_viewspace_dir * (pix_right_z - pix_center_pos.z); + vec3 pixel_top_delta = vec3(0.0, -pixel_size_at_center.y, 0.0) + normalized_viewspace_dir * (pix_top_z - pix_center_pos.z); + vec3 pixel_bottom_delta = vec3(0.0, +pixel_size_at_center.y, 0.0) + normalized_viewspace_dir * (pix_bottom_z - pix_center_pos.z); + + const float range_reduction = 4.0f; // this is to avoid various artifacts + const float modified_fallof_sq = range_reduction * fallof_sq; + + vec4 additional_obscurance; + additional_obscurance.x = calculate_pixel_obscurance(pixel_normal, pixel_left_delta, modified_fallof_sq); + additional_obscurance.y = calculate_pixel_obscurance(pixel_normal, pixel_right_delta, modified_fallof_sq); + additional_obscurance.z = calculate_pixel_obscurance(pixel_normal, pixel_top_delta, modified_fallof_sq); + additional_obscurance.w = calculate_pixel_obscurance(pixel_normal, pixel_bottom_delta, modified_fallof_sq); + + obscurance_sum += params.detail_intensity * dot(additional_obscurance, edgesLRTB); + } + } + + // Sharp normals also create edges - but this adds to the cost as well + if (!p_adaptive_base && (p_quality_level >= SSAO_NORMAL_BASED_EDGES_ENABLE_AT_QUALITY_PRESET)) { + vec3 neighbour_normal_left = load_normal(ivec2(full_res_coord), ivec2(-2, 0)); + vec3 neighbour_normal_right = load_normal(ivec2(full_res_coord), ivec2(2, 0)); + vec3 neighbour_normal_top = load_normal(ivec2(full_res_coord), ivec2(0, -2)); + vec3 neighbour_normal_bottom = load_normal(ivec2(full_res_coord), ivec2(0, 2)); + + const float dot_threshold = SSAO_NORMAL_BASED_EDGES_DOT_THRESHOLD; + + vec4 normal_edgesLRTB; + normal_edgesLRTB.x = clamp((dot(pixel_normal, neighbour_normal_left) + dot_threshold), 0.0, 1.0); + normal_edgesLRTB.y = clamp((dot(pixel_normal, neighbour_normal_right) + dot_threshold), 0.0, 1.0); + normal_edgesLRTB.z = clamp((dot(pixel_normal, neighbour_normal_top) + dot_threshold), 0.0, 1.0); + normal_edgesLRTB.w = clamp((dot(pixel_normal, neighbour_normal_bottom) + dot_threshold), 0.0, 1.0); + + edgesLRTB *= normal_edgesLRTB; + } + + const float global_mip_offset = SSAO_DEPTH_MIPS_GLOBAL_OFFSET; + float mip_offset = (p_quality_level < SSAO_DEPTH_MIPS_ENABLE_AT_QUALITY_PRESET) ? (0) : (log2(pixel_lookup_radius) + global_mip_offset); + + // Used to tilt the second set of samples so that the disk is effectively rotated by the normal + // effective at removing one set of artifacts, but too expensive for lower quality settings + vec2 norm_xy = vec2(pixel_normal.x, pixel_normal.y); + float norm_xy_length = length(norm_xy); + norm_xy /= vec2(norm_xy_length, -norm_xy_length); + norm_xy_length *= SSAO_TILT_SAMPLES_AMOUNT; + + // standard, non-adaptive approach + if ((p_quality_level != 3) || p_adaptive_base) { + for (int i = 0; i < number_of_taps; i++) { + SSAOTap(p_quality_level, obscurance_sum, weight_sum, i, rot_scale_matrix, pix_center_pos, pixel_normal, normalized_screen_pos, mip_offset, fallof_sq, 1.0, norm_xy, norm_xy_length); + } + } +#ifdef ADAPTIVE + else { + // add new ones if needed + vec2 full_res_uv = normalized_screen_pos + params.pass_uv_offset.xy; + float importance = textureLod(source_importance, full_res_uv, 0.0).x; + + // this is to normalize SSAO_DETAIL_AO_AMOUNT across all pixel regardless of importance + obscurance_sum *= (SSAO_ADAPTIVE_TAP_BASE_COUNT / float(SSAO_MAX_TAPS)) + (importance * SSAO_ADAPTIVE_TAP_FLEXIBLE_COUNT / float(SSAO_MAX_TAPS)); + + // load existing base values + vec2 base_values = imageLoad(source_ssao, ivec3(upos, params.pass)).xy; + weight_sum += base_values.y * float(SSAO_ADAPTIVE_TAP_BASE_COUNT * 4.0); + obscurance_sum += (base_values.x) * weight_sum; + + // increase importance around edges + float edge_count = dot(1.0 - edgesLRTB, vec4(1.0, 1.0, 1.0, 1.0)); + + float avg_total_importance = float(counter.sum) * params.load_counter_avg_div; + + float importance_limiter = clamp(params.adaptive_sample_limit / avg_total_importance, 0.0, 1.0); + importance *= importance_limiter; + + float additional_sample_count = SSAO_ADAPTIVE_TAP_FLEXIBLE_COUNT * importance; + + const float blend_range = 3.0; + const float blend_range_inv = 1.0 / blend_range; + + additional_sample_count += 0.5; + uint additional_samples = uint(additional_sample_count); + uint additional_samples_to = min(SSAO_MAX_TAPS, additional_samples + SSAO_ADAPTIVE_TAP_BASE_COUNT); + + for (uint i = SSAO_ADAPTIVE_TAP_BASE_COUNT; i < additional_samples_to; i++) { + additional_sample_count -= 1.0f; + float weight_mod = clamp(additional_sample_count * blend_range_inv, 0.0, 1.0); + SSAOTap(p_quality_level, obscurance_sum, weight_sum, int(i), rot_scale_matrix, pix_center_pos, pixel_normal, normalized_screen_pos, mip_offset, fallof_sq, weight_mod, norm_xy, norm_xy_length); + } + } +#endif + + // early out for adaptive base - just output weight (used for the next pass) + if (p_adaptive_base) { + float obscurance = obscurance_sum / weight_sum; + + r_shadow_term = obscurance; + r_edges = vec4(0.0); + r_weight = weight_sum; + return; + } + + // calculate weighted average + float obscurance = obscurance_sum / weight_sum; + + // calculate fadeout (1 close, gradient, 0 far) + float fade_out = clamp(pix_center_pos.z * params.fade_out_mul + params.fade_out_add, 0.0, 1.0); + + // Reduce the SSAO shadowing if we're on the edge to remove artifacts on edges (we don't care for the lower quality one) + if (!p_adaptive_base && (p_quality_level >= SSAO_DEPTH_BASED_EDGES_ENABLE_AT_QUALITY_PRESET)) { + // when there's more than 2 opposite edges, start fading out the occlusion to reduce aliasing artifacts + float edge_fadeout_factor = clamp((1.0 - edgesLRTB.x - edgesLRTB.y) * 0.35, 0.0, 1.0) + clamp((1.0 - edgesLRTB.z - edgesLRTB.w) * 0.35, 0.0, 1.0); + + fade_out *= clamp(1.0 - edge_fadeout_factor, 0.0, 1.0); + } + + // strength + obscurance = params.intensity * obscurance; + + // clamp + obscurance = min(obscurance, params.shadow_clamp); + + // fadeout + obscurance *= fade_out; + + // conceptually switch to occlusion with the meaning being visibility (grows with visibility, occlusion == 1 implies full visibility), + // to be in line with what is more commonly used. + float occlusion = 1.0 - obscurance; + + // modify the gradient + // note: this cannot be moved to a later pass because of loss of precision after storing in the render target + occlusion = pow(clamp(occlusion, 0.0, 1.0), params.shadow_power); + + // outputs! + r_shadow_term = occlusion; // Our final 'occlusion' term (0 means fully occluded, 1 means fully lit) + r_edges = edgesLRTB; // These are used to prevent blurring across edges, 1 means no edge, 0 means edge, 0.5 means half way there, etc. + r_weight = weight_sum; +} + +void main() { + float out_shadow_term; + float out_weight; + vec4 out_edges; + ivec2 ssC = ivec2(gl_GlobalInvocationID.xy); + if (any(greaterThanEqual(ssC, params.screen_size))) { //too large, do nothing + return; + } + + vec2 uv = vec2(gl_GlobalInvocationID) + vec2(0.5); +#ifdef SSAO_BASE + generate_SSAO_shadows_internal(out_shadow_term, out_edges, out_weight, uv, params.quality, true); + + imageStore(dest_image, ivec2(gl_GlobalInvocationID.xy), vec4(out_shadow_term, out_weight / (float(SSAO_ADAPTIVE_TAP_BASE_COUNT) * 4.0), 0.0, 0.0)); +#else + generate_SSAO_shadows_internal(out_shadow_term, out_edges, out_weight, uv, params.quality, false); // pass in quality levels + if (params.quality == 0) { + out_edges = vec4(1.0); + } + + imageStore(dest_image, ivec2(gl_GlobalInvocationID.xy), vec4(out_shadow_term, pack_edges(out_edges), 0.0, 0.0)); +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/effects/ssao_blur.glsl b/servers/rendering/renderer_rd/shaders/effects/ssao_blur.glsl new file mode 100644 index 0000000000..f42734c46d --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/ssao_blur.glsl @@ -0,0 +1,154 @@ +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// Copyright (c) 2016, Intel Corporation +// Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated +// documentation files (the "Software"), to deal in the Software without restriction, including without limitation +// the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to +// permit persons to whom the Software is furnished to do so, subject to the following conditions: +// The above copyright notice and this permission notice shall be included in all copies or substantial portions of +// the Software. +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO +// THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +// TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +// SOFTWARE. +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// File changes (yyyy-mm-dd) +// 2016-09-07: filip.strugar@intel.com: first commit +// 2020-12-05: clayjohn: convert to Vulkan and Godot +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// + +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +layout(set = 0, binding = 0) uniform sampler2D source_ssao; + +layout(rg8, set = 1, binding = 0) uniform restrict writeonly image2D dest_image; + +layout(push_constant, std430) uniform Params { + float edge_sharpness; + float pad; + vec2 half_screen_pixel_size; +} +params; + +vec4 unpack_edges(float p_packed_val) { + uint packed_val = uint(p_packed_val * 255.5); + vec4 edgesLRTB; + edgesLRTB.x = float((packed_val >> 6) & 0x03) / 3.0; + edgesLRTB.y = float((packed_val >> 4) & 0x03) / 3.0; + edgesLRTB.z = float((packed_val >> 2) & 0x03) / 3.0; + edgesLRTB.w = float((packed_val >> 0) & 0x03) / 3.0; + + return clamp(edgesLRTB + params.edge_sharpness, 0.0, 1.0); +} + +void add_sample(float p_ssao_value, float p_edge_value, inout float r_sum, inout float r_sum_weight) { + float weight = p_edge_value; + + r_sum += (weight * p_ssao_value); + r_sum_weight += weight; +} + +#ifdef MODE_WIDE +vec2 sample_blurred_wide(vec2 p_coord) { + vec2 vC = textureLodOffset(source_ssao, vec2(p_coord), 0.0, ivec2(0, 0)).xy; + vec2 vL = textureLodOffset(source_ssao, vec2(p_coord), 0.0, ivec2(-2, 0)).xy; + vec2 vT = textureLodOffset(source_ssao, vec2(p_coord), 0.0, ivec2(0, -2)).xy; + vec2 vR = textureLodOffset(source_ssao, vec2(p_coord), 0.0, ivec2(2, 0)).xy; + vec2 vB = textureLodOffset(source_ssao, vec2(p_coord), 0.0, ivec2(0, 2)).xy; + + float packed_edges = vC.y; + vec4 edgesLRTB = unpack_edges(packed_edges); + edgesLRTB.x *= unpack_edges(vL.y).y; + edgesLRTB.z *= unpack_edges(vT.y).w; + edgesLRTB.y *= unpack_edges(vR.y).x; + edgesLRTB.w *= unpack_edges(vB.y).z; + + float ssao_value = vC.x; + float ssao_valueL = vL.x; + float ssao_valueT = vT.x; + float ssao_valueR = vR.x; + float ssao_valueB = vB.x; + + float sum_weight = 0.8f; + float sum = ssao_value * sum_weight; + + add_sample(ssao_valueL, edgesLRTB.x, sum, sum_weight); + add_sample(ssao_valueR, edgesLRTB.y, sum, sum_weight); + add_sample(ssao_valueT, edgesLRTB.z, sum, sum_weight); + add_sample(ssao_valueB, edgesLRTB.w, sum, sum_weight); + + float ssao_avg = sum / sum_weight; + + ssao_value = ssao_avg; + + return vec2(ssao_value, packed_edges); +} +#endif + +#ifdef MODE_SMART +vec2 sample_blurred(vec3 p_pos, vec2 p_coord) { + float packed_edges = texelFetch(source_ssao, ivec2(p_pos.xy), 0).y; + vec4 edgesLRTB = unpack_edges(packed_edges); + + vec4 valuesUL = textureGather(source_ssao, vec2(p_coord - params.half_screen_pixel_size * 0.5)); + vec4 valuesBR = textureGather(source_ssao, vec2(p_coord + params.half_screen_pixel_size * 0.5)); + + float ssao_value = valuesUL.y; + float ssao_valueL = valuesUL.x; + float ssao_valueT = valuesUL.z; + float ssao_valueR = valuesBR.z; + float ssao_valueB = valuesBR.x; + + float sum_weight = 0.5; + float sum = ssao_value * sum_weight; + + add_sample(ssao_valueL, edgesLRTB.x, sum, sum_weight); + add_sample(ssao_valueR, edgesLRTB.y, sum, sum_weight); + + add_sample(ssao_valueT, edgesLRTB.z, sum, sum_weight); + add_sample(ssao_valueB, edgesLRTB.w, sum, sum_weight); + + float ssao_avg = sum / sum_weight; + + ssao_value = ssao_avg; + + return vec2(ssao_value, packed_edges); +} +#endif + +void main() { + // Pixel being shaded + ivec2 ssC = ivec2(gl_GlobalInvocationID.xy); + +#ifdef MODE_NON_SMART + + vec2 half_pixel = params.half_screen_pixel_size * 0.5; + + vec2 uv = (vec2(gl_GlobalInvocationID.xy) + vec2(0.5, 0.5)) * params.half_screen_pixel_size; + + vec2 center = textureLod(source_ssao, vec2(uv), 0.0).xy; + + vec4 vals; + vals.x = textureLod(source_ssao, vec2(uv + vec2(-half_pixel.x * 3, -half_pixel.y)), 0.0).x; + vals.y = textureLod(source_ssao, vec2(uv + vec2(+half_pixel.x, -half_pixel.y * 3)), 0.0).x; + vals.z = textureLod(source_ssao, vec2(uv + vec2(-half_pixel.x, +half_pixel.y * 3)), 0.0).x; + vals.w = textureLod(source_ssao, vec2(uv + vec2(+half_pixel.x * 3, +half_pixel.y)), 0.0).x; + + vec2 sampled = vec2(dot(vals, vec4(0.2)) + center.x * 0.2, center.y); + +#else +#ifdef MODE_SMART + vec2 sampled = sample_blurred(vec3(gl_GlobalInvocationID), (vec2(gl_GlobalInvocationID.xy) + vec2(0.5, 0.5)) * params.half_screen_pixel_size); +#else // MODE_WIDE + vec2 sampled = sample_blurred_wide((vec2(gl_GlobalInvocationID.xy) + vec2(0.5, 0.5)) * params.half_screen_pixel_size); +#endif + +#endif + imageStore(dest_image, ivec2(ssC), vec4(sampled, 0.0, 0.0)); +} diff --git a/servers/rendering/renderer_rd/shaders/effects/ssao_importance_map.glsl b/servers/rendering/renderer_rd/shaders/effects/ssao_importance_map.glsl new file mode 100644 index 0000000000..04f98964e8 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/ssao_importance_map.glsl @@ -0,0 +1,123 @@ +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// Copyright (c) 2016, Intel Corporation +// Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated +// documentation files (the "Software"), to deal in the Software without restriction, including without limitation +// the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to +// permit persons to whom the Software is furnished to do so, subject to the following conditions: +// The above copyright notice and this permission notice shall be included in all copies or substantial portions of +// the Software. +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO +// THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +// TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +// SOFTWARE. +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// File changes (yyyy-mm-dd) +// 2016-09-07: filip.strugar@intel.com: first commit +// 2020-12-05: clayjohn: convert to Vulkan and Godot +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// + +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +#ifdef GENERATE_MAP +layout(set = 0, binding = 0) uniform sampler2DArray source_texture; +#else +layout(set = 0, binding = 0) uniform sampler2D source_importance; +#endif +layout(r8, set = 1, binding = 0) uniform restrict writeonly image2D dest_image; + +#ifdef PROCESS_MAPB +layout(set = 2, binding = 0, std430) buffer Counter { + uint sum; +} +counter; +#endif + +layout(push_constant, std430) uniform Params { + vec2 half_screen_pixel_size; + float intensity; + float power; +} +params; + +void main() { + // Pixel being shaded + ivec2 ssC = ivec2(gl_GlobalInvocationID.xy); + +#ifdef GENERATE_MAP + // importance map stuff + uvec2 base_position = ssC * 2; + + vec2 base_uv = (vec2(base_position) + vec2(0.5f, 0.5f)) * params.half_screen_pixel_size; + + float minV = 1.0; + float maxV = 0.0; + for (int i = 0; i < 4; i++) { + vec4 vals = textureGather(source_texture, vec3(base_uv, i)); + + // apply the same modifications that would have been applied in the main shader + vals = params.intensity * vals; + + vals = 1 - vals; + + vals = pow(clamp(vals, 0.0, 1.0), vec4(params.power)); + + maxV = max(maxV, max(max(vals.x, vals.y), max(vals.z, vals.w))); + minV = min(minV, min(min(vals.x, vals.y), min(vals.z, vals.w))); + } + + float min_max_diff = maxV - minV; + + imageStore(dest_image, ssC, vec4(pow(clamp(min_max_diff * 2.0, 0.0, 1.0), 0.8))); +#endif + +#ifdef PROCESS_MAPA + vec2 uv = (vec2(ssC) + 0.5f) * params.half_screen_pixel_size * 2.0; + + float centre = textureLod(source_importance, uv, 0.0).x; + + vec2 half_pixel = params.half_screen_pixel_size; + + vec4 vals; + vals.x = textureLod(source_importance, uv + vec2(-half_pixel.x * 3, -half_pixel.y), 0.0).x; + vals.y = textureLod(source_importance, uv + vec2(+half_pixel.x, -half_pixel.y * 3), 0.0).x; + vals.z = textureLod(source_importance, uv + vec2(+half_pixel.x * 3, +half_pixel.y), 0.0).x; + vals.w = textureLod(source_importance, uv + vec2(-half_pixel.x, +half_pixel.y * 3), 0.0).x; + + float avg = dot(vals, vec4(0.25, 0.25, 0.25, 0.25)); + + imageStore(dest_image, ssC, vec4(avg)); +#endif + +#ifdef PROCESS_MAPB + vec2 uv = (vec2(ssC) + 0.5f) * params.half_screen_pixel_size * 2.0; + + float centre = textureLod(source_importance, uv, 0.0).x; + + vec2 half_pixel = params.half_screen_pixel_size; + + vec4 vals; + vals.x = textureLod(source_importance, uv + vec2(-half_pixel.x, -half_pixel.y * 3), 0.0).x; + vals.y = textureLod(source_importance, uv + vec2(+half_pixel.x * 3, -half_pixel.y), 0.0).x; + vals.z = textureLod(source_importance, uv + vec2(+half_pixel.x, +half_pixel.y * 3), 0.0).x; + vals.w = textureLod(source_importance, uv + vec2(-half_pixel.x * 3, +half_pixel.y), 0.0).x; + + float avg = dot(vals, vec4(0.25, 0.25, 0.25, 0.25)); + + imageStore(dest_image, ssC, vec4(avg)); + + // sum the average; to avoid overflowing we assume max AO resolution is not bigger than 16384x16384; so quarter res (used here) will be 4096x4096, which leaves us with 8 bits per pixel + uint sum = uint(clamp(avg, 0.0, 1.0) * 255.0 + 0.5); + + // save every 9th to avoid InterlockedAdd congestion - since we're blurring, this is good enough; compensated by multiplying load_counter_avg_div by 9 + if (((ssC.x % 3) + (ssC.y % 3)) == 0) { + atomicAdd(counter.sum, sum); + } +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/effects/ssao_interleave.glsl b/servers/rendering/renderer_rd/shaders/effects/ssao_interleave.glsl new file mode 100644 index 0000000000..f6a9a92fac --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/ssao_interleave.glsl @@ -0,0 +1,119 @@ +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// Copyright (c) 2016, Intel Corporation +// Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated +// documentation files (the "Software"), to deal in the Software without restriction, including without limitation +// the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to +// permit persons to whom the Software is furnished to do so, subject to the following conditions: +// The above copyright notice and this permission notice shall be included in all copies or substantial portions of +// the Software. +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO +// THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +// TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +// SOFTWARE. +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// File changes (yyyy-mm-dd) +// 2016-09-07: filip.strugar@intel.com: first commit +// 2020-12-05: clayjohn: convert to Vulkan and Godot +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +layout(rgba8, set = 0, binding = 0) uniform restrict writeonly image2D dest_image; +layout(set = 1, binding = 0) uniform sampler2DArray source_texture; + +layout(push_constant, std430) uniform Params { + float inv_sharpness; + uint size_modifier; + vec2 pixel_size; +} +params; + +vec4 unpack_edges(float p_packed_val) { + uint packed_val = uint(p_packed_val * 255.5); + vec4 edgesLRTB; + edgesLRTB.x = float((packed_val >> 6) & 0x03) / 3.0; + edgesLRTB.y = float((packed_val >> 4) & 0x03) / 3.0; + edgesLRTB.z = float((packed_val >> 2) & 0x03) / 3.0; + edgesLRTB.w = float((packed_val >> 0) & 0x03) / 3.0; + + return clamp(edgesLRTB + params.inv_sharpness, 0.0, 1.0); +} + +void main() { + ivec2 ssC = ivec2(gl_GlobalInvocationID.xy); + if (any(greaterThanEqual(ssC, ivec2(1.0 / params.pixel_size)))) { //too large, do nothing + return; + } + +#ifdef MODE_SMART + float ao; + uvec2 pix_pos = uvec2(gl_GlobalInvocationID.xy); + vec2 uv = (gl_GlobalInvocationID.xy + vec2(0.5)) * params.pixel_size; + + // calculate index in the four deinterleaved source array texture + int mx = int(pix_pos.x % 2); + int my = int(pix_pos.y % 2); + int index_center = mx + my * 2; // center index + int index_horizontal = (1 - mx) + my * 2; // neighbouring, horizontal + int index_vertical = mx + (1 - my) * 2; // neighbouring, vertical + int index_diagonal = (1 - mx) + (1 - my) * 2; // diagonal + + vec2 center_val = texelFetch(source_texture, ivec3(pix_pos / uvec2(params.size_modifier), index_center), 0).xy; + + ao = center_val.x; + + vec4 edgesLRTB = unpack_edges(center_val.y); + + // convert index shifts to sampling offsets + float fmx = float(mx); + float fmy = float(my); + + // in case of an edge, push sampling offsets away from the edge (towards pixel center) + float fmxe = (edgesLRTB.y - edgesLRTB.x); + float fmye = (edgesLRTB.w - edgesLRTB.z); + + // calculate final sampling offsets and sample using bilinear filter + vec2 uv_horizontal = (gl_GlobalInvocationID.xy + vec2(0.5) + vec2(fmx + fmxe - 0.5, 0.5 - fmy)) * params.pixel_size; + float ao_horizontal = textureLod(source_texture, vec3(uv_horizontal, index_horizontal), 0.0).x; + vec2 uv_vertical = (gl_GlobalInvocationID.xy + vec2(0.5) + vec2(0.5 - fmx, fmy - 0.5 + fmye)) * params.pixel_size; + float ao_vertical = textureLod(source_texture, vec3(uv_vertical, index_vertical), 0.0).x; + vec2 uv_diagonal = (gl_GlobalInvocationID.xy + vec2(0.5) + vec2(fmx - 0.5 + fmxe, fmy - 0.5 + fmye)) * params.pixel_size; + float ao_diagonal = textureLod(source_texture, vec3(uv_diagonal, index_diagonal), 0.0).x; + + // reduce weight for samples near edge - if the edge is on both sides, weight goes to 0 + vec4 blendWeights; + blendWeights.x = 1.0; + blendWeights.y = (edgesLRTB.x + edgesLRTB.y) * 0.5; + blendWeights.z = (edgesLRTB.z + edgesLRTB.w) * 0.5; + blendWeights.w = (blendWeights.y + blendWeights.z) * 0.5; + + // calculate weighted average + float blendWeightsSum = dot(blendWeights, vec4(1.0, 1.0, 1.0, 1.0)); + ao = dot(vec4(ao, ao_horizontal, ao_vertical, ao_diagonal), blendWeights) / blendWeightsSum; + + imageStore(dest_image, ivec2(gl_GlobalInvocationID.xy), vec4(ao)); +#else // !MODE_SMART + + vec2 uv = (gl_GlobalInvocationID.xy + vec2(0.5)) * params.pixel_size; +#ifdef MODE_HALF + float a = textureLod(source_texture, vec3(uv, 0), 0.0).x; + float d = textureLod(source_texture, vec3(uv, 3), 0.0).x; + float avg = (a + d) * 0.5; + +#else + float a = textureLod(source_texture, vec3(uv, 0), 0.0).x; + float b = textureLod(source_texture, vec3(uv, 1), 0.0).x; + float c = textureLod(source_texture, vec3(uv, 2), 0.0).x; + float d = textureLod(source_texture, vec3(uv, 3), 0.0).x; + float avg = (a + b + c + d) * 0.25; + +#endif + imageStore(dest_image, ivec2(gl_GlobalInvocationID.xy), vec4(avg)); +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/effects/ssil.glsl b/servers/rendering/renderer_rd/shaders/effects/ssil.glsl new file mode 100644 index 0000000000..513791dfbf --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/ssil.glsl @@ -0,0 +1,444 @@ +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// Copyright (c) 2016, Intel Corporation +// Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated +// documentation files (the "Software"), to deal in the Software without restriction, including without limitation +// the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to +// permit persons to whom the Software is furnished to do so, subject to the following conditions: +// The above copyright notice and this permission notice shall be included in all copies or substantial portions of +// the Software. +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO +// THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +// TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +// SOFTWARE. +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// File changes (yyyy-mm-dd) +// 2016-09-07: filip.strugar@intel.com: first commit +// 2020-12-05: clayjohn: convert to Vulkan and Godot +// 2021-05-27: clayjohn: convert SSAO to SSIL +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// + +#[compute] + +#version 450 + +#VERSION_DEFINES + +#define SSIL_MAIN_DISK_SAMPLE_COUNT (32) +const vec4 sample_pattern[SSIL_MAIN_DISK_SAMPLE_COUNT] = { + vec4(0.78488064, 0.56661671, 1.500000, -0.126083), vec4(0.26022232, -0.29575172, 1.500000, -1.064030), vec4(0.10459357, 0.08372527, 1.110000, -2.730563), vec4(-0.68286800, 0.04963045, 1.090000, -0.498827), + vec4(-0.13570161, -0.64190155, 1.250000, -0.532765), vec4(-0.26193795, -0.08205118, 0.670000, -1.783245), vec4(-0.61177456, 0.66664219, 0.710000, -0.044234), vec4(0.43675563, 0.25119025, 0.610000, -1.167283), + vec4(0.07884444, 0.86618668, 0.640000, -0.459002), vec4(-0.12790935, -0.29869005, 0.600000, -1.729424), vec4(-0.04031125, 0.02413622, 0.600000, -4.792042), vec4(0.16201244, -0.52851415, 0.790000, -1.067055), + vec4(-0.70991218, 0.47301072, 0.640000, -0.335236), vec4(0.03277707, -0.22349690, 0.600000, -1.982384), vec4(0.68921727, 0.36800742, 0.630000, -0.266718), vec4(0.29251814, 0.37775412, 0.610000, -1.422520), + vec4(-0.12224089, 0.96582592, 0.600000, -0.426142), vec4(0.11071457, -0.16131058, 0.600000, -2.165947), vec4(0.46562141, -0.59747696, 0.600000, -0.189760), vec4(-0.51548797, 0.11804193, 0.600000, -1.246800), + vec4(0.89141309, -0.42090443, 0.600000, 0.028192), vec4(-0.32402530, -0.01591529, 0.600000, -1.543018), vec4(0.60771245, 0.41635221, 0.600000, -0.605411), vec4(0.02379565, -0.08239821, 0.600000, -3.809046), + vec4(0.48951152, -0.23657045, 0.600000, -1.189011), vec4(-0.17611565, -0.81696892, 0.600000, -0.513724), vec4(-0.33930185, -0.20732205, 0.600000, -1.698047), vec4(-0.91974425, 0.05403209, 0.600000, 0.062246), + vec4(-0.15064627, -0.14949332, 0.600000, -1.896062), vec4(0.53180975, -0.35210401, 0.600000, -0.758838), vec4(0.41487166, 0.81442589, 0.600000, -0.505648), vec4(-0.24106961, -0.32721516, 0.600000, -1.665244) +}; + +// these values can be changed (up to SSIL_MAX_TAPS) with no changes required elsewhere; values for 4th and 5th preset are ignored but array needed to avoid compilation errors +// the actual number of texture samples is two times this value (each "tap" has two symmetrical depth texture samples) +const int num_taps[5] = { 3, 5, 12, 0, 0 }; + +#define SSIL_TILT_SAMPLES_ENABLE_AT_QUALITY_PRESET (99) // to disable simply set to 99 or similar +#define SSIL_TILT_SAMPLES_AMOUNT (0.4) +// +#define SSIL_HALOING_REDUCTION_ENABLE_AT_QUALITY_PRESET (1) // to disable simply set to 99 or similar +#define SSIL_HALOING_REDUCTION_AMOUNT (0.8) // values from 0.0 - 1.0, 1.0 means max weighting (will cause artifacts, 0.8 is more reasonable) +// +#define SSIL_DEPTH_MIPS_ENABLE_AT_QUALITY_PRESET (2) +#define SSIL_DEPTH_MIPS_GLOBAL_OFFSET (-4.3) // best noise/quality/performance tradeoff, found empirically +// +// !!warning!! the edge handling is hard-coded to 'disabled' on quality level 0, and enabled above, on the C++ side; while toggling it here will work for +// testing purposes, it will not yield performance gains (or correct results) +#define SSIL_DEPTH_BASED_EDGES_ENABLE_AT_QUALITY_PRESET (1) +// +#define SSIL_REDUCE_RADIUS_NEAR_SCREEN_BORDER_ENABLE_AT_QUALITY_PRESET (1) + +#define SSIL_MAX_TAPS 32 +#define SSIL_ADAPTIVE_TAP_BASE_COUNT 5 +#define SSIL_ADAPTIVE_TAP_FLEXIBLE_COUNT (SSIL_MAX_TAPS - SSIL_ADAPTIVE_TAP_BASE_COUNT) +#define SSIL_DEPTH_MIP_LEVELS 4 + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +layout(set = 0, binding = 0) uniform sampler2DArray source_depth_mipmaps; +layout(rgba8, set = 0, binding = 1) uniform restrict readonly image2D source_normal; +layout(set = 0, binding = 2) uniform Constants { //get into a lower set + vec4 rotation_matrices[20]; +} +constants; + +#ifdef ADAPTIVE +layout(rgba16, set = 1, binding = 0) uniform restrict readonly image2DArray source_ssil; +layout(set = 1, binding = 1) uniform sampler2D source_importance; +layout(set = 1, binding = 2, std430) buffer Counter { + uint sum; +} +counter; +#endif + +layout(rgba16, set = 2, binding = 0) uniform restrict writeonly image2D dest_image; +layout(r8, set = 2, binding = 1) uniform image2D edges_weights_image; + +layout(set = 3, binding = 0) uniform sampler2D last_frame; +layout(set = 3, binding = 1) uniform ProjectionConstants { + mat4 reprojection; +} +projection_constants; + +layout(push_constant, std430) uniform Params { + ivec2 screen_size; + int pass; + int quality; + + vec2 half_screen_pixel_size; + vec2 half_screen_pixel_size_x025; + + vec2 NDC_to_view_mul; + vec2 NDC_to_view_add; + + vec2 pad2; + float z_near; + float z_far; + + float radius; + float intensity; + int size_multiplier; + int pad; + + float fade_out_mul; + float fade_out_add; + float normal_rejection_amount; + float inv_radius_near_limit; + + bool is_orthogonal; + float neg_inv_radius; + float load_counter_avg_div; + float adaptive_sample_limit; + + ivec2 pass_coord_offset; + vec2 pass_uv_offset; +} +params; + +float pack_edges(vec4 p_edgesLRTB) { + p_edgesLRTB = round(clamp(p_edgesLRTB, 0.0, 1.0) * 3.05); + return dot(p_edgesLRTB, vec4(64.0 / 255.0, 16.0 / 255.0, 4.0 / 255.0, 1.0 / 255.0)); +} + +vec3 NDC_to_view_space(vec2 p_pos, float p_viewspace_depth) { + if (params.is_orthogonal) { + return vec3((params.NDC_to_view_mul * p_pos.xy + params.NDC_to_view_add), p_viewspace_depth); + } else { + return vec3((params.NDC_to_view_mul * p_pos.xy + params.NDC_to_view_add) * p_viewspace_depth, p_viewspace_depth); + } +} + +// calculate effect radius and fit our screen sampling pattern inside it +void calculate_radius_parameters(const float p_pix_center_length, const vec2 p_pixel_size_at_center, out float r_lookup_radius, out float r_radius, out float r_fallof_sq) { + r_radius = params.radius; + + // when too close, on-screen sampling disk will grow beyond screen size; limit this to avoid closeup temporal artifacts + const float too_close_limit = clamp(p_pix_center_length * params.inv_radius_near_limit, 0.0, 1.0) * 0.8 + 0.2; + + r_radius *= too_close_limit; + + // 0.85 is to reduce the radius to allow for more samples on a slope to still stay within influence + r_lookup_radius = (0.85 * r_radius) / p_pixel_size_at_center.x; + + // used to calculate falloff (both for AO samples and per-sample weights) + r_fallof_sq = -1.0 / (r_radius * r_radius); +} + +vec4 calculate_edges(const float p_center_z, const float p_left_z, const float p_right_z, const float p_top_z, const float p_bottom_z) { + // slope-sensitive depth-based edge detection + vec4 edgesLRTB = vec4(p_left_z, p_right_z, p_top_z, p_bottom_z) - p_center_z; + vec4 edgesLRTB_slope_adjusted = edgesLRTB + edgesLRTB.yxwz; + edgesLRTB = min(abs(edgesLRTB), abs(edgesLRTB_slope_adjusted)); + return clamp((1.3 - edgesLRTB / (p_center_z * 0.040)), 0.0, 1.0); +} + +vec3 decode_normal(vec3 p_encoded_normal) { + vec3 normal = p_encoded_normal * 2.0 - 1.0; + return normal; +} + +vec3 load_normal(ivec2 p_pos) { + vec3 encoded_normal = imageLoad(source_normal, p_pos).xyz; + encoded_normal.z = 1.0 - encoded_normal.z; + return decode_normal(encoded_normal); +} + +vec3 load_normal(ivec2 p_pos, ivec2 p_offset) { + vec3 encoded_normal = imageLoad(source_normal, p_pos + p_offset).xyz; + encoded_normal.z = 1.0 - encoded_normal.z; + return decode_normal(encoded_normal); +} + +// all vectors in viewspace +float calculate_pixel_obscurance(vec3 p_pixel_normal, vec3 p_hit_delta, float p_fallof_sq) { + float length_sq = dot(p_hit_delta, p_hit_delta); + float NdotD = dot(p_pixel_normal, p_hit_delta) / sqrt(length_sq); + + float falloff_mult = max(0.0, length_sq * p_fallof_sq + 1.0); + + return max(0, NdotD - 0.05) * falloff_mult; +} + +void SSIL_tap_inner(const int p_quality_level, inout vec3 r_color_sum, inout float r_obscurance_sum, inout float r_weight_sum, const vec2 p_sampling_uv, const float p_mip_level, const vec3 p_pix_center_pos, vec3 p_pixel_normal, const float p_fallof_sq, const float p_weight_mod) { + // get depth at sample + float viewspace_sample_z = textureLod(source_depth_mipmaps, vec3(p_sampling_uv, params.pass), p_mip_level).x; + vec3 sample_normal = load_normal(ivec2(p_sampling_uv * vec2(params.screen_size))); + + // convert to viewspace + vec3 hit_pos = NDC_to_view_space(p_sampling_uv.xy, viewspace_sample_z); + vec3 hit_delta = hit_pos - p_pix_center_pos; + + float obscurance = calculate_pixel_obscurance(p_pixel_normal, hit_delta, p_fallof_sq); + float weight = 1.0; + + if (p_quality_level >= SSIL_HALOING_REDUCTION_ENABLE_AT_QUALITY_PRESET) { + float reduct = max(0, -hit_delta.z); + reduct = clamp(reduct * params.neg_inv_radius + 2.0, 0.0, 1.0); + weight = SSIL_HALOING_REDUCTION_AMOUNT * reduct + (1.0 - SSIL_HALOING_REDUCTION_AMOUNT); + } + + // Translate sampling_uv to last screen's coordinates + const vec4 sample_pos = projection_constants.reprojection * vec4(p_sampling_uv * 2.0 - 1.0, (viewspace_sample_z - params.z_near) / (params.z_far - params.z_near) * 2.0 - 1.0, 1.0); + vec2 reprojected_sampling_uv = (sample_pos.xy / sample_pos.w) * 0.5 + 0.5; + + weight *= p_weight_mod; + + r_obscurance_sum += obscurance * weight; + + vec3 sample_color = textureLod(last_frame, reprojected_sampling_uv, 5.0).rgb; + // Reduce impact of fireflies by tonemapping before averaging: http://graphicrants.blogspot.com/2013/12/tone-mapping.html + sample_color /= (1.0 + dot(sample_color, vec3(0.299, 0.587, 0.114))); + r_color_sum += sample_color * obscurance * weight * mix(1.0, smoothstep(0.0, 0.1, -dot(sample_normal, normalize(hit_delta))), params.normal_rejection_amount); + r_weight_sum += weight; +} + +void SSILTap(const int p_quality_level, inout vec3 r_color_sum, inout float r_obscurance_sum, inout float r_weight_sum, const int p_tap_index, const mat2 p_rot_scale, const vec3 p_pix_center_pos, vec3 p_pixel_normal, const vec2 p_normalized_screen_pos, const float p_mip_offset, const float p_fallof_sq, float p_weight_mod, vec2 p_norm_xy, float p_norm_xy_length) { + vec2 sample_offset; + float sample_pow_2_len; + + // patterns + { + vec4 new_sample = sample_pattern[p_tap_index]; + sample_offset = new_sample.xy * p_rot_scale; + sample_pow_2_len = new_sample.w; // precalculated, same as: sample_pow_2_len = log2( length( new_sample.xy ) ); + p_weight_mod *= new_sample.z; + } + + // snap to pixel center (more correct obscurance math, avoids artifacts) + sample_offset = round(sample_offset); + + // calculate MIP based on the sample distance from the centre, similar to as described + // in http://graphics.cs.williams.edu/papers/SAOHPG12/. + float mip_level = (p_quality_level < SSIL_DEPTH_MIPS_ENABLE_AT_QUALITY_PRESET) ? (0) : (sample_pow_2_len + p_mip_offset); + + vec2 sampling_uv = sample_offset * params.half_screen_pixel_size + p_normalized_screen_pos; + + SSIL_tap_inner(p_quality_level, r_color_sum, r_obscurance_sum, r_weight_sum, sampling_uv, mip_level, p_pix_center_pos, p_pixel_normal, p_fallof_sq, p_weight_mod); + + // for the second tap, just use the mirrored offset + vec2 sample_offset_mirrored_uv = -sample_offset; + + // tilt the second set of samples so that the disk is effectively rotated by the normal + // effective at removing one set of artifacts, but too expensive for lower quality settings + if (p_quality_level >= SSIL_TILT_SAMPLES_ENABLE_AT_QUALITY_PRESET) { + float dot_norm = dot(sample_offset_mirrored_uv, p_norm_xy); + sample_offset_mirrored_uv -= dot_norm * p_norm_xy_length * p_norm_xy; + sample_offset_mirrored_uv = round(sample_offset_mirrored_uv); + } + + // snap to pixel center (more correct obscurance math, avoids artifacts) + vec2 sampling_mirrored_uv = sample_offset_mirrored_uv * params.half_screen_pixel_size + p_normalized_screen_pos; + + SSIL_tap_inner(p_quality_level, r_color_sum, r_obscurance_sum, r_weight_sum, sampling_mirrored_uv, mip_level, p_pix_center_pos, p_pixel_normal, p_fallof_sq, p_weight_mod); +} + +void generate_SSIL(out vec3 r_color, out vec4 r_edges, out float r_obscurance, out float r_weight, const vec2 p_pos, int p_quality_level, bool p_adaptive_base) { + vec2 pos_rounded = trunc(p_pos); + uvec2 upos = uvec2(pos_rounded); + + const int number_of_taps = (p_adaptive_base) ? (SSIL_ADAPTIVE_TAP_BASE_COUNT) : (num_taps[p_quality_level]); + float pix_z, pix_left_z, pix_top_z, pix_right_z, pix_bottom_z; + + vec4 valuesUL = textureGather(source_depth_mipmaps, vec3(pos_rounded * params.half_screen_pixel_size, params.pass)); + vec4 valuesBR = textureGather(source_depth_mipmaps, vec3((pos_rounded + vec2(1.0)) * params.half_screen_pixel_size, params.pass)); + + // get this pixel's viewspace depth + pix_z = valuesUL.y; + + // get left right top bottom neighbouring pixels for edge detection (gets compiled out on quality_level == 0) + pix_left_z = valuesUL.x; + pix_top_z = valuesUL.z; + pix_right_z = valuesBR.z; + pix_bottom_z = valuesBR.x; + + vec2 normalized_screen_pos = pos_rounded * params.half_screen_pixel_size + params.half_screen_pixel_size_x025; + vec3 pix_center_pos = NDC_to_view_space(normalized_screen_pos, pix_z); + + // Load this pixel's viewspace normal + uvec2 full_res_coord = upos * 2 * params.size_multiplier + params.pass_coord_offset.xy; + vec3 pixel_normal = load_normal(ivec2(full_res_coord)); + + const vec2 pixel_size_at_center = NDC_to_view_space(normalized_screen_pos.xy + params.half_screen_pixel_size, pix_center_pos.z).xy - pix_center_pos.xy; + + float pixel_lookup_radius; + float fallof_sq; + + // calculate effect radius and fit our screen sampling pattern inside it + float viewspace_radius; + calculate_radius_parameters(length(pix_center_pos), pixel_size_at_center, pixel_lookup_radius, viewspace_radius, fallof_sq); + + // calculate samples rotation/scaling + mat2 rot_scale_matrix; + uint pseudo_random_index; + + { + vec4 rotation_scale; + // reduce effect radius near the screen edges slightly; ideally, one would render a larger depth buffer (5% on each side) instead + if (!p_adaptive_base && (p_quality_level >= SSIL_REDUCE_RADIUS_NEAR_SCREEN_BORDER_ENABLE_AT_QUALITY_PRESET)) { + float near_screen_border = min(min(normalized_screen_pos.x, 1.0 - normalized_screen_pos.x), min(normalized_screen_pos.y, 1.0 - normalized_screen_pos.y)); + near_screen_border = clamp(10.0 * near_screen_border + 0.6, 0.0, 1.0); + pixel_lookup_radius *= near_screen_border; + } + + // load & update pseudo-random rotation matrix + pseudo_random_index = uint(pos_rounded.y * 2 + pos_rounded.x) % 5; + rotation_scale = constants.rotation_matrices[params.pass * 5 + pseudo_random_index]; + rot_scale_matrix = mat2(rotation_scale.x * pixel_lookup_radius, rotation_scale.y * pixel_lookup_radius, rotation_scale.z * pixel_lookup_radius, rotation_scale.w * pixel_lookup_radius); + } + + // the main obscurance & sample weight storage + vec3 color_sum = vec3(0.0); + float obscurance_sum = 0.0; + float weight_sum = 0.0; + + // edge mask for between this and left/right/top/bottom neighbour pixels - not used in quality level 0 so initialize to "no edge" (1 is no edge, 0 is edge) + vec4 edgesLRTB = vec4(1.0, 1.0, 1.0, 1.0); + + // Move center pixel slightly towards camera to avoid imprecision artifacts due to using of 16bit depth buffer; a lot smaller offsets needed when using 32bit floats + pix_center_pos *= 0.9992; + + if (!p_adaptive_base && (p_quality_level >= SSIL_DEPTH_BASED_EDGES_ENABLE_AT_QUALITY_PRESET)) { + edgesLRTB = calculate_edges(pix_z, pix_left_z, pix_right_z, pix_top_z, pix_bottom_z); + } + + const float global_mip_offset = SSIL_DEPTH_MIPS_GLOBAL_OFFSET; + float mip_offset = (p_quality_level < SSIL_DEPTH_MIPS_ENABLE_AT_QUALITY_PRESET) ? (0) : (log2(pixel_lookup_radius) + global_mip_offset); + + // Used to tilt the second set of samples so that the disk is effectively rotated by the normal + // effective at removing one set of artifacts, but too expensive for lower quality settings + vec2 norm_xy = vec2(pixel_normal.x, pixel_normal.y); + float norm_xy_length = length(norm_xy); + norm_xy /= vec2(norm_xy_length, -norm_xy_length); + norm_xy_length *= SSIL_TILT_SAMPLES_AMOUNT; + + // standard, non-adaptive approach + if ((p_quality_level != 3) || p_adaptive_base) { + for (int i = 0; i < number_of_taps; i++) { + SSILTap(p_quality_level, color_sum, obscurance_sum, weight_sum, i, rot_scale_matrix, pix_center_pos, pixel_normal, normalized_screen_pos, mip_offset, fallof_sq, 1.0, norm_xy, norm_xy_length); + } + } +#ifdef ADAPTIVE + else { + // add new ones if needed + vec2 full_res_uv = normalized_screen_pos + params.pass_uv_offset.xy; + float importance = textureLod(source_importance, full_res_uv, 0.0).x; + + //Need to store obscurance from base pass + // load existing base values + vec4 base_values = imageLoad(source_ssil, ivec3(upos, params.pass)); + weight_sum += imageLoad(edges_weights_image, ivec2(upos)).r * float(SSIL_ADAPTIVE_TAP_BASE_COUNT * 4.0); + color_sum += (base_values.rgb) * weight_sum; + obscurance_sum += (base_values.a) * weight_sum; + + // increase importance around edges + float edge_count = dot(1.0 - edgesLRTB, vec4(1.0, 1.0, 1.0, 1.0)); + + float avg_total_importance = float(counter.sum) * params.load_counter_avg_div; + + float importance_limiter = clamp(params.adaptive_sample_limit / avg_total_importance, 0.0, 1.0); + importance *= importance_limiter; + + float additional_sample_count = SSIL_ADAPTIVE_TAP_FLEXIBLE_COUNT * importance; + + const float blend_range = 3.0; + const float blend_range_inv = 1.0 / blend_range; + + additional_sample_count += 0.5; + uint additional_samples = uint(additional_sample_count); + uint additional_samples_to = min(SSIL_MAX_TAPS, additional_samples + SSIL_ADAPTIVE_TAP_BASE_COUNT); + + for (uint i = SSIL_ADAPTIVE_TAP_BASE_COUNT; i < additional_samples_to; i++) { + additional_sample_count -= 1.0f; + float weight_mod = clamp(additional_sample_count * blend_range_inv, 0.0, 1.0); + SSILTap(p_quality_level, color_sum, obscurance_sum, weight_sum, int(i), rot_scale_matrix, pix_center_pos, pixel_normal, normalized_screen_pos, mip_offset, fallof_sq, weight_mod, norm_xy, norm_xy_length); + } + } +#endif + + // Early out for adaptive base + if (p_adaptive_base) { + vec3 color = color_sum / weight_sum; + + r_color = color; + r_edges = vec4(0.0); + r_obscurance = obscurance_sum / weight_sum; + r_weight = weight_sum; + return; + } + + // Calculate weighted average + vec3 color = color_sum / weight_sum; + color /= 1.0 - dot(color, vec3(0.299, 0.587, 0.114)); + + // Calculate fadeout (1 close, gradient, 0 far) + float fade_out = clamp(pix_center_pos.z * params.fade_out_mul + params.fade_out_add, 0.0, 1.0); + + // Reduce the SSIL if we're on the edge to remove artifacts on edges (we don't care for the lower quality one) + if (!p_adaptive_base && (p_quality_level >= SSIL_DEPTH_BASED_EDGES_ENABLE_AT_QUALITY_PRESET)) { + // when there's more than 2 opposite edges, start fading out the occlusion to reduce aliasing artifacts + float edge_fadeout_factor = clamp((1.0 - edgesLRTB.x - edgesLRTB.y) * 0.35, 0.0, 1.0) + clamp((1.0 - edgesLRTB.z - edgesLRTB.w) * 0.35, 0.0, 1.0); + + fade_out *= clamp(1.0 - edge_fadeout_factor, 0.0, 1.0); + } + + color = params.intensity * color; + + color *= fade_out; + + // outputs! + r_color = color; + r_edges = edgesLRTB; // These are used to prevent blurring across edges, 1 means no edge, 0 means edge, 0.5 means half way there, etc. + r_obscurance = clamp((obscurance_sum / weight_sum) * params.intensity, 0.0, 1.0); + r_weight = weight_sum; +} + +void main() { + vec3 out_color; + float out_obscurance; + float out_weight; + vec4 out_edges; + ivec2 ssC = ivec2(gl_GlobalInvocationID.xy); + if (any(greaterThanEqual(ssC, params.screen_size))) { //too large, do nothing + return; + } + + vec2 uv = vec2(gl_GlobalInvocationID) + vec2(0.5); +#ifdef SSIL_BASE + generate_SSIL(out_color, out_edges, out_obscurance, out_weight, uv, params.quality, true); + + imageStore(dest_image, ssC, vec4(out_color, out_obscurance)); + imageStore(edges_weights_image, ssC, vec4(out_weight / (float(SSIL_ADAPTIVE_TAP_BASE_COUNT) * 4.0))); +#else + generate_SSIL(out_color, out_edges, out_obscurance, out_weight, uv, params.quality, false); // pass in quality levels + + imageStore(dest_image, ssC, vec4(out_color, out_obscurance)); + imageStore(edges_weights_image, ssC, vec4(pack_edges(out_edges))); +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/effects/ssil_blur.glsl b/servers/rendering/renderer_rd/shaders/effects/ssil_blur.glsl new file mode 100644 index 0000000000..47c56571f6 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/ssil_blur.glsl @@ -0,0 +1,144 @@ +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// Copyright (c) 2016, Intel Corporation +// Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated +// documentation files (the "Software"), to deal in the Software without restriction, including without limitation +// the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to +// permit persons to whom the Software is furnished to do so, subject to the following conditions: +// The above copyright notice and this permission notice shall be included in all copies or substantial portions of +// the Software. +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO +// THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +// TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +// SOFTWARE. +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// File changes (yyyy-mm-dd) +// 2016-09-07: filip.strugar@intel.com: first commit +// 2020-12-05: clayjohn: convert to Vulkan and Godot +// 2021-05-27: clayjohn: convert SSAO to SSIL +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// + +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +layout(set = 0, binding = 0) uniform sampler2D source_ssil; + +layout(rgba16, set = 1, binding = 0) uniform restrict writeonly image2D dest_image; + +layout(r8, set = 2, binding = 0) uniform restrict readonly image2D source_edges; + +layout(push_constant, std430) uniform Params { + float edge_sharpness; + float pad; + vec2 half_screen_pixel_size; +} +params; + +vec4 unpack_edges(float p_packed_val) { + uint packed_val = uint(p_packed_val * 255.5); + vec4 edgesLRTB; + edgesLRTB.x = float((packed_val >> 6) & 0x03) / 3.0; + edgesLRTB.y = float((packed_val >> 4) & 0x03) / 3.0; + edgesLRTB.z = float((packed_val >> 2) & 0x03) / 3.0; + edgesLRTB.w = float((packed_val >> 0) & 0x03) / 3.0; + + return clamp(edgesLRTB + params.edge_sharpness, 0.0, 1.0); +} + +void add_sample(vec4 p_ssil_value, float p_edge_value, inout vec4 r_sum, inout float r_sum_weight) { + float weight = p_edge_value; + + r_sum += (weight * p_ssil_value); + r_sum_weight += weight; +} + +#ifdef MODE_WIDE +vec4 sample_blurred_wide(ivec2 p_pos, vec2 p_coord) { + vec4 ssil_value = textureLodOffset(source_ssil, vec2(p_coord), 0.0, ivec2(0, 0)); + vec4 ssil_valueL = textureLodOffset(source_ssil, vec2(p_coord), 0.0, ivec2(-2, 0)); + vec4 ssil_valueT = textureLodOffset(source_ssil, vec2(p_coord), 0.0, ivec2(0, -2)); + vec4 ssil_valueR = textureLodOffset(source_ssil, vec2(p_coord), 0.0, ivec2(2, 0)); + vec4 ssil_valueB = textureLodOffset(source_ssil, vec2(p_coord), 0.0, ivec2(0, 2)); + + vec4 edgesLRTB = unpack_edges(imageLoad(source_edges, p_pos).r); + edgesLRTB.x *= unpack_edges(imageLoad(source_edges, p_pos + ivec2(-2, 0)).r).y; + edgesLRTB.z *= unpack_edges(imageLoad(source_edges, p_pos + ivec2(0, -2)).r).w; + edgesLRTB.y *= unpack_edges(imageLoad(source_edges, p_pos + ivec2(2, 0)).r).x; + edgesLRTB.w *= unpack_edges(imageLoad(source_edges, p_pos + ivec2(0, 2)).r).z; + + float sum_weight = 0.8; + vec4 sum = ssil_value * sum_weight; + + add_sample(ssil_valueL, edgesLRTB.x, sum, sum_weight); + add_sample(ssil_valueR, edgesLRTB.y, sum, sum_weight); + add_sample(ssil_valueT, edgesLRTB.z, sum, sum_weight); + add_sample(ssil_valueB, edgesLRTB.w, sum, sum_weight); + + vec4 ssil_avg = sum / sum_weight; + + ssil_value = ssil_avg; + + return ssil_value; +} +#endif + +#ifdef MODE_SMART +vec4 sample_blurred(ivec2 p_pos, vec2 p_coord) { + vec4 vC = textureLodOffset(source_ssil, vec2(p_coord), 0.0, ivec2(0, 0)); + vec4 vL = textureLodOffset(source_ssil, vec2(p_coord), 0.0, ivec2(-1, 0)); + vec4 vT = textureLodOffset(source_ssil, vec2(p_coord), 0.0, ivec2(0, -1)); + vec4 vR = textureLodOffset(source_ssil, vec2(p_coord), 0.0, ivec2(1, 0)); + vec4 vB = textureLodOffset(source_ssil, vec2(p_coord), 0.0, ivec2(0, 1)); + + float packed_edges = imageLoad(source_edges, p_pos).r; + vec4 edgesLRTB = unpack_edges(packed_edges); + + float sum_weight = 0.5; + vec4 sum = vC * sum_weight; + + add_sample(vL, edgesLRTB.x, sum, sum_weight); + add_sample(vR, edgesLRTB.y, sum, sum_weight); + add_sample(vT, edgesLRTB.z, sum, sum_weight); + add_sample(vB, edgesLRTB.w, sum, sum_weight); + + vec4 ssil_avg = sum / sum_weight; + + vec4 ssil_value = ssil_avg; + + return ssil_value; +} +#endif + +void main() { + // Pixel being shaded + ivec2 ssC = ivec2(gl_GlobalInvocationID.xy); + +#ifdef MODE_NON_SMART + + vec2 half_pixel = params.half_screen_pixel_size * 0.5; + + vec2 uv = (vec2(gl_GlobalInvocationID.xy) + vec2(0.5, 0.5)) * params.half_screen_pixel_size; + + vec4 centre = textureLod(source_ssil, uv, 0.0); + + vec4 value = textureLod(source_ssil, vec2(uv + vec2(-half_pixel.x * 3, -half_pixel.y)), 0.0) * 0.2; + value += textureLod(source_ssil, vec2(uv + vec2(+half_pixel.x, -half_pixel.y * 3)), 0.0) * 0.2; + value += textureLod(source_ssil, vec2(uv + vec2(-half_pixel.x, +half_pixel.y * 3)), 0.0) * 0.2; + value += textureLod(source_ssil, vec2(uv + vec2(+half_pixel.x * 3, +half_pixel.y)), 0.0) * 0.2; + + vec4 sampled = value + centre * 0.2; + +#else +#ifdef MODE_SMART + vec4 sampled = sample_blurred(ssC, (vec2(gl_GlobalInvocationID.xy) + vec2(0.5, 0.5)) * params.half_screen_pixel_size); +#else // MODE_WIDE + vec4 sampled = sample_blurred_wide(ssC, (vec2(gl_GlobalInvocationID.xy) + vec2(0.5, 0.5)) * params.half_screen_pixel_size); +#endif +#endif // MODE_NON_SMART + imageStore(dest_image, ssC, sampled); +} diff --git a/servers/rendering/renderer_rd/shaders/effects/ssil_importance_map.glsl b/servers/rendering/renderer_rd/shaders/effects/ssil_importance_map.glsl new file mode 100644 index 0000000000..6b6b02739d --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/ssil_importance_map.glsl @@ -0,0 +1,125 @@ +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// Copyright (c) 2016, Intel Corporation +// Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated +// documentation files (the "Software"), to deal in the Software without restriction, including without limitation +// the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to +// permit persons to whom the Software is furnished to do so, subject to the following conditions: +// The above copyright notice and this permission notice shall be included in all copies or substantial portions of +// the Software. +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO +// THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +// TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +// SOFTWARE. +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// File changes (yyyy-mm-dd) +// 2016-09-07: filip.strugar@intel.com: first commit +// 2020-12-05: clayjohn: convert to Vulkan and Godot +// 2021-05-27: clayjohn: convert SSAO to SSIL +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// + +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +#ifdef GENERATE_MAP +layout(set = 0, binding = 0) uniform sampler2DArray source_texture; +#else +layout(set = 0, binding = 0) uniform sampler2D source_importance; +#endif +layout(r8, set = 1, binding = 0) uniform restrict writeonly image2D dest_image; + +#ifdef PROCESS_MAPB +layout(set = 2, binding = 0, std430) buffer Counter { + uint sum; +} +counter; +#endif + +layout(push_constant, std430) uniform Params { + vec2 half_screen_pixel_size; + float intensity; + float pad; +} +params; + +void main() { + // Pixel being shaded + ivec2 ssC = ivec2(gl_GlobalInvocationID.xy); + +#ifdef GENERATE_MAP + // importance map stuff + uvec2 base_position = ssC * 2; + + float avg = 0.0; + float minV = 1.0; + float maxV = 0.0; + for (int i = 0; i < 4; i++) { + vec3 value_a = texelFetch(source_texture, ivec3(base_position, i), 0).rgb * params.intensity; + vec3 value_b = texelFetch(source_texture, ivec3(base_position, i) + ivec3(0, 1, 0), 0).rgb * params.intensity; + vec3 value_c = texelFetch(source_texture, ivec3(base_position, i) + ivec3(1, 0, 0), 0).rgb * params.intensity; + vec3 value_d = texelFetch(source_texture, ivec3(base_position, i) + ivec3(1, 1, 0), 0).rgb * params.intensity; + + // Calculate luminance (black and white value) + float a = dot(value_a, vec3(0.2125, 0.7154, 0.0721)); + float b = dot(value_b, vec3(0.2125, 0.7154, 0.0721)); + float c = dot(value_c, vec3(0.2125, 0.7154, 0.0721)); + float d = dot(value_d, vec3(0.2125, 0.7154, 0.0721)); + + maxV = max(maxV, max(max(a, b), max(c, d))); + minV = min(minV, min(min(a, b), min(c, d))); + } + + float min_max_diff = maxV - minV; + + imageStore(dest_image, ssC, vec4(pow(clamp(min_max_diff * 2.0, 0.0, 1.0), 0.6))); +#endif + +#ifdef PROCESS_MAPA + vec2 uv = (vec2(ssC) + 0.5) * params.half_screen_pixel_size * 2.0; + + float centre = textureLod(source_importance, uv, 0.0).x; + + vec2 half_pixel = params.half_screen_pixel_size; + + vec4 vals; + vals.x = textureLod(source_importance, uv + vec2(-half_pixel.x * 3, -half_pixel.y), 0.0).x; + vals.y = textureLod(source_importance, uv + vec2(+half_pixel.x, -half_pixel.y * 3), 0.0).x; + vals.z = textureLod(source_importance, uv + vec2(+half_pixel.x * 3, +half_pixel.y), 0.0).x; + vals.w = textureLod(source_importance, uv + vec2(-half_pixel.x, +half_pixel.y * 3), 0.0).x; + + float avg = dot(vals, vec4(0.25, 0.25, 0.25, 0.25)); + + imageStore(dest_image, ssC, vec4(avg)); +#endif + +#ifdef PROCESS_MAPB + vec2 uv = (vec2(ssC) + 0.5f) * params.half_screen_pixel_size * 2.0; + + float centre = textureLod(source_importance, uv, 0.0).x; + + vec2 half_pixel = params.half_screen_pixel_size; + + vec4 vals; + vals.x = textureLod(source_importance, uv + vec2(-half_pixel.x, -half_pixel.y * 3), 0.0).x; + vals.y = textureLod(source_importance, uv + vec2(+half_pixel.x * 3, -half_pixel.y), 0.0).x; + vals.z = textureLod(source_importance, uv + vec2(+half_pixel.x, +half_pixel.y * 3), 0.0).x; + vals.w = textureLod(source_importance, uv + vec2(-half_pixel.x * 3, +half_pixel.y), 0.0).x; + + float avg = dot(vals, vec4(0.25, 0.25, 0.25, 0.25)); + + imageStore(dest_image, ssC, vec4(avg)); + + // sum the average; to avoid overflowing we assume max AO resolution is not bigger than 16384x16384; so quarter res (used here) will be 4096x4096, which leaves us with 8 bits per pixel + uint sum = uint(clamp(avg, 0.0, 1.0) * 255.0 + 0.5); + + // save every 9th to avoid InterlockedAdd congestion - since we're blurring, this is good enough; compensated by multiplying load_counter_avg_div by 9 + if (((ssC.x % 3) + (ssC.y % 3)) == 0) { + atomicAdd(counter.sum, sum); + } +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/effects/ssil_interleave.glsl b/servers/rendering/renderer_rd/shaders/effects/ssil_interleave.glsl new file mode 100644 index 0000000000..9e86ac0cf0 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/ssil_interleave.glsl @@ -0,0 +1,122 @@ +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// Copyright (c) 2016, Intel Corporation +// Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated +// documentation files (the "Software"), to deal in the Software without restriction, including without limitation +// the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to +// permit persons to whom the Software is furnished to do so, subject to the following conditions: +// The above copyright notice and this permission notice shall be included in all copies or substantial portions of +// the Software. +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO +// THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +// TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +// SOFTWARE. +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// File changes (yyyy-mm-dd) +// 2016-09-07: filip.strugar@intel.com: first commit +// 2020-12-05: clayjohn: convert to Vulkan and Godot +// 2021-05-27: clayjohn: convert SSAO to SSIL +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// + +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +layout(rgba16, set = 0, binding = 0) uniform restrict writeonly image2D dest_image; +layout(set = 1, binding = 0) uniform sampler2DArray source_texture; +layout(r8, set = 2, binding = 0) uniform restrict readonly image2DArray source_edges; + +layout(push_constant, std430) uniform Params { + float inv_sharpness; + uint size_modifier; + vec2 pixel_size; +} +params; + +vec4 unpack_edges(float p_packed_val) { + uint packed_val = uint(p_packed_val * 255.5); + vec4 edgesLRTB; + edgesLRTB.x = float((packed_val >> 6) & 0x03) / 3.0; + edgesLRTB.y = float((packed_val >> 4) & 0x03) / 3.0; + edgesLRTB.z = float((packed_val >> 2) & 0x03) / 3.0; + edgesLRTB.w = float((packed_val >> 0) & 0x03) / 3.0; + + return clamp(edgesLRTB + params.inv_sharpness, 0.0, 1.0); +} + +void main() { + ivec2 ssC = ivec2(gl_GlobalInvocationID.xy); + if (any(greaterThanEqual(ssC, ivec2(1.0 / params.pixel_size)))) { //too large, do nothing + return; + } + +#ifdef MODE_SMART + uvec2 pix_pos = uvec2(gl_GlobalInvocationID.xy); + vec2 uv = (gl_GlobalInvocationID.xy + vec2(0.5)) * params.pixel_size; + + // calculate index in the four deinterleaved source array texture + int mx = int(pix_pos.x % 2); + int my = int(pix_pos.y % 2); + int index_center = mx + my * 2; // center index + int index_horizontal = (1 - mx) + my * 2; // neighbouring, horizontal + int index_vertical = mx + (1 - my) * 2; // neighbouring, vertical + int index_diagonal = (1 - mx) + (1 - my) * 2; // diagonal + + vec4 color = texelFetch(source_texture, ivec3(pix_pos / uvec2(params.size_modifier), index_center), 0); + + vec4 edgesLRTB = unpack_edges(imageLoad(source_edges, ivec3(pix_pos / uvec2(params.size_modifier), index_center)).r); + + // convert index shifts to sampling offsets + float fmx = float(mx); + float fmy = float(my); + + // in case of an edge, push sampling offsets away from the edge (towards pixel center) + float fmxe = (edgesLRTB.y - edgesLRTB.x); + float fmye = (edgesLRTB.w - edgesLRTB.z); + + // calculate final sampling offsets and sample using bilinear filter + vec2 uv_horizontal = (gl_GlobalInvocationID.xy + vec2(0.5) + vec2(fmx + fmxe - 0.5, 0.5 - fmy)) * params.pixel_size; + vec4 color_horizontal = textureLod(source_texture, vec3(uv_horizontal, index_horizontal), 0.0); + vec2 uv_vertical = (gl_GlobalInvocationID.xy + vec2(0.5) + vec2(0.5 - fmx, fmy - 0.5 + fmye)) * params.pixel_size; + vec4 color_vertical = textureLod(source_texture, vec3(uv_vertical, index_vertical), 0.0); + vec2 uv_diagonal = (gl_GlobalInvocationID.xy + vec2(0.5) + vec2(fmx - 0.5 + fmxe, fmy - 0.5 + fmye)) * params.pixel_size; + vec4 color_diagonal = textureLod(source_texture, vec3(uv_diagonal, index_diagonal), 0.0); + + // reduce weight for samples near edge - if the edge is on both sides, weight goes to 0 + vec4 blendWeights; + blendWeights.x = 1.0; + blendWeights.y = (edgesLRTB.x + edgesLRTB.y) * 0.5; + blendWeights.z = (edgesLRTB.z + edgesLRTB.w) * 0.5; + blendWeights.w = (blendWeights.y + blendWeights.z) * 0.5; + + // calculate weighted average + float blendWeightsSum = dot(blendWeights, vec4(1.0, 1.0, 1.0, 1.0)); + color += color_horizontal * blendWeights.y; + color += color_vertical * blendWeights.z; + color += color_diagonal * blendWeights.w; + color /= blendWeightsSum; + + imageStore(dest_image, ivec2(gl_GlobalInvocationID.xy), color); +#else // !MODE_SMART + + vec2 uv = (gl_GlobalInvocationID.xy + vec2(0.5)) * params.pixel_size; +#ifdef MODE_HALF + vec4 a = textureLod(source_texture, vec3(uv, 0), 0.0); + vec4 d = textureLod(source_texture, vec3(uv, 3), 0.0); + vec4 avg = (a + d) * 0.5; + +#else + vec4 a = textureLod(source_texture, vec3(uv, 0), 0.0); + vec4 b = textureLod(source_texture, vec3(uv, 1), 0.0); + vec4 c = textureLod(source_texture, vec3(uv, 2), 0.0); + vec4 d = textureLod(source_texture, vec3(uv, 3), 0.0); + vec4 avg = (a + b + c + d) * 0.25; + +#endif + imageStore(dest_image, ivec2(gl_GlobalInvocationID.xy), avg); +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/effects/subsurface_scattering.glsl b/servers/rendering/renderer_rd/shaders/effects/subsurface_scattering.glsl new file mode 100644 index 0000000000..fb35d3cde6 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/subsurface_scattering.glsl @@ -0,0 +1,189 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +#ifdef USE_25_SAMPLES +const int kernel_size = 13; + +const vec2 kernel[kernel_size] = vec2[]( + vec2(0.530605, 0.0), + vec2(0.0211412, 0.0208333), + vec2(0.0402784, 0.0833333), + vec2(0.0493588, 0.1875), + vec2(0.0410172, 0.333333), + vec2(0.0263642, 0.520833), + vec2(0.017924, 0.75), + vec2(0.0128496, 1.02083), + vec2(0.0094389, 1.33333), + vec2(0.00700976, 1.6875), + vec2(0.00500364, 2.08333), + vec2(0.00333804, 2.52083), + vec2(0.000973794, 3.0)); + +const vec4 skin_kernel[kernel_size] = vec4[]( + vec4(0.530605, 0.613514, 0.739601, 0), + vec4(0.0211412, 0.0459286, 0.0378196, 0.0208333), + vec4(0.0402784, 0.0657244, 0.04631, 0.0833333), + vec4(0.0493588, 0.0367726, 0.0219485, 0.1875), + vec4(0.0410172, 0.0199899, 0.0118481, 0.333333), + vec4(0.0263642, 0.0119715, 0.00684598, 0.520833), + vec4(0.017924, 0.00711691, 0.00347194, 0.75), + vec4(0.0128496, 0.00356329, 0.00132016, 1.02083), + vec4(0.0094389, 0.00139119, 0.000416598, 1.33333), + vec4(0.00700976, 0.00049366, 0.000151938, 1.6875), + vec4(0.00500364, 0.00020094, 5.28848e-005, 2.08333), + vec4(0.00333804, 7.85443e-005, 1.2945e-005, 2.52083), + vec4(0.000973794, 1.11862e-005, 9.43437e-007, 3)); + +#endif //USE_25_SAMPLES + +#ifdef USE_17_SAMPLES +const int kernel_size = 9; +const vec2 kernel[kernel_size] = vec2[]( + vec2(0.536343, 0.0), + vec2(0.0324462, 0.03125), + vec2(0.0582416, 0.125), + vec2(0.0571056, 0.28125), + vec2(0.0347317, 0.5), + vec2(0.0216301, 0.78125), + vec2(0.0144609, 1.125), + vec2(0.0100386, 1.53125), + vec2(0.00317394, 2.0)); + +const vec4 skin_kernel[kernel_size] = vec4[]( + vec4(0.536343, 0.624624, 0.748867, 0), + vec4(0.0324462, 0.0656718, 0.0532821, 0.03125), + vec4(0.0582416, 0.0659959, 0.0411329, 0.125), + vec4(0.0571056, 0.0287432, 0.0172844, 0.28125), + vec4(0.0347317, 0.0151085, 0.00871983, 0.5), + vec4(0.0216301, 0.00794618, 0.00376991, 0.78125), + vec4(0.0144609, 0.00317269, 0.00106399, 1.125), + vec4(0.0100386, 0.000914679, 0.000275702, 1.53125), + vec4(0.00317394, 0.000134823, 3.77269e-005, 2)); +#endif //USE_17_SAMPLES + +#ifdef USE_11_SAMPLES +const int kernel_size = 6; +const vec2 kernel[kernel_size] = vec2[]( + vec2(0.560479, 0.0), + vec2(0.0771802, 0.08), + vec2(0.0821904, 0.32), + vec2(0.03639, 0.72), + vec2(0.0192831, 1.28), + vec2(0.00471691, 2.0)); + +const vec4 skin_kernel[kernel_size] = vec4[]( + + vec4(0.560479, 0.669086, 0.784728, 0), + vec4(0.0771802, 0.113491, 0.0793803, 0.08), + vec4(0.0821904, 0.0358608, 0.0209261, 0.32), + vec4(0.03639, 0.0130999, 0.00643685, 0.72), + vec4(0.0192831, 0.00282018, 0.00084214, 1.28), + vec4(0.00471691, 0.000184771, 5.07565e-005, 2)); + +#endif //USE_11_SAMPLES + +layout(push_constant, std430) uniform Params { + ivec2 screen_size; + float camera_z_far; + float camera_z_near; + + bool vertical; + bool orthogonal; + float unit_size; + float scale; + + float depth_scale; + uint pad[3]; +} +params; + +layout(set = 0, binding = 0) uniform sampler2D source_image; +layout(rgba16f, set = 1, binding = 0) uniform restrict writeonly image2D dest_image; +layout(set = 2, binding = 0) uniform sampler2D source_depth; + +void do_filter(inout vec3 color_accum, inout vec3 divisor, vec2 uv, vec2 step, bool p_skin) { + // Accumulate the other samples: + for (int i = 1; i < kernel_size; i++) { + // Fetch color and depth for current sample: + vec2 offset = uv + kernel[i].y * step; + vec4 color = texture(source_image, offset); + + if (abs(color.a) < 0.001) { + break; //mix no more + } + + vec3 w; + if (p_skin) { + //skin + w = skin_kernel[i].rgb; + } else { + w = vec3(kernel[i].x); + } + + color_accum += color.rgb * w; + divisor += w; + } +} + +void main() { + // Pixel being shaded + ivec2 ssC = ivec2(gl_GlobalInvocationID.xy); + + if (any(greaterThanEqual(ssC, params.screen_size))) { //too large, do nothing + return; + } + + vec2 uv = (vec2(ssC) + 0.5) / vec2(params.screen_size); + + // Fetch color of current pixel: + vec4 base_color = texture(source_image, uv); + float strength = abs(base_color.a); + + if (strength > 0.0) { + vec2 dir = params.vertical ? vec2(0.0, 1.0) : vec2(1.0, 0.0); + + // Fetch linear depth of current pixel: + float depth = texture(source_depth, uv).r * 2.0 - 1.0; + float depth_scale; + + if (params.orthogonal) { + depth = ((depth + (params.camera_z_far + params.camera_z_near) / (params.camera_z_far - params.camera_z_near)) * (params.camera_z_far - params.camera_z_near)) / 2.0; + depth_scale = params.unit_size; //remember depth is negative by default in OpenGL + } else { + depth = 2.0 * params.camera_z_near * params.camera_z_far / (params.camera_z_far + params.camera_z_near - depth * (params.camera_z_far - params.camera_z_near)); + depth_scale = params.unit_size / depth; //remember depth is negative by default in OpenGL + } + + float scale = mix(params.scale, depth_scale, params.depth_scale); + + // Calculate the final step to fetch the surrounding pixels: + vec2 step = scale * dir; + step *= strength; + step /= 3.0; + // Accumulate the center sample: + + vec3 divisor; + bool skin = bool(base_color.a < 0.0); + + if (skin) { + //skin + divisor = skin_kernel[0].rgb; + } else { + divisor = vec3(kernel[0].x); + } + + vec3 color = base_color.rgb * divisor; + + do_filter(color, divisor, uv, step, skin); + do_filter(color, divisor, uv, -step, skin); + + base_color.rgb = color / divisor; + } + + imageStore(dest_image, ssC, base_color); +} diff --git a/servers/rendering/renderer_rd/shaders/effects/taa_resolve.glsl b/servers/rendering/renderer_rd/shaders/effects/taa_resolve.glsl new file mode 100644 index 0000000000..02566d8e35 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/taa_resolve.glsl @@ -0,0 +1,396 @@ +/////////////////////////////////////////////////////////////////////////////////// +// Copyright(c) 2016-2022 Panos Karabelas +// +// Permission is hereby granted, free of charge, to any person obtaining a copy +// of this software and associated documentation files (the "Software"), to deal +// in the Software without restriction, including without limitation the rights +// to use, copy, modify, merge, publish, distribute, sublicense, and / or sell +// copies of the Software, and to permit persons to whom the Software is furnished +// to do so, subject to the following conditions : +// +// The above copyright notice and this permission notice shall be included in +// all copies or substantial portions of the Software. +// +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS +// FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.IN NO EVENT SHALL THE AUTHORS OR +// COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER +// IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN +// CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +/////////////////////////////////////////////////////////////////////////////////// +// File changes (yyyy-mm-dd) +// 2022-05-06: Panos Karabelas: first commit +// 2020-12-05: Joan Fons: convert to Vulkan and Godot +/////////////////////////////////////////////////////////////////////////////////// + +#[compute] + +#version 450 + +#VERSION_DEFINES + +// Based on Spartan Engine's TAA implementation (without TAA upscale). +// <https://github.com/PanosK92/SpartanEngine/blob/a8338d0609b85dc32f3732a5c27fb4463816a3b9/Data/shaders/temporal_antialiasing.hlsl> + +#ifndef MOLTENVK_USED +#define USE_SUBGROUPS +#endif // MOLTENVK_USED + +#define GROUP_SIZE 8 +#define FLT_MIN 0.00000001 +#define FLT_MAX 32767.0 +#define RPC_9 0.11111111111 +#define RPC_16 0.0625 + +#ifdef USE_SUBGROUPS +layout(local_size_x = GROUP_SIZE, local_size_y = GROUP_SIZE, local_size_z = 1) in; +#endif + +layout(rgba16f, set = 0, binding = 0) uniform restrict readonly image2D color_buffer; +layout(set = 0, binding = 1) uniform sampler2D depth_buffer; +layout(rg16f, set = 0, binding = 2) uniform restrict readonly image2D velocity_buffer; +layout(rg16f, set = 0, binding = 3) uniform restrict readonly image2D last_velocity_buffer; +layout(set = 0, binding = 4) uniform sampler2D history_buffer; +layout(rgba16f, set = 0, binding = 5) uniform restrict writeonly image2D output_buffer; + +layout(push_constant, std430) uniform Params { + vec2 resolution; + float disocclusion_threshold; // 0.1 / max(params.resolution.x, params.resolution.y + float disocclusion_scale; +} +params; + +const ivec2 kOffsets3x3[9] = { + ivec2(-1, -1), + ivec2(0, -1), + ivec2(1, -1), + ivec2(-1, 0), + ivec2(0, 0), + ivec2(1, 0), + ivec2(-1, 1), + ivec2(0, 1), + ivec2(1, 1), +}; + +/*------------------------------------------------------------------------------ + THREAD GROUP SHARED MEMORY (LDS) +------------------------------------------------------------------------------*/ + +const int kBorderSize = 1; +const int kGroupSize = GROUP_SIZE; +const int kTileDimension = kGroupSize + kBorderSize * 2; +const int kTileDimension2 = kTileDimension * kTileDimension; + +vec3 reinhard(vec3 hdr) { + return hdr / (hdr + 1.0); +} +vec3 reinhard_inverse(vec3 sdr) { + return sdr / (1.0 - sdr); +} + +float get_depth(ivec2 thread_id) { + return texelFetch(depth_buffer, thread_id, 0).r; +} + +#ifdef USE_SUBGROUPS +shared vec3 tile_color[kTileDimension][kTileDimension]; +shared float tile_depth[kTileDimension][kTileDimension]; + +vec3 load_color(uvec2 group_thread_id) { + group_thread_id += kBorderSize; + return tile_color[group_thread_id.x][group_thread_id.y]; +} + +void store_color(uvec2 group_thread_id, vec3 color) { + tile_color[group_thread_id.x][group_thread_id.y] = color; +} + +float load_depth(uvec2 group_thread_id) { + group_thread_id += kBorderSize; + return tile_depth[group_thread_id.x][group_thread_id.y]; +} + +void store_depth(uvec2 group_thread_id, float depth) { + tile_depth[group_thread_id.x][group_thread_id.y] = depth; +} + +void store_color_depth(uvec2 group_thread_id, ivec2 thread_id) { + // out of bounds clamp + thread_id = clamp(thread_id, ivec2(0, 0), ivec2(params.resolution) - ivec2(1, 1)); + + store_color(group_thread_id, imageLoad(color_buffer, thread_id).rgb); + store_depth(group_thread_id, get_depth(thread_id)); +} + +void populate_group_shared_memory(uvec2 group_id, uint group_index) { + // Populate group shared memory + ivec2 group_top_left = ivec2(group_id) * kGroupSize - kBorderSize; + if (group_index < (kTileDimension2 >> 2)) { + ivec2 group_thread_id_1 = ivec2(group_index % kTileDimension, group_index / kTileDimension); + ivec2 group_thread_id_2 = ivec2((group_index + (kTileDimension2 >> 2)) % kTileDimension, (group_index + (kTileDimension2 >> 2)) / kTileDimension); + ivec2 group_thread_id_3 = ivec2((group_index + (kTileDimension2 >> 1)) % kTileDimension, (group_index + (kTileDimension2 >> 1)) / kTileDimension); + ivec2 group_thread_id_4 = ivec2((group_index + kTileDimension2 * 3 / 4) % kTileDimension, (group_index + kTileDimension2 * 3 / 4) / kTileDimension); + + store_color_depth(group_thread_id_1, group_top_left + group_thread_id_1); + store_color_depth(group_thread_id_2, group_top_left + group_thread_id_2); + store_color_depth(group_thread_id_3, group_top_left + group_thread_id_3); + store_color_depth(group_thread_id_4, group_top_left + group_thread_id_4); + } + + // Wait for group threads to load store data. + groupMemoryBarrier(); + barrier(); +} +#else +vec3 load_color(uvec2 screen_pos) { + return imageLoad(color_buffer, ivec2(screen_pos)).rgb; +} + +float load_depth(uvec2 screen_pos) { + return get_depth(ivec2(screen_pos)); +} +#endif + +/*------------------------------------------------------------------------------ + VELOCITY +------------------------------------------------------------------------------*/ + +void depth_test_min(uvec2 pos, inout float min_depth, inout uvec2 min_pos) { + float depth = load_depth(pos); + + if (depth < min_depth) { + min_depth = depth; + min_pos = pos; + } +} + +// Returns velocity with closest depth (3x3 neighborhood) +void get_closest_pixel_velocity_3x3(in uvec2 group_pos, uvec2 group_top_left, out vec2 velocity) { + float min_depth = 1.0; + uvec2 min_pos = group_pos; + + depth_test_min(group_pos + kOffsets3x3[0], min_depth, min_pos); + depth_test_min(group_pos + kOffsets3x3[1], min_depth, min_pos); + depth_test_min(group_pos + kOffsets3x3[2], min_depth, min_pos); + depth_test_min(group_pos + kOffsets3x3[3], min_depth, min_pos); + depth_test_min(group_pos + kOffsets3x3[4], min_depth, min_pos); + depth_test_min(group_pos + kOffsets3x3[5], min_depth, min_pos); + depth_test_min(group_pos + kOffsets3x3[6], min_depth, min_pos); + depth_test_min(group_pos + kOffsets3x3[7], min_depth, min_pos); + depth_test_min(group_pos + kOffsets3x3[8], min_depth, min_pos); + + // Velocity out + velocity = imageLoad(velocity_buffer, ivec2(group_top_left + min_pos)).xy; +} + +/*------------------------------------------------------------------------------ + HISTORY SAMPLING +------------------------------------------------------------------------------*/ + +vec3 sample_catmull_rom_9(sampler2D stex, vec2 uv, vec2 resolution) { + // Source: https://gist.github.com/TheRealMJP/c83b8c0f46b63f3a88a5986f4fa982b1 + // License: https://gist.github.com/TheRealMJP/bc503b0b87b643d3505d41eab8b332ae + + // We're going to sample a 4x4 grid of texels surrounding the target UV coordinate. We'll do this by rounding + // down the sample location to get the exact center of our "starting" texel. The starting texel will be at + // location [1, 1] in the grid, where [0, 0] is the top left corner. + vec2 sample_pos = uv * resolution; + vec2 texPos1 = floor(sample_pos - 0.5f) + 0.5f; + + // Compute the fractional offset from our starting texel to our original sample location, which we'll + // feed into the Catmull-Rom spline function to get our filter weights. + vec2 f = sample_pos - texPos1; + + // Compute the Catmull-Rom weights using the fractional offset that we calculated earlier. + // These equations are pre-expanded based on our knowledge of where the texels will be located, + // which lets us avoid having to evaluate a piece-wise function. + vec2 w0 = f * (-0.5f + f * (1.0f - 0.5f * f)); + vec2 w1 = 1.0f + f * f * (-2.5f + 1.5f * f); + vec2 w2 = f * (0.5f + f * (2.0f - 1.5f * f)); + vec2 w3 = f * f * (-0.5f + 0.5f * f); + + // Work out weighting factors and sampling offsets that will let us use bilinear filtering to + // simultaneously evaluate the middle 2 samples from the 4x4 grid. + vec2 w12 = w1 + w2; + vec2 offset12 = w2 / (w1 + w2); + + // Compute the final UV coordinates we'll use for sampling the texture + vec2 texPos0 = texPos1 - 1.0f; + vec2 texPos3 = texPos1 + 2.0f; + vec2 texPos12 = texPos1 + offset12; + + texPos0 /= resolution; + texPos3 /= resolution; + texPos12 /= resolution; + + vec3 result = vec3(0.0f, 0.0f, 0.0f); + + result += textureLod(stex, vec2(texPos0.x, texPos0.y), 0.0).xyz * w0.x * w0.y; + result += textureLod(stex, vec2(texPos12.x, texPos0.y), 0.0).xyz * w12.x * w0.y; + result += textureLod(stex, vec2(texPos3.x, texPos0.y), 0.0).xyz * w3.x * w0.y; + + result += textureLod(stex, vec2(texPos0.x, texPos12.y), 0.0).xyz * w0.x * w12.y; + result += textureLod(stex, vec2(texPos12.x, texPos12.y), 0.0).xyz * w12.x * w12.y; + result += textureLod(stex, vec2(texPos3.x, texPos12.y), 0.0).xyz * w3.x * w12.y; + + result += textureLod(stex, vec2(texPos0.x, texPos3.y), 0.0).xyz * w0.x * w3.y; + result += textureLod(stex, vec2(texPos12.x, texPos3.y), 0.0).xyz * w12.x * w3.y; + result += textureLod(stex, vec2(texPos3.x, texPos3.y), 0.0).xyz * w3.x * w3.y; + + return max(result, 0.0f); +} + +/*------------------------------------------------------------------------------ + HISTORY CLIPPING +------------------------------------------------------------------------------*/ + +// Based on "Temporal Reprojection Anti-Aliasing" - https://github.com/playdeadgames/temporal +vec3 clip_aabb(vec3 aabb_min, vec3 aabb_max, vec3 p, vec3 q) { + vec3 r = q - p; + vec3 rmax = (aabb_max - p.xyz); + vec3 rmin = (aabb_min - p.xyz); + + if (r.x > rmax.x + FLT_MIN) + r *= (rmax.x / r.x); + if (r.y > rmax.y + FLT_MIN) + r *= (rmax.y / r.y); + if (r.z > rmax.z + FLT_MIN) + r *= (rmax.z / r.z); + + if (r.x < rmin.x - FLT_MIN) + r *= (rmin.x / r.x); + if (r.y < rmin.y - FLT_MIN) + r *= (rmin.y / r.y); + if (r.z < rmin.z - FLT_MIN) + r *= (rmin.z / r.z); + + return p + r; +} + +// Clip history to the neighbourhood of the current sample +vec3 clip_history_3x3(uvec2 group_pos, vec3 color_history, vec2 velocity_closest) { + // Sample a 3x3 neighbourhood + vec3 s1 = load_color(group_pos + kOffsets3x3[0]); + vec3 s2 = load_color(group_pos + kOffsets3x3[1]); + vec3 s3 = load_color(group_pos + kOffsets3x3[2]); + vec3 s4 = load_color(group_pos + kOffsets3x3[3]); + vec3 s5 = load_color(group_pos + kOffsets3x3[4]); + vec3 s6 = load_color(group_pos + kOffsets3x3[5]); + vec3 s7 = load_color(group_pos + kOffsets3x3[6]); + vec3 s8 = load_color(group_pos + kOffsets3x3[7]); + vec3 s9 = load_color(group_pos + kOffsets3x3[8]); + + // Compute min and max (with an adaptive box size, which greatly reduces ghosting) + vec3 color_avg = (s1 + s2 + s3 + s4 + s5 + s6 + s7 + s8 + s9) * RPC_9; + vec3 color_avg2 = ((s1 * s1) + (s2 * s2) + (s3 * s3) + (s4 * s4) + (s5 * s5) + (s6 * s6) + (s7 * s7) + (s8 * s8) + (s9 * s9)) * RPC_9; + float box_size = mix(0.0f, 2.5f, smoothstep(0.02f, 0.0f, length(velocity_closest))); + vec3 dev = sqrt(abs(color_avg2 - (color_avg * color_avg))) * box_size; + vec3 color_min = color_avg - dev; + vec3 color_max = color_avg + dev; + + // Variance clipping + vec3 color = clip_aabb(color_min, color_max, clamp(color_avg, color_min, color_max), color_history); + + // Clamp to prevent NaNs + color = clamp(color, FLT_MIN, FLT_MAX); + + return color; +} + +/*------------------------------------------------------------------------------ + TAA +------------------------------------------------------------------------------*/ + +const vec3 lumCoeff = vec3(0.299f, 0.587f, 0.114f); + +float luminance(vec3 color) { + return max(dot(color, lumCoeff), 0.0001f); +} + +float get_factor_disocclusion(vec2 uv_reprojected, vec2 velocity) { + vec2 velocity_previous = imageLoad(last_velocity_buffer, ivec2(uv_reprojected * params.resolution)).xy; + vec2 velocity_texels = velocity * params.resolution; + vec2 prev_velocity_texels = velocity_previous * params.resolution; + float disocclusion = length(prev_velocity_texels - velocity_texels) - params.disocclusion_threshold; + return clamp(disocclusion * params.disocclusion_scale, 0.0, 1.0); +} + +vec3 temporal_antialiasing(uvec2 pos_group_top_left, uvec2 pos_group, uvec2 pos_screen, vec2 uv, sampler2D tex_history) { + // Get the velocity of the current pixel + vec2 velocity = imageLoad(velocity_buffer, ivec2(pos_screen)).xy; + + // Get reprojected uv + vec2 uv_reprojected = uv - velocity; + + // Get input color + vec3 color_input = load_color(pos_group); + + // Get history color (catmull-rom reduces a lot of the blurring that you get under motion) + vec3 color_history = sample_catmull_rom_9(tex_history, uv_reprojected, params.resolution).rgb; + + // Clip history to the neighbourhood of the current sample (fixes a lot of the ghosting). + vec2 velocity_closest = vec2(0.0); // This is best done by using the velocity with the closest depth. + get_closest_pixel_velocity_3x3(pos_group, pos_group_top_left, velocity_closest); + color_history = clip_history_3x3(pos_group, color_history, velocity_closest); + + // Compute blend factor + float blend_factor = RPC_16; // We want to be able to accumulate as many jitter samples as we generated, that is, 16. + { + // If re-projected UV is out of screen, converge to current color immediatel + float factor_screen = any(lessThan(uv_reprojected, vec2(0.0))) || any(greaterThan(uv_reprojected, vec2(1.0))) ? 1.0 : 0.0; + + // Increase blend factor when there is disocclusion (fixes a lot of the remaining ghosting). + float factor_disocclusion = get_factor_disocclusion(uv_reprojected, velocity); + + // Add to the blend factor + blend_factor = clamp(blend_factor + factor_screen + factor_disocclusion, 0.0, 1.0); + } + + // Resolve + vec3 color_resolved = vec3(0.0); + { + // Tonemap + color_history = reinhard(color_history); + color_input = reinhard(color_input); + + // Reduce flickering + float lum_color = luminance(color_input); + float lum_history = luminance(color_history); + float diff = abs(lum_color - lum_history) / max(lum_color, max(lum_history, 1.001)); + diff = 1.0 - diff; + diff = diff * diff; + blend_factor = mix(0.0, blend_factor, diff); + + // Lerp/blend + color_resolved = mix(color_history, color_input, blend_factor); + + // Inverse tonemap + color_resolved = reinhard_inverse(color_resolved); + } + + return color_resolved; +} + +void main() { +#ifdef USE_SUBGROUPS + populate_group_shared_memory(gl_WorkGroupID.xy, gl_LocalInvocationIndex); +#endif + + // Out of bounds check + if (any(greaterThanEqual(vec2(gl_GlobalInvocationID.xy), params.resolution))) { + return; + } + +#ifdef USE_SUBGROUPS + const uvec2 pos_group = gl_LocalInvocationID.xy; + const uvec2 pos_group_top_left = gl_WorkGroupID.xy * kGroupSize - kBorderSize; +#else + const uvec2 pos_group = gl_GlobalInvocationID.xy; + const uvec2 pos_group_top_left = uvec2(0, 0); +#endif + const uvec2 pos_screen = gl_GlobalInvocationID.xy; + const vec2 uv = (gl_GlobalInvocationID.xy + 0.5f) / params.resolution; + + vec3 result = temporal_antialiasing(pos_group_top_left, pos_group, pos_screen, uv, history_buffer); + imageStore(output_buffer, ivec2(gl_GlobalInvocationID.xy), vec4(result, 1.0)); +} diff --git a/servers/rendering/renderer_rd/shaders/effects/tonemap.glsl b/servers/rendering/renderer_rd/shaders/effects/tonemap.glsl new file mode 100644 index 0000000000..52aee8b648 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/tonemap.glsl @@ -0,0 +1,502 @@ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#ifdef MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#endif +#endif + +layout(location = 0) out vec2 uv_interp; + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0); +} + +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#ifdef MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#endif //MULTIVIEW + +layout(location = 0) in vec2 uv_interp; + +#ifdef SUBPASS +layout(input_attachment_index = 0, set = 0, binding = 0) uniform subpassInput input_color; +#elif defined(MULTIVIEW) +layout(set = 0, binding = 0) uniform sampler2DArray source_color; +#else +layout(set = 0, binding = 0) uniform sampler2D source_color; +#endif + +layout(set = 1, binding = 0) uniform sampler2D source_auto_exposure; +#ifdef MULTIVIEW +layout(set = 2, binding = 0) uniform sampler2DArray source_glow; +#else +layout(set = 2, binding = 0) uniform sampler2D source_glow; +#endif +layout(set = 2, binding = 1) uniform sampler2D glow_map; + +#ifdef USE_1D_LUT +layout(set = 3, binding = 0) uniform sampler2D source_color_correction; +#else +layout(set = 3, binding = 0) uniform sampler3D source_color_correction; +#endif + +layout(push_constant, std430) uniform Params { + vec3 bcs; + bool use_bcs; + + bool use_glow; + bool use_auto_exposure; + bool use_color_correction; + uint tonemapper; + + uvec2 glow_texture_size; + float glow_intensity; + float glow_map_strength; + + uint glow_mode; + float glow_levels[7]; + + float exposure; + float white; + float auto_exposure_scale; + float luminance_multiplier; + + vec2 pixel_size; + bool use_fxaa; + bool use_debanding; +} +params; + +layout(location = 0) out vec4 frag_color; + +#ifdef USE_GLOW_FILTER_BICUBIC +// w0, w1, w2, and w3 are the four cubic B-spline basis functions +float w0(float a) { + return (1.0f / 6.0f) * (a * (a * (-a + 3.0f) - 3.0f) + 1.0f); +} + +float w1(float a) { + return (1.0f / 6.0f) * (a * a * (3.0f * a - 6.0f) + 4.0f); +} + +float w2(float a) { + return (1.0f / 6.0f) * (a * (a * (-3.0f * a + 3.0f) + 3.0f) + 1.0f); +} + +float w3(float a) { + return (1.0f / 6.0f) * (a * a * a); +} + +// g0 and g1 are the two amplitude functions +float g0(float a) { + return w0(a) + w1(a); +} + +float g1(float a) { + return w2(a) + w3(a); +} + +// h0 and h1 are the two offset functions +float h0(float a) { + return -1.0f + w1(a) / (w0(a) + w1(a)); +} + +float h1(float a) { + return 1.0f + w3(a) / (w2(a) + w3(a)); +} + +#ifdef MULTIVIEW +vec4 texture2D_bicubic(sampler2DArray tex, vec2 uv, int p_lod) { + float lod = float(p_lod); + vec2 tex_size = vec2(params.glow_texture_size >> p_lod); + vec2 pixel_size = vec2(1.0f) / tex_size; + + uv = uv * tex_size + vec2(0.5f); + + vec2 iuv = floor(uv); + vec2 fuv = fract(uv); + + float g0x = g0(fuv.x); + float g1x = g1(fuv.x); + float h0x = h0(fuv.x); + float h1x = h1(fuv.x); + float h0y = h0(fuv.y); + float h1y = h1(fuv.y); + + vec3 p0 = vec3((vec2(iuv.x + h0x, iuv.y + h0y) - vec2(0.5f)) * pixel_size, ViewIndex); + vec3 p1 = vec3((vec2(iuv.x + h1x, iuv.y + h0y) - vec2(0.5f)) * pixel_size, ViewIndex); + vec3 p2 = vec3((vec2(iuv.x + h0x, iuv.y + h1y) - vec2(0.5f)) * pixel_size, ViewIndex); + vec3 p3 = vec3((vec2(iuv.x + h1x, iuv.y + h1y) - vec2(0.5f)) * pixel_size, ViewIndex); + + return (g0(fuv.y) * (g0x * textureLod(tex, p0, lod) + g1x * textureLod(tex, p1, lod))) + + (g1(fuv.y) * (g0x * textureLod(tex, p2, lod) + g1x * textureLod(tex, p3, lod))); +} + +#define GLOW_TEXTURE_SAMPLE(m_tex, m_uv, m_lod) texture2D_bicubic(m_tex, m_uv, m_lod) +#else // MULTIVIEW + +vec4 texture2D_bicubic(sampler2D tex, vec2 uv, int p_lod) { + float lod = float(p_lod); + vec2 tex_size = vec2(params.glow_texture_size >> p_lod); + vec2 pixel_size = vec2(1.0f) / tex_size; + + uv = uv * tex_size + vec2(0.5f); + + vec2 iuv = floor(uv); + vec2 fuv = fract(uv); + + float g0x = g0(fuv.x); + float g1x = g1(fuv.x); + float h0x = h0(fuv.x); + float h1x = h1(fuv.x); + float h0y = h0(fuv.y); + float h1y = h1(fuv.y); + + vec2 p0 = (vec2(iuv.x + h0x, iuv.y + h0y) - vec2(0.5f)) * pixel_size; + vec2 p1 = (vec2(iuv.x + h1x, iuv.y + h0y) - vec2(0.5f)) * pixel_size; + vec2 p2 = (vec2(iuv.x + h0x, iuv.y + h1y) - vec2(0.5f)) * pixel_size; + vec2 p3 = (vec2(iuv.x + h1x, iuv.y + h1y) - vec2(0.5f)) * pixel_size; + + return (g0(fuv.y) * (g0x * textureLod(tex, p0, lod) + g1x * textureLod(tex, p1, lod))) + + (g1(fuv.y) * (g0x * textureLod(tex, p2, lod) + g1x * textureLod(tex, p3, lod))); +} + +#define GLOW_TEXTURE_SAMPLE(m_tex, m_uv, m_lod) texture2D_bicubic(m_tex, m_uv, m_lod) +#endif // !MULTIVIEW + +#else // USE_GLOW_FILTER_BICUBIC + +#ifdef MULTIVIEW +#define GLOW_TEXTURE_SAMPLE(m_tex, m_uv, m_lod) textureLod(m_tex, vec3(m_uv, ViewIndex), float(m_lod)) +#else // MULTIVIEW +#define GLOW_TEXTURE_SAMPLE(m_tex, m_uv, m_lod) textureLod(m_tex, m_uv, float(m_lod)) +#endif // !MULTIVIEW + +#endif // !USE_GLOW_FILTER_BICUBIC + +vec3 tonemap_filmic(vec3 color, float white) { + // exposure bias: input scale (color *= bias, white *= bias) to make the brightness consistent with other tonemappers + // also useful to scale the input to the range that the tonemapper is designed for (some require very high input values) + // has no effect on the curve's general shape or visual properties + const float exposure_bias = 2.0f; + const float A = 0.22f * exposure_bias * exposure_bias; // bias baked into constants for performance + const float B = 0.30f * exposure_bias; + const float C = 0.10f; + const float D = 0.20f; + const float E = 0.01f; + const float F = 0.30f; + + vec3 color_tonemapped = ((color * (A * color + C * B) + D * E) / (color * (A * color + B) + D * F)) - E / F; + float white_tonemapped = ((white * (A * white + C * B) + D * E) / (white * (A * white + B) + D * F)) - E / F; + + return color_tonemapped / white_tonemapped; +} + +// Adapted from https://github.com/TheRealMJP/BakingLab/blob/master/BakingLab/ACES.hlsl +// (MIT License). +vec3 tonemap_aces(vec3 color, float white) { + const float exposure_bias = 1.8f; + const float A = 0.0245786f; + const float B = 0.000090537f; + const float C = 0.983729f; + const float D = 0.432951f; + const float E = 0.238081f; + + // Exposure bias baked into transform to save shader instructions. Equivalent to `color *= exposure_bias` + const mat3 rgb_to_rrt = mat3( + vec3(0.59719f * exposure_bias, 0.35458f * exposure_bias, 0.04823f * exposure_bias), + vec3(0.07600f * exposure_bias, 0.90834f * exposure_bias, 0.01566f * exposure_bias), + vec3(0.02840f * exposure_bias, 0.13383f * exposure_bias, 0.83777f * exposure_bias)); + + const mat3 odt_to_rgb = mat3( + vec3(1.60475f, -0.53108f, -0.07367f), + vec3(-0.10208f, 1.10813f, -0.00605f), + vec3(-0.00327f, -0.07276f, 1.07602f)); + + color *= rgb_to_rrt; + vec3 color_tonemapped = (color * (color + A) - B) / (color * (C * color + D) + E); + color_tonemapped *= odt_to_rgb; + + white *= exposure_bias; + float white_tonemapped = (white * (white + A) - B) / (white * (C * white + D) + E); + + return color_tonemapped / white_tonemapped; +} + +vec3 tonemap_reinhard(vec3 color, float white) { + return (white * color + color) / (color * white + white); +} + +vec3 linear_to_srgb(vec3 color) { + //if going to srgb, clamp from 0 to 1. + color = clamp(color, vec3(0.0), vec3(1.0)); + const vec3 a = vec3(0.055f); + return mix((vec3(1.0f) + a) * pow(color.rgb, vec3(1.0f / 2.4f)) - a, 12.92f * color.rgb, lessThan(color.rgb, vec3(0.0031308f))); +} + +#define TONEMAPPER_LINEAR 0 +#define TONEMAPPER_REINHARD 1 +#define TONEMAPPER_FILMIC 2 +#define TONEMAPPER_ACES 3 + +vec3 apply_tonemapping(vec3 color, float white) { // inputs are LINEAR, always outputs clamped [0;1] color + // Ensure color values passed to tonemappers are positive. + // They can be negative in the case of negative lights, which leads to undesired behavior. + if (params.tonemapper == TONEMAPPER_LINEAR) { + return color; + } else if (params.tonemapper == TONEMAPPER_REINHARD) { + return tonemap_reinhard(max(vec3(0.0f), color), white); + } else if (params.tonemapper == TONEMAPPER_FILMIC) { + return tonemap_filmic(max(vec3(0.0f), color), white); + } else { // TONEMAPPER_ACES + return tonemap_aces(max(vec3(0.0f), color), white); + } +} + +#ifdef MULTIVIEW +vec3 gather_glow(sampler2DArray tex, vec2 uv) { // sample all selected glow levels, view is added to uv later +#else +vec3 gather_glow(sampler2D tex, vec2 uv) { // sample all selected glow levels +#endif // defined(MULTIVIEW) + vec3 glow = vec3(0.0f); + + if (params.glow_levels[0] > 0.0001) { + glow += GLOW_TEXTURE_SAMPLE(tex, uv, 0).rgb * params.glow_levels[0]; + } + + if (params.glow_levels[1] > 0.0001) { + glow += GLOW_TEXTURE_SAMPLE(tex, uv, 1).rgb * params.glow_levels[1]; + } + + if (params.glow_levels[2] > 0.0001) { + glow += GLOW_TEXTURE_SAMPLE(tex, uv, 2).rgb * params.glow_levels[2]; + } + + if (params.glow_levels[3] > 0.0001) { + glow += GLOW_TEXTURE_SAMPLE(tex, uv, 3).rgb * params.glow_levels[3]; + } + + if (params.glow_levels[4] > 0.0001) { + glow += GLOW_TEXTURE_SAMPLE(tex, uv, 4).rgb * params.glow_levels[4]; + } + + if (params.glow_levels[5] > 0.0001) { + glow += GLOW_TEXTURE_SAMPLE(tex, uv, 5).rgb * params.glow_levels[5]; + } + + if (params.glow_levels[6] > 0.0001) { + glow += GLOW_TEXTURE_SAMPLE(tex, uv, 6).rgb * params.glow_levels[6]; + } + + return glow; +} + +#define GLOW_MODE_ADD 0 +#define GLOW_MODE_SCREEN 1 +#define GLOW_MODE_SOFTLIGHT 2 +#define GLOW_MODE_REPLACE 3 +#define GLOW_MODE_MIX 4 + +vec3 apply_glow(vec3 color, vec3 glow) { // apply glow using the selected blending mode + if (params.glow_mode == GLOW_MODE_ADD) { + return color + glow; + } else if (params.glow_mode == GLOW_MODE_SCREEN) { + //need color clamping + return max((color + glow) - (color * glow), vec3(0.0)); + } else if (params.glow_mode == GLOW_MODE_SOFTLIGHT) { + //need color clamping + glow = glow * vec3(0.5f) + vec3(0.5f); + + color.r = (glow.r <= 0.5f) ? (color.r - (1.0f - 2.0f * glow.r) * color.r * (1.0f - color.r)) : (((glow.r > 0.5f) && (color.r <= 0.25f)) ? (color.r + (2.0f * glow.r - 1.0f) * (4.0f * color.r * (4.0f * color.r + 1.0f) * (color.r - 1.0f) + 7.0f * color.r)) : (color.r + (2.0f * glow.r - 1.0f) * (sqrt(color.r) - color.r))); + color.g = (glow.g <= 0.5f) ? (color.g - (1.0f - 2.0f * glow.g) * color.g * (1.0f - color.g)) : (((glow.g > 0.5f) && (color.g <= 0.25f)) ? (color.g + (2.0f * glow.g - 1.0f) * (4.0f * color.g * (4.0f * color.g + 1.0f) * (color.g - 1.0f) + 7.0f * color.g)) : (color.g + (2.0f * glow.g - 1.0f) * (sqrt(color.g) - color.g))); + color.b = (glow.b <= 0.5f) ? (color.b - (1.0f - 2.0f * glow.b) * color.b * (1.0f - color.b)) : (((glow.b > 0.5f) && (color.b <= 0.25f)) ? (color.b + (2.0f * glow.b - 1.0f) * (4.0f * color.b * (4.0f * color.b + 1.0f) * (color.b - 1.0f) + 7.0f * color.b)) : (color.b + (2.0f * glow.b - 1.0f) * (sqrt(color.b) - color.b))); + return color; + } else { //replace + return glow; + } +} + +vec3 apply_bcs(vec3 color, vec3 bcs) { + color = mix(vec3(0.0f), color, bcs.x); + color = mix(vec3(0.5f), color, bcs.y); + color = mix(vec3(dot(vec3(1.0f), color) * 0.33333f), color, bcs.z); + + return color; +} +#ifdef USE_1D_LUT +vec3 apply_color_correction(vec3 color) { + color.r = texture(source_color_correction, vec2(color.r, 0.0f)).r; + color.g = texture(source_color_correction, vec2(color.g, 0.0f)).g; + color.b = texture(source_color_correction, vec2(color.b, 0.0f)).b; + return color; +} +#else +vec3 apply_color_correction(vec3 color) { + return textureLod(source_color_correction, color, 0.0).rgb; +} +#endif + +#ifndef SUBPASS +vec3 do_fxaa(vec3 color, float exposure, vec2 uv_interp) { + const float FXAA_REDUCE_MIN = (1.0 / 128.0); + const float FXAA_REDUCE_MUL = (1.0 / 8.0); + const float FXAA_SPAN_MAX = 8.0; + +#ifdef MULTIVIEW + vec3 rgbNW = textureLod(source_color, vec3(uv_interp + vec2(-0.5, -0.5) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbNE = textureLod(source_color, vec3(uv_interp + vec2(0.5, -0.5) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbSW = textureLod(source_color, vec3(uv_interp + vec2(-0.5, 0.5) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbSE = textureLod(source_color, vec3(uv_interp + vec2(0.5, 0.5) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier; +#else + vec3 rgbNW = textureLod(source_color, uv_interp + vec2(-0.5, -0.5) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbNE = textureLod(source_color, uv_interp + vec2(0.5, -0.5) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbSW = textureLod(source_color, uv_interp + vec2(-0.5, 0.5) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbSE = textureLod(source_color, uv_interp + vec2(0.5, 0.5) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier; +#endif + vec3 rgbM = color; + vec3 luma = vec3(0.299, 0.587, 0.114); + float lumaNW = dot(rgbNW, luma); + float lumaNE = dot(rgbNE, luma); + float lumaSW = dot(rgbSW, luma); + float lumaSE = dot(rgbSE, luma); + float lumaM = dot(rgbM, luma); + float lumaMin = min(lumaM, min(min(lumaNW, lumaNE), min(lumaSW, lumaSE))); + float lumaMax = max(lumaM, max(max(lumaNW, lumaNE), max(lumaSW, lumaSE))); + + vec2 dir; + dir.x = -((lumaNW + lumaNE) - (lumaSW + lumaSE)); + dir.y = ((lumaNW + lumaSW) - (lumaNE + lumaSE)); + + float dirReduce = max((lumaNW + lumaNE + lumaSW + lumaSE) * + (0.25 * FXAA_REDUCE_MUL), + FXAA_REDUCE_MIN); + + float rcpDirMin = 1.0 / (min(abs(dir.x), abs(dir.y)) + dirReduce); + dir = min(vec2(FXAA_SPAN_MAX, FXAA_SPAN_MAX), + max(vec2(-FXAA_SPAN_MAX, -FXAA_SPAN_MAX), + dir * rcpDirMin)) * + params.pixel_size; + +#ifdef MULTIVIEW + vec3 rgbA = 0.5 * exposure * (textureLod(source_color, vec3(uv_interp + dir * (1.0 / 3.0 - 0.5), ViewIndex), 0.0).xyz + textureLod(source_color, vec3(uv_interp + dir * (2.0 / 3.0 - 0.5), ViewIndex), 0.0).xyz) * params.luminance_multiplier; + vec3 rgbB = rgbA * 0.5 + 0.25 * exposure * (textureLod(source_color, vec3(uv_interp + dir * -0.5, ViewIndex), 0.0).xyz + textureLod(source_color, vec3(uv_interp + dir * 0.5, ViewIndex), 0.0).xyz) * params.luminance_multiplier; +#else + vec3 rgbA = 0.5 * exposure * (textureLod(source_color, uv_interp + dir * (1.0 / 3.0 - 0.5), 0.0).xyz + textureLod(source_color, uv_interp + dir * (2.0 / 3.0 - 0.5), 0.0).xyz) * params.luminance_multiplier; + vec3 rgbB = rgbA * 0.5 + 0.25 * exposure * (textureLod(source_color, uv_interp + dir * -0.5, 0.0).xyz + textureLod(source_color, uv_interp + dir * 0.5, 0.0).xyz) * params.luminance_multiplier; +#endif + + float lumaB = dot(rgbB, luma); + if ((lumaB < lumaMin) || (lumaB > lumaMax)) { + return rgbA; + } else { + return rgbB; + } +} +#endif // !SUBPASS + +// From https://alex.vlachos.com/graphics/Alex_Vlachos_Advanced_VR_Rendering_GDC2015.pdf +// and https://www.shadertoy.com/view/MslGR8 (5th one starting from the bottom) +// NOTE: `frag_coord` is in pixels (i.e. not normalized UV). +vec3 screen_space_dither(vec2 frag_coord) { + // Iestyn's RGB dither (7 asm instructions) from Portal 2 X360, slightly modified for VR. + vec3 dither = vec3(dot(vec2(171.0, 231.0), frag_coord)); + dither.rgb = fract(dither.rgb / vec3(103.0, 71.0, 97.0)); + + // Subtract 0.5 to avoid slightly brightening the whole viewport. + return (dither.rgb - 0.5) / 255.0; +} + +void main() { +#ifdef SUBPASS + // SUBPASS and MULTIVIEW can be combined but in that case we're already reading from the correct layer + vec4 color = subpassLoad(input_color); +#elif defined(MULTIVIEW) + vec4 color = textureLod(source_color, vec3(uv_interp, ViewIndex), 0.0f); +#else + vec4 color = textureLod(source_color, uv_interp, 0.0f); +#endif + color.rgb *= params.luminance_multiplier; + + // Exposure + + float exposure = params.exposure; + +#ifndef SUBPASS + if (params.use_auto_exposure) { + exposure *= 1.0 / (texelFetch(source_auto_exposure, ivec2(0, 0), 0).r * params.luminance_multiplier / params.auto_exposure_scale); + } +#endif + + color.rgb *= exposure; + + // Early Tonemap & SRGB Conversion +#ifndef SUBPASS + if (params.use_fxaa) { + // FXAA must be performed before glow to preserve the "bleed" effect of glow. + color.rgb = do_fxaa(color.rgb, exposure, uv_interp); + } + + if (params.use_glow && params.glow_mode == GLOW_MODE_MIX) { + vec3 glow = gather_glow(source_glow, uv_interp) * params.luminance_multiplier; + if (params.glow_map_strength > 0.001) { + glow = mix(glow, texture(glow_map, uv_interp).rgb * glow, params.glow_map_strength); + } + color.rgb = mix(color.rgb, glow, params.glow_intensity); + } +#endif + + color.rgb = apply_tonemapping(color.rgb, params.white); + + color.rgb = linear_to_srgb(color.rgb); // regular linear -> SRGB conversion + +#ifndef SUBPASS + // Glow + if (params.use_glow && params.glow_mode != GLOW_MODE_MIX) { + vec3 glow = gather_glow(source_glow, uv_interp) * params.glow_intensity * params.luminance_multiplier; + if (params.glow_map_strength > 0.001) { + glow = mix(glow, texture(glow_map, uv_interp).rgb * glow, params.glow_map_strength); + } + + // high dynamic range -> SRGB + glow = apply_tonemapping(glow, params.white); + glow = linear_to_srgb(glow); + + color.rgb = apply_glow(color.rgb, glow); + } +#endif + + // Additional effects + + if (params.use_bcs) { + color.rgb = apply_bcs(color.rgb, params.bcs); + } + + if (params.use_color_correction) { + color.rgb = apply_color_correction(color.rgb); + } + + if (params.use_debanding) { + // Debanding should be done at the end of tonemapping, but before writing to the LDR buffer. + // Otherwise, we're adding noise to an already-quantized image. + color.rgb += screen_space_dither(gl_FragCoord.xy); + } + + frag_color = color; +} diff --git a/servers/rendering/renderer_rd/shaders/effects/vrs.glsl b/servers/rendering/renderer_rd/shaders/effects/vrs.glsl new file mode 100644 index 0000000000..5ef83c0b44 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/vrs.glsl @@ -0,0 +1,72 @@ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#ifdef MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#endif //MULTIVIEW + +#ifdef MULTIVIEW +layout(location = 0) out vec3 uv_interp; +#else +layout(location = 0) out vec2 uv_interp; +#endif + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp.xy = base_arr[gl_VertexIndex]; +#ifdef MULTIVIEW + uv_interp.z = ViewIndex; +#endif + + gl_Position = vec4(uv_interp.xy * 2.0 - 1.0, 0.0, 1.0); +} + +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#ifdef MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#endif //MULTIVIEW + +#ifdef MULTIVIEW +layout(location = 0) in vec3 uv_interp; +layout(set = 0, binding = 0) uniform sampler2DArray source_color; +#else /* MULTIVIEW */ +layout(location = 0) in vec2 uv_interp; +layout(set = 0, binding = 0) uniform sampler2D source_color; +#endif /* MULTIVIEW */ + +layout(location = 0) out uint frag_color; + +void main() { +#ifdef MULTIVIEW + vec3 uv = uv_interp; +#else + vec2 uv = uv_interp; +#endif + +#ifdef MULTIVIEW + vec4 color = textureLod(source_color, uv, 0.0); +#else /* MULTIVIEW */ + vec4 color = textureLod(source_color, uv, 0.0); +#endif /* MULTIVIEW */ + + // See if we can change the sampler to one that returns int... + frag_color = uint(color.r * 256.0); +} diff --git a/servers/rendering/renderer_rd/shaders/environment/SCsub b/servers/rendering/renderer_rd/shaders/environment/SCsub new file mode 100644 index 0000000000..f06a2d86e2 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/environment/SCsub @@ -0,0 +1,17 @@ +#!/usr/bin/env python + +Import("env") + +if "RD_GLSL" in env["BUILDERS"]: + # find all include files + gl_include_files = [str(f) for f in Glob("*_inc.glsl")] + [str(f) for f in Glob("../*_inc.glsl")] + + # find all shader code(all glsl files excluding our include files) + glsl_files = [str(f) for f in Glob("*.glsl") if str(f) not in gl_include_files] + + # make sure we recompile shaders if include files change + env.Depends([f + ".gen.h" for f in glsl_files], gl_include_files + ["#glsl_builders.py"]) + + # compile shaders + for glsl_file in glsl_files: + env.RD_GLSL(glsl_file) diff --git a/servers/rendering/renderer_rd/shaders/environment/gi.glsl b/servers/rendering/renderer_rd/shaders/environment/gi.glsl new file mode 100644 index 0000000000..ab927df678 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/environment/gi.glsl @@ -0,0 +1,769 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +#define M_PI 3.141592 + +/* Specialization Constants (Toggles) */ + +layout(constant_id = 0) const bool sc_half_res = false; +layout(constant_id = 1) const bool sc_use_full_projection_matrix = false; +layout(constant_id = 2) const bool sc_use_vrs = false; + +#define SDFGI_MAX_CASCADES 8 + +//set 0 for SDFGI and render buffers + +layout(set = 0, binding = 1) uniform texture3D sdf_cascades[SDFGI_MAX_CASCADES]; +layout(set = 0, binding = 2) uniform texture3D light_cascades[SDFGI_MAX_CASCADES]; +layout(set = 0, binding = 3) uniform texture3D aniso0_cascades[SDFGI_MAX_CASCADES]; +layout(set = 0, binding = 4) uniform texture3D aniso1_cascades[SDFGI_MAX_CASCADES]; +layout(set = 0, binding = 5) uniform texture3D occlusion_texture; + +layout(set = 0, binding = 6) uniform sampler linear_sampler; +layout(set = 0, binding = 7) uniform sampler linear_sampler_with_mipmaps; + +struct ProbeCascadeData { + vec3 position; + float to_probe; + ivec3 probe_world_offset; + float to_cell; // 1/bounds * grid_size + vec3 pad; + float exposure_normalization; +}; + +layout(rgba16f, set = 0, binding = 9) uniform restrict writeonly image2D ambient_buffer; +layout(rgba16f, set = 0, binding = 10) uniform restrict writeonly image2D reflection_buffer; + +layout(set = 0, binding = 11) uniform texture2DArray lightprobe_texture; + +layout(set = 0, binding = 12) uniform texture2D depth_buffer; +layout(set = 0, binding = 13) uniform texture2D normal_roughness_buffer; +layout(set = 0, binding = 14) uniform utexture2D voxel_gi_buffer; + +layout(set = 0, binding = 15, std140) uniform SDFGI { + vec3 grid_size; + uint max_cascades; + + bool use_occlusion; + int probe_axis_size; + float probe_to_uvw; + float normal_bias; + + vec3 lightprobe_tex_pixel_size; + float energy; + + vec3 lightprobe_uv_offset; + float y_mult; + + vec3 occlusion_clamp; + uint pad3; + + vec3 occlusion_renormalize; + uint pad4; + + vec3 cascade_probe_size; + uint pad5; + + ProbeCascadeData cascades[SDFGI_MAX_CASCADES]; +} +sdfgi; + +#define MAX_VOXEL_GI_INSTANCES 8 + +struct VoxelGIData { + mat4 xform; // 64 - 64 + + vec3 bounds; // 12 - 76 + float dynamic_range; // 4 - 80 + + float bias; // 4 - 84 + float normal_bias; // 4 - 88 + bool blend_ambient; // 4 - 92 + uint mipmaps; // 4 - 96 + + vec3 pad; // 12 - 108 + float exposure_normalization; // 4 - 112 +}; + +layout(set = 0, binding = 16, std140) uniform VoxelGIs { + VoxelGIData data[MAX_VOXEL_GI_INSTANCES]; +} +voxel_gi_instances; + +layout(set = 0, binding = 17) uniform texture3D voxel_gi_textures[MAX_VOXEL_GI_INSTANCES]; + +layout(set = 0, binding = 18, std140) uniform SceneData { + mat4x4 inv_projection[2]; + mat4x4 cam_transform; + vec4 eye_offset[2]; + + ivec2 screen_size; + float pad1; + float pad2; +} +scene_data; + +layout(r8ui, set = 0, binding = 19) uniform restrict readonly uimage2D vrs_buffer; + +layout(push_constant, std430) uniform Params { + uint max_voxel_gi_instances; + bool high_quality_vct; + bool orthogonal; + uint view_index; + + vec4 proj_info; + + float z_near; + float z_far; + float pad2; + float pad3; +} +params; + +vec2 octahedron_wrap(vec2 v) { + vec2 signVal; + signVal.x = v.x >= 0.0 ? 1.0 : -1.0; + signVal.y = v.y >= 0.0 ? 1.0 : -1.0; + return (1.0 - abs(v.yx)) * signVal; +} + +vec2 octahedron_encode(vec3 n) { + // https://twitter.com/Stubbesaurus/status/937994790553227264 + n /= (abs(n.x) + abs(n.y) + abs(n.z)); + n.xy = n.z >= 0.0 ? n.xy : octahedron_wrap(n.xy); + n.xy = n.xy * 0.5 + 0.5; + return n.xy; +} + +vec4 blend_color(vec4 src, vec4 dst) { + vec4 res; + float sa = 1.0 - src.a; + res.a = dst.a * sa + src.a; + if (res.a == 0.0) { + res.rgb = vec3(0); + } else { + res.rgb = (dst.rgb * dst.a * sa + src.rgb * src.a) / res.a; + } + return res; +} + +vec3 reconstruct_position(ivec2 screen_pos) { + if (sc_use_full_projection_matrix) { + vec4 pos; + pos.xy = (2.0 * vec2(screen_pos) / vec2(scene_data.screen_size)) - 1.0; + pos.z = texelFetch(sampler2D(depth_buffer, linear_sampler), screen_pos, 0).r * 2.0 - 1.0; + pos.w = 1.0; + + pos = scene_data.inv_projection[params.view_index] * pos; + + return pos.xyz / pos.w; + } else { + vec3 pos; + pos.z = texelFetch(sampler2D(depth_buffer, linear_sampler), screen_pos, 0).r; + + pos.z = pos.z * 2.0 - 1.0; + if (params.orthogonal) { + pos.z = ((pos.z + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0; + } else { + pos.z = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - pos.z * (params.z_far - params.z_near)); + } + pos.z = -pos.z; + + pos.xy = vec2(screen_pos) * params.proj_info.xy + params.proj_info.zw; + if (!params.orthogonal) { + pos.xy *= pos.z; + } + + return pos; + } +} + +void sdfvoxel_gi_process(uint cascade, vec3 cascade_pos, vec3 cam_pos, vec3 cam_normal, vec3 cam_specular_normal, float roughness, out vec3 diffuse_light, out vec3 specular_light) { + cascade_pos += cam_normal * sdfgi.normal_bias; + + vec3 base_pos = floor(cascade_pos); + //cascade_pos += mix(vec3(0.0),vec3(0.01),lessThan(abs(cascade_pos-base_pos),vec3(0.01))) * cam_normal; + ivec3 probe_base_pos = ivec3(base_pos); + + vec4 diffuse_accum = vec4(0.0); + vec3 specular_accum; + + ivec3 tex_pos = ivec3(probe_base_pos.xy, int(cascade)); + tex_pos.x += probe_base_pos.z * sdfgi.probe_axis_size; + tex_pos.xy = tex_pos.xy * (SDFGI_OCT_SIZE + 2) + ivec2(1); + + vec3 diffuse_posf = (vec3(tex_pos) + vec3(octahedron_encode(cam_normal) * float(SDFGI_OCT_SIZE), 0.0)) * sdfgi.lightprobe_tex_pixel_size; + + vec3 specular_posf = (vec3(tex_pos) + vec3(octahedron_encode(cam_specular_normal) * float(SDFGI_OCT_SIZE), 0.0)) * sdfgi.lightprobe_tex_pixel_size; + + specular_accum = vec3(0.0); + + vec4 light_accum = vec4(0.0); + float weight_accum = 0.0; + + for (uint j = 0; j < 8; j++) { + ivec3 offset = (ivec3(j) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1); + ivec3 probe_posi = probe_base_pos; + probe_posi += offset; + + // Compute weight + + vec3 probe_pos = vec3(probe_posi); + vec3 probe_to_pos = cascade_pos - probe_pos; + vec3 probe_dir = normalize(-probe_to_pos); + + vec3 trilinear = vec3(1.0) - abs(probe_to_pos); + float weight = trilinear.x * trilinear.y * trilinear.z * max(0.005, dot(cam_normal, probe_dir)); + + // Compute lightprobe occlusion + + if (sdfgi.use_occlusion) { + ivec3 occ_indexv = abs((sdfgi.cascades[cascade].probe_world_offset + probe_posi) & ivec3(1, 1, 1)) * ivec3(1, 2, 4); + vec4 occ_mask = mix(vec4(0.0), vec4(1.0), equal(ivec4(occ_indexv.x | occ_indexv.y), ivec4(0, 1, 2, 3))); + + vec3 occ_pos = clamp(cascade_pos, probe_pos - sdfgi.occlusion_clamp, probe_pos + sdfgi.occlusion_clamp) * sdfgi.probe_to_uvw; + occ_pos.z += float(cascade); + if (occ_indexv.z != 0) { //z bit is on, means index is >=4, so make it switch to the other half of textures + occ_pos.x += 1.0; + } + + occ_pos *= sdfgi.occlusion_renormalize; + float occlusion = dot(textureLod(sampler3D(occlusion_texture, linear_sampler), occ_pos, 0.0), occ_mask); + + weight *= max(occlusion, 0.01); + } + + // Compute lightprobe texture position + + vec3 diffuse; + vec3 pos_uvw = diffuse_posf; + pos_uvw.xy += vec2(offset.xy) * sdfgi.lightprobe_uv_offset.xy; + pos_uvw.x += float(offset.z) * sdfgi.lightprobe_uv_offset.z; + diffuse = textureLod(sampler2DArray(lightprobe_texture, linear_sampler), pos_uvw, 0.0).rgb; + + diffuse_accum += vec4(diffuse * weight * sdfgi.cascades[cascade].exposure_normalization, weight); + + { + vec3 specular = vec3(0.0); + vec3 pos_uvw = specular_posf; + pos_uvw.xy += vec2(offset.xy) * sdfgi.lightprobe_uv_offset.xy; + pos_uvw.x += float(offset.z) * sdfgi.lightprobe_uv_offset.z; + if (roughness < 0.99) { + specular = textureLod(sampler2DArray(lightprobe_texture, linear_sampler), pos_uvw + vec3(0, 0, float(sdfgi.max_cascades)), 0.0).rgb; + } + if (roughness > 0.2) { + specular = mix(specular, textureLod(sampler2DArray(lightprobe_texture, linear_sampler), pos_uvw, 0.0).rgb, (roughness - 0.2) * 1.25); + } + + specular_accum += specular * weight * sdfgi.cascades[cascade].exposure_normalization; + } + } + + if (diffuse_accum.a > 0.0) { + diffuse_accum.rgb /= diffuse_accum.a; + } + + diffuse_light = diffuse_accum.rgb; + + if (diffuse_accum.a > 0.0) { + specular_accum /= diffuse_accum.a; + } + + specular_light = specular_accum; +} + +void sdfgi_process(vec3 vertex, vec3 normal, vec3 reflection, float roughness, out vec4 ambient_light, out vec4 reflection_light) { + //make vertex orientation the world one, but still align to camera + vertex.y *= sdfgi.y_mult; + normal.y *= sdfgi.y_mult; + reflection.y *= sdfgi.y_mult; + + //renormalize + normal = normalize(normal); + reflection = normalize(reflection); + + vec3 cam_pos = vertex; + vec3 cam_normal = normal; + + vec4 light_accum = vec4(0.0); + float weight_accum = 0.0; + + vec4 light_blend_accum = vec4(0.0); + float weight_blend_accum = 0.0; + + float blend = -1.0; + + // helper constants, compute once + + uint cascade = 0xFFFFFFFF; + vec3 cascade_pos; + vec3 cascade_normal; + + for (uint i = 0; i < sdfgi.max_cascades; i++) { + cascade_pos = (cam_pos - sdfgi.cascades[i].position) * sdfgi.cascades[i].to_probe; + + if (any(lessThan(cascade_pos, vec3(0.0))) || any(greaterThanEqual(cascade_pos, sdfgi.cascade_probe_size))) { + continue; //skip cascade + } + + cascade = i; + break; + } + + if (cascade < SDFGI_MAX_CASCADES) { + ambient_light = vec4(0, 0, 0, 1); + reflection_light = vec4(0, 0, 0, 1); + + float blend; + vec3 diffuse, specular; + sdfvoxel_gi_process(cascade, cascade_pos, cam_pos, cam_normal, reflection, roughness, diffuse, specular); + + { + //process blend + float blend_from = (float(sdfgi.probe_axis_size - 1) / 2.0) - 2.5; + float blend_to = blend_from + 2.0; + + vec3 inner_pos = cam_pos * sdfgi.cascades[cascade].to_probe; + + float len = length(inner_pos); + + inner_pos = abs(normalize(inner_pos)); + len *= max(inner_pos.x, max(inner_pos.y, inner_pos.z)); + + if (len >= blend_from) { + blend = smoothstep(blend_from, blend_to, len); + } else { + blend = 0.0; + } + } + + if (blend > 0.0) { + //blend + if (cascade == sdfgi.max_cascades - 1) { + ambient_light.a = 1.0 - blend; + reflection_light.a = 1.0 - blend; + + } else { + vec3 diffuse2, specular2; + cascade_pos = (cam_pos - sdfgi.cascades[cascade + 1].position) * sdfgi.cascades[cascade + 1].to_probe; + sdfvoxel_gi_process(cascade + 1, cascade_pos, cam_pos, cam_normal, reflection, roughness, diffuse2, specular2); + diffuse = mix(diffuse, diffuse2, blend); + specular = mix(specular, specular2, blend); + } + } + + ambient_light.rgb = diffuse; + + if (roughness < 0.2) { + vec3 pos_to_uvw = 1.0 / sdfgi.grid_size; + vec4 light_accum = vec4(0.0); + + float blend_size = (sdfgi.grid_size.x / float(sdfgi.probe_axis_size - 1)) * 0.5; + + float radius_sizes[SDFGI_MAX_CASCADES]; + cascade = 0xFFFF; + + float base_distance = length(cam_pos); + for (uint i = 0; i < sdfgi.max_cascades; i++) { + radius_sizes[i] = (1.0 / sdfgi.cascades[i].to_cell) * (sdfgi.grid_size.x * 0.5 - blend_size); + if (cascade == 0xFFFF && base_distance < radius_sizes[i]) { + cascade = i; + } + } + + cascade = min(cascade, sdfgi.max_cascades - 1); + + float max_distance = radius_sizes[sdfgi.max_cascades - 1]; + vec3 ray_pos = cam_pos; + vec3 ray_dir = reflection; + + { + float prev_radius = cascade > 0 ? radius_sizes[cascade - 1] : 0.0; + float base_blend = (base_distance - prev_radius) / (radius_sizes[cascade] - prev_radius); + float bias = (1.0 + base_blend) * 1.1; + vec3 abs_ray_dir = abs(ray_dir); + //ray_pos += ray_dir * (bias / sdfgi.cascades[cascade].to_cell); //bias to avoid self occlusion + ray_pos += (ray_dir * 1.0 / max(abs_ray_dir.x, max(abs_ray_dir.y, abs_ray_dir.z)) + cam_normal * 1.4) * bias / sdfgi.cascades[cascade].to_cell; + } + float softness = 0.2 + min(1.0, roughness * 5.0) * 4.0; //approximation to roughness so it does not seem like a hard fade + uint i = 0; + bool found = false; + while (true) { + if (length(ray_pos) >= max_distance || light_accum.a > 0.99) { + break; + } + if (!found && i >= cascade && length(ray_pos) < radius_sizes[i]) { + uint next_i = min(i + 1, sdfgi.max_cascades - 1); + cascade = max(i, cascade); //never go down + + vec3 pos = ray_pos - sdfgi.cascades[i].position; + pos *= sdfgi.cascades[i].to_cell * pos_to_uvw; + + float fdistance = textureLod(sampler3D(sdf_cascades[i], linear_sampler), pos, 0.0).r * 255.0 - 1.1; + + vec4 hit_light = vec4(0.0); + if (fdistance < softness) { + hit_light.rgb = textureLod(sampler3D(light_cascades[i], linear_sampler), pos, 0.0).rgb; + hit_light.rgb *= 0.5; //approximation given value read is actually meant for anisotropy + hit_light.a = clamp(1.0 - (fdistance / softness), 0.0, 1.0); + hit_light.rgb *= hit_light.a; + } + + fdistance /= sdfgi.cascades[i].to_cell; + + if (i < (sdfgi.max_cascades - 1)) { + pos = ray_pos - sdfgi.cascades[next_i].position; + pos *= sdfgi.cascades[next_i].to_cell * pos_to_uvw; + + float fdistance2 = textureLod(sampler3D(sdf_cascades[next_i], linear_sampler), pos, 0.0).r * 255.0 - 1.1; + + vec4 hit_light2 = vec4(0.0); + if (fdistance2 < softness) { + hit_light2.rgb = textureLod(sampler3D(light_cascades[next_i], linear_sampler), pos, 0.0).rgb; + hit_light2.rgb *= 0.5; //approximation given value read is actually meant for anisotropy + hit_light2.a = clamp(1.0 - (fdistance2 / softness), 0.0, 1.0); + hit_light2.rgb *= hit_light2.a; + } + + float prev_radius = i == 0 ? 0.0 : radius_sizes[max(0, i - 1)]; + float blend = clamp((length(ray_pos) - prev_radius) / (radius_sizes[i] - prev_radius), 0.0, 1.0); + + fdistance2 /= sdfgi.cascades[next_i].to_cell; + + hit_light = mix(hit_light, hit_light2, blend); + fdistance = mix(fdistance, fdistance2, blend); + } + + light_accum += hit_light; + ray_pos += ray_dir * fdistance; + found = true; + } + i++; + if (i == sdfgi.max_cascades) { + i = 0; + found = false; + } + } + + vec3 light = light_accum.rgb / max(light_accum.a, 0.00001); + float alpha = min(1.0, light_accum.a); + + float b = min(1.0, roughness * 5.0); + + float sa = 1.0 - b; + + reflection_light.a = alpha * sa + b; + if (reflection_light.a == 0) { + specular = vec3(0.0); + } else { + specular = (light * alpha * sa + specular * b) / reflection_light.a; + } + } + + reflection_light.rgb = specular; + + ambient_light.rgb *= sdfgi.energy; + reflection_light.rgb *= sdfgi.energy; + } else { + ambient_light = vec4(0); + reflection_light = vec4(0); + } +} + +//standard voxel cone trace +vec4 voxel_cone_trace(texture3D probe, vec3 cell_size, vec3 pos, vec3 direction, float tan_half_angle, float max_distance, float p_bias) { + float dist = p_bias; + vec4 color = vec4(0.0); + + while (dist < max_distance && color.a < 0.95) { + float diameter = max(1.0, 2.0 * tan_half_angle * dist); + vec3 uvw_pos = (pos + dist * direction) * cell_size; + float half_diameter = diameter * 0.5; + //check if outside, then break + if (any(greaterThan(abs(uvw_pos - 0.5), vec3(0.5f + half_diameter * cell_size)))) { + break; + } + vec4 scolor = textureLod(sampler3D(probe, linear_sampler_with_mipmaps), uvw_pos, log2(diameter)); + float a = (1.0 - color.a); + color += a * scolor; + dist += half_diameter; + } + + return color; +} + +vec4 voxel_cone_trace_45_degrees(texture3D probe, vec3 cell_size, vec3 pos, vec3 direction, float max_distance, float p_bias) { + float dist = p_bias; + vec4 color = vec4(0.0); + float radius = max(0.5, dist); + float lod_level = log2(radius * 2.0); + + while (dist < max_distance && color.a < 0.95) { + vec3 uvw_pos = (pos + dist * direction) * cell_size; + + //check if outside, then break + if (any(greaterThan(abs(uvw_pos - 0.5), vec3(0.5f + radius * cell_size)))) { + break; + } + vec4 scolor = textureLod(sampler3D(probe, linear_sampler_with_mipmaps), uvw_pos, lod_level); + lod_level += 1.0; + + float a = (1.0 - color.a); + scolor *= a; + color += scolor; + dist += radius; + radius = max(0.5, dist); + } + return color; +} + +void voxel_gi_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3 normal_xform, float roughness, inout vec4 out_spec, inout vec4 out_diff, inout float out_blend) { + position = (voxel_gi_instances.data[index].xform * vec4(position, 1.0)).xyz; + ref_vec = normalize((voxel_gi_instances.data[index].xform * vec4(ref_vec, 0.0)).xyz); + normal = normalize((voxel_gi_instances.data[index].xform * vec4(normal, 0.0)).xyz); + + position += normal * voxel_gi_instances.data[index].normal_bias; + + //this causes corrupted pixels, i have no idea why.. + if (any(bvec2(any(lessThan(position, vec3(0.0))), any(greaterThan(position, voxel_gi_instances.data[index].bounds))))) { + return; + } + + mat3 dir_xform = mat3(voxel_gi_instances.data[index].xform) * normal_xform; + + vec3 blendv = abs(position / voxel_gi_instances.data[index].bounds * 2.0 - 1.0); + float blend = clamp(1.0 - max(blendv.x, max(blendv.y, blendv.z)), 0.0, 1.0); + //float blend=1.0; + + float max_distance = length(voxel_gi_instances.data[index].bounds); + vec3 cell_size = 1.0 / voxel_gi_instances.data[index].bounds; + + //irradiance + + vec4 light = vec4(0.0); + + if (params.high_quality_vct) { + const uint cone_dir_count = 6; + vec3 cone_dirs[cone_dir_count] = vec3[]( + vec3(0.0, 0.0, 1.0), + vec3(0.866025, 0.0, 0.5), + vec3(0.267617, 0.823639, 0.5), + vec3(-0.700629, 0.509037, 0.5), + vec3(-0.700629, -0.509037, 0.5), + vec3(0.267617, -0.823639, 0.5)); + + float cone_weights[cone_dir_count] = float[](0.25, 0.15, 0.15, 0.15, 0.15, 0.15); + float cone_angle_tan = 0.577; + + for (uint i = 0; i < cone_dir_count; i++) { + vec3 dir = normalize(dir_xform * cone_dirs[i]); + light += cone_weights[i] * voxel_cone_trace(voxel_gi_textures[index], cell_size, position, dir, cone_angle_tan, max_distance, voxel_gi_instances.data[index].bias); + } + } else { + const uint cone_dir_count = 4; + vec3 cone_dirs[cone_dir_count] = vec3[]( + vec3(0.707107, 0.0, 0.707107), + vec3(0.0, 0.707107, 0.707107), + vec3(-0.707107, 0.0, 0.707107), + vec3(0.0, -0.707107, 0.707107)); + + float cone_weights[cone_dir_count] = float[](0.25, 0.25, 0.25, 0.25); + for (int i = 0; i < cone_dir_count; i++) { + vec3 dir = normalize(dir_xform * cone_dirs[i]); + light += cone_weights[i] * voxel_cone_trace_45_degrees(voxel_gi_textures[index], cell_size, position, dir, max_distance, voxel_gi_instances.data[index].bias); + } + } + + light.rgb *= voxel_gi_instances.data[index].dynamic_range * voxel_gi_instances.data[index].exposure_normalization; + if (!voxel_gi_instances.data[index].blend_ambient) { + light.a = 1.0; + } + + out_diff += light * blend; + + //radiance + vec4 irr_light = voxel_cone_trace(voxel_gi_textures[index], cell_size, position, ref_vec, tan(roughness * 0.5 * M_PI * 0.99), max_distance, voxel_gi_instances.data[index].bias); + irr_light.rgb *= voxel_gi_instances.data[index].dynamic_range * voxel_gi_instances.data[index].exposure_normalization; + if (!voxel_gi_instances.data[index].blend_ambient) { + irr_light.a = 1.0; + } + + out_spec += irr_light * blend; + + out_blend += blend; +} + +vec4 fetch_normal_and_roughness(ivec2 pos) { + vec4 normal_roughness = texelFetch(sampler2D(normal_roughness_buffer, linear_sampler), pos, 0); + normal_roughness.xyz = normalize(normal_roughness.xyz * 2.0 - 1.0); + return normal_roughness; +} + +void process_gi(ivec2 pos, vec3 vertex, inout vec4 ambient_light, inout vec4 reflection_light) { + vec4 normal_roughness = fetch_normal_and_roughness(pos); + + vec3 normal = normal_roughness.xyz; + + if (normal.length() > 0.5) { + //valid normal, can do GI + float roughness = normal_roughness.w; + vec3 view = -normalize(mat3(scene_data.cam_transform) * (vertex - scene_data.eye_offset[gl_GlobalInvocationID.z].xyz)); + vertex = mat3(scene_data.cam_transform) * vertex; + normal = normalize(mat3(scene_data.cam_transform) * normal); + vec3 reflection = normalize(reflect(-view, normal)); + +#ifdef USE_SDFGI + sdfgi_process(vertex, normal, reflection, roughness, ambient_light, reflection_light); +#endif + +#ifdef USE_VOXEL_GI_INSTANCES + { + uvec2 voxel_gi_tex = texelFetch(usampler2D(voxel_gi_buffer, linear_sampler), pos, 0).rg; + roughness *= roughness; + //find arbitrary tangent and bitangent, then build a matrix + vec3 v0 = abs(normal.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 1.0, 0.0); + vec3 tangent = normalize(cross(v0, normal)); + vec3 bitangent = normalize(cross(tangent, normal)); + mat3 normal_mat = mat3(tangent, bitangent, normal); + + vec4 amb_accum = vec4(0.0); + vec4 spec_accum = vec4(0.0); + float blend_accum = 0.0; + + for (uint i = 0; i < params.max_voxel_gi_instances; i++) { + if (any(equal(uvec2(i), voxel_gi_tex))) { + voxel_gi_compute(i, vertex, normal, reflection, normal_mat, roughness, spec_accum, amb_accum, blend_accum); + } + } + if (blend_accum > 0.0) { + amb_accum /= blend_accum; + spec_accum /= blend_accum; + } + +#ifdef USE_SDFGI + reflection_light = blend_color(spec_accum, reflection_light); + ambient_light = blend_color(amb_accum, ambient_light); +#else + reflection_light = spec_accum; + ambient_light = amb_accum; +#endif + } +#endif + } +} + +void main() { + ivec2 pos = ivec2(gl_GlobalInvocationID.xy); + + uint vrs_x, vrs_y; + if (sc_use_vrs) { + ivec2 vrs_pos; + + // Currently we use a 16x16 texel, possibly some day make this configurable. + if (sc_half_res) { + vrs_pos = pos >> 3; + } else { + vrs_pos = pos >> 4; + } + + uint vrs_texel = imageLoad(vrs_buffer, vrs_pos).r; + // note, valid values for vrs_x and vrs_y are 1, 2 and 4. + vrs_x = 1 << ((vrs_texel >> 2) & 3); + vrs_y = 1 << (vrs_texel & 3); + + if (mod(pos.x, vrs_x) != 0) { + return; + } + + if (mod(pos.y, vrs_y) != 0) { + return; + } + } + + if (sc_half_res) { + pos <<= 1; + } + + if (any(greaterThanEqual(pos, scene_data.screen_size))) { //too large, do nothing + return; + } + + vec4 ambient_light = vec4(0.0); + vec4 reflection_light = vec4(0.0); + + vec3 vertex = reconstruct_position(pos); + vertex.y = -vertex.y; + + process_gi(pos, vertex, ambient_light, reflection_light); + + if (sc_half_res) { + pos >>= 1; + } + + imageStore(ambient_buffer, pos, ambient_light); + imageStore(reflection_buffer, pos, reflection_light); + + if (sc_use_vrs) { + if (vrs_x > 1) { + imageStore(ambient_buffer, pos + ivec2(1, 0), ambient_light); + imageStore(reflection_buffer, pos + ivec2(1, 0), reflection_light); + } + + if (vrs_x > 2) { + imageStore(ambient_buffer, pos + ivec2(2, 0), ambient_light); + imageStore(reflection_buffer, pos + ivec2(2, 0), reflection_light); + + imageStore(ambient_buffer, pos + ivec2(3, 0), ambient_light); + imageStore(reflection_buffer, pos + ivec2(3, 0), reflection_light); + } + + if (vrs_y > 1) { + imageStore(ambient_buffer, pos + ivec2(0, 1), ambient_light); + imageStore(reflection_buffer, pos + ivec2(0, 1), reflection_light); + } + + if (vrs_y > 1 && vrs_x > 1) { + imageStore(ambient_buffer, pos + ivec2(1, 1), ambient_light); + imageStore(reflection_buffer, pos + ivec2(1, 1), reflection_light); + } + + if (vrs_y > 1 && vrs_x > 2) { + imageStore(ambient_buffer, pos + ivec2(2, 1), ambient_light); + imageStore(reflection_buffer, pos + ivec2(2, 1), reflection_light); + + imageStore(ambient_buffer, pos + ivec2(3, 1), ambient_light); + imageStore(reflection_buffer, pos + ivec2(3, 1), reflection_light); + } + + if (vrs_y > 2) { + imageStore(ambient_buffer, pos + ivec2(0, 2), ambient_light); + imageStore(reflection_buffer, pos + ivec2(0, 2), reflection_light); + imageStore(ambient_buffer, pos + ivec2(0, 3), ambient_light); + imageStore(reflection_buffer, pos + ivec2(0, 3), reflection_light); + } + + if (vrs_y > 2 && vrs_x > 1) { + imageStore(ambient_buffer, pos + ivec2(1, 2), ambient_light); + imageStore(reflection_buffer, pos + ivec2(1, 2), reflection_light); + imageStore(ambient_buffer, pos + ivec2(1, 3), ambient_light); + imageStore(reflection_buffer, pos + ivec2(1, 3), reflection_light); + } + + if (vrs_y > 2 && vrs_x > 2) { + imageStore(ambient_buffer, pos + ivec2(2, 2), ambient_light); + imageStore(reflection_buffer, pos + ivec2(2, 2), reflection_light); + imageStore(ambient_buffer, pos + ivec2(2, 3), ambient_light); + imageStore(reflection_buffer, pos + ivec2(2, 3), reflection_light); + + imageStore(ambient_buffer, pos + ivec2(3, 2), ambient_light); + imageStore(reflection_buffer, pos + ivec2(3, 2), reflection_light); + imageStore(ambient_buffer, pos + ivec2(3, 3), ambient_light); + imageStore(reflection_buffer, pos + ivec2(3, 3), reflection_light); + } + } +} diff --git a/servers/rendering/renderer_rd/shaders/environment/sdfgi_debug.glsl b/servers/rendering/renderer_rd/shaders/environment/sdfgi_debug.glsl new file mode 100644 index 0000000000..177dab16c7 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/environment/sdfgi_debug.glsl @@ -0,0 +1,185 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +#define MAX_CASCADES 8 + +layout(set = 0, binding = 1) uniform texture3D sdf_cascades[MAX_CASCADES]; +layout(set = 0, binding = 2) uniform texture3D light_cascades[MAX_CASCADES]; +layout(set = 0, binding = 3) uniform texture3D aniso0_cascades[MAX_CASCADES]; +layout(set = 0, binding = 4) uniform texture3D aniso1_cascades[MAX_CASCADES]; +layout(set = 0, binding = 5) uniform texture3D occlusion_texture; + +layout(set = 0, binding = 8) uniform sampler linear_sampler; + +struct CascadeData { + vec3 offset; //offset of (0,0,0) in world coordinates + float to_cell; // 1/bounds * grid_size + ivec3 probe_world_offset; + uint pad; + vec4 pad2; +}; + +layout(set = 0, binding = 9, std140) uniform Cascades { + CascadeData data[MAX_CASCADES]; +} +cascades; + +layout(rgba16f, set = 0, binding = 10) uniform restrict writeonly image2D screen_buffer; + +layout(set = 0, binding = 11) uniform texture2DArray lightprobe_texture; + +layout(push_constant, std430) uniform Params { + vec3 grid_size; + uint max_cascades; + + ivec2 screen_size; + float y_mult; + + float z_near; + + mat3x4 inv_projection; + // We pack these more tightly than mat3 and vec3, which will require some reconstruction trickery. + float cam_basis[3][3]; + float cam_origin[3]; +} +params; + +vec3 linear_to_srgb(vec3 color) { + //if going to srgb, clamp from 0 to 1. + color = clamp(color, vec3(0.0), vec3(1.0)); + const vec3 a = vec3(0.055f); + return mix((vec3(1.0f) + a) * pow(color.rgb, vec3(1.0f / 2.4f)) - a, 12.92f * color.rgb, lessThan(color.rgb, vec3(0.0031308f))); +} + +vec2 octahedron_wrap(vec2 v) { + vec2 signVal; + signVal.x = v.x >= 0.0 ? 1.0 : -1.0; + signVal.y = v.y >= 0.0 ? 1.0 : -1.0; + return (1.0 - abs(v.yx)) * signVal; +} + +vec2 octahedron_encode(vec3 n) { + // https://twitter.com/Stubbesaurus/status/937994790553227264 + n /= (abs(n.x) + abs(n.y) + abs(n.z)); + n.xy = n.z >= 0.0 ? n.xy : octahedron_wrap(n.xy); + n.xy = n.xy * 0.5 + 0.5; + return n.xy; +} + +void main() { + // Pixel being shaded + ivec2 screen_pos = ivec2(gl_GlobalInvocationID.xy); + if (any(greaterThanEqual(screen_pos, params.screen_size))) { //too large, do nothing + return; + } + + vec3 ray_pos; + vec3 ray_dir; + { + ray_pos = vec3(params.cam_origin[0], params.cam_origin[1], params.cam_origin[2]); + + ray_dir.xy = ((vec2(screen_pos) / vec2(params.screen_size)) * 2.0 - 1.0); + ray_dir.z = params.z_near; + + ray_dir = (vec4(ray_dir, 1.0) * mat4(params.inv_projection)).xyz; + + mat3 cam_basis; + { + vec3 c0 = vec3(params.cam_basis[0][0], params.cam_basis[0][1], params.cam_basis[0][2]); + vec3 c1 = vec3(params.cam_basis[1][0], params.cam_basis[1][1], params.cam_basis[1][2]); + vec3 c2 = vec3(params.cam_basis[2][0], params.cam_basis[2][1], params.cam_basis[2][2]); + cam_basis = mat3(c0, c1, c2); + } + ray_dir = normalize(cam_basis * ray_dir); + } + + ray_pos.y *= params.y_mult; + ray_dir.y *= params.y_mult; + ray_dir = normalize(ray_dir); + + vec3 pos_to_uvw = 1.0 / params.grid_size; + + vec3 light = vec3(0.0); + float blend = 0.0; + +#if 1 + // No interpolation + + vec3 inv_dir = 1.0 / ray_dir; + + float rough = 0.5; + bool hit = false; + + for (uint i = 0; i < params.max_cascades; i++) { + //convert to local bounds + vec3 pos = ray_pos - cascades.data[i].offset; + pos *= cascades.data[i].to_cell; + + // Should never happen for debug, since we start mostly at the bounds center, + // but add anyway. + //if (any(lessThan(pos,vec3(0.0))) || any(greaterThanEqual(pos,params.grid_size))) { + // continue; //already past bounds for this cascade, goto next + //} + + //find maximum advance distance (until reaching bounds) + vec3 t0 = -pos * inv_dir; + vec3 t1 = (params.grid_size - pos) * inv_dir; + vec3 tmax = max(t0, t1); + float max_advance = min(tmax.x, min(tmax.y, tmax.z)); + + float advance = 0.0; + vec3 uvw; + hit = false; + + while (advance < max_advance) { + //read how much to advance from SDF + uvw = (pos + ray_dir * advance) * pos_to_uvw; + + float distance = texture(sampler3D(sdf_cascades[i], linear_sampler), uvw).r * 255.0 - 1.7; + + if (distance < 0.001) { + //consider hit + hit = true; + break; + } + + advance += distance; + } + + if (!hit) { + pos += ray_dir * min(advance, max_advance); + pos /= cascades.data[i].to_cell; + pos += cascades.data[i].offset; + ray_pos = pos; + continue; + } + + //compute albedo, emission and normal at hit point + + const float EPSILON = 0.001; + vec3 hit_normal = normalize(vec3( + texture(sampler3D(sdf_cascades[i], linear_sampler), uvw + vec3(EPSILON, 0.0, 0.0)).r - texture(sampler3D(sdf_cascades[i], linear_sampler), uvw - vec3(EPSILON, 0.0, 0.0)).r, + texture(sampler3D(sdf_cascades[i], linear_sampler), uvw + vec3(0.0, EPSILON, 0.0)).r - texture(sampler3D(sdf_cascades[i], linear_sampler), uvw - vec3(0.0, EPSILON, 0.0)).r, + texture(sampler3D(sdf_cascades[i], linear_sampler), uvw + vec3(0.0, 0.0, EPSILON)).r - texture(sampler3D(sdf_cascades[i], linear_sampler), uvw - vec3(0.0, 0.0, EPSILON)).r)); + + vec3 hit_light = texture(sampler3D(light_cascades[i], linear_sampler), uvw).rgb; + vec4 aniso0 = texture(sampler3D(aniso0_cascades[i], linear_sampler), uvw); + vec3 hit_aniso0 = aniso0.rgb; + vec3 hit_aniso1 = vec3(aniso0.a, texture(sampler3D(aniso1_cascades[i], linear_sampler), uvw).rg); + + hit_light *= (dot(max(vec3(0.0), (hit_normal * hit_aniso0)), vec3(1.0)) + dot(max(vec3(0.0), (-hit_normal * hit_aniso1)), vec3(1.0))); + + light = hit_light; + + break; + } + +#endif + + imageStore(screen_buffer, screen_pos, vec4(linear_to_srgb(light), 1.0)); +} diff --git a/servers/rendering/renderer_rd/shaders/environment/sdfgi_debug_probes.glsl b/servers/rendering/renderer_rd/shaders/environment/sdfgi_debug_probes.glsl new file mode 100644 index 0000000000..a0ef169f03 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/environment/sdfgi_debug_probes.glsl @@ -0,0 +1,268 @@ +#[vertex] + +#version 450 + +#if defined(USE_MULTIVIEW) && defined(has_VK_KHR_multiview) +#extension GL_EXT_multiview : enable +#endif + +#ifdef USE_MULTIVIEW +#ifdef has_VK_KHR_multiview +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +// !BAS! This needs to become an input once we implement our fallback! +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#else // USE_MULTIVIEW +// Set to zero, not supported in non stereo +#define ViewIndex 0 +#endif //USE_MULTIVIEW + +#VERSION_DEFINES + +#define MAX_CASCADES 8 +#define MAX_VIEWS 2 + +layout(push_constant, std430) uniform Params { + uint band_power; + uint sections_in_band; + uint band_mask; + float section_arc; + + vec3 grid_size; + uint cascade; + + uint pad; + float y_mult; + uint probe_debug_index; + int probe_axis_size; +} +params; + +// https://in4k.untergrund.net/html_articles/hugi_27_-_coding_corner_polaris_sphere_tessellation_101.htm + +vec3 get_sphere_vertex(uint p_vertex_id) { + float x_angle = float(p_vertex_id & 1u) + (p_vertex_id >> params.band_power); + + float y_angle = + float((p_vertex_id & params.band_mask) >> 1) + ((p_vertex_id >> params.band_power) * params.sections_in_band); + + x_angle *= params.section_arc * 0.5f; // remember - 180AA x rot not 360 + y_angle *= -params.section_arc; + + vec3 point = vec3(sin(x_angle) * sin(y_angle), cos(x_angle), sin(x_angle) * cos(y_angle)); + + return point; +} + +#ifdef MODE_PROBES + +layout(location = 0) out vec3 normal_interp; +layout(location = 1) out flat uint probe_index; + +#endif + +#ifdef MODE_VISIBILITY + +layout(location = 0) out float visibility; + +#endif + +struct CascadeData { + vec3 offset; //offset of (0,0,0) in world coordinates + float to_cell; // 1/bounds * grid_size + ivec3 probe_world_offset; + uint pad; + vec4 pad2; +}; + +layout(set = 0, binding = 1, std140) uniform Cascades { + CascadeData data[MAX_CASCADES]; +} +cascades; + +layout(set = 0, binding = 4) uniform texture3D occlusion_texture; +layout(set = 0, binding = 3) uniform sampler linear_sampler; + +layout(set = 0, binding = 5, std140) uniform SceneData { + mat4 projection[MAX_VIEWS]; +} +scene_data; + +void main() { +#ifdef MODE_PROBES + probe_index = gl_InstanceIndex; + + normal_interp = get_sphere_vertex(gl_VertexIndex); + + vec3 vertex = normal_interp * 0.2; + + float probe_cell_size = float(params.grid_size / float(params.probe_axis_size - 1)) / cascades.data[params.cascade].to_cell; + + ivec3 probe_cell; + probe_cell.x = int(probe_index % params.probe_axis_size); + probe_cell.y = int(probe_index / (params.probe_axis_size * params.probe_axis_size)); + probe_cell.z = int((probe_index / params.probe_axis_size) % params.probe_axis_size); + + vertex += (cascades.data[params.cascade].offset + vec3(probe_cell) * probe_cell_size) / vec3(1.0, params.y_mult, 1.0); + + gl_Position = scene_data.projection[ViewIndex] * vec4(vertex, 1.0); +#endif + +#ifdef MODE_VISIBILITY + + int probe_index = int(params.probe_debug_index); + + vec3 vertex = get_sphere_vertex(gl_VertexIndex) * 0.01; + + float probe_cell_size = float(params.grid_size / float(params.probe_axis_size - 1)) / cascades.data[params.cascade].to_cell; + + ivec3 probe_cell; + probe_cell.x = int(probe_index % params.probe_axis_size); + probe_cell.y = int((probe_index % (params.probe_axis_size * params.probe_axis_size)) / params.probe_axis_size); + probe_cell.z = int(probe_index / (params.probe_axis_size * params.probe_axis_size)); + + vertex += (cascades.data[params.cascade].offset + vec3(probe_cell) * probe_cell_size) / vec3(1.0, params.y_mult, 1.0); + + int probe_voxels = int(params.grid_size.x) / int(params.probe_axis_size - 1); + int occluder_index = int(gl_InstanceIndex); + + int diameter = probe_voxels * 2; + ivec3 occluder_pos; + occluder_pos.x = int(occluder_index % diameter); + occluder_pos.y = int(occluder_index / (diameter * diameter)); + occluder_pos.z = int((occluder_index / diameter) % diameter); + + float cell_size = 1.0 / cascades.data[params.cascade].to_cell; + + ivec3 occluder_offset = occluder_pos - ivec3(diameter / 2); + vertex += ((vec3(occluder_offset) + vec3(0.5)) * cell_size) / vec3(1.0, params.y_mult, 1.0); + + ivec3 global_cell = probe_cell + cascades.data[params.cascade].probe_world_offset; + uint occlusion_layer = 0; + if ((global_cell.x & 1) != 0) { + occlusion_layer |= 1; + } + if ((global_cell.y & 1) != 0) { + occlusion_layer |= 2; + } + if ((global_cell.z & 1) != 0) { + occlusion_layer |= 4; + } + ivec3 tex_pos = probe_cell * probe_voxels + occluder_offset; + + const vec4 layer_axis[4] = vec4[]( + vec4(1, 0, 0, 0), + vec4(0, 1, 0, 0), + vec4(0, 0, 1, 0), + vec4(0, 0, 0, 1)); + + tex_pos.z += int(params.cascade) * int(params.grid_size); + if (occlusion_layer >= 4) { + tex_pos.x += int(params.grid_size.x); + occlusion_layer &= 3; + } + + visibility = dot(texelFetch(sampler3D(occlusion_texture, linear_sampler), tex_pos, 0), layer_axis[occlusion_layer]); + + gl_Position = scene_data.projection[ViewIndex] * vec4(vertex, 1.0); + +#endif +} + +#[fragment] + +#version 450 + +#if defined(USE_MULTIVIEW) && defined(has_VK_KHR_multiview) +#extension GL_EXT_multiview : enable +#endif + +#ifdef USE_MULTIVIEW +#ifdef has_VK_KHR_multiview +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +// !BAS! This needs to become an input once we implement our fallback! +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#else // USE_MULTIVIEW +// Set to zero, not supported in non stereo +#define ViewIndex 0 +#endif //USE_MULTIVIEW + +#VERSION_DEFINES + +#define MAX_VIEWS 2 + +layout(location = 0) out vec4 frag_color; + +layout(set = 0, binding = 2) uniform texture2DArray lightprobe_texture; +layout(set = 0, binding = 3) uniform sampler linear_sampler; + +layout(push_constant, std430) uniform Params { + uint band_power; + uint sections_in_band; + uint band_mask; + float section_arc; + + vec3 grid_size; + uint cascade; + + uint pad; + float y_mult; + uint probe_debug_index; + int probe_axis_size; +} +params; + +#ifdef MODE_PROBES + +layout(location = 0) in vec3 normal_interp; +layout(location = 1) in flat uint probe_index; + +#endif + +#ifdef MODE_VISIBILITY +layout(location = 0) in float visibility; +#endif + +vec2 octahedron_wrap(vec2 v) { + vec2 signVal; + signVal.x = v.x >= 0.0 ? 1.0 : -1.0; + signVal.y = v.y >= 0.0 ? 1.0 : -1.0; + return (1.0 - abs(v.yx)) * signVal; +} + +vec2 octahedron_encode(vec3 n) { + // https://twitter.com/Stubbesaurus/status/937994790553227264 + n /= (abs(n.x) + abs(n.y) + abs(n.z)); + n.xy = n.z >= 0.0 ? n.xy : octahedron_wrap(n.xy); + n.xy = n.xy * 0.5 + 0.5; + return n.xy; +} + +void main() { +#ifdef MODE_PROBES + + ivec3 tex_pos; + tex_pos.x = int(probe_index) % params.probe_axis_size; //x + tex_pos.y = int(probe_index) / (params.probe_axis_size * params.probe_axis_size); + tex_pos.x += params.probe_axis_size * ((int(probe_index) / params.probe_axis_size) % params.probe_axis_size); //z + tex_pos.z = int(params.cascade); + + vec3 tex_pos_ofs = vec3(octahedron_encode(normal_interp) * float(OCT_SIZE), 0.0); + vec3 tex_posf = vec3(vec2(tex_pos.xy * (OCT_SIZE + 2) + ivec2(1)), float(tex_pos.z)) + tex_pos_ofs; + + tex_posf.xy /= vec2(ivec2(params.probe_axis_size * params.probe_axis_size * (OCT_SIZE + 2), params.probe_axis_size * (OCT_SIZE + 2))); + + vec4 indirect_light = textureLod(sampler2DArray(lightprobe_texture, linear_sampler), tex_posf, 0.0); + + frag_color = indirect_light; + +#endif + +#ifdef MODE_VISIBILITY + + frag_color = vec4(vec3(1, visibility, visibility), 1.0); +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/environment/sdfgi_direct_light.glsl b/servers/rendering/renderer_rd/shaders/environment/sdfgi_direct_light.glsl new file mode 100644 index 0000000000..9f7449b8aa --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/environment/sdfgi_direct_light.glsl @@ -0,0 +1,507 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in; + +#define MAX_CASCADES 8 + +layout(set = 0, binding = 1) uniform texture3D sdf_cascades[MAX_CASCADES]; +layout(set = 0, binding = 2) uniform sampler linear_sampler; + +layout(set = 0, binding = 3, std430) restrict readonly buffer DispatchData { + uint x; + uint y; + uint z; + uint total_count; +} +dispatch_data; + +struct ProcessVoxel { + uint position; // xyz 7 bit packed, extra 11 bits for neighbors. + uint albedo; // rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neighbors. + uint light; // rgbe8985 encoded total saved light, extra 2 bits for neighbors. + uint light_aniso; // 55555 light anisotropy, extra 2 bits for neighbors. + //total neighbours: 26 +}; + +#ifdef MODE_PROCESS_STATIC +layout(set = 0, binding = 4, std430) restrict buffer ProcessVoxels { +#else +layout(set = 0, binding = 4, std430) restrict buffer readonly ProcessVoxels { +#endif + ProcessVoxel data[]; +} +process_voxels; + +layout(r32ui, set = 0, binding = 5) uniform restrict uimage3D dst_light; +layout(rgba8, set = 0, binding = 6) uniform restrict image3D dst_aniso0; +layout(rg8, set = 0, binding = 7) uniform restrict image3D dst_aniso1; + +struct CascadeData { + vec3 offset; //offset of (0,0,0) in world coordinates + float to_cell; // 1/bounds * grid_size + ivec3 probe_world_offset; + uint pad; + vec4 pad2; +}; + +layout(set = 0, binding = 8, std140) uniform Cascades { + CascadeData data[MAX_CASCADES]; +} +cascades; + +#define LIGHT_TYPE_DIRECTIONAL 0 +#define LIGHT_TYPE_OMNI 1 +#define LIGHT_TYPE_SPOT 2 + +struct Light { + vec3 color; + float energy; + + vec3 direction; + bool has_shadow; + + vec3 position; + float attenuation; + + uint type; + float cos_spot_angle; + float inv_spot_attenuation; + float radius; +}; + +layout(set = 0, binding = 9, std140) buffer restrict readonly Lights { + Light data[]; +} +lights; + +layout(set = 0, binding = 10) uniform texture2DArray lightprobe_texture; +layout(set = 0, binding = 11) uniform texture3D occlusion_texture; + +layout(push_constant, std430) uniform Params { + vec3 grid_size; + uint max_cascades; + + uint cascade; + uint light_count; + uint process_offset; + uint process_increment; + + int probe_axis_size; + float bounce_feedback; + float y_mult; + bool use_occlusion; +} +params; + +vec2 octahedron_wrap(vec2 v) { + vec2 signVal; + signVal.x = v.x >= 0.0 ? 1.0 : -1.0; + signVal.y = v.y >= 0.0 ? 1.0 : -1.0; + return (1.0 - abs(v.yx)) * signVal; +} + +vec2 octahedron_encode(vec3 n) { + // https://twitter.com/Stubbesaurus/status/937994790553227264 + n /= (abs(n.x) + abs(n.y) + abs(n.z)); + n.xy = n.z >= 0.0 ? n.xy : octahedron_wrap(n.xy); + n.xy = n.xy * 0.5 + 0.5; + return n.xy; +} + +float get_omni_attenuation(float distance, float inv_range, float decay) { + float nd = distance * inv_range; + nd *= nd; + nd *= nd; // nd^4 + nd = max(1.0 - nd, 0.0); + nd *= nd; // nd^2 + return nd * pow(max(distance, 0.0001), -decay); +} + +void main() { + uint voxel_index = uint(gl_GlobalInvocationID.x); + + //used for skipping voxels every N frames + if (params.process_increment > 1) { + voxel_index *= params.process_increment; + voxel_index += params.process_offset; + } + + if (voxel_index >= dispatch_data.total_count) { + return; + } + + uint voxel_position = process_voxels.data[voxel_index].position; + + //keep for storing to texture + ivec3 positioni = ivec3((uvec3(voxel_position, voxel_position, voxel_position) >> uvec3(0, 7, 14)) & uvec3(0x7F)); + + vec3 position = vec3(positioni) + vec3(0.5); + position /= cascades.data[params.cascade].to_cell; + position += cascades.data[params.cascade].offset; + + uint voxel_albedo = process_voxels.data[voxel_index].albedo; + + vec3 albedo = vec3(uvec3(voxel_albedo >> 10, voxel_albedo >> 5, voxel_albedo) & uvec3(0x1F)) / float(0x1F); + vec3 light_accum[6] = vec3[](vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0)); + uint valid_aniso = (voxel_albedo >> 15) & 0x3F; + + const vec3 aniso_dir[6] = vec3[]( + vec3(1, 0, 0), + vec3(0, 1, 0), + vec3(0, 0, 1), + vec3(-1, 0, 0), + vec3(0, -1, 0), + vec3(0, 0, -1)); + + // Add indirect light first, in order to save computation resources +#ifdef MODE_PROCESS_DYNAMIC + if (params.bounce_feedback > 0.001) { + vec3 feedback = (params.bounce_feedback < 1.0) ? (albedo * params.bounce_feedback) : mix(albedo, vec3(1.0), params.bounce_feedback - 1.0); + vec3 pos = (vec3(positioni) + vec3(0.5)) * float(params.probe_axis_size - 1) / params.grid_size; + ivec3 probe_base_pos = ivec3(pos); + + float weight_accum[6] = float[](0, 0, 0, 0, 0, 0); + + ivec3 tex_pos = ivec3(probe_base_pos.xy, int(params.cascade)); + tex_pos.x += probe_base_pos.z * int(params.probe_axis_size); + + tex_pos.xy = tex_pos.xy * (OCT_SIZE + 2) + ivec2(1); + + vec3 base_tex_posf = vec3(tex_pos); + vec2 tex_pixel_size = 1.0 / vec2(ivec2((OCT_SIZE + 2) * params.probe_axis_size * params.probe_axis_size, (OCT_SIZE + 2) * params.probe_axis_size)); + vec3 probe_uv_offset = vec3(ivec3(OCT_SIZE + 2, OCT_SIZE + 2, (OCT_SIZE + 2) * params.probe_axis_size)) * tex_pixel_size.xyx; + + for (uint j = 0; j < 8; j++) { + ivec3 offset = (ivec3(j) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1); + ivec3 probe_posi = probe_base_pos; + probe_posi += offset; + + // Compute weight + + vec3 probe_pos = vec3(probe_posi); + vec3 probe_to_pos = pos - probe_pos; + vec3 probe_dir = normalize(-probe_to_pos); + + // Compute lightprobe texture position + + vec3 trilinear = vec3(1.0) - abs(probe_to_pos); + + for (uint k = 0; k < 6; k++) { + if (bool(valid_aniso & (1 << k))) { + vec3 n = aniso_dir[k]; + float weight = trilinear.x * trilinear.y * trilinear.z * max(0, dot(n, probe_dir)); + + if (weight > 0.0 && params.use_occlusion) { + ivec3 occ_indexv = abs((cascades.data[params.cascade].probe_world_offset + probe_posi) & ivec3(1, 1, 1)) * ivec3(1, 2, 4); + vec4 occ_mask = mix(vec4(0.0), vec4(1.0), equal(ivec4(occ_indexv.x | occ_indexv.y), ivec4(0, 1, 2, 3))); + + vec3 occ_pos = (vec3(positioni) + aniso_dir[k] + vec3(0.5)) / params.grid_size; + occ_pos.z += float(params.cascade); + if (occ_indexv.z != 0) { //z bit is on, means index is >=4, so make it switch to the other half of textures + occ_pos.x += 1.0; + } + occ_pos *= vec3(0.5, 1.0, 1.0 / float(params.max_cascades)); //renormalize + float occlusion = dot(textureLod(sampler3D(occlusion_texture, linear_sampler), occ_pos, 0.0), occ_mask); + + weight *= occlusion; + } + + if (weight > 0.0) { + vec3 tex_posf = base_tex_posf + vec3(octahedron_encode(n) * float(OCT_SIZE), 0.0); + tex_posf.xy *= tex_pixel_size; + + vec3 pos_uvw = tex_posf; + pos_uvw.xy += vec2(offset.xy) * probe_uv_offset.xy; + pos_uvw.x += float(offset.z) * probe_uv_offset.z; + vec3 indirect_light = textureLod(sampler2DArray(lightprobe_texture, linear_sampler), pos_uvw, 0.0).rgb; + + light_accum[k] += indirect_light * weight; + weight_accum[k] += weight; + } + } + } + } + + for (uint k = 0; k < 6; k++) { + if (weight_accum[k] > 0.0) { + light_accum[k] /= weight_accum[k]; + light_accum[k] *= feedback; + } + } + } + +#endif + + { + uint rgbe = process_voxels.data[voxel_index].light; + + //read rgbe8985 + float r = float((rgbe & 0xff) << 1); + float g = float((rgbe >> 8) & 0x1ff); + float b = float(((rgbe >> 17) & 0xff) << 1); + float e = float((rgbe >> 25) & 0x1F); + float m = pow(2.0, e - 15.0 - 9.0); + + vec3 l = vec3(r, g, b) * m; + + uint aniso = process_voxels.data[voxel_index].light_aniso; + for (uint i = 0; i < 6; i++) { + float strength = ((aniso >> (i * 5)) & 0x1F) / float(0x1F); + light_accum[i] += l * strength; + } + } + + // Raytrace light + + vec3 pos_to_uvw = 1.0 / params.grid_size; + vec3 uvw_ofs = pos_to_uvw * 0.5; + + for (uint i = 0; i < params.light_count; i++) { + float attenuation = 1.0; + vec3 direction; + float light_distance = 1e20; + + switch (lights.data[i].type) { + case LIGHT_TYPE_DIRECTIONAL: { + direction = -lights.data[i].direction; + } break; + case LIGHT_TYPE_OMNI: { + vec3 rel_vec = lights.data[i].position - position; + direction = normalize(rel_vec); + light_distance = length(rel_vec); + rel_vec.y /= params.y_mult; + attenuation = get_omni_attenuation(light_distance, 1.0 / lights.data[i].radius, lights.data[i].attenuation); + + } break; + case LIGHT_TYPE_SPOT: { + vec3 rel_vec = lights.data[i].position - position; + direction = normalize(rel_vec); + light_distance = length(rel_vec); + rel_vec.y /= params.y_mult; + attenuation = get_omni_attenuation(light_distance, 1.0 / lights.data[i].radius, lights.data[i].attenuation); + + float cos_spot_angle = lights.data[i].cos_spot_angle; + float cos_angle = dot(-direction, lights.data[i].direction); + + if (cos_angle < cos_spot_angle) { + continue; + } + + float scos = max(cos_angle, cos_spot_angle); + float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - cos_spot_angle)); + attenuation *= 1.0 - pow(spot_rim, lights.data[i].inv_spot_attenuation); + } break; + } + + if (attenuation < 0.001) { + continue; + } + + bool hit = false; + + vec3 ray_pos = position; + vec3 ray_dir = direction; + vec3 inv_dir = 1.0 / ray_dir; + + //this is how to properly bias outgoing rays + float cell_size = 1.0 / cascades.data[params.cascade].to_cell; + ray_pos += sign(direction) * cell_size * 0.48; // go almost to the box edge but remain inside + ray_pos += ray_dir * 0.4 * cell_size; //apply a small bias from there + + for (uint j = params.cascade; j < params.max_cascades; j++) { + //convert to local bounds + vec3 pos = ray_pos - cascades.data[j].offset; + pos *= cascades.data[j].to_cell; + float local_distance = light_distance * cascades.data[j].to_cell; + + if (any(lessThan(pos, vec3(0.0))) || any(greaterThanEqual(pos, params.grid_size))) { + continue; //already past bounds for this cascade, goto next + } + + //find maximum advance distance (until reaching bounds) + vec3 t0 = -pos * inv_dir; + vec3 t1 = (params.grid_size - pos) * inv_dir; + vec3 tmax = max(t0, t1); + float max_advance = min(tmax.x, min(tmax.y, tmax.z)); + + max_advance = min(local_distance, max_advance); + + float advance = 0.0; + float occlusion = 1.0; + + while (advance < max_advance) { + //read how much to advance from SDF + vec3 uvw = (pos + ray_dir * advance) * pos_to_uvw; + + float distance = texture(sampler3D(sdf_cascades[j], linear_sampler), uvw).r * 255.0 - 1.0; + if (distance < 0.001) { + //consider hit + hit = true; + break; + } + + occlusion = min(occlusion, distance); + + advance += distance; + } + + if (hit) { + attenuation *= occlusion; + break; + } + + if (advance >= local_distance) { + break; //past light distance, abandon search + } + //change ray origin to collision with bounds + pos += ray_dir * max_advance; + pos /= cascades.data[j].to_cell; + pos += cascades.data[j].offset; + light_distance -= max_advance / cascades.data[j].to_cell; + ray_pos = pos; + } + + if (!hit) { + vec3 light = albedo * lights.data[i].color.rgb * lights.data[i].energy * attenuation; + + for (int j = 0; j < 6; j++) { + if (bool(valid_aniso & (1 << j))) { + light_accum[j] += max(0.0, dot(aniso_dir[j], direction)) * light; + } + } + } + } + + // Store the light in the light texture + + float lumas[6]; + vec3 light_total = vec3(0); + + for (int i = 0; i < 6; i++) { + light_total += light_accum[i]; + lumas[i] = max(light_accum[i].r, max(light_accum[i].g, light_accum[i].b)); + } + + float luma_total = max(light_total.r, max(light_total.g, light_total.b)); + + uint light_total_rgbe; + + { + //compress to RGBE9995 to save space + + const float pow2to9 = 512.0f; + const float B = 15.0f; + const float N = 9.0f; + const float LN2 = 0.6931471805599453094172321215; + + float cRed = clamp(light_total.r, 0.0, 65408.0); + float cGreen = clamp(light_total.g, 0.0, 65408.0); + float cBlue = clamp(light_total.b, 0.0, 65408.0); + + float cMax = max(cRed, max(cGreen, cBlue)); + + float expp = max(-B - 1.0f, floor(log(cMax) / LN2)) + 1.0f + B; + + float sMax = floor((cMax / pow(2.0f, expp - B - N)) + 0.5f); + + float exps = expp + 1.0f; + + if (0.0 <= sMax && sMax < pow2to9) { + exps = expp; + } + + float sRed = floor((cRed / pow(2.0f, exps - B - N)) + 0.5f); + float sGreen = floor((cGreen / pow(2.0f, exps - B - N)) + 0.5f); + float sBlue = floor((cBlue / pow(2.0f, exps - B - N)) + 0.5f); +#ifdef MODE_PROCESS_STATIC + //since its self-save, use RGBE8985 + light_total_rgbe = ((uint(sRed) & 0x1FF) >> 1) | ((uint(sGreen) & 0x1FF) << 8) | (((uint(sBlue) & 0x1FF) >> 1) << 17) | ((uint(exps) & 0x1F) << 25); + +#else + light_total_rgbe = (uint(sRed) & 0x1FF) | ((uint(sGreen) & 0x1FF) << 9) | ((uint(sBlue) & 0x1FF) << 18) | ((uint(exps) & 0x1F) << 27); +#endif + } + +#ifdef MODE_PROCESS_DYNAMIC + + vec4 aniso0; + aniso0.r = lumas[0] / luma_total; + aniso0.g = lumas[1] / luma_total; + aniso0.b = lumas[2] / luma_total; + aniso0.a = lumas[3] / luma_total; + + vec2 aniso1; + aniso1.r = lumas[4] / luma_total; + aniso1.g = lumas[5] / luma_total; + + //save to 3D textures + imageStore(dst_aniso0, positioni, aniso0); + imageStore(dst_aniso1, positioni, vec4(aniso1, 0.0, 0.0)); + imageStore(dst_light, positioni, uvec4(light_total_rgbe)); + + //also fill neighbours, so light interpolation during the indirect pass works + + //recover the neighbour list from the leftover bits + uint neighbours = (voxel_albedo >> 21) | ((voxel_position >> 21) << 11) | ((process_voxels.data[voxel_index].light >> 30) << 22) | ((process_voxels.data[voxel_index].light_aniso >> 30) << 24); + + const uint max_neighbours = 26; + const ivec3 neighbour_positions[max_neighbours] = ivec3[]( + ivec3(-1, -1, -1), + ivec3(-1, -1, 0), + ivec3(-1, -1, 1), + ivec3(-1, 0, -1), + ivec3(-1, 0, 0), + ivec3(-1, 0, 1), + ivec3(-1, 1, -1), + ivec3(-1, 1, 0), + ivec3(-1, 1, 1), + ivec3(0, -1, -1), + ivec3(0, -1, 0), + ivec3(0, -1, 1), + ivec3(0, 0, -1), + ivec3(0, 0, 1), + ivec3(0, 1, -1), + ivec3(0, 1, 0), + ivec3(0, 1, 1), + ivec3(1, -1, -1), + ivec3(1, -1, 0), + ivec3(1, -1, 1), + ivec3(1, 0, -1), + ivec3(1, 0, 0), + ivec3(1, 0, 1), + ivec3(1, 1, -1), + ivec3(1, 1, 0), + ivec3(1, 1, 1)); + + for (uint i = 0; i < max_neighbours; i++) { + if (bool(neighbours & (1 << i))) { + ivec3 neighbour_pos = positioni + neighbour_positions[i]; + imageStore(dst_light, neighbour_pos, uvec4(light_total_rgbe)); + imageStore(dst_aniso0, neighbour_pos, aniso0); + imageStore(dst_aniso1, neighbour_pos, vec4(aniso1, 0.0, 0.0)); + } + } + +#endif + +#ifdef MODE_PROCESS_STATIC + + //save back the anisotropic + + uint light = process_voxels.data[voxel_index].light & (3 << 30); + light |= light_total_rgbe; + process_voxels.data[voxel_index].light = light; //replace + + uint light_aniso = process_voxels.data[voxel_index].light_aniso & (3 << 30); + for (int i = 0; i < 6; i++) { + light_aniso |= min(31, uint((lumas[i] / luma_total) * 31.0)) << (i * 5); + } + + process_voxels.data[voxel_index].light_aniso = light_aniso; + +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/environment/sdfgi_integrate.glsl b/servers/rendering/renderer_rd/shaders/environment/sdfgi_integrate.glsl new file mode 100644 index 0000000000..4bdb0dcc72 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/environment/sdfgi_integrate.glsl @@ -0,0 +1,613 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +#define MAX_CASCADES 8 + +layout(set = 0, binding = 1) uniform texture3D sdf_cascades[MAX_CASCADES]; +layout(set = 0, binding = 2) uniform texture3D light_cascades[MAX_CASCADES]; +layout(set = 0, binding = 3) uniform texture3D aniso0_cascades[MAX_CASCADES]; +layout(set = 0, binding = 4) uniform texture3D aniso1_cascades[MAX_CASCADES]; + +layout(set = 0, binding = 6) uniform sampler linear_sampler; + +struct CascadeData { + vec3 offset; //offset of (0,0,0) in world coordinates + float to_cell; // 1/bounds * grid_size + ivec3 probe_world_offset; + uint pad; + vec4 pad2; +}; + +layout(set = 0, binding = 7, std140) uniform Cascades { + CascadeData data[MAX_CASCADES]; +} +cascades; + +layout(r32ui, set = 0, binding = 8) uniform restrict uimage2DArray lightprobe_texture_data; +layout(rgba16i, set = 0, binding = 9) uniform restrict iimage2DArray lightprobe_history_texture; +layout(rgba32i, set = 0, binding = 10) uniform restrict iimage2D lightprobe_average_texture; + +//used for scrolling +layout(rgba16i, set = 0, binding = 11) uniform restrict iimage2DArray lightprobe_history_scroll_texture; +layout(rgba32i, set = 0, binding = 12) uniform restrict iimage2D lightprobe_average_scroll_texture; + +layout(rgba32i, set = 0, binding = 13) uniform restrict iimage2D lightprobe_average_parent_texture; + +layout(rgba16f, set = 0, binding = 14) uniform restrict writeonly image2DArray lightprobe_ambient_texture; + +#ifdef USE_CUBEMAP_ARRAY +layout(set = 1, binding = 0) uniform textureCubeArray sky_irradiance; +#else +layout(set = 1, binding = 0) uniform textureCube sky_irradiance; +#endif +layout(set = 1, binding = 1) uniform sampler linear_sampler_mipmaps; + +#define HISTORY_BITS 10 + +#define SKY_MODE_DISABLED 0 +#define SKY_MODE_COLOR 1 +#define SKY_MODE_SKY 2 + +layout(push_constant, std430) uniform Params { + vec3 grid_size; + uint max_cascades; + + uint probe_axis_size; + uint cascade; + uint history_index; + uint history_size; + + uint ray_count; + float ray_bias; + ivec2 image_size; + + ivec3 world_offset; + uint sky_mode; + + ivec3 scroll; + float sky_energy; + + vec3 sky_color; + float y_mult; + + bool store_ambient_texture; + uint pad[3]; +} +params; + +const float PI = 3.14159265f; +const float GOLDEN_ANGLE = PI * (3.0 - sqrt(5.0)); + +vec3 vogel_hemisphere(uint p_index, uint p_count, float p_offset) { + float r = sqrt(float(p_index) + 0.5f) / sqrt(float(p_count)); + float theta = float(p_index) * GOLDEN_ANGLE + p_offset; + float y = cos(r * PI * 0.5); + float l = sin(r * PI * 0.5); + return vec3(l * cos(theta), l * sin(theta), y * (float(p_index & 1) * 2.0 - 1.0)); +} + +uvec3 hash3(uvec3 x) { + x = ((x >> 16) ^ x) * 0x45d9f3b; + x = ((x >> 16) ^ x) * 0x45d9f3b; + x = (x >> 16) ^ x; + return x; +} + +float hashf3(vec3 co) { + return fract(sin(dot(co, vec3(12.9898, 78.233, 137.13451))) * 43758.5453); +} + +vec3 octahedron_encode(vec2 f) { + // https://twitter.com/Stubbesaurus/status/937994790553227264 + f = f * 2.0 - 1.0; + vec3 n = vec3(f.x, f.y, 1.0f - abs(f.x) - abs(f.y)); + float t = clamp(-n.z, 0.0, 1.0); + n.x += n.x >= 0 ? -t : t; + n.y += n.y >= 0 ? -t : t; + return normalize(n); +} + +uint rgbe_encode(vec3 color) { + const float pow2to9 = 512.0f; + const float B = 15.0f; + const float N = 9.0f; + const float LN2 = 0.6931471805599453094172321215; + + float cRed = clamp(color.r, 0.0, 65408.0); + float cGreen = clamp(color.g, 0.0, 65408.0); + float cBlue = clamp(color.b, 0.0, 65408.0); + + float cMax = max(cRed, max(cGreen, cBlue)); + + float expp = max(-B - 1.0f, floor(log(cMax) / LN2)) + 1.0f + B; + + float sMax = floor((cMax / pow(2.0f, expp - B - N)) + 0.5f); + + float exps = expp + 1.0f; + + if (0.0 <= sMax && sMax < pow2to9) { + exps = expp; + } + + float sRed = floor((cRed / pow(2.0f, exps - B - N)) + 0.5f); + float sGreen = floor((cGreen / pow(2.0f, exps - B - N)) + 0.5f); + float sBlue = floor((cBlue / pow(2.0f, exps - B - N)) + 0.5f); + return (uint(sRed) & 0x1FF) | ((uint(sGreen) & 0x1FF) << 9) | ((uint(sBlue) & 0x1FF) << 18) | ((uint(exps) & 0x1F) << 27); +} + +struct SH { +#if (SH_SIZE == 16) + float c[48]; +#else + float c[28]; +#endif +}; + +shared SH sh_accum[64]; //8x8 + +void main() { + ivec2 pos = ivec2(gl_GlobalInvocationID.xy); + if (any(greaterThanEqual(pos, params.image_size))) { //too large, do nothing + return; + } + + uint probe_index = gl_LocalInvocationID.x + gl_LocalInvocationID.y * 8; + +#ifdef MODE_PROCESS + + float probe_cell_size = float(params.grid_size.x / float(params.probe_axis_size - 1)) / cascades.data[params.cascade].to_cell; + + ivec3 probe_cell; + probe_cell.x = pos.x % int(params.probe_axis_size); + probe_cell.y = pos.y; + probe_cell.z = pos.x / int(params.probe_axis_size); + + vec3 probe_pos = cascades.data[params.cascade].offset + vec3(probe_cell) * probe_cell_size; + vec3 pos_to_uvw = 1.0 / params.grid_size; + + for (uint i = 0; i < SH_SIZE * 3; i++) { + sh_accum[probe_index].c[i] = 0.0; + } + + // quickly ensure each probe has a different "offset" for the vogel function, based on integer world position + uvec3 h3 = hash3(uvec3(params.world_offset + probe_cell)); + float offset = hashf3(vec3(h3 & uvec3(0xFFFFF))); + + //for a more homogeneous hemisphere, alternate based on history frames + uint ray_offset = params.history_index; + uint ray_mult = params.history_size; + uint ray_total = ray_mult * params.ray_count; + + for (uint i = 0; i < params.ray_count; i++) { + vec3 ray_dir = vogel_hemisphere(ray_offset + i * ray_mult, ray_total, offset); + ray_dir.y *= params.y_mult; + ray_dir = normalize(ray_dir); + + //needs to be visible + vec3 ray_pos = probe_pos; + vec3 inv_dir = 1.0 / ray_dir; + + bool hit = false; + uint hit_cascade; + + float bias = params.ray_bias; + vec3 abs_ray_dir = abs(ray_dir); + ray_pos += ray_dir * 1.0 / max(abs_ray_dir.x, max(abs_ray_dir.y, abs_ray_dir.z)) * bias / cascades.data[params.cascade].to_cell; + vec3 uvw; + + for (uint j = params.cascade; j < params.max_cascades; j++) { + //convert to local bounds + vec3 pos = ray_pos - cascades.data[j].offset; + pos *= cascades.data[j].to_cell; + + if (any(lessThan(pos, vec3(0.0))) || any(greaterThanEqual(pos, params.grid_size))) { + continue; //already past bounds for this cascade, goto next + } + + //find maximum advance distance (until reaching bounds) + vec3 t0 = -pos * inv_dir; + vec3 t1 = (params.grid_size - pos) * inv_dir; + vec3 tmax = max(t0, t1); + float max_advance = min(tmax.x, min(tmax.y, tmax.z)); + + float advance = 0.0; + + while (advance < max_advance) { + //read how much to advance from SDF + uvw = (pos + ray_dir * advance) * pos_to_uvw; + + float distance = texture(sampler3D(sdf_cascades[j], linear_sampler), uvw).r * 255.0 - 1.0; + if (distance < 0.05) { + //consider hit + hit = true; + break; + } + + advance += distance; + } + + if (hit) { + hit_cascade = j; + break; + } + + //change ray origin to collision with bounds + pos += ray_dir * max_advance; + pos /= cascades.data[j].to_cell; + pos += cascades.data[j].offset; + ray_pos = pos; + } + + vec4 light; + if (hit) { + //avoid reading different texture from different threads + for (uint j = params.cascade; j < params.max_cascades; j++) { + if (j == hit_cascade) { + const float EPSILON = 0.001; + vec3 hit_normal = normalize(vec3( + texture(sampler3D(sdf_cascades[hit_cascade], linear_sampler), uvw + vec3(EPSILON, 0.0, 0.0)).r - texture(sampler3D(sdf_cascades[hit_cascade], linear_sampler), uvw - vec3(EPSILON, 0.0, 0.0)).r, + texture(sampler3D(sdf_cascades[hit_cascade], linear_sampler), uvw + vec3(0.0, EPSILON, 0.0)).r - texture(sampler3D(sdf_cascades[hit_cascade], linear_sampler), uvw - vec3(0.0, EPSILON, 0.0)).r, + texture(sampler3D(sdf_cascades[hit_cascade], linear_sampler), uvw + vec3(0.0, 0.0, EPSILON)).r - texture(sampler3D(sdf_cascades[hit_cascade], linear_sampler), uvw - vec3(0.0, 0.0, EPSILON)).r)); + + vec3 hit_light = texture(sampler3D(light_cascades[hit_cascade], linear_sampler), uvw).rgb; + vec4 aniso0 = texture(sampler3D(aniso0_cascades[hit_cascade], linear_sampler), uvw); + vec3 hit_aniso0 = aniso0.rgb; + vec3 hit_aniso1 = vec3(aniso0.a, texture(sampler3D(aniso1_cascades[hit_cascade], linear_sampler), uvw).rg); + + //one liner magic + light.rgb = hit_light * (dot(max(vec3(0.0), (hit_normal * hit_aniso0)), vec3(1.0)) + dot(max(vec3(0.0), (-hit_normal * hit_aniso1)), vec3(1.0))); + light.a = 1.0; + } + } + + } else if (params.sky_mode == SKY_MODE_SKY) { +#ifdef USE_CUBEMAP_ARRAY + light.rgb = textureLod(samplerCubeArray(sky_irradiance, linear_sampler_mipmaps), vec4(ray_dir, 0.0), 2.0).rgb; // Use second mipmap because we don't usually throw a lot of rays, so this compensates. +#else + light.rgb = textureLod(samplerCube(sky_irradiance, linear_sampler_mipmaps), ray_dir, 2.0).rgb; // Use second mipmap because we don't usually throw a lot of rays, so this compensates. +#endif + light.rgb *= params.sky_energy; + light.a = 0.0; + + } else if (params.sky_mode == SKY_MODE_COLOR) { + light.rgb = params.sky_color; + light.rgb *= params.sky_energy; + light.a = 0.0; + } else { + light = vec4(0, 0, 0, 0); + } + + vec3 ray_dir2 = ray_dir * ray_dir; + +#define SH_ACCUM(m_idx, m_value) \ + { \ + vec3 l = light.rgb * (m_value); \ + sh_accum[probe_index].c[m_idx * 3 + 0] += l.r; \ + sh_accum[probe_index].c[m_idx * 3 + 1] += l.g; \ + sh_accum[probe_index].c[m_idx * 3 + 2] += l.b; \ + } + SH_ACCUM(0, 0.282095); //l0 + SH_ACCUM(1, 0.488603 * ray_dir.y); //l1n1 + SH_ACCUM(2, 0.488603 * ray_dir.z); //l1n0 + SH_ACCUM(3, 0.488603 * ray_dir.x); //l1p1 + SH_ACCUM(4, 1.092548 * ray_dir.x * ray_dir.y); //l2n2 + SH_ACCUM(5, 1.092548 * ray_dir.y * ray_dir.z); //l2n1 + SH_ACCUM(6, 0.315392 * (3.0 * ray_dir2.z - 1.0)); //l20 + SH_ACCUM(7, 1.092548 * ray_dir.x * ray_dir.z); //l2p1 + SH_ACCUM(8, 0.546274 * (ray_dir2.x - ray_dir2.y)); //l2p2 +#if (SH_SIZE == 16) + SH_ACCUM(9, 0.590043 * ray_dir.y * (3.0f * ray_dir2.x - ray_dir2.y)); + SH_ACCUM(10, 2.890611 * ray_dir.y * ray_dir.x * ray_dir.z); + SH_ACCUM(11, 0.646360 * ray_dir.y * (-1.0f + 5.0f * ray_dir2.z)); + SH_ACCUM(12, 0.373176 * (5.0f * ray_dir2.z * ray_dir.z - 3.0f * ray_dir.z)); + SH_ACCUM(13, 0.457045 * ray_dir.x * (-1.0f + 5.0f * ray_dir2.z)); + SH_ACCUM(14, 1.445305 * (ray_dir2.x - ray_dir2.y) * ray_dir.z); + SH_ACCUM(15, 0.590043 * ray_dir.x * (ray_dir2.x - 3.0f * ray_dir2.y)); + +#endif + } + + for (uint i = 0; i < SH_SIZE; i++) { + // store in history texture + ivec3 prev_pos = ivec3(pos.x, pos.y * SH_SIZE + i, int(params.history_index)); + ivec2 average_pos = prev_pos.xy; + + vec4 value = vec4(sh_accum[probe_index].c[i * 3 + 0], sh_accum[probe_index].c[i * 3 + 1], sh_accum[probe_index].c[i * 3 + 2], 1.0) * 4.0 / float(params.ray_count); + + ivec4 ivalue = clamp(ivec4(value * float(1 << HISTORY_BITS)), -32768, 32767); //clamp to 16 bits, so higher values don't break average + + ivec4 prev_value = imageLoad(lightprobe_history_texture, prev_pos); + ivec4 average = imageLoad(lightprobe_average_texture, average_pos); + + average -= prev_value; + average += ivalue; + + imageStore(lightprobe_history_texture, prev_pos, ivalue); + imageStore(lightprobe_average_texture, average_pos, average); + + if (params.store_ambient_texture && i == 0) { + ivec3 ambient_pos = ivec3(pos, int(params.cascade)); + vec4 ambient_light = (vec4(average) / float(params.history_size)) / float(1 << HISTORY_BITS); + ambient_light *= 0.88622; // SHL0 + imageStore(lightprobe_ambient_texture, ambient_pos, ambient_light); + } + } +#endif // MODE PROCESS + +#ifdef MODE_STORE + + // converting to octahedral in this step is required because + // octahedral is much faster to read from the screen than spherical harmonics, + // despite the very slight quality loss + + ivec2 sh_pos = (pos / OCT_SIZE) * ivec2(1, SH_SIZE); + ivec2 oct_pos = (pos / OCT_SIZE) * (OCT_SIZE + 2) + ivec2(1); + ivec2 local_pos = pos % OCT_SIZE; + + //compute the octahedral normal for this texel + vec3 normal = octahedron_encode(vec2(local_pos) / float(OCT_SIZE)); + + // read the spherical harmonic + + vec3 normal2 = normal * normal; + float c[SH_SIZE] = float[]( + + 0.282095, //l0 + 0.488603 * normal.y, //l1n1 + 0.488603 * normal.z, //l1n0 + 0.488603 * normal.x, //l1p1 + 1.092548 * normal.x * normal.y, //l2n2 + 1.092548 * normal.y * normal.z, //l2n1 + 0.315392 * (3.0 * normal2.z - 1.0), //l20 + 1.092548 * normal.x * normal.z, //l2p1 + 0.546274 * (normal2.x - normal2.y) //l2p2 +#if (SH_SIZE == 16) + , + 0.590043 * normal.y * (3.0f * normal2.x - normal2.y), + 2.890611 * normal.y * normal.x * normal.z, + 0.646360 * normal.y * (-1.0f + 5.0f * normal2.z), + 0.373176 * (5.0f * normal2.z * normal.z - 3.0f * normal.z), + 0.457045 * normal.x * (-1.0f + 5.0f * normal2.z), + 1.445305 * (normal2.x - normal2.y) * normal.z, + 0.590043 * normal.x * (normal2.x - 3.0f * normal2.y) + +#endif + ); + + const float l_mult[SH_SIZE] = float[]( + 1.0, + 2.0 / 3.0, + 2.0 / 3.0, + 2.0 / 3.0, + 1.0 / 4.0, + 1.0 / 4.0, + 1.0 / 4.0, + 1.0 / 4.0, + 1.0 / 4.0 +#if (SH_SIZE == 16) + , // l4 does not contribute to irradiance + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 +#endif + ); + + vec3 irradiance = vec3(0.0); + vec3 radiance = vec3(0.0); + + for (uint i = 0; i < SH_SIZE; i++) { + // store in history texture + ivec2 average_pos = sh_pos + ivec2(0, i); + ivec4 average = imageLoad(lightprobe_average_texture, average_pos); + + vec4 sh = (vec4(average) / float(params.history_size)) / float(1 << HISTORY_BITS); + + vec3 m = sh.rgb * c[i] * 4.0; + + irradiance += m * l_mult[i]; + radiance += m; + } + + //encode RGBE9995 for the final texture + + uint irradiance_rgbe = rgbe_encode(irradiance); + uint radiance_rgbe = rgbe_encode(radiance); + + //store in octahedral map + + ivec3 texture_pos = ivec3(oct_pos, int(params.cascade)); + ivec3 copy_to[4] = ivec3[](ivec3(-2, -2, -2), ivec3(-2, -2, -2), ivec3(-2, -2, -2), ivec3(-2, -2, -2)); + copy_to[0] = texture_pos + ivec3(local_pos, 0); + + if (local_pos == ivec2(0, 0)) { + copy_to[1] = texture_pos + ivec3(OCT_SIZE - 1, -1, 0); + copy_to[2] = texture_pos + ivec3(-1, OCT_SIZE - 1, 0); + copy_to[3] = texture_pos + ivec3(OCT_SIZE, OCT_SIZE, 0); + } else if (local_pos == ivec2(OCT_SIZE - 1, 0)) { + copy_to[1] = texture_pos + ivec3(0, -1, 0); + copy_to[2] = texture_pos + ivec3(OCT_SIZE, OCT_SIZE - 1, 0); + copy_to[3] = texture_pos + ivec3(-1, OCT_SIZE, 0); + } else if (local_pos == ivec2(0, OCT_SIZE - 1)) { + copy_to[1] = texture_pos + ivec3(-1, 0, 0); + copy_to[2] = texture_pos + ivec3(OCT_SIZE - 1, OCT_SIZE, 0); + copy_to[3] = texture_pos + ivec3(OCT_SIZE, -1, 0); + } else if (local_pos == ivec2(OCT_SIZE - 1, OCT_SIZE - 1)) { + copy_to[1] = texture_pos + ivec3(0, OCT_SIZE, 0); + copy_to[2] = texture_pos + ivec3(OCT_SIZE, 0, 0); + copy_to[3] = texture_pos + ivec3(-1, -1, 0); + } else if (local_pos.y == 0) { + copy_to[1] = texture_pos + ivec3(OCT_SIZE - local_pos.x - 1, local_pos.y - 1, 0); + } else if (local_pos.x == 0) { + copy_to[1] = texture_pos + ivec3(local_pos.x - 1, OCT_SIZE - local_pos.y - 1, 0); + } else if (local_pos.y == OCT_SIZE - 1) { + copy_to[1] = texture_pos + ivec3(OCT_SIZE - local_pos.x - 1, local_pos.y + 1, 0); + } else if (local_pos.x == OCT_SIZE - 1) { + copy_to[1] = texture_pos + ivec3(local_pos.x + 1, OCT_SIZE - local_pos.y - 1, 0); + } + + for (int i = 0; i < 4; i++) { + if (copy_to[i] == ivec3(-2, -2, -2)) { + continue; + } + imageStore(lightprobe_texture_data, copy_to[i], uvec4(irradiance_rgbe)); + imageStore(lightprobe_texture_data, copy_to[i] + ivec3(0, 0, int(params.max_cascades)), uvec4(radiance_rgbe)); + } + +#endif + +#ifdef MODE_SCROLL + + ivec3 probe_cell; + probe_cell.x = pos.x % int(params.probe_axis_size); + probe_cell.y = pos.y; + probe_cell.z = pos.x / int(params.probe_axis_size); + + ivec3 read_probe = probe_cell - params.scroll; + + if (all(greaterThanEqual(read_probe, ivec3(0))) && all(lessThan(read_probe, ivec3(params.probe_axis_size)))) { + // can scroll + ivec2 tex_pos; + tex_pos = read_probe.xy; + tex_pos.x += read_probe.z * int(params.probe_axis_size); + + //scroll + for (uint j = 0; j < params.history_size; j++) { + for (int i = 0; i < SH_SIZE; i++) { + // copy from history texture + ivec3 src_pos = ivec3(tex_pos.x, tex_pos.y * SH_SIZE + i, int(j)); + ivec3 dst_pos = ivec3(pos.x, pos.y * SH_SIZE + i, int(j)); + ivec4 value = imageLoad(lightprobe_history_texture, src_pos); + imageStore(lightprobe_history_scroll_texture, dst_pos, value); + } + } + + for (int i = 0; i < SH_SIZE; i++) { + // copy from average texture + ivec2 src_pos = ivec2(tex_pos.x, tex_pos.y * SH_SIZE + i); + ivec2 dst_pos = ivec2(pos.x, pos.y * SH_SIZE + i); + ivec4 value = imageLoad(lightprobe_average_texture, src_pos); + imageStore(lightprobe_average_scroll_texture, dst_pos, value); + } + } else if (params.cascade < params.max_cascades - 1) { + //can't scroll, must look for position in parent cascade + + //to global coords + float cell_to_probe = float(params.grid_size.x / float(params.probe_axis_size - 1)); + + float probe_cell_size = cell_to_probe / cascades.data[params.cascade].to_cell; + vec3 probe_pos = cascades.data[params.cascade].offset + vec3(probe_cell) * probe_cell_size; + + //to parent local coords + float probe_cell_size_next = cell_to_probe / cascades.data[params.cascade + 1].to_cell; + probe_pos -= cascades.data[params.cascade + 1].offset; + probe_pos /= probe_cell_size_next; + + ivec3 probe_posi = ivec3(probe_pos); + //add up all light, no need to use occlusion here, since occlusion will do its work afterwards + + vec4 average_light[SH_SIZE] = vec4[](vec4(0), vec4(0), vec4(0), vec4(0), vec4(0), vec4(0), vec4(0), vec4(0), vec4(0) +#if (SH_SIZE == 16) + , + vec4(0), vec4(0), vec4(0), vec4(0), vec4(0), vec4(0), vec4(0) +#endif + ); + float total_weight = 0.0; + + for (int i = 0; i < 8; i++) { + ivec3 offset = probe_posi + ((ivec3(i) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1)); + + vec3 trilinear = vec3(1.0) - abs(probe_pos - vec3(offset)); + float weight = trilinear.x * trilinear.y * trilinear.z; + + ivec2 tex_pos; + tex_pos = offset.xy; + tex_pos.x += offset.z * int(params.probe_axis_size); + + for (int j = 0; j < SH_SIZE; j++) { + // copy from history texture + ivec2 src_pos = ivec2(tex_pos.x, tex_pos.y * SH_SIZE + j); + ivec4 average = imageLoad(lightprobe_average_parent_texture, src_pos); + vec4 value = (vec4(average) / float(params.history_size)) / float(1 << HISTORY_BITS); + average_light[j] += value * weight; + } + + total_weight += weight; + } + + if (total_weight > 0.0) { + total_weight = 1.0 / total_weight; + } + //store the averaged values everywhere + + for (int i = 0; i < SH_SIZE; i++) { + ivec4 ivalue = clamp(ivec4(average_light[i] * total_weight * float(1 << HISTORY_BITS)), ivec4(-32768), ivec4(32767)); //clamp to 16 bits, so higher values don't break average + // copy from history texture + ivec3 dst_pos = ivec3(pos.x, pos.y * SH_SIZE + i, 0); + for (uint j = 0; j < params.history_size; j++) { + dst_pos.z = int(j); + imageStore(lightprobe_history_scroll_texture, dst_pos, ivalue); + } + + ivalue *= int(params.history_size); //average needs to have all history added up + imageStore(lightprobe_average_scroll_texture, dst_pos.xy, ivalue); + } + + } else { + //scroll at the edge of the highest cascade, just copy what is there, + //since its the closest we have anyway + + for (uint j = 0; j < params.history_size; j++) { + ivec2 tex_pos; + tex_pos = probe_cell.xy; + tex_pos.x += probe_cell.z * int(params.probe_axis_size); + + for (int i = 0; i < SH_SIZE; i++) { + // copy from history texture + ivec3 src_pos = ivec3(tex_pos.x, tex_pos.y * SH_SIZE + i, int(j)); + ivec3 dst_pos = ivec3(pos.x, pos.y * SH_SIZE + i, int(j)); + ivec4 value = imageLoad(lightprobe_history_texture, dst_pos); + imageStore(lightprobe_history_scroll_texture, dst_pos, value); + } + } + + for (int i = 0; i < SH_SIZE; i++) { + // copy from average texture + ivec2 spos = ivec2(pos.x, pos.y * SH_SIZE + i); + ivec4 average = imageLoad(lightprobe_average_texture, spos); + imageStore(lightprobe_average_scroll_texture, spos, average); + } + } + +#endif + +#ifdef MODE_SCROLL_STORE + + //do not update probe texture, as these will be updated later + + for (uint j = 0; j < params.history_size; j++) { + for (int i = 0; i < SH_SIZE; i++) { + // copy from history texture + ivec3 spos = ivec3(pos.x, pos.y * SH_SIZE + i, int(j)); + ivec4 value = imageLoad(lightprobe_history_scroll_texture, spos); + imageStore(lightprobe_history_texture, spos, value); + } + } + + for (int i = 0; i < SH_SIZE; i++) { + // copy from average texture + ivec2 spos = ivec2(pos.x, pos.y * SH_SIZE + i); + ivec4 average = imageLoad(lightprobe_average_scroll_texture, spos); + imageStore(lightprobe_average_texture, spos, average); + } + +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/environment/sdfgi_preprocess.glsl b/servers/rendering/renderer_rd/shaders/environment/sdfgi_preprocess.glsl new file mode 100644 index 0000000000..bce98f4054 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/environment/sdfgi_preprocess.glsl @@ -0,0 +1,1056 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +#ifdef MODE_JUMPFLOOD_OPTIMIZED +#define GROUP_SIZE 8 + +layout(local_size_x = GROUP_SIZE, local_size_y = GROUP_SIZE, local_size_z = GROUP_SIZE) in; + +#elif defined(MODE_OCCLUSION) || defined(MODE_SCROLL) +//buffer layout +layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in; + +#else +//grid layout +layout(local_size_x = 4, local_size_y = 4, local_size_z = 4) in; + +#endif + +#if defined(MODE_INITIALIZE_JUMP_FLOOD) || defined(MODE_INITIALIZE_JUMP_FLOOD_HALF) +layout(r16ui, set = 0, binding = 1) uniform restrict readonly uimage3D src_color; +layout(rgba8ui, set = 0, binding = 2) uniform restrict writeonly uimage3D dst_positions; +#endif + +#ifdef MODE_UPSCALE_JUMP_FLOOD +layout(r16ui, set = 0, binding = 1) uniform restrict readonly uimage3D src_color; +layout(rgba8ui, set = 0, binding = 2) uniform restrict readonly uimage3D src_positions_half; +layout(rgba8ui, set = 0, binding = 3) uniform restrict writeonly uimage3D dst_positions; +#endif + +#if defined(MODE_JUMPFLOOD) || defined(MODE_JUMPFLOOD_OPTIMIZED) +layout(rgba8ui, set = 0, binding = 1) uniform restrict readonly uimage3D src_positions; +layout(rgba8ui, set = 0, binding = 2) uniform restrict writeonly uimage3D dst_positions; +#endif + +#ifdef MODE_JUMPFLOOD_OPTIMIZED + +shared uvec4 group_positions[(GROUP_SIZE + 2) * (GROUP_SIZE + 2) * (GROUP_SIZE + 2)]; //4x4x4 with margins + +void group_store(ivec3 p_pos, uvec4 p_value) { + uint offset = uint(p_pos.z * (GROUP_SIZE + 2) * (GROUP_SIZE + 2) + p_pos.y * (GROUP_SIZE + 2) + p_pos.x); + group_positions[offset] = p_value; +} + +uvec4 group_load(ivec3 p_pos) { + uint offset = uint(p_pos.z * (GROUP_SIZE + 2) * (GROUP_SIZE + 2) + p_pos.y * (GROUP_SIZE + 2) + p_pos.x); + return group_positions[offset]; +} + +#endif + +#ifdef MODE_OCCLUSION + +layout(r16ui, set = 0, binding = 1) uniform restrict readonly uimage3D src_color; +layout(r8, set = 0, binding = 2) uniform restrict image3D dst_occlusion[8]; +layout(r32ui, set = 0, binding = 3) uniform restrict readonly uimage3D src_facing; + +const uvec2 group_size_offset[11] = uvec2[](uvec2(1, 0), uvec2(3, 1), uvec2(6, 4), uvec2(10, 10), uvec2(15, 20), uvec2(21, 35), uvec2(28, 56), uvec2(36, 84), uvec2(42, 120), uvec2(46, 162), uvec2(48, 208)); +const uint group_pos[256] = uint[](0, + 65536, 256, 1, + 131072, 65792, 512, 65537, 257, 2, + 196608, 131328, 66048, 768, 131073, 65793, 513, 65538, 258, 3, + 262144, 196864, 131584, 66304, 1024, 196609, 131329, 66049, 769, 131074, 65794, 514, 65539, 259, 4, + 327680, 262400, 197120, 131840, 66560, 1280, 262145, 196865, 131585, 66305, 1025, 196610, 131330, 66050, 770, 131075, 65795, 515, 65540, 260, 5, + 393216, 327936, 262656, 197376, 132096, 66816, 1536, 327681, 262401, 197121, 131841, 66561, 1281, 262146, 196866, 131586, 66306, 1026, 196611, 131331, 66051, 771, 131076, 65796, 516, 65541, 261, 6, + 458752, 393472, 328192, 262912, 197632, 132352, 67072, 1792, 393217, 327937, 262657, 197377, 132097, 66817, 1537, 327682, 262402, 197122, 131842, 66562, 1282, 262147, 196867, 131587, 66307, 1027, 196612, 131332, 66052, 772, 131077, 65797, 517, 65542, 262, 7, + 459008, 393728, 328448, 263168, 197888, 132608, 67328, 458753, 393473, 328193, 262913, 197633, 132353, 67073, 1793, 393218, 327938, 262658, 197378, 132098, 66818, 1538, 327683, 262403, 197123, 131843, 66563, 1283, 262148, 196868, 131588, 66308, 1028, 196613, 131333, 66053, 773, 131078, 65798, 518, 65543, 263, + 459264, 393984, 328704, 263424, 198144, 132864, 459009, 393729, 328449, 263169, 197889, 132609, 67329, 458754, 393474, 328194, 262914, 197634, 132354, 67074, 1794, 393219, 327939, 262659, 197379, 132099, 66819, 1539, 327684, 262404, 197124, 131844, 66564, 1284, 262149, 196869, 131589, 66309, 1029, 196614, 131334, 66054, 774, 131079, 65799, 519, + 459520, 394240, 328960, 263680, 198400, 459265, 393985, 328705, 263425, 198145, 132865, 459010, 393730, 328450, 263170, 197890, 132610, 67330, 458755, 393475, 328195, 262915, 197635, 132355, 67075, 1795, 393220, 327940, 262660, 197380, 132100, 66820, 1540, 327685, 262405, 197125, 131845, 66565, 1285, 262150, 196870, 131590, 66310, 1030, 196615, 131335, 66055, 775); + +shared uint occlusion_facing[((OCCLUSION_SIZE * 2) * (OCCLUSION_SIZE * 2) * (OCCLUSION_SIZE * 2)) / 4]; + +uint get_facing(ivec3 p_pos) { + uint ofs = uint(p_pos.z * OCCLUSION_SIZE * 2 * OCCLUSION_SIZE * 2 + p_pos.y * OCCLUSION_SIZE * 2 + p_pos.x); + uint v = occlusion_facing[ofs / 4]; + return (v >> ((ofs % 4) * 8)) & 0xFF; +} + +#endif + +#ifdef MODE_STORE + +layout(rgba8ui, set = 0, binding = 1) uniform restrict readonly uimage3D src_positions; +layout(r16ui, set = 0, binding = 2) uniform restrict readonly uimage3D src_albedo; +layout(r8, set = 0, binding = 3) uniform restrict readonly image3D src_occlusion[8]; +layout(r32ui, set = 0, binding = 4) uniform restrict readonly uimage3D src_light; +layout(r32ui, set = 0, binding = 5) uniform restrict readonly uimage3D src_light_aniso; +layout(r32ui, set = 0, binding = 6) uniform restrict readonly uimage3D src_facing; + +layout(r8, set = 0, binding = 7) uniform restrict writeonly image3D dst_sdf; +layout(r16ui, set = 0, binding = 8) uniform restrict writeonly uimage3D dst_occlusion; + +layout(set = 0, binding = 10, std430) restrict buffer DispatchData { + uint x; + uint y; + uint z; + uint total_count; +} +dispatch_data; + +struct ProcessVoxel { + uint position; // xyz 7 bit packed, extra 11 bits for neighbors. + uint albedo; //rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neighbours + uint light; //rgbe8985 encoded total saved light, extra 2 bits for neighbours + uint light_aniso; //55555 light anisotropy, extra 2 bits for neighbours + //total neighbours: 26 +}; + +layout(set = 0, binding = 11, std430) restrict buffer writeonly ProcessVoxels { + ProcessVoxel data[]; +} +dst_process_voxels; + +shared ProcessVoxel store_positions[4 * 4 * 4]; +shared uint store_position_count; +shared uint store_from_index; +#endif + +#ifdef MODE_SCROLL + +layout(r16ui, set = 0, binding = 1) uniform restrict writeonly uimage3D dst_albedo; +layout(r32ui, set = 0, binding = 2) uniform restrict writeonly uimage3D dst_facing; +layout(r32ui, set = 0, binding = 3) uniform restrict writeonly uimage3D dst_light; +layout(r32ui, set = 0, binding = 4) uniform restrict writeonly uimage3D dst_light_aniso; + +layout(set = 0, binding = 5, std430) restrict buffer readonly DispatchData { + uint x; + uint y; + uint z; + uint total_count; +} +dispatch_data; + +struct ProcessVoxel { + uint position; // xyz 7 bit packed, extra 11 bits for neighbors. + uint albedo; //rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neighbours + uint light; //rgbe8985 encoded total saved light, extra 2 bits for neighbours + uint light_aniso; //55555 light anisotropy, extra 2 bits for neighbours + //total neighbours: 26 +}; + +layout(set = 0, binding = 6, std430) restrict buffer readonly ProcessVoxels { + ProcessVoxel data[]; +} +src_process_voxels; + +#endif + +#ifdef MODE_SCROLL_OCCLUSION + +layout(r8, set = 0, binding = 1) uniform restrict image3D dst_occlusion[8]; +layout(r16ui, set = 0, binding = 2) uniform restrict readonly uimage3D src_occlusion; + +#endif + +layout(push_constant, std430) uniform Params { + ivec3 scroll; + + int grid_size; + + ivec3 probe_offset; + int step_size; + + bool half_size; + uint occlusion_index; + int cascade; + uint pad; +} +params; + +void main() { +#ifdef MODE_SCROLL + + // Pixel being shaded + int index = int(gl_GlobalInvocationID.x); + if (index >= dispatch_data.total_count) { //too big + return; + } + + ivec3 read_pos = (ivec3(src_process_voxels.data[index].position) >> ivec3(0, 7, 14)) & ivec3(0x7F); + ivec3 write_pos = read_pos + params.scroll; + + if (any(lessThan(write_pos, ivec3(0))) || any(greaterThanEqual(write_pos, ivec3(params.grid_size)))) { + return; // Fits outside the 3D texture, don't do anything. + } + + uint albedo = ((src_process_voxels.data[index].albedo & 0x7FFF) << 1) | 1; //add solid bit + imageStore(dst_albedo, write_pos, uvec4(albedo)); + + uint facing = (src_process_voxels.data[index].albedo >> 15) & 0x3F; //6 anisotropic facing bits + imageStore(dst_facing, write_pos, uvec4(facing)); + + uint light = src_process_voxels.data[index].light & 0x3fffffff; //30 bits of RGBE8985 + imageStore(dst_light, write_pos, uvec4(light)); + + uint light_aniso = src_process_voxels.data[index].light_aniso & 0x3fffffff; //30 bits of 6 anisotropic 5 bits values + imageStore(dst_light_aniso, write_pos, uvec4(light_aniso)); + +#endif + +#ifdef MODE_SCROLL_OCCLUSION + + ivec3 pos = ivec3(gl_GlobalInvocationID.xyz); + if (any(greaterThanEqual(pos, ivec3(params.grid_size) - abs(params.scroll)))) { //too large, do nothing + return; + } + + ivec3 read_pos = pos + max(ivec3(0), -params.scroll); + ivec3 write_pos = pos + max(ivec3(0), params.scroll); + + read_pos.z += params.cascade * params.grid_size; + uint occlusion = imageLoad(src_occlusion, read_pos).r; + read_pos.x += params.grid_size; + occlusion |= imageLoad(src_occlusion, read_pos).r << 16; + + const uint occlusion_shift[8] = uint[](12, 8, 4, 0, 28, 24, 20, 16); + + for (uint i = 0; i < 8; i++) { + float o = float((occlusion >> occlusion_shift[i]) & 0xF) / 15.0; + imageStore(dst_occlusion[i], write_pos, vec4(o)); + } + +#endif + +#ifdef MODE_INITIALIZE_JUMP_FLOOD + + ivec3 pos = ivec3(gl_GlobalInvocationID.xyz); + + uint c = imageLoad(src_color, pos).r; + uvec4 v; + if (bool(c & 0x1)) { + //bit set means this is solid + v.xyz = uvec3(pos); + v.w = 255; //not zero means used + } else { + v.xyz = uvec3(0); + v.w = 0; // zero means unused + } + + imageStore(dst_positions, pos, v); +#endif + +#ifdef MODE_INITIALIZE_JUMP_FLOOD_HALF + + ivec3 pos = ivec3(gl_GlobalInvocationID.xyz); + ivec3 base_pos = pos * 2; + + //since we store in half size, lets kind of randomize what we store, so + //the half size jump flood has a bit better chance to find something + uvec4 closest[8]; + int closest_count = 0; + + for (uint i = 0; i < 8; i++) { + ivec3 src_pos = base_pos + ((ivec3(i) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1)); + uint c = imageLoad(src_color, src_pos).r; + if (bool(c & 1)) { + uvec4 v = uvec4(uvec3(src_pos), 255); + closest[closest_count] = v; + closest_count++; + } + } + + if (closest_count == 0) { + imageStore(dst_positions, pos, uvec4(0)); + } else { + ivec3 indexv = (pos & ivec3(1, 1, 1)) * ivec3(1, 2, 4); + int index = (indexv.x | indexv.y | indexv.z) % closest_count; + imageStore(dst_positions, pos, closest[index]); + } + +#endif + +#ifdef MODE_JUMPFLOOD + + //regular jumpflood, efficient for large steps, inefficient for small steps + ivec3 pos = ivec3(gl_GlobalInvocationID.xyz); + + vec3 posf = vec3(pos); + + if (params.half_size) { + posf = posf * 2.0 + 0.5; + } + + uvec4 p = imageLoad(src_positions, pos); + + if (!params.half_size && p == uvec4(uvec3(pos), 255)) { + imageStore(dst_positions, pos, p); + return; //points to itself and valid, nothing better can be done, just pass + } + + float p_dist; + + if (p.w != 0) { + p_dist = distance(posf, vec3(p.xyz)); + } else { + p_dist = 0.0; //should not matter + } + + const uint offset_count = 26; + const ivec3 offsets[offset_count] = ivec3[]( + ivec3(-1, -1, -1), + ivec3(-1, -1, 0), + ivec3(-1, -1, 1), + ivec3(-1, 0, -1), + ivec3(-1, 0, 0), + ivec3(-1, 0, 1), + ivec3(-1, 1, -1), + ivec3(-1, 1, 0), + ivec3(-1, 1, 1), + ivec3(0, -1, -1), + ivec3(0, -1, 0), + ivec3(0, -1, 1), + ivec3(0, 0, -1), + ivec3(0, 0, 1), + ivec3(0, 1, -1), + ivec3(0, 1, 0), + ivec3(0, 1, 1), + ivec3(1, -1, -1), + ivec3(1, -1, 0), + ivec3(1, -1, 1), + ivec3(1, 0, -1), + ivec3(1, 0, 0), + ivec3(1, 0, 1), + ivec3(1, 1, -1), + ivec3(1, 1, 0), + ivec3(1, 1, 1)); + + for (uint i = 0; i < offset_count; i++) { + ivec3 ofs = pos + offsets[i] * params.step_size; + if (any(lessThan(ofs, ivec3(0))) || any(greaterThanEqual(ofs, ivec3(params.grid_size)))) { + continue; + } + uvec4 q = imageLoad(src_positions, ofs); + + if (q.w == 0) { + continue; //was not initialized yet, ignore + } + + float q_dist = distance(posf, vec3(q.xyz)); + if (p.w == 0 || q_dist < p_dist) { + p = q; //just replace because current is unused + p_dist = q_dist; + } + } + + imageStore(dst_positions, pos, p); +#endif + +#ifdef MODE_JUMPFLOOD_OPTIMIZED + //optimized version using shared compute memory + + ivec3 group_offset = ivec3(gl_WorkGroupID.xyz) % params.step_size; + ivec3 group_pos = group_offset + (ivec3(gl_WorkGroupID.xyz) / params.step_size) * ivec3(GROUP_SIZE * params.step_size); + + //load data into local group memory + + if (all(lessThan(ivec3(gl_LocalInvocationID.xyz), ivec3((GROUP_SIZE + 2) / 2)))) { + //use this thread for loading, this method uses less threads for this but its simpler and less divergent + ivec3 base_pos = ivec3(gl_LocalInvocationID.xyz) * 2; + for (uint i = 0; i < 8; i++) { + ivec3 load_pos = base_pos + ((ivec3(i) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1)); + ivec3 load_global_pos = group_pos + (load_pos - ivec3(1)) * params.step_size; + uvec4 q; + if (all(greaterThanEqual(load_global_pos, ivec3(0))) && all(lessThan(load_global_pos, ivec3(params.grid_size)))) { + q = imageLoad(src_positions, load_global_pos); + } else { + q = uvec4(0); //unused + } + + group_store(load_pos, q); + } + } + + ivec3 global_pos = group_pos + ivec3(gl_LocalInvocationID.xyz) * params.step_size; + + if (any(lessThan(global_pos, ivec3(0))) || any(greaterThanEqual(global_pos, ivec3(params.grid_size)))) { + return; //do nothing else, end here because outside range + } + + //sync + groupMemoryBarrier(); + barrier(); + + ivec3 local_pos = ivec3(gl_LocalInvocationID.xyz) + ivec3(1); + + const uint offset_count = 27; + const ivec3 offsets[offset_count] = ivec3[]( + ivec3(-1, -1, -1), + ivec3(-1, -1, 0), + ivec3(-1, -1, 1), + ivec3(-1, 0, -1), + ivec3(-1, 0, 0), + ivec3(-1, 0, 1), + ivec3(-1, 1, -1), + ivec3(-1, 1, 0), + ivec3(-1, 1, 1), + ivec3(0, -1, -1), + ivec3(0, -1, 0), + ivec3(0, -1, 1), + ivec3(0, 0, -1), + ivec3(0, 0, 0), + ivec3(0, 0, 1), + ivec3(0, 1, -1), + ivec3(0, 1, 0), + ivec3(0, 1, 1), + ivec3(1, -1, -1), + ivec3(1, -1, 0), + ivec3(1, -1, 1), + ivec3(1, 0, -1), + ivec3(1, 0, 0), + ivec3(1, 0, 1), + ivec3(1, 1, -1), + ivec3(1, 1, 0), + ivec3(1, 1, 1)); + + //only makes sense if point is inside screen + uvec4 closest = uvec4(0); + float closest_dist = 0.0; + + vec3 posf = vec3(global_pos); + + if (params.half_size) { + posf = posf * 2.0 + 0.5; + } + + for (uint i = 0; i < offset_count; i++) { + uvec4 point = group_load(local_pos + offsets[i]); + + if (point.w == 0) { + continue; //was not initialized yet, ignore + } + + float dist = distance(posf, vec3(point.xyz)); + if (closest.w == 0 || dist < closest_dist) { + closest = point; + closest_dist = dist; + } + } + + imageStore(dst_positions, global_pos, closest); + +#endif + +#ifdef MODE_UPSCALE_JUMP_FLOOD + + ivec3 pos = ivec3(gl_GlobalInvocationID.xyz); + + uint c = imageLoad(src_color, pos).r; + uvec4 v; + if (bool(c & 1)) { + //bit set means this is solid + v.xyz = uvec3(pos); + v.w = 255; //not zero means used + } else { + v = imageLoad(src_positions_half, pos >> 1); + float d = length(vec3(ivec3(v.xyz) - pos)); + + ivec3 vbase = ivec3(v.xyz - (v.xyz & uvec3(1))); + + //search around if there is a better candidate from the same block + for (int i = 0; i < 8; i++) { + ivec3 bits = ((ivec3(i) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1)); + ivec3 p = vbase + bits; + + float d2 = length(vec3(p - pos)); + if (d2 < d) { //check valid distance before test so we avoid a read + uint c2 = imageLoad(src_color, p).r; + if (bool(c2 & 1)) { + v.xyz = uvec3(p); + d = d2; + } + } + } + + //could validate better position.. + } + + imageStore(dst_positions, pos, v); + +#endif + +#ifdef MODE_OCCLUSION + + uint invocation_idx = uint(gl_LocalInvocationID.x); + ivec3 region = ivec3(gl_WorkGroupID); + + ivec3 region_offset = -ivec3(OCCLUSION_SIZE); + region_offset += region * OCCLUSION_SIZE * 2; + region_offset += params.probe_offset * OCCLUSION_SIZE; + + if (params.scroll != ivec3(0)) { + //validate scroll region + ivec3 region_offset_to = region_offset + ivec3(OCCLUSION_SIZE * 2); + uvec3 scroll_mask = uvec3(notEqual(params.scroll, ivec3(0))); //save which axes acre scrolling + ivec3 scroll_from = mix(ivec3(0), ivec3(params.grid_size) + params.scroll, lessThan(params.scroll, ivec3(0))); + ivec3 scroll_to = mix(ivec3(params.grid_size), params.scroll, greaterThan(params.scroll, ivec3(0))); + + if ((uvec3(lessThanEqual(region_offset_to, scroll_from)) | uvec3(greaterThanEqual(region_offset, scroll_to))) * scroll_mask == scroll_mask) { //all axes that scroll are out, exit + return; //region outside scroll bounds, quit + } + } + +#define OCC_HALF_SIZE (OCCLUSION_SIZE / 2) + + ivec3 local_ofs = ivec3(uvec3(invocation_idx % OCC_HALF_SIZE, (invocation_idx % (OCC_HALF_SIZE * OCC_HALF_SIZE)) / OCC_HALF_SIZE, invocation_idx / (OCC_HALF_SIZE * OCC_HALF_SIZE))) * 4; + + /* for(int i=0;i<64;i++) { + ivec3 offset = region_offset + local_ofs + ((ivec3(i) >> ivec3(0,2,4)) & ivec3(3,3,3)); + uint facig = + if (all(greaterThanEqual(offset,ivec3(0))) && all(lessThan(offset,ivec3(params.grid_size)))) {*/ + + for (int i = 0; i < 16; i++) { //skip x, so it can be packed + + ivec3 offset = local_ofs + ((ivec3(i * 4) >> ivec3(0, 2, 4)) & ivec3(3, 3, 3)); + + uint facing_pack = 0; + for (int j = 0; j < 4; j++) { + ivec3 foffset = region_offset + offset + ivec3(j, 0, 0); + if (all(greaterThanEqual(foffset, ivec3(0))) && all(lessThan(foffset, ivec3(params.grid_size)))) { + uint f = imageLoad(src_facing, foffset).r; + facing_pack |= f << (j * 8); + } + } + + occlusion_facing[(offset.z * (OCCLUSION_SIZE * 2 * OCCLUSION_SIZE * 2) + offset.y * (OCCLUSION_SIZE * 2) + offset.x) / 4] = facing_pack; + } + + //sync occlusion saved + groupMemoryBarrier(); + barrier(); + + //process occlusion + +#define OCC_STEPS (OCCLUSION_SIZE * 3 - 2) +#define OCC_HALF_STEPS (OCC_STEPS / 2) + + for (int step = 0; step < OCC_STEPS; step++) { + bool shrink = step >= OCC_HALF_STEPS; + int occ_step = shrink ? OCC_HALF_STEPS - (step - OCC_HALF_STEPS) - 1 : step; + + if (invocation_idx < group_size_offset[occ_step].x) { + uint pv = group_pos[group_size_offset[occ_step].y + invocation_idx]; + ivec3 proc_abs = (ivec3(int(pv)) >> ivec3(0, 8, 16)) & ivec3(0xFF); + + if (shrink) { + proc_abs = ivec3(OCCLUSION_SIZE) - proc_abs - ivec3(1); + } + + for (int i = 0; i < 8; i++) { + ivec3 bits = ((ivec3(i) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1)); + ivec3 proc_sign = bits * 2 - 1; + ivec3 local_offset = ivec3(OCCLUSION_SIZE) + proc_abs * proc_sign - (ivec3(1) - bits); + ivec3 offset = local_offset + region_offset; + if (all(greaterThanEqual(offset, ivec3(0))) && all(lessThan(offset, ivec3(params.grid_size)))) { + float occ; + + uint facing = get_facing(local_offset); + + if (facing != 0) { //solid + occ = 0.0; + } else if (step == 0) { +#if 0 + occ = 0.0; + if (get_facing(local_offset - ivec3(proc_sign.x,0,0))==0) { + occ+=1.0; + } + if (get_facing(local_offset - ivec3(0,proc_sign.y,0))==0) { + occ+=1.0; + } + if (get_facing(local_offset - ivec3(0,0,proc_sign.z))==0) { + occ+=1.0; + } + /* + if (get_facing(local_offset - proc_sign)==0) { + occ+=1.0; + }*/ + + occ/=3.0; +#endif + occ = 1.0; + + } else { + ivec3 read_dir = -proc_sign; + + ivec3 major_axis; + if (proc_abs.x < proc_abs.y) { + if (proc_abs.z < proc_abs.y) { + major_axis = ivec3(0, 1, 0); + } else { + major_axis = ivec3(0, 0, 1); + } + } else { + if (proc_abs.z < proc_abs.x) { + major_axis = ivec3(1, 0, 0); + } else { + major_axis = ivec3(0, 0, 1); + } + } + + float avg = 0.0; + occ = 0.0; + + ivec3 read_x = offset + ivec3(read_dir.x, 0, 0) + (proc_abs.x == 0 ? major_axis * read_dir : ivec3(0)); + ivec3 read_y = offset + ivec3(0, read_dir.y, 0) + (proc_abs.y == 0 ? major_axis * read_dir : ivec3(0)); + ivec3 read_z = offset + ivec3(0, 0, read_dir.z) + (proc_abs.z == 0 ? major_axis * read_dir : ivec3(0)); + + uint facing_x = get_facing(read_x - region_offset); + if (facing_x == 0) { + if (all(greaterThanEqual(read_x, ivec3(0))) && all(lessThan(read_x, ivec3(params.grid_size)))) { + occ += imageLoad(dst_occlusion[params.occlusion_index], read_x).r; + avg += 1.0; + } + } else { + if (proc_abs.x != 0) { //do not occlude from voxels in the opposite octant + avg += 1.0; + } + } + + uint facing_y = get_facing(read_y - region_offset); + if (facing_y == 0) { + if (all(greaterThanEqual(read_y, ivec3(0))) && all(lessThan(read_y, ivec3(params.grid_size)))) { + occ += imageLoad(dst_occlusion[params.occlusion_index], read_y).r; + avg += 1.0; + } + } else { + if (proc_abs.y != 0) { + avg += 1.0; + } + } + + uint facing_z = get_facing(read_z - region_offset); + if (facing_z == 0) { + if (all(greaterThanEqual(read_z, ivec3(0))) && all(lessThan(read_z, ivec3(params.grid_size)))) { + occ += imageLoad(dst_occlusion[params.occlusion_index], read_z).r; + avg += 1.0; + } + } else { + if (proc_abs.z != 0) { + avg += 1.0; + } + } + + if (avg > 0.0) { + occ /= avg; + } + } + + imageStore(dst_occlusion[params.occlusion_index], offset, vec4(occ)); + } + } + } + + groupMemoryBarrier(); + barrier(); + } +#if 1 + //bias solid voxels away + + for (int i = 0; i < 64; i++) { + ivec3 local_offset = local_ofs + ((ivec3(i) >> ivec3(0, 2, 4)) & ivec3(3, 3, 3)); + ivec3 offset = region_offset + local_offset; + + if (all(greaterThanEqual(offset, ivec3(0))) && all(lessThan(offset, ivec3(params.grid_size)))) { + uint facing = get_facing(local_offset); + + if (facing != 0) { + //only work on solids + + ivec3 proc_pos = local_offset - ivec3(OCCLUSION_SIZE); + proc_pos += mix(ivec3(0), ivec3(1), greaterThanEqual(proc_pos, ivec3(0))); + + float avg = 0.0; + float occ = 0.0; + + ivec3 read_dir = -sign(proc_pos); + ivec3 read_dir_x = ivec3(read_dir.x, 0, 0); + ivec3 read_dir_y = ivec3(0, read_dir.y, 0); + ivec3 read_dir_z = ivec3(0, 0, read_dir.z); + //solid +#if 0 + + uvec3 facing_pos_base = (uvec3(facing) >> uvec3(0,1,2)) & uvec3(1,1,1); + uvec3 facing_neg_base = (uvec3(facing) >> uvec3(3,4,5)) & uvec3(1,1,1); + uvec3 facing_pos= facing_pos_base &((~facing_neg_base)&uvec3(1,1,1)); + uvec3 facing_neg= facing_neg_base &((~facing_pos_base)&uvec3(1,1,1)); +#else + uvec3 facing_pos = (uvec3(facing) >> uvec3(0, 1, 2)) & uvec3(1, 1, 1); + uvec3 facing_neg = (uvec3(facing) >> uvec3(3, 4, 5)) & uvec3(1, 1, 1); +#endif + bvec3 read_valid = bvec3(mix(facing_neg, facing_pos, greaterThan(read_dir, ivec3(0)))); + + //sides + if (read_valid.x) { + ivec3 read_offset = local_offset + read_dir_x; + uint f = get_facing(read_offset); + if (f == 0) { + read_offset += region_offset; + if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) { + occ += imageLoad(dst_occlusion[params.occlusion_index], read_offset).r; + avg += 1.0; + } + } + } + + if (read_valid.y) { + ivec3 read_offset = local_offset + read_dir_y; + uint f = get_facing(read_offset); + if (f == 0) { + read_offset += region_offset; + if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) { + occ += imageLoad(dst_occlusion[params.occlusion_index], read_offset).r; + avg += 1.0; + } + } + } + + if (read_valid.z) { + ivec3 read_offset = local_offset + read_dir_z; + uint f = get_facing(read_offset); + if (f == 0) { + read_offset += region_offset; + if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) { + occ += imageLoad(dst_occlusion[params.occlusion_index], read_offset).r; + avg += 1.0; + } + } + } + + //adjacents + + if (all(read_valid.yz)) { + ivec3 read_offset = local_offset + read_dir_y + read_dir_z; + uint f = get_facing(read_offset); + if (f == 0) { + read_offset += region_offset; + if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) { + occ += imageLoad(dst_occlusion[params.occlusion_index], read_offset).r; + avg += 1.0; + } + } + } + + if (all(read_valid.xz)) { + ivec3 read_offset = local_offset + read_dir_x + read_dir_z; + uint f = get_facing(read_offset); + if (f == 0) { + read_offset += region_offset; + if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) { + occ += imageLoad(dst_occlusion[params.occlusion_index], read_offset).r; + avg += 1.0; + } + } + } + + if (all(read_valid.xy)) { + ivec3 read_offset = local_offset + read_dir_x + read_dir_y; + uint f = get_facing(read_offset); + if (f == 0) { + read_offset += region_offset; + if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) { + occ += imageLoad(dst_occlusion[params.occlusion_index], read_offset).r; + avg += 1.0; + } + } + } + + //diagonal + + if (all(read_valid)) { + ivec3 read_offset = local_offset + read_dir; + uint f = get_facing(read_offset); + if (f == 0) { + read_offset += region_offset; + if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) { + occ += imageLoad(dst_occlusion[params.occlusion_index], read_offset).r; + avg += 1.0; + } + } + } + + if (avg > 0.0) { + occ /= avg; + } + + imageStore(dst_occlusion[params.occlusion_index], offset, vec4(occ)); + } + } + } + +#endif + +#if 1 + groupMemoryBarrier(); + barrier(); + + for (int i = 0; i < 64; i++) { + ivec3 local_offset = local_ofs + ((ivec3(i) >> ivec3(0, 2, 4)) & ivec3(3, 3, 3)); + ivec3 offset = region_offset + local_offset; + + if (all(greaterThanEqual(offset, ivec3(0))) && all(lessThan(offset, ivec3(params.grid_size)))) { + uint facing = get_facing(local_offset); + + if (facing == 0) { + ivec3 proc_pos = local_offset - ivec3(OCCLUSION_SIZE); + proc_pos += mix(ivec3(0), ivec3(1), greaterThanEqual(proc_pos, ivec3(0))); + + ivec3 proc_abs = abs(proc_pos); + + ivec3 read_dir = sign(proc_pos); //opposite direction + ivec3 read_dir_x = ivec3(read_dir.x, 0, 0); + ivec3 read_dir_y = ivec3(0, read_dir.y, 0); + ivec3 read_dir_z = ivec3(0, 0, read_dir.z); + //solid + uvec3 read_mask = mix(uvec3(1, 2, 4), uvec3(8, 16, 32), greaterThan(read_dir, ivec3(0))); //match positive with negative normals + uvec3 block_mask = mix(uvec3(1, 2, 4), uvec3(8, 16, 32), lessThan(read_dir, ivec3(0))); //match positive with negative normals + + block_mask = uvec3(0); + + float visible = 0.0; + float occlude_total = 0.0; + + if (proc_abs.x < OCCLUSION_SIZE) { + ivec3 read_offset = local_offset + read_dir_x; + uint x_mask = get_facing(read_offset); + if (x_mask != 0) { + read_offset += region_offset; + if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) { + occlude_total += 1.0; + if (bool(x_mask & read_mask.x) && !bool(x_mask & block_mask.x)) { + visible += 1.0; + } + } + } + } + + if (proc_abs.y < OCCLUSION_SIZE) { + ivec3 read_offset = local_offset + read_dir_y; + uint y_mask = get_facing(read_offset); + if (y_mask != 0) { + read_offset += region_offset; + if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) { + occlude_total += 1.0; + if (bool(y_mask & read_mask.y) && !bool(y_mask & block_mask.y)) { + visible += 1.0; + } + } + } + } + + if (proc_abs.z < OCCLUSION_SIZE) { + ivec3 read_offset = local_offset + read_dir_z; + uint z_mask = get_facing(read_offset); + if (z_mask != 0) { + read_offset += region_offset; + if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) { + occlude_total += 1.0; + if (bool(z_mask & read_mask.z) && !bool(z_mask & block_mask.z)) { + visible += 1.0; + } + } + } + } + + //if near the cartesian plane, test in opposite direction too + + read_mask = mix(uvec3(1, 2, 4), uvec3(8, 16, 32), lessThan(read_dir, ivec3(0))); //match negative with positive normals + block_mask = mix(uvec3(1, 2, 4), uvec3(8, 16, 32), greaterThan(read_dir, ivec3(0))); //match negative with positive normals + block_mask = uvec3(0); + + if (proc_abs.x == 1) { + ivec3 read_offset = local_offset - read_dir_x; + uint x_mask = get_facing(read_offset); + if (x_mask != 0) { + read_offset += region_offset; + if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) { + occlude_total += 1.0; + if (bool(x_mask & read_mask.x) && !bool(x_mask & block_mask.x)) { + visible += 1.0; + } + } + } + } + + if (proc_abs.y == 1) { + ivec3 read_offset = local_offset - read_dir_y; + uint y_mask = get_facing(read_offset); + if (y_mask != 0) { + read_offset += region_offset; + if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) { + occlude_total += 1.0; + if (bool(y_mask & read_mask.y) && !bool(y_mask & block_mask.y)) { + visible += 1.0; + } + } + } + } + + if (proc_abs.z == 1) { + ivec3 read_offset = local_offset - read_dir_z; + uint z_mask = get_facing(read_offset); + if (z_mask != 0) { + read_offset += region_offset; + if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) { + occlude_total += 1.0; + if (bool(z_mask & read_mask.z) && !bool(z_mask & block_mask.z)) { + visible += 1.0; + } + } + } + } + + if (occlude_total > 0.0) { + float occ = imageLoad(dst_occlusion[params.occlusion_index], offset).r; + occ *= visible / occlude_total; + imageStore(dst_occlusion[params.occlusion_index], offset, vec4(occ)); + } + } + } + } + +#endif + + /* + for(int i=0;i<8;i++) { + ivec3 local_offset = local_pos + ((ivec3(i) >> ivec3(2,1,0)) & ivec3(1,1,1)) * OCCLUSION_SIZE; + ivec3 offset = local_offset - ivec3(OCCLUSION_SIZE); //looking around probe, so starts negative + offset += region * OCCLUSION_SIZE * 2; //offset by region + offset += params.probe_offset * OCCLUSION_SIZE; // offset by probe offset + if (all(greaterThanEqual(offset,ivec3(0))) && all(lessThan(offset,ivec3(params.grid_size)))) { + imageStore(dst_occlusion[params.occlusion_index],offset,vec4( occlusion_data[ to_linear(local_offset) ] )); + //imageStore(dst_occlusion[params.occlusion_index],offset,vec4( occlusion_solid[ to_linear(local_offset) ] )); + } + } +*/ + +#endif + +#ifdef MODE_STORE + + ivec3 local = ivec3(gl_LocalInvocationID.xyz); + ivec3 pos = ivec3(gl_GlobalInvocationID.xyz); + // store SDF + uvec4 p = imageLoad(src_positions, pos); + + bool solid = false; + float d; + if (ivec3(p.xyz) == pos) { + //solid block + d = 0; + solid = true; + } else { + //distance block + d = 1.0 + length(vec3(p.xyz) - vec3(pos)); + } + + d /= 255.0; + + imageStore(dst_sdf, pos, vec4(d)); + + // STORE OCCLUSION + + uint occlusion = 0; + const uint occlusion_shift[8] = uint[](12, 8, 4, 0, 28, 24, 20, 16); + for (int i = 0; i < 8; i++) { + float occ = imageLoad(src_occlusion[i], pos).r; + occlusion |= uint(clamp(occ * 15.0, 0.0, 15.0)) << occlusion_shift[i]; + } + { + ivec3 occ_pos = pos; + occ_pos.z += params.cascade * params.grid_size; + imageStore(dst_occlusion, occ_pos, uvec4(occlusion & 0xFFFF)); + occ_pos.x += params.grid_size; + imageStore(dst_occlusion, occ_pos, uvec4(occlusion >> 16)); + } + + // STORE POSITIONS + + if (local == ivec3(0)) { + store_position_count = 0; //base one stores as zero, the others wait + } + + groupMemoryBarrier(); + barrier(); + + if (solid) { + uint index = atomicAdd(store_position_count, 1); + // At least do the conversion work in parallel + store_positions[index].position = uint(pos.x | (pos.y << 7) | (pos.z << 14)); + + //see around which voxels point to this one, add them to the list + uint bit_index = 0; + uint neighbour_bits = 0; + for (int i = -1; i <= 1; i++) { + for (int j = -1; j <= 1; j++) { + for (int k = -1; k <= 1; k++) { + if (i == 0 && j == 0 && k == 0) { + continue; + } + ivec3 npos = pos + ivec3(i, j, k); + if (all(greaterThanEqual(npos, ivec3(0))) && all(lessThan(npos, ivec3(params.grid_size)))) { + p = imageLoad(src_positions, npos); + if (ivec3(p.xyz) == pos) { + neighbour_bits |= (1 << bit_index); + } + } + bit_index++; + } + } + } + + uint rgb = imageLoad(src_albedo, pos).r; + uint facing = imageLoad(src_facing, pos).r; + + store_positions[index].albedo = rgb >> 1; //store as it comes (555) to avoid precision loss (and move away the alpha bit) + store_positions[index].albedo |= (facing & 0x3F) << 15; // store facing in bits 15-21 + + store_positions[index].albedo |= neighbour_bits << 21; //store lower 11 bits of neighbours with remaining albedo + store_positions[index].position |= (neighbour_bits >> 11) << 21; //store 11 bits more of neighbours with position + + store_positions[index].light = imageLoad(src_light, pos).r; + store_positions[index].light_aniso = imageLoad(src_light_aniso, pos).r; + //add neighbours + store_positions[index].light |= (neighbour_bits >> 22) << 30; //store 2 bits more of neighbours with light + store_positions[index].light_aniso |= (neighbour_bits >> 24) << 30; //store 2 bits more of neighbours with aniso + } + + groupMemoryBarrier(); + barrier(); + + // global increment only once per group, to reduce pressure + + if (local == ivec3(0) && store_position_count > 0) { + store_from_index = atomicAdd(dispatch_data.total_count, store_position_count); + uint group_count = (store_from_index + store_position_count - 1) / 64 + 1; + atomicMax(dispatch_data.x, group_count); + } + + groupMemoryBarrier(); + barrier(); + + uint read_index = uint(local.z * 4 * 4 + local.y * 4 + local.x); + uint write_index = store_from_index + read_index; + + if (read_index < store_position_count) { + dst_process_voxels.data[write_index] = store_positions[read_index]; + } + + if (pos == ivec3(0)) { + //this thread clears y and z + dispatch_data.y = 1; + dispatch_data.z = 1; + } +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/environment/sky.glsl b/servers/rendering/renderer_rd/shaders/environment/sky.glsl new file mode 100644 index 0000000000..d523461600 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/environment/sky.glsl @@ -0,0 +1,268 @@ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#define MAX_VIEWS 2 + +#if defined(USE_MULTIVIEW) && defined(has_VK_KHR_multiview) +#extension GL_EXT_multiview : enable +#endif + +layout(location = 0) out vec2 uv_interp; + +layout(push_constant, std430) uniform Params { + mat3 orientation; + vec4 projections[MAX_VIEWS]; + vec3 position; + float time; + vec3 pad; + float luminance_multiplier; +} +params; + +void main() { + vec2 base_arr[4] = vec2[](vec2(-1.0, -1.0), vec2(-1.0, 1.0), vec2(1.0, 1.0), vec2(1.0, -1.0)); + uv_interp = base_arr[gl_VertexIndex]; + gl_Position = vec4(uv_interp, 1.0, 1.0); +} + +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#ifdef USE_MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +// !BAS! This needs to become an input once we implement our fallback! +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#else // USE_MULTIVIEW +// Set to zero, not supported in non stereo +#define ViewIndex 0 +#endif //USE_MULTIVIEW + +#define M_PI 3.14159265359 +#define MAX_VIEWS 2 + +layout(location = 0) in vec2 uv_interp; + +layout(push_constant, std430) uniform Params { + mat3 orientation; + vec4 projections[MAX_VIEWS]; + vec3 position; + float time; + vec3 pad; + float luminance_multiplier; +} +params; + +#define SAMPLER_NEAREST_CLAMP 0 +#define SAMPLER_LINEAR_CLAMP 1 +#define SAMPLER_NEAREST_WITH_MIPMAPS_CLAMP 2 +#define SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP 3 +#define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_CLAMP 4 +#define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_CLAMP 5 +#define SAMPLER_NEAREST_REPEAT 6 +#define SAMPLER_LINEAR_REPEAT 7 +#define SAMPLER_NEAREST_WITH_MIPMAPS_REPEAT 8 +#define SAMPLER_LINEAR_WITH_MIPMAPS_REPEAT 9 +#define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_REPEAT 10 +#define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_REPEAT 11 + +layout(set = 0, binding = 0) uniform sampler material_samplers[12]; + +layout(set = 0, binding = 1, std430) restrict readonly buffer GlobalShaderUniformData { + vec4 data[]; +} +global_shader_uniforms; + +layout(set = 0, binding = 2, std140) uniform SceneData { + bool volumetric_fog_enabled; // 4 - 4 + float volumetric_fog_inv_length; // 4 - 8 + float volumetric_fog_detail_spread; // 4 - 12 + float volumetric_fog_sky_affect; // 4 - 16 + + bool fog_enabled; // 4 - 20 + float fog_sky_affect; // 4 - 24 + float fog_density; // 4 - 28 + float fog_sun_scatter; // 4 - 32 + + vec3 fog_light_color; // 12 - 44 + float fog_aerial_perspective; // 4 - 48 + + float z_far; // 4 - 52 + uint directional_light_count; // 4 - 56 + uint pad1; // 4 - 60 + uint pad2; // 4 - 64 +} +scene_data; + +struct DirectionalLightData { + vec4 direction_energy; + vec4 color_size; + bool enabled; +}; + +layout(set = 0, binding = 3, std140) uniform DirectionalLights { + DirectionalLightData data[MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS]; +} +directional_lights; + +#ifdef MATERIAL_UNIFORMS_USED +layout(set = 1, binding = 0, std140) uniform MaterialUniforms{ +#MATERIAL_UNIFORMS +} material; +#endif + +layout(set = 2, binding = 0) uniform textureCube radiance; +#ifdef USE_CUBEMAP_PASS +layout(set = 2, binding = 1) uniform textureCube half_res; +layout(set = 2, binding = 2) uniform textureCube quarter_res; +#else +layout(set = 2, binding = 1) uniform texture2D half_res; +layout(set = 2, binding = 2) uniform texture2D quarter_res; +#endif + +layout(set = 3, binding = 0) uniform texture3D volumetric_fog_texture; + +#ifdef USE_CUBEMAP_PASS +#define AT_CUBEMAP_PASS true +#else +#define AT_CUBEMAP_PASS false +#endif + +#ifdef USE_HALF_RES_PASS +#define AT_HALF_RES_PASS true +#else +#define AT_HALF_RES_PASS false +#endif + +#ifdef USE_QUARTER_RES_PASS +#define AT_QUARTER_RES_PASS true +#else +#define AT_QUARTER_RES_PASS false +#endif + +#GLOBALS + +layout(location = 0) out vec4 frag_color; + +#ifdef USE_DEBANDING +// https://www.iryoku.com/next-generation-post-processing-in-call-of-duty-advanced-warfare +vec3 interleaved_gradient_noise(vec2 pos) { + const vec3 magic = vec3(0.06711056f, 0.00583715f, 52.9829189f); + float res = fract(magic.z * fract(dot(pos, magic.xy))) * 2.0 - 1.0; + return vec3(res, -res, res) / 255.0; +} +#endif + +vec4 volumetric_fog_process(vec2 screen_uv) { + vec3 fog_pos = vec3(screen_uv, 1.0); + + return texture(sampler3D(volumetric_fog_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), fog_pos); +} + +vec4 fog_process(vec3 view, vec3 sky_color) { + vec3 fog_color = mix(scene_data.fog_light_color, sky_color, scene_data.fog_aerial_perspective); + + if (scene_data.fog_sun_scatter > 0.001) { + vec4 sun_scatter = vec4(0.0); + float sun_total = 0.0; + for (uint i = 0; i < scene_data.directional_light_count; i++) { + vec3 light_color = directional_lights.data[i].color_size.xyz * directional_lights.data[i].direction_energy.w; + float light_amount = pow(max(dot(view, directional_lights.data[i].direction_energy.xyz), 0.0), 8.0); + fog_color += light_color * light_amount * scene_data.fog_sun_scatter; + } + } + + return vec4(fog_color, 1.0); +} + +void main() { + vec3 cube_normal; + cube_normal.z = -1.0; + cube_normal.x = (cube_normal.z * (-uv_interp.x - params.projections[ViewIndex].x)) / params.projections[ViewIndex].y; + cube_normal.y = -(cube_normal.z * (-uv_interp.y - params.projections[ViewIndex].z)) / params.projections[ViewIndex].w; + cube_normal = mat3(params.orientation) * cube_normal; + cube_normal = normalize(cube_normal); + + vec2 uv = uv_interp * 0.5 + 0.5; + + vec2 panorama_coords = vec2(atan(cube_normal.x, -cube_normal.z), acos(cube_normal.y)); + + if (panorama_coords.x < 0.0) { + panorama_coords.x += M_PI * 2.0; + } + + panorama_coords /= vec2(M_PI * 2.0, M_PI); + + vec3 color = vec3(0.0, 0.0, 0.0); + float alpha = 1.0; // Only available to subpasses + vec4 half_res_color = vec4(1.0); + vec4 quarter_res_color = vec4(1.0); + vec4 custom_fog = vec4(0.0); + +#ifdef USE_CUBEMAP_PASS +#ifdef USES_HALF_RES_COLOR + half_res_color = texture(samplerCube(half_res, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), cube_normal) / params.luminance_multiplier; +#endif +#ifdef USES_QUARTER_RES_COLOR + quarter_res_color = texture(samplerCube(quarter_res, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), cube_normal) / params.luminance_multiplier; +#endif +#else +#ifdef USES_HALF_RES_COLOR + half_res_color = textureLod(sampler2D(half_res, material_samplers[SAMPLER_LINEAR_CLAMP]), uv, 0.0) / params.luminance_multiplier; +#endif +#ifdef USES_QUARTER_RES_COLOR + quarter_res_color = textureLod(sampler2D(quarter_res, material_samplers[SAMPLER_LINEAR_CLAMP]), uv, 0.0) / params.luminance_multiplier; +#endif +#endif + + { + +#CODE : SKY + + } + + frag_color.rgb = color; + frag_color.a = alpha; + +#if !defined(DISABLE_FOG) && !defined(USE_CUBEMAP_PASS) + + // Draw "fixed" fog before volumetric fog to ensure volumetric fog can appear in front of the sky. + if (scene_data.fog_enabled) { + vec4 fog = fog_process(cube_normal, frag_color.rgb); + frag_color.rgb = mix(frag_color.rgb, fog.rgb, fog.a * scene_data.fog_sky_affect); + } + + if (scene_data.volumetric_fog_enabled) { + vec4 fog = volumetric_fog_process(uv); + frag_color.rgb = mix(frag_color.rgb, fog.rgb, fog.a * scene_data.volumetric_fog_sky_affect); + } + + if (custom_fog.a > 0.0) { + frag_color.rgb = mix(frag_color.rgb, custom_fog.rgb, custom_fog.a); + } + +#endif // DISABLE_FOG + + // Blending is disabled for Sky, so alpha doesn't blend. + // Alpha is used for subsurface scattering so make sure it doesn't get applied to Sky. + if (!AT_CUBEMAP_PASS && !AT_HALF_RES_PASS && !AT_QUARTER_RES_PASS) { + frag_color.a = 0.0; + } + + // For mobile renderer we're multiplying by 0.5 as we're using a UNORM buffer. + // For both mobile and clustered, we also bake in the exposure value for the environment and camera. + frag_color.rgb = frag_color.rgb * params.luminance_multiplier; + +#ifdef USE_DEBANDING + frag_color.rgb += interleaved_gradient_noise(gl_FragCoord.xy); +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/environment/volumetric_fog.glsl b/servers/rendering/renderer_rd/shaders/environment/volumetric_fog.glsl new file mode 100644 index 0000000000..4658afd02d --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/environment/volumetric_fog.glsl @@ -0,0 +1,309 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 4, local_size_y = 4, local_size_z = 4) in; + +#define SAMPLER_NEAREST_CLAMP 0 +#define SAMPLER_LINEAR_CLAMP 1 +#define SAMPLER_NEAREST_WITH_MIPMAPS_CLAMP 2 +#define SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP 3 +#define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_CLAMP 4 +#define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_CLAMP 5 +#define SAMPLER_NEAREST_REPEAT 6 +#define SAMPLER_LINEAR_REPEAT 7 +#define SAMPLER_NEAREST_WITH_MIPMAPS_REPEAT 8 +#define SAMPLER_LINEAR_WITH_MIPMAPS_REPEAT 9 +#define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_REPEAT 10 +#define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_REPEAT 11 + +#define DENSITY_SCALE 1024.0 + +#include "../cluster_data_inc.glsl" +#include "../light_data_inc.glsl" + +#define M_PI 3.14159265359 + +layout(set = 0, binding = 1) uniform sampler material_samplers[12]; + +layout(set = 0, binding = 2, std430) restrict readonly buffer GlobalShaderUniformData { + vec4 data[]; +} +global_shader_uniforms; + +layout(push_constant, std430) uniform Params { + vec3 position; + float pad; + + vec3 extents; + float pad2; + + ivec3 corner; + uint shape; + + mat4 transform; +} +params; + +#ifdef MOLTENVK_USED +layout(set = 1, binding = 1) volatile buffer emissive_only_map_buffer { + uint emissive_only_map[]; +}; +#else +layout(r32ui, set = 1, binding = 1) uniform volatile uimage3D emissive_only_map; +#endif + +layout(set = 1, binding = 2, std140) uniform SceneParams { + vec2 fog_frustum_size_begin; + vec2 fog_frustum_size_end; + + float fog_frustum_end; + float z_near; // + float z_far; // + float time; + + ivec3 fog_volume_size; + uint directional_light_count; // + + bool use_temporal_reprojection; + uint temporal_frame; + float detail_spread; + float temporal_blend; + + mat4 to_prev_view; + mat4 transform; +} +scene_params; + +#ifdef MOLTENVK_USED +layout(set = 1, binding = 3) volatile buffer density_only_map_buffer { + uint density_only_map[]; +}; +layout(set = 1, binding = 4) volatile buffer light_only_map_buffer { + uint light_only_map[]; +}; +#else +layout(r32ui, set = 1, binding = 3) uniform volatile uimage3D density_only_map; +layout(r32ui, set = 1, binding = 4) uniform volatile uimage3D light_only_map; +#endif + +#ifdef MATERIAL_UNIFORMS_USED +layout(set = 2, binding = 0, std140) uniform MaterialUniforms{ +#MATERIAL_UNIFORMS +} material; +#endif + +#GLOBALS + +float get_depth_at_pos(float cell_depth_size, int z) { + float d = float(z) * cell_depth_size + cell_depth_size * 0.5; //center of voxels + d = pow(d, scene_params.detail_spread); + return scene_params.fog_frustum_end * d; +} + +#define TEMPORAL_FRAMES 16 + +const vec3 halton_map[TEMPORAL_FRAMES] = vec3[]( + vec3(0.5, 0.33333333, 0.2), + vec3(0.25, 0.66666667, 0.4), + vec3(0.75, 0.11111111, 0.6), + vec3(0.125, 0.44444444, 0.8), + vec3(0.625, 0.77777778, 0.04), + vec3(0.375, 0.22222222, 0.24), + vec3(0.875, 0.55555556, 0.44), + vec3(0.0625, 0.88888889, 0.64), + vec3(0.5625, 0.03703704, 0.84), + vec3(0.3125, 0.37037037, 0.08), + vec3(0.8125, 0.7037037, 0.28), + vec3(0.1875, 0.14814815, 0.48), + vec3(0.6875, 0.48148148, 0.68), + vec3(0.4375, 0.81481481, 0.88), + vec3(0.9375, 0.25925926, 0.12), + vec3(0.03125, 0.59259259, 0.32)); + +void main() { + vec3 fog_cell_size = 1.0 / vec3(scene_params.fog_volume_size); + + ivec3 pos = ivec3(gl_GlobalInvocationID.xyz) + params.corner; + if (any(greaterThanEqual(pos, scene_params.fog_volume_size))) { + return; //do not compute + } +#ifdef MOLTENVK_USED + uint lpos = pos.z * scene_params.fog_volume_size.x * scene_params.fog_volume_size.y + pos.y * scene_params.fog_volume_size.x + pos.x; +#endif + + vec3 posf = vec3(pos); + + vec3 fog_unit_pos = posf * fog_cell_size + fog_cell_size * 0.5; //center of voxels + fog_unit_pos.z = pow(fog_unit_pos.z, scene_params.detail_spread); + + vec3 view_pos; + view_pos.xy = (fog_unit_pos.xy * 2.0 - 1.0) * mix(scene_params.fog_frustum_size_begin, scene_params.fog_frustum_size_end, vec2(fog_unit_pos.z)); + view_pos.z = -scene_params.fog_frustum_end * fog_unit_pos.z; + view_pos.y = -view_pos.y; + + if (scene_params.use_temporal_reprojection) { + vec3 prev_view = (scene_params.to_prev_view * vec4(view_pos, 1.0)).xyz; + //undo transform into prev view + prev_view.y = -prev_view.y; + //z back to unit size + prev_view.z /= -scene_params.fog_frustum_end; + //xy back to unit size + prev_view.xy /= mix(scene_params.fog_frustum_size_begin, scene_params.fog_frustum_size_end, vec2(prev_view.z)); + prev_view.xy = prev_view.xy * 0.5 + 0.5; + //z back to unspread value + prev_view.z = pow(prev_view.z, 1.0 / scene_params.detail_spread); + + if (all(greaterThan(prev_view, vec3(0.0))) && all(lessThan(prev_view, vec3(1.0)))) { + //reprojectinon fits + // Since we can reproject, now we must jitter the current view pos. + // This is done here because cells that can't reproject should not jitter. + + fog_unit_pos = posf * fog_cell_size + fog_cell_size * halton_map[scene_params.temporal_frame]; //center of voxels, offset by halton table + fog_unit_pos.z = pow(fog_unit_pos.z, scene_params.detail_spread); + + view_pos.xy = (fog_unit_pos.xy * 2.0 - 1.0) * mix(scene_params.fog_frustum_size_begin, scene_params.fog_frustum_size_end, vec2(fog_unit_pos.z)); + view_pos.z = -scene_params.fog_frustum_end * fog_unit_pos.z; + view_pos.y = -view_pos.y; + } + } + + float density = 0.0; + vec3 emission = vec3(0.0); + vec3 albedo = vec3(0.0); + + float cell_depth_size = abs(view_pos.z - get_depth_at_pos(fog_cell_size.z, pos.z + 1)); + + vec4 world = scene_params.transform * vec4(view_pos, 1.0); + world.xyz /= world.w; + + vec3 uvw = fog_unit_pos; + + vec4 local_pos = params.transform * world; + local_pos.xyz /= local_pos.w; + + float sdf = -1.0; + if (params.shape == 0) { + // Ellipsoid + // https://www.shadertoy.com/view/tdS3DG + float k0 = length(local_pos.xyz / params.extents); + float k1 = length(local_pos.xyz / (params.extents * params.extents)); + sdf = k0 * (k0 - 1.0) / k1; + } else if (params.shape == 1) { + // Cone + // https://iquilezles.org/www/articles/distfunctions/distfunctions.htm + + // Compute the cone angle automatically to fit within the volume's extents. + float inv_height = 1.0 / max(0.001, params.extents.y); + float radius = 1.0 / max(0.001, (min(params.extents.x, params.extents.z) * 0.5)); + float hypotenuse = sqrt(radius * radius + inv_height * inv_height); + float rsin = radius / hypotenuse; + float rcos = inv_height / hypotenuse; + vec2 c = vec2(rsin, rcos); + + float q = length(local_pos.xz); + sdf = max(dot(c, vec2(q, local_pos.y - params.extents.y)), -params.extents.y - local_pos.y); + } else if (params.shape == 2) { + // Cylinder + // https://iquilezles.org/www/articles/distfunctions/distfunctions.htm + vec2 d = abs(vec2(length(local_pos.xz), local_pos.y)) - vec2(min(params.extents.x, params.extents.z), params.extents.y); + sdf = min(max(d.x, d.y), 0.0) + length(max(d, 0.0)); + } else if (params.shape == 3) { + // Box + // https://iquilezles.org/www/articles/distfunctions/distfunctions.htm + vec3 q = abs(local_pos.xyz) - params.extents; + sdf = length(max(q, 0.0)) + min(max(q.x, max(q.y, q.z)), 0.0); + } + + float cull_mask = 1.0; //used to cull cells that do not contribute + if (params.shape <= 3) { +#ifndef SDF_USED + cull_mask = 1.0 - smoothstep(-0.1, 0.0, sdf); +#endif + uvw = clamp((local_pos.xyz + params.extents) / (2.0 * params.extents), 0.0, 1.0); + } + + if (cull_mask > 0.0) { + { +#CODE : FOG + } + +#ifdef DENSITY_USED + density *= cull_mask; + if (abs(density) > 0.001) { + int final_density = int(density * DENSITY_SCALE); +#ifdef MOLTENVK_USED + atomicAdd(density_only_map[lpos], uint(final_density)); +#else + imageAtomicAdd(density_only_map, pos, uint(final_density)); +#endif + +#ifdef EMISSION_USED + { + emission *= clamp(density, 0.0, 1.0); + emission = clamp(emission, vec3(0.0), vec3(4.0)); + // Scale to fit into R11G11B10 with a range of 0-4 + uvec3 emission_u = uvec3(emission.r * 511.0, emission.g * 511.0, emission.b * 255.0); + // R and G have 11 bits each and B has 10. Then pack them into a 32 bit uint + uint final_emission = emission_u.r << 21 | emission_u.g << 10 | emission_u.b; +#ifdef MOLTENVK_USED + uint prev_emission = atomicAdd(emissive_only_map[lpos], final_emission); +#else + uint prev_emission = imageAtomicAdd(emissive_only_map, pos, final_emission); +#endif + + // Adding can lead to colors overflowing, so validate + uvec3 prev_emission_u = uvec3(prev_emission >> 21, (prev_emission << 11) >> 21, prev_emission % 1024); + uint add_emission = final_emission + prev_emission; + uvec3 add_emission_u = uvec3(add_emission >> 21, (add_emission << 11) >> 21, add_emission % 1024); + + bvec3 overflowing = lessThan(add_emission_u, prev_emission_u + emission_u); + + if (any(overflowing)) { + uvec3 overflow_factor = mix(uvec3(0), uvec3(2047 << 21, 2047 << 10, 1023), overflowing); + uint force_max = overflow_factor.r | overflow_factor.g | overflow_factor.b; +#ifdef MOLTENVK_USED + atomicOr(emissive_only_map[lpos], force_max); +#else + imageAtomicOr(emissive_only_map, pos, force_max); +#endif + } + } +#endif +#ifdef ALBEDO_USED + { + vec3 scattering = albedo * clamp(density, 0.0, 1.0); + scattering = clamp(scattering, vec3(0.0), vec3(1.0)); + uvec3 scattering_u = uvec3(scattering.r * 2047.0, scattering.g * 2047.0, scattering.b * 1023.0); + // R and G have 11 bits each and B has 10. Then pack them into a 32 bit uint + uint final_scattering = scattering_u.r << 21 | scattering_u.g << 10 | scattering_u.b; +#ifdef MOLTENVK_USED + uint prev_scattering = atomicAdd(light_only_map[lpos], final_scattering); +#else + uint prev_scattering = imageAtomicAdd(light_only_map, pos, final_scattering); +#endif + + // Adding can lead to colors overflowing, so validate + uvec3 prev_scattering_u = uvec3(prev_scattering >> 21, (prev_scattering << 11) >> 21, prev_scattering % 1024); + uint add_scattering = final_scattering + prev_scattering; + uvec3 add_scattering_u = uvec3(add_scattering >> 21, (add_scattering << 11) >> 21, add_scattering % 1024); + + bvec3 overflowing = lessThan(add_scattering_u, prev_scattering_u + scattering_u); + + if (any(overflowing)) { + uvec3 overflow_factor = mix(uvec3(0), uvec3(2047 << 21, 2047 << 10, 1023), overflowing); + uint force_max = overflow_factor.r | overflow_factor.g | overflow_factor.b; +#ifdef MOLTENVK_USED + atomicOr(light_only_map[lpos], force_max); +#else + imageAtomicOr(light_only_map, pos, force_max); +#endif + } + } +#endif // ALBEDO_USED + } +#endif // DENSITY_USED + } +} diff --git a/servers/rendering/renderer_rd/shaders/environment/volumetric_fog_process.glsl b/servers/rendering/renderer_rd/shaders/environment/volumetric_fog_process.glsl new file mode 100644 index 0000000000..28507e6c12 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/environment/volumetric_fog_process.glsl @@ -0,0 +1,792 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +/* Do not use subgroups here, seems there is not much advantage and causes glitches +#if defined(has_GL_KHR_shader_subgroup_ballot) && defined(has_GL_KHR_shader_subgroup_arithmetic) +#extension GL_KHR_shader_subgroup_ballot: enable +#extension GL_KHR_shader_subgroup_arithmetic: enable + +#define USE_SUBGROUPS +#endif +*/ + +#ifdef MODE_DENSITY +layout(local_size_x = 4, local_size_y = 4, local_size_z = 4) in; +#else +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; +#endif + +#include "../cluster_data_inc.glsl" +#include "../light_data_inc.glsl" + +#define M_PI 3.14159265359 + +#define DENSITY_SCALE 1024.0 + +layout(set = 0, binding = 1) uniform texture2D shadow_atlas; +layout(set = 0, binding = 2) uniform texture2D directional_shadow_atlas; + +layout(set = 0, binding = 3, std430) restrict readonly buffer OmniLights { + LightData data[]; +} +omni_lights; + +layout(set = 0, binding = 4, std430) restrict readonly buffer SpotLights { + LightData data[]; +} +spot_lights; + +layout(set = 0, binding = 5, std140) uniform DirectionalLights { + DirectionalLightData data[MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS]; +} +directional_lights; + +layout(set = 0, binding = 6, std430) buffer restrict readonly ClusterBuffer { + uint data[]; +} +cluster_buffer; + +layout(set = 0, binding = 7) uniform sampler linear_sampler; + +#ifdef MODE_DENSITY +layout(rgba16f, set = 0, binding = 8) uniform restrict writeonly image3D density_map; +#endif + +#ifdef MODE_FOG +layout(rgba16f, set = 0, binding = 8) uniform restrict readonly image3D density_map; +layout(rgba16f, set = 0, binding = 9) uniform restrict writeonly image3D fog_map; +#endif + +#ifdef MODE_COPY +layout(rgba16f, set = 0, binding = 8) uniform restrict readonly image3D source_map; +layout(rgba16f, set = 0, binding = 9) uniform restrict writeonly image3D dest_map; +#endif + +#ifdef MODE_FILTER +layout(rgba16f, set = 0, binding = 8) uniform restrict readonly image3D source_map; +layout(rgba16f, set = 0, binding = 9) uniform restrict writeonly image3D dest_map; +#endif + +layout(set = 0, binding = 10) uniform sampler shadow_sampler; + +#define MAX_VOXEL_GI_INSTANCES 8 + +struct VoxelGIData { + mat4 xform; // 64 - 64 + + vec3 bounds; // 12 - 76 + float dynamic_range; // 4 - 80 + + float bias; // 4 - 84 + float normal_bias; // 4 - 88 + bool blend_ambient; // 4 - 92 + uint mipmaps; // 4 - 96 + + vec3 pad; // 12 - 108 + float exposure_normalization; // 4 - 112 +}; + +layout(set = 0, binding = 11, std140) uniform VoxelGIs { + VoxelGIData data[MAX_VOXEL_GI_INSTANCES]; +} +voxel_gi_instances; + +layout(set = 0, binding = 12) uniform texture3D voxel_gi_textures[MAX_VOXEL_GI_INSTANCES]; + +layout(set = 0, binding = 13) uniform sampler linear_sampler_with_mipmaps; + +#ifdef ENABLE_SDFGI + +// SDFGI Integration on set 1 +#define SDFGI_MAX_CASCADES 8 + +struct SDFVoxelGICascadeData { + vec3 position; + float to_probe; + ivec3 probe_world_offset; + float to_cell; // 1/bounds * grid_size + vec3 pad; + float exposure_normalization; +}; + +layout(set = 1, binding = 0, std140) uniform SDFGI { + vec3 grid_size; + uint max_cascades; + + bool use_occlusion; + int probe_axis_size; + float probe_to_uvw; + float normal_bias; + + vec3 lightprobe_tex_pixel_size; + float energy; + + vec3 lightprobe_uv_offset; + float y_mult; + + vec3 occlusion_clamp; + uint pad3; + + vec3 occlusion_renormalize; + uint pad4; + + vec3 cascade_probe_size; + uint pad5; + + SDFVoxelGICascadeData cascades[SDFGI_MAX_CASCADES]; +} +sdfgi; + +layout(set = 1, binding = 1) uniform texture2DArray sdfgi_ambient_texture; + +layout(set = 1, binding = 2) uniform texture3D sdfgi_occlusion_texture; + +#endif //SDFGI + +layout(set = 0, binding = 14, std140) uniform Params { + vec2 fog_frustum_size_begin; + vec2 fog_frustum_size_end; + + float fog_frustum_end; + float ambient_inject; + float z_far; + int filter_axis; + + vec3 ambient_color; + float sky_contribution; + + ivec3 fog_volume_size; + uint directional_light_count; + + vec3 base_emission; + float base_density; + + vec3 base_scattering; + float phase_g; + + float detail_spread; + float gi_inject; + uint max_voxel_gi_instances; + uint cluster_type_size; + + vec2 screen_size; + uint cluster_shift; + uint cluster_width; + + uint max_cluster_element_count_div_32; + bool use_temporal_reprojection; + uint temporal_frame; + float temporal_blend; + + mat3x4 cam_rotation; + mat4 to_prev_view; + + mat3 radiance_inverse_xform; +} +params; +#ifndef MODE_COPY +layout(set = 0, binding = 15) uniform texture3D prev_density_texture; + +#ifdef MOLTENVK_USED +layout(set = 0, binding = 16) buffer density_only_map_buffer { + uint density_only_map[]; +}; +layout(set = 0, binding = 17) buffer light_only_map_buffer { + uint light_only_map[]; +}; +layout(set = 0, binding = 18) buffer emissive_only_map_buffer { + uint emissive_only_map[]; +}; +#else +layout(r32ui, set = 0, binding = 16) uniform uimage3D density_only_map; +layout(r32ui, set = 0, binding = 17) uniform uimage3D light_only_map; +layout(r32ui, set = 0, binding = 18) uniform uimage3D emissive_only_map; +#endif + +#ifdef USE_RADIANCE_CUBEMAP_ARRAY +layout(set = 0, binding = 19) uniform textureCubeArray sky_texture; +#else +layout(set = 0, binding = 19) uniform textureCube sky_texture; +#endif +#endif // MODE_COPY + +float get_depth_at_pos(float cell_depth_size, int z) { + float d = float(z) * cell_depth_size + cell_depth_size * 0.5; //center of voxels + d = pow(d, params.detail_spread); + return params.fog_frustum_end * d; +} + +vec3 hash3f(uvec3 x) { + x = ((x >> 16) ^ x) * 0x45d9f3b; + x = ((x >> 16) ^ x) * 0x45d9f3b; + x = (x >> 16) ^ x; + return vec3(x & 0xFFFFF) / vec3(float(0xFFFFF)); +} + +float get_omni_attenuation(float dist, float inv_range, float decay) { + float nd = dist * inv_range; + nd *= nd; + nd *= nd; // nd^4 + nd = max(1.0 - nd, 0.0); + nd *= nd; // nd^2 + return nd * pow(max(dist, 0.0001), -decay); +} + +void cluster_get_item_range(uint p_offset, out uint item_min, out uint item_max, out uint item_from, out uint item_to) { + uint item_min_max = cluster_buffer.data[p_offset]; + item_min = item_min_max & 0xFFFF; + item_max = item_min_max >> 16; + + item_from = item_min >> 5; + item_to = (item_max == 0) ? 0 : ((item_max - 1) >> 5) + 1; //side effect of how it is stored, as item_max 0 means no elements +} + +uint cluster_get_range_clip_mask(uint i, uint z_min, uint z_max) { + int local_min = clamp(int(z_min) - int(i) * 32, 0, 31); + int mask_width = min(int(z_max) - int(z_min), 32 - local_min); + return bitfieldInsert(uint(0), uint(0xFFFFFFFF), local_min, mask_width); +} + +float henyey_greenstein(float cos_theta, float g) { + const float k = 0.0795774715459; // 1 / (4 * PI) + return k * (1.0 - g * g) / (pow(1.0 + g * g - 2.0 * g * cos_theta, 1.5)); +} + +#define TEMPORAL_FRAMES 16 + +const vec3 halton_map[TEMPORAL_FRAMES] = vec3[]( + vec3(0.5, 0.33333333, 0.2), + vec3(0.25, 0.66666667, 0.4), + vec3(0.75, 0.11111111, 0.6), + vec3(0.125, 0.44444444, 0.8), + vec3(0.625, 0.77777778, 0.04), + vec3(0.375, 0.22222222, 0.24), + vec3(0.875, 0.55555556, 0.44), + vec3(0.0625, 0.88888889, 0.64), + vec3(0.5625, 0.03703704, 0.84), + vec3(0.3125, 0.37037037, 0.08), + vec3(0.8125, 0.7037037, 0.28), + vec3(0.1875, 0.14814815, 0.48), + vec3(0.6875, 0.48148148, 0.68), + vec3(0.4375, 0.81481481, 0.88), + vec3(0.9375, 0.25925926, 0.12), + vec3(0.03125, 0.59259259, 0.32)); + +// Higher values will make light in volumetric fog fade out sooner when it's occluded by shadow. +const float INV_FOG_FADE = 10.0; + +void main() { + vec3 fog_cell_size = 1.0 / vec3(params.fog_volume_size); + +#ifdef MODE_DENSITY + + ivec3 pos = ivec3(gl_GlobalInvocationID.xyz); + if (any(greaterThanEqual(pos, params.fog_volume_size))) { + return; //do not compute + } +#ifdef MOLTENVK_USED + uint lpos = pos.z * params.fog_volume_size.x * params.fog_volume_size.y + pos.y * params.fog_volume_size.x + pos.x; +#endif + + vec3 posf = vec3(pos); + + //posf += mix(vec3(0.0),vec3(1.0),0.3) * hash3f(uvec3(pos)) * 2.0 - 1.0; + + vec3 fog_unit_pos = posf * fog_cell_size + fog_cell_size * 0.5; //center of voxels + + uvec2 screen_pos = uvec2(fog_unit_pos.xy * params.screen_size); + uvec2 cluster_pos = screen_pos >> params.cluster_shift; + uint cluster_offset = (params.cluster_width * cluster_pos.y + cluster_pos.x) * (params.max_cluster_element_count_div_32 + 32); + //positions in screen are too spread apart, no hopes for optimizing with subgroups + + fog_unit_pos.z = pow(fog_unit_pos.z, params.detail_spread); + + vec3 view_pos; + view_pos.xy = (fog_unit_pos.xy * 2.0 - 1.0) * mix(params.fog_frustum_size_begin, params.fog_frustum_size_end, vec2(fog_unit_pos.z)); + view_pos.z = -params.fog_frustum_end * fog_unit_pos.z; + view_pos.y = -view_pos.y; + + vec4 reprojected_density = vec4(0.0); + float reproject_amount = 0.0; + + if (params.use_temporal_reprojection) { + vec3 prev_view = (params.to_prev_view * vec4(view_pos, 1.0)).xyz; + //undo transform into prev view + prev_view.y = -prev_view.y; + //z back to unit size + prev_view.z /= -params.fog_frustum_end; + //xy back to unit size + prev_view.xy /= mix(params.fog_frustum_size_begin, params.fog_frustum_size_end, vec2(prev_view.z)); + prev_view.xy = prev_view.xy * 0.5 + 0.5; + //z back to unspread value + prev_view.z = pow(prev_view.z, 1.0 / params.detail_spread); + + if (all(greaterThan(prev_view, vec3(0.0))) && all(lessThan(prev_view, vec3(1.0)))) { + //reprojectinon fits + + reprojected_density = textureLod(sampler3D(prev_density_texture, linear_sampler), prev_view, 0.0); + reproject_amount = params.temporal_blend; + + // Since we can reproject, now we must jitter the current view pos. + // This is done here because cells that can't reproject should not jitter. + + fog_unit_pos = posf * fog_cell_size + fog_cell_size * halton_map[params.temporal_frame]; //center of voxels, offset by halton table + + screen_pos = uvec2(fog_unit_pos.xy * params.screen_size); + cluster_pos = screen_pos >> params.cluster_shift; + cluster_offset = (params.cluster_width * cluster_pos.y + cluster_pos.x) * (params.max_cluster_element_count_div_32 + 32); + //positions in screen are too spread apart, no hopes for optimizing with subgroups + + fog_unit_pos.z = pow(fog_unit_pos.z, params.detail_spread); + + view_pos.xy = (fog_unit_pos.xy * 2.0 - 1.0) * mix(params.fog_frustum_size_begin, params.fog_frustum_size_end, vec2(fog_unit_pos.z)); + view_pos.z = -params.fog_frustum_end * fog_unit_pos.z; + view_pos.y = -view_pos.y; + } + } + + uint cluster_z = uint(clamp((abs(view_pos.z) / params.z_far) * 32.0, 0.0, 31.0)); + + vec3 total_light = vec3(0.0); + + float total_density = params.base_density; +#ifdef MOLTENVK_USED + uint local_density = density_only_map[lpos]; +#else + uint local_density = imageLoad(density_only_map, pos).x; +#endif + + total_density += float(int(local_density)) / DENSITY_SCALE; + total_density = max(0.0, total_density); + +#ifdef MOLTENVK_USED + uint scattering_u = light_only_map[lpos]; +#else + uint scattering_u = imageLoad(light_only_map, pos).x; +#endif + vec3 scattering = vec3(scattering_u >> 21, (scattering_u << 11) >> 21, scattering_u % 1024) / vec3(2047.0, 2047.0, 1023.0); + scattering += params.base_scattering * params.base_density; + +#ifdef MOLTENVK_USED + uint emission_u = emissive_only_map[lpos]; +#else + uint emission_u = imageLoad(emissive_only_map, pos).x; +#endif + vec3 emission = vec3(emission_u >> 21, (emission_u << 11) >> 21, emission_u % 1024) / vec3(511.0, 511.0, 255.0); + emission += params.base_emission * params.base_density; + + float cell_depth_size = abs(view_pos.z - get_depth_at_pos(fog_cell_size.z, pos.z + 1)); + //compute directional lights + + if (total_density > 0.00005) { + for (uint i = 0; i < params.directional_light_count; i++) { + if (directional_lights.data[i].volumetric_fog_energy > 0.001) { + vec3 shadow_attenuation = vec3(1.0); + + if (directional_lights.data[i].shadow_opacity > 0.001) { + float depth_z = -view_pos.z; + + vec4 pssm_coord; + vec3 light_dir = directional_lights.data[i].direction; + vec4 v = vec4(view_pos, 1.0); + float z_range; + + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); + pssm_coord /= pssm_coord.w; + z_range = directional_lights.data[i].shadow_z_range.x; + + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + pssm_coord /= pssm_coord.w; + z_range = directional_lights.data[i].shadow_z_range.y; + + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + pssm_coord /= pssm_coord.w; + z_range = directional_lights.data[i].shadow_z_range.z; + + } else { + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + pssm_coord /= pssm_coord.w; + z_range = directional_lights.data[i].shadow_z_range.w; + } + + float depth = texture(sampler2D(directional_shadow_atlas, linear_sampler), pssm_coord.xy).r; + float shadow = exp(min(0.0, (depth - pssm_coord.z)) * z_range * INV_FOG_FADE); + + shadow = mix(shadow, 1.0, smoothstep(directional_lights.data[i].fade_from, directional_lights.data[i].fade_to, view_pos.z)); //done with negative values for performance + + shadow_attenuation = mix(vec3(1.0 - directional_lights.data[i].shadow_opacity), vec3(1.0), shadow); + } + + total_light += shadow_attenuation * directional_lights.data[i].color * directional_lights.data[i].energy * henyey_greenstein(dot(normalize(view_pos), normalize(directional_lights.data[i].direction)), params.phase_g) * directional_lights.data[i].volumetric_fog_energy; + } + } + + // Compute light from sky + if (params.ambient_inject > 0.0) { + vec3 isotropic = vec3(0.0); + vec3 anisotropic = vec3(0.0); + if (params.sky_contribution > 0.0) { + float mip_bias = 2.0 + total_density * (MAX_SKY_LOD - 2.0); // Not physically based, but looks nice + vec3 scatter_direction = (params.radiance_inverse_xform * normalize(view_pos)) * sign(params.phase_g); +#ifdef USE_RADIANCE_CUBEMAP_ARRAY + isotropic = texture(samplerCubeArray(sky_texture, linear_sampler_with_mipmaps), vec4(0.0, 1.0, 0.0, mip_bias)).rgb; + anisotropic = texture(samplerCubeArray(sky_texture, linear_sampler_with_mipmaps), vec4(scatter_direction, mip_bias)).rgb; +#else + isotropic = textureLod(samplerCube(sky_texture, linear_sampler_with_mipmaps), vec3(0.0, 1.0, 0.0), mip_bias).rgb; + anisotropic = textureLod(samplerCube(sky_texture, linear_sampler_with_mipmaps), vec3(scatter_direction), mip_bias).rgb; +#endif //USE_RADIANCE_CUBEMAP_ARRAY + } + + total_light += mix(params.ambient_color, mix(isotropic, anisotropic, abs(params.phase_g)), params.sky_contribution) * params.ambient_inject; + } + + //compute lights from cluster + + { //omni lights + + uint cluster_omni_offset = cluster_offset; + + uint item_min; + uint item_max; + uint item_from; + uint item_to; + + cluster_get_item_range(cluster_omni_offset + params.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); + +#ifdef USE_SUBGROUPS + item_from = subgroupBroadcastFirst(subgroupMin(item_from)); + item_to = subgroupBroadcastFirst(subgroupMax(item_to)); +#endif + + for (uint i = item_from; i < item_to; i++) { + uint mask = cluster_buffer.data[cluster_omni_offset + i]; + mask &= cluster_get_range_clip_mask(i, item_min, item_max); +#ifdef USE_SUBGROUPS + uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); +#else + uint merged_mask = mask; +#endif + + while (merged_mask != 0) { + uint bit = findMSB(merged_mask); + merged_mask &= ~(1 << bit); +#ifdef USE_SUBGROUPS + if (((1 << bit) & mask) == 0) { //do not process if not originally here + continue; + } +#endif + uint light_index = 32 * i + bit; + + //if (!bool(omni_omni_lights.data[light_index].mask & draw_call.layer_mask)) { + // continue; //not masked + //} + + vec3 light_pos = omni_lights.data[light_index].position; + float d = distance(omni_lights.data[light_index].position, view_pos); + float shadow_attenuation = 1.0; + + if (omni_lights.data[light_index].volumetric_fog_energy > 0.001 && d * omni_lights.data[light_index].inv_radius < 1.0) { + float attenuation = get_omni_attenuation(d, omni_lights.data[light_index].inv_radius, omni_lights.data[light_index].attenuation); + + vec3 light = omni_lights.data[light_index].color; + + if (omni_lights.data[light_index].shadow_opacity > 0.001) { + //has shadow + vec4 uv_rect = omni_lights.data[light_index].atlas_rect; + vec2 flip_offset = omni_lights.data[light_index].direction.xy; + + vec3 local_vert = (omni_lights.data[light_index].shadow_matrix * vec4(view_pos, 1.0)).xyz; + + float shadow_len = length(local_vert); //need to remember shadow len from here + vec3 shadow_sample = normalize(local_vert); + + if (shadow_sample.z >= 0.0) { + uv_rect.xy += flip_offset; + } + + shadow_sample.z = 1.0 + abs(shadow_sample.z); + vec3 pos = vec3(shadow_sample.xy / shadow_sample.z, shadow_len - omni_lights.data[light_index].shadow_bias); + pos.z *= omni_lights.data[light_index].inv_radius; + + pos.xy = pos.xy * 0.5 + 0.5; + pos.xy = uv_rect.xy + pos.xy * uv_rect.zw; + + float depth = texture(sampler2D(shadow_atlas, linear_sampler), pos.xy).r; + + shadow_attenuation = mix(1.0 - omni_lights.data[light_index].shadow_opacity, 1.0, exp(min(0.0, (depth - pos.z)) / omni_lights.data[light_index].inv_radius * INV_FOG_FADE)); + } + total_light += light * attenuation * shadow_attenuation * henyey_greenstein(dot(normalize(light_pos - view_pos), normalize(view_pos)), params.phase_g) * omni_lights.data[light_index].volumetric_fog_energy; + } + } + } + } + + { //spot lights + + uint cluster_spot_offset = cluster_offset + params.cluster_type_size; + + uint item_min; + uint item_max; + uint item_from; + uint item_to; + + cluster_get_item_range(cluster_spot_offset + params.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); + +#ifdef USE_SUBGROUPS + item_from = subgroupBroadcastFirst(subgroupMin(item_from)); + item_to = subgroupBroadcastFirst(subgroupMax(item_to)); +#endif + + for (uint i = item_from; i < item_to; i++) { + uint mask = cluster_buffer.data[cluster_spot_offset + i]; + mask &= cluster_get_range_clip_mask(i, item_min, item_max); +#ifdef USE_SUBGROUPS + uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); +#else + uint merged_mask = mask; +#endif + + while (merged_mask != 0) { + uint bit = findMSB(merged_mask); + merged_mask &= ~(1 << bit); +#ifdef USE_SUBGROUPS + if (((1 << bit) & mask) == 0) { //do not process if not originally here + continue; + } +#endif + + //if (!bool(omni_lights.data[light_index].mask & draw_call.layer_mask)) { + // continue; //not masked + //} + + uint light_index = 32 * i + bit; + + vec3 light_pos = spot_lights.data[light_index].position; + vec3 light_rel_vec = spot_lights.data[light_index].position - view_pos; + float d = length(light_rel_vec); + float shadow_attenuation = 1.0; + + if (spot_lights.data[light_index].volumetric_fog_energy > 0.001 && d * spot_lights.data[light_index].inv_radius < 1.0) { + float attenuation = get_omni_attenuation(d, spot_lights.data[light_index].inv_radius, spot_lights.data[light_index].attenuation); + + vec3 spot_dir = spot_lights.data[light_index].direction; + float scos = max(dot(-normalize(light_rel_vec), spot_dir), spot_lights.data[light_index].cone_angle); + float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - spot_lights.data[light_index].cone_angle)); + attenuation *= 1.0 - pow(spot_rim, spot_lights.data[light_index].cone_attenuation); + + vec3 light = spot_lights.data[light_index].color; + + if (spot_lights.data[light_index].shadow_opacity > 0.001) { + //has shadow + vec4 uv_rect = spot_lights.data[light_index].atlas_rect; + vec2 flip_offset = spot_lights.data[light_index].direction.xy; + + vec3 local_vert = (spot_lights.data[light_index].shadow_matrix * vec4(view_pos, 1.0)).xyz; + + float shadow_len = length(local_vert); //need to remember shadow len from here + vec3 shadow_sample = normalize(local_vert); + + if (shadow_sample.z >= 0.0) { + uv_rect.xy += flip_offset; + } + + shadow_sample.z = 1.0 + abs(shadow_sample.z); + vec3 pos = vec3(shadow_sample.xy / shadow_sample.z, shadow_len - spot_lights.data[light_index].shadow_bias); + pos.z *= spot_lights.data[light_index].inv_radius; + + pos.xy = pos.xy * 0.5 + 0.5; + pos.xy = uv_rect.xy + pos.xy * uv_rect.zw; + + float depth = texture(sampler2D(shadow_atlas, linear_sampler), pos.xy).r; + + shadow_attenuation = mix(1.0 - spot_lights.data[light_index].shadow_opacity, 1.0, exp(min(0.0, (depth - pos.z)) / spot_lights.data[light_index].inv_radius * INV_FOG_FADE)); + } + total_light += light * attenuation * shadow_attenuation * henyey_greenstein(dot(normalize(light_rel_vec), normalize(view_pos)), params.phase_g) * spot_lights.data[light_index].volumetric_fog_energy; + } + } + } + } + + vec3 world_pos = mat3(params.cam_rotation) * view_pos; + + for (uint i = 0; i < params.max_voxel_gi_instances; i++) { + vec3 position = (voxel_gi_instances.data[i].xform * vec4(world_pos, 1.0)).xyz; + + //this causes corrupted pixels, i have no idea why.. + if (all(bvec2(all(greaterThanEqual(position, vec3(0.0))), all(lessThan(position, voxel_gi_instances.data[i].bounds))))) { + position /= voxel_gi_instances.data[i].bounds; + + vec4 light = vec4(0.0); + for (uint j = 0; j < voxel_gi_instances.data[i].mipmaps; j++) { + vec4 slight = textureLod(sampler3D(voxel_gi_textures[i], linear_sampler_with_mipmaps), position, float(j)); + float a = (1.0 - light.a); + light += a * slight; + } + + light.rgb *= voxel_gi_instances.data[i].dynamic_range * params.gi_inject * voxel_gi_instances.data[i].exposure_normalization; + + total_light += light.rgb; + } + } + + //sdfgi +#ifdef ENABLE_SDFGI + + { + float blend = -1.0; + vec3 ambient_total = vec3(0.0); + + for (uint i = 0; i < sdfgi.max_cascades; i++) { + vec3 cascade_pos = (world_pos - sdfgi.cascades[i].position) * sdfgi.cascades[i].to_probe; + + if (any(lessThan(cascade_pos, vec3(0.0))) || any(greaterThanEqual(cascade_pos, sdfgi.cascade_probe_size))) { + continue; //skip cascade + } + + vec3 base_pos = floor(cascade_pos); + ivec3 probe_base_pos = ivec3(base_pos); + + vec4 ambient_accum = vec4(0.0); + + ivec3 tex_pos = ivec3(probe_base_pos.xy, int(i)); + tex_pos.x += probe_base_pos.z * sdfgi.probe_axis_size; + + for (uint j = 0; j < 8; j++) { + ivec3 offset = (ivec3(j) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1); + ivec3 probe_posi = probe_base_pos; + probe_posi += offset; + + // Compute weight + + vec3 probe_pos = vec3(probe_posi); + vec3 probe_to_pos = cascade_pos - probe_pos; + + vec3 trilinear = vec3(1.0) - abs(probe_to_pos); + float weight = trilinear.x * trilinear.y * trilinear.z; + + // Compute lightprobe occlusion + + if (sdfgi.use_occlusion) { + ivec3 occ_indexv = abs((sdfgi.cascades[i].probe_world_offset + probe_posi) & ivec3(1, 1, 1)) * ivec3(1, 2, 4); + vec4 occ_mask = mix(vec4(0.0), vec4(1.0), equal(ivec4(occ_indexv.x | occ_indexv.y), ivec4(0, 1, 2, 3))); + + vec3 occ_pos = clamp(cascade_pos, probe_pos - sdfgi.occlusion_clamp, probe_pos + sdfgi.occlusion_clamp) * sdfgi.probe_to_uvw; + occ_pos.z += float(i); + if (occ_indexv.z != 0) { //z bit is on, means index is >=4, so make it switch to the other half of textures + occ_pos.x += 1.0; + } + + occ_pos *= sdfgi.occlusion_renormalize; + float occlusion = dot(textureLod(sampler3D(sdfgi_occlusion_texture, linear_sampler), occ_pos, 0.0), occ_mask); + + weight *= max(occlusion, 0.01); + } + + // Compute ambient texture position + + ivec3 uvw = tex_pos; + uvw.xy += offset.xy; + uvw.x += offset.z * sdfgi.probe_axis_size; + + vec3 ambient = texelFetch(sampler2DArray(sdfgi_ambient_texture, linear_sampler), uvw, 0).rgb; + + ambient_accum.rgb += ambient * weight * sdfgi.cascades[i].exposure_normalization; + ambient_accum.a += weight; + } + + if (ambient_accum.a > 0) { + ambient_accum.rgb /= ambient_accum.a; + } + ambient_total = ambient_accum.rgb; + break; + } + + total_light += ambient_total * params.gi_inject; + } + +#endif + } + + vec4 final_density = vec4(total_light * scattering + emission, total_density); + + final_density = mix(final_density, reprojected_density, reproject_amount); + + imageStore(density_map, pos, final_density); +#ifdef MOLTENVK_USED + density_only_map[lpos] = 0; + light_only_map[lpos] = 0; + emissive_only_map[lpos] = 0; +#else + imageStore(density_only_map, pos, uvec4(0)); + imageStore(light_only_map, pos, uvec4(0)); + imageStore(emissive_only_map, pos, uvec4(0)); +#endif +#endif + +#ifdef MODE_FOG + + ivec3 pos = ivec3(gl_GlobalInvocationID.xy, 0); + + if (any(greaterThanEqual(pos, params.fog_volume_size))) { + return; //do not compute + } + + vec4 fog_accum = vec4(0.0, 0.0, 0.0, 1.0); + float prev_z = 0.0; + + for (int i = 0; i < params.fog_volume_size.z; i++) { + //compute fog position + ivec3 fog_pos = pos + ivec3(0, 0, i); + //get fog value + vec4 fog = imageLoad(density_map, fog_pos); + + //get depth at cell pos + float z = get_depth_at_pos(fog_cell_size.z, i); + //get distance from previous pos + float d = abs(prev_z - z); + //compute transmittance using beer's law + float transmittance = exp(-d * fog.a); + + fog_accum.rgb += ((fog.rgb - fog.rgb * transmittance) / max(fog.a, 0.00001)) * fog_accum.a; + fog_accum.a *= transmittance; + + prev_z = z; + + imageStore(fog_map, fog_pos, vec4(fog_accum.rgb, 1.0 - fog_accum.a)); + } + +#endif + +#ifdef MODE_FILTER + + ivec3 pos = ivec3(gl_GlobalInvocationID.xyz); + + const float gauss[7] = float[](0.071303, 0.131514, 0.189879, 0.214607, 0.189879, 0.131514, 0.071303); + + const ivec3 filter_dir[3] = ivec3[](ivec3(1, 0, 0), ivec3(0, 1, 0), ivec3(0, 0, 1)); + ivec3 offset = filter_dir[params.filter_axis]; + + vec4 accum = vec4(0.0); + for (int i = -3; i <= 3; i++) { + accum += imageLoad(source_map, clamp(pos + offset * i, ivec3(0), params.fog_volume_size - ivec3(1))) * gauss[i + 3]; + } + + imageStore(dest_map, pos, accum); + +#endif +#ifdef MODE_COPY + ivec3 pos = ivec3(gl_GlobalInvocationID.xyz); + if (any(greaterThanEqual(pos, params.fog_volume_size))) { + return; //do not compute + } + + imageStore(dest_map, pos, imageLoad(source_map, pos)); + +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/environment/voxel_gi.glsl b/servers/rendering/renderer_rd/shaders/environment/voxel_gi.glsl new file mode 100644 index 0000000000..577c6d0cd0 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/environment/voxel_gi.glsl @@ -0,0 +1,616 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +#ifdef MODE_DYNAMIC +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; +#else +layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in; +#endif + +#ifndef MODE_DYNAMIC + +#define NO_CHILDREN 0xFFFFFFFF + +struct CellChildren { + uint children[8]; +}; + +layout(set = 0, binding = 1, std430) buffer CellChildrenBuffer { + CellChildren data[]; +} +cell_children; + +struct CellData { + uint position; // xyz 10 bits + uint albedo; //rgb albedo + uint emission; //rgb normalized with e as multiplier + uint normal; //RGB normal encoded +}; + +layout(set = 0, binding = 2, std430) buffer CellDataBuffer { + CellData data[]; +} +cell_data; + +#endif // MODE DYNAMIC + +#define LIGHT_TYPE_DIRECTIONAL 0 +#define LIGHT_TYPE_OMNI 1 +#define LIGHT_TYPE_SPOT 2 + +#if defined(MODE_COMPUTE_LIGHT) || defined(MODE_DYNAMIC_LIGHTING) + +struct Light { + uint type; + float energy; + float radius; + float attenuation; + + vec3 color; + float cos_spot_angle; + + vec3 position; + float inv_spot_attenuation; + + vec3 direction; + bool has_shadow; +}; + +layout(set = 0, binding = 3, std140) uniform Lights { + Light data[MAX_LIGHTS]; +} +lights; + +#endif // MODE COMPUTE LIGHT + +#ifdef MODE_SECOND_BOUNCE + +layout(set = 0, binding = 5) uniform texture3D color_texture; + +#endif // MODE_SECOND_BOUNCE + +#ifndef MODE_DYNAMIC + +layout(push_constant, std430) uniform Params { + ivec3 limits; + uint stack_size; + + float emission_scale; + float propagation; + float dynamic_range; + + uint light_count; + uint cell_offset; + uint cell_count; + float aniso_strength; + uint pad; +} +params; + +layout(set = 0, binding = 4, std430) buffer Outputs { + vec4 data[]; +} +outputs; + +#endif // MODE DYNAMIC + +layout(set = 0, binding = 9) uniform texture3D texture_sdf; +layout(set = 0, binding = 10) uniform sampler texture_sampler; + +#ifdef MODE_WRITE_TEXTURE + +layout(rgba8, set = 0, binding = 5) uniform restrict writeonly image3D color_tex; + +#endif + +#ifdef MODE_DYNAMIC + +layout(push_constant, std430) uniform Params { + ivec3 limits; + uint light_count; //when not lighting + ivec3 x_dir; + float z_base; + ivec3 y_dir; + float z_sign; + ivec3 z_dir; + float pos_multiplier; + ivec2 rect_pos; + ivec2 rect_size; + ivec2 prev_rect_ofs; + ivec2 prev_rect_size; + bool flip_x; + bool flip_y; + float dynamic_range; + bool on_mipmap; + float propagation; + float pad[3]; +} +params; + +#ifdef MODE_DYNAMIC_LIGHTING + +layout(rgba8, set = 0, binding = 5) uniform restrict readonly image2D source_albedo; +layout(rgba8, set = 0, binding = 6) uniform restrict readonly image2D source_normal; +layout(rgba8, set = 0, binding = 7) uniform restrict readonly image2D source_orm; +//layout (set=0,binding=8) uniform texture2D source_depth; +layout(rgba16f, set = 0, binding = 11) uniform restrict image2D emission; +layout(r32f, set = 0, binding = 12) uniform restrict image2D depth; + +#endif + +#ifdef MODE_DYNAMIC_SHRINK + +layout(rgba16f, set = 0, binding = 5) uniform restrict readonly image2D source_light; +layout(r32f, set = 0, binding = 6) uniform restrict readonly image2D source_depth; + +#ifdef MODE_DYNAMIC_SHRINK_WRITE + +layout(rgba16f, set = 0, binding = 7) uniform restrict writeonly image2D light; +layout(r32f, set = 0, binding = 8) uniform restrict writeonly image2D depth; + +#endif // MODE_DYNAMIC_SHRINK_WRITE + +#ifdef MODE_DYNAMIC_SHRINK_PLOT + +layout(rgba8, set = 0, binding = 11) uniform restrict image3D color_texture; + +#endif //MODE_DYNAMIC_SHRINK_PLOT + +#endif // MODE_DYNAMIC_SHRINK + +//layout (rgba8,set=0,binding=5) uniform restrict writeonly image3D color_tex; + +#endif // MODE DYNAMIC + +#if defined(MODE_COMPUTE_LIGHT) || defined(MODE_DYNAMIC_LIGHTING) + +float raymarch(float distance, float distance_adv, vec3 from, vec3 direction) { + vec3 cell_size = 1.0 / vec3(params.limits); + float occlusion = 1.0; + while (distance > 0.5) { //use this to avoid precision errors + float advance = texture(sampler3D(texture_sdf, texture_sampler), from * cell_size).r * 255.0 - 1.0; + if (advance < 0.0) { + occlusion = 0.0; + break; + } + + occlusion = min(advance, occlusion); + + advance = max(distance_adv, advance - mod(advance, distance_adv)); //should always advance in multiples of distance_adv + + from += direction * advance; + distance -= advance; + } + + return occlusion; //max(0.0,distance); +} + +float get_omni_attenuation(float distance, float inv_range, float decay) { + float nd = distance * inv_range; + nd *= nd; + nd *= nd; // nd^4 + nd = max(1.0 - nd, 0.0); + nd *= nd; // nd^2 + return nd * pow(max(distance, 0.0001), -decay); +} + +bool compute_light_vector(uint light, vec3 pos, out float attenuation, out vec3 light_pos) { + if (lights.data[light].type == LIGHT_TYPE_DIRECTIONAL) { + light_pos = pos - lights.data[light].direction * length(vec3(params.limits)); + attenuation = 1.0; + + } else { + light_pos = lights.data[light].position; + float distance = length(pos - light_pos); + if (distance >= lights.data[light].radius) { + return false; + } + + attenuation = get_omni_attenuation(distance, 1.0 / lights.data[light].radius, lights.data[light].attenuation); + + if (lights.data[light].type == LIGHT_TYPE_SPOT) { + vec3 rel = normalize(pos - light_pos); + float cos_spot_angle = lights.data[light].cos_spot_angle; + float cos_angle = dot(rel, lights.data[light].direction); + if (cos_angle < cos_spot_angle) { + return false; + } + + float scos = max(cos_angle, cos_spot_angle); + float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - cos_spot_angle)); + attenuation *= 1.0 - pow(spot_rim, lights.data[light].inv_spot_attenuation); + } + } + + return true; +} + +float get_normal_advance(vec3 p_normal) { + vec3 normal = p_normal; + vec3 unorm = abs(normal); + + if ((unorm.x >= unorm.y) && (unorm.x >= unorm.z)) { + // x code + unorm = normal.x > 0.0 ? vec3(1.0, 0.0, 0.0) : vec3(-1.0, 0.0, 0.0); + } else if ((unorm.y > unorm.x) && (unorm.y >= unorm.z)) { + // y code + unorm = normal.y > 0.0 ? vec3(0.0, 1.0, 0.0) : vec3(0.0, -1.0, 0.0); + } else if ((unorm.z > unorm.x) && (unorm.z > unorm.y)) { + // z code + unorm = normal.z > 0.0 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 0.0, -1.0); + } else { + // oh-no we messed up code + // has to be + unorm = vec3(1.0, 0.0, 0.0); + } + + return 1.0 / dot(normal, unorm); +} + +void clip_segment(vec4 plane, vec3 begin, inout vec3 end) { + vec3 segment = begin - end; + float den = dot(plane.xyz, segment); + + //printf("den is %i\n",den); + if (den < 0.0001) { + return; + } + + float dist = (dot(plane.xyz, begin) - plane.w) / den; + + if (dist < 0.0001 || dist > 1.0001) { + return; + } + + end = begin + segment * -dist; +} + +bool compute_light_at_pos(uint index, vec3 pos, vec3 normal, inout vec3 light, inout vec3 light_dir) { + float attenuation; + vec3 light_pos; + + if (!compute_light_vector(index, pos, attenuation, light_pos)) { + return false; + } + + light_dir = normalize(pos - light_pos); + + if (attenuation < 0.01 || (length(normal) > 0.2 && dot(normal, light_dir) >= 0)) { + return false; //not facing the light, or attenuation is near zero + } + + if (lights.data[index].has_shadow) { + float distance_adv = get_normal_advance(light_dir); + + vec3 to = pos; + if (length(normal) > 0.2) { + to += normal * distance_adv * 0.51; + } else { + to -= sign(light_dir) * 0.45; //go near the edge towards the light direction to avoid self occlusion + } + + //clip + clip_segment(mix(vec4(-1.0, 0.0, 0.0, 0.0), vec4(1.0, 0.0, 0.0, float(params.limits.x - 1)), bvec4(light_dir.x < 0.0)), to, light_pos); + clip_segment(mix(vec4(0.0, -1.0, 0.0, 0.0), vec4(0.0, 1.0, 0.0, float(params.limits.y - 1)), bvec4(light_dir.y < 0.0)), to, light_pos); + clip_segment(mix(vec4(0.0, 0.0, -1.0, 0.0), vec4(0.0, 0.0, 1.0, float(params.limits.z - 1)), bvec4(light_dir.z < 0.0)), to, light_pos); + + float distance = length(to - light_pos); + if (distance < 0.1) { + return false; // hit + } + + distance += distance_adv - mod(distance, distance_adv); //make it reach the center of the box always + light_pos = to - light_dir * distance; + + //from -= sign(light_dir)*0.45; //go near the edge towards the light direction to avoid self occlusion + + /*float dist = raymarch(distance,distance_adv,light_pos,light_dir); + + if (dist > distance_adv) { + return false; + } + + attenuation *= 1.0 - smoothstep(0.1*distance_adv,distance_adv,dist); + */ + + float occlusion = raymarch(distance, distance_adv, light_pos, light_dir); + + if (occlusion == 0.0) { + return false; + } + + attenuation *= occlusion; //1.0 - smoothstep(0.1*distance_adv,distance_adv,dist); + } + + light = lights.data[index].color * attenuation * lights.data[index].energy; + return true; +} + +#endif // MODE COMPUTE LIGHT + +void main() { +#ifndef MODE_DYNAMIC + + uint cell_index = gl_GlobalInvocationID.x; + if (cell_index >= params.cell_count) { + return; + } + cell_index += params.cell_offset; + + uvec3 posu = uvec3(cell_data.data[cell_index].position & 0x7FF, (cell_data.data[cell_index].position >> 11) & 0x3FF, cell_data.data[cell_index].position >> 21); + vec4 albedo = unpackUnorm4x8(cell_data.data[cell_index].albedo); + +#endif + + /////////////////COMPUTE LIGHT/////////////////////////////// + +#ifdef MODE_COMPUTE_LIGHT + + vec3 pos = vec3(posu) + vec3(0.5); + + vec3 emission = vec3(uvec3(cell_data.data[cell_index].emission & 0x1ff, (cell_data.data[cell_index].emission >> 9) & 0x1ff, (cell_data.data[cell_index].emission >> 18) & 0x1ff)) * pow(2.0, float(cell_data.data[cell_index].emission >> 27) - 15.0 - 9.0); + vec3 normal = unpackSnorm4x8(cell_data.data[cell_index].normal).xyz; + + vec3 accum = vec3(0.0); + + for (uint i = 0; i < params.light_count; i++) { + vec3 light; + vec3 light_dir; + if (!compute_light_at_pos(i, pos, normal.xyz, light, light_dir)) { + continue; + } + + light *= albedo.rgb; + + if (length(normal) > 0.2) { + accum += max(0.0, dot(normal, -light_dir)) * light; + } else { + //all directions + accum += light; + } + } + + outputs.data[cell_index] = vec4(accum + emission, 0.0); + +#endif //MODE_COMPUTE_LIGHT + + /////////////////SECOND BOUNCE/////////////////////////////// + +#ifdef MODE_SECOND_BOUNCE + vec3 pos = vec3(posu) + vec3(0.5); + ivec3 ipos = ivec3(posu); + vec4 normal = unpackSnorm4x8(cell_data.data[cell_index].normal); + + vec3 accum = outputs.data[cell_index].rgb; + + if (length(normal.xyz) > 0.2) { + vec3 v0 = abs(normal.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 1.0, 0.0); + vec3 tangent = normalize(cross(v0, normal.xyz)); + vec3 bitangent = normalize(cross(tangent, normal.xyz)); + mat3 normal_mat = mat3(tangent, bitangent, normal.xyz); + +#define MAX_CONE_DIRS 6 + + vec3 cone_dirs[MAX_CONE_DIRS] = vec3[]( + vec3(0.0, 0.0, 1.0), + vec3(0.866025, 0.0, 0.5), + vec3(0.267617, 0.823639, 0.5), + vec3(-0.700629, 0.509037, 0.5), + vec3(-0.700629, -0.509037, 0.5), + vec3(0.267617, -0.823639, 0.5)); + + float cone_weights[MAX_CONE_DIRS] = float[](0.25, 0.15, 0.15, 0.15, 0.15, 0.15); + float tan_half_angle = 0.577; + + for (int i = 0; i < MAX_CONE_DIRS; i++) { + vec3 direction = normal_mat * cone_dirs[i]; + vec4 color = vec4(0.0); + { + float dist = 1.5; + float max_distance = length(vec3(params.limits)); + vec3 cell_size = 1.0 / vec3(params.limits); + + while (dist < max_distance && color.a < 0.95) { + float diameter = max(1.0, 2.0 * tan_half_angle * dist); + vec3 uvw_pos = (pos + dist * direction) * cell_size; + float half_diameter = diameter * 0.5; + //check if outside, then break + //if ( any(greaterThan(abs(uvw_pos - 0.5),vec3(0.5f + half_diameter * cell_size)) ) ) { + // break; + //} + + float log2_diameter = log2(diameter); + vec4 scolor = textureLod(sampler3D(color_texture, texture_sampler), uvw_pos, log2_diameter); + float a = (1.0 - color.a); + color += a * scolor; + dist += half_diameter; + } + } + color *= cone_weights[i] * vec4(albedo.rgb, 1.0) * params.dynamic_range; //restore range + accum += color.rgb; + } + } + + outputs.data[cell_index] = vec4(accum, 0.0); + +#endif // MODE_SECOND_BOUNCE + + /////////////////UPDATE MIPMAPS/////////////////////////////// + +#ifdef MODE_UPDATE_MIPMAPS + + { + vec3 light_accum = vec3(0.0); + float count = 0.0; + for (uint i = 0; i < 8; i++) { + uint child_index = cell_children.data[cell_index].children[i]; + if (child_index == NO_CHILDREN) { + continue; + } + light_accum += outputs.data[child_index].rgb; + + count += 1.0; + } + + float divisor = mix(8.0, count, params.propagation); + outputs.data[cell_index] = vec4(light_accum / divisor, 0.0); + } +#endif + + ///////////////////WRITE TEXTURE///////////////////////////// + +#ifdef MODE_WRITE_TEXTURE + { + imageStore(color_tex, ivec3(posu), vec4(outputs.data[cell_index].rgb / params.dynamic_range, albedo.a)); + } +#endif + + ///////////////////DYNAMIC LIGHTING///////////////////////////// + +#ifdef MODE_DYNAMIC + + ivec2 pos_xy = ivec2(gl_GlobalInvocationID.xy); + if (any(greaterThanEqual(pos_xy, params.rect_size))) { + return; //out of bounds + } + + ivec2 uv_xy = pos_xy; + if (params.flip_x) { + uv_xy.x = params.rect_size.x - pos_xy.x - 1; + } + if (params.flip_y) { + uv_xy.y = params.rect_size.y - pos_xy.y - 1; + } + +#ifdef MODE_DYNAMIC_LIGHTING + + { + float z = params.z_base + imageLoad(depth, uv_xy).x * params.z_sign; + + ivec3 pos = params.x_dir * (params.rect_pos.x + pos_xy.x) + params.y_dir * (params.rect_pos.y + pos_xy.y) + abs(params.z_dir) * int(z); + + vec3 normal = imageLoad(source_normal, uv_xy).xyz * 2.0 - 1.0; + normal = vec3(params.x_dir) * normal.x * mix(1.0, -1.0, params.flip_x) + vec3(params.y_dir) * normal.y * mix(1.0, -1.0, params.flip_y) - vec3(params.z_dir) * normal.z; + + vec4 albedo = imageLoad(source_albedo, uv_xy); + + //determine the position in space + + vec3 accum = vec3(0.0); + for (uint i = 0; i < params.light_count; i++) { + vec3 light; + vec3 light_dir; + if (!compute_light_at_pos(i, vec3(pos) * params.pos_multiplier, normal, light, light_dir)) { + continue; + } + + light *= albedo.rgb; + + accum += max(0.0, dot(normal, -light_dir)) * light; + } + + accum += imageLoad(emission, uv_xy).xyz; + + imageStore(emission, uv_xy, vec4(accum, albedo.a)); + imageStore(depth, uv_xy, vec4(z)); + } + +#endif // MODE DYNAMIC LIGHTING + +#ifdef MODE_DYNAMIC_SHRINK + + { + vec4 accum = vec4(0.0); + float accum_z = 0.0; + float count = 0.0; + + for (int i = 0; i < 4; i++) { + ivec2 ofs = pos_xy * 2 + ivec2(i & 1, i >> 1) - params.prev_rect_ofs; + if (any(lessThan(ofs, ivec2(0))) || any(greaterThanEqual(ofs, params.prev_rect_size))) { + continue; + } + if (params.flip_x) { + ofs.x = params.prev_rect_size.x - ofs.x - 1; + } + if (params.flip_y) { + ofs.y = params.prev_rect_size.y - ofs.y - 1; + } + + vec4 light = imageLoad(source_light, ofs); + if (light.a == 0.0) { //ignore empty + continue; + } + accum += light; + float z = imageLoad(source_depth, ofs).x; + accum_z += z * 0.5; //shrink half too + count += 1.0; + } + + if (params.on_mipmap) { + accum.rgb /= mix(8.0, count, params.propagation); + accum.a /= 8.0; + } else { + accum /= 4.0; + } + + if (count == 0.0) { + accum_z = 0.0; //avoid nan + } else { + accum_z /= count; + } + +#ifdef MODE_DYNAMIC_SHRINK_WRITE + + imageStore(light, uv_xy, accum); + imageStore(depth, uv_xy, vec4(accum_z)); +#endif + +#ifdef MODE_DYNAMIC_SHRINK_PLOT + + if (accum.a < 0.001) { + return; //do not blit if alpha is too low + } + + ivec3 pos = params.x_dir * (params.rect_pos.x + pos_xy.x) + params.y_dir * (params.rect_pos.y + pos_xy.y) + abs(params.z_dir) * int(accum_z); + + float z_frac = fract(accum_z); + + for (int i = 0; i < 2; i++) { + ivec3 pos3d = pos + abs(params.z_dir) * i; + if (any(lessThan(pos3d, ivec3(0))) || any(greaterThanEqual(pos3d, params.limits))) { + //skip if offlimits + continue; + } + vec4 color_blit = accum * (i == 0 ? 1.0 - z_frac : z_frac); + vec4 color = imageLoad(color_texture, pos3d); + color.rgb *= params.dynamic_range; + +#if 0 + color.rgb = mix(color.rgb,color_blit.rgb,color_blit.a); + color.a+=color_blit.a; +#else + + float sa = 1.0 - color_blit.a; + vec4 result; + result.a = color.a * sa + color_blit.a; + if (result.a == 0.0) { + result = vec4(0.0); + } else { + result.rgb = (color.rgb * color.a * sa + color_blit.rgb * color_blit.a) / result.a; + color = result; + } + +#endif + color.rgb /= params.dynamic_range; + imageStore(color_texture, pos3d, color); + //imageStore(color_texture,pos3d,vec4(1,1,1,1)); + } +#endif // MODE_DYNAMIC_SHRINK_PLOT + } +#endif + +#endif // MODE DYNAMIC +} diff --git a/servers/rendering/renderer_rd/shaders/environment/voxel_gi_debug.glsl b/servers/rendering/renderer_rd/shaders/environment/voxel_gi_debug.glsl new file mode 100644 index 0000000000..fd7a2bf8ad --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/environment/voxel_gi_debug.glsl @@ -0,0 +1,168 @@ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +struct CellData { + uint position; // xyz 10 bits + uint albedo; //rgb albedo + uint emission; //rgb normalized with e as multiplier + uint normal; //RGB normal encoded +}; + +layout(set = 0, binding = 1, std140) buffer CellDataBuffer { + CellData data[]; +} +cell_data; + +layout(set = 0, binding = 2) uniform texture3D color_tex; + +layout(set = 0, binding = 3) uniform sampler tex_sampler; + +layout(push_constant, std430) uniform Params { + mat4 projection; + uint cell_offset; + float dynamic_range; + float alpha; + uint level; + ivec3 bounds; + uint pad; +} +params; + +layout(location = 0) out vec4 color_interp; + +void main() { + const vec3 cube_triangles[36] = vec3[]( + vec3(-1.0f, -1.0f, -1.0f), + vec3(-1.0f, -1.0f, 1.0f), + vec3(-1.0f, 1.0f, 1.0f), + vec3(1.0f, 1.0f, -1.0f), + vec3(-1.0f, -1.0f, -1.0f), + vec3(-1.0f, 1.0f, -1.0f), + vec3(1.0f, -1.0f, 1.0f), + vec3(-1.0f, -1.0f, -1.0f), + vec3(1.0f, -1.0f, -1.0f), + vec3(1.0f, 1.0f, -1.0f), + vec3(1.0f, -1.0f, -1.0f), + vec3(-1.0f, -1.0f, -1.0f), + vec3(-1.0f, -1.0f, -1.0f), + vec3(-1.0f, 1.0f, 1.0f), + vec3(-1.0f, 1.0f, -1.0f), + vec3(1.0f, -1.0f, 1.0f), + vec3(-1.0f, -1.0f, 1.0f), + vec3(-1.0f, -1.0f, -1.0f), + vec3(-1.0f, 1.0f, 1.0f), + vec3(-1.0f, -1.0f, 1.0f), + vec3(1.0f, -1.0f, 1.0f), + vec3(1.0f, 1.0f, 1.0f), + vec3(1.0f, -1.0f, -1.0f), + vec3(1.0f, 1.0f, -1.0f), + vec3(1.0f, -1.0f, -1.0f), + vec3(1.0f, 1.0f, 1.0f), + vec3(1.0f, -1.0f, 1.0f), + vec3(1.0f, 1.0f, 1.0f), + vec3(1.0f, 1.0f, -1.0f), + vec3(-1.0f, 1.0f, -1.0f), + vec3(1.0f, 1.0f, 1.0f), + vec3(-1.0f, 1.0f, -1.0f), + vec3(-1.0f, 1.0f, 1.0f), + vec3(1.0f, 1.0f, 1.0f), + vec3(-1.0f, 1.0f, 1.0f), + vec3(1.0f, -1.0f, 1.0f)); + + vec3 vertex = cube_triangles[gl_VertexIndex] * 0.5 + 0.5; +#ifdef MODE_DEBUG_LIGHT_FULL + uvec3 posu = uvec3(gl_InstanceIndex % params.bounds.x, (gl_InstanceIndex / params.bounds.x) % params.bounds.y, gl_InstanceIndex / (params.bounds.y * params.bounds.x)); +#else + uint cell_index = gl_InstanceIndex + params.cell_offset; + + uvec3 posu = uvec3(cell_data.data[cell_index].position & 0x7FF, (cell_data.data[cell_index].position >> 11) & 0x3FF, cell_data.data[cell_index].position >> 21); +#endif + +#ifdef MODE_DEBUG_EMISSION + color_interp.xyz = vec3(uvec3(cell_data.data[cell_index].emission & 0x1ff, (cell_data.data[cell_index].emission >> 9) & 0x1ff, (cell_data.data[cell_index].emission >> 18) & 0x1ff)) * pow(2.0, float(cell_data.data[cell_index].emission >> 27) - 15.0 - 9.0); +#endif + +#ifdef MODE_DEBUG_COLOR + color_interp.xyz = unpackUnorm4x8(cell_data.data[cell_index].albedo).xyz; +#endif + +#ifdef MODE_DEBUG_LIGHT + color_interp = texelFetch(sampler3D(color_tex, tex_sampler), ivec3(posu), int(params.level)); + color_interp.xyz *params.dynamic_range; +#endif + + float scale = (1 << params.level); + + gl_Position = params.projection * vec4((vec3(posu) + vertex) * scale, 1.0); + +#ifdef MODE_DEBUG_LIGHT_FULL + if (color_interp.a == 0.0) { + gl_Position = vec4(0.0); //force clip and not draw + } +#else + color_interp.a = params.alpha; +#endif +} + +#[fragment] + +#version 450 + +#VERSION_DEFINES + +layout(location = 0) in vec4 color_interp; +layout(location = 0) out vec4 frag_color; + +void main() { + frag_color = color_interp; + +#ifdef MODE_DEBUG_LIGHT_FULL + + //there really is no alpha, so use dither + + int x = int(gl_FragCoord.x) % 4; + int y = int(gl_FragCoord.y) % 4; + int index = x + y * 4; + float limit = 0.0; + if (x < 8) { + if (index == 0) + limit = 0.0625; + if (index == 1) + limit = 0.5625; + if (index == 2) + limit = 0.1875; + if (index == 3) + limit = 0.6875; + if (index == 4) + limit = 0.8125; + if (index == 5) + limit = 0.3125; + if (index == 6) + limit = 0.9375; + if (index == 7) + limit = 0.4375; + if (index == 8) + limit = 0.25; + if (index == 9) + limit = 0.75; + if (index == 10) + limit = 0.125; + if (index == 11) + limit = 0.625; + if (index == 12) + limit = 1.0; + if (index == 13) + limit = 0.5; + if (index == 14) + limit = 0.875; + if (index == 15) + limit = 0.375; + } + if (frag_color.a < limit) { + discard; + } +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/environment/voxel_gi_sdf.glsl b/servers/rendering/renderer_rd/shaders/environment/voxel_gi_sdf.glsl new file mode 100644 index 0000000000..47a611a543 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/environment/voxel_gi_sdf.glsl @@ -0,0 +1,180 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 4, local_size_y = 4, local_size_z = 4) in; + +#define MAX_DISTANCE 100000.0 + +#define NO_CHILDREN 0xFFFFFFFF + +struct CellChildren { + uint children[8]; +}; + +layout(set = 0, binding = 1, std430) buffer CellChildrenBuffer { + CellChildren data[]; +} +cell_children; + +struct CellData { + uint position; // xyz 10 bits + uint albedo; //rgb albedo + uint emission; //rgb normalized with e as multiplier + uint normal; //RGB normal encoded +}; + +layout(set = 0, binding = 2, std430) buffer CellDataBuffer { + CellData data[]; +} +cell_data; + +layout(r8ui, set = 0, binding = 3) uniform restrict writeonly uimage3D sdf_tex; + +layout(push_constant, std430) uniform Params { + uint offset; + uint end; + uint pad0; + uint pad1; +} +params; + +void main() { + vec3 pos = vec3(gl_GlobalInvocationID); + float closest_dist = MAX_DISTANCE; + + for (uint i = params.offset; i < params.end; i++) { + vec3 posu = vec3(uvec3(cell_data.data[i].position & 0x7FF, (cell_data.data[i].position >> 11) & 0x3FF, cell_data.data[i].position >> 21)); + float dist = length(pos - posu); + if (dist < closest_dist) { + closest_dist = dist; + } + } + + uint dist_8; + + if (closest_dist < 0.0001) { // same cell + dist_8 = 0; //equals to -1 + } else { + dist_8 = clamp(uint(closest_dist), 0, 254) + 1; //conservative, 0 is 1, so <1 is considered solid + } + + imageStore(sdf_tex, ivec3(gl_GlobalInvocationID), uvec4(dist_8)); + //imageStore(sdf_tex,pos,uvec4(pos*2,0)); +} + +#if 0 +layout(push_constant, std430) uniform Params { + ivec3 limits; + uint stack_size; +} +params; + +float distance_to_aabb(ivec3 pos, ivec3 aabb_pos, ivec3 aabb_size) { + vec3 delta = vec3(max(ivec3(0), max(aabb_pos - pos, pos - (aabb_pos + aabb_size - ivec3(1))))); + return length(delta); +} + +void main() { + ivec3 pos = ivec3(gl_GlobalInvocationID); + + uint stack[10] = uint[](0, 0, 0, 0, 0, 0, 0, 0, 0, 0); + uint stack_indices[10] = uint[](0, 0, 0, 0, 0, 0, 0, 0, 0, 0); + ivec3 stack_positions[10] = ivec3[](ivec3(0), ivec3(0), ivec3(0), ivec3(0), ivec3(0), ivec3(0), ivec3(0), ivec3(0), ivec3(0), ivec3(0)); + + const uint cell_orders[8] = uint[]( + 0x11f58d1, + 0xe2e70a, + 0xd47463, + 0xbb829c, + 0x8d11f5, + 0x70ae2e, + 0x463d47, + 0x29cbb8); + + bool cell_found = false; + bool cell_found_exact = false; + ivec3 closest_cell_pos; + float closest_distance = MAX_DISTANCE; + int stack_pos = 0; + + while (true) { + uint index = stack_indices[stack_pos] >> 24; + + if (index == 8) { + //go up + if (stack_pos == 0) { + break; //done going through octree + } + stack_pos--; + continue; + } + + stack_indices[stack_pos] = (stack_indices[stack_pos] & ((1 << 24) - 1)) | ((index + 1) << 24); + + uint cell_index = (stack_indices[stack_pos] >> (index * 3)) & 0x7; + uint child_cell = cell_children.data[stack[stack_pos]].children[cell_index]; + + if (child_cell == NO_CHILDREN) { + continue; + } + + ivec3 child_cell_size = params.limits >> (stack_pos + 1); + ivec3 child_cell_pos = stack_positions[stack_pos]; + + child_cell_pos += mix(ivec3(0), child_cell_size, bvec3(uvec3(index & 1, index & 2, index & 4) != uvec3(0))); + + bool is_leaf = stack_pos == (params.stack_size - 2); + + if (child_cell_pos == pos && is_leaf) { + //we may actually end up in the exact cell. + //if this happens, just abort + cell_found_exact = true; + break; + } + + if (cell_found) { + //discard by distance + float distance = distance_to_aabb(pos, child_cell_pos, child_cell_size); + if (distance >= closest_distance) { + continue; //pointless, just test next child + } else if (is_leaf) { + //closer than what we have AND end of stack, save and continue + closest_cell_pos = child_cell_pos; + closest_distance = distance; + continue; + } + } else if (is_leaf) { + //first solid cell we find, save and continue + closest_distance = distance_to_aabb(pos, child_cell_pos, child_cell_size); + closest_cell_pos = child_cell_pos; + cell_found = true; + continue; + } + + bvec3 direction = greaterThan((pos - (child_cell_pos + (child_cell_size >> 1))), ivec3(0)); + uint cell_order = 0; + cell_order |= mix(0, 1, direction.x); + cell_order |= mix(0, 2, direction.y); + cell_order |= mix(0, 4, direction.z); + + stack[stack_pos + 1] = child_cell; + stack_indices[stack_pos + 1] = cell_orders[cell_order]; //start counting + stack_positions[stack_pos + 1] = child_cell_pos; + stack_pos++; //go up stack + } + + uint dist_8; + + if (cell_found_exact) { + dist_8 = 0; //equals to -1 + } else { + float closest_distance = length(vec3(pos - closest_cell_pos)); + dist_8 = clamp(uint(closest_distance), 0, 254) + 1; //conservative, 0 is 1, so <1 is considered solid + } + + imageStore(sdf_tex, pos, uvec4(dist_8)); +} +#endif diff --git a/servers/rendering/renderer_rd/shaders/forward_clustered/SCsub b/servers/rendering/renderer_rd/shaders/forward_clustered/SCsub new file mode 100644 index 0000000000..f06a2d86e2 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/forward_clustered/SCsub @@ -0,0 +1,17 @@ +#!/usr/bin/env python + +Import("env") + +if "RD_GLSL" in env["BUILDERS"]: + # find all include files + gl_include_files = [str(f) for f in Glob("*_inc.glsl")] + [str(f) for f in Glob("../*_inc.glsl")] + + # find all shader code(all glsl files excluding our include files) + glsl_files = [str(f) for f in Glob("*.glsl") if str(f) not in gl_include_files] + + # make sure we recompile shaders if include files change + env.Depends([f + ".gen.h" for f in glsl_files], gl_include_files + ["#glsl_builders.py"]) + + # compile shaders + for glsl_file in glsl_files: + env.RD_GLSL(glsl_file) diff --git a/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered.glsl b/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered.glsl new file mode 100644 index 0000000000..0bdf0e50aa --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered.glsl @@ -0,0 +1,2187 @@ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#include "scene_forward_clustered_inc.glsl" + +#define SHADER_IS_SRGB false + +/* INPUT ATTRIBS */ + +layout(location = 0) in vec3 vertex_attrib; + +//only for pure render depth when normal is not used + +#ifdef NORMAL_USED +layout(location = 1) in vec2 normal_attrib; +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) +layout(location = 2) in vec2 tangent_attrib; +#endif + +#if defined(COLOR_USED) +layout(location = 3) in vec4 color_attrib; +#endif + +#ifdef UV_USED +layout(location = 4) in vec2 uv_attrib; +#endif + +#if defined(UV2_USED) || defined(USE_LIGHTMAP) || defined(MODE_RENDER_MATERIAL) +layout(location = 5) in vec2 uv2_attrib; +#endif + +#if defined(CUSTOM0_USED) +layout(location = 6) in vec4 custom0_attrib; +#endif + +#if defined(CUSTOM1_USED) +layout(location = 7) in vec4 custom1_attrib; +#endif + +#if defined(CUSTOM2_USED) +layout(location = 8) in vec4 custom2_attrib; +#endif + +#if defined(CUSTOM3_USED) +layout(location = 9) in vec4 custom3_attrib; +#endif + +#if defined(BONES_USED) || defined(USE_PARTICLE_TRAILS) +layout(location = 10) in uvec4 bone_attrib; +#endif + +#if defined(WEIGHTS_USED) || defined(USE_PARTICLE_TRAILS) +layout(location = 11) in vec4 weight_attrib; +#endif + +vec3 oct_to_vec3(vec2 e) { + vec3 v = vec3(e.xy, 1.0 - abs(e.x) - abs(e.y)); + float t = max(-v.z, 0.0); + v.xy += t * -sign(v.xy); + return v; +} + +/* Varyings */ + +layout(location = 0) out vec3 vertex_interp; + +#ifdef NORMAL_USED +layout(location = 1) out vec3 normal_interp; +#endif + +#if defined(COLOR_USED) +layout(location = 2) out vec4 color_interp; +#endif + +#ifdef UV_USED +layout(location = 3) out vec2 uv_interp; +#endif + +#if defined(UV2_USED) || defined(USE_LIGHTMAP) +layout(location = 4) out vec2 uv2_interp; +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) +layout(location = 5) out vec3 tangent_interp; +layout(location = 6) out vec3 binormal_interp; +#endif + +#ifdef MOTION_VECTORS +layout(location = 7) out vec4 screen_position; +layout(location = 8) out vec4 prev_screen_position; +#endif + +#ifdef MATERIAL_UNIFORMS_USED +layout(set = MATERIAL_UNIFORM_SET, binding = 0, std140) uniform MaterialUniforms{ +#MATERIAL_UNIFORMS +} material; +#endif + +float global_time; + +#ifdef MODE_DUAL_PARABOLOID + +layout(location = 9) out float dp_clip; + +#endif + +layout(location = 10) out flat uint instance_index_interp; + +#ifdef USE_MULTIVIEW +#ifdef has_VK_KHR_multiview +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +// !BAS! This needs to become an input once we implement our fallback! +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#else // USE_MULTIVIEW +// Set to zero, not supported in non stereo +#define ViewIndex 0 +#endif //USE_MULTIVIEW + +invariant gl_Position; + +#GLOBALS + +#ifdef USE_DOUBLE_PRECISION +// Helper functions for emulating double precision when adding floats. +vec3 quick_two_sum(vec3 a, vec3 b, out vec3 out_p) { + vec3 s = a + b; + out_p = b - (s - a); + return s; +} + +vec3 two_sum(vec3 a, vec3 b, out vec3 out_p) { + vec3 s = a + b; + vec3 v = s - a; + out_p = (a - (s - v)) + (b - v); + return s; +} + +vec3 double_add_vec3(vec3 base_a, vec3 prec_a, vec3 base_b, vec3 prec_b, out vec3 out_precision) { + vec3 s, t, se, te; + s = two_sum(base_a, base_b, se); + t = two_sum(prec_a, prec_b, te); + se += t; + s = quick_two_sum(s, se, se); + se += te; + s = quick_two_sum(s, se, out_precision); + return s; +} +#endif + +void vertex_shader(in uint instance_index, in bool is_multimesh, in uint multimesh_offset, in SceneData scene_data, in mat4 model_matrix, out vec4 screen_pos) { + vec4 instance_custom = vec4(0.0); +#if defined(COLOR_USED) + color_interp = color_attrib; +#endif + + mat4 inv_view_matrix = scene_data.inv_view_matrix; + +#ifdef USE_DOUBLE_PRECISION + vec3 model_precision = vec3(model_matrix[0][3], model_matrix[1][3], model_matrix[2][3]); + model_matrix[0][3] = 0.0; + model_matrix[1][3] = 0.0; + model_matrix[2][3] = 0.0; + vec3 view_precision = vec3(inv_view_matrix[0][3], inv_view_matrix[1][3], inv_view_matrix[2][3]); + inv_view_matrix[0][3] = 0.0; + inv_view_matrix[1][3] = 0.0; + inv_view_matrix[2][3] = 0.0; +#endif + + mat3 model_normal_matrix; + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_NON_UNIFORM_SCALE)) { + model_normal_matrix = transpose(inverse(mat3(model_matrix))); + } else { + model_normal_matrix = mat3(model_matrix); + } + + mat4 matrix; + mat4 read_model_matrix = model_matrix; + + if (is_multimesh) { + //multimesh, instances are for it + +#ifdef USE_PARTICLE_TRAILS + uint trail_size = (instances.data[instance_index].flags >> INSTANCE_FLAGS_PARTICLE_TRAIL_SHIFT) & INSTANCE_FLAGS_PARTICLE_TRAIL_MASK; + uint stride = 3 + 1 + 1; //particles always uses this format + + uint offset = trail_size * stride * gl_InstanceIndex; + +#ifdef COLOR_USED + vec4 pcolor; +#endif + { + uint boffset = offset + bone_attrib.x * stride; + matrix = mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], transforms.data[boffset + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weight_attrib.x; +#ifdef COLOR_USED + pcolor = transforms.data[boffset + 3] * weight_attrib.x; +#endif + } + if (weight_attrib.y > 0.001) { + uint boffset = offset + bone_attrib.y * stride; + matrix += mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], transforms.data[boffset + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weight_attrib.y; +#ifdef COLOR_USED + pcolor += transforms.data[boffset + 3] * weight_attrib.y; +#endif + } + if (weight_attrib.z > 0.001) { + uint boffset = offset + bone_attrib.z * stride; + matrix += mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], transforms.data[boffset + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weight_attrib.z; +#ifdef COLOR_USED + pcolor += transforms.data[boffset + 3] * weight_attrib.z; +#endif + } + if (weight_attrib.w > 0.001) { + uint boffset = offset + bone_attrib.w * stride; + matrix += mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], transforms.data[boffset + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weight_attrib.w; +#ifdef COLOR_USED + pcolor += transforms.data[boffset + 3] * weight_attrib.w; +#endif + } + + instance_custom = transforms.data[offset + 4]; + +#ifdef COLOR_USED + color_interp *= pcolor; +#endif + +#else + uint stride = 0; + { + //TODO implement a small lookup table for the stride + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH_FORMAT_2D)) { + stride += 2; + } else { + stride += 3; + } + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH_HAS_COLOR)) { + stride += 1; + } + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH_HAS_CUSTOM_DATA)) { + stride += 1; + } + } + + uint offset = stride * (gl_InstanceIndex + multimesh_offset); + + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH_FORMAT_2D)) { + matrix = mat4(transforms.data[offset + 0], transforms.data[offset + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0)); + offset += 2; + } else { + matrix = mat4(transforms.data[offset + 0], transforms.data[offset + 1], transforms.data[offset + 2], vec4(0.0, 0.0, 0.0, 1.0)); + offset += 3; + } + + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH_HAS_COLOR)) { +#ifdef COLOR_USED + color_interp *= transforms.data[offset]; +#endif + offset += 1; + } + + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH_HAS_CUSTOM_DATA)) { + instance_custom = transforms.data[offset]; + } + +#endif + //transpose + matrix = transpose(matrix); +#if !defined(USE_DOUBLE_PRECISION) || defined(SKIP_TRANSFORM_USED) || defined(VERTEX_WORLD_COORDS_USED) || defined(MODEL_MATRIX_USED) + // Normally we can bake the multimesh transform into the model matrix, but when using double precision + // we avoid baking it in so we can emulate high precision. + read_model_matrix = model_matrix * matrix; +#if !defined(USE_DOUBLE_PRECISION) || defined(SKIP_TRANSFORM_USED) || defined(VERTEX_WORLD_COORDS_USED) + model_matrix = read_model_matrix; +#endif // !defined(USE_DOUBLE_PRECISION) || defined(SKIP_TRANSFORM_USED) || defined(VERTEX_WORLD_COORDS_USED) +#endif // !defined(USE_DOUBLE_PRECISION) || defined(SKIP_TRANSFORM_USED) || defined(VERTEX_WORLD_COORDS_USED) || defined(MODEL_MATRIX_USED) + model_normal_matrix = model_normal_matrix * mat3(matrix); + } + + vec3 vertex = vertex_attrib; +#ifdef NORMAL_USED + vec3 normal = oct_to_vec3(normal_attrib * 2.0 - 1.0); +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) + vec2 signed_tangent_attrib = tangent_attrib * 2.0 - 1.0; + vec3 tangent = oct_to_vec3(vec2(signed_tangent_attrib.x, abs(signed_tangent_attrib.y) * 2.0 - 1.0)); + float binormalf = sign(signed_tangent_attrib.y); + vec3 binormal = normalize(cross(normal, tangent) * binormalf); +#endif + +#ifdef UV_USED + uv_interp = uv_attrib; +#endif + +#if defined(UV2_USED) || defined(USE_LIGHTMAP) + uv2_interp = uv2_attrib; +#endif + +#ifdef OVERRIDE_POSITION + vec4 position; +#endif + +#ifdef USE_MULTIVIEW + mat4 projection_matrix = scene_data.projection_matrix_view[ViewIndex]; + mat4 inv_projection_matrix = scene_data.inv_projection_matrix_view[ViewIndex]; +#else + mat4 projection_matrix = scene_data.projection_matrix; + mat4 inv_projection_matrix = scene_data.inv_projection_matrix; +#endif //USE_MULTIVIEW + +//using world coordinates +#if !defined(SKIP_TRANSFORM_USED) && defined(VERTEX_WORLD_COORDS_USED) + + vertex = (model_matrix * vec4(vertex, 1.0)).xyz; + +#ifdef NORMAL_USED + normal = model_normal_matrix * normal; +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) + + tangent = model_normal_matrix * tangent; + binormal = model_normal_matrix * binormal; + +#endif +#endif + + float roughness = 1.0; + + mat4 modelview = scene_data.view_matrix * model_matrix; + mat3 modelview_normal = mat3(scene_data.view_matrix) * model_normal_matrix; + + { +#CODE : VERTEX + } + +// using local coordinates (default) +#if !defined(SKIP_TRANSFORM_USED) && !defined(VERTEX_WORLD_COORDS_USED) + +#ifdef USE_DOUBLE_PRECISION + // We separate the basis from the origin because the basis is fine with single point precision. + // Then we combine the translations from the model matrix and the view matrix using emulated doubles. + // We add the result to the vertex and ignore the final lost precision. + vec3 model_origin = model_matrix[3].xyz; + if (is_multimesh) { + vertex = mat3(matrix) * vertex; + model_origin = double_add_vec3(model_origin, model_precision, matrix[3].xyz, vec3(0.0), model_precision); + } + vertex = mat3(model_matrix) * vertex; + vec3 temp_precision; // Will be ignored. + vertex += double_add_vec3(model_origin, model_precision, scene_data.inv_view_matrix[3].xyz, view_precision, temp_precision); + vertex = mat3(scene_data.view_matrix) * vertex; +#else + vertex = (modelview * vec4(vertex, 1.0)).xyz; +#endif +#ifdef NORMAL_USED + normal = modelview_normal * normal; +#endif + +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) + + binormal = modelview_normal * binormal; + tangent = modelview_normal * tangent; +#endif + +//using world coordinates +#if !defined(SKIP_TRANSFORM_USED) && defined(VERTEX_WORLD_COORDS_USED) + + vertex = (scene_data.view_matrix * vec4(vertex, 1.0)).xyz; +#ifdef NORMAL_USED + normal = (scene_data.view_matrix * vec4(normal, 0.0)).xyz; +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) + binormal = (scene_data.view_matrix * vec4(binormal, 0.0)).xyz; + tangent = (scene_data.view_matrix * vec4(tangent, 0.0)).xyz; +#endif +#endif + + vertex_interp = vertex; + +#ifdef NORMAL_USED + normal_interp = normal; +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) + tangent_interp = tangent; + binormal_interp = binormal; +#endif + +#ifdef MODE_RENDER_DEPTH + +#ifdef MODE_DUAL_PARABOLOID + + vertex_interp.z *= scene_data.dual_paraboloid_side; + + dp_clip = vertex_interp.z; //this attempts to avoid noise caused by objects sent to the other parabolloid side due to bias + + //for dual paraboloid shadow mapping, this is the fastest but least correct way, as it curves straight edges + + vec3 vtx = vertex_interp; + float distance = length(vtx); + vtx = normalize(vtx); + vtx.xy /= 1.0 - vtx.z; + vtx.z = (distance / scene_data.z_far); + vtx.z = vtx.z * 2.0 - 1.0; + vertex_interp = vtx; + +#endif + +#endif //MODE_RENDER_DEPTH + +#ifdef OVERRIDE_POSITION + gl_Position = position; +#else + gl_Position = projection_matrix * vec4(vertex_interp, 1.0); +#endif + +#ifdef MOTION_VECTORS + screen_pos = gl_Position; +#endif + +#ifdef MODE_RENDER_DEPTH + if (scene_data.pancake_shadows) { + if (gl_Position.z <= 0.00001) { + gl_Position.z = 0.00001; + } + } +#endif +#ifdef MODE_RENDER_MATERIAL + if (scene_data.material_uv2_mode) { + vec2 uv_offset = unpackHalf2x16(draw_call.uv_offset); + gl_Position.xy = (uv2_attrib.xy + uv_offset) * 2.0 - 1.0; + gl_Position.z = 0.00001; + gl_Position.w = 1.0; + } +#endif +} + +void main() { + uint instance_index = draw_call.instance_index; + + bool is_multimesh = bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH); + if (!is_multimesh) { + instance_index += gl_InstanceIndex; + } + + instance_index_interp = instance_index; + + mat4 model_matrix = instances.data[instance_index].transform; +#if defined(MOTION_VECTORS) + global_time = scene_data_block.prev_data.time; + vertex_shader(instance_index, is_multimesh, draw_call.multimesh_motion_vectors_previous_offset, scene_data_block.prev_data, instances.data[instance_index].prev_transform, prev_screen_position); + global_time = scene_data_block.data.time; + vertex_shader(instance_index, is_multimesh, draw_call.multimesh_motion_vectors_current_offset, scene_data_block.data, model_matrix, screen_position); +#else + global_time = scene_data_block.data.time; + vec4 screen_position; + vertex_shader(instance_index, is_multimesh, draw_call.multimesh_motion_vectors_current_offset, scene_data_block.data, model_matrix, screen_position); +#endif +} + +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#define SHADER_IS_SRGB false + +/* Specialization Constants (Toggles) */ + +layout(constant_id = 0) const bool sc_use_forward_gi = false; +layout(constant_id = 1) const bool sc_use_light_projector = false; +layout(constant_id = 2) const bool sc_use_light_soft_shadows = false; +layout(constant_id = 3) const bool sc_use_directional_soft_shadows = false; + +/* Specialization Constants (Values) */ + +layout(constant_id = 6) const uint sc_soft_shadow_samples = 4; +layout(constant_id = 7) const uint sc_penumbra_shadow_samples = 4; + +layout(constant_id = 8) const uint sc_directional_soft_shadow_samples = 4; +layout(constant_id = 9) const uint sc_directional_penumbra_shadow_samples = 4; + +layout(constant_id = 10) const bool sc_decal_use_mipmaps = true; +layout(constant_id = 11) const bool sc_projector_use_mipmaps = true; + +// not used in clustered renderer but we share some code with the mobile renderer that requires this. +const float sc_luminance_multiplier = 1.0; + +#include "scene_forward_clustered_inc.glsl" + +/* Varyings */ + +layout(location = 0) in vec3 vertex_interp; + +#ifdef NORMAL_USED +layout(location = 1) in vec3 normal_interp; +#endif + +#if defined(COLOR_USED) +layout(location = 2) in vec4 color_interp; +#endif + +#ifdef UV_USED +layout(location = 3) in vec2 uv_interp; +#endif + +#if defined(UV2_USED) || defined(USE_LIGHTMAP) +layout(location = 4) in vec2 uv2_interp; +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) +layout(location = 5) in vec3 tangent_interp; +layout(location = 6) in vec3 binormal_interp; +#endif + +#ifdef MOTION_VECTORS +layout(location = 7) in vec4 screen_position; +layout(location = 8) in vec4 prev_screen_position; +#endif + +#ifdef MODE_DUAL_PARABOLOID + +layout(location = 9) in float dp_clip; + +#endif + +layout(location = 10) in flat uint instance_index_interp; + +#ifdef USE_MULTIVIEW +#ifdef has_VK_KHR_multiview +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +// !BAS! This needs to become an input once we implement our fallback! +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#else // USE_MULTIVIEW +// Set to zero, not supported in non stereo +#define ViewIndex 0 +#endif //USE_MULTIVIEW + +//defines to keep compatibility with vertex + +#ifdef USE_MULTIVIEW +#define projection_matrix scene_data.projection_matrix_view[ViewIndex] +#define inv_projection_matrix scene_data.inv_projection_matrix_view[ViewIndex] +#else +#define projection_matrix scene_data.projection_matrix +#define inv_projection_matrix scene_data.inv_projection_matrix +#endif + +#define global_time scene_data_block.data.time + +#if defined(ENABLE_SSS) && defined(ENABLE_TRANSMITTANCE) +//both required for transmittance to be enabled +#define LIGHT_TRANSMITTANCE_USED +#endif + +#ifdef MATERIAL_UNIFORMS_USED +layout(set = MATERIAL_UNIFORM_SET, binding = 0, std140) uniform MaterialUniforms{ + +#MATERIAL_UNIFORMS + +} material; +#endif + +#GLOBALS + +#ifdef MODE_RENDER_DEPTH + +#ifdef MODE_RENDER_MATERIAL + +layout(location = 0) out vec4 albedo_output_buffer; +layout(location = 1) out vec4 normal_output_buffer; +layout(location = 2) out vec4 orm_output_buffer; +layout(location = 3) out vec4 emission_output_buffer; +layout(location = 4) out float depth_output_buffer; + +#endif // MODE_RENDER_MATERIAL + +#ifdef MODE_RENDER_NORMAL_ROUGHNESS +layout(location = 0) out vec4 normal_roughness_output_buffer; + +#ifdef MODE_RENDER_VOXEL_GI +layout(location = 1) out uvec2 voxel_gi_buffer; +#endif + +#endif //MODE_RENDER_NORMAL +#else // RENDER DEPTH + +#ifdef MODE_SEPARATE_SPECULAR + +layout(location = 0) out vec4 diffuse_buffer; //diffuse (rgb) and roughness +layout(location = 1) out vec4 specular_buffer; //specular and SSS (subsurface scatter) +#else + +layout(location = 0) out vec4 frag_color; +#endif // MODE_SEPARATE_SPECULAR + +#endif // RENDER DEPTH + +#ifdef MOTION_VECTORS +layout(location = 2) out vec2 motion_vector; +#endif + +#include "../scene_forward_aa_inc.glsl" + +#if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + +// Default to SPECULAR_SCHLICK_GGX. +#if !defined(SPECULAR_DISABLED) && !defined(SPECULAR_SCHLICK_GGX) && !defined(SPECULAR_TOON) +#define SPECULAR_SCHLICK_GGX +#endif + +#include "../scene_forward_lights_inc.glsl" + +#include "../scene_forward_gi_inc.glsl" + +#endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + +#ifndef MODE_RENDER_DEPTH + +vec4 volumetric_fog_process(vec2 screen_uv, float z) { + vec3 fog_pos = vec3(screen_uv, z * implementation_data.volumetric_fog_inv_length); + if (fog_pos.z < 0.0) { + return vec4(0.0); + } else if (fog_pos.z < 1.0) { + fog_pos.z = pow(fog_pos.z, implementation_data.volumetric_fog_detail_spread); + } + + return texture(sampler3D(volumetric_fog_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), fog_pos); +} + +vec4 fog_process(vec3 vertex) { + vec3 fog_color = scene_data_block.data.fog_light_color; + + if (scene_data_block.data.fog_aerial_perspective > 0.0) { + vec3 sky_fog_color = vec3(0.0); + vec3 cube_view = scene_data_block.data.radiance_inverse_xform * vertex; + // mip_level always reads from the second mipmap and higher so the fog is always slightly blurred + float mip_level = mix(1.0 / MAX_ROUGHNESS_LOD, 1.0, 1.0 - (abs(vertex.z) - scene_data_block.data.z_near) / (scene_data_block.data.z_far - scene_data_block.data.z_near)); +#ifdef USE_RADIANCE_CUBEMAP_ARRAY + float lod, blend; + blend = modf(mip_level * MAX_ROUGHNESS_LOD, lod); + sky_fog_color = texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(cube_view, lod)).rgb; + sky_fog_color = mix(sky_fog_color, texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(cube_view, lod + 1)).rgb, blend); +#else + sky_fog_color = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), cube_view, mip_level * MAX_ROUGHNESS_LOD).rgb; +#endif //USE_RADIANCE_CUBEMAP_ARRAY + fog_color = mix(fog_color, sky_fog_color, scene_data_block.data.fog_aerial_perspective); + } + + if (scene_data_block.data.fog_sun_scatter > 0.001) { + vec4 sun_scatter = vec4(0.0); + float sun_total = 0.0; + vec3 view = normalize(vertex); + + for (uint i = 0; i < scene_data_block.data.directional_light_count; i++) { + vec3 light_color = directional_lights.data[i].color * directional_lights.data[i].energy; + float light_amount = pow(max(dot(view, directional_lights.data[i].direction), 0.0), 8.0); + fog_color += light_color * light_amount * scene_data_block.data.fog_sun_scatter; + } + } + + float fog_amount = 1.0 - exp(min(0.0, -length(vertex) * scene_data_block.data.fog_density)); + + if (abs(scene_data_block.data.fog_height_density) >= 0.0001) { + float y = (scene_data_block.data.inv_view_matrix * vec4(vertex, 1.0)).y; + + float y_dist = y - scene_data_block.data.fog_height; + + float vfog_amount = 1.0 - exp(min(0.0, y_dist * scene_data_block.data.fog_height_density)); + + fog_amount = max(vfog_amount, fog_amount); + } + + return vec4(fog_color, fog_amount); +} + +void cluster_get_item_range(uint p_offset, out uint item_min, out uint item_max, out uint item_from, out uint item_to) { + uint item_min_max = cluster_buffer.data[p_offset]; + item_min = item_min_max & 0xFFFFu; + item_max = item_min_max >> 16; + + item_from = item_min >> 5; + item_to = (item_max == 0) ? 0 : ((item_max - 1) >> 5) + 1; //side effect of how it is stored, as item_max 0 means no elements +} + +uint cluster_get_range_clip_mask(uint i, uint z_min, uint z_max) { + int local_min = clamp(int(z_min) - int(i) * 32, 0, 31); + int mask_width = min(int(z_max) - int(z_min), 32 - local_min); + return bitfieldInsert(uint(0), uint(0xFFFFFFFF), local_min, mask_width); +} + +#endif //!MODE_RENDER DEPTH + +void fragment_shader(in SceneData scene_data) { + uint instance_index = instance_index_interp; + + //lay out everything, whatever is unused is optimized away anyway + vec3 vertex = vertex_interp; +#ifdef USE_MULTIVIEW + vec3 view = -normalize(vertex_interp - scene_data.eye_offset[ViewIndex].xyz); +#else + vec3 view = -normalize(vertex_interp); +#endif + vec3 albedo = vec3(1.0); + vec3 backlight = vec3(0.0); + vec4 transmittance_color = vec4(0.0, 0.0, 0.0, 1.0); + float transmittance_depth = 0.0; + float transmittance_boost = 0.0; + float metallic = 0.0; + float specular = 0.5; + vec3 emission = vec3(0.0); + float roughness = 1.0; + float rim = 0.0; + float rim_tint = 0.0; + float clearcoat = 0.0; + float clearcoat_roughness = 0.0; + float anisotropy = 0.0; + vec2 anisotropy_flow = vec2(1.0, 0.0); + vec4 fog = vec4(0.0); +#if defined(CUSTOM_RADIANCE_USED) + vec4 custom_radiance = vec4(0.0); +#endif +#if defined(CUSTOM_IRRADIANCE_USED) + vec4 custom_irradiance = vec4(0.0); +#endif + + float ao = 1.0; + float ao_light_affect = 0.0; + + float alpha = float(instances.data[instance_index].flags >> INSTANCE_FLAGS_FADE_SHIFT) / float(255.0); + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) + vec3 binormal = normalize(binormal_interp); + vec3 tangent = normalize(tangent_interp); +#else + vec3 binormal = vec3(0.0); + vec3 tangent = vec3(0.0); +#endif + +#ifdef NORMAL_USED + vec3 normal = normalize(normal_interp); + +#if defined(DO_SIDE_CHECK) + if (!gl_FrontFacing) { + normal = -normal; + } +#endif + +#endif //NORMAL_USED + +#ifdef UV_USED + vec2 uv = uv_interp; +#endif + +#if defined(UV2_USED) || defined(USE_LIGHTMAP) + vec2 uv2 = uv2_interp; +#endif + +#if defined(COLOR_USED) + vec4 color = color_interp; +#endif + +#if defined(NORMAL_MAP_USED) + + vec3 normal_map = vec3(0.5); +#endif + + float normal_map_depth = 1.0; + + vec2 screen_uv = gl_FragCoord.xy * scene_data.screen_pixel_size; + + float sss_strength = 0.0; + +#ifdef ALPHA_SCISSOR_USED + float alpha_scissor_threshold = 1.0; +#endif // ALPHA_SCISSOR_USED + +#ifdef ALPHA_HASH_USED + float alpha_hash_scale = 1.0; +#endif // ALPHA_HASH_USED + +#ifdef ALPHA_ANTIALIASING_EDGE_USED + float alpha_antialiasing_edge = 0.0; + vec2 alpha_texture_coordinate = vec2(0.0, 0.0); +#endif // ALPHA_ANTIALIASING_EDGE_USED + + mat4 inv_view_matrix = scene_data.inv_view_matrix; + mat4 read_model_matrix = instances.data[instance_index].transform; +#ifdef USE_DOUBLE_PRECISION + read_model_matrix[0][3] = 0.0; + read_model_matrix[1][3] = 0.0; + read_model_matrix[2][3] = 0.0; + inv_view_matrix[0][3] = 0.0; + inv_view_matrix[1][3] = 0.0; + inv_view_matrix[2][3] = 0.0; +#endif + + { +#CODE : FRAGMENT + } + +#ifdef LIGHT_TRANSMITTANCE_USED + transmittance_color.a *= sss_strength; +#endif + +#ifndef USE_SHADOW_TO_OPACITY + +#ifdef ALPHA_SCISSOR_USED + if (alpha < alpha_scissor_threshold) { + discard; + } +#endif // ALPHA_SCISSOR_USED + +// alpha hash can be used in unison with alpha antialiasing +#ifdef ALPHA_HASH_USED + if (alpha < compute_alpha_hash_threshold(vertex, alpha_hash_scale)) { + discard; + } +#endif // ALPHA_HASH_USED + +// If we are not edge antialiasing, we need to remove the output alpha channel from scissor and hash +#if (defined(ALPHA_SCISSOR_USED) || defined(ALPHA_HASH_USED)) && !defined(ALPHA_ANTIALIASING_EDGE_USED) + alpha = 1.0; +#endif + +#ifdef ALPHA_ANTIALIASING_EDGE_USED +// If alpha scissor is used, we must further the edge threshold, otherwise we won't get any edge feather +#ifdef ALPHA_SCISSOR_USED + alpha_antialiasing_edge = clamp(alpha_scissor_threshold + alpha_antialiasing_edge, 0.0, 1.0); +#endif + alpha = compute_alpha_antialiasing_edge(alpha, alpha_texture_coordinate, alpha_antialiasing_edge); +#endif // ALPHA_ANTIALIASING_EDGE_USED + +#ifdef USE_OPAQUE_PREPASS + if (alpha < scene_data.opaque_prepass_threshold) { + discard; + } +#endif // USE_OPAQUE_PREPASS + +#endif // !USE_SHADOW_TO_OPACITY + +#ifdef NORMAL_MAP_USED + + normal_map.xy = normal_map.xy * 2.0 - 1.0; + normal_map.z = sqrt(max(0.0, 1.0 - dot(normal_map.xy, normal_map.xy))); //always ignore Z, as it can be RG packed, Z may be pos/neg, etc. + + normal = normalize(mix(normal, tangent * normal_map.x + binormal * normal_map.y + normal * normal_map.z, normal_map_depth)); + +#endif + +#ifdef LIGHT_ANISOTROPY_USED + + if (anisotropy > 0.01) { + //rotation matrix + mat3 rot = mat3(tangent, binormal, normal); + //make local to space + tangent = normalize(rot * vec3(anisotropy_flow.x, anisotropy_flow.y, 0.0)); + binormal = normalize(rot * vec3(-anisotropy_flow.y, anisotropy_flow.x, 0.0)); + } + +#endif + +#ifdef ENABLE_CLIP_ALPHA + if (albedo.a < 0.99) { + //used for doublepass and shadowmapping + discard; + } +#endif + + /////////////////////// FOG ////////////////////// +#ifndef MODE_RENDER_DEPTH + +#ifndef CUSTOM_FOG_USED + // fog must be processed as early as possible and then packed. + // to maximize VGPR usage + // Draw "fixed" fog before volumetric fog to ensure volumetric fog can appear in front of the sky. + + if (scene_data.fog_enabled) { + fog = fog_process(vertex); + } + + if (implementation_data.volumetric_fog_enabled) { + vec4 volumetric_fog = volumetric_fog_process(screen_uv, -vertex.z); + if (scene_data.fog_enabled) { + //must use the full blending equation here to blend fogs + vec4 res; + float sa = 1.0 - volumetric_fog.a; + res.a = fog.a * sa + volumetric_fog.a; + if (res.a == 0.0) { + res.rgb = vec3(0.0); + } else { + res.rgb = (fog.rgb * fog.a * sa + volumetric_fog.rgb * volumetric_fog.a) / res.a; + } + fog = res; + } else { + fog = volumetric_fog; + } + } +#endif //!CUSTOM_FOG_USED + + uint fog_rg = packHalf2x16(fog.rg); + uint fog_ba = packHalf2x16(fog.ba); + +#endif //!MODE_RENDER_DEPTH + + /////////////////////// DECALS //////////////////////////////// + +#ifndef MODE_RENDER_DEPTH + + uvec2 cluster_pos = uvec2(gl_FragCoord.xy) >> implementation_data.cluster_shift; + uint cluster_offset = (implementation_data.cluster_width * cluster_pos.y + cluster_pos.x) * (implementation_data.max_cluster_element_count_div_32 + 32); + + uint cluster_z = uint(clamp((-vertex.z / scene_data.z_far) * 32.0, 0.0, 31.0)); + + //used for interpolating anything cluster related + vec3 vertex_ddx = dFdx(vertex); + vec3 vertex_ddy = dFdy(vertex); + + { // process decals + + uint cluster_decal_offset = cluster_offset + implementation_data.cluster_type_size * 2; + + uint item_min; + uint item_max; + uint item_from; + uint item_to; + + cluster_get_item_range(cluster_decal_offset + implementation_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); + +#ifdef USE_SUBGROUPS + item_from = subgroupBroadcastFirst(subgroupMin(item_from)); + item_to = subgroupBroadcastFirst(subgroupMax(item_to)); +#endif + + for (uint i = item_from; i < item_to; i++) { + uint mask = cluster_buffer.data[cluster_decal_offset + i]; + mask &= cluster_get_range_clip_mask(i, item_min, item_max); +#ifdef USE_SUBGROUPS + uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); +#else + uint merged_mask = mask; +#endif + + while (merged_mask != 0) { + uint bit = findMSB(merged_mask); + merged_mask &= ~(1u << bit); +#ifdef USE_SUBGROUPS + if (((1u << bit) & mask) == 0) { //do not process if not originally here + continue; + } +#endif + uint decal_index = 32 * i + bit; + + if (!bool(decals.data[decal_index].mask & instances.data[instance_index].layer_mask)) { + continue; //not masked + } + + vec3 uv_local = (decals.data[decal_index].xform * vec4(vertex, 1.0)).xyz; + if (any(lessThan(uv_local, vec3(0.0, -1.0, 0.0))) || any(greaterThan(uv_local, vec3(1.0)))) { + continue; //out of decal + } + + float fade = pow(1.0 - (uv_local.y > 0.0 ? uv_local.y : -uv_local.y), uv_local.y > 0.0 ? decals.data[decal_index].upper_fade : decals.data[decal_index].lower_fade); + + if (decals.data[decal_index].normal_fade > 0.0) { + fade *= smoothstep(decals.data[decal_index].normal_fade, 1.0, dot(normal_interp, decals.data[decal_index].normal) * 0.5 + 0.5); + } + + //we need ddx/ddy for mipmaps, so simulate them + vec2 ddx = (decals.data[decal_index].xform * vec4(vertex_ddx, 0.0)).xz; + vec2 ddy = (decals.data[decal_index].xform * vec4(vertex_ddy, 0.0)).xz; + + if (decals.data[decal_index].albedo_rect != vec4(0.0)) { + //has albedo + vec4 decal_albedo; + if (sc_decal_use_mipmaps) { + decal_albedo = textureGrad(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].albedo_rect.zw + decals.data[decal_index].albedo_rect.xy, ddx * decals.data[decal_index].albedo_rect.zw, ddy * decals.data[decal_index].albedo_rect.zw); + } else { + decal_albedo = textureLod(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].albedo_rect.zw + decals.data[decal_index].albedo_rect.xy, 0.0); + } + decal_albedo *= decals.data[decal_index].modulate; + decal_albedo.a *= fade; + albedo = mix(albedo, decal_albedo.rgb, decal_albedo.a * decals.data[decal_index].albedo_mix); + + if (decals.data[decal_index].normal_rect != vec4(0.0)) { + vec3 decal_normal; + if (sc_decal_use_mipmaps) { + decal_normal = textureGrad(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].normal_rect.zw + decals.data[decal_index].normal_rect.xy, ddx * decals.data[decal_index].normal_rect.zw, ddy * decals.data[decal_index].normal_rect.zw).xyz; + } else { + decal_normal = textureLod(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].normal_rect.zw + decals.data[decal_index].normal_rect.xy, 0.0).xyz; + } + decal_normal.xy = decal_normal.xy * vec2(2.0, -2.0) - vec2(1.0, -1.0); //users prefer flipped y normal maps in most authoring software + decal_normal.z = sqrt(max(0.0, 1.0 - dot(decal_normal.xy, decal_normal.xy))); + //convert to view space, use xzy because y is up + decal_normal = (decals.data[decal_index].normal_xform * decal_normal.xzy).xyz; + + normal = normalize(mix(normal, decal_normal, decal_albedo.a)); + } + + if (decals.data[decal_index].orm_rect != vec4(0.0)) { + vec3 decal_orm; + if (sc_decal_use_mipmaps) { + decal_orm = textureGrad(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].orm_rect.zw + decals.data[decal_index].orm_rect.xy, ddx * decals.data[decal_index].orm_rect.zw, ddy * decals.data[decal_index].orm_rect.zw).xyz; + } else { + decal_orm = textureLod(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].orm_rect.zw + decals.data[decal_index].orm_rect.xy, 0.0).xyz; + } + ao = mix(ao, decal_orm.r, decal_albedo.a); + roughness = mix(roughness, decal_orm.g, decal_albedo.a); + metallic = mix(metallic, decal_orm.b, decal_albedo.a); + } + } + + if (decals.data[decal_index].emission_rect != vec4(0.0)) { + //emission is additive, so its independent from albedo + if (sc_decal_use_mipmaps) { + emission += textureGrad(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].emission_rect.zw + decals.data[decal_index].emission_rect.xy, ddx * decals.data[decal_index].emission_rect.zw, ddy * decals.data[decal_index].emission_rect.zw).xyz * decals.data[decal_index].modulate.rgb * decals.data[decal_index].emission_energy * fade; + } else { + emission += textureLod(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].emission_rect.zw + decals.data[decal_index].emission_rect.xy, 0.0).xyz * decals.data[decal_index].modulate.rgb * decals.data[decal_index].emission_energy * fade; + } + } + } + } + } + + //pack albedo until needed again, saves 2 VGPRs in the meantime + +#endif //not render depth + /////////////////////// LIGHTING ////////////////////////////// + +#ifdef NORMAL_USED + if (scene_data.roughness_limiter_enabled) { + //https://www.jp.square-enix.com/tech/library/pdf/ImprovedGeometricSpecularAA.pdf + float roughness2 = roughness * roughness; + vec3 dndu = dFdx(normal), dndv = dFdy(normal); + float variance = scene_data.roughness_limiter_amount * (dot(dndu, dndu) + dot(dndv, dndv)); + float kernelRoughness2 = min(2.0 * variance, scene_data.roughness_limiter_limit); //limit effect + float filteredRoughness2 = min(1.0, roughness2 + kernelRoughness2); + roughness = sqrt(filteredRoughness2); + } +#endif + //apply energy conservation + + vec3 specular_light = vec3(0.0, 0.0, 0.0); + vec3 diffuse_light = vec3(0.0, 0.0, 0.0); + vec3 ambient_light = vec3(0.0, 0.0, 0.0); + +#ifndef MODE_UNSHADED + // Used in regular draw pass and when drawing SDFs for SDFGI and materials for VoxelGI. + emission *= scene_data.emissive_exposure_normalization; +#endif + +#if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + + if (scene_data.use_reflection_cubemap) { +#ifdef LIGHT_ANISOTROPY_USED + // https://google.github.io/filament/Filament.html#lighting/imagebasedlights/anisotropy + vec3 anisotropic_direction = anisotropy >= 0.0 ? binormal : tangent; + vec3 anisotropic_tangent = cross(anisotropic_direction, view); + vec3 anisotropic_normal = cross(anisotropic_tangent, anisotropic_direction); + vec3 bent_normal = normalize(mix(normal, anisotropic_normal, abs(anisotropy) * clamp(5.0 * roughness, 0.0, 1.0))); + vec3 ref_vec = reflect(-view, bent_normal); + ref_vec = mix(ref_vec, bent_normal, roughness * roughness); +#else + vec3 ref_vec = reflect(-view, normal); + ref_vec = mix(ref_vec, normal, roughness * roughness); +#endif + + float horizon = min(1.0 + dot(ref_vec, normal), 1.0); + ref_vec = scene_data.radiance_inverse_xform * ref_vec; +#ifdef USE_RADIANCE_CUBEMAP_ARRAY + + float lod, blend; + blend = modf(roughness * MAX_ROUGHNESS_LOD, lod); + specular_light = texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(ref_vec, lod)).rgb; + specular_light = mix(specular_light, texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(ref_vec, lod + 1)).rgb, blend); + +#else + specular_light = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), ref_vec, roughness * MAX_ROUGHNESS_LOD).rgb; + +#endif //USE_RADIANCE_CUBEMAP_ARRAY + specular_light *= scene_data.IBL_exposure_normalization; + specular_light *= horizon * horizon; + specular_light *= scene_data.ambient_light_color_energy.a; + } + +#if defined(CUSTOM_RADIANCE_USED) + specular_light = mix(specular_light, custom_radiance.rgb, custom_radiance.a); +#endif + +#ifndef USE_LIGHTMAP + //lightmap overrides everything + if (scene_data.use_ambient_light) { + ambient_light = scene_data.ambient_light_color_energy.rgb; + + if (scene_data.use_ambient_cubemap) { + vec3 ambient_dir = scene_data.radiance_inverse_xform * normal; +#ifdef USE_RADIANCE_CUBEMAP_ARRAY + vec3 cubemap_ambient = texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(ambient_dir, MAX_ROUGHNESS_LOD)).rgb; +#else + vec3 cubemap_ambient = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), ambient_dir, MAX_ROUGHNESS_LOD).rgb; +#endif //USE_RADIANCE_CUBEMAP_ARRAY + cubemap_ambient *= scene_data.IBL_exposure_normalization; + ambient_light = mix(ambient_light, cubemap_ambient * scene_data.ambient_light_color_energy.a, scene_data.ambient_color_sky_mix); + } + } +#endif // USE_LIGHTMAP +#if defined(CUSTOM_IRRADIANCE_USED) + ambient_light = mix(ambient_light, custom_irradiance.rgb, custom_irradiance.a); +#endif + +#ifdef LIGHT_CLEARCOAT_USED + + if (scene_data.use_reflection_cubemap) { + vec3 n = normalize(normal_interp); // We want to use geometric normal, not normal_map + float NoV = max(dot(n, view), 0.0001); + vec3 ref_vec = reflect(-view, n); + // The clear coat layer assumes an IOR of 1.5 (4% reflectance) + float Fc = clearcoat * (0.04 + 0.96 * SchlickFresnel(NoV)); + float attenuation = 1.0 - Fc; + ambient_light *= attenuation; + specular_light *= attenuation; + + ref_vec = mix(ref_vec, n, clearcoat_roughness * clearcoat_roughness); + float horizon = min(1.0 + dot(ref_vec, normal), 1.0); + ref_vec = scene_data.radiance_inverse_xform * ref_vec; + float roughness_lod = mix(0.001, 0.1, clearcoat_roughness) * MAX_ROUGHNESS_LOD; +#ifdef USE_RADIANCE_CUBEMAP_ARRAY + + float lod, blend; + blend = modf(roughness_lod, lod); + vec3 clearcoat_light = texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(ref_vec, lod)).rgb; + clearcoat_light = mix(clearcoat_light, texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(ref_vec, lod + 1)).rgb, blend); + +#else + vec3 clearcoat_light = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), ref_vec, roughness_lod).rgb; + +#endif //USE_RADIANCE_CUBEMAP_ARRAY + specular_light += clearcoat_light * horizon * horizon * Fc * scene_data.ambient_light_color_energy.a; + } +#endif +#endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + + //radiance + +/// GI /// +#if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + +#ifdef USE_LIGHTMAP + + //lightmap + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_LIGHTMAP_CAPTURE)) { //has lightmap capture + uint index = instances.data[instance_index].gi_offset; + + vec3 wnormal = mat3(scene_data.inv_view_matrix) * normal; + const float c1 = 0.429043; + const float c2 = 0.511664; + const float c3 = 0.743125; + const float c4 = 0.886227; + const float c5 = 0.247708; + ambient_light += (c1 * lightmap_captures.data[index].sh[8].rgb * (wnormal.x * wnormal.x - wnormal.y * wnormal.y) + + c3 * lightmap_captures.data[index].sh[6].rgb * wnormal.z * wnormal.z + + c4 * lightmap_captures.data[index].sh[0].rgb - + c5 * lightmap_captures.data[index].sh[6].rgb + + 2.0 * c1 * lightmap_captures.data[index].sh[4].rgb * wnormal.x * wnormal.y + + 2.0 * c1 * lightmap_captures.data[index].sh[7].rgb * wnormal.x * wnormal.z + + 2.0 * c1 * lightmap_captures.data[index].sh[5].rgb * wnormal.y * wnormal.z + + 2.0 * c2 * lightmap_captures.data[index].sh[3].rgb * wnormal.x + + 2.0 * c2 * lightmap_captures.data[index].sh[1].rgb * wnormal.y + + 2.0 * c2 * lightmap_captures.data[index].sh[2].rgb * wnormal.z) * + scene_data.emissive_exposure_normalization; + + } else if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_LIGHTMAP)) { // has actual lightmap + bool uses_sh = bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_SH_LIGHTMAP); + uint ofs = instances.data[instance_index].gi_offset & 0xFFFF; + vec3 uvw; + uvw.xy = uv2 * instances.data[instance_index].lightmap_uv_scale.zw + instances.data[instance_index].lightmap_uv_scale.xy; + uvw.z = float((instances.data[instance_index].gi_offset >> 16) & 0xFFFF); + + if (uses_sh) { + uvw.z *= 4.0; //SH textures use 4 times more data + vec3 lm_light_l0 = textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw + vec3(0.0, 0.0, 0.0), 0.0).rgb; + vec3 lm_light_l1n1 = textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw + vec3(0.0, 0.0, 1.0), 0.0).rgb; + vec3 lm_light_l1_0 = textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw + vec3(0.0, 0.0, 2.0), 0.0).rgb; + vec3 lm_light_l1p1 = textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw + vec3(0.0, 0.0, 3.0), 0.0).rgb; + + uint idx = instances.data[instance_index].gi_offset >> 20; + vec3 n = normalize(lightmaps.data[idx].normal_xform * normal); + float en = lightmaps.data[idx].exposure_normalization; + + ambient_light += lm_light_l0 * 0.282095f * en; + ambient_light += lm_light_l1n1 * 0.32573 * n.y * en; + ambient_light += lm_light_l1_0 * 0.32573 * n.z * en; + ambient_light += lm_light_l1p1 * 0.32573 * n.x * en; + if (metallic > 0.01) { // since the more direct bounced light is lost, we can kind of fake it with this trick + vec3 r = reflect(normalize(-vertex), normal); + specular_light += lm_light_l1n1 * 0.32573 * r.y * en; + specular_light += lm_light_l1_0 * 0.32573 * r.z * en; + specular_light += lm_light_l1p1 * 0.32573 * r.x * en; + } + + } else { + uint idx = instances.data[instance_index].gi_offset >> 20; + ambient_light += textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw, 0.0).rgb * lightmaps.data[idx].exposure_normalization; + } + } +#else + + if (sc_use_forward_gi && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_SDFGI)) { //has lightmap capture + + //make vertex orientation the world one, but still align to camera + vec3 cam_pos = mat3(scene_data.inv_view_matrix) * vertex; + vec3 cam_normal = mat3(scene_data.inv_view_matrix) * normal; + vec3 cam_reflection = mat3(scene_data.inv_view_matrix) * reflect(-view, normal); + + //apply y-mult + cam_pos.y *= sdfgi.y_mult; + cam_normal.y *= sdfgi.y_mult; + cam_normal = normalize(cam_normal); + cam_reflection.y *= sdfgi.y_mult; + cam_normal = normalize(cam_normal); + cam_reflection = normalize(cam_reflection); + + vec4 light_accum = vec4(0.0); + float weight_accum = 0.0; + + vec4 light_blend_accum = vec4(0.0); + float weight_blend_accum = 0.0; + + float blend = -1.0; + + // helper constants, compute once + + uint cascade = 0xFFFFFFFF; + vec3 cascade_pos; + vec3 cascade_normal; + + for (uint i = 0; i < sdfgi.max_cascades; i++) { + cascade_pos = (cam_pos - sdfgi.cascades[i].position) * sdfgi.cascades[i].to_probe; + + if (any(lessThan(cascade_pos, vec3(0.0))) || any(greaterThanEqual(cascade_pos, sdfgi.cascade_probe_size))) { + continue; //skip cascade + } + + cascade = i; + break; + } + + if (cascade < SDFGI_MAX_CASCADES) { + bool use_specular = true; + float blend; + vec3 diffuse, specular; + sdfgi_process(cascade, cascade_pos, cam_pos, cam_normal, cam_reflection, use_specular, roughness, diffuse, specular, blend); + + if (blend > 0.0) { + //blend + if (cascade == sdfgi.max_cascades - 1) { + diffuse = mix(diffuse, ambient_light, blend); + if (use_specular) { + specular = mix(specular, specular_light, blend); + } + } else { + vec3 diffuse2, specular2; + float blend2; + cascade_pos = (cam_pos - sdfgi.cascades[cascade + 1].position) * sdfgi.cascades[cascade + 1].to_probe; + sdfgi_process(cascade + 1, cascade_pos, cam_pos, cam_normal, cam_reflection, use_specular, roughness, diffuse2, specular2, blend2); + diffuse = mix(diffuse, diffuse2, blend); + if (use_specular) { + specular = mix(specular, specular2, blend); + } + } + } + + ambient_light = diffuse; + if (use_specular) { + specular_light = specular; + } + } + } + + if (sc_use_forward_gi && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_VOXEL_GI)) { // process voxel_gi_instances + + uint index1 = instances.data[instance_index].gi_offset & 0xFFFF; + vec3 ref_vec = normalize(reflect(-view, normal)); + ref_vec = mix(ref_vec, normal, roughness * roughness); + //find arbitrary tangent and bitangent, then build a matrix + vec3 v0 = abs(normal.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 1.0, 0.0); + vec3 tangent = normalize(cross(v0, normal)); + vec3 bitangent = normalize(cross(tangent, normal)); + mat3 normal_mat = mat3(tangent, bitangent, normal); + + vec4 amb_accum = vec4(0.0); + vec4 spec_accum = vec4(0.0); + voxel_gi_compute(index1, vertex, normal, ref_vec, normal_mat, roughness * roughness, ambient_light, specular_light, spec_accum, amb_accum); + + uint index2 = instances.data[instance_index].gi_offset >> 16; + + if (index2 != 0xFFFF) { + voxel_gi_compute(index2, vertex, normal, ref_vec, normal_mat, roughness * roughness, ambient_light, specular_light, spec_accum, amb_accum); + } + + if (amb_accum.a > 0.0) { + amb_accum.rgb /= amb_accum.a; + } + + if (spec_accum.a > 0.0) { + spec_accum.rgb /= spec_accum.a; + } + + specular_light = spec_accum.rgb; + ambient_light = amb_accum.rgb; + } + + if (!sc_use_forward_gi && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_GI_BUFFERS)) { //use GI buffers + + vec2 coord; + + if (implementation_data.gi_upscale_for_msaa) { + vec2 base_coord = screen_uv; + vec2 closest_coord = base_coord; +#ifdef USE_MULTIVIEW + float closest_ang = dot(normal, textureLod(sampler2DArray(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), vec3(base_coord, ViewIndex), 0.0).xyz * 2.0 - 1.0); +#else // USE_MULTIVIEW + float closest_ang = dot(normal, textureLod(sampler2D(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), base_coord, 0.0).xyz * 2.0 - 1.0); +#endif // USE_MULTIVIEW + + for (int i = 0; i < 4; i++) { + const vec2 neighbours[4] = vec2[](vec2(-1, 0), vec2(1, 0), vec2(0, -1), vec2(0, 1)); + vec2 neighbour_coord = base_coord + neighbours[i] * scene_data.screen_pixel_size; +#ifdef USE_MULTIVIEW + float neighbour_ang = dot(normal, textureLod(sampler2DArray(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), vec3(neighbour_coord, ViewIndex), 0.0).xyz * 2.0 - 1.0); +#else // USE_MULTIVIEW + float neighbour_ang = dot(normal, textureLod(sampler2D(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), neighbour_coord, 0.0).xyz * 2.0 - 1.0); +#endif // USE_MULTIVIEW + if (neighbour_ang > closest_ang) { + closest_ang = neighbour_ang; + closest_coord = neighbour_coord; + } + } + + coord = closest_coord; + + } else { + coord = screen_uv; + } + +#ifdef USE_MULTIVIEW + vec4 buffer_ambient = textureLod(sampler2DArray(ambient_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), vec3(coord, ViewIndex), 0.0); + vec4 buffer_reflection = textureLod(sampler2DArray(reflection_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), vec3(coord, ViewIndex), 0.0); +#else // USE_MULTIVIEW + vec4 buffer_ambient = textureLod(sampler2D(ambient_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), coord, 0.0); + vec4 buffer_reflection = textureLod(sampler2D(reflection_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), coord, 0.0); +#endif // USE_MULTIVIEW + + ambient_light = mix(ambient_light, buffer_ambient.rgb, buffer_ambient.a); + specular_light = mix(specular_light, buffer_reflection.rgb, buffer_reflection.a); + } +#endif // !USE_LIGHTMAP + + if (bool(implementation_data.ss_effects_flags & SCREEN_SPACE_EFFECTS_FLAGS_USE_SSAO)) { + float ssao = texture(sampler2D(ao_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), screen_uv).r; + ao = min(ao, ssao); + ao_light_affect = mix(ao_light_affect, max(ao_light_affect, implementation_data.ssao_light_affect), implementation_data.ssao_ao_affect); + } + + { // process reflections + + vec4 reflection_accum = vec4(0.0, 0.0, 0.0, 0.0); + vec4 ambient_accum = vec4(0.0, 0.0, 0.0, 0.0); + + uint cluster_reflection_offset = cluster_offset + implementation_data.cluster_type_size * 3; + + uint item_min; + uint item_max; + uint item_from; + uint item_to; + + cluster_get_item_range(cluster_reflection_offset + implementation_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); + +#ifdef USE_SUBGROUPS + item_from = subgroupBroadcastFirst(subgroupMin(item_from)); + item_to = subgroupBroadcastFirst(subgroupMax(item_to)); +#endif + +#ifdef LIGHT_ANISOTROPY_USED + // https://google.github.io/filament/Filament.html#lighting/imagebasedlights/anisotropy + vec3 anisotropic_direction = anisotropy >= 0.0 ? binormal : tangent; + vec3 anisotropic_tangent = cross(anisotropic_direction, view); + vec3 anisotropic_normal = cross(anisotropic_tangent, anisotropic_direction); + vec3 bent_normal = normalize(mix(normal, anisotropic_normal, abs(anisotropy) * clamp(5.0 * roughness, 0.0, 1.0))); +#else + vec3 bent_normal = normal; +#endif + vec3 ref_vec = normalize(reflect(-view, bent_normal)); + ref_vec = mix(ref_vec, bent_normal, roughness * roughness); + + for (uint i = item_from; i < item_to; i++) { + uint mask = cluster_buffer.data[cluster_reflection_offset + i]; + mask &= cluster_get_range_clip_mask(i, item_min, item_max); +#ifdef USE_SUBGROUPS + uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); +#else + uint merged_mask = mask; +#endif + + while (merged_mask != 0) { + uint bit = findMSB(merged_mask); + merged_mask &= ~(1u << bit); +#ifdef USE_SUBGROUPS + if (((1u << bit) & mask) == 0) { //do not process if not originally here + continue; + } +#endif + uint reflection_index = 32 * i + bit; + + if (!bool(reflections.data[reflection_index].mask & instances.data[instance_index].layer_mask)) { + continue; //not masked + } + + reflection_process(reflection_index, vertex, ref_vec, normal, roughness, ambient_light, specular_light, ambient_accum, reflection_accum); + } + } + + if (reflection_accum.a > 0.0) { + specular_light = reflection_accum.rgb / reflection_accum.a; + } + +#if !defined(USE_LIGHTMAP) + if (ambient_accum.a > 0.0) { + ambient_light = ambient_accum.rgb / ambient_accum.a; + } +#endif + } + + //finalize ambient light here + { +#if defined(AMBIENT_LIGHT_DISABLED) + ambient_light = vec3(0.0, 0.0, 0.0); +#else + ambient_light *= albedo.rgb; + ambient_light *= ao; + + if (bool(implementation_data.ss_effects_flags & SCREEN_SPACE_EFFECTS_FLAGS_USE_SSIL)) { + vec4 ssil = textureLod(sampler2D(ssil_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), screen_uv, 0.0); + ambient_light *= 1.0 - ssil.a; + ambient_light += ssil.rgb * albedo.rgb; + } +#endif // AMBIENT_LIGHT_DISABLED + } + + // convert ao to direct light ao + ao = mix(1.0, ao, ao_light_affect); + + //this saves some VGPRs + vec3 f0 = F0(metallic, specular, albedo); + + { +#if defined(DIFFUSE_TOON) + //simplify for toon, as + specular_light *= specular * metallic * albedo * 2.0; +#else + + // scales the specular reflections, needs to be computed before lighting happens, + // but after environment, GI, and reflection probes are added + // Environment brdf approximation (Lazarov 2013) + // see https://www.unrealengine.com/en-US/blog/physically-based-shading-on-mobile + const vec4 c0 = vec4(-1.0, -0.0275, -0.572, 0.022); + const vec4 c1 = vec4(1.0, 0.0425, 1.04, -0.04); + vec4 r = roughness * c0 + c1; + float ndotv = clamp(dot(normal, view), 0.0, 1.0); + float a004 = min(r.x * r.x, exp2(-9.28 * ndotv)) * r.x + r.y; + vec2 env = vec2(-1.04, 1.04) * a004 + r.zw; + + specular_light *= env.x * f0 + env.y * clamp(50.0 * f0.g, 0.0, 1.0); +#endif + } + +#endif //GI !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + +#if !defined(MODE_RENDER_DEPTH) + //this saves some VGPRs + uint orms = packUnorm4x8(vec4(ao, roughness, metallic, specular)); +#endif + +// LIGHTING +#if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + + { // Directional light. + + // Do shadow and lighting in two passes to reduce register pressure. +#ifndef SHADOWS_DISABLED + uint shadow0 = 0; + uint shadow1 = 0; + + for (uint i = 0; i < 8; i++) { + if (i >= scene_data.directional_light_count) { + break; + } + + if (!bool(directional_lights.data[i].mask & instances.data[instance_index].layer_mask)) { + continue; //not masked + } + + if (directional_lights.data[i].bake_mode == LIGHT_BAKE_STATIC && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_LIGHTMAP)) { + continue; // Statically baked light and object uses lightmap, skip + } + + float shadow = 1.0; + + if (directional_lights.data[i].shadow_opacity > 0.001) { + float depth_z = -vertex.z; + vec3 light_dir = directional_lights.data[i].direction; + vec3 base_normal_bias = normalize(normal_interp) * (1.0 - max(0.0, dot(light_dir, -normalize(normal_interp)))); + +#define BIAS_FUNC(m_var, m_idx) \ + m_var.xyz += light_dir * directional_lights.data[i].shadow_bias[m_idx]; \ + vec3 normal_bias = base_normal_bias * directional_lights.data[i].shadow_normal_bias[m_idx]; \ + normal_bias -= light_dir * dot(light_dir, normal_bias); \ + m_var.xyz += normal_bias; + + //version with soft shadows, more expensive + if (sc_use_directional_soft_shadows && directional_lights.data[i].softshadow_angle > 0) { + uint blend_count = 0; + const uint blend_max = directional_lights.data[i].blend_splits ? 2 : 1; + + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 0) + + vec4 pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); + pssm_coord /= pssm_coord.w; + + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.x; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale1 * test_radius; + shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + blend_count++; + } + + if (blend_count < blend_max && depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 1) + + vec4 pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + pssm_coord /= pssm_coord.w; + + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.y; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale2 * test_radius; + float s = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + + if (blend_count == 0) { + shadow = s; + } else { + //blend + float blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); + shadow = mix(shadow, s, blend); + } + + blend_count++; + } + + if (blend_count < blend_max && depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 2) + + vec4 pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + pssm_coord /= pssm_coord.w; + + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.z; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale3 * test_radius; + float s = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + + if (blend_count == 0) { + shadow = s; + } else { + //blend + float blend = smoothstep(directional_lights.data[i].shadow_split_offsets.x, directional_lights.data[i].shadow_split_offsets.y, depth_z); + shadow = mix(shadow, s, blend); + } + + blend_count++; + } + + if (blend_count < blend_max) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 3) + + vec4 pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + pssm_coord /= pssm_coord.w; + + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.w; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale4 * test_radius; + float s = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + + if (blend_count == 0) { + shadow = s; + } else { + //blend + float blend = smoothstep(directional_lights.data[i].shadow_split_offsets.y, directional_lights.data[i].shadow_split_offsets.z, depth_z); + shadow = mix(shadow, s, blend); + } + } + + } else { //no soft shadows + + vec4 pssm_coord; + float blur_factor; + + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 0) + + pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); + blur_factor = 1.0; + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 1) + + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.y; + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 2) + + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.z; + } else { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 3) + + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.w; + } + + pssm_coord /= pssm_coord.w; + + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale * blur_factor, pssm_coord); + + if (directional_lights.data[i].blend_splits) { + float pssm_blend; + float blur_factor2; + + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 1) + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + pssm_blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor2 = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.y; + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 2) + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.x, directional_lights.data[i].shadow_split_offsets.y, depth_z); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor2 = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.z; + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 3) + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.y, directional_lights.data[i].shadow_split_offsets.z, depth_z); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor2 = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.w; + } else { + pssm_blend = 0.0; //if no blend, same coord will be used (divide by z will result in same value, and already cached) + blur_factor2 = 1.0; + } + + pssm_coord /= pssm_coord.w; + + float shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale * blur_factor2, pssm_coord); + shadow = mix(shadow, shadow2, pssm_blend); + } + } + + shadow = mix(shadow, 1.0, smoothstep(directional_lights.data[i].fade_from, directional_lights.data[i].fade_to, vertex.z)); //done with negative values for performance + +#undef BIAS_FUNC + } // shadows + + if (i < 4) { + shadow0 |= uint(clamp(shadow * 255.0, 0.0, 255.0)) << (i * 8); + } else { + shadow1 |= uint(clamp(shadow * 255.0, 0.0, 255.0)) << ((i - 4) * 8); + } + } +#endif // SHADOWS_DISABLED + + for (uint i = 0; i < 8; i++) { + if (i >= scene_data.directional_light_count) { + break; + } + + if (!bool(directional_lights.data[i].mask & instances.data[instance_index].layer_mask)) { + continue; //not masked + } + +#ifdef LIGHT_TRANSMITTANCE_USED + float transmittance_z = transmittance_depth; + + if (directional_lights.data[i].shadow_opacity > 0.001) { + float depth_z = -vertex.z; + + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.x, 1.0); + vec4 trans_coord = directional_lights.data[i].shadow_matrix1 * trans_vertex; + trans_coord /= trans_coord.w; + + float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; + shadow_z *= directional_lights.data[i].shadow_z_range.x; + float z = trans_coord.z * directional_lights.data[i].shadow_z_range.x; + + transmittance_z = z - shadow_z; + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.y, 1.0); + vec4 trans_coord = directional_lights.data[i].shadow_matrix2 * trans_vertex; + trans_coord /= trans_coord.w; + + float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; + shadow_z *= directional_lights.data[i].shadow_z_range.y; + float z = trans_coord.z * directional_lights.data[i].shadow_z_range.y; + + transmittance_z = z - shadow_z; + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.z, 1.0); + vec4 trans_coord = directional_lights.data[i].shadow_matrix3 * trans_vertex; + trans_coord /= trans_coord.w; + + float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; + shadow_z *= directional_lights.data[i].shadow_z_range.z; + float z = trans_coord.z * directional_lights.data[i].shadow_z_range.z; + + transmittance_z = z - shadow_z; + + } else { + vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.w, 1.0); + vec4 trans_coord = directional_lights.data[i].shadow_matrix4 * trans_vertex; + trans_coord /= trans_coord.w; + + float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; + shadow_z *= directional_lights.data[i].shadow_z_range.w; + float z = trans_coord.z * directional_lights.data[i].shadow_z_range.w; + + transmittance_z = z - shadow_z; + } + } +#endif + + float shadow = 1.0; +#ifndef SHADOWS_DISABLED + if (i < 4) { + shadow = float(shadow0 >> (i * 8u) & 0xFFu) / 255.0; + } else { + shadow = float(shadow1 >> ((i - 4u) * 8u) & 0xFFu) / 255.0; + } + + shadow = shadow * directional_lights.data[i].shadow_opacity + 1.0 - directional_lights.data[i].shadow_opacity; +#endif + + blur_shadow(shadow); + + float size_A = sc_use_light_soft_shadows ? directional_lights.data[i].size : 0.0; + + light_compute(normal, directional_lights.data[i].direction, normalize(view), size_A, directional_lights.data[i].color * directional_lights.data[i].energy, shadow, f0, orms, 1.0, albedo, alpha, +#ifdef LIGHT_BACKLIGHT_USED + backlight, +#endif +#ifdef LIGHT_TRANSMITTANCE_USED + transmittance_color, + transmittance_depth, + transmittance_boost, + transmittance_z, +#endif +#ifdef LIGHT_RIM_USED + rim, rim_tint, +#endif +#ifdef LIGHT_CLEARCOAT_USED + clearcoat, clearcoat_roughness, normalize(normal_interp), +#endif +#ifdef LIGHT_ANISOTROPY_USED + binormal, + tangent, anisotropy, +#endif + diffuse_light, + specular_light); + } + } + + { //omni lights + + uint cluster_omni_offset = cluster_offset; + + uint item_min; + uint item_max; + uint item_from; + uint item_to; + + cluster_get_item_range(cluster_omni_offset + implementation_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); + +#ifdef USE_SUBGROUPS + item_from = subgroupBroadcastFirst(subgroupMin(item_from)); + item_to = subgroupBroadcastFirst(subgroupMax(item_to)); +#endif + + for (uint i = item_from; i < item_to; i++) { + uint mask = cluster_buffer.data[cluster_omni_offset + i]; + mask &= cluster_get_range_clip_mask(i, item_min, item_max); +#ifdef USE_SUBGROUPS + uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); +#else + uint merged_mask = mask; +#endif + + while (merged_mask != 0) { + uint bit = findMSB(merged_mask); + merged_mask &= ~(1u << bit); +#ifdef USE_SUBGROUPS + if (((1u << bit) & mask) == 0) { //do not process if not originally here + continue; + } +#endif + uint light_index = 32 * i + bit; + + if (!bool(omni_lights.data[light_index].mask & instances.data[instance_index].layer_mask)) { + continue; //not masked + } + + if (omni_lights.data[light_index].bake_mode == LIGHT_BAKE_STATIC && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_LIGHTMAP)) { + continue; // Statically baked light and object uses lightmap, skip + } + + float shadow = light_process_omni_shadow(light_index, vertex, normal); + + shadow = blur_shadow(shadow); + + light_process_omni(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, albedo, alpha, +#ifdef LIGHT_BACKLIGHT_USED + backlight, +#endif +#ifdef LIGHT_TRANSMITTANCE_USED + transmittance_color, + transmittance_depth, + transmittance_boost, +#endif +#ifdef LIGHT_RIM_USED + rim, + rim_tint, +#endif +#ifdef LIGHT_CLEARCOAT_USED + clearcoat, clearcoat_roughness, normalize(normal_interp), +#endif +#ifdef LIGHT_ANISOTROPY_USED + tangent, binormal, anisotropy, +#endif + diffuse_light, specular_light); + } + } + } + + { //spot lights + + uint cluster_spot_offset = cluster_offset + implementation_data.cluster_type_size; + + uint item_min; + uint item_max; + uint item_from; + uint item_to; + + cluster_get_item_range(cluster_spot_offset + implementation_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); + +#ifdef USE_SUBGROUPS + item_from = subgroupBroadcastFirst(subgroupMin(item_from)); + item_to = subgroupBroadcastFirst(subgroupMax(item_to)); +#endif + + for (uint i = item_from; i < item_to; i++) { + uint mask = cluster_buffer.data[cluster_spot_offset + i]; + mask &= cluster_get_range_clip_mask(i, item_min, item_max); +#ifdef USE_SUBGROUPS + uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); +#else + uint merged_mask = mask; +#endif + + while (merged_mask != 0) { + uint bit = findMSB(merged_mask); + merged_mask &= ~(1u << bit); +#ifdef USE_SUBGROUPS + if (((1u << bit) & mask) == 0) { //do not process if not originally here + continue; + } +#endif + + uint light_index = 32 * i + bit; + + if (!bool(spot_lights.data[light_index].mask & instances.data[instance_index].layer_mask)) { + continue; //not masked + } + + if (spot_lights.data[light_index].bake_mode == LIGHT_BAKE_STATIC && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_LIGHTMAP)) { + continue; // Statically baked light and object uses lightmap, skip + } + + float shadow = light_process_spot_shadow(light_index, vertex, normal); + + shadow = blur_shadow(shadow); + + light_process_spot(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, albedo, alpha, +#ifdef LIGHT_BACKLIGHT_USED + backlight, +#endif +#ifdef LIGHT_TRANSMITTANCE_USED + transmittance_color, + transmittance_depth, + transmittance_boost, +#endif +#ifdef LIGHT_RIM_USED + rim, + rim_tint, +#endif +#ifdef LIGHT_CLEARCOAT_USED + clearcoat, clearcoat_roughness, normalize(normal_interp), +#endif +#ifdef LIGHT_ANISOTROPY_USED + tangent, + binormal, anisotropy, +#endif + diffuse_light, specular_light); + } + } + } + +#ifdef USE_SHADOW_TO_OPACITY + alpha = min(alpha, clamp(length(ambient_light), 0.0, 1.0)); + +#if defined(ALPHA_SCISSOR_USED) + if (alpha < alpha_scissor) { + discard; + } +#endif // ALPHA_SCISSOR_USED + +#ifdef USE_OPAQUE_PREPASS + + if (alpha < scene_data.opaque_prepass_threshold) { + discard; + } + +#endif // USE_OPAQUE_PREPASS + +#endif // USE_SHADOW_TO_OPACITY + +#endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + +#ifdef MODE_RENDER_DEPTH + +#ifdef MODE_RENDER_SDF + + { + vec3 local_pos = (implementation_data.sdf_to_bounds * vec4(vertex, 1.0)).xyz; + ivec3 grid_pos = implementation_data.sdf_offset + ivec3(local_pos * vec3(implementation_data.sdf_size)); + + uint albedo16 = 0x1; //solid flag + albedo16 |= clamp(uint(albedo.r * 31.0), 0, 31) << 11; + albedo16 |= clamp(uint(albedo.g * 31.0), 0, 31) << 6; + albedo16 |= clamp(uint(albedo.b * 31.0), 0, 31) << 1; + + imageStore(albedo_volume_grid, grid_pos, uvec4(albedo16)); + + uint facing_bits = 0; + const vec3 aniso_dir[6] = vec3[]( + vec3(1, 0, 0), + vec3(0, 1, 0), + vec3(0, 0, 1), + vec3(-1, 0, 0), + vec3(0, -1, 0), + vec3(0, 0, -1)); + + vec3 cam_normal = mat3(scene_data.inv_view_matrix) * normalize(normal_interp); + + float closest_dist = -1e20; + + for (uint i = 0; i < 6; i++) { + float d = dot(cam_normal, aniso_dir[i]); + if (d > closest_dist) { + closest_dist = d; + facing_bits = (1 << i); + } + } + +#ifdef MOLTENVK_USED + imageStore(geom_facing_grid, grid_pos, uvec4(imageLoad(geom_facing_grid, grid_pos).r | facing_bits)); //store facing bits +#else + imageAtomicOr(geom_facing_grid, grid_pos, facing_bits); //store facing bits +#endif + + if (length(emission) > 0.001) { + float lumas[6]; + vec3 light_total = vec3(0); + + for (int i = 0; i < 6; i++) { + float strength = max(0.0, dot(cam_normal, aniso_dir[i])); + vec3 light = emission * strength; + light_total += light; + lumas[i] = max(light.r, max(light.g, light.b)); + } + + float luma_total = max(light_total.r, max(light_total.g, light_total.b)); + + uint light_aniso = 0; + + for (int i = 0; i < 6; i++) { + light_aniso |= min(31, uint((lumas[i] / luma_total) * 31.0)) << (i * 5); + } + + //compress to RGBE9995 to save space + + const float pow2to9 = 512.0f; + const float B = 15.0f; + const float N = 9.0f; + const float LN2 = 0.6931471805599453094172321215; + + float cRed = clamp(light_total.r, 0.0, 65408.0); + float cGreen = clamp(light_total.g, 0.0, 65408.0); + float cBlue = clamp(light_total.b, 0.0, 65408.0); + + float cMax = max(cRed, max(cGreen, cBlue)); + + float expp = max(-B - 1.0f, floor(log(cMax) / LN2)) + 1.0f + B; + + float sMax = floor((cMax / pow(2.0f, expp - B - N)) + 0.5f); + + float exps = expp + 1.0f; + + if (0.0 <= sMax && sMax < pow2to9) { + exps = expp; + } + + float sRed = floor((cRed / pow(2.0f, exps - B - N)) + 0.5f); + float sGreen = floor((cGreen / pow(2.0f, exps - B - N)) + 0.5f); + float sBlue = floor((cBlue / pow(2.0f, exps - B - N)) + 0.5f); + //store as 8985 to have 2 extra neighbour bits + uint light_rgbe = ((uint(sRed) & 0x1FFu) >> 1) | ((uint(sGreen) & 0x1FFu) << 8) | (((uint(sBlue) & 0x1FFu) >> 1) << 17) | ((uint(exps) & 0x1Fu) << 25); + + imageStore(emission_grid, grid_pos, uvec4(light_rgbe)); + imageStore(emission_aniso_grid, grid_pos, uvec4(light_aniso)); + } + } + +#endif + +#ifdef MODE_RENDER_MATERIAL + + albedo_output_buffer.rgb = albedo; + albedo_output_buffer.a = alpha; + + normal_output_buffer.rgb = normal * 0.5 + 0.5; + normal_output_buffer.a = 0.0; + depth_output_buffer.r = -vertex.z; + + orm_output_buffer.r = ao; + orm_output_buffer.g = roughness; + orm_output_buffer.b = metallic; + orm_output_buffer.a = sss_strength; + + emission_output_buffer.rgb = emission; + emission_output_buffer.a = 0.0; +#endif + +#ifdef MODE_RENDER_NORMAL_ROUGHNESS + normal_roughness_output_buffer = vec4(normal * 0.5 + 0.5, roughness); + +#ifdef MODE_RENDER_VOXEL_GI + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_VOXEL_GI)) { // process voxel_gi_instances + uint index1 = instances.data[instance_index].gi_offset & 0xFFFF; + uint index2 = instances.data[instance_index].gi_offset >> 16; + voxel_gi_buffer.x = index1 & 0xFFu; + voxel_gi_buffer.y = index2 & 0xFFu; + } else { + voxel_gi_buffer.x = 0xFF; + voxel_gi_buffer.y = 0xFF; + } +#endif + +#endif //MODE_RENDER_NORMAL_ROUGHNESS + +//nothing happens, so a tree-ssa optimizer will result in no fragment shader :) +#else + + // multiply by albedo + diffuse_light *= albedo; // ambient must be multiplied by albedo at the end + + // apply direct light AO + ao = unpackUnorm4x8(orms).x; + specular_light *= ao; + diffuse_light *= ao; + + // apply metallic + metallic = unpackUnorm4x8(orms).z; + diffuse_light *= 1.0 - metallic; + ambient_light *= 1.0 - metallic; + + //restore fog + fog = vec4(unpackHalf2x16(fog_rg), unpackHalf2x16(fog_ba)); + +#ifdef MODE_SEPARATE_SPECULAR + +#ifdef MODE_UNSHADED + diffuse_buffer = vec4(albedo.rgb, 0.0); + specular_buffer = vec4(0.0); + +#else + +#ifdef SSS_MODE_SKIN + sss_strength = -sss_strength; +#endif + diffuse_buffer = vec4(emission + diffuse_light + ambient_light, sss_strength); + specular_buffer = vec4(specular_light, metallic); +#endif + + diffuse_buffer.rgb = mix(diffuse_buffer.rgb, fog.rgb, fog.a); + specular_buffer.rgb = mix(specular_buffer.rgb, vec3(0.0), fog.a); + +#else //MODE_SEPARATE_SPECULAR + +#ifdef MODE_UNSHADED + frag_color = vec4(albedo, alpha); +#else + frag_color = vec4(emission + ambient_light + diffuse_light + specular_light, alpha); +//frag_color = vec4(1.0); +#endif //USE_NO_SHADING + + // Draw "fixed" fog before volumetric fog to ensure volumetric fog can appear in front of the sky. + frag_color.rgb = mix(frag_color.rgb, fog.rgb, fog.a); + +#endif //MODE_SEPARATE_SPECULAR + +#endif //MODE_RENDER_DEPTH +#ifdef MOTION_VECTORS + vec2 position_clip = (screen_position.xy / screen_position.w) - scene_data.taa_jitter; + vec2 prev_position_clip = (prev_screen_position.xy / prev_screen_position.w) - scene_data_block.prev_data.taa_jitter; + + vec2 position_uv = position_clip * vec2(0.5, 0.5); + vec2 prev_position_uv = prev_position_clip * vec2(0.5, 0.5); + + motion_vector = position_uv - prev_position_uv; +#endif +} + +void main() { +#ifdef MODE_DUAL_PARABOLOID + + if (dp_clip > 0.0) + discard; +#endif + + fragment_shader(scene_data_block.data); +} diff --git a/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered_inc.glsl b/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered_inc.glsl new file mode 100644 index 0000000000..3a45ab0059 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered_inc.glsl @@ -0,0 +1,322 @@ +#define M_PI 3.14159265359 +#define ROUGHNESS_MAX_LOD 5 + +#define MAX_VOXEL_GI_INSTANCES 8 +#define MAX_VIEWS 2 + +#ifndef MOLTENVK_USED +#if defined(has_GL_KHR_shader_subgroup_ballot) && defined(has_GL_KHR_shader_subgroup_arithmetic) + +#extension GL_KHR_shader_subgroup_ballot : enable +#extension GL_KHR_shader_subgroup_arithmetic : enable + +#define USE_SUBGROUPS +#endif +#endif // MOLTENVK_USED + +#if defined(USE_MULTIVIEW) && defined(has_VK_KHR_multiview) +#extension GL_EXT_multiview : enable +#endif + +#include "../cluster_data_inc.glsl" +#include "../decal_data_inc.glsl" +#include "../scene_data_inc.glsl" + +#if !defined(MODE_RENDER_DEPTH) || defined(MODE_RENDER_MATERIAL) || defined(MODE_RENDER_SDF) || defined(MODE_RENDER_NORMAL_ROUGHNESS) || defined(MODE_RENDER_VOXEL_GI) || defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) +#ifndef NORMAL_USED +#define NORMAL_USED +#endif +#endif + +layout(push_constant, std430) uniform DrawCall { + uint instance_index; + uint uv_offset; + uint multimesh_motion_vectors_current_offset; + uint multimesh_motion_vectors_previous_offset; +} +draw_call; + +#define SDFGI_MAX_CASCADES 8 + +/* Set 0: Base Pass (never changes) */ + +#include "../light_data_inc.glsl" + +#define SAMPLER_NEAREST_CLAMP 0 +#define SAMPLER_LINEAR_CLAMP 1 +#define SAMPLER_NEAREST_WITH_MIPMAPS_CLAMP 2 +#define SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP 3 +#define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_CLAMP 4 +#define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_CLAMP 5 +#define SAMPLER_NEAREST_REPEAT 6 +#define SAMPLER_LINEAR_REPEAT 7 +#define SAMPLER_NEAREST_WITH_MIPMAPS_REPEAT 8 +#define SAMPLER_LINEAR_WITH_MIPMAPS_REPEAT 9 +#define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_REPEAT 10 +#define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_REPEAT 11 + +layout(set = 0, binding = 1) uniform sampler material_samplers[12]; + +layout(set = 0, binding = 2) uniform sampler shadow_sampler; + +layout(set = 0, binding = 3) uniform sampler decal_sampler; + +layout(set = 0, binding = 4) uniform sampler light_projector_sampler; + +#define INSTANCE_FLAGS_NON_UNIFORM_SCALE (1 << 4) +#define INSTANCE_FLAGS_USE_GI_BUFFERS (1 << 5) +#define INSTANCE_FLAGS_USE_SDFGI (1 << 6) +#define INSTANCE_FLAGS_USE_LIGHTMAP_CAPTURE (1 << 7) +#define INSTANCE_FLAGS_USE_LIGHTMAP (1 << 8) +#define INSTANCE_FLAGS_USE_SH_LIGHTMAP (1 << 9) +#define INSTANCE_FLAGS_USE_VOXEL_GI (1 << 10) +#define INSTANCE_FLAGS_PARTICLES (1 << 11) +#define INSTANCE_FLAGS_MULTIMESH (1 << 12) +#define INSTANCE_FLAGS_MULTIMESH_FORMAT_2D (1 << 13) +#define INSTANCE_FLAGS_MULTIMESH_HAS_COLOR (1 << 14) +#define INSTANCE_FLAGS_MULTIMESH_HAS_CUSTOM_DATA (1 << 15) +#define INSTANCE_FLAGS_PARTICLE_TRAIL_SHIFT 16 +#define INSTANCE_FLAGS_FADE_SHIFT 24 +//3 bits of stride +#define INSTANCE_FLAGS_PARTICLE_TRAIL_MASK 0xFF + +#define SCREEN_SPACE_EFFECTS_FLAGS_USE_SSAO 1 +#define SCREEN_SPACE_EFFECTS_FLAGS_USE_SSIL 2 + +layout(set = 0, binding = 5, std430) restrict readonly buffer OmniLights { + LightData data[]; +} +omni_lights; + +layout(set = 0, binding = 6, std430) restrict readonly buffer SpotLights { + LightData data[]; +} +spot_lights; + +layout(set = 0, binding = 7, std430) restrict readonly buffer ReflectionProbeData { + ReflectionData data[]; +} +reflections; + +layout(set = 0, binding = 8, std140) uniform DirectionalLights { + DirectionalLightData data[MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS]; +} +directional_lights; + +#define LIGHTMAP_FLAG_USE_DIRECTION 1 +#define LIGHTMAP_FLAG_USE_SPECULAR_DIRECTION 2 + +struct Lightmap { + mat3 normal_xform; + vec3 pad; + float exposure_normalization; +}; + +layout(set = 0, binding = 9, std140) restrict readonly buffer Lightmaps { + Lightmap data[]; +} +lightmaps; + +struct LightmapCapture { + vec4 sh[9]; +}; + +layout(set = 0, binding = 10, std140) restrict readonly buffer LightmapCaptures { + LightmapCapture data[]; +} +lightmap_captures; + +layout(set = 0, binding = 11) uniform texture2D decal_atlas; +layout(set = 0, binding = 12) uniform texture2D decal_atlas_srgb; + +layout(set = 0, binding = 13, std430) restrict readonly buffer Decals { + DecalData data[]; +} +decals; + +layout(set = 0, binding = 14, std430) restrict readonly buffer GlobalShaderUniformData { + vec4 data[]; +} +global_shader_uniforms; + +struct SDFVoxelGICascadeData { + vec3 position; + float to_probe; + ivec3 probe_world_offset; + float to_cell; // 1/bounds * grid_size + vec3 pad; + float exposure_normalization; +}; + +layout(set = 0, binding = 15, std140) uniform SDFGI { + vec3 grid_size; + uint max_cascades; + + bool use_occlusion; + int probe_axis_size; + float probe_to_uvw; + float normal_bias; + + vec3 lightprobe_tex_pixel_size; + float energy; + + vec3 lightprobe_uv_offset; + float y_mult; + + vec3 occlusion_clamp; + uint pad3; + + vec3 occlusion_renormalize; + uint pad4; + + vec3 cascade_probe_size; + uint pad5; + + SDFVoxelGICascadeData cascades[SDFGI_MAX_CASCADES]; +} +sdfgi; + +/* Set 1: Render Pass (changes per render pass) */ + +layout(set = 1, binding = 0, std140) uniform SceneDataBlock { + SceneData data; + SceneData prev_data; +} +scene_data_block; + +struct ImplementationData { + uint cluster_shift; + uint cluster_width; + uint cluster_type_size; + uint max_cluster_element_count_div_32; + + uint ss_effects_flags; + float ssao_light_affect; + float ssao_ao_affect; + uint pad1; + + mat4 sdf_to_bounds; + + ivec3 sdf_offset; + uint pad2; + + ivec3 sdf_size; + bool gi_upscale_for_msaa; + + bool volumetric_fog_enabled; + float volumetric_fog_inv_length; + float volumetric_fog_detail_spread; + uint volumetric_fog_pad; +}; + +layout(set = 1, binding = 1, std140) uniform ImplementationDataBlock { + ImplementationData data; +} +implementation_data_block; + +#define implementation_data implementation_data_block.data + +struct InstanceData { + mat4 transform; + mat4 prev_transform; + uint flags; + uint instance_uniforms_ofs; //base offset in global buffer for instance variables + uint gi_offset; //GI information when using lightmapping (VCT or lightmap index) + uint layer_mask; + vec4 lightmap_uv_scale; +}; + +layout(set = 1, binding = 2, std430) buffer restrict readonly InstanceDataBuffer { + InstanceData data[]; +} +instances; + +#ifdef USE_RADIANCE_CUBEMAP_ARRAY + +layout(set = 1, binding = 3) uniform textureCubeArray radiance_cubemap; + +#else + +layout(set = 1, binding = 3) uniform textureCube radiance_cubemap; + +#endif + +layout(set = 1, binding = 4) uniform textureCubeArray reflection_atlas; + +layout(set = 1, binding = 5) uniform texture2D shadow_atlas; + +layout(set = 1, binding = 6) uniform texture2D directional_shadow_atlas; + +layout(set = 1, binding = 7) uniform texture2DArray lightmap_textures[MAX_LIGHTMAP_TEXTURES]; + +layout(set = 1, binding = 8) uniform texture3D voxel_gi_textures[MAX_VOXEL_GI_INSTANCES]; + +layout(set = 1, binding = 9, std430) buffer restrict readonly ClusterBuffer { + uint data[]; +} +cluster_buffer; + +#ifdef MODE_RENDER_SDF + +layout(r16ui, set = 1, binding = 10) uniform restrict writeonly uimage3D albedo_volume_grid; +layout(r32ui, set = 1, binding = 11) uniform restrict writeonly uimage3D emission_grid; +layout(r32ui, set = 1, binding = 12) uniform restrict writeonly uimage3D emission_aniso_grid; +layout(r32ui, set = 1, binding = 13) uniform restrict uimage3D geom_facing_grid; + +//still need to be present for shaders that use it, so remap them to something +#define depth_buffer shadow_atlas +#define color_buffer shadow_atlas +#define normal_roughness_buffer shadow_atlas + +#else + +layout(set = 1, binding = 10) uniform texture2D depth_buffer; +layout(set = 1, binding = 11) uniform texture2D color_buffer; + +#ifdef USE_MULTIVIEW +layout(set = 1, binding = 12) uniform texture2DArray normal_roughness_buffer; +layout(set = 1, binding = 14) uniform texture2DArray ambient_buffer; +layout(set = 1, binding = 15) uniform texture2DArray reflection_buffer; +#else // USE_MULTIVIEW +layout(set = 1, binding = 12) uniform texture2D normal_roughness_buffer; +layout(set = 1, binding = 14) uniform texture2D ambient_buffer; +layout(set = 1, binding = 15) uniform texture2D reflection_buffer; +#endif +layout(set = 1, binding = 13) uniform texture2D ao_buffer; +layout(set = 1, binding = 16) uniform texture2DArray sdfgi_lightprobe_texture; +layout(set = 1, binding = 17) uniform texture3D sdfgi_occlusion_cascades; + +struct VoxelGIData { + mat4 xform; // 64 - 64 + + vec3 bounds; // 12 - 76 + float dynamic_range; // 4 - 80 + + float bias; // 4 - 84 + float normal_bias; // 4 - 88 + bool blend_ambient; // 4 - 92 + uint mipmaps; // 4 - 96 + + vec3 pad; // 12 - 108 + float exposure_normalization; // 4 - 112 +}; + +layout(set = 1, binding = 18, std140) uniform VoxelGIs { + VoxelGIData data[MAX_VOXEL_GI_INSTANCES]; +} +voxel_gi_instances; + +layout(set = 1, binding = 19) uniform texture3D volumetric_fog_texture; + +layout(set = 1, binding = 20) uniform texture2D ssil_buffer; + +#endif + +/* Set 2 Skeleton & Instancing (can change per item) */ + +layout(set = 2, binding = 0, std430) restrict readonly buffer Transforms { + vec4 data[]; +} +transforms; + +/* Set 3 User Material */ diff --git a/servers/rendering/renderer_rd/shaders/forward_mobile/SCsub b/servers/rendering/renderer_rd/shaders/forward_mobile/SCsub new file mode 100644 index 0000000000..f06a2d86e2 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/forward_mobile/SCsub @@ -0,0 +1,17 @@ +#!/usr/bin/env python + +Import("env") + +if "RD_GLSL" in env["BUILDERS"]: + # find all include files + gl_include_files = [str(f) for f in Glob("*_inc.glsl")] + [str(f) for f in Glob("../*_inc.glsl")] + + # find all shader code(all glsl files excluding our include files) + glsl_files = [str(f) for f in Glob("*.glsl") if str(f) not in gl_include_files] + + # make sure we recompile shaders if include files change + env.Depends([f + ".gen.h" for f in glsl_files], gl_include_files + ["#glsl_builders.py"]) + + # compile shaders + for glsl_file in glsl_files: + env.RD_GLSL(glsl_file) diff --git a/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile.glsl b/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile.glsl new file mode 100644 index 0000000000..9aeaa6d978 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile.glsl @@ -0,0 +1,1733 @@ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +/* Include our forward mobile UBOs definitions etc. */ +#include "scene_forward_mobile_inc.glsl" + +#define SHADER_IS_SRGB false + +/* INPUT ATTRIBS */ + +layout(location = 0) in vec3 vertex_attrib; + +//only for pure render depth when normal is not used + +#ifdef NORMAL_USED +layout(location = 1) in vec2 normal_attrib; +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) +layout(location = 2) in vec2 tangent_attrib; +#endif + +#if defined(COLOR_USED) +layout(location = 3) in vec4 color_attrib; +#endif + +#ifdef UV_USED +layout(location = 4) in vec2 uv_attrib; +#endif + +#if defined(UV2_USED) || defined(USE_LIGHTMAP) || defined(MODE_RENDER_MATERIAL) +layout(location = 5) in vec2 uv2_attrib; +#endif // MODE_RENDER_MATERIAL + +#if defined(CUSTOM0_USED) +layout(location = 6) in vec4 custom0_attrib; +#endif + +#if defined(CUSTOM1_USED) +layout(location = 7) in vec4 custom1_attrib; +#endif + +#if defined(CUSTOM2_USED) +layout(location = 8) in vec4 custom2_attrib; +#endif + +#if defined(CUSTOM3_USED) +layout(location = 9) in vec4 custom3_attrib; +#endif + +#if defined(BONES_USED) || defined(USE_PARTICLE_TRAILS) +layout(location = 10) in uvec4 bone_attrib; +#endif + +#if defined(WEIGHTS_USED) || defined(USE_PARTICLE_TRAILS) +layout(location = 11) in vec4 weight_attrib; +#endif + +vec3 oct_to_vec3(vec2 e) { + vec3 v = vec3(e.xy, 1.0 - abs(e.x) - abs(e.y)); + float t = max(-v.z, 0.0); + v.xy += t * -sign(v.xy); + return v; +} + +/* Varyings */ + +layout(location = 0) highp out vec3 vertex_interp; + +#ifdef NORMAL_USED +layout(location = 1) mediump out vec3 normal_interp; +#endif + +#if defined(COLOR_USED) +layout(location = 2) mediump out vec4 color_interp; +#endif + +#ifdef UV_USED +layout(location = 3) mediump out vec2 uv_interp; +#endif + +#if defined(UV2_USED) || defined(USE_LIGHTMAP) +layout(location = 4) mediump out vec2 uv2_interp; +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) +layout(location = 5) mediump out vec3 tangent_interp; +layout(location = 6) mediump out vec3 binormal_interp; +#endif + +#ifdef MATERIAL_UNIFORMS_USED +layout(set = MATERIAL_UNIFORM_SET, binding = 0, std140) uniform MaterialUniforms{ + +#MATERIAL_UNIFORMS + +} material; +#endif + +#ifdef MODE_DUAL_PARABOLOID + +layout(location = 9) out highp float dp_clip; + +#endif + +#ifdef USE_MULTIVIEW +#ifdef has_VK_KHR_multiview +#define ViewIndex gl_ViewIndex +#else +// !BAS! This needs to become an input once we implement our fallback! +#define ViewIndex 0 +#endif +#else +// Set to zero, not supported in non stereo +#define ViewIndex 0 +#endif //USE_MULTIVIEW + +invariant gl_Position; + +#GLOBALS + +#define scene_data scene_data_block.data + +#ifdef USE_DOUBLE_PRECISION +// Helper functions for emulating double precision when adding floats. +vec3 quick_two_sum(vec3 a, vec3 b, out vec3 out_p) { + vec3 s = a + b; + out_p = b - (s - a); + return s; +} + +vec3 two_sum(vec3 a, vec3 b, out vec3 out_p) { + vec3 s = a + b; + vec3 v = s - a; + out_p = (a - (s - v)) + (b - v); + return s; +} + +vec3 double_add_vec3(vec3 base_a, vec3 prec_a, vec3 base_b, vec3 prec_b, out vec3 out_precision) { + vec3 s, t, se, te; + s = two_sum(base_a, base_b, se); + t = two_sum(prec_a, prec_b, te); + se += t; + s = quick_two_sum(s, se, se); + se += te; + s = quick_two_sum(s, se, out_precision); + return s; +} +#endif + +void main() { + vec4 instance_custom = vec4(0.0); +#if defined(COLOR_USED) + color_interp = color_attrib; +#endif + + bool is_multimesh = bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH); + + mat4 model_matrix = draw_call.transform; + mat4 inv_view_matrix = scene_data.inv_view_matrix; +#ifdef USE_DOUBLE_PRECISION + vec3 model_precision = vec3(model_matrix[0][3], model_matrix[1][3], model_matrix[2][3]); + model_matrix[0][3] = 0.0; + model_matrix[1][3] = 0.0; + model_matrix[2][3] = 0.0; + vec3 view_precision = vec3(inv_view_matrix[0][3], inv_view_matrix[1][3], inv_view_matrix[2][3]); + inv_view_matrix[0][3] = 0.0; + inv_view_matrix[1][3] = 0.0; + inv_view_matrix[2][3] = 0.0; +#endif + + mat3 model_normal_matrix; + if (bool(draw_call.flags & INSTANCE_FLAGS_NON_UNIFORM_SCALE)) { + model_normal_matrix = transpose(inverse(mat3(model_matrix))); + } else { + model_normal_matrix = mat3(model_matrix); + } + + mat4 matrix; + mat4 read_model_matrix = model_matrix; + + if (is_multimesh) { + //multimesh, instances are for it + +#ifdef USE_PARTICLE_TRAILS + uint trail_size = (draw_call.flags >> INSTANCE_FLAGS_PARTICLE_TRAIL_SHIFT) & INSTANCE_FLAGS_PARTICLE_TRAIL_MASK; + uint stride = 3 + 1 + 1; //particles always uses this format + + uint offset = trail_size * stride * gl_InstanceIndex; + +#ifdef COLOR_USED + vec4 pcolor; +#endif + { + uint boffset = offset + bone_attrib.x * stride; + matrix = mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], transforms.data[boffset + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weight_attrib.x; +#ifdef COLOR_USED + pcolor = transforms.data[boffset + 3] * weight_attrib.x; +#endif + } + if (weight_attrib.y > 0.001) { + uint boffset = offset + bone_attrib.y * stride; + matrix += mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], transforms.data[boffset + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weight_attrib.y; +#ifdef COLOR_USED + pcolor += transforms.data[boffset + 3] * weight_attrib.y; +#endif + } + if (weight_attrib.z > 0.001) { + uint boffset = offset + bone_attrib.z * stride; + matrix += mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], transforms.data[boffset + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weight_attrib.z; +#ifdef COLOR_USED + pcolor += transforms.data[boffset + 3] * weight_attrib.z; +#endif + } + if (weight_attrib.w > 0.001) { + uint boffset = offset + bone_attrib.w * stride; + matrix += mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], transforms.data[boffset + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weight_attrib.w; +#ifdef COLOR_USED + pcolor += transforms.data[boffset + 3] * weight_attrib.w; +#endif + } + + instance_custom = transforms.data[offset + 4]; + +#ifdef COLOR_USED + color_interp *= pcolor; +#endif + +#else + uint stride = 0; + { + //TODO implement a small lookup table for the stride + if (bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH_FORMAT_2D)) { + stride += 2; + } else { + stride += 3; + } + if (bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH_HAS_COLOR)) { + stride += 1; + } + if (bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH_HAS_CUSTOM_DATA)) { + stride += 1; + } + } + + uint offset = stride * gl_InstanceIndex; + + if (bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH_FORMAT_2D)) { + matrix = mat4(transforms.data[offset + 0], transforms.data[offset + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0)); + offset += 2; + } else { + matrix = mat4(transforms.data[offset + 0], transforms.data[offset + 1], transforms.data[offset + 2], vec4(0.0, 0.0, 0.0, 1.0)); + offset += 3; + } + + if (bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH_HAS_COLOR)) { +#ifdef COLOR_USED + color_interp *= transforms.data[offset]; +#endif + offset += 1; + } + + if (bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH_HAS_CUSTOM_DATA)) { + instance_custom = transforms.data[offset]; + } + +#endif + //transpose + matrix = transpose(matrix); + +#if !defined(USE_DOUBLE_PRECISION) || defined(SKIP_TRANSFORM_USED) || defined(VERTEX_WORLD_COORDS_USED) || defined(MODEL_MATRIX_USED) + // Normally we can bake the multimesh transform into the model matrix, but when using double precision + // we avoid baking it in so we can emulate high precision. + read_model_matrix = model_matrix * matrix; +#if !defined(USE_DOUBLE_PRECISION) || defined(SKIP_TRANSFORM_USED) || defined(VERTEX_WORLD_COORDS_USED) + model_matrix = read_model_matrix; +#endif // !defined(USE_DOUBLE_PRECISION) || defined(SKIP_TRANSFORM_USED) || defined(VERTEX_WORLD_COORDS_USED) +#endif // !defined(USE_DOUBLE_PRECISION) || defined(SKIP_TRANSFORM_USED) || defined(VERTEX_WORLD_COORDS_USED) || defined(MODEL_MATRIX_USED) + model_normal_matrix = model_normal_matrix * mat3(matrix); + } + + vec3 vertex = vertex_attrib; +#ifdef NORMAL_USED + vec3 normal = oct_to_vec3(normal_attrib * 2.0 - 1.0); +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) + vec2 signed_tangent_attrib = tangent_attrib * 2.0 - 1.0; + vec3 tangent = oct_to_vec3(vec2(signed_tangent_attrib.x, abs(signed_tangent_attrib.y) * 2.0 - 1.0)); + float binormalf = sign(signed_tangent_attrib.y); + vec3 binormal = normalize(cross(normal, tangent) * binormalf); +#endif + +#ifdef UV_USED + uv_interp = uv_attrib; +#endif + +#if defined(UV2_USED) || defined(USE_LIGHTMAP) + uv2_interp = uv2_attrib; +#endif + +#ifdef OVERRIDE_POSITION + vec4 position; +#endif + +#ifdef USE_MULTIVIEW + mat4 projection_matrix = scene_data.projection_matrix_view[ViewIndex]; + mat4 inv_projection_matrix = scene_data.inv_projection_matrix_view[ViewIndex]; +#else + mat4 projection_matrix = scene_data.projection_matrix; + mat4 inv_projection_matrix = scene_data.inv_projection_matrix; +#endif //USE_MULTIVIEW + +//using world coordinates +#if !defined(SKIP_TRANSFORM_USED) && defined(VERTEX_WORLD_COORDS_USED) + + vertex = (model_matrix * vec4(vertex, 1.0)).xyz; + +#ifdef NORMAL_USED + normal = model_normal_matrix * normal; +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) + + tangent = model_normal_matrix * tangent; + binormal = model_normal_matrix * binormal; + +#endif +#endif + + float roughness = 1.0; + + mat4 modelview = scene_data.view_matrix * model_matrix; + mat3 modelview_normal = mat3(scene_data.view_matrix) * model_normal_matrix; + + { +#CODE : VERTEX + } + + /* output */ + +// using local coordinates (default) +#if !defined(SKIP_TRANSFORM_USED) && !defined(VERTEX_WORLD_COORDS_USED) + +#ifdef USE_DOUBLE_PRECISION + // We separate the basis from the origin because the basis is fine with single point precision. + // Then we combine the translations from the model matrix and the view matrix using emulated doubles. + // We add the result to the vertex and ignore the final lost precision. + vec3 model_origin = model_matrix[3].xyz; + if (is_multimesh) { + vertex = mat3(matrix) * vertex; + model_origin = double_add_vec3(model_origin, model_precision, matrix[3].xyz, vec3(0.0), model_precision); + } + vertex = mat3(model_matrix) * vertex; + vec3 temp_precision; + vertex += double_add_vec3(model_origin, model_precision, scene_data.inv_view_matrix[3].xyz, view_precision, temp_precision); + vertex = mat3(scene_data.view_matrix) * vertex; +#else + vertex = (modelview * vec4(vertex, 1.0)).xyz; +#endif +#ifdef NORMAL_USED + normal = modelview_normal * normal; +#endif + +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) + + binormal = modelview_normal * binormal; + tangent = modelview_normal * tangent; +#endif + +//using world coordinates +#if !defined(SKIP_TRANSFORM_USED) && defined(VERTEX_WORLD_COORDS_USED) + + vertex = (scene_data.view_matrix * vec4(vertex, 1.0)).xyz; +#ifdef NORMAL_USED + normal = (scene_data.view_matrix * vec4(normal, 0.0)).xyz; +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) + binormal = (scene_data.view_matrix * vec4(binormal, 0.0)).xyz; + tangent = (scene_data.view_matrix * vec4(tangent, 0.0)).xyz; +#endif +#endif + + vertex_interp = vertex; +#ifdef NORMAL_USED + normal_interp = normal; +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) + tangent_interp = tangent; + binormal_interp = binormal; +#endif + +#ifdef MODE_RENDER_DEPTH + +#ifdef MODE_DUAL_PARABOLOID + + vertex_interp.z *= scene_data.dual_paraboloid_side; + + dp_clip = vertex_interp.z; //this attempts to avoid noise caused by objects sent to the other parabolloid side due to bias + + //for dual paraboloid shadow mapping, this is the fastest but least correct way, as it curves straight edges + + vec3 vtx = vertex_interp; + float distance = length(vtx); + vtx = normalize(vtx); + vtx.xy /= 1.0 - vtx.z; + vtx.z = (distance / scene_data.z_far); + vtx.z = vtx.z * 2.0 - 1.0; + vertex_interp = vtx; + +#endif + +#endif //MODE_RENDER_DEPTH + +#ifdef OVERRIDE_POSITION + gl_Position = position; +#else + gl_Position = projection_matrix * vec4(vertex_interp, 1.0); +#endif // OVERRIDE_POSITION + +#ifdef MODE_RENDER_DEPTH + if (scene_data.pancake_shadows) { + if (gl_Position.z <= 0.00001) { + gl_Position.z = 0.00001; + } + } +#endif // MODE_RENDER_DEPTH +#ifdef MODE_RENDER_MATERIAL + if (scene_data.material_uv2_mode) { + vec2 uv_offset = draw_call.lightmap_uv_scale.xy; // we are abusing lightmap_uv_scale here, we shouldn't have a lightmap during a depth pass... + gl_Position.xy = (uv2_attrib.xy + uv_offset) * 2.0 - 1.0; + gl_Position.z = 0.00001; + gl_Position.w = 1.0; + } +#endif // MODE_RENDER_MATERIAL +} + +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#define SHADER_IS_SRGB false + +/* Specialization Constants */ + +#if !defined(MODE_RENDER_DEPTH) + +#if !defined(MODE_UNSHADED) + +layout(constant_id = 0) const bool sc_use_light_projector = false; +layout(constant_id = 1) const bool sc_use_light_soft_shadows = false; +layout(constant_id = 2) const bool sc_use_directional_soft_shadows = false; + +layout(constant_id = 3) const uint sc_soft_shadow_samples = 4; +layout(constant_id = 4) const uint sc_penumbra_shadow_samples = 4; + +layout(constant_id = 5) const uint sc_directional_soft_shadow_samples = 4; +layout(constant_id = 6) const uint sc_directional_penumbra_shadow_samples = 4; + +layout(constant_id = 8) const bool sc_projector_use_mipmaps = true; + +layout(constant_id = 9) const bool sc_disable_omni_lights = false; +layout(constant_id = 10) const bool sc_disable_spot_lights = false; +layout(constant_id = 11) const bool sc_disable_reflection_probes = false; +layout(constant_id = 12) const bool sc_disable_directional_lights = false; + +#endif //!MODE_UNSHADED + +layout(constant_id = 7) const bool sc_decal_use_mipmaps = true; +layout(constant_id = 13) const bool sc_disable_decals = false; +layout(constant_id = 14) const bool sc_disable_fog = false; + +#endif //!MODE_RENDER_DEPTH + +layout(constant_id = 15) const float sc_luminance_multiplier = 2.0; + +/* Include our forward mobile UBOs definitions etc. */ +#include "scene_forward_mobile_inc.glsl" + +/* Varyings */ + +layout(location = 0) highp in vec3 vertex_interp; + +#ifdef NORMAL_USED +layout(location = 1) mediump in vec3 normal_interp; +#endif + +#if defined(COLOR_USED) +layout(location = 2) mediump in vec4 color_interp; +#endif + +#ifdef UV_USED +layout(location = 3) mediump in vec2 uv_interp; +#endif + +#if defined(UV2_USED) || defined(USE_LIGHTMAP) +layout(location = 4) mediump in vec2 uv2_interp; +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) +layout(location = 5) mediump in vec3 tangent_interp; +layout(location = 6) mediump in vec3 binormal_interp; +#endif + +#ifdef MODE_DUAL_PARABOLOID + +layout(location = 9) highp in float dp_clip; + +#endif + +#ifdef USE_MULTIVIEW +#ifdef has_VK_KHR_multiview +#define ViewIndex gl_ViewIndex +#else +// !BAS! This needs to become an input once we implement our fallback! +#define ViewIndex 0 +#endif +#else +// Set to zero, not supported in non stereo +#define ViewIndex 0 +#endif //USE_MULTIVIEW + +//defines to keep compatibility with vertex + +#ifdef USE_MULTIVIEW +#define projection_matrix scene_data.projection_matrix_view[ViewIndex] +#define inv_projection_matrix scene_data.inv_projection_matrix_view[ViewIndex] +#else +#define projection_matrix scene_data.projection_matrix +#define inv_projection_matrix scene_data.inv_projection_matrix +#endif + +#if defined(ENABLE_SSS) && defined(ENABLE_TRANSMITTANCE) +//both required for transmittance to be enabled +#define LIGHT_TRANSMITTANCE_USED +#endif + +#ifdef MATERIAL_UNIFORMS_USED +layout(set = MATERIAL_UNIFORM_SET, binding = 0, std140) uniform MaterialUniforms{ + +#MATERIAL_UNIFORMS + +} material; +#endif + +#GLOBALS + +/* clang-format on */ + +#ifdef MODE_RENDER_DEPTH + +#ifdef MODE_RENDER_MATERIAL + +layout(location = 0) out vec4 albedo_output_buffer; +layout(location = 1) out vec4 normal_output_buffer; +layout(location = 2) out vec4 orm_output_buffer; +layout(location = 3) out vec4 emission_output_buffer; +layout(location = 4) out float depth_output_buffer; + +#endif // MODE_RENDER_MATERIAL + +#else // RENDER DEPTH + +#ifdef MODE_MULTIPLE_RENDER_TARGETS + +layout(location = 0) out vec4 diffuse_buffer; //diffuse (rgb) and roughness +layout(location = 1) out vec4 specular_buffer; //specular and SSS (subsurface scatter) +#else + +layout(location = 0) out mediump vec4 frag_color; +#endif // MODE_MULTIPLE_RENDER_TARGETS + +#endif // RENDER DEPTH + +#include "../scene_forward_aa_inc.glsl" + +#if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + +// Default to SPECULAR_SCHLICK_GGX. +#if !defined(SPECULAR_DISABLED) && !defined(SPECULAR_SCHLICK_GGX) && !defined(SPECULAR_TOON) +#define SPECULAR_SCHLICK_GGX +#endif + +#include "../scene_forward_lights_inc.glsl" + +#endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + +#ifndef MODE_RENDER_DEPTH + +/* + Only supporting normal fog here. +*/ + +vec4 fog_process(vec3 vertex) { + vec3 fog_color = scene_data_block.data.fog_light_color; + + if (scene_data_block.data.fog_aerial_perspective > 0.0) { + vec3 sky_fog_color = vec3(0.0); + vec3 cube_view = scene_data_block.data.radiance_inverse_xform * vertex; + // mip_level always reads from the second mipmap and higher so the fog is always slightly blurred + float mip_level = mix(1.0 / MAX_ROUGHNESS_LOD, 1.0, 1.0 - (abs(vertex.z) - scene_data_block.data.z_near) / (scene_data_block.data.z_far - scene_data_block.data.z_near)); +#ifdef USE_RADIANCE_CUBEMAP_ARRAY + float lod, blend; + blend = modf(mip_level * MAX_ROUGHNESS_LOD, lod); + sky_fog_color = texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(cube_view, lod)).rgb; + sky_fog_color = mix(sky_fog_color, texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(cube_view, lod + 1)).rgb, blend); +#else + sky_fog_color = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), cube_view, mip_level * MAX_ROUGHNESS_LOD).rgb; +#endif //USE_RADIANCE_CUBEMAP_ARRAY + fog_color = mix(fog_color, sky_fog_color, scene_data_block.data.fog_aerial_perspective); + } + + if (scene_data_block.data.fog_sun_scatter > 0.001) { + vec4 sun_scatter = vec4(0.0); + float sun_total = 0.0; + vec3 view = normalize(vertex); + + for (uint i = 0; i < scene_data_block.data.directional_light_count; i++) { + vec3 light_color = directional_lights.data[i].color * directional_lights.data[i].energy; + float light_amount = pow(max(dot(view, directional_lights.data[i].direction), 0.0), 8.0); + fog_color += light_color * light_amount * scene_data_block.data.fog_sun_scatter; + } + } + + float fog_amount = 1.0 - exp(min(0.0, -length(vertex) * scene_data_block.data.fog_density)); + + if (abs(scene_data_block.data.fog_height_density) >= 0.0001) { + float y = (scene_data_block.data.inv_view_matrix * vec4(vertex, 1.0)).y; + + float y_dist = y - scene_data_block.data.fog_height; + + float vfog_amount = 1.0 - exp(min(0.0, y_dist * scene_data_block.data.fog_height_density)); + + fog_amount = max(vfog_amount, fog_amount); + } + + return vec4(fog_color, fog_amount); +} + +#endif //!MODE_RENDER DEPTH + +#define scene_data scene_data_block.data + +void main() { +#ifdef MODE_DUAL_PARABOLOID + + if (dp_clip > 0.0) + discard; +#endif + + //lay out everything, whatever is unused is optimized away anyway + vec3 vertex = vertex_interp; +#ifdef USE_MULTIVIEW + vec3 view = -normalize(vertex_interp - scene_data.eye_offset[ViewIndex].xyz); +#else + vec3 view = -normalize(vertex_interp); +#endif + vec3 albedo = vec3(1.0); + vec3 backlight = vec3(0.0); + vec4 transmittance_color = vec4(0.0); + float transmittance_depth = 0.0; + float transmittance_boost = 0.0; + float metallic = 0.0; + float specular = 0.5; + vec3 emission = vec3(0.0); + float roughness = 1.0; + float rim = 0.0; + float rim_tint = 0.0; + float clearcoat = 0.0; + float clearcoat_roughness = 0.0; + float anisotropy = 0.0; + vec2 anisotropy_flow = vec2(1.0, 0.0); + vec4 fog = vec4(0.0); +#if defined(CUSTOM_RADIANCE_USED) + vec4 custom_radiance = vec4(0.0); +#endif +#if defined(CUSTOM_IRRADIANCE_USED) + vec4 custom_irradiance = vec4(0.0); +#endif + + float ao = 1.0; + float ao_light_affect = 0.0; + + float alpha = 1.0; + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) + vec3 binormal = normalize(binormal_interp); + vec3 tangent = normalize(tangent_interp); +#else + vec3 binormal = vec3(0.0); + vec3 tangent = vec3(0.0); +#endif + +#ifdef NORMAL_USED + vec3 normal = normalize(normal_interp); + +#if defined(DO_SIDE_CHECK) + if (!gl_FrontFacing) { + normal = -normal; + } +#endif + +#endif //NORMAL_USED + +#ifdef UV_USED + vec2 uv = uv_interp; +#endif + +#if defined(UV2_USED) || defined(USE_LIGHTMAP) + vec2 uv2 = uv2_interp; +#endif + +#if defined(COLOR_USED) + vec4 color = color_interp; +#endif + +#if defined(NORMAL_MAP_USED) + + vec3 normal_map = vec3(0.5); +#endif + + float normal_map_depth = 1.0; + + vec2 screen_uv = gl_FragCoord.xy * scene_data.screen_pixel_size; + + float sss_strength = 0.0; + +#ifdef ALPHA_SCISSOR_USED + float alpha_scissor_threshold = 1.0; +#endif // ALPHA_SCISSOR_USED + +#ifdef ALPHA_HASH_USED + float alpha_hash_scale = 1.0; +#endif // ALPHA_HASH_USED + +#ifdef ALPHA_ANTIALIASING_EDGE_USED + float alpha_antialiasing_edge = 0.0; + vec2 alpha_texture_coordinate = vec2(0.0, 0.0); +#endif // ALPHA_ANTIALIASING_EDGE_USED + + mat4 inv_view_matrix = scene_data.inv_view_matrix; + mat4 read_model_matrix = draw_call.transform; +#ifdef USE_DOUBLE_PRECISION + read_model_matrix[0][3] = 0.0; + read_model_matrix[1][3] = 0.0; + read_model_matrix[2][3] = 0.0; + inv_view_matrix[0][3] = 0.0; + inv_view_matrix[1][3] = 0.0; + inv_view_matrix[2][3] = 0.0; +#endif + + { +#CODE : FRAGMENT + } + +#ifdef LIGHT_TRANSMITTANCE_USED +#ifdef SSS_MODE_SKIN + transmittance_color.a = sss_strength; +#else + transmittance_color.a *= sss_strength; +#endif +#endif + +#ifndef USE_SHADOW_TO_OPACITY + +#ifdef ALPHA_SCISSOR_USED + if (alpha < alpha_scissor_threshold) { + discard; + } +#endif // ALPHA_SCISSOR_USED + +// alpha hash can be used in unison with alpha antialiasing +#ifdef ALPHA_HASH_USED + if (alpha < compute_alpha_hash_threshold(vertex, alpha_hash_scale)) { + discard; + } +#endif // ALPHA_HASH_USED + +// If we are not edge antialiasing, we need to remove the output alpha channel from scissor and hash +#if (defined(ALPHA_SCISSOR_USED) || defined(ALPHA_HASH_USED)) && !defined(ALPHA_ANTIALIASING_EDGE_USED) + alpha = 1.0; +#endif + +#ifdef ALPHA_ANTIALIASING_EDGE_USED +// If alpha scissor is used, we must further the edge threshold, otherwise we won't get any edge feather +#ifdef ALPHA_SCISSOR_USED + alpha_antialiasing_edge = clamp(alpha_scissor_threshold + alpha_antialiasing_edge, 0.0, 1.0); +#endif + alpha = compute_alpha_antialiasing_edge(alpha, alpha_texture_coordinate, alpha_antialiasing_edge); +#endif // ALPHA_ANTIALIASING_EDGE_USED + +#ifdef USE_OPAQUE_PREPASS + if (alpha < scene_data.opaque_prepass_threshold) { + discard; + } +#endif // USE_OPAQUE_PREPASS + +#endif // !USE_SHADOW_TO_OPACITY + +#ifdef NORMAL_MAP_USED + + normal_map.xy = normal_map.xy * 2.0 - 1.0; + normal_map.z = sqrt(max(0.0, 1.0 - dot(normal_map.xy, normal_map.xy))); //always ignore Z, as it can be RG packed, Z may be pos/neg, etc. + + normal = normalize(mix(normal, tangent * normal_map.x + binormal * normal_map.y + normal * normal_map.z, normal_map_depth)); + +#endif + +#ifdef LIGHT_ANISOTROPY_USED + + if (anisotropy > 0.01) { + //rotation matrix + mat3 rot = mat3(tangent, binormal, normal); + //make local to space + tangent = normalize(rot * vec3(anisotropy_flow.x, anisotropy_flow.y, 0.0)); + binormal = normalize(rot * vec3(-anisotropy_flow.y, anisotropy_flow.x, 0.0)); + } + +#endif + +#ifdef ENABLE_CLIP_ALPHA + if (albedo.a < 0.99) { + //used for doublepass and shadowmapping + discard; + } +#endif + + /////////////////////// FOG ////////////////////// +#ifndef MODE_RENDER_DEPTH + +#ifndef CUSTOM_FOG_USED + // fog must be processed as early as possible and then packed. + // to maximize VGPR usage + // Draw "fixed" fog before volumetric fog to ensure volumetric fog can appear in front of the sky. + + if (!sc_disable_fog && scene_data.fog_enabled) { + fog = fog_process(vertex); + } + +#endif //!CUSTOM_FOG_USED + + uint fog_rg = packHalf2x16(fog.rg); + uint fog_ba = packHalf2x16(fog.ba); + +#endif //!MODE_RENDER_DEPTH + + /////////////////////// DECALS //////////////////////////////// + +#ifndef MODE_RENDER_DEPTH + + vec3 vertex_ddx = dFdx(vertex); + vec3 vertex_ddy = dFdy(vertex); + + if (!sc_disable_decals) { //Decals + // must implement + + uint decal_indices = draw_call.decals.x; + for (uint i = 0; i < 8; i++) { + uint decal_index = decal_indices & 0xFF; + if (i == 4) { + decal_indices = draw_call.decals.y; + } else { + decal_indices = decal_indices >> 8; + } + + if (decal_index == 0xFF) { + break; + } + + vec3 uv_local = (decals.data[decal_index].xform * vec4(vertex, 1.0)).xyz; + if (any(lessThan(uv_local, vec3(0.0, -1.0, 0.0))) || any(greaterThan(uv_local, vec3(1.0)))) { + continue; //out of decal + } + + float fade = pow(1.0 - (uv_local.y > 0.0 ? uv_local.y : -uv_local.y), uv_local.y > 0.0 ? decals.data[decal_index].upper_fade : decals.data[decal_index].lower_fade); + + if (decals.data[decal_index].normal_fade > 0.0) { + fade *= smoothstep(decals.data[decal_index].normal_fade, 1.0, dot(normal_interp, decals.data[decal_index].normal) * 0.5 + 0.5); + } + + //we need ddx/ddy for mipmaps, so simulate them + vec2 ddx = (decals.data[decal_index].xform * vec4(vertex_ddx, 0.0)).xz; + vec2 ddy = (decals.data[decal_index].xform * vec4(vertex_ddy, 0.0)).xz; + + if (decals.data[decal_index].albedo_rect != vec4(0.0)) { + //has albedo + vec4 decal_albedo; + if (sc_decal_use_mipmaps) { + decal_albedo = textureGrad(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].albedo_rect.zw + decals.data[decal_index].albedo_rect.xy, ddx * decals.data[decal_index].albedo_rect.zw, ddy * decals.data[decal_index].albedo_rect.zw); + } else { + decal_albedo = textureLod(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].albedo_rect.zw + decals.data[decal_index].albedo_rect.xy, 0.0); + } + decal_albedo *= decals.data[decal_index].modulate; + decal_albedo.a *= fade; + albedo = mix(albedo, decal_albedo.rgb, decal_albedo.a * decals.data[decal_index].albedo_mix); + + if (decals.data[decal_index].normal_rect != vec4(0.0)) { + vec3 decal_normal; + if (sc_decal_use_mipmaps) { + decal_normal = textureGrad(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].normal_rect.zw + decals.data[decal_index].normal_rect.xy, ddx * decals.data[decal_index].normal_rect.zw, ddy * decals.data[decal_index].normal_rect.zw).xyz; + } else { + decal_normal = textureLod(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].normal_rect.zw + decals.data[decal_index].normal_rect.xy, 0.0).xyz; + } + decal_normal.xy = decal_normal.xy * vec2(2.0, -2.0) - vec2(1.0, -1.0); //users prefer flipped y normal maps in most authoring software + decal_normal.z = sqrt(max(0.0, 1.0 - dot(decal_normal.xy, decal_normal.xy))); + //convert to view space, use xzy because y is up + decal_normal = (decals.data[decal_index].normal_xform * decal_normal.xzy).xyz; + + normal = normalize(mix(normal, decal_normal, decal_albedo.a)); + } + + if (decals.data[decal_index].orm_rect != vec4(0.0)) { + vec3 decal_orm; + if (sc_decal_use_mipmaps) { + decal_orm = textureGrad(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].orm_rect.zw + decals.data[decal_index].orm_rect.xy, ddx * decals.data[decal_index].orm_rect.zw, ddy * decals.data[decal_index].orm_rect.zw).xyz; + } else { + decal_orm = textureLod(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].orm_rect.zw + decals.data[decal_index].orm_rect.xy, 0.0).xyz; + } + ao = mix(ao, decal_orm.r, decal_albedo.a); + roughness = mix(roughness, decal_orm.g, decal_albedo.a); + metallic = mix(metallic, decal_orm.b, decal_albedo.a); + } + } + + if (decals.data[decal_index].emission_rect != vec4(0.0)) { + //emission is additive, so its independent from albedo + if (sc_decal_use_mipmaps) { + emission += textureGrad(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].emission_rect.zw + decals.data[decal_index].emission_rect.xy, ddx * decals.data[decal_index].emission_rect.zw, ddy * decals.data[decal_index].emission_rect.zw).xyz * decals.data[decal_index].emission_energy * fade; + } else { + emission += textureLod(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].emission_rect.zw + decals.data[decal_index].emission_rect.xy, 0.0).xyz * decals.data[decal_index].emission_energy * fade; + } + } + } + } //Decals +#endif //!MODE_RENDER_DEPTH + + /////////////////////// LIGHTING ////////////////////////////// + +#ifdef NORMAL_USED + if (scene_data.roughness_limiter_enabled) { + //https://www.jp.square-enix.com/tech/library/pdf/ImprovedGeometricSpecularAA.pdf + float roughness2 = roughness * roughness; + vec3 dndu = dFdx(normal), dndv = dFdy(normal); + float variance = scene_data.roughness_limiter_amount * (dot(dndu, dndu) + dot(dndv, dndv)); + float kernelRoughness2 = min(2.0 * variance, scene_data.roughness_limiter_limit); //limit effect + float filteredRoughness2 = min(1.0, roughness2 + kernelRoughness2); + roughness = sqrt(filteredRoughness2); + } +#endif // NORMAL_USED + //apply energy conservation + + vec3 specular_light = vec3(0.0, 0.0, 0.0); + vec3 diffuse_light = vec3(0.0, 0.0, 0.0); + vec3 ambient_light = vec3(0.0, 0.0, 0.0); + +#ifndef MODE_UNSHADED + // Used in regular draw pass and when drawing SDFs for SDFGI and materials for VoxelGI. + emission *= scene_data.emissive_exposure_normalization; +#endif + +#if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + + if (scene_data.use_reflection_cubemap) { +#ifdef LIGHT_ANISOTROPY_USED + // https://google.github.io/filament/Filament.html#lighting/imagebasedlights/anisotropy + vec3 anisotropic_direction = anisotropy >= 0.0 ? binormal : tangent; + vec3 anisotropic_tangent = cross(anisotropic_direction, view); + vec3 anisotropic_normal = cross(anisotropic_tangent, anisotropic_direction); + vec3 bent_normal = normalize(mix(normal, anisotropic_normal, abs(anisotropy) * clamp(5.0 * roughness, 0.0, 1.0))); + vec3 ref_vec = reflect(-view, bent_normal); + ref_vec = mix(ref_vec, bent_normal, roughness * roughness); +#else + vec3 ref_vec = reflect(-view, normal); + ref_vec = mix(ref_vec, normal, roughness * roughness); +#endif + float horizon = min(1.0 + dot(ref_vec, normal), 1.0); + ref_vec = scene_data.radiance_inverse_xform * ref_vec; +#ifdef USE_RADIANCE_CUBEMAP_ARRAY + + float lod, blend; + blend = modf(roughness * MAX_ROUGHNESS_LOD, lod); + specular_light = texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(ref_vec, lod)).rgb; + specular_light = mix(specular_light, texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(ref_vec, lod + 1)).rgb, blend); + +#else // USE_RADIANCE_CUBEMAP_ARRAY + specular_light = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), ref_vec, roughness * MAX_ROUGHNESS_LOD).rgb; + +#endif //USE_RADIANCE_CUBEMAP_ARRAY + specular_light *= sc_luminance_multiplier; + specular_light *= scene_data.IBL_exposure_normalization; + specular_light *= horizon * horizon; + specular_light *= scene_data.ambient_light_color_energy.a; + } + +#if defined(CUSTOM_RADIANCE_USED) + specular_light = mix(specular_light, custom_radiance.rgb, custom_radiance.a); +#endif // CUSTOM_RADIANCE_USED + +#ifndef USE_LIGHTMAP + //lightmap overrides everything + if (scene_data.use_ambient_light) { + ambient_light = scene_data.ambient_light_color_energy.rgb; + + if (scene_data.use_ambient_cubemap) { + vec3 ambient_dir = scene_data.radiance_inverse_xform * normal; +#ifdef USE_RADIANCE_CUBEMAP_ARRAY + vec3 cubemap_ambient = texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(ambient_dir, MAX_ROUGHNESS_LOD)).rgb; +#else + vec3 cubemap_ambient = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), ambient_dir, MAX_ROUGHNESS_LOD).rgb; +#endif //USE_RADIANCE_CUBEMAP_ARRAY + cubemap_ambient *= sc_luminance_multiplier; + cubemap_ambient *= scene_data.IBL_exposure_normalization; + ambient_light = mix(ambient_light, cubemap_ambient * scene_data.ambient_light_color_energy.a, scene_data.ambient_color_sky_mix); + } + } +#endif // !USE_LIGHTMAP + +#if defined(CUSTOM_IRRADIANCE_USED) + ambient_light = mix(ambient_light, custom_irradiance.rgb, custom_irradiance.a); +#endif // CUSTOM_IRRADIANCE_USED +#ifdef LIGHT_CLEARCOAT_USED + + if (scene_data.use_reflection_cubemap) { + vec3 n = normalize(normal_interp); // We want to use geometric normal, not normal_map + float NoV = max(dot(n, view), 0.0001); + vec3 ref_vec = reflect(-view, n); + ref_vec = mix(ref_vec, n, clearcoat_roughness * clearcoat_roughness); + // The clear coat layer assumes an IOR of 1.5 (4% reflectance) + float Fc = clearcoat * (0.04 + 0.96 * SchlickFresnel(NoV)); + float attenuation = 1.0 - Fc; + ambient_light *= attenuation; + specular_light *= attenuation; + + float horizon = min(1.0 + dot(ref_vec, normal), 1.0); + ref_vec = scene_data.radiance_inverse_xform * ref_vec; + float roughness_lod = mix(0.001, 0.1, clearcoat_roughness) * MAX_ROUGHNESS_LOD; +#ifdef USE_RADIANCE_CUBEMAP_ARRAY + + float lod, blend; + blend = modf(roughness_lod, lod); + vec3 clearcoat_light = texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(ref_vec, lod)).rgb; + clearcoat_light = mix(clearcoat_light, texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(ref_vec, lod + 1)).rgb, blend); + +#else + vec3 clearcoat_light = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), ref_vec, roughness_lod).rgb; + +#endif //USE_RADIANCE_CUBEMAP_ARRAY + specular_light += clearcoat_light * horizon * horizon * Fc * scene_data.ambient_light_color_energy.a; + } +#endif +#endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + + //radiance + +#if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + +#ifdef USE_LIGHTMAP + + //lightmap + if (bool(draw_call.flags & INSTANCE_FLAGS_USE_LIGHTMAP_CAPTURE)) { //has lightmap capture + uint index = draw_call.gi_offset; + + vec3 wnormal = mat3(scene_data.inv_view_matrix) * normal; + const float c1 = 0.429043; + const float c2 = 0.511664; + const float c3 = 0.743125; + const float c4 = 0.886227; + const float c5 = 0.247708; + ambient_light += (c1 * lightmap_captures.data[index].sh[8].rgb * (wnormal.x * wnormal.x - wnormal.y * wnormal.y) + + c3 * lightmap_captures.data[index].sh[6].rgb * wnormal.z * wnormal.z + + c4 * lightmap_captures.data[index].sh[0].rgb - + c5 * lightmap_captures.data[index].sh[6].rgb + + 2.0 * c1 * lightmap_captures.data[index].sh[4].rgb * wnormal.x * wnormal.y + + 2.0 * c1 * lightmap_captures.data[index].sh[7].rgb * wnormal.x * wnormal.z + + 2.0 * c1 * lightmap_captures.data[index].sh[5].rgb * wnormal.y * wnormal.z + + 2.0 * c2 * lightmap_captures.data[index].sh[3].rgb * wnormal.x + + 2.0 * c2 * lightmap_captures.data[index].sh[1].rgb * wnormal.y + + 2.0 * c2 * lightmap_captures.data[index].sh[2].rgb * wnormal.z) * + scene_data.emissive_exposure_normalization; + + } else if (bool(draw_call.flags & INSTANCE_FLAGS_USE_LIGHTMAP)) { // has actual lightmap + bool uses_sh = bool(draw_call.flags & INSTANCE_FLAGS_USE_SH_LIGHTMAP); + uint ofs = draw_call.gi_offset & 0xFFFF; + vec3 uvw; + uvw.xy = uv2 * draw_call.lightmap_uv_scale.zw + draw_call.lightmap_uv_scale.xy; + uvw.z = float((draw_call.gi_offset >> 16) & 0xFFFF); + + uint idx = draw_call.gi_offset >> 20; + + if (uses_sh) { + uvw.z *= 4.0; //SH textures use 4 times more data + vec3 lm_light_l0 = textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw + vec3(0.0, 0.0, 0.0), 0.0).rgb; + vec3 lm_light_l1n1 = textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw + vec3(0.0, 0.0, 1.0), 0.0).rgb; + vec3 lm_light_l1_0 = textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw + vec3(0.0, 0.0, 2.0), 0.0).rgb; + vec3 lm_light_l1p1 = textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw + vec3(0.0, 0.0, 3.0), 0.0).rgb; + + vec3 n = normalize(lightmaps.data[idx].normal_xform * normal); + float exposure_normalization = lightmaps.data[idx].exposure_normalization; + + ambient_light += lm_light_l0 * 0.282095f; + ambient_light += lm_light_l1n1 * 0.32573 * n.y * exposure_normalization; + ambient_light += lm_light_l1_0 * 0.32573 * n.z * exposure_normalization; + ambient_light += lm_light_l1p1 * 0.32573 * n.x * exposure_normalization; + if (metallic > 0.01) { // since the more direct bounced light is lost, we can kind of fake it with this trick + vec3 r = reflect(normalize(-vertex), normal); + specular_light += lm_light_l1n1 * 0.32573 * r.y * exposure_normalization; + specular_light += lm_light_l1_0 * 0.32573 * r.z * exposure_normalization; + specular_light += lm_light_l1p1 * 0.32573 * r.x * exposure_normalization; + } + + } else { + ambient_light += textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw, 0.0).rgb * lightmaps.data[idx].exposure_normalization; + } + } + + // No GI nor non low end mode... + +#endif // USE_LIGHTMAP + + // skipping ssao, do we remove ssao totally? + + if (!sc_disable_reflection_probes) { //Reflection probes + vec4 reflection_accum = vec4(0.0, 0.0, 0.0, 0.0); + vec4 ambient_accum = vec4(0.0, 0.0, 0.0, 0.0); + + uint reflection_indices = draw_call.reflection_probes.x; + +#ifdef LIGHT_ANISOTROPY_USED + // https://google.github.io/filament/Filament.html#lighting/imagebasedlights/anisotropy + vec3 anisotropic_direction = anisotropy >= 0.0 ? binormal : tangent; + vec3 anisotropic_tangent = cross(anisotropic_direction, view); + vec3 anisotropic_normal = cross(anisotropic_tangent, anisotropic_direction); + vec3 bent_normal = normalize(mix(normal, anisotropic_normal, abs(anisotropy) * clamp(5.0 * roughness, 0.0, 1.0))); +#else + vec3 bent_normal = normal; +#endif + vec3 ref_vec = normalize(reflect(-view, bent_normal)); + ref_vec = mix(ref_vec, bent_normal, roughness * roughness); + + for (uint i = 0; i < 8; i++) { + uint reflection_index = reflection_indices & 0xFF; + if (i == 4) { + reflection_indices = draw_call.reflection_probes.y; + } else { + reflection_indices = reflection_indices >> 8; + } + + if (reflection_index == 0xFF) { + break; + } + + reflection_process(reflection_index, vertex, ref_vec, bent_normal, roughness, ambient_light, specular_light, ambient_accum, reflection_accum); + } + + if (reflection_accum.a > 0.0) { + specular_light = reflection_accum.rgb / reflection_accum.a; + } + +#if !defined(USE_LIGHTMAP) + if (ambient_accum.a > 0.0) { + ambient_light = ambient_accum.rgb / ambient_accum.a; + } +#endif + } //Reflection probes + + // finalize ambient light here + { +#if defined(AMBIENT_LIGHT_DISABLED) + ambient_light = vec3(0.0, 0.0, 0.0); +#else + ambient_light *= albedo.rgb; + ambient_light *= ao; +#endif // AMBIENT_LIGHT_DISABLED + } + + // convert ao to direct light ao + ao = mix(1.0, ao, ao_light_affect); + + //this saves some VGPRs + vec3 f0 = F0(metallic, specular, albedo); + + { +#if defined(DIFFUSE_TOON) + //simplify for toon, as + specular_light *= specular * metallic * albedo * 2.0; +#else + + // scales the specular reflections, needs to be computed before lighting happens, + // but after environment, GI, and reflection probes are added + // Environment brdf approximation (Lazarov 2013) + // see https://www.unrealengine.com/en-US/blog/physically-based-shading-on-mobile + const vec4 c0 = vec4(-1.0, -0.0275, -0.572, 0.022); + const vec4 c1 = vec4(1.0, 0.0425, 1.04, -0.04); + vec4 r = roughness * c0 + c1; + float ndotv = clamp(dot(normal, view), 0.0, 1.0); + float a004 = min(r.x * r.x, exp2(-9.28 * ndotv)) * r.x + r.y; + vec2 env = vec2(-1.04, 1.04) * a004 + r.zw; + + specular_light *= env.x * f0 + env.y; +#endif + } + +#endif // !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + +#if !defined(MODE_RENDER_DEPTH) + //this saves some VGPRs + uint orms = packUnorm4x8(vec4(ao, roughness, metallic, specular)); +#endif + +// LIGHTING +#if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + + if (!sc_disable_directional_lights) { //directional light +#ifndef SHADOWS_DISABLED + // Do shadow and lighting in two passes to reduce register pressure + uint shadow0 = 0; + uint shadow1 = 0; + + for (uint i = 0; i < 8; i++) { + if (i >= scene_data.directional_light_count) { + break; + } + + if (!bool(directional_lights.data[i].mask & draw_call.layer_mask)) { + continue; //not masked + } + + float shadow = 1.0; + + // Directional light shadow code is basically the same as forward clustered at this point in time minus `LIGHT_TRANSMITTANCE_USED` support. + // Not sure if there is a reason to change this seeing directional lights are part of our global data + // Should think about whether we may want to move this code into an include file or function?? + +#ifdef USE_SOFT_SHADOWS + //version with soft shadows, more expensive + if (directional_lights.data[i].shadow_opacity > 0.001) { + float depth_z = -vertex.z; + + vec4 pssm_coord; + vec3 light_dir = directional_lights.data[i].direction; + +#define BIAS_FUNC(m_var, m_idx) \ + m_var.xyz += light_dir * directional_lights.data[i].shadow_bias[m_idx]; \ + vec3 normal_bias = normalize(normal_interp) * (1.0 - max(0.0, dot(light_dir, -normalize(normal_interp)))) * directional_lights.data[i].shadow_normal_bias[m_idx]; \ + normal_bias -= light_dir * dot(light_dir, normal_bias); \ + m_var.xyz += normal_bias; + + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 0) + + pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); + pssm_coord /= pssm_coord.w; + + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.x; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale1 * test_radius; + shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 1) + + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + pssm_coord /= pssm_coord.w; + + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.y; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale2 * test_radius; + shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 2) + + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + pssm_coord /= pssm_coord.w; + + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.z; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale3 * test_radius; + shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + } else { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 3) + + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + pssm_coord /= pssm_coord.w; + + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.w; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale4 * test_radius; + shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + } + + if (directional_lights.data[i].blend_splits) { + float pssm_blend; + float shadow2; + + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 1) + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + pssm_coord /= pssm_coord.w; + + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.y; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale2 * test_radius; + shadow2 = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + + pssm_blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 2) + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + pssm_coord /= pssm_coord.w; + + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.z; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale3 * test_radius; + shadow2 = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + + pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.x, directional_lights.data[i].shadow_split_offsets.y, depth_z); + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 3) + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + pssm_coord /= pssm_coord.w; + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.w; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale4 * test_radius; + shadow2 = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + + pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.y, directional_lights.data[i].shadow_split_offsets.z, depth_z); + } else { + pssm_blend = 0.0; //if no blend, same coord will be used (divide by z will result in same value, and already cached) + } + + pssm_blend = sqrt(pssm_blend); + + shadow = mix(shadow, shadow2, pssm_blend); + } + + shadow = mix(shadow, 1.0, smoothstep(directional_lights.data[i].fade_from, directional_lights.data[i].fade_to, vertex.z)); //done with negative values for performance + +#undef BIAS_FUNC + } +#else + // Soft shadow disabled version + + if (directional_lights.data[i].shadow_opacity > 0.001) { + float depth_z = -vertex.z; + + vec4 pssm_coord; + float blur_factor; + vec3 light_dir = directional_lights.data[i].direction; + vec3 base_normal_bias = normalize(normal_interp) * (1.0 - max(0.0, dot(light_dir, -normalize(normal_interp)))); + +#define BIAS_FUNC(m_var, m_idx) \ + m_var.xyz += light_dir * directional_lights.data[i].shadow_bias[m_idx]; \ + vec3 normal_bias = base_normal_bias * directional_lights.data[i].shadow_normal_bias[m_idx]; \ + normal_bias -= light_dir * dot(light_dir, normal_bias); \ + m_var.xyz += normal_bias; + + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 0) + + pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); + blur_factor = 1.0; + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 1) + + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.y; + ; + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 2) + + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.z; + } else { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 3) + + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.w; + } + + pssm_coord /= pssm_coord.w; + + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale * blur_factor, pssm_coord); + + if (directional_lights.data[i].blend_splits) { + float pssm_blend; + float blur_factor2; + + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 1) + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + pssm_blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor2 = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.y; + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 2) + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.x, directional_lights.data[i].shadow_split_offsets.y, depth_z); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor2 = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.z; + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 3) + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.y, directional_lights.data[i].shadow_split_offsets.z, depth_z); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor2 = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.w; + } else { + pssm_blend = 0.0; //if no blend, same coord will be used (divide by z will result in same value, and already cached) + blur_factor2 = 1.0; + } + + pssm_coord /= pssm_coord.w; + + float shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale * blur_factor2, pssm_coord); + shadow = mix(shadow, shadow2, pssm_blend); + } + + shadow = mix(shadow, 1.0, smoothstep(directional_lights.data[i].fade_from, directional_lights.data[i].fade_to, vertex.z)); //done with negative values for performance + +#undef BIAS_FUNC + } +#endif + + if (i < 4) { + shadow0 |= uint(clamp(shadow * 255.0, 0.0, 255.0)) << (i * 8); + } else { + shadow1 |= uint(clamp(shadow * 255.0, 0.0, 255.0)) << ((i - 4) * 8); + } + } + +#endif // SHADOWS_DISABLED + + for (uint i = 0; i < 8; i++) { + if (i >= scene_data.directional_light_count) { + break; + } + + if (!bool(directional_lights.data[i].mask & draw_call.layer_mask)) { + continue; //not masked + } + + // We're not doing light transmittence + + float shadow = 1.0; +#ifndef SHADOWS_DISABLED + if (i < 4) { + shadow = float(shadow0 >> (i * 8) & 0xFF) / 255.0; + } else { + shadow = float(shadow1 >> ((i - 4) * 8) & 0xFF) / 255.0; + } +#endif + blur_shadow(shadow); + + light_compute(normal, directional_lights.data[i].direction, normalize(view), 0.0, directional_lights.data[i].color * directional_lights.data[i].energy, shadow, f0, orms, 1.0, albedo, alpha, +#ifdef LIGHT_BACKLIGHT_USED + backlight, +#endif +/* not supported here +#ifdef LIGHT_TRANSMITTANCE_USED + transmittance_color, + transmittance_depth, + transmittance_boost, + transmittance_z, +#endif +*/ +#ifdef LIGHT_RIM_USED + rim, rim_tint, +#endif +#ifdef LIGHT_CLEARCOAT_USED + clearcoat, clearcoat_roughness, normalize(normal_interp), +#endif +#ifdef LIGHT_ANISOTROPY_USED + binormal, tangent, anisotropy, +#endif +#ifdef USE_SOFT_SHADOW + directional_lights.data[i].size, +#endif + diffuse_light, + specular_light); + } + } //directional light + + if (!sc_disable_omni_lights) { //omni lights + uint light_indices = draw_call.omni_lights.x; + for (uint i = 0; i < 8; i++) { + uint light_index = light_indices & 0xFF; + if (i == 4) { + light_indices = draw_call.omni_lights.y; + } else { + light_indices = light_indices >> 8; + } + + if (light_index == 0xFF) { + break; + } + + float shadow = light_process_omni_shadow(light_index, vertex, normal); + + shadow = blur_shadow(shadow); + + light_process_omni(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, albedo, alpha, +#ifdef LIGHT_BACKLIGHT_USED + backlight, +#endif +/* +#ifdef LIGHT_TRANSMITTANCE_USED + transmittance_color, + transmittance_depth, + transmittance_boost, +#endif +*/ +#ifdef LIGHT_RIM_USED + rim, + rim_tint, +#endif +#ifdef LIGHT_CLEARCOAT_USED + clearcoat, clearcoat_roughness, normalize(normal_interp), +#endif +#ifdef LIGHT_ANISOTROPY_USED + tangent, + binormal, anisotropy, +#endif + diffuse_light, specular_light); + } + } //omni lights + + if (!sc_disable_spot_lights) { //spot lights + + uint light_indices = draw_call.spot_lights.x; + for (uint i = 0; i < 8; i++) { + uint light_index = light_indices & 0xFF; + if (i == 4) { + light_indices = draw_call.spot_lights.y; + } else { + light_indices = light_indices >> 8; + } + + if (light_index == 0xFF) { + break; + } + + float shadow = light_process_spot_shadow(light_index, vertex, normal); + + shadow = blur_shadow(shadow); + + light_process_spot(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, albedo, alpha, +#ifdef LIGHT_BACKLIGHT_USED + backlight, +#endif +/* +#ifdef LIGHT_TRANSMITTANCE_USED + transmittance_color, + transmittance_depth, + transmittance_boost, +#endif +*/ +#ifdef LIGHT_RIM_USED + rim, + rim_tint, +#endif +#ifdef LIGHT_CLEARCOAT_USED + clearcoat, clearcoat_roughness, normalize(normal_interp), +#endif +#ifdef LIGHT_ANISOTROPY_USED + tangent, + binormal, anisotropy, +#endif + diffuse_light, specular_light); + } + } //spot lights + +#ifdef USE_SHADOW_TO_OPACITY + alpha = min(alpha, clamp(length(ambient_light), 0.0, 1.0)); + +#if defined(ALPHA_SCISSOR_USED) + if (alpha < alpha_scissor) { + discard; + } +#endif // ALPHA_SCISSOR_USED + +#ifdef USE_OPAQUE_PREPASS + + if (alpha < scene_data.opaque_prepass_threshold) { + discard; + } + +#endif // USE_OPAQUE_PREPASS + +#endif // USE_SHADOW_TO_OPACITY + +#endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + +#ifdef MODE_RENDER_DEPTH + +#ifdef MODE_RENDER_MATERIAL + + albedo_output_buffer.rgb = albedo; + albedo_output_buffer.a = alpha; + + normal_output_buffer.rgb = normal * 0.5 + 0.5; + normal_output_buffer.a = 0.0; + depth_output_buffer.r = -vertex.z; + + orm_output_buffer.r = ao; + orm_output_buffer.g = roughness; + orm_output_buffer.b = metallic; + orm_output_buffer.a = sss_strength; + + emission_output_buffer.rgb = emission; + emission_output_buffer.a = 0.0; +#endif // MODE_RENDER_MATERIAL + +#else // MODE_RENDER_DEPTH + + // multiply by albedo + diffuse_light *= albedo; // ambient must be multiplied by albedo at the end + + // apply direct light AO + ao = unpackUnorm4x8(orms).x; + specular_light *= ao; + diffuse_light *= ao; + + // apply metallic + metallic = unpackUnorm4x8(orms).z; + diffuse_light *= 1.0 - metallic; + ambient_light *= 1.0 - metallic; + + //restore fog + fog = vec4(unpackHalf2x16(fog_rg), unpackHalf2x16(fog_ba)); + +#ifdef MODE_MULTIPLE_RENDER_TARGETS + +#ifdef MODE_UNSHADED + diffuse_buffer = vec4(albedo.rgb, 0.0); + specular_buffer = vec4(0.0); + +#else // MODE_UNSHADED + +#ifdef SSS_MODE_SKIN + sss_strength = -sss_strength; +#endif // SSS_MODE_SKIN + diffuse_buffer = vec4(emission + diffuse_light + ambient_light, sss_strength); + specular_buffer = vec4(specular_light, metallic); +#endif // MODE_UNSHADED + + diffuse_buffer.rgb = mix(diffuse_buffer.rgb, fog.rgb, fog.a); + specular_buffer.rgb = mix(specular_buffer.rgb, vec3(0.0), fog.a); + +#else //MODE_MULTIPLE_RENDER_TARGETS + +#ifdef MODE_UNSHADED + frag_color = vec4(albedo, alpha); +#else // MODE_UNSHADED + frag_color = vec4(emission + ambient_light + diffuse_light + specular_light, alpha); +#endif // MODE_UNSHADED + + // Draw "fixed" fog before volumetric fog to ensure volumetric fog can appear in front of the sky. + frag_color.rgb = mix(frag_color.rgb, fog.rgb, fog.a); + + // On mobile we use a UNORM buffer with 10bpp which results in a range from 0.0 - 1.0 resulting in HDR breaking + // We divide by sc_luminance_multiplier to support a range from 0.0 - 2.0 both increasing precision on bright and darker images + frag_color.rgb = frag_color.rgb / sc_luminance_multiplier; + +#endif //MODE_MULTIPLE_RENDER_TARGETS + +#endif //MODE_RENDER_DEPTH +} diff --git a/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile_inc.glsl b/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile_inc.glsl new file mode 100644 index 0000000000..631ff0575b --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile_inc.glsl @@ -0,0 +1,167 @@ +#define M_PI 3.14159265359 +#define MAX_VIEWS 2 + +#if defined(USE_MULTIVIEW) && defined(has_VK_KHR_multiview) +#extension GL_EXT_multiview : enable +#endif + +#include "../decal_data_inc.glsl" +#include "../scene_data_inc.glsl" + +#if !defined(MODE_RENDER_DEPTH) || defined(MODE_RENDER_MATERIAL) || defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) +#ifndef NORMAL_USED +#define NORMAL_USED +#endif +#endif + +/* don't exceed 128 bytes!! */ +/* put instance data into our push content, not a array */ +layout(push_constant, std430) uniform DrawCall { + highp mat4 transform; // 64 - 64 + uint flags; // 04 - 68 + uint instance_uniforms_ofs; //base offset in global buffer for instance variables // 04 - 72 + uint gi_offset; //GI information when using lightmapping (VCT or lightmap index) // 04 - 76 + uint layer_mask; // 04 - 80 + highp vec4 lightmap_uv_scale; // 16 - 96 doubles as uv_offset when needed + + uvec2 reflection_probes; // 08 - 104 + uvec2 omni_lights; // 08 - 112 + uvec2 spot_lights; // 08 - 120 + uvec2 decals; // 08 - 128 +} +draw_call; + +/* Set 0: Base Pass (never changes) */ + +#include "../light_data_inc.glsl" + +#define SAMPLER_NEAREST_CLAMP 0 +#define SAMPLER_LINEAR_CLAMP 1 +#define SAMPLER_NEAREST_WITH_MIPMAPS_CLAMP 2 +#define SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP 3 +#define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_CLAMP 4 +#define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_CLAMP 5 +#define SAMPLER_NEAREST_REPEAT 6 +#define SAMPLER_LINEAR_REPEAT 7 +#define SAMPLER_NEAREST_WITH_MIPMAPS_REPEAT 8 +#define SAMPLER_LINEAR_WITH_MIPMAPS_REPEAT 9 +#define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_REPEAT 10 +#define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_REPEAT 11 + +layout(set = 0, binding = 1) uniform sampler material_samplers[12]; + +layout(set = 0, binding = 2) uniform sampler shadow_sampler; + +layout(set = 0, binding = 3) uniform sampler decal_sampler; +layout(set = 0, binding = 4) uniform sampler light_projector_sampler; + +#define INSTANCE_FLAGS_NON_UNIFORM_SCALE (1 << 4) +#define INSTANCE_FLAGS_USE_GI_BUFFERS (1 << 5) +#define INSTANCE_FLAGS_USE_SDFGI (1 << 6) +#define INSTANCE_FLAGS_USE_LIGHTMAP_CAPTURE (1 << 7) +#define INSTANCE_FLAGS_USE_LIGHTMAP (1 << 8) +#define INSTANCE_FLAGS_USE_SH_LIGHTMAP (1 << 9) +#define INSTANCE_FLAGS_USE_VOXEL_GI (1 << 10) +#define INSTANCE_FLAGS_PARTICLES (1 << 11) +#define INSTANCE_FLAGS_MULTIMESH (1 << 12) +#define INSTANCE_FLAGS_MULTIMESH_FORMAT_2D (1 << 13) +#define INSTANCE_FLAGS_MULTIMESH_HAS_COLOR (1 << 14) +#define INSTANCE_FLAGS_MULTIMESH_HAS_CUSTOM_DATA (1 << 15) +#define INSTANCE_FLAGS_PARTICLE_TRAIL_SHIFT 16 +//3 bits of stride +#define INSTANCE_FLAGS_PARTICLE_TRAIL_MASK 0xFF + +layout(set = 0, binding = 5, std430) restrict readonly buffer OmniLights { + LightData data[]; +} +omni_lights; + +layout(set = 0, binding = 6, std430) restrict readonly buffer SpotLights { + LightData data[]; +} +spot_lights; + +layout(set = 0, binding = 7, std430) restrict readonly buffer ReflectionProbeData { + ReflectionData data[]; +} +reflections; + +layout(set = 0, binding = 8, std140) uniform DirectionalLights { + DirectionalLightData data[MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS]; +} +directional_lights; + +#define LIGHTMAP_FLAG_USE_DIRECTION 1 +#define LIGHTMAP_FLAG_USE_SPECULAR_DIRECTION 2 + +struct Lightmap { + mediump mat3 normal_xform; + vec3 pad; + float exposure_normalization; +}; + +layout(set = 0, binding = 9, std140) restrict readonly buffer Lightmaps { + Lightmap data[]; +} +lightmaps; + +struct LightmapCapture { + mediump vec4 sh[9]; +}; + +layout(set = 0, binding = 10, std140) restrict readonly buffer LightmapCaptures { + LightmapCapture data[]; +} +lightmap_captures; + +layout(set = 0, binding = 11) uniform mediump texture2D decal_atlas; +layout(set = 0, binding = 12) uniform mediump texture2D decal_atlas_srgb; + +layout(set = 0, binding = 13, std430) restrict readonly buffer Decals { + DecalData data[]; +} +decals; + +layout(set = 0, binding = 14, std430) restrict readonly buffer GlobalShaderUniformData { + highp vec4 data[]; +} +global_shader_uniforms; + +/* Set 1: Render Pass (changes per render pass) */ + +layout(set = 1, binding = 0, std140) uniform SceneDataBlock { + SceneData data; + SceneData prev_data; +} +scene_data_block; + +#ifdef USE_RADIANCE_CUBEMAP_ARRAY + +layout(set = 1, binding = 2) uniform mediump textureCubeArray radiance_cubemap; + +#else + +layout(set = 1, binding = 2) uniform mediump textureCube radiance_cubemap; + +#endif + +layout(set = 1, binding = 3) uniform mediump textureCubeArray reflection_atlas; + +layout(set = 1, binding = 4) uniform highp texture2D shadow_atlas; + +layout(set = 1, binding = 5) uniform highp texture2D directional_shadow_atlas; + +// this needs to change to providing just the lightmap we're using.. +layout(set = 1, binding = 6) uniform texture2DArray lightmap_textures[MAX_LIGHTMAP_TEXTURES]; + +layout(set = 1, binding = 9) uniform highp texture2D depth_buffer; +layout(set = 1, binding = 10) uniform mediump texture2D color_buffer; + +/* Set 2 Skeleton & Instancing (can change per item) */ + +layout(set = 2, binding = 0, std430) restrict readonly buffer Transforms { + highp vec4 data[]; +} +transforms; + +/* Set 3 User Material */ diff --git a/servers/rendering/renderer_rd/shaders/giprobe_write.glsl b/servers/rendering/renderer_rd/shaders/giprobe_write.glsl new file mode 100644 index 0000000000..6c73864bf6 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/giprobe_write.glsl @@ -0,0 +1,275 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in; + +#define NO_CHILDREN 0xFFFFFFFF + +struct CellChildren { + uint children[8]; +}; + +layout(set = 0, binding = 1, std430) buffer CellChildrenBuffer { + CellChildren data[]; +} +cell_children; + +struct CellData { + uint position; // xyz 10 bits + uint albedo; //rgb albedo + uint emission; //rgb normalized with e as multiplier + uint normal; //RGB normal encoded +}; + +layout(set = 0, binding = 2, std430) buffer CellDataBuffer { + CellData data[]; +} +cell_data; + +#define LIGHT_TYPE_DIRECTIONAL 0 +#define LIGHT_TYPE_OMNI 1 +#define LIGHT_TYPE_SPOT 2 + +#ifdef MODE_COMPUTE_LIGHT + +struct Light { + uint type; + float energy; + float radius; + float attenuation; + + vec3 color; + float cos_spot_angle; + + vec3 position; + float inv_spot_attenuation; + + vec3 direction; + bool has_shadow; +}; + +layout(set = 0, binding = 3, std140) uniform Lights { + Light data[MAX_LIGHTS]; +} +lights; + +#endif + +layout(push_constant, std430) uniform Params { + ivec3 limits; + uint stack_size; + + float emission_scale; + float propagation; + float dynamic_range; + + uint light_count; + uint cell_offset; + uint cell_count; + uint pad[2]; +} +params; + +layout(set = 0, binding = 4, std140) uniform Outputs { + vec4 data[]; +} +output; + +#ifdef MODE_COMPUTE_LIGHT + +uint raymarch(float distance, float distance_adv, vec3 from, vec3 direction) { + uint result = NO_CHILDREN; + + ivec3 size = ivec3(max(max(params.limits.x, params.limits.y), params.limits.z)); + + while (distance > -distance_adv) { //use this to avoid precision errors + uint cell = 0; + + ivec3 pos = ivec3(from); + + if (all(greaterThanEqual(pos, ivec3(0))) && all(lessThan(pos, size))) { + ivec3 ofs = ivec3(0); + ivec3 half_size = size / 2; + + for (int i = 0; i < params.stack_size - 1; i++) { + bvec3 greater = greaterThanEqual(pos, ofs + half_size); + + ofs += mix(ivec3(0), half_size, greater); + + uint child = 0; //wonder if this can be done faster + if (greater.x) { + child |= 1; + } + if (greater.y) { + child |= 2; + } + if (greater.z) { + child |= 4; + } + + cell = cell_children.data[cell].children[child]; + if (cell == NO_CHILDREN) { + break; + } + + half_size >>= ivec3(1); + } + + if (cell != NO_CHILDREN) { + return cell; //found cell! + } + } + + from += direction * distance_adv; + distance -= distance_adv; + } + + return NO_CHILDREN; +} + +bool compute_light_vector(uint light, uint cell, vec3 pos, out float attenuation, out vec3 light_pos) { + if (lights.data[light].type == LIGHT_TYPE_DIRECTIONAL) { + light_pos = pos - lights.data[light].direction * length(vec3(params.limits)); + attenuation = 1.0; + } else { + light_pos = lights.data[light].position; + float distance = length(pos - light_pos); + if (distance >= lights.data[light].radius) { + return false; + } + + attenuation = pow(clamp(1.0 - distance / lights.data[light].radius, 0.0001, 1.0), lights.data[light].attenuation); + + if (lights.data[light].type == LIGHT_TYPE_SPOT) { + vec3 rel = normalize(pos - light_pos); + float cos_spot_angle = lights.data[light].cos_spot_angle; + float cos_angle = dot(rel, lights.data[light].direction); + if (cos_angle < cos_spot_angle) { + return false; + } + + float scos = max(cos_angle, cos_spot_angle); + float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - cos_spot_angle)); + attenuation *= 1.0 - pow(spot_rim, lights.data[light].inv_spot_attenuation); + } + } + + return true; +} + +float get_normal_advance(vec3 p_normal) { + vec3 normal = p_normal; + vec3 unorm = abs(normal); + + if ((unorm.x >= unorm.y) && (unorm.x >= unorm.z)) { + // x code + unorm = normal.x > 0.0 ? vec3(1.0, 0.0, 0.0) : vec3(-1.0, 0.0, 0.0); + } else if ((unorm.y > unorm.x) && (unorm.y >= unorm.z)) { + // y code + unorm = normal.y > 0.0 ? vec3(0.0, 1.0, 0.0) : vec3(0.0, -1.0, 0.0); + } else if ((unorm.z > unorm.x) && (unorm.z > unorm.y)) { + // z code + unorm = normal.z > 0.0 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 0.0, -1.0); + } else { + // oh-no we messed up code + // has to be + unorm = vec3(1.0, 0.0, 0.0); + } + + return 1.0 / dot(normal, unorm); +} + +#endif + +void main() { + uint cell_index = gl_GlobalInvocationID.x; + if (cell_index >= params.cell_count) { + return; + } + cell_index += params.cell_offset; + + uvec3 posu = uvec3(cell_data.data[cell_index].position & 0x7FF, (cell_data.data[cell_index].position >> 11) & 0x3FF, cell_data.data[cell_index].position >> 21); + vec4 albedo = unpackUnorm4x8(cell_data.data[cell_index].albedo); + +#ifdef MODE_COMPUTE_LIGHT + + vec3 pos = vec3(posu) + vec3(0.5); + + vec3 emission = vec3(ivec3(cell_data.data[cell_index].emission & 0x3FF, (cell_data.data[cell_index].emission >> 10) & 0x7FF, cell_data.data[cell_index].emission >> 21)) * params.emission_scale; + vec4 normal = unpackSnorm4x8(cell_data.data[cell_index].normal); + + vec3 accum = vec3(0.0); + + for (uint i = 0; i < params.light_count; i++) { + float attenuation; + vec3 light_pos; + + if (!compute_light_vector(i, cell_index, pos, attenuation, light_pos)) { + continue; + } + + vec3 light_dir = pos - light_pos; + float distance = length(light_dir); + light_dir = normalize(light_dir); + + if (length(normal.xyz) > 0.2 && dot(normal.xyz, light_dir) >= 0) { + continue; //not facing the light + } + + if (lights.data[i].has_shadow) { + float distance_adv = get_normal_advance(light_dir); + + distance += distance_adv - mod(distance, distance_adv); //make it reach the center of the box always + + vec3 from = pos - light_dir * distance; //approximate + from -= sign(light_dir) * 0.45; //go near the edge towards the light direction to avoid self occlusion + + uint result = raymarch(distance, distance_adv, from, light_dir); + + if (result != cell_index) { + continue; //was occluded + } + } + + vec3 light = lights.data[i].color * albedo.rgb * attenuation * lights.data[i].energy; + + if (length(normal.xyz) > 0.2) { + accum += max(0.0, dot(normal.xyz, -light_dir)) * light + emission; + } else { + //all directions + accum += light + emission; + } + } + + output.data[cell_index] = vec4(accum, 0.0); + +#endif //MODE_COMPUTE_LIGHT + +#ifdef MODE_UPDATE_MIPMAPS + + { + vec3 light_accum = vec3(0.0); + float count = 0.0; + for (uint i = 0; i < 8; i++) { + uint child_index = cell_children.data[cell_index].children[i]; + if (child_index == NO_CHILDREN) { + continue; + } + light_accum += output.data[child_index].rgb; + + count += 1.0; + } + + float divisor = mix(8.0, count, params.propagation); + output.data[cell_index] = vec4(light_accum / divisor, 0.0); + } +#endif + +#ifdef MODE_WRITE_TEXTURE + { + } +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/light_data_inc.glsl b/servers/rendering/renderer_rd/shaders/light_data_inc.glsl new file mode 100644 index 0000000000..7488a3f2c7 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/light_data_inc.glsl @@ -0,0 +1,83 @@ +#define LIGHT_BAKE_DISABLED 0 +#define LIGHT_BAKE_STATIC 1 +#define LIGHT_BAKE_DYNAMIC 2 + +struct LightData { //this structure needs to be as packed as possible + highp vec3 position; + highp float inv_radius; + + mediump vec3 direction; + highp float size; + + mediump vec3 color; + mediump float attenuation; + + mediump float cone_attenuation; + mediump float cone_angle; + mediump float specular_amount; + mediump float shadow_opacity; + + highp vec4 atlas_rect; // rect in the shadow atlas + highp mat4 shadow_matrix; + highp float shadow_bias; + highp float shadow_normal_bias; + highp float transmittance_bias; + highp float soft_shadow_size; // for spot, it's the size in uv coordinates of the light, for omni it's the span angle + highp float soft_shadow_scale; // scales the shadow kernel for blurrier shadows + uint mask; + mediump float volumetric_fog_energy; + uint bake_mode; + highp vec4 projector_rect; //projector rect in srgb decal atlas +}; + +#define REFLECTION_AMBIENT_DISABLED 0 +#define REFLECTION_AMBIENT_ENVIRONMENT 1 +#define REFLECTION_AMBIENT_COLOR 2 + +struct ReflectionData { + highp vec3 box_extents; + mediump float index; + highp vec3 box_offset; + uint mask; + mediump vec3 ambient; // ambient color + mediump float intensity; + bool exterior; + bool box_project; + uint ambient_mode; + float exposure_normalization; + //0-8 is intensity,8-9 is ambient, mode + highp mat4 local_matrix; // up to here for spot and omni, rest is for directional + // notes: for ambientblend, use distance to edge to blend between already existing global environment +}; + +struct DirectionalLightData { + mediump vec3 direction; + highp float energy; // needs to be highp to avoid NaNs being created with high energy values (i.e. when using physical light units and over-exposing the image) + mediump vec3 color; + mediump float size; + mediump float specular; + uint mask; + highp float softshadow_angle; + highp float soft_shadow_scale; + bool blend_splits; + mediump float shadow_opacity; + highp float fade_from; + highp float fade_to; + uvec2 pad; + uint bake_mode; + mediump float volumetric_fog_energy; + highp vec4 shadow_bias; + highp vec4 shadow_normal_bias; + highp vec4 shadow_transmittance_bias; + highp vec4 shadow_z_range; + highp vec4 shadow_range_begin; + highp vec4 shadow_split_offsets; + highp mat4 shadow_matrix1; + highp mat4 shadow_matrix2; + highp mat4 shadow_matrix3; + highp mat4 shadow_matrix4; + highp vec2 uv_scale1; + highp vec2 uv_scale2; + highp vec2 uv_scale3; + highp vec2 uv_scale4; +}; diff --git a/servers/rendering/renderer_rd/shaders/luminance_reduce.glsl b/servers/rendering/renderer_rd/shaders/luminance_reduce.glsl new file mode 100644 index 0000000000..0ee4cf6e31 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/luminance_reduce.glsl @@ -0,0 +1,82 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +#define BLOCK_SIZE 8 + +layout(local_size_x = BLOCK_SIZE, local_size_y = BLOCK_SIZE, local_size_z = 1) in; + +shared float tmp_data[BLOCK_SIZE * BLOCK_SIZE]; + +#ifdef READ_TEXTURE + +//use for main texture +layout(set = 0, binding = 0) uniform sampler2D source_texture; + +#else + +//use for intermediate textures +layout(r32f, set = 0, binding = 0) uniform restrict readonly image2D source_luminance; + +#endif + +layout(r32f, set = 1, binding = 0) uniform restrict writeonly image2D dest_luminance; + +#ifdef WRITE_LUMINANCE +layout(set = 2, binding = 0) uniform sampler2D prev_luminance; +#endif + +layout(push_constant, std430) uniform Params { + ivec2 source_size; + float max_luminance; + float min_luminance; + float exposure_adjust; + float pad[3]; +} +params; + +void main() { + uint t = gl_LocalInvocationID.y * BLOCK_SIZE + gl_LocalInvocationID.x; + ivec2 pos = ivec2(gl_GlobalInvocationID.xy); + + if (any(lessThan(pos, params.source_size))) { +#ifdef READ_TEXTURE + vec3 v = texelFetch(source_texture, pos, 0).rgb; + tmp_data[t] = max(v.r, max(v.g, v.b)); +#else + tmp_data[t] = imageLoad(source_luminance, pos).r; +#endif + } else { + tmp_data[t] = 0.0; + } + + groupMemoryBarrier(); + barrier(); + + uint size = (BLOCK_SIZE * BLOCK_SIZE) >> 1; + + do { + if (t < size) { + tmp_data[t] += tmp_data[t + size]; + } + groupMemoryBarrier(); + barrier(); + + size >>= 1; + } while (size >= 1); + + if (t == 0) { + //compute rect size + ivec2 rect_size = min(params.source_size - pos, ivec2(BLOCK_SIZE)); + float avg = tmp_data[0] / float(rect_size.x * rect_size.y); + //float avg = tmp_data[0] / float(BLOCK_SIZE*BLOCK_SIZE); + pos /= ivec2(BLOCK_SIZE); +#ifdef WRITE_LUMINANCE + float prev_lum = texelFetch(prev_luminance, ivec2(0, 0), 0).r; //1 pixel previous exposure + avg = clamp(prev_lum + (avg - prev_lum) * params.exposure_adjust, params.min_luminance, params.max_luminance); +#endif + imageStore(dest_luminance, pos, vec4(avg)); + } +} diff --git a/servers/rendering/renderer_rd/shaders/luminance_reduce_raster.glsl b/servers/rendering/renderer_rd/shaders/luminance_reduce_raster.glsl new file mode 100644 index 0000000000..29ebd74a90 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/luminance_reduce_raster.glsl @@ -0,0 +1,74 @@ +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#include "luminance_reduce_raster_inc.glsl" + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + + gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#include "luminance_reduce_raster_inc.glsl" + +layout(location = 0) in vec2 uv_interp; +/* clang-format on */ + +layout(set = 0, binding = 0) uniform sampler2D source_exposure; + +#ifdef FINAL_PASS +layout(set = 1, binding = 0) uniform sampler2D prev_luminance; +#endif + +layout(location = 0) out highp float luminance; + +void main() { + ivec2 dest_pos = ivec2(uv_interp * settings.dest_size); + ivec2 src_pos = ivec2(uv_interp * settings.source_size); + + ivec2 next_pos = (dest_pos + ivec2(1)) * settings.source_size / settings.dest_size; + next_pos = max(next_pos, src_pos + ivec2(1)); //so it at least reads one pixel + + highp vec3 source_color = vec3(0.0); + for (int i = src_pos.x; i < next_pos.x; i++) { + for (int j = src_pos.y; j < next_pos.y; j++) { + source_color += texelFetch(source_exposure, ivec2(i, j), 0).rgb; + } + } + + source_color /= float((next_pos.x - src_pos.x) * (next_pos.y - src_pos.y)); + +#ifdef FIRST_PASS + luminance = max(source_color.r, max(source_color.g, source_color.b)); + + // This formula should be more "accurate" but gave an overexposed result when testing. + // Leaving it here so we can revisit it if we want. + // luminance = source_color.r * 0.21 + source_color.g * 0.71 + source_color.b * 0.07; +#else + luminance = source_color.r; +#endif + +#ifdef FINAL_PASS + // Obtain our target luminance + luminance = clamp(luminance, settings.min_luminance, settings.max_luminance); + + // Now smooth to our transition + highp float prev_lum = texelFetch(prev_luminance, ivec2(0, 0), 0).r; //1 pixel previous luminance + luminance = prev_lum + (luminance - prev_lum) * clamp(settings.exposure_adjust, 0.0, 1.0); +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/luminance_reduce_raster_inc.glsl b/servers/rendering/renderer_rd/shaders/luminance_reduce_raster_inc.glsl new file mode 100644 index 0000000000..b8860f6518 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/luminance_reduce_raster_inc.glsl @@ -0,0 +1,11 @@ + +layout(push_constant, std430) uniform PushConstant { + ivec2 source_size; + ivec2 dest_size; + + float exposure_adjust; + float min_luminance; + float max_luminance; + uint pad1; +} +settings; diff --git a/servers/rendering/renderer_rd/shaders/particles.glsl b/servers/rendering/renderer_rd/shaders/particles.glsl new file mode 100644 index 0000000000..3a6dd579b9 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/particles.glsl @@ -0,0 +1,658 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in; + +#define SAMPLER_NEAREST_CLAMP 0 +#define SAMPLER_LINEAR_CLAMP 1 +#define SAMPLER_NEAREST_WITH_MIPMAPS_CLAMP 2 +#define SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP 3 +#define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_CLAMP 4 +#define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_CLAMP 5 +#define SAMPLER_NEAREST_REPEAT 6 +#define SAMPLER_LINEAR_REPEAT 7 +#define SAMPLER_NEAREST_WITH_MIPMAPS_REPEAT 8 +#define SAMPLER_LINEAR_WITH_MIPMAPS_REPEAT 9 +#define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_REPEAT 10 +#define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_REPEAT 11 + +#define SDF_MAX_LENGTH 16384.0 + +/* SET 0: GLOBAL DATA */ + +layout(set = 0, binding = 1) uniform sampler material_samplers[12]; + +layout(set = 0, binding = 2, std430) restrict readonly buffer GlobalShaderUniformData { + vec4 data[]; +} +global_shader_uniforms; + +/* Set 1: FRAME AND PARTICLE DATA */ + +// a frame history is kept for trail deterministic behavior + +#define MAX_ATTRACTORS 32 + +#define ATTRACTOR_TYPE_SPHERE 0 +#define ATTRACTOR_TYPE_BOX 1 +#define ATTRACTOR_TYPE_VECTOR_FIELD 2 + +struct Attractor { + mat4 transform; + vec3 extents; //exents or radius + uint type; + uint texture_index; //texture index for vector field + float strength; + float attenuation; + float directionality; +}; + +#define MAX_COLLIDERS 32 + +#define COLLIDER_TYPE_SPHERE 0 +#define COLLIDER_TYPE_BOX 1 +#define COLLIDER_TYPE_SDF 2 +#define COLLIDER_TYPE_HEIGHT_FIELD 3 +#define COLLIDER_TYPE_2D_SDF 4 + +struct Collider { + mat4 transform; + vec3 extents; //exents or radius + uint type; + + uint texture_index; //texture index for vector field + float scale; + uint pad[2]; +}; + +struct FrameParams { + bool emitting; + float system_phase; + float prev_system_phase; + uint cycle; + + float explosiveness; + float randomness; + float time; + float delta; + + uint frame; + uint pad0; + uint pad1; + uint pad2; + + uint random_seed; + uint attractor_count; + uint collider_count; + float particle_size; + + mat4 emission_transform; + + Attractor attractors[MAX_ATTRACTORS]; + Collider colliders[MAX_COLLIDERS]; +}; + +layout(set = 1, binding = 0, std430) restrict buffer FrameHistory { + FrameParams data[]; +} +frame_history; + +#define PARTICLE_FLAG_ACTIVE uint(1) +#define PARTICLE_FLAG_STARTED uint(2) +#define PARTICLE_FLAG_TRAILED uint(4) +#define PARTICLE_FRAME_MASK uint(0xFFFF) +#define PARTICLE_FRAME_SHIFT uint(16) + +struct ParticleData { + mat4 xform; + vec3 velocity; + uint flags; + vec4 color; + vec4 custom; +#ifdef USERDATA1_USED + vec4 userdata1; +#endif +#ifdef USERDATA2_USED + vec4 userdata2; +#endif +#ifdef USERDATA3_USED + vec4 userdata3; +#endif +#ifdef USERDATA4_USED + vec4 userdata4; +#endif +#ifdef USERDATA5_USED + vec4 userdata5; +#endif +#ifdef USERDATA6_USED + vec4 userdata6; +#endif +}; + +layout(set = 1, binding = 1, std430) restrict buffer Particles { + ParticleData data[]; +} +particles; + +#define EMISSION_FLAG_HAS_POSITION 1 +#define EMISSION_FLAG_HAS_ROTATION_SCALE 2 +#define EMISSION_FLAG_HAS_VELOCITY 4 +#define EMISSION_FLAG_HAS_COLOR 8 +#define EMISSION_FLAG_HAS_CUSTOM 16 + +struct ParticleEmission { + mat4 xform; + vec3 velocity; + uint flags; + vec4 color; + vec4 custom; +}; + +layout(set = 1, binding = 2, std430) restrict buffer SourceEmission { + int particle_count; + uint pad0; + uint pad1; + uint pad2; + ParticleEmission data[]; +} +src_particles; + +layout(set = 1, binding = 3, std430) restrict buffer DestEmission { + int particle_count; + int particle_max; + uint pad1; + uint pad2; + ParticleEmission data[]; +} +dst_particles; + +/* SET 2: COLLIDER/ATTRACTOR TEXTURES */ + +#define MAX_3D_TEXTURES 7 + +layout(set = 2, binding = 0) uniform texture3D sdf_vec_textures[MAX_3D_TEXTURES]; +layout(set = 2, binding = 1) uniform texture2D height_field_texture; + +/* SET 3: MATERIAL */ + +#ifdef MATERIAL_UNIFORMS_USED +layout(set = 3, binding = 0, std140) uniform MaterialUniforms{ + +#MATERIAL_UNIFORMS + +} material; +#endif + +layout(push_constant, std430) uniform Params { + float lifetime; + bool clear; + uint total_particles; + uint trail_size; + bool use_fractional_delta; + bool sub_emitter_mode; + bool can_emit; + bool trail_pass; +} +params; + +uint hash(uint x) { + x = ((x >> uint(16)) ^ x) * uint(0x45d9f3b); + x = ((x >> uint(16)) ^ x) * uint(0x45d9f3b); + x = (x >> uint(16)) ^ x; + return x; +} + +bool emit_subparticle(mat4 p_xform, vec3 p_velocity, vec4 p_color, vec4 p_custom, uint p_flags) { + if (!params.can_emit) { + return false; + } + + bool valid = false; + + int dst_index = atomicAdd(dst_particles.particle_count, 1); + + if (dst_index >= dst_particles.particle_max) { + atomicAdd(dst_particles.particle_count, -1); + return false; + } + + dst_particles.data[dst_index].xform = p_xform; + dst_particles.data[dst_index].velocity = p_velocity; + dst_particles.data[dst_index].color = p_color; + dst_particles.data[dst_index].custom = p_custom; + dst_particles.data[dst_index].flags = p_flags; + + return true; +} + +vec3 safe_normalize(vec3 direction) { + const float EPSILON = 0.001; + if (length(direction) < EPSILON) { + return vec3(0.0); + } + return normalize(direction); +} + +#GLOBALS + +void main() { + uint particle = gl_GlobalInvocationID.x; + + if (params.trail_size > 1) { + if (params.trail_pass) { + particle += (particle / (params.trail_size - 1)) + 1; + } else { + particle *= params.trail_size; + } + } + + if (particle >= params.total_particles * params.trail_size) { + return; //discard + } + + uint index = particle / params.trail_size; + uint frame = (particle % params.trail_size); + +#define FRAME frame_history.data[frame] +#define PARTICLE particles.data[particle] + + bool apply_forces = true; + bool apply_velocity = true; + float local_delta = FRAME.delta; + + float mass = 1.0; + + bool restart = false; + + bool restart_position = false; + bool restart_rotation_scale = false; + bool restart_velocity = false; + bool restart_color = false; + bool restart_custom = false; + + if (params.clear) { + PARTICLE.color = vec4(1.0); + PARTICLE.custom = vec4(0.0); + PARTICLE.velocity = vec3(0.0); + PARTICLE.flags = 0; + PARTICLE.xform = mat4( + vec4(1.0, 0.0, 0.0, 0.0), + vec4(0.0, 1.0, 0.0, 0.0), + vec4(0.0, 0.0, 1.0, 0.0), + vec4(0.0, 0.0, 0.0, 1.0)); + } + + //clear started flag if set + + if (params.trail_pass) { + //trail started + uint src_idx = index * params.trail_size; + if (bool(particles.data[src_idx].flags & PARTICLE_FLAG_STARTED)) { + //save start conditions for trails + PARTICLE.color = particles.data[src_idx].color; + PARTICLE.custom = particles.data[src_idx].custom; + PARTICLE.velocity = particles.data[src_idx].velocity; + PARTICLE.flags = PARTICLE_FLAG_TRAILED | ((frame_history.data[0].frame & PARTICLE_FRAME_MASK) << PARTICLE_FRAME_SHIFT); //mark it as trailed, save in which frame it will start + PARTICLE.xform = particles.data[src_idx].xform; + } + + if (bool(PARTICLE.flags & PARTICLE_FLAG_TRAILED) && ((PARTICLE.flags >> PARTICLE_FRAME_SHIFT) == (FRAME.frame & PARTICLE_FRAME_MASK))) { //check this is trailed and see if it should start now + // we just assume that this is the first frame of the particle, the rest is deterministic + PARTICLE.flags = PARTICLE_FLAG_ACTIVE | (particles.data[src_idx].flags & (PARTICLE_FRAME_MASK << PARTICLE_FRAME_SHIFT)); + return; //- this appears like it should be correct, but it seems not to be.. wonder why. + } + } else { + PARTICLE.flags &= ~PARTICLE_FLAG_STARTED; + } + + bool collided = false; + vec3 collision_normal = vec3(0.0); + float collision_depth = 0.0; + + vec3 attractor_force = vec3(0.0); + +#if !defined(DISABLE_VELOCITY) + + if (bool(PARTICLE.flags & PARTICLE_FLAG_ACTIVE)) { + PARTICLE.xform[3].xyz += PARTICLE.velocity * local_delta; + } +#endif + + if (!params.trail_pass && params.sub_emitter_mode) { + if (!bool(PARTICLE.flags & PARTICLE_FLAG_ACTIVE)) { + int src_index = atomicAdd(src_particles.particle_count, -1) - 1; + + if (src_index >= 0) { + PARTICLE.flags = (PARTICLE_FLAG_ACTIVE | PARTICLE_FLAG_STARTED | (FRAME.cycle << PARTICLE_FRAME_SHIFT)); + restart = true; + + if (bool(src_particles.data[src_index].flags & EMISSION_FLAG_HAS_POSITION)) { + PARTICLE.xform[3] = src_particles.data[src_index].xform[3]; + } else { + PARTICLE.xform[3] = vec4(0, 0, 0, 1); + restart_position = true; + } + if (bool(src_particles.data[src_index].flags & EMISSION_FLAG_HAS_ROTATION_SCALE)) { + PARTICLE.xform[0] = src_particles.data[src_index].xform[0]; + PARTICLE.xform[1] = src_particles.data[src_index].xform[1]; + PARTICLE.xform[2] = src_particles.data[src_index].xform[2]; + } else { + PARTICLE.xform[0] = vec4(1, 0, 0, 0); + PARTICLE.xform[1] = vec4(0, 1, 0, 0); + PARTICLE.xform[2] = vec4(0, 0, 1, 0); + restart_rotation_scale = true; + } + if (bool(src_particles.data[src_index].flags & EMISSION_FLAG_HAS_VELOCITY)) { + PARTICLE.velocity = src_particles.data[src_index].velocity; + } else { + PARTICLE.velocity = vec3(0); + restart_velocity = true; + } + if (bool(src_particles.data[src_index].flags & EMISSION_FLAG_HAS_COLOR)) { + PARTICLE.color = src_particles.data[src_index].color; + } else { + PARTICLE.color = vec4(1); + restart_color = true; + } + + if (bool(src_particles.data[src_index].flags & EMISSION_FLAG_HAS_CUSTOM)) { + PARTICLE.custom = src_particles.data[src_index].custom; + } else { + PARTICLE.custom = vec4(0); + restart_custom = true; + } + } + } + + } else if (FRAME.emitting) { + float restart_phase = float(index) / float(params.total_particles); + + if (FRAME.randomness > 0.0) { + uint seed = FRAME.cycle; + if (restart_phase >= FRAME.system_phase) { + seed -= uint(1); + } + seed *= uint(params.total_particles); + seed += uint(index); + float random = float(hash(seed) % uint(65536)) / 65536.0; + restart_phase += FRAME.randomness * random * 1.0 / float(params.total_particles); + } + + restart_phase *= (1.0 - FRAME.explosiveness); + + if (FRAME.system_phase > FRAME.prev_system_phase) { + // restart_phase >= prev_system_phase is used so particles emit in the first frame they are processed + + if (restart_phase >= FRAME.prev_system_phase && restart_phase < FRAME.system_phase) { + restart = true; + if (params.use_fractional_delta) { + local_delta = (FRAME.system_phase - restart_phase) * params.lifetime; + } + } + + } else if (FRAME.delta > 0.0) { + if (restart_phase >= FRAME.prev_system_phase) { + restart = true; + if (params.use_fractional_delta) { + local_delta = (1.0 - restart_phase + FRAME.system_phase) * params.lifetime; + } + + } else if (restart_phase < FRAME.system_phase) { + restart = true; + if (params.use_fractional_delta) { + local_delta = (FRAME.system_phase - restart_phase) * params.lifetime; + } + } + } + + if (params.trail_pass) { + restart = false; + } + + if (restart) { + PARTICLE.flags = FRAME.emitting ? (PARTICLE_FLAG_ACTIVE | PARTICLE_FLAG_STARTED | (FRAME.cycle << PARTICLE_FRAME_SHIFT)) : 0; + restart_position = true; + restart_rotation_scale = true; + restart_velocity = true; + restart_color = true; + restart_custom = true; + } + } + + bool particle_active = bool(PARTICLE.flags & PARTICLE_FLAG_ACTIVE); + + uint particle_number = (PARTICLE.flags >> PARTICLE_FRAME_SHIFT) * uint(params.total_particles) + index; + + if (restart && particle_active) { +#CODE : START + } + + if (particle_active) { + for (uint i = 0; i < FRAME.attractor_count; i++) { + vec3 dir; + float amount; + vec3 rel_vec = PARTICLE.xform[3].xyz - FRAME.attractors[i].transform[3].xyz; + vec3 local_pos = rel_vec * mat3(FRAME.attractors[i].transform); + + switch (FRAME.attractors[i].type) { + case ATTRACTOR_TYPE_SPHERE: { + dir = safe_normalize(rel_vec); + float d = length(local_pos) / FRAME.attractors[i].extents.x; + if (d > 1.0) { + continue; + } + amount = max(0.0, 1.0 - d); + } break; + case ATTRACTOR_TYPE_BOX: { + dir = safe_normalize(rel_vec); + + vec3 abs_pos = abs(local_pos / FRAME.attractors[i].extents); + float d = max(abs_pos.x, max(abs_pos.y, abs_pos.z)); + if (d > 1.0) { + continue; + } + amount = max(0.0, 1.0 - d); + + } break; + case ATTRACTOR_TYPE_VECTOR_FIELD: { + vec3 uvw_pos = (local_pos / FRAME.attractors[i].extents + 1.0) * 0.5; + if (any(lessThan(uvw_pos, vec3(0.0))) || any(greaterThan(uvw_pos, vec3(1.0)))) { + continue; + } + vec3 s = texture(sampler3D(sdf_vec_textures[FRAME.attractors[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos).xyz * 2.0 - 1.0; + dir = mat3(FRAME.attractors[i].transform) * safe_normalize(s); //revert direction + amount = length(s); + + } break; + } + amount = pow(amount, FRAME.attractors[i].attenuation); + dir = safe_normalize(mix(dir, FRAME.attractors[i].transform[2].xyz, FRAME.attractors[i].directionality)); + attractor_force -= amount * dir * FRAME.attractors[i].strength; + } + + float particle_size = FRAME.particle_size; + +#ifdef USE_COLLISION_SCALE + + particle_size *= dot(vec3(length(PARTICLE.xform[0].xyz), length(PARTICLE.xform[1].xyz), length(PARTICLE.xform[2].xyz)), vec3(0.33333333333)); + +#endif + + if (FRAME.collider_count == 1 && FRAME.colliders[0].type == COLLIDER_TYPE_2D_SDF) { + //2D collision + + vec2 pos = PARTICLE.xform[3].xy; + vec4 to_sdf_x = FRAME.colliders[0].transform[0]; + vec4 to_sdf_y = FRAME.colliders[0].transform[1]; + vec2 sdf_pos = vec2(dot(vec4(pos, 0, 1), to_sdf_x), dot(vec4(pos, 0, 1), to_sdf_y)); + + vec4 sdf_to_screen = vec4(FRAME.colliders[0].extents, FRAME.colliders[0].scale); + + vec2 uv_pos = sdf_pos * sdf_to_screen.xy + sdf_to_screen.zw; + + if (all(greaterThan(uv_pos, vec2(0.0))) && all(lessThan(uv_pos, vec2(1.0)))) { + vec2 pos2 = pos + vec2(0, particle_size); + vec2 sdf_pos2 = vec2(dot(vec4(pos2, 0, 1), to_sdf_x), dot(vec4(pos2, 0, 1), to_sdf_y)); + float sdf_particle_size = distance(sdf_pos, sdf_pos2); + + float d = texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uv_pos).r * SDF_MAX_LENGTH; + + d -= sdf_particle_size; + + if (d < 0.0) { + const float EPSILON = 0.001; + vec2 n = normalize(vec2( + texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uv_pos + vec2(EPSILON, 0.0)).r - texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uv_pos - vec2(EPSILON, 0.0)).r, + texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uv_pos + vec2(0.0, EPSILON)).r - texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uv_pos - vec2(0.0, EPSILON)).r)); + + collided = true; + sdf_pos2 = sdf_pos + n * d; + pos2 = vec2(dot(vec4(sdf_pos2, 0, 1), FRAME.colliders[0].transform[2]), dot(vec4(sdf_pos2, 0, 1), FRAME.colliders[0].transform[3])); + + n = pos - pos2; + + collision_normal = normalize(vec3(n, 0.0)); + collision_depth = length(n); + } + } + + } else { + for (uint i = 0; i < FRAME.collider_count; i++) { + vec3 normal; + float depth; + bool col = false; + + vec3 rel_vec = PARTICLE.xform[3].xyz - FRAME.colliders[i].transform[3].xyz; + vec3 local_pos = rel_vec * mat3(FRAME.colliders[i].transform); + + switch (FRAME.colliders[i].type) { + case COLLIDER_TYPE_SPHERE: { + float d = length(rel_vec) - (particle_size + FRAME.colliders[i].extents.x); + + if (d < 0.0) { + col = true; + depth = -d; + normal = normalize(rel_vec); + } + + } break; + case COLLIDER_TYPE_BOX: { + vec3 abs_pos = abs(local_pos); + vec3 sgn_pos = sign(local_pos); + + if (any(greaterThan(abs_pos, FRAME.colliders[i].extents))) { + //point outside box + + vec3 closest = min(abs_pos, FRAME.colliders[i].extents); + vec3 rel = abs_pos - closest; + depth = length(rel) - particle_size; + if (depth < 0.0) { + col = true; + normal = mat3(FRAME.colliders[i].transform) * (normalize(rel) * sgn_pos); + depth = -depth; + } + } else { + //point inside box + vec3 axis_len = FRAME.colliders[i].extents - abs_pos; + // there has to be a faster way to do this? + if (all(lessThan(axis_len.xx, axis_len.yz))) { + normal = vec3(1, 0, 0); + } else if (all(lessThan(axis_len.yy, axis_len.xz))) { + normal = vec3(0, 1, 0); + } else { + normal = vec3(0, 0, 1); + } + + col = true; + depth = dot(normal * axis_len, vec3(1)) + particle_size; + normal = mat3(FRAME.colliders[i].transform) * (normal * sgn_pos); + } + + } break; + case COLLIDER_TYPE_SDF: { + vec3 apos = abs(local_pos); + float extra_dist = 0.0; + if (any(greaterThan(apos, FRAME.colliders[i].extents))) { //outside + vec3 mpos = min(apos, FRAME.colliders[i].extents); + extra_dist = distance(mpos, apos); + } + + if (extra_dist > particle_size) { + continue; + } + + vec3 uvw_pos = (local_pos / FRAME.colliders[i].extents) * 0.5 + 0.5; + float s = texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos).r; + s *= FRAME.colliders[i].scale; + s += extra_dist; + if (s < particle_size) { + col = true; + depth = particle_size - s; + const float EPSILON = 0.001; + normal = mat3(FRAME.colliders[i].transform) * + normalize( + vec3( + texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos + vec3(EPSILON, 0.0, 0.0)).r - texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos - vec3(EPSILON, 0.0, 0.0)).r, + texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos + vec3(0.0, EPSILON, 0.0)).r - texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos - vec3(0.0, EPSILON, 0.0)).r, + texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos + vec3(0.0, 0.0, EPSILON)).r - texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos - vec3(0.0, 0.0, EPSILON)).r)); + } + + } break; + case COLLIDER_TYPE_HEIGHT_FIELD: { + vec3 local_pos_bottom = local_pos; + local_pos_bottom.y -= particle_size; + + if (any(greaterThan(abs(local_pos_bottom), FRAME.colliders[i].extents))) { + continue; + } + const float DELTA = 1.0 / 8192.0; + + vec3 uvw_pos = vec3(local_pos_bottom / FRAME.colliders[i].extents) * 0.5 + 0.5; + + float y = 1.0 - texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos.xz).r; + + if (y > uvw_pos.y) { + //inside heightfield + + vec3 pos1 = (vec3(uvw_pos.x, y, uvw_pos.z) * 2.0 - 1.0) * FRAME.colliders[i].extents; + vec3 pos2 = (vec3(uvw_pos.x + DELTA, 1.0 - texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos.xz + vec2(DELTA, 0)).r, uvw_pos.z) * 2.0 - 1.0) * FRAME.colliders[i].extents; + vec3 pos3 = (vec3(uvw_pos.x, 1.0 - texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos.xz + vec2(0, DELTA)).r, uvw_pos.z + DELTA) * 2.0 - 1.0) * FRAME.colliders[i].extents; + + normal = normalize(cross(pos1 - pos2, pos1 - pos3)); + float local_y = (vec3(local_pos / FRAME.colliders[i].extents) * 0.5 + 0.5).y; + + col = true; + depth = dot(normal, pos1) - dot(normal, local_pos_bottom); + } + + } break; + } + + if (col) { + if (!collided) { + collided = true; + collision_normal = normal; + collision_depth = depth; + } else { + vec3 c = collision_normal * collision_depth; + c += normal * max(0.0, depth - dot(normal, c)); + collision_normal = normalize(c); + collision_depth = length(c); + } + } + } + } + } + + if (particle_active) { +#CODE : PROCESS + } + + PARTICLE.flags &= ~PARTICLE_FLAG_ACTIVE; + if (particle_active) { + PARTICLE.flags |= PARTICLE_FLAG_ACTIVE; + } +} diff --git a/servers/rendering/renderer_rd/shaders/particles_copy.glsl b/servers/rendering/renderer_rd/shaders/particles_copy.glsl new file mode 100644 index 0000000000..afbd5a9caa --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/particles_copy.glsl @@ -0,0 +1,233 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in; + +#define PARTICLE_FLAG_ACTIVE uint(1) +#define PARTICLE_FLAG_STARTED uint(2) +#define PARTICLE_FLAG_TRAILED uint(4) + +struct ParticleData { + mat4 xform; + vec3 velocity; + uint flags; + vec4 color; + vec4 custom; +#ifdef USERDATA_COUNT + vec4 userdata[USERDATA_COUNT]; +#endif +}; + +layout(set = 0, binding = 1, std430) restrict readonly buffer Particles { + ParticleData data[]; +} +particles; + +layout(set = 0, binding = 2, std430) restrict writeonly buffer Transforms { + vec4 data[]; +} +instances; + +#ifdef USE_SORT_BUFFER + +layout(set = 1, binding = 0, std430) restrict buffer SortBuffer { + vec2 data[]; +} +sort_buffer; + +#endif // USE_SORT_BUFFER + +layout(set = 2, binding = 0, std430) restrict readonly buffer TrailBindPoses { + mat4 data[]; +} +trail_bind_poses; + +layout(push_constant, std430) uniform Params { + vec3 sort_direction; + uint total_particles; + + uint trail_size; + uint trail_total; + float frame_delta; + float frame_remainder; + + vec3 align_up; + uint align_mode; + + bool order_by_lifetime; + uint lifetime_split; + bool lifetime_reverse; + bool copy_mode_2d; + + mat4 inv_emission_transform; +} +params; + +#define TRANSFORM_ALIGN_DISABLED 0 +#define TRANSFORM_ALIGN_Z_BILLBOARD 1 +#define TRANSFORM_ALIGN_Y_TO_VELOCITY 2 +#define TRANSFORM_ALIGN_Z_BILLBOARD_Y_TO_VELOCITY 3 + +void main() { +#ifdef MODE_FILL_SORT_BUFFER + + uint particle = gl_GlobalInvocationID.x; + if (particle >= params.total_particles) { + return; //discard + } + + uint src_particle = particle; + if (params.trail_size > 1) { + src_particle = src_particle * params.trail_size + params.trail_size / 2; //use trail center for sorting + } + sort_buffer.data[particle].x = dot(params.sort_direction, particles.data[src_particle].xform[3].xyz); + sort_buffer.data[particle].y = float(particle); +#endif + +#ifdef MODE_FILL_INSTANCES + + uint particle = gl_GlobalInvocationID.x; + + if (particle >= params.total_particles) { + return; //discard + } + +#ifdef USE_SORT_BUFFER + + if (params.trail_size > 1) { + particle = uint(sort_buffer.data[particle / params.trail_size].y) + (particle % params.trail_size); + } else { + particle = uint(sort_buffer.data[particle].y); //use index from sort buffer + } +#else + if (params.order_by_lifetime) { + if (params.trail_size > 1) { + uint limit = (params.total_particles / params.trail_size) - params.lifetime_split; + + uint base_index = particle / params.trail_size; + uint base_offset = particle % params.trail_size; + + if (params.lifetime_reverse) { + base_index = (params.total_particles / params.trail_size) - base_index - 1; + } + + if (base_index < limit) { + base_index = params.lifetime_split + base_index; + } else { + base_index -= limit; + } + + particle = base_index * params.trail_size + base_offset; + + } else { + uint limit = params.total_particles - params.lifetime_split; + + if (params.lifetime_reverse) { + particle = params.total_particles - particle - 1; + } + + if (particle < limit) { + particle = params.lifetime_split + particle; + } else { + particle -= limit; + } + } + } +#endif // USE_SORT_BUFFER + + mat4 txform; + + if (bool(particles.data[particle].flags & PARTICLE_FLAG_ACTIVE) || bool(particles.data[particle].flags & PARTICLE_FLAG_TRAILED)) { + txform = particles.data[particle].xform; + if (params.trail_size > 1) { + // Since the steps don't fit precisely in the history frames, must do a tiny bit of + // interpolation to get them close to their intended location. + uint part_ofs = particle % params.trail_size; + float natural_ofs = fract((float(part_ofs) / float(params.trail_size)) * float(params.trail_total)) * params.frame_delta; + + txform[3].xyz -= particles.data[particle].velocity * natural_ofs; + } + + switch (params.align_mode) { + case TRANSFORM_ALIGN_DISABLED: { + } break; //nothing + case TRANSFORM_ALIGN_Z_BILLBOARD: { + mat3 local = mat3(normalize(cross(params.align_up, params.sort_direction)), params.align_up, params.sort_direction); + local = local * mat3(txform); + txform[0].xyz = local[0]; + txform[1].xyz = local[1]; + txform[2].xyz = local[2]; + + } break; + case TRANSFORM_ALIGN_Y_TO_VELOCITY: { + vec3 v = particles.data[particle].velocity; + float s = (length(txform[0]) + length(txform[1]) + length(txform[2])) / 3.0; + if (length(v) > 0.0) { + txform[1].xyz = normalize(v); + } else { + txform[1].xyz = normalize(txform[1].xyz); + } + + txform[0].xyz = normalize(cross(txform[1].xyz, txform[2].xyz)); + txform[2].xyz = vec3(0.0, 0.0, 1.0) * s; + txform[0].xyz *= s; + txform[1].xyz *= s; + } break; + case TRANSFORM_ALIGN_Z_BILLBOARD_Y_TO_VELOCITY: { + vec3 v = particles.data[particle].velocity; + vec3 sv = v - params.sort_direction * dot(params.sort_direction, v); //screen velocity + float s = (length(txform[0]) + length(txform[1]) + length(txform[2])) / 3.0; + + if (length(sv) == 0) { + sv = params.align_up; + } + + sv = normalize(sv); + + txform[0].xyz = normalize(cross(sv, params.sort_direction)) * s; + txform[1].xyz = sv * s; + txform[2].xyz = params.sort_direction * s; + + } break; + } + + txform[3].xyz += particles.data[particle].velocity * params.frame_remainder; + + if (params.trail_size > 1) { + uint part_ofs = particle % params.trail_size; + txform = txform * trail_bind_poses.data[part_ofs]; + } + + if (params.copy_mode_2d) { + // In global mode, bring 2D particles to local coordinates + // as they will be drawn with the node position as origin. + txform = params.inv_emission_transform * txform; + } + + txform = transpose(txform); + } else { + txform = mat4(vec4(0.0), vec4(0.0), vec4(0.0), vec4(0.0)); //zero scale, becomes invisible + } + + if (params.copy_mode_2d) { + uint write_offset = gl_GlobalInvocationID.x * (2 + 1 + 1); //xform + color + custom + + instances.data[write_offset + 0] = txform[0]; + instances.data[write_offset + 1] = txform[1]; + instances.data[write_offset + 2] = particles.data[particle].color; + instances.data[write_offset + 3] = particles.data[particle].custom; + } else { + uint write_offset = gl_GlobalInvocationID.x * (3 + 1 + 1); //xform + color + custom + + instances.data[write_offset + 0] = txform[0]; + instances.data[write_offset + 1] = txform[1]; + instances.data[write_offset + 2] = txform[2]; + instances.data[write_offset + 3] = particles.data[particle].color; + instances.data[write_offset + 4] = particles.data[particle].custom; + } + +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/roughness_limiter.glsl b/servers/rendering/renderer_rd/shaders/roughness_limiter.glsl new file mode 100644 index 0000000000..59027df8e9 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/roughness_limiter.glsl @@ -0,0 +1,70 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +layout(set = 0, binding = 0) uniform sampler2D source_normal; +layout(r8, set = 1, binding = 0) uniform restrict writeonly image2D dest_roughness; + +layout(push_constant, std430) uniform Params { + ivec2 screen_size; + float curve; + uint pad; +} +params; + +#define HALF_PI 1.5707963267948966 + +void main() { + // Pixel being shaded + ivec2 pos = ivec2(gl_GlobalInvocationID.xy); + if (any(greaterThan(pos, params.screen_size))) { //too large, do nothing + return; + } + + vec3 normal_accum = vec3(0.0); + float accum = 0.0; + for (int i = 0; i <= 1; i++) { + for (int j = 0; j <= 1; j++) { + normal_accum += normalize(texelFetch(source_normal, pos + ivec2(i, j), 0).xyz * 2.0 - 1.0); + accum += 1.0; + } + } + + normal_accum /= accum; + + float r = length(normal_accum); + + float limit; + + if (r < 1.0) { + float threshold = 0.4; + + /* + //Formula from Filament, does not make sense to me. + + float r2 = r * r; + float kappa = (3.0f * r - r * r2) / (1.0f - r2); + float variance = 0.25f / kappa; + limit = sqrt(min(2.0f * variance, threshold * threshold)); + */ + /* + //Formula based on probability distribution graph + + float width = acos(max(0.0,r)); // convert to angle (width) + float roughness = pow(width,1.7)*0.854492; //approximate (crappy) formula to convert to roughness + limit = min(sqrt(roughness), threshold); //convert to perceptual roughness and apply threshold + */ + + limit = min(sqrt(pow(acos(max(0.0, r)) / HALF_PI, params.curve)), threshold); //convert to perceptual roughness and apply threshold + + //limit = 0.5; + } else { + limit = 0.0; + } + + imageStore(dest_roughness, pos, vec4(limit)); +} diff --git a/servers/rendering/renderer_rd/shaders/scene_data_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_data_inc.glsl new file mode 100644 index 0000000000..048257e9ef --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/scene_data_inc.glsl @@ -0,0 +1,69 @@ +// Scene data stores all our 3D rendering globals for a frame such as our matrices +// where this information is independent of the different RD implementations. +// This enables us to use this UBO in our main scene render shaders but also in +// effects that need access to this data. + +struct SceneData { + highp mat4 projection_matrix; + highp mat4 inv_projection_matrix; + highp mat4 inv_view_matrix; + highp mat4 view_matrix; + + // only used for multiview + highp mat4 projection_matrix_view[MAX_VIEWS]; + highp mat4 inv_projection_matrix_view[MAX_VIEWS]; + highp vec4 eye_offset[MAX_VIEWS]; + + highp vec2 viewport_size; + highp vec2 screen_pixel_size; + + // Use vec4s because std140 doesn't play nice with vec2s, z and w are wasted. + highp vec4 directional_penumbra_shadow_kernel[32]; + highp vec4 directional_soft_shadow_kernel[32]; + highp vec4 penumbra_shadow_kernel[32]; + highp vec4 soft_shadow_kernel[32]; + + mediump mat3 radiance_inverse_xform; + + mediump vec4 ambient_light_color_energy; + + mediump float ambient_color_sky_mix; + bool use_ambient_light; + bool use_ambient_cubemap; + bool use_reflection_cubemap; + + highp vec2 shadow_atlas_pixel_size; + highp vec2 directional_shadow_pixel_size; + + uint directional_light_count; + mediump float dual_paraboloid_side; + highp float z_far; + highp float z_near; + + bool roughness_limiter_enabled; + mediump float roughness_limiter_amount; + mediump float roughness_limiter_limit; + mediump float opaque_prepass_threshold; + + bool fog_enabled; + highp float fog_density; + highp float fog_height; + highp float fog_height_density; + + mediump vec3 fog_light_color; + mediump float fog_sun_scatter; + + mediump float fog_aerial_perspective; + highp float time; + mediump float reflection_multiplier; // one normally, zero when rendering reflections + bool material_uv2_mode; + + vec2 taa_jitter; + float emissive_exposure_normalization; + float IBL_exposure_normalization; + + bool pancake_shadows; + uint pad1; + uint pad2; + uint pad3; +}; diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_aa_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_aa_inc.glsl new file mode 100644 index 0000000000..97c913d489 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/scene_forward_aa_inc.glsl @@ -0,0 +1,57 @@ +#ifdef ALPHA_HASH_USED + +float hash_2d(vec2 p) { + return fract(1.0e4 * sin(17.0 * p.x + 0.1 * p.y) * + (0.1 + abs(sin(13.0 * p.y + p.x)))); +} + +float hash_3d(vec3 p) { + return hash_2d(vec2(hash_2d(p.xy), p.z)); +} + +float compute_alpha_hash_threshold(vec3 pos, float hash_scale) { + vec3 dx = dFdx(pos); + vec3 dy = dFdx(pos); + float delta_max_sqr = max(length(dx), length(dy)); + float pix_scale = 1.0 / (hash_scale * delta_max_sqr); + + vec2 pix_scales = + vec2(exp2(floor(log2(pix_scale))), exp2(ceil(log2(pix_scale)))); + + vec2 a_thresh = vec2(hash_3d(floor(pix_scales.x * pos.xyz)), + hash_3d(floor(pix_scales.y * pos.xyz))); + + float lerp_factor = fract(log2(pix_scale)); + + float a_interp = (1.0 - lerp_factor) * a_thresh.x + lerp_factor * a_thresh.y; + + float min_lerp = min(lerp_factor, 1.0 - lerp_factor); + + vec3 cases = vec3(a_interp * a_interp / (2.0 * min_lerp * (1.0 - min_lerp)), + (a_interp - 0.5 * min_lerp) / (1.0 - min_lerp), + 1.0 - ((1.0 - a_interp) * (1.0 - a_interp) / (2.0 * min_lerp * (1.0 - min_lerp)))); + + float alpha_hash_threshold = + (lerp_factor < (1.0 - min_lerp)) ? ((lerp_factor < min_lerp) ? cases.x : cases.y) : cases.z; + + return clamp(alpha_hash_threshold, 0.0, 1.0); +} + +#endif // ALPHA_HASH_USED + +#ifdef ALPHA_ANTIALIASING_EDGE_USED + +float calc_mip_level(vec2 texture_coord) { + vec2 dx = dFdx(texture_coord); + vec2 dy = dFdy(texture_coord); + float delta_max_sqr = max(dot(dx, dx), dot(dy, dy)); + return max(0.0, 0.5 * log2(delta_max_sqr)); +} + +float compute_alpha_antialiasing_edge(float input_alpha, vec2 texture_coord, float alpha_edge) { + input_alpha *= 1.0 + max(0, calc_mip_level(texture_coord)) * 0.25; // 0.25 mip scale, magic number + input_alpha = (input_alpha - alpha_edge) / max(fwidth(input_alpha), 0.0001) + 0.5; + return clamp(input_alpha, 0.0, 1.0); +} + +#endif // ALPHA_ANTIALIASING_USED diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_gi_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_gi_inc.glsl new file mode 100644 index 0000000000..ae5e1b7251 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/scene_forward_gi_inc.glsl @@ -0,0 +1,242 @@ +// Functions related to gi/sdfgi for our forward renderer + +//standard voxel cone trace +vec4 voxel_cone_trace(texture3D probe, vec3 cell_size, vec3 pos, vec3 direction, float tan_half_angle, float max_distance, float p_bias) { + float dist = p_bias; + vec4 color = vec4(0.0); + + while (dist < max_distance && color.a < 0.95) { + float diameter = max(1.0, 2.0 * tan_half_angle * dist); + vec3 uvw_pos = (pos + dist * direction) * cell_size; + float half_diameter = diameter * 0.5; + //check if outside, then break + if (any(greaterThan(abs(uvw_pos - 0.5), vec3(0.5f + half_diameter * cell_size)))) { + break; + } + vec4 scolor = textureLod(sampler3D(probe, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uvw_pos, log2(diameter)); + float a = (1.0 - color.a); + color += a * scolor; + dist += half_diameter; + } + + return color; +} + +vec4 voxel_cone_trace_45_degrees(texture3D probe, vec3 cell_size, vec3 pos, vec3 direction, float tan_half_angle, float max_distance, float p_bias) { + float dist = p_bias; + vec4 color = vec4(0.0); + float radius = max(0.5, tan_half_angle * dist); + float lod_level = log2(radius * 2.0); + + while (dist < max_distance && color.a < 0.95) { + vec3 uvw_pos = (pos + dist * direction) * cell_size; + + //check if outside, then break + if (any(greaterThan(abs(uvw_pos - 0.5), vec3(0.5f + radius * cell_size)))) { + break; + } + vec4 scolor = textureLod(sampler3D(probe, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uvw_pos, lod_level); + lod_level += 1.0; + + float a = (1.0 - color.a); + scolor *= a; + color += scolor; + dist += radius; + radius = max(0.5, tan_half_angle * dist); + } + + return color; +} + +void voxel_gi_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3 normal_xform, float roughness, vec3 ambient, vec3 environment, inout vec4 out_spec, inout vec4 out_diff) { + position = (voxel_gi_instances.data[index].xform * vec4(position, 1.0)).xyz; + ref_vec = normalize((voxel_gi_instances.data[index].xform * vec4(ref_vec, 0.0)).xyz); + normal = normalize((voxel_gi_instances.data[index].xform * vec4(normal, 0.0)).xyz); + + position += normal * voxel_gi_instances.data[index].normal_bias; + + //this causes corrupted pixels, i have no idea why.. + if (any(bvec2(any(lessThan(position, vec3(0.0))), any(greaterThan(position, voxel_gi_instances.data[index].bounds))))) { + return; + } + + vec3 blendv = abs(position / voxel_gi_instances.data[index].bounds * 2.0 - 1.0); + float blend = clamp(1.0 - max(blendv.x, max(blendv.y, blendv.z)), 0.0, 1.0); + //float blend=1.0; + + float max_distance = length(voxel_gi_instances.data[index].bounds); + vec3 cell_size = 1.0 / voxel_gi_instances.data[index].bounds; + + //radiance + +#define MAX_CONE_DIRS 4 + + vec3 cone_dirs[MAX_CONE_DIRS] = vec3[]( + vec3(0.707107, 0.0, 0.707107), + vec3(0.0, 0.707107, 0.707107), + vec3(-0.707107, 0.0, 0.707107), + vec3(0.0, -0.707107, 0.707107)); + + float cone_weights[MAX_CONE_DIRS] = float[](0.25, 0.25, 0.25, 0.25); + float cone_angle_tan = 0.98269; + + vec3 light = vec3(0.0); + + for (int i = 0; i < MAX_CONE_DIRS; i++) { + vec3 dir = normalize((voxel_gi_instances.data[index].xform * vec4(normal_xform * cone_dirs[i], 0.0)).xyz); + + vec4 cone_light = voxel_cone_trace_45_degrees(voxel_gi_textures[index], cell_size, position, dir, cone_angle_tan, max_distance, voxel_gi_instances.data[index].bias); + + if (voxel_gi_instances.data[index].blend_ambient) { + cone_light.rgb = mix(ambient, cone_light.rgb, min(1.0, cone_light.a / 0.95)); + } + + light += cone_weights[i] * cone_light.rgb; + } + + light *= voxel_gi_instances.data[index].dynamic_range * voxel_gi_instances.data[index].exposure_normalization; + out_diff += vec4(light * blend, blend); + + //irradiance + vec4 irr_light = voxel_cone_trace(voxel_gi_textures[index], cell_size, position, ref_vec, tan(roughness * 0.5 * M_PI * 0.99), max_distance, voxel_gi_instances.data[index].bias); + if (voxel_gi_instances.data[index].blend_ambient) { + irr_light.rgb = mix(environment, irr_light.rgb, min(1.0, irr_light.a / 0.95)); + } + irr_light.rgb *= voxel_gi_instances.data[index].dynamic_range * voxel_gi_instances.data[index].exposure_normalization; + //irr_light=vec3(0.0); + + out_spec += vec4(irr_light.rgb * blend, blend); +} + +vec2 octahedron_wrap(vec2 v) { + vec2 signVal; + signVal.x = v.x >= 0.0 ? 1.0 : -1.0; + signVal.y = v.y >= 0.0 ? 1.0 : -1.0; + return (1.0 - abs(v.yx)) * signVal; +} + +vec2 octahedron_encode(vec3 n) { + // https://twitter.com/Stubbesaurus/status/937994790553227264 + n /= (abs(n.x) + abs(n.y) + abs(n.z)); + n.xy = n.z >= 0.0 ? n.xy : octahedron_wrap(n.xy); + n.xy = n.xy * 0.5 + 0.5; + return n.xy; +} + +void sdfgi_process(uint cascade, vec3 cascade_pos, vec3 cam_pos, vec3 cam_normal, vec3 cam_specular_normal, bool use_specular, float roughness, out vec3 diffuse_light, out vec3 specular_light, out float blend) { + cascade_pos += cam_normal * sdfgi.normal_bias; + + vec3 base_pos = floor(cascade_pos); + //cascade_pos += mix(vec3(0.0),vec3(0.01),lessThan(abs(cascade_pos-base_pos),vec3(0.01))) * cam_normal; + ivec3 probe_base_pos = ivec3(base_pos); + + vec4 diffuse_accum = vec4(0.0); + vec3 specular_accum; + + ivec3 tex_pos = ivec3(probe_base_pos.xy, int(cascade)); + tex_pos.x += probe_base_pos.z * sdfgi.probe_axis_size; + tex_pos.xy = tex_pos.xy * (SDFGI_OCT_SIZE + 2) + ivec2(1); + + vec3 diffuse_posf = (vec3(tex_pos) + vec3(octahedron_encode(cam_normal) * float(SDFGI_OCT_SIZE), 0.0)) * sdfgi.lightprobe_tex_pixel_size; + + vec3 specular_posf; + + if (use_specular) { + specular_accum = vec3(0.0); + specular_posf = (vec3(tex_pos) + vec3(octahedron_encode(cam_specular_normal) * float(SDFGI_OCT_SIZE), 0.0)) * sdfgi.lightprobe_tex_pixel_size; + } + + vec4 light_accum = vec4(0.0); + float weight_accum = 0.0; + + for (uint j = 0; j < 8; j++) { + ivec3 offset = (ivec3(j) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1); + ivec3 probe_posi = probe_base_pos; + probe_posi += offset; + + // Compute weight + + vec3 probe_pos = vec3(probe_posi); + vec3 probe_to_pos = cascade_pos - probe_pos; + vec3 probe_dir = normalize(-probe_to_pos); + + vec3 trilinear = vec3(1.0) - abs(probe_to_pos); + float weight = trilinear.x * trilinear.y * trilinear.z * max(0.005, dot(cam_normal, probe_dir)); + + // Compute lightprobe occlusion + + if (sdfgi.use_occlusion) { + ivec3 occ_indexv = abs((sdfgi.cascades[cascade].probe_world_offset + probe_posi) & ivec3(1, 1, 1)) * ivec3(1, 2, 4); + vec4 occ_mask = mix(vec4(0.0), vec4(1.0), equal(ivec4(occ_indexv.x | occ_indexv.y), ivec4(0, 1, 2, 3))); + + vec3 occ_pos = clamp(cascade_pos, probe_pos - sdfgi.occlusion_clamp, probe_pos + sdfgi.occlusion_clamp) * sdfgi.probe_to_uvw; + occ_pos.z += float(cascade); + if (occ_indexv.z != 0) { //z bit is on, means index is >=4, so make it switch to the other half of textures + occ_pos.x += 1.0; + } + + occ_pos *= sdfgi.occlusion_renormalize; + float occlusion = dot(textureLod(sampler3D(sdfgi_occlusion_cascades, material_samplers[SAMPLER_LINEAR_CLAMP]), occ_pos, 0.0), occ_mask); + + weight *= max(occlusion, 0.01); + } + + // Compute lightprobe texture position + + vec3 diffuse; + vec3 pos_uvw = diffuse_posf; + pos_uvw.xy += vec2(offset.xy) * sdfgi.lightprobe_uv_offset.xy; + pos_uvw.x += float(offset.z) * sdfgi.lightprobe_uv_offset.z; + diffuse = textureLod(sampler2DArray(sdfgi_lightprobe_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), pos_uvw, 0.0).rgb; + + diffuse_accum += vec4(diffuse * weight * sdfgi.cascades[cascade].exposure_normalization, weight); + + if (use_specular) { + vec3 specular = vec3(0.0); + vec3 pos_uvw = specular_posf; + pos_uvw.xy += vec2(offset.xy) * sdfgi.lightprobe_uv_offset.xy; + pos_uvw.x += float(offset.z) * sdfgi.lightprobe_uv_offset.z; + if (roughness < 0.99) { + specular = textureLod(sampler2DArray(sdfgi_lightprobe_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), pos_uvw + vec3(0, 0, float(sdfgi.max_cascades)), 0.0).rgb; + } + if (roughness > 0.5) { + specular = mix(specular, textureLod(sampler2DArray(sdfgi_lightprobe_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), pos_uvw, 0.0).rgb, (roughness - 0.5) * 2.0); + } + + specular_accum += specular * weight * sdfgi.cascades[cascade].exposure_normalization; + } + } + + if (diffuse_accum.a > 0.0) { + diffuse_accum.rgb /= diffuse_accum.a; + } + + diffuse_light = diffuse_accum.rgb; + + if (use_specular) { + if (diffuse_accum.a > 0.0) { + specular_accum /= diffuse_accum.a; + } + + specular_light = specular_accum; + } + + { + //process blend + float blend_from = (float(sdfgi.probe_axis_size - 1) / 2.0) - 2.5; + float blend_to = blend_from + 2.0; + + vec3 inner_pos = cam_pos * sdfgi.cascades[cascade].to_probe; + + float len = length(inner_pos); + + inner_pos = abs(normalize(inner_pos)); + len *= max(inner_pos.x, max(inner_pos.y, inner_pos.z)); + + if (len >= blend_from) { + blend = smoothstep(blend_from, blend_to, len); + } else { + blend = 0.0; + } + } +} diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl new file mode 100644 index 0000000000..2fba1351f7 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl @@ -0,0 +1,976 @@ +// Functions related to lighting + +float D_GGX(float cos_theta_m, float alpha) { + float a = cos_theta_m * alpha; + float k = alpha / (1.0 - cos_theta_m * cos_theta_m + a * a); + return k * k * (1.0 / M_PI); +} + +// From Earl Hammon, Jr. "PBR Diffuse Lighting for GGX+Smith Microsurfaces" https://www.gdcvault.com/play/1024478/PBR-Diffuse-Lighting-for-GGX +float V_GGX(float NdotL, float NdotV, float alpha) { + return 0.5 / mix(2.0 * NdotL * NdotV, NdotL + NdotV, alpha); +} + +float D_GGX_anisotropic(float cos_theta_m, float alpha_x, float alpha_y, float cos_phi, float sin_phi) { + float alpha2 = alpha_x * alpha_y; + highp vec3 v = vec3(alpha_y * cos_phi, alpha_x * sin_phi, alpha2 * cos_theta_m); + highp float v2 = dot(v, v); + float w2 = alpha2 / v2; + float D = alpha2 * w2 * w2 * (1.0 / M_PI); + return D; +} + +float V_GGX_anisotropic(float alpha_x, float alpha_y, float TdotV, float TdotL, float BdotV, float BdotL, float NdotV, float NdotL) { + float Lambda_V = NdotL * length(vec3(alpha_x * TdotV, alpha_y * BdotV, NdotV)); + float Lambda_L = NdotV * length(vec3(alpha_x * TdotL, alpha_y * BdotL, NdotL)); + return 0.5 / (Lambda_V + Lambda_L); +} + +float SchlickFresnel(float u) { + float m = 1.0 - u; + float m2 = m * m; + return m2 * m2 * m; // pow(m,5) +} + +vec3 F0(float metallic, float specular, vec3 albedo) { + float dielectric = 0.16 * specular * specular; + // use albedo * metallic as colored specular reflectance at 0 angle for metallic materials; + // see https://google.github.io/filament/Filament.md.html + return mix(vec3(dielectric), albedo, vec3(metallic)); +} + +void light_compute(vec3 N, vec3 L, vec3 V, float A, vec3 light_color, float attenuation, vec3 f0, uint orms, float specular_amount, vec3 albedo, inout float alpha, +#ifdef LIGHT_BACKLIGHT_USED + vec3 backlight, +#endif +#ifdef LIGHT_TRANSMITTANCE_USED + vec4 transmittance_color, + float transmittance_depth, + float transmittance_boost, + float transmittance_z, +#endif +#ifdef LIGHT_RIM_USED + float rim, float rim_tint, +#endif +#ifdef LIGHT_CLEARCOAT_USED + float clearcoat, float clearcoat_roughness, vec3 vertex_normal, +#endif +#ifdef LIGHT_ANISOTROPY_USED + vec3 B, vec3 T, float anisotropy, +#endif + inout vec3 diffuse_light, inout vec3 specular_light) { + + vec4 orms_unpacked = unpackUnorm4x8(orms); + + float roughness = orms_unpacked.y; + float metallic = orms_unpacked.z; + +#if defined(LIGHT_CODE_USED) + // light is written by the light shader + + vec3 normal = N; + vec3 light = L; + vec3 view = V; + +#CODE : LIGHT + +#else + + float NdotL = min(A + dot(N, L), 1.0); + float cNdotL = max(NdotL, 0.0); // clamped NdotL + float NdotV = dot(N, V); + float cNdotV = max(NdotV, 1e-4); + +#if defined(DIFFUSE_BURLEY) || defined(SPECULAR_SCHLICK_GGX) || defined(LIGHT_CLEARCOAT_USED) + vec3 H = normalize(V + L); +#endif + +#if defined(SPECULAR_SCHLICK_GGX) + float cNdotH = clamp(A + dot(N, H), 0.0, 1.0); +#endif + +#if defined(DIFFUSE_BURLEY) || defined(SPECULAR_SCHLICK_GGX) || defined(LIGHT_CLEARCOAT_USED) + float cLdotH = clamp(A + dot(L, H), 0.0, 1.0); +#endif + + if (metallic < 1.0) { + float diffuse_brdf_NL; // BRDF times N.L for calculating diffuse radiance + +#if defined(DIFFUSE_LAMBERT_WRAP) + // Energy conserving lambert wrap shader. + // https://web.archive.org/web/20210228210901/http://blog.stevemcauley.com/2011/12/03/energy-conserving-wrapped-diffuse/ + diffuse_brdf_NL = max(0.0, (NdotL + roughness) / ((1.0 + roughness) * (1.0 + roughness))) * (1.0 / M_PI); +#elif defined(DIFFUSE_TOON) + + diffuse_brdf_NL = smoothstep(-roughness, max(roughness, 0.01), NdotL) * (1.0 / M_PI); + +#elif defined(DIFFUSE_BURLEY) + + { + float FD90_minus_1 = 2.0 * cLdotH * cLdotH * roughness - 0.5; + float FdV = 1.0 + FD90_minus_1 * SchlickFresnel(cNdotV); + float FdL = 1.0 + FD90_minus_1 * SchlickFresnel(cNdotL); + diffuse_brdf_NL = (1.0 / M_PI) * FdV * FdL * cNdotL; + /* + float energyBias = mix(roughness, 0.0, 0.5); + float energyFactor = mix(roughness, 1.0, 1.0 / 1.51); + float fd90 = energyBias + 2.0 * VoH * VoH * roughness; + float f0 = 1.0; + float lightScatter = f0 + (fd90 - f0) * pow(1.0 - cNdotL, 5.0); + float viewScatter = f0 + (fd90 - f0) * pow(1.0 - cNdotV, 5.0); + + diffuse_brdf_NL = lightScatter * viewScatter * energyFactor; + */ + } +#else + // lambert + diffuse_brdf_NL = cNdotL * (1.0 / M_PI); +#endif + + diffuse_light += light_color * diffuse_brdf_NL * attenuation; + +#if defined(LIGHT_BACKLIGHT_USED) + diffuse_light += light_color * (vec3(1.0 / M_PI) - diffuse_brdf_NL) * backlight * attenuation; +#endif + +#if defined(LIGHT_RIM_USED) + float rim_light = pow(max(0.0, 1.0 - cNdotV), max(0.0, (1.0 - roughness) * 16.0)); + diffuse_light += rim_light * rim * mix(vec3(1.0), albedo, rim_tint) * light_color; +#endif + +#ifdef LIGHT_TRANSMITTANCE_USED + + { +#ifdef SSS_MODE_SKIN + float scale = 8.25 / transmittance_depth; + float d = scale * abs(transmittance_z); + float dd = -d * d; + vec3 profile = vec3(0.233, 0.455, 0.649) * exp(dd / 0.0064) + + vec3(0.1, 0.336, 0.344) * exp(dd / 0.0484) + + vec3(0.118, 0.198, 0.0) * exp(dd / 0.187) + + vec3(0.113, 0.007, 0.007) * exp(dd / 0.567) + + vec3(0.358, 0.004, 0.0) * exp(dd / 1.99) + + vec3(0.078, 0.0, 0.0) * exp(dd / 7.41); + + diffuse_light += profile * transmittance_color.a * light_color * clamp(transmittance_boost - NdotL, 0.0, 1.0) * (1.0 / M_PI); +#else + + float scale = 8.25 / transmittance_depth; + float d = scale * abs(transmittance_z); + float dd = -d * d; + diffuse_light += exp(dd) * transmittance_color.rgb * transmittance_color.a * light_color * clamp(transmittance_boost - NdotL, 0.0, 1.0) * (1.0 / M_PI); +#endif + } +#else + +#endif //LIGHT_TRANSMITTANCE_USED + } + + if (roughness > 0.0) { // FIXME: roughness == 0 should not disable specular light entirely + + // D + +#if defined(SPECULAR_TOON) + + vec3 R = normalize(-reflect(L, N)); + float RdotV = dot(R, V); + float mid = 1.0 - roughness; + mid *= mid; + float intensity = smoothstep(mid - roughness * 0.5, mid + roughness * 0.5, RdotV) * mid; + diffuse_light += light_color * intensity * attenuation * specular_amount; // write to diffuse_light, as in toon shading you generally want no reflection + +#elif defined(SPECULAR_DISABLED) + // none.. + +#elif defined(SPECULAR_SCHLICK_GGX) + // shlick+ggx as default + float alpha_ggx = roughness * roughness; +#if defined(LIGHT_ANISOTROPY_USED) + + float aspect = sqrt(1.0 - anisotropy * 0.9); + float ax = alpha_ggx / aspect; + float ay = alpha_ggx * aspect; + float XdotH = dot(T, H); + float YdotH = dot(B, H); + float D = D_GGX_anisotropic(cNdotH, ax, ay, XdotH, YdotH); + float G = V_GGX_anisotropic(ax, ay, dot(T, V), dot(T, L), dot(B, V), dot(B, L), cNdotV, cNdotL); +#else // LIGHT_ANISOTROPY_USED + float D = D_GGX(cNdotH, alpha_ggx); + float G = V_GGX(cNdotL, cNdotV, alpha_ggx); +#endif // LIGHT_ANISOTROPY_USED + // F + float cLdotH5 = SchlickFresnel(cLdotH); + // Calculate Fresnel using specular occlusion term from Filament: + // https://google.github.io/filament/Filament.html#lighting/occlusion/specularocclusion + float f90 = clamp(dot(f0, vec3(50.0 * 0.33)), 0.0, 1.0); + vec3 F = f0 + (f90 - f0) * cLdotH5; + + vec3 specular_brdf_NL = cNdotL * D * F * G; + + specular_light += specular_brdf_NL * light_color * attenuation * specular_amount; +#endif + +#if defined(LIGHT_CLEARCOAT_USED) + // Clearcoat ignores normal_map, use vertex normal instead + float ccNdotL = max(min(A + dot(vertex_normal, L), 1.0), 0.0); + float ccNdotH = clamp(A + dot(vertex_normal, H), 0.0, 1.0); + float ccNdotV = max(dot(vertex_normal, V), 1e-4); + +#if !defined(SPECULAR_SCHLICK_GGX) + float cLdotH5 = SchlickFresnel(cLdotH); +#endif + float Dr = D_GGX(ccNdotH, mix(0.001, 0.1, clearcoat_roughness)); + float Gr = 0.25 / (cLdotH * cLdotH); + float Fr = mix(.04, 1.0, cLdotH5); + float clearcoat_specular_brdf_NL = clearcoat * Gr * Fr * Dr * cNdotL; + + specular_light += clearcoat_specular_brdf_NL * light_color * attenuation * specular_amount; + // TODO: Clearcoat adds light to the scene right now (it is non-energy conserving), both diffuse and specular need to be scaled by (1.0 - FR) + // but to do so we need to rearrange this entire function +#endif // LIGHT_CLEARCOAT_USED + } + +#ifdef USE_SHADOW_TO_OPACITY + alpha = min(alpha, clamp(1.0 - attenuation, 0.0, 1.0)); +#endif + +#endif //defined(LIGHT_CODE_USED) +} + +#ifndef SHADOWS_DISABLED + +// Interleaved Gradient Noise +// https://www.iryoku.com/next-generation-post-processing-in-call-of-duty-advanced-warfare +float quick_hash(vec2 pos) { + const vec3 magic = vec3(0.06711056f, 0.00583715f, 52.9829189f); + return fract(magic.z * fract(dot(pos, magic.xy))); +} + +float sample_directional_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, vec4 coord) { + vec2 pos = coord.xy; + float depth = coord.z; + + //if only one sample is taken, take it from the center + if (sc_directional_soft_shadow_samples == 0) { + return textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos, depth, 1.0)); + } + + mat2 disk_rotation; + { + float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; + float sr = sin(r); + float cr = cos(r); + disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); + } + + float avg = 0.0; + + for (uint i = 0; i < sc_directional_soft_shadow_samples; i++) { + avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + shadow_pixel_size * (disk_rotation * scene_data_block.data.directional_soft_shadow_kernel[i].xy), depth, 1.0)); + } + + return avg * (1.0 / float(sc_directional_soft_shadow_samples)); +} + +float sample_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, vec3 coord) { + vec2 pos = coord.xy; + float depth = coord.z; + + //if only one sample is taken, take it from the center + if (sc_soft_shadow_samples == 0) { + return textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos, depth, 1.0)); + } + + mat2 disk_rotation; + { + float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; + float sr = sin(r); + float cr = cos(r); + disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); + } + + float avg = 0.0; + + for (uint i = 0; i < sc_soft_shadow_samples; i++) { + avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + shadow_pixel_size * (disk_rotation * scene_data_block.data.soft_shadow_kernel[i].xy), depth, 1.0)); + } + + return avg * (1.0 / float(sc_soft_shadow_samples)); +} + +float sample_omni_pcf_shadow(texture2D shadow, float blur_scale, vec2 coord, vec4 uv_rect, vec2 flip_offset, float depth) { + //if only one sample is taken, take it from the center + if (sc_soft_shadow_samples == 0) { + vec2 pos = coord * 0.5 + 0.5; + pos = uv_rect.xy + pos * uv_rect.zw; + return textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos, depth, 1.0)); + } + + mat2 disk_rotation; + { + float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; + float sr = sin(r); + float cr = cos(r); + disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); + } + + float avg = 0.0; + vec2 offset_scale = blur_scale * 2.0 * scene_data_block.data.shadow_atlas_pixel_size / uv_rect.zw; + + for (uint i = 0; i < sc_soft_shadow_samples; i++) { + vec2 offset = offset_scale * (disk_rotation * scene_data_block.data.soft_shadow_kernel[i].xy); + vec2 sample_coord = coord + offset; + + float sample_coord_length_sqaured = dot(sample_coord, sample_coord); + bool do_flip = sample_coord_length_sqaured > 1.0; + + if (do_flip) { + float len = sqrt(sample_coord_length_sqaured); + sample_coord = sample_coord * (2.0 / len - 1.0); + } + + sample_coord = sample_coord * 0.5 + 0.5; + sample_coord = uv_rect.xy + sample_coord * uv_rect.zw; + + if (do_flip) { + sample_coord += flip_offset; + } + avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(sample_coord, depth, 1.0)); + } + + return avg * (1.0 / float(sc_soft_shadow_samples)); +} + +float sample_directional_soft_shadow(texture2D shadow, vec3 pssm_coord, vec2 tex_scale) { + //find blocker + float blocker_count = 0.0; + float blocker_average = 0.0; + + mat2 disk_rotation; + { + float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; + float sr = sin(r); + float cr = cos(r); + disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); + } + + for (uint i = 0; i < sc_directional_penumbra_shadow_samples; i++) { + vec2 suv = pssm_coord.xy + (disk_rotation * scene_data_block.data.directional_penumbra_shadow_kernel[i].xy) * tex_scale; + float d = textureLod(sampler2D(shadow, material_samplers[SAMPLER_LINEAR_CLAMP]), suv, 0.0).r; + if (d < pssm_coord.z) { + blocker_average += d; + blocker_count += 1.0; + } + } + + if (blocker_count > 0.0) { + //blockers found, do soft shadow + blocker_average /= blocker_count; + float penumbra = (pssm_coord.z - blocker_average) / blocker_average; + tex_scale *= penumbra; + + float s = 0.0; + for (uint i = 0; i < sc_directional_penumbra_shadow_samples; i++) { + vec2 suv = pssm_coord.xy + (disk_rotation * scene_data_block.data.directional_penumbra_shadow_kernel[i].xy) * tex_scale; + s += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(suv, pssm_coord.z, 1.0)); + } + + return s / float(sc_directional_penumbra_shadow_samples); + + } else { + //no blockers found, so no shadow + return 1.0; + } +} + +#endif // SHADOWS_DISABLED + +float get_omni_attenuation(float distance, float inv_range, float decay) { + float nd = distance * inv_range; + nd *= nd; + nd *= nd; // nd^4 + nd = max(1.0 - nd, 0.0); + nd *= nd; // nd^2 + return nd * pow(max(distance, 0.0001), -decay); +} + +float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { +#ifndef SHADOWS_DISABLED + if (omni_lights.data[idx].shadow_opacity > 0.001) { + // there is a shadowmap + vec2 texel_size = scene_data_block.data.shadow_atlas_pixel_size; + vec4 base_uv_rect = omni_lights.data[idx].atlas_rect; + base_uv_rect.xy += texel_size; + base_uv_rect.zw -= texel_size * 2.0; + + // Omni lights use direction.xy to store to store the offset between the two paraboloid regions + vec2 flip_offset = omni_lights.data[idx].direction.xy; + + vec3 local_vert = (omni_lights.data[idx].shadow_matrix * vec4(vertex, 1.0)).xyz; + + float shadow_len = length(local_vert); //need to remember shadow len from here + vec3 shadow_dir = normalize(local_vert); + + vec3 local_normal = normalize(mat3(omni_lights.data[idx].shadow_matrix) * normal); + vec3 normal_bias = local_normal * omni_lights.data[idx].shadow_normal_bias * (1.0 - abs(dot(local_normal, shadow_dir))); + + float shadow; + + if (sc_use_light_soft_shadows && omni_lights.data[idx].soft_shadow_size > 0.0) { + //soft shadow + + //find blocker + + float blocker_count = 0.0; + float blocker_average = 0.0; + + mat2 disk_rotation; + { + float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; + float sr = sin(r); + float cr = cos(r); + disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); + } + + vec3 basis_normal = shadow_dir; + vec3 v0 = abs(basis_normal.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 1.0, 0.0); + vec3 tangent = normalize(cross(v0, basis_normal)); + vec3 bitangent = normalize(cross(tangent, basis_normal)); + float z_norm = shadow_len * omni_lights.data[idx].inv_radius; + + tangent *= omni_lights.data[idx].soft_shadow_size * omni_lights.data[idx].soft_shadow_scale; + bitangent *= omni_lights.data[idx].soft_shadow_size * omni_lights.data[idx].soft_shadow_scale; + + for (uint i = 0; i < sc_penumbra_shadow_samples; i++) { + vec2 disk = disk_rotation * scene_data_block.data.penumbra_shadow_kernel[i].xy; + + vec3 pos = local_vert + tangent * disk.x + bitangent * disk.y; + + pos = normalize(pos); + + vec4 uv_rect = base_uv_rect; + + if (pos.z >= 0.0) { + uv_rect.xy += flip_offset; + } + + pos.z = 1.0 + abs(pos.z); + pos.xy /= pos.z; + + pos.xy = pos.xy * 0.5 + 0.5; + pos.xy = uv_rect.xy + pos.xy * uv_rect.zw; + + float d = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), pos.xy, 0.0).r; + if (d < z_norm) { + blocker_average += d; + blocker_count += 1.0; + } + } + + if (blocker_count > 0.0) { + //blockers found, do soft shadow + blocker_average /= blocker_count; + float penumbra = (z_norm - blocker_average) / blocker_average; + tangent *= penumbra; + bitangent *= penumbra; + + z_norm -= omni_lights.data[idx].inv_radius * omni_lights.data[idx].shadow_bias; + + shadow = 0.0; + for (uint i = 0; i < sc_penumbra_shadow_samples; i++) { + vec2 disk = disk_rotation * scene_data_block.data.penumbra_shadow_kernel[i].xy; + vec3 pos = local_vert + tangent * disk.x + bitangent * disk.y; + + pos = normalize(pos); + pos = normalize(pos + normal_bias); + + vec4 uv_rect = base_uv_rect; + + if (pos.z >= 0.0) { + uv_rect.xy += flip_offset; + } + + pos.z = 1.0 + abs(pos.z); + pos.xy /= pos.z; + + pos.xy = pos.xy * 0.5 + 0.5; + pos.xy = uv_rect.xy + pos.xy * uv_rect.zw; + shadow += textureProj(sampler2DShadow(shadow_atlas, shadow_sampler), vec4(pos.xy, z_norm, 1.0)); + } + + shadow /= float(sc_penumbra_shadow_samples); + shadow = mix(1.0, shadow, omni_lights.data[idx].shadow_opacity); + + } else { + //no blockers found, so no shadow + shadow = 1.0; + } + } else { + vec4 uv_rect = base_uv_rect; + + vec3 shadow_sample = normalize(shadow_dir + normal_bias); + if (shadow_sample.z >= 0.0) { + uv_rect.xy += flip_offset; + flip_offset *= -1.0; + } + + shadow_sample.z = 1.0 + abs(shadow_sample.z); + vec2 pos = shadow_sample.xy / shadow_sample.z; + float depth = shadow_len - omni_lights.data[idx].shadow_bias; + depth *= omni_lights.data[idx].inv_radius; + shadow = mix(1.0, sample_omni_pcf_shadow(shadow_atlas, omni_lights.data[idx].soft_shadow_scale / shadow_sample.z, pos, uv_rect, flip_offset, depth), omni_lights.data[idx].shadow_opacity); + } + + return shadow; + } +#endif + + return 1.0; +} + +void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 vertex_ddx, vec3 vertex_ddy, vec3 f0, uint orms, float shadow, vec3 albedo, inout float alpha, +#ifdef LIGHT_BACKLIGHT_USED + vec3 backlight, +#endif +#ifdef LIGHT_TRANSMITTANCE_USED + vec4 transmittance_color, + float transmittance_depth, + float transmittance_boost, +#endif +#ifdef LIGHT_RIM_USED + float rim, float rim_tint, +#endif +#ifdef LIGHT_CLEARCOAT_USED + float clearcoat, float clearcoat_roughness, vec3 vertex_normal, +#endif +#ifdef LIGHT_ANISOTROPY_USED + vec3 binormal, vec3 tangent, float anisotropy, +#endif + inout vec3 diffuse_light, inout vec3 specular_light) { + vec3 light_rel_vec = omni_lights.data[idx].position - vertex; + float light_length = length(light_rel_vec); + float omni_attenuation = get_omni_attenuation(light_length, omni_lights.data[idx].inv_radius, omni_lights.data[idx].attenuation); + float light_attenuation = omni_attenuation; + vec3 color = omni_lights.data[idx].color; + + float size_A = 0.0; + + if (sc_use_light_soft_shadows && omni_lights.data[idx].size > 0.0) { + float t = omni_lights.data[idx].size / max(0.001, light_length); + size_A = max(0.0, 1.0 - 1 / sqrt(1 + t * t)); + } + +#ifdef LIGHT_TRANSMITTANCE_USED + float transmittance_z = transmittance_depth; //no transmittance by default + transmittance_color.a *= light_attenuation; + { + vec4 clamp_rect = omni_lights.data[idx].atlas_rect; + + //redo shadowmapping, but shrink the model a bit to avoid arctifacts + vec4 splane = (omni_lights.data[idx].shadow_matrix * vec4(vertex - normalize(normal_interp) * omni_lights.data[idx].transmittance_bias, 1.0)); + + float shadow_len = length(splane.xyz); + splane.xyz = normalize(splane.xyz); + + if (splane.z >= 0.0) { + splane.z += 1.0; + clamp_rect.y += clamp_rect.w; + } else { + splane.z = 1.0 - splane.z; + } + + splane.xy /= splane.z; + + splane.xy = splane.xy * 0.5 + 0.5; + splane.z = shadow_len * omni_lights.data[idx].inv_radius; + splane.xy = clamp_rect.xy + splane.xy * clamp_rect.zw; + // splane.xy = clamp(splane.xy,clamp_rect.xy + scene_data_block.data.shadow_atlas_pixel_size,clamp_rect.xy + clamp_rect.zw - scene_data_block.data.shadow_atlas_pixel_size ); + splane.w = 1.0; //needed? i think it should be 1 already + + float shadow_z = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), splane.xy, 0.0).r; + transmittance_z = (splane.z - shadow_z) / omni_lights.data[idx].inv_radius; + } +#endif + + if (sc_use_light_projector && omni_lights.data[idx].projector_rect != vec4(0.0)) { + vec3 local_v = (omni_lights.data[idx].shadow_matrix * vec4(vertex, 1.0)).xyz; + local_v = normalize(local_v); + + vec4 atlas_rect = omni_lights.data[idx].projector_rect; + + if (local_v.z >= 0.0) { + atlas_rect.y += atlas_rect.w; + } + + local_v.z = 1.0 + abs(local_v.z); + + local_v.xy /= local_v.z; + local_v.xy = local_v.xy * 0.5 + 0.5; + vec2 proj_uv = local_v.xy * atlas_rect.zw; + + if (sc_projector_use_mipmaps) { + vec2 proj_uv_ddx; + vec2 proj_uv_ddy; + { + vec3 local_v_ddx = (omni_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddx, 1.0)).xyz; + local_v_ddx = normalize(local_v_ddx); + + if (local_v_ddx.z >= 0.0) { + local_v_ddx.z += 1.0; + } else { + local_v_ddx.z = 1.0 - local_v_ddx.z; + } + + local_v_ddx.xy /= local_v_ddx.z; + local_v_ddx.xy = local_v_ddx.xy * 0.5 + 0.5; + + proj_uv_ddx = local_v_ddx.xy * atlas_rect.zw - proj_uv; + + vec3 local_v_ddy = (omni_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddy, 1.0)).xyz; + local_v_ddy = normalize(local_v_ddy); + + if (local_v_ddy.z >= 0.0) { + local_v_ddy.z += 1.0; + } else { + local_v_ddy.z = 1.0 - local_v_ddy.z; + } + + local_v_ddy.xy /= local_v_ddy.z; + local_v_ddy.xy = local_v_ddy.xy * 0.5 + 0.5; + + proj_uv_ddy = local_v_ddy.xy * atlas_rect.zw - proj_uv; + } + + vec4 proj = textureGrad(sampler2D(decal_atlas_srgb, light_projector_sampler), proj_uv + atlas_rect.xy, proj_uv_ddx, proj_uv_ddy); + color *= proj.rgb * proj.a; + } else { + vec4 proj = textureLod(sampler2D(decal_atlas_srgb, light_projector_sampler), proj_uv + atlas_rect.xy, 0.0); + color *= proj.rgb * proj.a; + } + } + + light_attenuation *= shadow; + + light_compute(normal, normalize(light_rel_vec), eye_vec, size_A, color, light_attenuation, f0, orms, omni_lights.data[idx].specular_amount, albedo, alpha, +#ifdef LIGHT_BACKLIGHT_USED + backlight, +#endif +#ifdef LIGHT_TRANSMITTANCE_USED + transmittance_color, + transmittance_depth, + transmittance_boost, + transmittance_z, +#endif +#ifdef LIGHT_RIM_USED + rim * omni_attenuation, rim_tint, +#endif +#ifdef LIGHT_CLEARCOAT_USED + clearcoat, clearcoat_roughness, vertex_normal, +#endif +#ifdef LIGHT_ANISOTROPY_USED + binormal, tangent, anisotropy, +#endif + diffuse_light, + specular_light); +} + +float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { +#ifndef SHADOWS_DISABLED + if (spot_lights.data[idx].shadow_opacity > 0.001) { + vec3 light_rel_vec = spot_lights.data[idx].position - vertex; + float light_length = length(light_rel_vec); + vec3 spot_dir = spot_lights.data[idx].direction; + + vec3 shadow_dir = light_rel_vec / light_length; + vec3 normal_bias = normal * light_length * spot_lights.data[idx].shadow_normal_bias * (1.0 - abs(dot(normal, shadow_dir))); + + //there is a shadowmap + vec4 v = vec4(vertex + normal_bias, 1.0); + + vec4 splane = (spot_lights.data[idx].shadow_matrix * v); + splane.z -= spot_lights.data[idx].shadow_bias / (light_length * spot_lights.data[idx].inv_radius); + splane /= splane.w; + + float shadow; + if (sc_use_light_soft_shadows && spot_lights.data[idx].soft_shadow_size > 0.0) { + //soft shadow + + //find blocker + float z_norm = dot(spot_dir, -light_rel_vec) * spot_lights.data[idx].inv_radius; + + vec2 shadow_uv = splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy; + + float blocker_count = 0.0; + float blocker_average = 0.0; + + mat2 disk_rotation; + { + float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; + float sr = sin(r); + float cr = cos(r); + disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); + } + + float uv_size = spot_lights.data[idx].soft_shadow_size * z_norm * spot_lights.data[idx].soft_shadow_scale; + vec2 clamp_max = spot_lights.data[idx].atlas_rect.xy + spot_lights.data[idx].atlas_rect.zw; + for (uint i = 0; i < sc_penumbra_shadow_samples; i++) { + vec2 suv = shadow_uv + (disk_rotation * scene_data_block.data.penumbra_shadow_kernel[i].xy) * uv_size; + suv = clamp(suv, spot_lights.data[idx].atlas_rect.xy, clamp_max); + float d = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), suv, 0.0).r; + if (d < splane.z) { + blocker_average += d; + blocker_count += 1.0; + } + } + + if (blocker_count > 0.0) { + //blockers found, do soft shadow + blocker_average /= blocker_count; + float penumbra = (z_norm - blocker_average) / blocker_average; + uv_size *= penumbra; + + shadow = 0.0; + for (uint i = 0; i < sc_penumbra_shadow_samples; i++) { + vec2 suv = shadow_uv + (disk_rotation * scene_data_block.data.penumbra_shadow_kernel[i].xy) * uv_size; + suv = clamp(suv, spot_lights.data[idx].atlas_rect.xy, clamp_max); + shadow += textureProj(sampler2DShadow(shadow_atlas, shadow_sampler), vec4(suv, splane.z, 1.0)); + } + + shadow /= float(sc_penumbra_shadow_samples); + shadow = mix(1.0, shadow, spot_lights.data[idx].shadow_opacity); + + } else { + //no blockers found, so no shadow + shadow = 1.0; + } + } else { + //hard shadow + vec3 shadow_uv = vec3(splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy, splane.z); + shadow = mix(1.0, sample_pcf_shadow(shadow_atlas, spot_lights.data[idx].soft_shadow_scale * scene_data_block.data.shadow_atlas_pixel_size, shadow_uv), spot_lights.data[idx].shadow_opacity); + } + + return shadow; + } + +#endif // SHADOWS_DISABLED + + return 1.0; +} + +vec2 normal_to_panorama(vec3 n) { + n = normalize(n); + vec2 panorama_coords = vec2(atan(n.x, n.z), acos(-n.y)); + + if (panorama_coords.x < 0.0) { + panorama_coords.x += M_PI * 2.0; + } + + panorama_coords /= vec2(M_PI * 2.0, M_PI); + return panorama_coords; +} + +void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 vertex_ddx, vec3 vertex_ddy, vec3 f0, uint orms, float shadow, vec3 albedo, inout float alpha, +#ifdef LIGHT_BACKLIGHT_USED + vec3 backlight, +#endif +#ifdef LIGHT_TRANSMITTANCE_USED + vec4 transmittance_color, + float transmittance_depth, + float transmittance_boost, +#endif +#ifdef LIGHT_RIM_USED + float rim, float rim_tint, +#endif +#ifdef LIGHT_CLEARCOAT_USED + float clearcoat, float clearcoat_roughness, vec3 vertex_normal, +#endif +#ifdef LIGHT_ANISOTROPY_USED + vec3 binormal, vec3 tangent, float anisotropy, +#endif + inout vec3 diffuse_light, + inout vec3 specular_light) { + vec3 light_rel_vec = spot_lights.data[idx].position - vertex; + float light_length = length(light_rel_vec); + float spot_attenuation = get_omni_attenuation(light_length, spot_lights.data[idx].inv_radius, spot_lights.data[idx].attenuation); + vec3 spot_dir = spot_lights.data[idx].direction; + float scos = max(dot(-normalize(light_rel_vec), spot_dir), spot_lights.data[idx].cone_angle); + float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - spot_lights.data[idx].cone_angle)); + spot_attenuation *= 1.0 - pow(spot_rim, spot_lights.data[idx].cone_attenuation); + float light_attenuation = spot_attenuation; + vec3 color = spot_lights.data[idx].color; + float specular_amount = spot_lights.data[idx].specular_amount; + + float size_A = 0.0; + + if (sc_use_light_soft_shadows && spot_lights.data[idx].size > 0.0) { + float t = spot_lights.data[idx].size / max(0.001, light_length); + size_A = max(0.0, 1.0 - 1 / sqrt(1 + t * t)); + } + +#ifdef LIGHT_TRANSMITTANCE_USED + float transmittance_z = transmittance_depth; + transmittance_color.a *= light_attenuation; + { + vec4 splane = (spot_lights.data[idx].shadow_matrix * vec4(vertex - normalize(normal_interp) * spot_lights.data[idx].transmittance_bias, 1.0)); + splane /= splane.w; + splane.xy = splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy; + + float shadow_z = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), splane.xy, 0.0).r; + + shadow_z = shadow_z * 2.0 - 1.0; + float z_far = 1.0 / spot_lights.data[idx].inv_radius; + float z_near = 0.01; + shadow_z = 2.0 * z_near * z_far / (z_far + z_near - shadow_z * (z_far - z_near)); + + //distance to light plane + float z = dot(spot_dir, -light_rel_vec); + transmittance_z = z - shadow_z; + } +#endif //LIGHT_TRANSMITTANCE_USED + + if (sc_use_light_projector && spot_lights.data[idx].projector_rect != vec4(0.0)) { + vec4 splane = (spot_lights.data[idx].shadow_matrix * vec4(vertex, 1.0)); + splane /= splane.w; + + vec2 proj_uv = splane.xy * spot_lights.data[idx].projector_rect.zw; + + if (sc_projector_use_mipmaps) { + //ensure we have proper mipmaps + vec4 splane_ddx = (spot_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddx, 1.0)); + splane_ddx /= splane_ddx.w; + vec2 proj_uv_ddx = splane_ddx.xy * spot_lights.data[idx].projector_rect.zw - proj_uv; + + vec4 splane_ddy = (spot_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddy, 1.0)); + splane_ddy /= splane_ddy.w; + vec2 proj_uv_ddy = splane_ddy.xy * spot_lights.data[idx].projector_rect.zw - proj_uv; + + vec4 proj = textureGrad(sampler2D(decal_atlas_srgb, light_projector_sampler), proj_uv + spot_lights.data[idx].projector_rect.xy, proj_uv_ddx, proj_uv_ddy); + color *= proj.rgb * proj.a; + } else { + vec4 proj = textureLod(sampler2D(decal_atlas_srgb, light_projector_sampler), proj_uv + spot_lights.data[idx].projector_rect.xy, 0.0); + color *= proj.rgb * proj.a; + } + } + light_attenuation *= shadow; + + light_compute(normal, normalize(light_rel_vec), eye_vec, size_A, color, light_attenuation, f0, orms, spot_lights.data[idx].specular_amount, albedo, alpha, +#ifdef LIGHT_BACKLIGHT_USED + backlight, +#endif +#ifdef LIGHT_TRANSMITTANCE_USED + transmittance_color, + transmittance_depth, + transmittance_boost, + transmittance_z, +#endif +#ifdef LIGHT_RIM_USED + rim * spot_attenuation, rim_tint, +#endif +#ifdef LIGHT_CLEARCOAT_USED + clearcoat, clearcoat_roughness, vertex_normal, +#endif +#ifdef LIGHT_ANISOTROPY_USED + binormal, tangent, anisotropy, +#endif + diffuse_light, specular_light); +} + +void reflection_process(uint ref_index, vec3 vertex, vec3 ref_vec, vec3 normal, float roughness, vec3 ambient_light, vec3 specular_light, inout vec4 ambient_accum, inout vec4 reflection_accum) { + vec3 box_extents = reflections.data[ref_index].box_extents; + vec3 local_pos = (reflections.data[ref_index].local_matrix * vec4(vertex, 1.0)).xyz; + + if (any(greaterThan(abs(local_pos), box_extents))) { //out of the reflection box + return; + } + + vec3 inner_pos = abs(local_pos / box_extents); + float blend = max(inner_pos.x, max(inner_pos.y, inner_pos.z)); + //make blend more rounded + blend = mix(length(inner_pos), blend, blend); + blend *= blend; + blend = max(0.0, 1.0 - blend); + + if (reflections.data[ref_index].intensity > 0.0) { // compute reflection + + vec3 local_ref_vec = (reflections.data[ref_index].local_matrix * vec4(ref_vec, 0.0)).xyz; + + if (reflections.data[ref_index].box_project) { //box project + + vec3 nrdir = normalize(local_ref_vec); + vec3 rbmax = (box_extents - local_pos) / nrdir; + vec3 rbmin = (-box_extents - local_pos) / nrdir; + + vec3 rbminmax = mix(rbmin, rbmax, greaterThan(nrdir, vec3(0.0, 0.0, 0.0))); + + float fa = min(min(rbminmax.x, rbminmax.y), rbminmax.z); + vec3 posonbox = local_pos + nrdir * fa; + local_ref_vec = posonbox - reflections.data[ref_index].box_offset; + } + + vec4 reflection; + + reflection.rgb = textureLod(samplerCubeArray(reflection_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(local_ref_vec, reflections.data[ref_index].index), roughness * MAX_ROUGHNESS_LOD).rgb * sc_luminance_multiplier; + reflection.rgb *= reflections.data[ref_index].exposure_normalization; + if (reflections.data[ref_index].exterior) { + reflection.rgb = mix(specular_light, reflection.rgb, blend); + } + + reflection.rgb *= reflections.data[ref_index].intensity; //intensity + reflection.a = blend; + reflection.rgb *= reflection.a; + + reflection_accum += reflection; + } + + switch (reflections.data[ref_index].ambient_mode) { + case REFLECTION_AMBIENT_DISABLED: { + //do nothing + } break; + case REFLECTION_AMBIENT_ENVIRONMENT: { + //do nothing + vec3 local_amb_vec = (reflections.data[ref_index].local_matrix * vec4(normal, 0.0)).xyz; + + vec4 ambient_out; + + ambient_out.rgb = textureLod(samplerCubeArray(reflection_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(local_amb_vec, reflections.data[ref_index].index), MAX_ROUGHNESS_LOD).rgb; + ambient_out.rgb *= reflections.data[ref_index].exposure_normalization; + ambient_out.a = blend; + if (reflections.data[ref_index].exterior) { + ambient_out.rgb = mix(ambient_light, ambient_out.rgb, blend); + } + + ambient_out.rgb *= ambient_out.a; + ambient_accum += ambient_out; + } break; + case REFLECTION_AMBIENT_COLOR: { + vec4 ambient_out; + ambient_out.a = blend; + ambient_out.rgb = reflections.data[ref_index].ambient; + if (reflections.data[ref_index].exterior) { + ambient_out.rgb = mix(ambient_light, ambient_out.rgb, blend); + } + ambient_out.rgb *= ambient_out.a; + ambient_accum += ambient_out; + } break; + } +} + +float blur_shadow(float shadow) { + return shadow; +#if 0 + //disabling for now, will investigate later + float interp_shadow = shadow; + if (gl_HelperInvocation) { + interp_shadow = -4.0; // technically anything below -4 will do but just to make sure + } + + uvec2 fc2 = uvec2(gl_FragCoord.xy); + interp_shadow -= dFdx(interp_shadow) * (float(fc2.x & 1) - 0.5); + interp_shadow -= dFdy(interp_shadow) * (float(fc2.y & 1) - 0.5); + + if (interp_shadow >= 0.0) { + shadow = interp_shadow; + } + return shadow; +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/skeleton.glsl b/servers/rendering/renderer_rd/shaders/skeleton.glsl new file mode 100644 index 0000000000..75bea9300b --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/skeleton.glsl @@ -0,0 +1,286 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in; + +layout(set = 0, binding = 1, std430) buffer restrict writeonly DstVertexData { + uint data[]; +} +dst_vertices; + +layout(set = 0, binding = 2, std430) buffer restrict readonly BlendShapeWeights { + float data[]; +} +blend_shape_weights; + +layout(set = 1, binding = 0, std430) buffer restrict readonly SrcVertexData { + uint data[]; +} +src_vertices; + +layout(set = 1, binding = 1, std430) buffer restrict readonly BoneWeightData { + uint data[]; +} +src_bone_weights; + +layout(set = 1, binding = 2, std430) buffer restrict readonly BlendShapeData { + uint data[]; +} +src_blend_shapes; + +layout(set = 2, binding = 0, std430) buffer restrict readonly SkeletonData { + vec4 data[]; +} +bone_transforms; + +layout(push_constant, std430) uniform Params { + bool has_normal; + bool has_tangent; + bool has_skeleton; + bool has_blend_shape; + + uint vertex_count; + uint vertex_stride; + uint skin_stride; + uint skin_weight_offset; + + uint blend_shape_count; + bool normalized_blend_shapes; + uint pad0; + uint pad1; +} +params; + +vec2 uint_to_vec2(uint base) { + uvec2 decode = (uvec2(base) >> uvec2(0, 16)) & uvec2(0xFFFF, 0xFFFF); + return vec2(decode) / vec2(65535.0, 65535.0) * 2.0 - 1.0; +} + +vec3 oct_to_vec3(vec2 oct) { + vec3 v = vec3(oct.xy, 1.0 - abs(oct.x) - abs(oct.y)); + float t = max(-v.z, 0.0); + v.xy += t * -sign(v.xy); + return v; +} + +vec3 decode_uint_oct_to_norm(uint base) { + return oct_to_vec3(uint_to_vec2(base)); +} + +vec4 decode_uint_oct_to_tang(uint base) { + vec2 oct_sign_encoded = uint_to_vec2(base); + // Binormal sign encoded in y component + vec2 oct = vec2(oct_sign_encoded.x, abs(oct_sign_encoded.y) * 2.0 - 1.0); + return vec4(oct_to_vec3(oct), sign(oct_sign_encoded.y)); +} + +vec2 signNotZero(vec2 v) { + return mix(vec2(-1.0), vec2(1.0), greaterThanEqual(v.xy, vec2(0.0))); +} + +uint vec2_to_uint(vec2 base) { + uvec2 enc = uvec2(clamp(ivec2(base * vec2(65535, 65535)), ivec2(0), ivec2(0xFFFF, 0xFFFF))) << uvec2(0, 16); + return enc.x | enc.y; +} + +vec2 vec3_to_oct(vec3 e) { + e /= abs(e.x) + abs(e.y) + abs(e.z); + vec2 oct = e.z >= 0.0f ? e.xy : (vec2(1.0f) - abs(e.yx)) * signNotZero(e.xy); + return oct * 0.5f + 0.5f; +} + +uint encode_norm_to_uint_oct(vec3 base) { + return vec2_to_uint(vec3_to_oct(base)); +} + +uint encode_tang_to_uint_oct(vec4 base) { + vec2 oct = vec3_to_oct(base.xyz); + // Encode binormal sign in y component + oct.y = oct.y * 0.5f + 0.5f; + oct.y = base.w >= 0.0f ? oct.y : 1 - oct.y; + return vec2_to_uint(oct); +} + +void main() { + uint index = gl_GlobalInvocationID.x; + if (index >= params.vertex_count) { + return; + } + + uint src_offset = index * params.vertex_stride; + +#ifdef MODE_2D + vec2 vertex = uintBitsToFloat(uvec2(src_vertices.data[src_offset + 0], src_vertices.data[src_offset + 1])); + + if (params.has_blend_shape) { + float blend_total = 0.0; + vec2 blend_vertex = vec2(0.0); + + for (uint i = 0; i < params.blend_shape_count; i++) { + float w = blend_shape_weights.data[i]; + if (abs(w) > 0.0001) { + uint base_offset = (params.vertex_count * i + index) * params.vertex_stride; + + blend_vertex += uintBitsToFloat(uvec2(src_blend_shapes.data[base_offset + 0], src_blend_shapes.data[base_offset + 1])) * w; + + base_offset += 2; + + blend_total += w; + } + } + + if (params.normalized_blend_shapes) { + vertex = (1.0 - blend_total) * vertex; + } + + vertex += blend_vertex; + } + + if (params.has_skeleton) { + uint skin_offset = params.skin_stride * index; + + uvec2 bones = uvec2(src_bone_weights.data[skin_offset + 0], src_bone_weights.data[skin_offset + 1]); + uvec2 bones_01 = uvec2(bones.x & 0xFFFF, bones.x >> 16) * 3; //pre-add xform offset + uvec2 bones_23 = uvec2(bones.y & 0xFFFF, bones.y >> 16) * 3; + + skin_offset += params.skin_weight_offset; + + uvec2 weights = uvec2(src_bone_weights.data[skin_offset + 0], src_bone_weights.data[skin_offset + 1]); + + vec2 weights_01 = unpackUnorm2x16(weights.x); + vec2 weights_23 = unpackUnorm2x16(weights.y); + + mat4 m = mat4(bone_transforms.data[bones_01.x], bone_transforms.data[bones_01.x + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0)) * weights_01.x; + m += mat4(bone_transforms.data[bones_01.y], bone_transforms.data[bones_01.y + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0)) * weights_01.y; + m += mat4(bone_transforms.data[bones_23.x], bone_transforms.data[bones_23.x + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0)) * weights_23.x; + m += mat4(bone_transforms.data[bones_23.y], bone_transforms.data[bones_23.y + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0)) * weights_23.y; + + //reverse order because its transposed + vertex = (vec4(vertex, 0.0, 1.0) * m).xy; + } +#else + vec3 vertex; + vec3 normal; + vec4 tangent; + + vertex = uintBitsToFloat(uvec3(src_vertices.data[src_offset + 0], src_vertices.data[src_offset + 1], src_vertices.data[src_offset + 2])); + + src_offset += 3; + + if (params.has_normal) { + normal = decode_uint_oct_to_norm(src_vertices.data[src_offset]); + src_offset++; + } + + if (params.has_tangent) { + tangent = decode_uint_oct_to_tang(src_vertices.data[src_offset]); + } + + if (params.has_blend_shape) { + float blend_total = 0.0; + vec3 blend_vertex = vec3(0.0); + vec3 blend_normal = vec3(0.0); + vec3 blend_tangent = vec3(0.0); + + for (uint i = 0; i < params.blend_shape_count; i++) { + float w = blend_shape_weights.data[i]; + if (abs(w) > 0.0001) { + uint base_offset = (params.vertex_count * i + index) * params.vertex_stride; + + blend_vertex += uintBitsToFloat(uvec3(src_blend_shapes.data[base_offset + 0], src_blend_shapes.data[base_offset + 1], src_blend_shapes.data[base_offset + 2])) * w; + + base_offset += 3; + + if (params.has_normal) { + blend_normal += decode_uint_oct_to_norm(src_blend_shapes.data[base_offset]) * w; + base_offset++; + } + + if (params.has_tangent) { + blend_tangent += decode_uint_oct_to_tang(src_blend_shapes.data[base_offset]).rgb * w; + } + + blend_total += w; + } + } + + if (params.normalized_blend_shapes) { + vertex = (1.0 - blend_total) * vertex; + normal = (1.0 - blend_total) * normal; + tangent.rgb = (1.0 - blend_total) * tangent.rgb; + } + + vertex += blend_vertex; + normal = normalize(normal + blend_normal); + tangent.rgb = normalize(tangent.rgb + blend_tangent); + } + + if (params.has_skeleton) { + uint skin_offset = params.skin_stride * index; + + uvec2 bones = uvec2(src_bone_weights.data[skin_offset + 0], src_bone_weights.data[skin_offset + 1]); + uvec2 bones_01 = uvec2(bones.x & 0xFFFF, bones.x >> 16) * 3; //pre-add xform offset + uvec2 bones_23 = uvec2(bones.y & 0xFFFF, bones.y >> 16) * 3; + + skin_offset += params.skin_weight_offset; + + uvec2 weights = uvec2(src_bone_weights.data[skin_offset + 0], src_bone_weights.data[skin_offset + 1]); + + vec2 weights_01 = unpackUnorm2x16(weights.x); + vec2 weights_23 = unpackUnorm2x16(weights.y); + + mat4 m = mat4(bone_transforms.data[bones_01.x], bone_transforms.data[bones_01.x + 1], bone_transforms.data[bones_01.x + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weights_01.x; + m += mat4(bone_transforms.data[bones_01.y], bone_transforms.data[bones_01.y + 1], bone_transforms.data[bones_01.y + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weights_01.y; + m += mat4(bone_transforms.data[bones_23.x], bone_transforms.data[bones_23.x + 1], bone_transforms.data[bones_23.x + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weights_23.x; + m += mat4(bone_transforms.data[bones_23.y], bone_transforms.data[bones_23.y + 1], bone_transforms.data[bones_23.y + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weights_23.y; + + if (params.skin_weight_offset == 4) { + //using 8 bones/weights + skin_offset = params.skin_stride * index + 2; + + bones = uvec2(src_bone_weights.data[skin_offset + 0], src_bone_weights.data[skin_offset + 1]); + bones_01 = uvec2(bones.x & 0xFFFF, bones.x >> 16) * 3; //pre-add xform offset + bones_23 = uvec2(bones.y & 0xFFFF, bones.y >> 16) * 3; + + skin_offset += params.skin_weight_offset; + + weights = uvec2(src_bone_weights.data[skin_offset + 0], src_bone_weights.data[skin_offset + 1]); + + weights_01 = unpackUnorm2x16(weights.x); + weights_23 = unpackUnorm2x16(weights.y); + + m += mat4(bone_transforms.data[bones_01.x], bone_transforms.data[bones_01.x + 1], bone_transforms.data[bones_01.x + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weights_01.x; + m += mat4(bone_transforms.data[bones_01.y], bone_transforms.data[bones_01.y + 1], bone_transforms.data[bones_01.y + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weights_01.y; + m += mat4(bone_transforms.data[bones_23.x], bone_transforms.data[bones_23.x + 1], bone_transforms.data[bones_23.x + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weights_23.x; + m += mat4(bone_transforms.data[bones_23.y], bone_transforms.data[bones_23.y + 1], bone_transforms.data[bones_23.y + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weights_23.y; + } + + //reverse order because its transposed + vertex = (vec4(vertex, 1.0) * m).xyz; + normal = normalize((vec4(normal, 0.0) * m).xyz); + tangent.xyz = normalize((vec4(tangent.xyz, 0.0) * m).xyz); + } + + uint dst_offset = index * params.vertex_stride; + + uvec3 uvertex = floatBitsToUint(vertex); + dst_vertices.data[dst_offset + 0] = uvertex.x; + dst_vertices.data[dst_offset + 1] = uvertex.y; + dst_vertices.data[dst_offset + 2] = uvertex.z; + + dst_offset += 3; + + if (params.has_normal) { + dst_vertices.data[dst_offset] = encode_norm_to_uint_oct(normal); + dst_offset++; + } + + if (params.has_tangent) { + dst_vertices.data[dst_offset] = encode_tang_to_uint_oct(tangent); + } + +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/sort.glsl b/servers/rendering/renderer_rd/shaders/sort.glsl new file mode 100644 index 0000000000..48cf69012a --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/sort.glsl @@ -0,0 +1,203 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +// Original version here: +// https://github.com/GPUOpen-LibrariesAndSDKs/GPUParticles11/blob/master/gpuparticles11/src/Shaders + +// +// Copyright (c) 2016 Advanced Micro Devices, Inc. All rights reserved. +// +// Permission is hereby granted, free of charge, to any person obtaining a copy +// of this software and associated documentation files (the "Software"), to deal +// in the Software without restriction, including without limitation the rights +// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell +// copies of the Software, and to permit persons to whom the Software is +// furnished to do so, subject to the following conditions: +// +// The above copyright notice and this permission notice shall be included in +// all copies or substantial portions of the Software. +// +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN +// THE SOFTWARE. +// + +#define SORT_SIZE 512 +#define NUM_THREADS (SORT_SIZE / 2) +#define INVERSION (16 * 2 + 8 * 3) +#define ITERATIONS 1 + +layout(local_size_x = NUM_THREADS, local_size_y = 1, local_size_z = 1) in; + +#ifndef MODE_SORT_STEP + +shared vec2 g_LDS[SORT_SIZE]; + +#endif + +layout(set = 1, binding = 0, std430) restrict buffer SortBuffer { + vec2 data[]; +} +sort_buffer; + +layout(push_constant, std430) uniform Params { + uint total_elements; + uint pad[3]; + ivec4 job_params; +} +params; + +void main() { +#ifdef MODE_SORT_BLOCK + + uvec3 Gid = gl_WorkGroupID; + uvec3 DTid = gl_GlobalInvocationID; + uvec3 GTid = gl_LocalInvocationID; + uint GI = gl_LocalInvocationIndex; + + int GlobalBaseIndex = int((Gid.x * SORT_SIZE) + GTid.x); + int LocalBaseIndex = int(GI); + int numElementsInThreadGroup = int(min(SORT_SIZE, params.total_elements - (Gid.x * SORT_SIZE))); + + // Load shared data + + int i; + for (i = 0; i < 2 * ITERATIONS; ++i) { + if (GI + i * NUM_THREADS < numElementsInThreadGroup) + g_LDS[LocalBaseIndex + i * NUM_THREADS] = sort_buffer.data[GlobalBaseIndex + i * NUM_THREADS]; + } + + groupMemoryBarrier(); + barrier(); + + // Bitonic sort + for (int nMergeSize = 2; nMergeSize <= SORT_SIZE; nMergeSize = nMergeSize * 2) { + for (int nMergeSubSize = nMergeSize >> 1; nMergeSubSize > 0; nMergeSubSize = nMergeSubSize >> 1) { + for (i = 0; i < ITERATIONS; ++i) { + int tmp_index = int(GI + NUM_THREADS * i); + int index_low = tmp_index & (nMergeSubSize - 1); + int index_high = 2 * (tmp_index - index_low); + int index = index_high + index_low; + + int nSwapElem = nMergeSubSize == nMergeSize >> 1 ? index_high + (2 * nMergeSubSize - 1) - index_low : index_high + nMergeSubSize + index_low; + if (nSwapElem < numElementsInThreadGroup) { + vec2 a = g_LDS[index]; + vec2 b = g_LDS[nSwapElem]; + + if (a.x > b.x) { + g_LDS[index] = b; + g_LDS[nSwapElem] = a; + } + } + groupMemoryBarrier(); + barrier(); + } + } + } + + // Store shared data + for (i = 0; i < 2 * ITERATIONS; ++i) { + if (GI + i * NUM_THREADS < numElementsInThreadGroup) { + sort_buffer.data[GlobalBaseIndex + i * NUM_THREADS] = g_LDS[LocalBaseIndex + i * NUM_THREADS]; + } + } + +#endif + +#ifdef MODE_SORT_STEP + + uvec3 Gid = gl_WorkGroupID; + uvec3 GTid = gl_LocalInvocationID; + + ivec4 tgp; + + tgp.x = int(Gid.x) * 256; + tgp.y = 0; + tgp.z = int(params.total_elements); + tgp.w = min(512, max(0, tgp.z - int(Gid.x) * 512)); + + uint localID = int(tgp.x) + GTid.x; // calculate threadID within this sortable-array + + uint index_low = localID & (params.job_params.x - 1); + uint index_high = 2 * (localID - index_low); + + uint index = tgp.y + index_high + index_low; + uint nSwapElem = tgp.y + index_high + params.job_params.y + params.job_params.z * index_low; + + if (nSwapElem < tgp.y + tgp.z) { + vec2 a = sort_buffer.data[index]; + vec2 b = sort_buffer.data[nSwapElem]; + + if (a.x > b.x) { + sort_buffer.data[index] = b; + sort_buffer.data[nSwapElem] = a; + } + } + +#endif + +#ifdef MODE_SORT_INNER + + uvec3 Gid = gl_WorkGroupID; + uvec3 DTid = gl_GlobalInvocationID; + uvec3 GTid = gl_LocalInvocationID; + uint GI = gl_LocalInvocationIndex; + + ivec4 tgp; + + tgp.x = int(Gid.x * 256); + tgp.y = 0; + tgp.z = int(params.total_elements.x); + tgp.w = int(min(512, max(0, params.total_elements - Gid.x * 512))); + + int GlobalBaseIndex = int(tgp.y + tgp.x * 2 + GTid.x); + int LocalBaseIndex = int(GI); + int i; + + // Load shared data + for (i = 0; i < 2; ++i) { + if (GI + i * NUM_THREADS < tgp.w) + g_LDS[LocalBaseIndex + i * NUM_THREADS] = sort_buffer.data[GlobalBaseIndex + i * NUM_THREADS]; + } + + groupMemoryBarrier(); + barrier(); + + // sort threadgroup shared memory + for (int nMergeSubSize = SORT_SIZE >> 1; nMergeSubSize > 0; nMergeSubSize = nMergeSubSize >> 1) { + int tmp_index = int(GI); + int index_low = tmp_index & (nMergeSubSize - 1); + int index_high = 2 * (tmp_index - index_low); + int index = index_high + index_low; + + int nSwapElem = index_high + nMergeSubSize + index_low; + + if (nSwapElem < tgp.w) { + vec2 a = g_LDS[index]; + vec2 b = g_LDS[nSwapElem]; + + if (a.x > b.x) { + g_LDS[index] = b; + g_LDS[nSwapElem] = a; + } + } + groupMemoryBarrier(); + barrier(); + } + + // Store shared data + for (i = 0; i < 2; ++i) { + if (GI + i * NUM_THREADS < tgp.w) { + sort_buffer.data[GlobalBaseIndex + i * NUM_THREADS] = g_LDS[LocalBaseIndex + i * NUM_THREADS]; + } + } + +#endif +} diff --git a/servers/rendering/renderer_rd/storage_rd/SCsub b/servers/rendering/renderer_rd/storage_rd/SCsub new file mode 100644 index 0000000000..86681f9c74 --- /dev/null +++ b/servers/rendering/renderer_rd/storage_rd/SCsub @@ -0,0 +1,5 @@ +#!/usr/bin/env python + +Import("env") + +env.add_source_files(env.servers_sources, "*.cpp") diff --git a/servers/rendering/renderer_rd/storage_rd/forward_id_storage.cpp b/servers/rendering/renderer_rd/storage_rd/forward_id_storage.cpp new file mode 100644 index 0000000000..c7f106eba0 --- /dev/null +++ b/servers/rendering/renderer_rd/storage_rd/forward_id_storage.cpp @@ -0,0 +1,43 @@ +/*************************************************************************/ +/* forward_id_storage.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "forward_id_storage.h" + +using namespace RendererRD; + +ForwardIDStorage *ForwardIDStorage::singleton = nullptr; + +ForwardIDStorage::ForwardIDStorage() { + singleton = this; +} + +ForwardIDStorage::~ForwardIDStorage() { + singleton = nullptr; +} diff --git a/servers/rendering/renderer_rd/storage_rd/forward_id_storage.h b/servers/rendering/renderer_rd/storage_rd/forward_id_storage.h new file mode 100644 index 0000000000..f6a74383d4 --- /dev/null +++ b/servers/rendering/renderer_rd/storage_rd/forward_id_storage.h @@ -0,0 +1,68 @@ +/*************************************************************************/ +/* forward_id_storage.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef FORWARD_ID_STORAGE_H +#define FORWARD_ID_STORAGE_H + +#include "servers/rendering/storage/utilities.h" + +class RendererSceneRenderRD; + +namespace RendererRD { + +typedef int32_t ForwardID; + +enum ForwardIDType { + FORWARD_ID_TYPE_OMNI_LIGHT, + FORWARD_ID_TYPE_SPOT_LIGHT, + FORWARD_ID_TYPE_REFLECTION_PROBE, + FORWARD_ID_TYPE_DECAL, + FORWARD_ID_MAX, +}; + +class ForwardIDStorage { +private: + static ForwardIDStorage *singleton; + +public: + static ForwardIDStorage *get_singleton() { return singleton; } + + ForwardIDStorage(); + virtual ~ForwardIDStorage(); + + virtual RendererRD::ForwardID allocate_forward_id(RendererRD::ForwardIDType p_type) { return -1; } + virtual void free_forward_id(RendererRD::ForwardIDType p_type, RendererRD::ForwardID p_id) {} + virtual void map_forward_id(RendererRD::ForwardIDType p_type, RendererRD::ForwardID p_id, uint32_t p_index) {} + virtual bool uses_forward_ids() const { return false; } +}; + +} // namespace RendererRD + +#endif // FORWARD_ID_STORAGE_H diff --git a/servers/rendering/renderer_rd/storage_rd/light_storage.cpp b/servers/rendering/renderer_rd/storage_rd/light_storage.cpp new file mode 100644 index 0000000000..c83473ef07 --- /dev/null +++ b/servers/rendering/renderer_rd/storage_rd/light_storage.cpp @@ -0,0 +1,2446 @@ +/*************************************************************************/ +/* light_storage.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "light_storage.h" +#include "core/config/project_settings.h" +#include "servers/rendering/renderer_rd/renderer_scene_render_rd.h" +#include "texture_storage.h" + +using namespace RendererRD; + +LightStorage *LightStorage::singleton = nullptr; + +LightStorage *LightStorage::get_singleton() { + return singleton; +} + +LightStorage::LightStorage() { + singleton = this; + + TextureStorage *texture_storage = TextureStorage::get_singleton(); + + directional_shadow.size = GLOBAL_GET("rendering/lights_and_shadows/directional_shadow/size"); + directional_shadow.use_16_bits = GLOBAL_GET("rendering/lights_and_shadows/directional_shadow/16_bits"); + + using_lightmap_array = true; // high end + if (using_lightmap_array) { + uint64_t textures_per_stage = RD::get_singleton()->limit_get(RD::LIMIT_MAX_TEXTURES_PER_SHADER_STAGE); + + if (textures_per_stage <= 256) { + lightmap_textures.resize(32); + } else { + lightmap_textures.resize(1024); + } + + for (int i = 0; i < lightmap_textures.size(); i++) { + lightmap_textures.write[i] = texture_storage->texture_rd_get_default(TextureStorage::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE); + } + } + + lightmap_probe_capture_update_speed = GLOBAL_GET("rendering/lightmapping/probe_capture/update_speed"); +} + +LightStorage::~LightStorage() { + free_reflection_data(); + free_light_data(); + + for (const KeyValue<int, ShadowCubemap> &E : shadow_cubemaps) { + RD::get_singleton()->free(E.value.cubemap); + } + + singleton = nullptr; +} + +bool LightStorage::free(RID p_rid) { + if (owns_reflection_probe(p_rid)) { + reflection_probe_free(p_rid); + return true; + } else if (owns_reflection_atlas(p_rid)) { + reflection_atlas_free(p_rid); + return true; + } else if (owns_reflection_probe_instance(p_rid)) { + reflection_probe_instance_free(p_rid); + return true; + } else if (owns_light(p_rid)) { + light_free(p_rid); + return true; + } else if (owns_light_instance(p_rid)) { + light_instance_free(p_rid); + return true; + } else if (owns_lightmap(p_rid)) { + lightmap_free(p_rid); + return true; + } else if (owns_lightmap_instance(p_rid)) { + lightmap_instance_free(p_rid); + return true; + } else if (owns_shadow_atlas(p_rid)) { + shadow_atlas_free(p_rid); + return true; + } + + return false; +} + +/* LIGHT */ + +void LightStorage::_light_initialize(RID p_light, RS::LightType p_type) { + Light light; + light.type = p_type; + + light.param[RS::LIGHT_PARAM_ENERGY] = 1.0; + light.param[RS::LIGHT_PARAM_INDIRECT_ENERGY] = 1.0; + light.param[RS::LIGHT_PARAM_VOLUMETRIC_FOG_ENERGY] = 1.0; + light.param[RS::LIGHT_PARAM_SPECULAR] = 0.5; + light.param[RS::LIGHT_PARAM_RANGE] = 1.0; + light.param[RS::LIGHT_PARAM_SIZE] = 0.0; + light.param[RS::LIGHT_PARAM_ATTENUATION] = 1.0; + light.param[RS::LIGHT_PARAM_SPOT_ANGLE] = 45; + light.param[RS::LIGHT_PARAM_SPOT_ATTENUATION] = 1.0; + light.param[RS::LIGHT_PARAM_SHADOW_MAX_DISTANCE] = 0; + light.param[RS::LIGHT_PARAM_SHADOW_SPLIT_1_OFFSET] = 0.1; + light.param[RS::LIGHT_PARAM_SHADOW_SPLIT_2_OFFSET] = 0.3; + light.param[RS::LIGHT_PARAM_SHADOW_SPLIT_3_OFFSET] = 0.6; + light.param[RS::LIGHT_PARAM_SHADOW_FADE_START] = 0.8; + light.param[RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS] = 1.0; + light.param[RS::LIGHT_PARAM_SHADOW_BIAS] = 0.02; + light.param[RS::LIGHT_PARAM_SHADOW_OPACITY] = 1.0; + light.param[RS::LIGHT_PARAM_SHADOW_BLUR] = 0; + light.param[RS::LIGHT_PARAM_SHADOW_PANCAKE_SIZE] = 20.0; + light.param[RS::LIGHT_PARAM_TRANSMITTANCE_BIAS] = 0.05; + light.param[RS::LIGHT_PARAM_INTENSITY] = p_type == RS::LIGHT_DIRECTIONAL ? 100000.0 : 1000.0; + + light_owner.initialize_rid(p_light, light); +} + +RID LightStorage::directional_light_allocate() { + return light_owner.allocate_rid(); +} + +void LightStorage::directional_light_initialize(RID p_light) { + _light_initialize(p_light, RS::LIGHT_DIRECTIONAL); +} + +RID LightStorage::omni_light_allocate() { + return light_owner.allocate_rid(); +} + +void LightStorage::omni_light_initialize(RID p_light) { + _light_initialize(p_light, RS::LIGHT_OMNI); +} + +RID LightStorage::spot_light_allocate() { + return light_owner.allocate_rid(); +} + +void LightStorage::spot_light_initialize(RID p_light) { + _light_initialize(p_light, RS::LIGHT_SPOT); +} + +void LightStorage::light_free(RID p_rid) { + light_set_projector(p_rid, RID()); //clear projector + + // delete the texture + Light *light = light_owner.get_or_null(p_rid); + light->dependency.deleted_notify(p_rid); + light_owner.free(p_rid); +} + +void LightStorage::light_set_color(RID p_light, const Color &p_color) { + Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND(!light); + + light->color = p_color; +} + +void LightStorage::light_set_param(RID p_light, RS::LightParam p_param, float p_value) { + Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND(!light); + ERR_FAIL_INDEX(p_param, RS::LIGHT_PARAM_MAX); + + if (light->param[p_param] == p_value) { + return; + } + + switch (p_param) { + case RS::LIGHT_PARAM_RANGE: + case RS::LIGHT_PARAM_SPOT_ANGLE: + case RS::LIGHT_PARAM_SHADOW_MAX_DISTANCE: + case RS::LIGHT_PARAM_SHADOW_SPLIT_1_OFFSET: + case RS::LIGHT_PARAM_SHADOW_SPLIT_2_OFFSET: + case RS::LIGHT_PARAM_SHADOW_SPLIT_3_OFFSET: + case RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS: + case RS::LIGHT_PARAM_SHADOW_PANCAKE_SIZE: + case RS::LIGHT_PARAM_SHADOW_BIAS: { + light->version++; + light->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_LIGHT); + } break; + case RS::LIGHT_PARAM_SIZE: { + if ((light->param[p_param] > CMP_EPSILON) != (p_value > CMP_EPSILON)) { + //changing from no size to size and the opposite + light->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_LIGHT_SOFT_SHADOW_AND_PROJECTOR); + } + } break; + default: { + } + } + + light->param[p_param] = p_value; +} + +void LightStorage::light_set_shadow(RID p_light, bool p_enabled) { + Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND(!light); + light->shadow = p_enabled; + + light->version++; + light->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_LIGHT); +} + +void LightStorage::light_set_projector(RID p_light, RID p_texture) { + TextureStorage *texture_storage = TextureStorage::get_singleton(); + Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND(!light); + + if (light->projector == p_texture) { + return; + } + + if (light->type != RS::LIGHT_DIRECTIONAL && light->projector.is_valid()) { + texture_storage->texture_remove_from_decal_atlas(light->projector, light->type == RS::LIGHT_OMNI); + } + + light->projector = p_texture; + + if (light->type != RS::LIGHT_DIRECTIONAL) { + if (light->projector.is_valid()) { + texture_storage->texture_add_to_decal_atlas(light->projector, light->type == RS::LIGHT_OMNI); + } + light->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_LIGHT_SOFT_SHADOW_AND_PROJECTOR); + } +} + +void LightStorage::light_set_negative(RID p_light, bool p_enable) { + Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND(!light); + + light->negative = p_enable; +} + +void LightStorage::light_set_cull_mask(RID p_light, uint32_t p_mask) { + Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND(!light); + + light->cull_mask = p_mask; + + light->version++; + light->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_LIGHT); +} + +void LightStorage::light_set_distance_fade(RID p_light, bool p_enabled, float p_begin, float p_shadow, float p_length) { + Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND(!light); + + light->distance_fade = p_enabled; + light->distance_fade_begin = p_begin; + light->distance_fade_shadow = p_shadow; + light->distance_fade_length = p_length; +} + +void LightStorage::light_set_reverse_cull_face_mode(RID p_light, bool p_enabled) { + Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND(!light); + + light->reverse_cull = p_enabled; + + light->version++; + light->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_LIGHT); +} + +void LightStorage::light_set_bake_mode(RID p_light, RS::LightBakeMode p_bake_mode) { + Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND(!light); + + light->bake_mode = p_bake_mode; + + light->version++; + light->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_LIGHT); +} + +void LightStorage::light_set_max_sdfgi_cascade(RID p_light, uint32_t p_cascade) { + Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND(!light); + + light->max_sdfgi_cascade = p_cascade; + + light->version++; + light->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_LIGHT); +} + +void LightStorage::light_omni_set_shadow_mode(RID p_light, RS::LightOmniShadowMode p_mode) { + Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND(!light); + + light->omni_shadow_mode = p_mode; + + light->version++; + light->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_LIGHT); +} + +RS::LightOmniShadowMode LightStorage::light_omni_get_shadow_mode(RID p_light) { + const Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND_V(!light, RS::LIGHT_OMNI_SHADOW_CUBE); + + return light->omni_shadow_mode; +} + +void LightStorage::light_directional_set_shadow_mode(RID p_light, RS::LightDirectionalShadowMode p_mode) { + Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND(!light); + + light->directional_shadow_mode = p_mode; + light->version++; + light->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_LIGHT); +} + +void LightStorage::light_directional_set_blend_splits(RID p_light, bool p_enable) { + Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND(!light); + + light->directional_blend_splits = p_enable; + light->version++; + light->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_LIGHT); +} + +bool LightStorage::light_directional_get_blend_splits(RID p_light) const { + const Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND_V(!light, false); + + return light->directional_blend_splits; +} + +void LightStorage::light_directional_set_sky_mode(RID p_light, RS::LightDirectionalSkyMode p_mode) { + Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND(!light); + + light->directional_sky_mode = p_mode; +} + +RS::LightDirectionalSkyMode LightStorage::light_directional_get_sky_mode(RID p_light) const { + const Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND_V(!light, RS::LIGHT_DIRECTIONAL_SKY_MODE_LIGHT_AND_SKY); + + return light->directional_sky_mode; +} + +RS::LightDirectionalShadowMode LightStorage::light_directional_get_shadow_mode(RID p_light) { + const Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND_V(!light, RS::LIGHT_DIRECTIONAL_SHADOW_ORTHOGONAL); + + return light->directional_shadow_mode; +} + +uint32_t LightStorage::light_get_max_sdfgi_cascade(RID p_light) { + const Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND_V(!light, 0); + + return light->max_sdfgi_cascade; +} + +RS::LightBakeMode LightStorage::light_get_bake_mode(RID p_light) { + const Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND_V(!light, RS::LIGHT_BAKE_DISABLED); + + return light->bake_mode; +} + +uint64_t LightStorage::light_get_version(RID p_light) const { + const Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND_V(!light, 0); + + return light->version; +} + +AABB LightStorage::light_get_aabb(RID p_light) const { + const Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND_V(!light, AABB()); + + switch (light->type) { + case RS::LIGHT_SPOT: { + float len = light->param[RS::LIGHT_PARAM_RANGE]; + float size = Math::tan(Math::deg_to_rad(light->param[RS::LIGHT_PARAM_SPOT_ANGLE])) * len; + return AABB(Vector3(-size, -size, -len), Vector3(size * 2, size * 2, len)); + }; + case RS::LIGHT_OMNI: { + float r = light->param[RS::LIGHT_PARAM_RANGE]; + return AABB(-Vector3(r, r, r), Vector3(r, r, r) * 2); + }; + case RS::LIGHT_DIRECTIONAL: { + return AABB(); + }; + } + + ERR_FAIL_V(AABB()); +} + +Dependency *LightStorage::light_get_dependency(RID p_light) const { + Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_NULL_V(light, nullptr); + + return &light->dependency; +} + +/* LIGHT INSTANCE API */ + +RID LightStorage::light_instance_create(RID p_light) { + RID li = light_instance_owner.make_rid(LightInstance()); + + LightInstance *light_instance = light_instance_owner.get_or_null(li); + + light_instance->self = li; + light_instance->light = p_light; + light_instance->light_type = light_get_type(p_light); + if (light_instance->light_type != RS::LIGHT_DIRECTIONAL) { + light_instance->forward_id = ForwardIDStorage::get_singleton()->allocate_forward_id(light_instance->light_type == RS::LIGHT_OMNI ? FORWARD_ID_TYPE_OMNI_LIGHT : FORWARD_ID_TYPE_SPOT_LIGHT); + } + + return li; +} + +void LightStorage::light_instance_free(RID p_light) { + LightInstance *light_instance = light_instance_owner.get_or_null(p_light); + + //remove from shadow atlases.. + for (const RID &E : light_instance->shadow_atlases) { + ShadowAtlas *shadow_atlas = shadow_atlas_owner.get_or_null(E); + ERR_CONTINUE(!shadow_atlas->shadow_owners.has(p_light)); + uint32_t key = shadow_atlas->shadow_owners[p_light]; + uint32_t q = (key >> QUADRANT_SHIFT) & 0x3; + uint32_t s = key & SHADOW_INDEX_MASK; + + shadow_atlas->quadrants[q].shadows.write[s].owner = RID(); + + if (key & OMNI_LIGHT_FLAG) { + // Omni lights use two atlas spots, make sure to clear the other as well + shadow_atlas->quadrants[q].shadows.write[s + 1].owner = RID(); + } + + shadow_atlas->shadow_owners.erase(p_light); + } + + if (light_instance->light_type != RS::LIGHT_DIRECTIONAL) { + ForwardIDStorage::get_singleton()->free_forward_id(light_instance->light_type == RS::LIGHT_OMNI ? FORWARD_ID_TYPE_OMNI_LIGHT : FORWARD_ID_TYPE_SPOT_LIGHT, light_instance->forward_id); + } + light_instance_owner.free(p_light); +} + +void LightStorage::light_instance_set_transform(RID p_light_instance, const Transform3D &p_transform) { + LightInstance *light_instance = light_instance_owner.get_or_null(p_light_instance); + ERR_FAIL_COND(!light_instance); + + light_instance->transform = p_transform; +} + +void LightStorage::light_instance_set_aabb(RID p_light_instance, const AABB &p_aabb) { + LightInstance *light_instance = light_instance_owner.get_or_null(p_light_instance); + ERR_FAIL_COND(!light_instance); + + light_instance->aabb = p_aabb; +} + +void LightStorage::light_instance_set_shadow_transform(RID p_light_instance, const Projection &p_projection, const Transform3D &p_transform, float p_far, float p_split, int p_pass, float p_shadow_texel_size, float p_bias_scale, float p_range_begin, const Vector2 &p_uv_scale) { + LightInstance *light_instance = light_instance_owner.get_or_null(p_light_instance); + ERR_FAIL_COND(!light_instance); + + ERR_FAIL_INDEX(p_pass, 6); + + light_instance->shadow_transform[p_pass].camera = p_projection; + light_instance->shadow_transform[p_pass].transform = p_transform; + light_instance->shadow_transform[p_pass].farplane = p_far; + light_instance->shadow_transform[p_pass].split = p_split; + light_instance->shadow_transform[p_pass].bias_scale = p_bias_scale; + light_instance->shadow_transform[p_pass].range_begin = p_range_begin; + light_instance->shadow_transform[p_pass].shadow_texel_size = p_shadow_texel_size; + light_instance->shadow_transform[p_pass].uv_scale = p_uv_scale; +} + +void LightStorage::light_instance_mark_visible(RID p_light_instance) { + LightInstance *light_instance = light_instance_owner.get_or_null(p_light_instance); + ERR_FAIL_COND(!light_instance); + + light_instance->last_scene_pass = RendererSceneRenderRD::get_singleton()->get_scene_pass(); +} + +/* LIGHT DATA */ + +void LightStorage::free_light_data() { + if (directional_light_buffer.is_valid()) { + RD::get_singleton()->free(directional_light_buffer); + directional_light_buffer = RID(); + } + + if (omni_light_buffer.is_valid()) { + RD::get_singleton()->free(omni_light_buffer); + omni_light_buffer = RID(); + } + + if (spot_light_buffer.is_valid()) { + RD::get_singleton()->free(spot_light_buffer); + spot_light_buffer = RID(); + } + + if (directional_lights != nullptr) { + memdelete_arr(directional_lights); + directional_lights = nullptr; + } + + if (omni_lights != nullptr) { + memdelete_arr(omni_lights); + omni_lights = nullptr; + } + + if (spot_lights != nullptr) { + memdelete_arr(spot_lights); + spot_lights = nullptr; + } + + if (omni_light_sort != nullptr) { + memdelete_arr(omni_light_sort); + omni_light_sort = nullptr; + } + + if (spot_light_sort != nullptr) { + memdelete_arr(spot_light_sort); + spot_light_sort = nullptr; + } +} + +void LightStorage::set_max_lights(const uint32_t p_max_lights) { + max_lights = p_max_lights; + + uint32_t light_buffer_size = max_lights * sizeof(LightData); + omni_lights = memnew_arr(LightData, max_lights); + omni_light_buffer = RD::get_singleton()->storage_buffer_create(light_buffer_size); + omni_light_sort = memnew_arr(LightInstanceDepthSort, max_lights); + spot_lights = memnew_arr(LightData, max_lights); + spot_light_buffer = RD::get_singleton()->storage_buffer_create(light_buffer_size); + spot_light_sort = memnew_arr(LightInstanceDepthSort, max_lights); + //defines += "\n#define MAX_LIGHT_DATA_STRUCTS " + itos(max_lights) + "\n"; + + max_directional_lights = RendererSceneRender::MAX_DIRECTIONAL_LIGHTS; + uint32_t directional_light_buffer_size = max_directional_lights * sizeof(DirectionalLightData); + directional_lights = memnew_arr(DirectionalLightData, max_directional_lights); + directional_light_buffer = RD::get_singleton()->uniform_buffer_create(directional_light_buffer_size); +} + +void LightStorage::update_light_buffers(RenderDataRD *p_render_data, const PagedArray<RID> &p_lights, const Transform3D &p_camera_transform, RID p_shadow_atlas, bool p_using_shadows, uint32_t &r_directional_light_count, uint32_t &r_positional_light_count, bool &r_directional_light_soft_shadows) { + ForwardIDStorage *forward_id_storage = ForwardIDStorage::get_singleton(); + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + + Transform3D inverse_transform = p_camera_transform.affine_inverse(); + + r_directional_light_count = 0; + r_positional_light_count = 0; + + Plane camera_plane(-p_camera_transform.basis.get_column(Vector3::AXIS_Z).normalized(), p_camera_transform.origin); + + omni_light_count = 0; + spot_light_count = 0; + + r_directional_light_soft_shadows = false; + + for (int i = 0; i < (int)p_lights.size(); i++) { + LightInstance *light_instance = light_instance_owner.get_or_null(p_lights[i]); + if (!light_instance) { + continue; + } + Light *light = light_owner.get_or_null(light_instance->light); + + ERR_CONTINUE(light == nullptr); + + switch (light->type) { + case RS::LIGHT_DIRECTIONAL: { + if (r_directional_light_count >= max_directional_lights || light->directional_sky_mode == RS::LIGHT_DIRECTIONAL_SKY_MODE_SKY_ONLY) { + continue; + } + + DirectionalLightData &light_data = directional_lights[r_directional_light_count]; + + Transform3D light_transform = light_instance->transform; + + Vector3 direction = inverse_transform.basis.xform(light_transform.basis.xform(Vector3(0, 0, 1))).normalized(); + + light_data.direction[0] = direction.x; + light_data.direction[1] = direction.y; + light_data.direction[2] = direction.z; + + float sign = light->negative ? -1 : 1; + + light_data.energy = sign * light->param[RS::LIGHT_PARAM_ENERGY]; + + if (RendererSceneRenderRD::get_singleton()->is_using_physical_light_units()) { + light_data.energy *= light->param[RS::LIGHT_PARAM_INTENSITY]; + } else { + light_data.energy *= Math_PI; + } + + if (p_render_data->camera_attributes.is_valid()) { + light_data.energy *= RSG::camera_attributes->camera_attributes_get_exposure_normalization_factor(p_render_data->camera_attributes); + } + + Color linear_col = light->color.srgb_to_linear(); + light_data.color[0] = linear_col.r; + light_data.color[1] = linear_col.g; + light_data.color[2] = linear_col.b; + + light_data.specular = light->param[RS::LIGHT_PARAM_SPECULAR]; + light_data.volumetric_fog_energy = light->param[RS::LIGHT_PARAM_VOLUMETRIC_FOG_ENERGY]; + light_data.mask = light->cull_mask; + + float size = light->param[RS::LIGHT_PARAM_SIZE]; + + light_data.size = 1.0 - Math::cos(Math::deg_to_rad(size)); //angle to cosine offset + + if (RendererSceneRenderRD::get_singleton()->get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_PSSM_SPLITS) { + WARN_PRINT_ONCE("The DirectionalLight3D PSSM splits debug draw mode is not reimplemented yet."); + } + + light_data.shadow_opacity = (p_using_shadows && light->shadow) + ? light->param[RS::LIGHT_PARAM_SHADOW_OPACITY] + : 0.0; + + float angular_diameter = light->param[RS::LIGHT_PARAM_SIZE]; + if (angular_diameter > 0.0) { + // I know tan(0) is 0, but let's not risk it with numerical precision. + // technically this will keep expanding until reaching the sun, but all we care + // is expand until we reach the radius of the near plane (there can't be more occluders than that) + angular_diameter = Math::tan(Math::deg_to_rad(angular_diameter)); + if (light->shadow && light->param[RS::LIGHT_PARAM_SHADOW_BLUR] > 0.0) { + // Only enable PCSS-like soft shadows if blurring is enabled. + // Otherwise, performance would decrease with no visual difference. + r_directional_light_soft_shadows = true; + } + } else { + angular_diameter = 0.0; + } + + if (light_data.shadow_opacity > 0.001) { + RS::LightDirectionalShadowMode smode = light->directional_shadow_mode; + + int limit = smode == RS::LIGHT_DIRECTIONAL_SHADOW_ORTHOGONAL ? 0 : (smode == RS::LIGHT_DIRECTIONAL_SHADOW_PARALLEL_2_SPLITS ? 1 : 3); + light_data.blend_splits = (smode != RS::LIGHT_DIRECTIONAL_SHADOW_ORTHOGONAL) && light->directional_blend_splits; + for (int j = 0; j < 4; j++) { + Rect2 atlas_rect = light_instance->shadow_transform[j].atlas_rect; + Projection matrix = light_instance->shadow_transform[j].camera; + float split = light_instance->shadow_transform[MIN(limit, j)].split; + + Projection bias; + bias.set_light_bias(); + Projection rectm; + rectm.set_light_atlas_rect(atlas_rect); + + Transform3D modelview = (inverse_transform * light_instance->shadow_transform[j].transform).inverse(); + + Projection shadow_mtx = rectm * bias * matrix * modelview; + light_data.shadow_split_offsets[j] = split; + float bias_scale = light_instance->shadow_transform[j].bias_scale; + light_data.shadow_bias[j] = light->param[RS::LIGHT_PARAM_SHADOW_BIAS] / 100.0 * bias_scale; + light_data.shadow_normal_bias[j] = light->param[RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS] * light_instance->shadow_transform[j].shadow_texel_size; + light_data.shadow_transmittance_bias[j] = light->param[RS::LIGHT_PARAM_TRANSMITTANCE_BIAS] * bias_scale; + light_data.shadow_z_range[j] = light_instance->shadow_transform[j].farplane; + light_data.shadow_range_begin[j] = light_instance->shadow_transform[j].range_begin; + RendererRD::MaterialStorage::store_camera(shadow_mtx, light_data.shadow_matrices[j]); + + Vector2 uv_scale = light_instance->shadow_transform[j].uv_scale; + uv_scale *= atlas_rect.size; //adapt to atlas size + switch (j) { + case 0: { + light_data.uv_scale1[0] = uv_scale.x; + light_data.uv_scale1[1] = uv_scale.y; + } break; + case 1: { + light_data.uv_scale2[0] = uv_scale.x; + light_data.uv_scale2[1] = uv_scale.y; + } break; + case 2: { + light_data.uv_scale3[0] = uv_scale.x; + light_data.uv_scale3[1] = uv_scale.y; + } break; + case 3: { + light_data.uv_scale4[0] = uv_scale.x; + light_data.uv_scale4[1] = uv_scale.y; + } break; + } + } + + float fade_start = light->param[RS::LIGHT_PARAM_SHADOW_FADE_START]; + light_data.fade_from = -light_data.shadow_split_offsets[3] * MIN(fade_start, 0.999); //using 1.0 would break smoothstep + light_data.fade_to = -light_data.shadow_split_offsets[3]; + + light_data.soft_shadow_scale = light->param[RS::LIGHT_PARAM_SHADOW_BLUR]; + light_data.softshadow_angle = angular_diameter; + light_data.bake_mode = light->bake_mode; + + if (angular_diameter <= 0.0) { + light_data.soft_shadow_scale *= RendererSceneRenderRD::get_singleton()->directional_shadow_quality_radius_get(); // Only use quality radius for PCF + } + } + + r_directional_light_count++; + } break; + case RS::LIGHT_OMNI: { + if (omni_light_count >= max_lights) { + continue; + } + + Transform3D light_transform = light_instance->transform; + const real_t distance = camera_plane.distance_to(light_transform.origin); + + if (light->distance_fade) { + const float fade_begin = light->distance_fade_begin; + const float fade_length = light->distance_fade_length; + + if (distance > fade_begin) { + if (distance > fade_begin + fade_length) { + // Out of range, don't draw this light to improve performance. + continue; + } + } + } + + omni_light_sort[omni_light_count].light_instance = light_instance; + omni_light_sort[omni_light_count].light = light; + omni_light_sort[omni_light_count].depth = distance; + omni_light_count++; + } break; + case RS::LIGHT_SPOT: { + if (spot_light_count >= max_lights) { + continue; + } + + Transform3D light_transform = light_instance->transform; + const real_t distance = camera_plane.distance_to(light_transform.origin); + + if (light->distance_fade) { + const float fade_begin = light->distance_fade_begin; + const float fade_length = light->distance_fade_length; + + if (distance > fade_begin) { + if (distance > fade_begin + fade_length) { + // Out of range, don't draw this light to improve performance. + continue; + } + } + } + + spot_light_sort[spot_light_count].light_instance = light_instance; + spot_light_sort[spot_light_count].light = light; + spot_light_sort[spot_light_count].depth = distance; + spot_light_count++; + } break; + } + + light_instance->last_pass = RSG::rasterizer->get_frame_number(); + } + + if (omni_light_count) { + SortArray<LightInstanceDepthSort> sorter; + sorter.sort(omni_light_sort, omni_light_count); + } + + if (spot_light_count) { + SortArray<LightInstanceDepthSort> sorter; + sorter.sort(spot_light_sort, spot_light_count); + } + + bool using_forward_ids = forward_id_storage->uses_forward_ids(); + + for (uint32_t i = 0; i < (omni_light_count + spot_light_count); i++) { + uint32_t index = (i < omni_light_count) ? i : i - (omni_light_count); + LightData &light_data = (i < omni_light_count) ? omni_lights[index] : spot_lights[index]; + RS::LightType type = (i < omni_light_count) ? RS::LIGHT_OMNI : RS::LIGHT_SPOT; + LightInstance *light_instance = (i < omni_light_count) ? omni_light_sort[index].light_instance : spot_light_sort[index].light_instance; + Light *light = (i < omni_light_count) ? omni_light_sort[index].light : spot_light_sort[index].light; + + if (using_forward_ids) { + forward_id_storage->map_forward_id(type == RS::LIGHT_OMNI ? RendererRD::FORWARD_ID_TYPE_OMNI_LIGHT : RendererRD::FORWARD_ID_TYPE_SPOT_LIGHT, light_instance->forward_id, index); + } + + Transform3D light_transform = light_instance->transform; + + float sign = light->negative ? -1 : 1; + Color linear_col = light->color.srgb_to_linear(); + + light_data.attenuation = light->param[RS::LIGHT_PARAM_ATTENUATION]; + + // Reuse fade begin, fade length and distance for shadow LOD determination later. + float fade_begin = 0.0; + float fade_shadow = 0.0; + float fade_length = 0.0; + real_t distance = 0.0; + + float fade = 1.0; + float shadow_opacity_fade = 1.0; + if (light->distance_fade) { + fade_begin = light->distance_fade_begin; + fade_shadow = light->distance_fade_shadow; + fade_length = light->distance_fade_length; + distance = camera_plane.distance_to(light_transform.origin); + + // Use `smoothstep()` to make opacity changes more gradual and less noticeable to the player. + if (distance > fade_begin) { + fade = Math::smoothstep(0.0f, 1.0f, 1.0f - float(distance - fade_begin) / fade_length); + } + + if (distance > fade_shadow) { + shadow_opacity_fade = Math::smoothstep(0.0f, 1.0f, 1.0f - float(distance - fade_shadow) / fade_length); + } + } + + float energy = sign * light->param[RS::LIGHT_PARAM_ENERGY] * fade; + + if (RendererSceneRenderRD::get_singleton()->is_using_physical_light_units()) { + energy *= light->param[RS::LIGHT_PARAM_INTENSITY]; + + // Convert from Luminous Power to Luminous Intensity + if (type == RS::LIGHT_OMNI) { + energy *= 1.0 / (Math_PI * 4.0); + } else { + // Spot Lights are not physically accurate, Luminous Intensity should change in relation to the cone angle. + // We make this assumption to keep them easy to control. + energy *= 1.0 / Math_PI; + } + } else { + energy *= Math_PI; + } + + if (p_render_data->camera_attributes.is_valid()) { + energy *= RSG::camera_attributes->camera_attributes_get_exposure_normalization_factor(p_render_data->camera_attributes); + } + + light_data.color[0] = linear_col.r * energy; + light_data.color[1] = linear_col.g * energy; + light_data.color[2] = linear_col.b * energy; + light_data.specular_amount = light->param[RS::LIGHT_PARAM_SPECULAR] * 2.0; + light_data.volumetric_fog_energy = light->param[RS::LIGHT_PARAM_VOLUMETRIC_FOG_ENERGY]; + light_data.bake_mode = light->bake_mode; + + float radius = MAX(0.001, light->param[RS::LIGHT_PARAM_RANGE]); + light_data.inv_radius = 1.0 / radius; + + Vector3 pos = inverse_transform.xform(light_transform.origin); + + light_data.position[0] = pos.x; + light_data.position[1] = pos.y; + light_data.position[2] = pos.z; + + Vector3 direction = inverse_transform.basis.xform(light_transform.basis.xform(Vector3(0, 0, -1))).normalized(); + + light_data.direction[0] = direction.x; + light_data.direction[1] = direction.y; + light_data.direction[2] = direction.z; + + float size = light->param[RS::LIGHT_PARAM_SIZE]; + + light_data.size = size; + + light_data.inv_spot_attenuation = 1.0f / light->param[RS::LIGHT_PARAM_SPOT_ATTENUATION]; + float spot_angle = light->param[RS::LIGHT_PARAM_SPOT_ANGLE]; + light_data.cos_spot_angle = Math::cos(Math::deg_to_rad(spot_angle)); + + light_data.mask = light->cull_mask; + + light_data.atlas_rect[0] = 0; + light_data.atlas_rect[1] = 0; + light_data.atlas_rect[2] = 0; + light_data.atlas_rect[3] = 0; + + RID projector = light->projector; + + if (projector.is_valid()) { + Rect2 rect = texture_storage->decal_atlas_get_texture_rect(projector); + + if (type == RS::LIGHT_SPOT) { + light_data.projector_rect[0] = rect.position.x; + light_data.projector_rect[1] = rect.position.y + rect.size.height; //flip because shadow is flipped + light_data.projector_rect[2] = rect.size.width; + light_data.projector_rect[3] = -rect.size.height; + } else { + light_data.projector_rect[0] = rect.position.x; + light_data.projector_rect[1] = rect.position.y; + light_data.projector_rect[2] = rect.size.width; + light_data.projector_rect[3] = rect.size.height * 0.5; //used by dp, so needs to be half + } + } else { + light_data.projector_rect[0] = 0; + light_data.projector_rect[1] = 0; + light_data.projector_rect[2] = 0; + light_data.projector_rect[3] = 0; + } + + const bool needs_shadow = + p_using_shadows && + owns_shadow_atlas(p_shadow_atlas) && + shadow_atlas_owns_light_instance(p_shadow_atlas, light_instance->self) && + light->shadow; + + bool in_shadow_range = true; + if (needs_shadow && light->distance_fade) { + if (distance > light->distance_fade_shadow + light->distance_fade_length) { + // Out of range, don't draw shadows to improve performance. + in_shadow_range = false; + } + } + + if (needs_shadow && in_shadow_range) { + // fill in the shadow information + + light_data.shadow_opacity = light->param[RS::LIGHT_PARAM_SHADOW_OPACITY] * shadow_opacity_fade; + + float shadow_texel_size = light_instance_get_shadow_texel_size(light_instance->self, p_shadow_atlas); + light_data.shadow_normal_bias = light->param[RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS] * shadow_texel_size * 10.0; + + if (type == RS::LIGHT_SPOT) { + light_data.shadow_bias = light->param[RS::LIGHT_PARAM_SHADOW_BIAS] / 100.0; + } else { //omni + light_data.shadow_bias = light->param[RS::LIGHT_PARAM_SHADOW_BIAS]; + } + + light_data.transmittance_bias = light->param[RS::LIGHT_PARAM_TRANSMITTANCE_BIAS]; + + Vector2i omni_offset; + Rect2 rect = light_instance_get_shadow_atlas_rect(light_instance->self, p_shadow_atlas, omni_offset); + + light_data.atlas_rect[0] = rect.position.x; + light_data.atlas_rect[1] = rect.position.y; + light_data.atlas_rect[2] = rect.size.width; + light_data.atlas_rect[3] = rect.size.height; + + light_data.soft_shadow_scale = light->param[RS::LIGHT_PARAM_SHADOW_BLUR]; + + if (type == RS::LIGHT_OMNI) { + Transform3D proj = (inverse_transform * light_transform).inverse(); + + RendererRD::MaterialStorage::store_transform(proj, light_data.shadow_matrix); + + if (size > 0.0 && light_data.soft_shadow_scale > 0.0) { + // Only enable PCSS-like soft shadows if blurring is enabled. + // Otherwise, performance would decrease with no visual difference. + light_data.soft_shadow_size = size; + } else { + light_data.soft_shadow_size = 0.0; + light_data.soft_shadow_scale *= RendererSceneRenderRD::get_singleton()->shadows_quality_radius_get(); // Only use quality radius for PCF + } + + light_data.direction[0] = omni_offset.x * float(rect.size.width); + light_data.direction[1] = omni_offset.y * float(rect.size.height); + } else if (type == RS::LIGHT_SPOT) { + Transform3D modelview = (inverse_transform * light_transform).inverse(); + Projection bias; + bias.set_light_bias(); + + Projection cm = light_instance->shadow_transform[0].camera; + Projection shadow_mtx = bias * cm * modelview; + RendererRD::MaterialStorage::store_camera(shadow_mtx, light_data.shadow_matrix); + + if (size > 0.0 && light_data.soft_shadow_scale > 0.0) { + // Only enable PCSS-like soft shadows if blurring is enabled. + // Otherwise, performance would decrease with no visual difference. + float half_np = cm.get_z_near() * Math::tan(Math::deg_to_rad(spot_angle)); + light_data.soft_shadow_size = (size * 0.5 / radius) / (half_np / cm.get_z_near()) * rect.size.width; + } else { + light_data.soft_shadow_size = 0.0; + light_data.soft_shadow_scale *= RendererSceneRenderRD::get_singleton()->shadows_quality_radius_get(); // Only use quality radius for PCF + } + } + } else { + light_data.shadow_opacity = 0.0; + } + + light_instance->cull_mask = light->cull_mask; + + // hook for subclass to do further processing. + RendererSceneRenderRD::get_singleton()->setup_added_light(type, light_transform, radius, spot_angle); + + r_positional_light_count++; + } + + //update without barriers + if (omni_light_count) { + RD::get_singleton()->buffer_update(omni_light_buffer, 0, sizeof(LightData) * omni_light_count, omni_lights, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); + } + + if (spot_light_count) { + RD::get_singleton()->buffer_update(spot_light_buffer, 0, sizeof(LightData) * spot_light_count, spot_lights, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); + } + + if (r_directional_light_count) { + RD::get_singleton()->buffer_update(directional_light_buffer, 0, sizeof(DirectionalLightData) * r_directional_light_count, directional_lights, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); + } +} + +/* REFLECTION PROBE */ + +RID LightStorage::reflection_probe_allocate() { + return reflection_probe_owner.allocate_rid(); +} + +void LightStorage::reflection_probe_initialize(RID p_reflection_probe) { + reflection_probe_owner.initialize_rid(p_reflection_probe, ReflectionProbe()); +} + +void LightStorage::reflection_probe_free(RID p_rid) { + ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_rid); + reflection_probe->dependency.deleted_notify(p_rid); + reflection_probe_owner.free(p_rid); +}; + +void LightStorage::reflection_probe_set_update_mode(RID p_probe, RS::ReflectionProbeUpdateMode p_mode) { + ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND(!reflection_probe); + + reflection_probe->update_mode = p_mode; + reflection_probe->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_REFLECTION_PROBE); +} + +void LightStorage::reflection_probe_set_intensity(RID p_probe, float p_intensity) { + ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND(!reflection_probe); + + reflection_probe->intensity = p_intensity; +} + +void LightStorage::reflection_probe_set_ambient_mode(RID p_probe, RS::ReflectionProbeAmbientMode p_mode) { + ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND(!reflection_probe); + + reflection_probe->ambient_mode = p_mode; +} + +void LightStorage::reflection_probe_set_ambient_color(RID p_probe, const Color &p_color) { + ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND(!reflection_probe); + + reflection_probe->ambient_color = p_color; +} + +void LightStorage::reflection_probe_set_ambient_energy(RID p_probe, float p_energy) { + ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND(!reflection_probe); + + reflection_probe->ambient_color_energy = p_energy; +} + +void LightStorage::reflection_probe_set_max_distance(RID p_probe, float p_distance) { + ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND(!reflection_probe); + + reflection_probe->max_distance = p_distance; + + reflection_probe->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_REFLECTION_PROBE); +} + +void LightStorage::reflection_probe_set_extents(RID p_probe, const Vector3 &p_extents) { + ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND(!reflection_probe); + + if (reflection_probe->extents == p_extents) { + return; + } + reflection_probe->extents = p_extents; + reflection_probe->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_REFLECTION_PROBE); +} + +void LightStorage::reflection_probe_set_origin_offset(RID p_probe, const Vector3 &p_offset) { + ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND(!reflection_probe); + + reflection_probe->origin_offset = p_offset; + reflection_probe->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_REFLECTION_PROBE); +} + +void LightStorage::reflection_probe_set_as_interior(RID p_probe, bool p_enable) { + ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND(!reflection_probe); + + reflection_probe->interior = p_enable; + reflection_probe->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_REFLECTION_PROBE); +} + +void LightStorage::reflection_probe_set_enable_box_projection(RID p_probe, bool p_enable) { + ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND(!reflection_probe); + + reflection_probe->box_projection = p_enable; +} + +void LightStorage::reflection_probe_set_enable_shadows(RID p_probe, bool p_enable) { + ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND(!reflection_probe); + + reflection_probe->enable_shadows = p_enable; + reflection_probe->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_REFLECTION_PROBE); +} + +void LightStorage::reflection_probe_set_cull_mask(RID p_probe, uint32_t p_layers) { + ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND(!reflection_probe); + + reflection_probe->cull_mask = p_layers; + reflection_probe->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_REFLECTION_PROBE); +} + +void LightStorage::reflection_probe_set_resolution(RID p_probe, int p_resolution) { + ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND(!reflection_probe); + ERR_FAIL_COND(p_resolution < 32); + + reflection_probe->resolution = p_resolution; +} + +void LightStorage::reflection_probe_set_mesh_lod_threshold(RID p_probe, float p_ratio) { + ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND(!reflection_probe); + + reflection_probe->mesh_lod_threshold = p_ratio; + + reflection_probe->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_REFLECTION_PROBE); +} + +void LightStorage::reflection_probe_set_baked_exposure(RID p_probe, float p_exposure) { + ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND(!reflection_probe); + + reflection_probe->baked_exposure = p_exposure; +} + +AABB LightStorage::reflection_probe_get_aabb(RID p_probe) const { + const ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND_V(!reflection_probe, AABB()); + + AABB aabb; + aabb.position = -reflection_probe->extents; + aabb.size = reflection_probe->extents * 2.0; + + return aabb; +} + +RS::ReflectionProbeUpdateMode LightStorage::reflection_probe_get_update_mode(RID p_probe) const { + const ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND_V(!reflection_probe, RS::REFLECTION_PROBE_UPDATE_ALWAYS); + + return reflection_probe->update_mode; +} + +uint32_t LightStorage::reflection_probe_get_cull_mask(RID p_probe) const { + const ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND_V(!reflection_probe, 0); + + return reflection_probe->cull_mask; +} + +Vector3 LightStorage::reflection_probe_get_extents(RID p_probe) const { + const ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND_V(!reflection_probe, Vector3()); + + return reflection_probe->extents; +} + +Vector3 LightStorage::reflection_probe_get_origin_offset(RID p_probe) const { + const ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND_V(!reflection_probe, Vector3()); + + return reflection_probe->origin_offset; +} + +bool LightStorage::reflection_probe_renders_shadows(RID p_probe) const { + const ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND_V(!reflection_probe, false); + + return reflection_probe->enable_shadows; +} + +float LightStorage::reflection_probe_get_origin_max_distance(RID p_probe) const { + const ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND_V(!reflection_probe, 0); + + return reflection_probe->max_distance; +} + +float LightStorage::reflection_probe_get_mesh_lod_threshold(RID p_probe) const { + const ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND_V(!reflection_probe, 0); + + return reflection_probe->mesh_lod_threshold; +} + +int LightStorage::reflection_probe_get_resolution(RID p_probe) const { + const ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND_V(!reflection_probe, 0); + + return reflection_probe->resolution; +} + +float LightStorage::reflection_probe_get_baked_exposure(RID p_probe) const { + const ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND_V(!reflection_probe, 1.0); + + return reflection_probe->baked_exposure; +} + +float LightStorage::reflection_probe_get_intensity(RID p_probe) const { + const ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND_V(!reflection_probe, 0); + + return reflection_probe->intensity; +} + +bool LightStorage::reflection_probe_is_interior(RID p_probe) const { + const ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND_V(!reflection_probe, false); + + return reflection_probe->interior; +} + +bool LightStorage::reflection_probe_is_box_projection(RID p_probe) const { + const ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND_V(!reflection_probe, false); + + return reflection_probe->box_projection; +} + +RS::ReflectionProbeAmbientMode LightStorage::reflection_probe_get_ambient_mode(RID p_probe) const { + const ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND_V(!reflection_probe, RS::REFLECTION_PROBE_AMBIENT_DISABLED); + return reflection_probe->ambient_mode; +} + +Color LightStorage::reflection_probe_get_ambient_color(RID p_probe) const { + const ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND_V(!reflection_probe, Color()); + + return reflection_probe->ambient_color; +} +float LightStorage::reflection_probe_get_ambient_color_energy(RID p_probe) const { + const ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_COND_V(!reflection_probe, 0); + + return reflection_probe->ambient_color_energy; +} + +Dependency *LightStorage::reflection_probe_get_dependency(RID p_probe) const { + ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); + ERR_FAIL_NULL_V(reflection_probe, nullptr); + + return &reflection_probe->dependency; +} + +/* REFLECTION ATLAS */ + +RID LightStorage::reflection_atlas_create() { + ReflectionAtlas ra; + ra.count = GLOBAL_GET("rendering/reflections/reflection_atlas/reflection_count"); + ra.size = GLOBAL_GET("rendering/reflections/reflection_atlas/reflection_size"); + ra.cluster_builder = nullptr; + + return reflection_atlas_owner.make_rid(ra); +} + +void LightStorage::reflection_atlas_free(RID p_ref_atlas) { + reflection_atlas_set_size(p_ref_atlas, 0, 0); + ReflectionAtlas *ra = reflection_atlas_owner.get_or_null(p_ref_atlas); + if (ra->cluster_builder) { + memdelete(ra->cluster_builder); + } + reflection_atlas_owner.free(p_ref_atlas); +} + +void LightStorage::reflection_atlas_set_size(RID p_ref_atlas, int p_reflection_size, int p_reflection_count) { + ReflectionAtlas *ra = reflection_atlas_owner.get_or_null(p_ref_atlas); + ERR_FAIL_COND(!ra); + + if (ra->size == p_reflection_size && ra->count == p_reflection_count) { + return; //no changes + } + + if (ra->cluster_builder) { + // only if we're using our cluster + ra->cluster_builder->setup(Size2i(ra->size, ra->size), max_cluster_elements, RID(), RID(), RID()); + } + + ra->size = p_reflection_size; + ra->count = p_reflection_count; + + if (ra->reflection.is_valid()) { + //clear and invalidate everything + RD::get_singleton()->free(ra->reflection); + ra->reflection = RID(); + RD::get_singleton()->free(ra->depth_buffer); + ra->depth_buffer = RID(); + for (int i = 0; i < ra->reflections.size(); i++) { + ra->reflections.write[i].data.clear_reflection_data(); + if (ra->reflections[i].owner.is_null()) { + continue; + } + reflection_probe_release_atlas_index(ra->reflections[i].owner); + //rp->atlasindex clear + } + + ra->reflections.clear(); + } +} + +int LightStorage::reflection_atlas_get_size(RID p_ref_atlas) const { + ReflectionAtlas *ra = reflection_atlas_owner.get_or_null(p_ref_atlas); + ERR_FAIL_COND_V(!ra, 0); + + return ra->size; +} + +/* REFLECTION PROBE INSTANCE */ + +RID LightStorage::reflection_probe_instance_create(RID p_probe) { + ReflectionProbeInstance rpi; + rpi.probe = p_probe; + rpi.forward_id = ForwardIDStorage::get_singleton()->allocate_forward_id(FORWARD_ID_TYPE_REFLECTION_PROBE); + + return reflection_probe_instance_owner.make_rid(rpi); +} + +void LightStorage::reflection_probe_instance_free(RID p_instance) { + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.get_or_null(p_instance); + ForwardIDStorage::get_singleton()->free_forward_id(FORWARD_ID_TYPE_REFLECTION_PROBE, rpi->forward_id); + reflection_probe_release_atlas_index(p_instance); + reflection_probe_instance_owner.free(p_instance); +} + +void LightStorage::reflection_probe_instance_set_transform(RID p_instance, const Transform3D &p_transform) { + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.get_or_null(p_instance); + ERR_FAIL_COND(!rpi); + + rpi->transform = p_transform; + rpi->dirty = true; +} + +void LightStorage::reflection_probe_release_atlas_index(RID p_instance) { + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.get_or_null(p_instance); + ERR_FAIL_COND(!rpi); + + if (rpi->atlas.is_null()) { + return; //nothing to release + } + ReflectionAtlas *atlas = reflection_atlas_owner.get_or_null(rpi->atlas); + ERR_FAIL_COND(!atlas); + ERR_FAIL_INDEX(rpi->atlas_index, atlas->reflections.size()); + atlas->reflections.write[rpi->atlas_index].owner = RID(); + rpi->atlas_index = -1; + rpi->atlas = RID(); +} + +bool LightStorage::reflection_probe_instance_needs_redraw(RID p_instance) { + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.get_or_null(p_instance); + ERR_FAIL_COND_V(!rpi, false); + + if (rpi->rendering) { + return false; + } + + if (rpi->dirty) { + return true; + } + + if (LightStorage::get_singleton()->reflection_probe_get_update_mode(rpi->probe) == RS::REFLECTION_PROBE_UPDATE_ALWAYS) { + return true; + } + + return rpi->atlas_index == -1; +} + +bool LightStorage::reflection_probe_instance_has_reflection(RID p_instance) { + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.get_or_null(p_instance); + ERR_FAIL_COND_V(!rpi, false); + + return rpi->atlas.is_valid(); +} + +bool LightStorage::reflection_probe_instance_begin_render(RID p_instance, RID p_reflection_atlas) { + ReflectionAtlas *atlas = reflection_atlas_owner.get_or_null(p_reflection_atlas); + + ERR_FAIL_COND_V(!atlas, false); + + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.get_or_null(p_instance); + ERR_FAIL_COND_V(!rpi, false); + + RD::get_singleton()->draw_command_begin_label("Reflection probe render"); + + if (LightStorage::get_singleton()->reflection_probe_get_update_mode(rpi->probe) == RS::REFLECTION_PROBE_UPDATE_ALWAYS && atlas->reflection.is_valid() && atlas->size != 256) { + WARN_PRINT("ReflectionProbes set to UPDATE_ALWAYS must have an atlas size of 256. Please update the atlas size in the ProjectSettings."); + reflection_atlas_set_size(p_reflection_atlas, 256, atlas->count); + } + + if (LightStorage::get_singleton()->reflection_probe_get_update_mode(rpi->probe) == RS::REFLECTION_PROBE_UPDATE_ALWAYS && atlas->reflection.is_valid() && atlas->reflections[0].data.layers[0].mipmaps.size() != 8) { + // Invalidate reflection atlas, need to regenerate + RD::get_singleton()->free(atlas->reflection); + atlas->reflection = RID(); + + for (int i = 0; i < atlas->reflections.size(); i++) { + if (atlas->reflections[i].owner.is_null()) { + continue; + } + reflection_probe_release_atlas_index(atlas->reflections[i].owner); + } + + atlas->reflections.clear(); + } + + if (atlas->reflection.is_null()) { + int mipmaps = MIN(RendererSceneRenderRD::get_singleton()->get_sky()->roughness_layers, Image::get_image_required_mipmaps(atlas->size, atlas->size, Image::FORMAT_RGBAH) + 1); + mipmaps = LightStorage::get_singleton()->reflection_probe_get_update_mode(rpi->probe) == RS::REFLECTION_PROBE_UPDATE_ALWAYS ? 8 : mipmaps; // always use 8 mipmaps with real time filtering + { + //reflection atlas was unused, create: + RD::TextureFormat tf; + tf.array_layers = 6 * atlas->count; + tf.format = RendererSceneRenderRD::get_singleton()->_render_buffers_get_color_format(); + tf.texture_type = RD::TEXTURE_TYPE_CUBE_ARRAY; + tf.mipmaps = mipmaps; + tf.width = atlas->size; + tf.height = atlas->size; + tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | (RendererSceneRenderRD::get_singleton()->_render_buffers_can_be_storage() ? RD::TEXTURE_USAGE_STORAGE_BIT : 0); + + atlas->reflection = RD::get_singleton()->texture_create(tf, RD::TextureView()); + } + { + RD::TextureFormat tf; + tf.format = RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_D32_SFLOAT, RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) ? RD::DATA_FORMAT_D32_SFLOAT : RD::DATA_FORMAT_X8_D24_UNORM_PACK32; + tf.width = atlas->size; + tf.height = atlas->size; + tf.usage_bits = RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; + atlas->depth_buffer = RD::get_singleton()->texture_create(tf, RD::TextureView()); + } + atlas->reflections.resize(atlas->count); + for (int i = 0; i < atlas->count; i++) { + atlas->reflections.write[i].data.update_reflection_data(atlas->size, mipmaps, false, atlas->reflection, i * 6, LightStorage::get_singleton()->reflection_probe_get_update_mode(rpi->probe) == RS::REFLECTION_PROBE_UPDATE_ALWAYS, RendererSceneRenderRD::get_singleton()->get_sky()->roughness_layers, RendererSceneRenderRD::get_singleton()->_render_buffers_get_color_format()); + for (int j = 0; j < 6; j++) { + atlas->reflections.write[i].fbs[j] = RendererSceneRenderRD::get_singleton()->reflection_probe_create_framebuffer(atlas->reflections.write[i].data.layers[0].mipmaps[0].views[j], atlas->depth_buffer); + } + } + + Vector<RID> fb; + fb.push_back(atlas->depth_buffer); + atlas->depth_fb = RD::get_singleton()->framebuffer_create(fb); + } + + if (rpi->atlas_index == -1) { + for (int i = 0; i < atlas->reflections.size(); i++) { + if (atlas->reflections[i].owner.is_null()) { + rpi->atlas_index = i; + break; + } + } + //find the one used last + if (rpi->atlas_index == -1) { + //everything is in use, find the one least used via LRU + uint64_t pass_min = 0; + + for (int i = 0; i < atlas->reflections.size(); i++) { + ReflectionProbeInstance *rpi2 = reflection_probe_instance_owner.get_or_null(atlas->reflections[i].owner); + if (rpi2->last_pass < pass_min) { + pass_min = rpi2->last_pass; + rpi->atlas_index = i; + } + } + } + } + + if (rpi->atlas_index != -1) { // should we fail if this is still -1 ? + atlas->reflections.write[rpi->atlas_index].owner = p_instance; + } + + rpi->atlas = p_reflection_atlas; + rpi->rendering = true; + rpi->dirty = false; + rpi->processing_layer = 1; + rpi->processing_side = 0; + + RD::get_singleton()->draw_command_end_label(); + + return true; +} + +bool LightStorage::reflection_probe_instance_postprocess_step(RID p_instance) { + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.get_or_null(p_instance); + ERR_FAIL_COND_V(!rpi, false); + ERR_FAIL_COND_V(!rpi->rendering, false); + ERR_FAIL_COND_V(rpi->atlas.is_null(), false); + + ReflectionAtlas *atlas = reflection_atlas_owner.get_or_null(rpi->atlas); + if (!atlas || rpi->atlas_index == -1) { + //does not belong to an atlas anymore, cancel (was removed from atlas or atlas changed while rendering) + rpi->rendering = false; + return false; + } + + if (LightStorage::get_singleton()->reflection_probe_get_update_mode(rpi->probe) == RS::REFLECTION_PROBE_UPDATE_ALWAYS) { + // Using real time reflections, all roughness is done in one step + atlas->reflections.write[rpi->atlas_index].data.create_reflection_fast_filter(false); + rpi->rendering = false; + rpi->processing_side = 0; + rpi->processing_layer = 1; + return true; + } + + if (rpi->processing_layer > 1) { + atlas->reflections.write[rpi->atlas_index].data.create_reflection_importance_sample(false, 10, rpi->processing_layer, RendererSceneRenderRD::get_singleton()->get_sky()->sky_ggx_samples_quality); + rpi->processing_layer++; + if (rpi->processing_layer == atlas->reflections[rpi->atlas_index].data.layers[0].mipmaps.size()) { + rpi->rendering = false; + rpi->processing_side = 0; + rpi->processing_layer = 1; + return true; + } + return false; + + } else { + atlas->reflections.write[rpi->atlas_index].data.create_reflection_importance_sample(false, rpi->processing_side, rpi->processing_layer, RendererSceneRenderRD::get_singleton()->get_sky()->sky_ggx_samples_quality); + } + + rpi->processing_side++; + if (rpi->processing_side == 6) { + rpi->processing_side = 0; + rpi->processing_layer++; + } + + return false; +} + +uint32_t LightStorage::reflection_probe_instance_get_resolution(RID p_instance) { + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.get_or_null(p_instance); + ERR_FAIL_COND_V(!rpi, 0); + + ReflectionAtlas *atlas = reflection_atlas_owner.get_or_null(rpi->atlas); + ERR_FAIL_COND_V(!atlas, 0); + return atlas->size; +} + +RID LightStorage::reflection_probe_instance_get_framebuffer(RID p_instance, int p_index) { + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.get_or_null(p_instance); + ERR_FAIL_COND_V(!rpi, RID()); + ERR_FAIL_INDEX_V(p_index, 6, RID()); + + ReflectionAtlas *atlas = reflection_atlas_owner.get_or_null(rpi->atlas); + ERR_FAIL_COND_V(!atlas, RID()); + return atlas->reflections[rpi->atlas_index].fbs[p_index]; +} + +RID LightStorage::reflection_probe_instance_get_depth_framebuffer(RID p_instance, int p_index) { + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.get_or_null(p_instance); + ERR_FAIL_COND_V(!rpi, RID()); + ERR_FAIL_INDEX_V(p_index, 6, RID()); + + ReflectionAtlas *atlas = reflection_atlas_owner.get_or_null(rpi->atlas); + ERR_FAIL_COND_V(!atlas, RID()); + return atlas->depth_fb; +} + +ClusterBuilderRD *LightStorage::reflection_probe_instance_get_cluster_builder(RID p_instance, ClusterBuilderSharedDataRD *p_cluster_builder_shared) { + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.get_or_null(p_instance); + ReflectionAtlas *ra = reflection_atlas_owner.get_or_null(rpi->atlas); + if (!ra) { + ERR_PRINT("reflection probe has no reflection atlas! Bug?"); + return nullptr; + } else { + if (ra->cluster_builder == nullptr) { + ra->cluster_builder = memnew(ClusterBuilderRD); + ra->cluster_builder->set_shared(p_cluster_builder_shared); + ra->cluster_builder->setup(Size2i(ra->size, ra->size), get_max_cluster_elements(), RID(), RID(), RID()); + } + return ra->cluster_builder; + } +} + +/* REFLECTION DATA */ + +void LightStorage::free_reflection_data() { + if (reflection_buffer.is_valid()) { + RD::get_singleton()->free(reflection_buffer); + reflection_buffer = RID(); + } + + if (reflections != nullptr) { + memdelete_arr(reflections); + reflections = nullptr; + } + + if (reflection_sort != nullptr) { + memdelete_arr(reflection_sort); + reflection_sort = nullptr; + } +} + +void LightStorage::set_max_reflection_probes(const uint32_t p_max_reflection_probes) { + max_reflections = p_max_reflection_probes; + reflections = memnew_arr(ReflectionData, max_reflections); + reflection_sort = memnew_arr(ReflectionProbeInstanceSort, max_reflections); + reflection_buffer = RD::get_singleton()->storage_buffer_create(sizeof(ReflectionData) * max_reflections); +} + +void LightStorage::update_reflection_probe_buffer(RenderDataRD *p_render_data, const PagedArray<RID> &p_reflections, const Transform3D &p_camera_inverse_transform, RID p_environment) { + ForwardIDStorage *forward_id_storage = ForwardIDStorage::get_singleton(); + + reflection_count = 0; + + for (uint32_t i = 0; i < (uint32_t)p_reflections.size(); i++) { + if (reflection_count == max_reflections) { + break; + } + + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.get_or_null(p_reflections[i]); + if (!rpi) { + continue; + } + + Transform3D transform = rpi->transform; + + reflection_sort[reflection_count].probe_instance = rpi; + reflection_sort[reflection_count].depth = -p_camera_inverse_transform.xform(transform.origin).z; + reflection_count++; + } + + if (reflection_count > 0) { + SortArray<ReflectionProbeInstanceSort> sort_array; + sort_array.sort(reflection_sort, reflection_count); + } + + bool using_forward_ids = forward_id_storage->uses_forward_ids(); + for (uint32_t i = 0; i < reflection_count; i++) { + ReflectionProbeInstance *rpi = reflection_sort[i].probe_instance; + + if (using_forward_ids) { + forward_id_storage->map_forward_id(FORWARD_ID_TYPE_REFLECTION_PROBE, rpi->forward_id, i); + } + + ReflectionProbe *probe = reflection_probe_owner.get_or_null(rpi->probe); + + ReflectionData &reflection_ubo = reflections[i]; + + Vector3 extents = probe->extents; + + rpi->cull_mask = probe->cull_mask; + + reflection_ubo.box_extents[0] = extents.x; + reflection_ubo.box_extents[1] = extents.y; + reflection_ubo.box_extents[2] = extents.z; + reflection_ubo.index = rpi->atlas_index; + + Vector3 origin_offset = probe->origin_offset; + + reflection_ubo.box_offset[0] = origin_offset.x; + reflection_ubo.box_offset[1] = origin_offset.y; + reflection_ubo.box_offset[2] = origin_offset.z; + reflection_ubo.mask = probe->cull_mask; + + reflection_ubo.intensity = probe->intensity; + reflection_ubo.ambient_mode = probe->ambient_mode; + + reflection_ubo.exterior = !probe->interior; + reflection_ubo.box_project = probe->box_projection; + reflection_ubo.exposure_normalization = 1.0; + + if (p_render_data->camera_attributes.is_valid()) { + float exposure = RSG::camera_attributes->camera_attributes_get_exposure_normalization_factor(p_render_data->camera_attributes); + reflection_ubo.exposure_normalization = exposure / probe->baked_exposure; + } + + Color ambient_linear = probe->ambient_color.srgb_to_linear(); + float interior_ambient_energy = probe->ambient_color_energy; + reflection_ubo.ambient[0] = ambient_linear.r * interior_ambient_energy; + reflection_ubo.ambient[1] = ambient_linear.g * interior_ambient_energy; + reflection_ubo.ambient[2] = ambient_linear.b * interior_ambient_energy; + + Transform3D transform = rpi->transform; + Transform3D proj = (p_camera_inverse_transform * transform).inverse(); + MaterialStorage::store_transform(proj, reflection_ubo.local_matrix); + + // hook for subclass to do further processing. + RendererSceneRenderRD::get_singleton()->setup_added_reflection_probe(transform, extents); + + rpi->last_pass = RSG::rasterizer->get_frame_number(); + } + + if (reflection_count) { + RD::get_singleton()->buffer_update(reflection_buffer, 0, reflection_count * sizeof(ReflectionData), reflections, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); + } +} + +/* LIGHTMAP API */ + +RID LightStorage::lightmap_allocate() { + return lightmap_owner.allocate_rid(); +} + +void LightStorage::lightmap_initialize(RID p_lightmap) { + lightmap_owner.initialize_rid(p_lightmap, Lightmap()); +} + +void LightStorage::lightmap_free(RID p_rid) { + lightmap_set_textures(p_rid, RID(), false); + Lightmap *lightmap = lightmap_owner.get_or_null(p_rid); + lightmap->dependency.deleted_notify(p_rid); + lightmap_owner.free(p_rid); +} + +void LightStorage::lightmap_set_textures(RID p_lightmap, RID p_light, bool p_uses_spherical_haromics) { + TextureStorage *texture_storage = TextureStorage::get_singleton(); + + Lightmap *lm = lightmap_owner.get_or_null(p_lightmap); + ERR_FAIL_COND(!lm); + + lightmap_array_version++; + + //erase lightmap users + if (lm->light_texture.is_valid()) { + TextureStorage::Texture *t = texture_storage->get_singleton()->get_texture(lm->light_texture); + if (t) { + t->lightmap_users.erase(p_lightmap); + } + } + + TextureStorage::Texture *t = texture_storage->get_singleton()->get_texture(p_light); + lm->light_texture = p_light; + lm->uses_spherical_harmonics = p_uses_spherical_haromics; + + RID default_2d_array = texture_storage->texture_rd_get_default(TextureStorage::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE); + if (!t) { + if (using_lightmap_array) { + if (lm->array_index >= 0) { + lightmap_textures.write[lm->array_index] = default_2d_array; + lm->array_index = -1; + } + } + + return; + } + + t->lightmap_users.insert(p_lightmap); + + if (using_lightmap_array) { + if (lm->array_index < 0) { + //not in array, try to put in array + for (int i = 0; i < lightmap_textures.size(); i++) { + if (lightmap_textures[i] == default_2d_array) { + lm->array_index = i; + break; + } + } + } + ERR_FAIL_COND_MSG(lm->array_index < 0, "Maximum amount of lightmaps in use (" + itos(lightmap_textures.size()) + ") has been exceeded, lightmap will nod display properly."); + + lightmap_textures.write[lm->array_index] = t->rd_texture; + } +} + +void LightStorage::lightmap_set_probe_bounds(RID p_lightmap, const AABB &p_bounds) { + Lightmap *lm = lightmap_owner.get_or_null(p_lightmap); + ERR_FAIL_COND(!lm); + lm->bounds = p_bounds; +} + +void LightStorage::lightmap_set_probe_interior(RID p_lightmap, bool p_interior) { + Lightmap *lm = lightmap_owner.get_or_null(p_lightmap); + ERR_FAIL_COND(!lm); + lm->interior = p_interior; +} + +void LightStorage::lightmap_set_probe_capture_data(RID p_lightmap, const PackedVector3Array &p_points, const PackedColorArray &p_point_sh, const PackedInt32Array &p_tetrahedra, const PackedInt32Array &p_bsp_tree) { + Lightmap *lm = lightmap_owner.get_or_null(p_lightmap); + ERR_FAIL_COND(!lm); + + if (p_points.size()) { + ERR_FAIL_COND(p_points.size() * 9 != p_point_sh.size()); + ERR_FAIL_COND((p_tetrahedra.size() % 4) != 0); + ERR_FAIL_COND((p_bsp_tree.size() % 6) != 0); + } + + lm->points = p_points; + lm->bsp_tree = p_bsp_tree; + lm->point_sh = p_point_sh; + lm->tetrahedra = p_tetrahedra; +} + +void LightStorage::lightmap_set_baked_exposure_normalization(RID p_lightmap, float p_exposure) { + Lightmap *lm = lightmap_owner.get_or_null(p_lightmap); + ERR_FAIL_COND(!lm); + + lm->baked_exposure = p_exposure; +} + +PackedVector3Array LightStorage::lightmap_get_probe_capture_points(RID p_lightmap) const { + Lightmap *lm = lightmap_owner.get_or_null(p_lightmap); + ERR_FAIL_COND_V(!lm, PackedVector3Array()); + + return lm->points; +} + +PackedColorArray LightStorage::lightmap_get_probe_capture_sh(RID p_lightmap) const { + Lightmap *lm = lightmap_owner.get_or_null(p_lightmap); + ERR_FAIL_COND_V(!lm, PackedColorArray()); + return lm->point_sh; +} + +PackedInt32Array LightStorage::lightmap_get_probe_capture_tetrahedra(RID p_lightmap) const { + Lightmap *lm = lightmap_owner.get_or_null(p_lightmap); + ERR_FAIL_COND_V(!lm, PackedInt32Array()); + return lm->tetrahedra; +} + +PackedInt32Array LightStorage::lightmap_get_probe_capture_bsp_tree(RID p_lightmap) const { + Lightmap *lm = lightmap_owner.get_or_null(p_lightmap); + ERR_FAIL_COND_V(!lm, PackedInt32Array()); + return lm->bsp_tree; +} + +void LightStorage::lightmap_set_probe_capture_update_speed(float p_speed) { + lightmap_probe_capture_update_speed = p_speed; +} + +Dependency *LightStorage::lightmap_get_dependency(RID p_lightmap) const { + Lightmap *lm = lightmap_owner.get_or_null(p_lightmap); + ERR_FAIL_NULL_V(lm, nullptr); + + return &lm->dependency; +} + +void LightStorage::lightmap_tap_sh_light(RID p_lightmap, const Vector3 &p_point, Color *r_sh) { + Lightmap *lm = lightmap_owner.get_or_null(p_lightmap); + ERR_FAIL_COND(!lm); + + for (int i = 0; i < 9; i++) { + r_sh[i] = Color(0, 0, 0, 0); + } + + if (!lm->points.size() || !lm->bsp_tree.size() || !lm->tetrahedra.size()) { + return; + } + + static_assert(sizeof(Lightmap::BSP) == 24); + + const Lightmap::BSP *bsp = (const Lightmap::BSP *)lm->bsp_tree.ptr(); + int32_t node = 0; + while (node >= 0) { + if (Plane(bsp[node].plane[0], bsp[node].plane[1], bsp[node].plane[2], bsp[node].plane[3]).is_point_over(p_point)) { +#ifdef DEBUG_ENABLED + ERR_FAIL_COND(bsp[node].over >= 0 && bsp[node].over < node); +#endif + + node = bsp[node].over; + } else { +#ifdef DEBUG_ENABLED + ERR_FAIL_COND(bsp[node].under >= 0 && bsp[node].under < node); +#endif + node = bsp[node].under; + } + } + + if (node == Lightmap::BSP::EMPTY_LEAF) { + return; //nothing could be done + } + + node = ABS(node) - 1; + + uint32_t *tetrahedron = (uint32_t *)&lm->tetrahedra[node * 4]; + Vector3 points[4] = { lm->points[tetrahedron[0]], lm->points[tetrahedron[1]], lm->points[tetrahedron[2]], lm->points[tetrahedron[3]] }; + const Color *sh_colors[4]{ &lm->point_sh[tetrahedron[0] * 9], &lm->point_sh[tetrahedron[1] * 9], &lm->point_sh[tetrahedron[2] * 9], &lm->point_sh[tetrahedron[3] * 9] }; + Color barycentric = Geometry3D::tetrahedron_get_barycentric_coords(points[0], points[1], points[2], points[3], p_point); + + for (int i = 0; i < 4; i++) { + float c = CLAMP(barycentric[i], 0.0, 1.0); + for (int j = 0; j < 9; j++) { + r_sh[j] += sh_colors[i][j] * c; + } + } +} + +bool LightStorage::lightmap_is_interior(RID p_lightmap) const { + const Lightmap *lm = lightmap_owner.get_or_null(p_lightmap); + ERR_FAIL_COND_V(!lm, false); + return lm->interior; +} + +AABB LightStorage::lightmap_get_aabb(RID p_lightmap) const { + const Lightmap *lm = lightmap_owner.get_or_null(p_lightmap); + ERR_FAIL_COND_V(!lm, AABB()); + return lm->bounds; +} + +/* LIGHTMAP INSTANCE */ + +RID LightStorage::lightmap_instance_create(RID p_lightmap) { + LightmapInstance li; + li.lightmap = p_lightmap; + return lightmap_instance_owner.make_rid(li); +} + +void LightStorage::lightmap_instance_free(RID p_lightmap) { + lightmap_instance_owner.free(p_lightmap); +} + +void LightStorage::lightmap_instance_set_transform(RID p_lightmap, const Transform3D &p_transform) { + LightmapInstance *li = lightmap_instance_owner.get_or_null(p_lightmap); + ERR_FAIL_COND(!li); + li->transform = p_transform; +} + +/* SHADOW ATLAS API */ + +RID LightStorage::shadow_atlas_create() { + return shadow_atlas_owner.make_rid(ShadowAtlas()); +} + +void LightStorage::shadow_atlas_free(RID p_atlas) { + shadow_atlas_set_size(p_atlas, 0); + shadow_atlas_owner.free(p_atlas); +} + +void LightStorage::_update_shadow_atlas(ShadowAtlas *shadow_atlas) { + if (shadow_atlas->size > 0 && shadow_atlas->depth.is_null()) { + RD::TextureFormat tf; + tf.format = shadow_atlas->use_16_bits ? RD::DATA_FORMAT_D16_UNORM : RD::DATA_FORMAT_D32_SFLOAT; + tf.width = shadow_atlas->size; + tf.height = shadow_atlas->size; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT; + + shadow_atlas->depth = RD::get_singleton()->texture_create(tf, RD::TextureView()); + Vector<RID> fb_tex; + fb_tex.push_back(shadow_atlas->depth); + shadow_atlas->fb = RD::get_singleton()->framebuffer_create(fb_tex); + } +} + +void LightStorage::shadow_atlas_set_size(RID p_atlas, int p_size, bool p_16_bits) { + ShadowAtlas *shadow_atlas = shadow_atlas_owner.get_or_null(p_atlas); + ERR_FAIL_COND(!shadow_atlas); + ERR_FAIL_COND(p_size < 0); + p_size = next_power_of_2(p_size); + + if (p_size == shadow_atlas->size && p_16_bits == shadow_atlas->use_16_bits) { + return; + } + + // erasing atlas + if (shadow_atlas->depth.is_valid()) { + RD::get_singleton()->free(shadow_atlas->depth); + shadow_atlas->depth = RID(); + } + for (int i = 0; i < 4; i++) { + //clear subdivisions + shadow_atlas->quadrants[i].shadows.clear(); + shadow_atlas->quadrants[i].shadows.resize(1 << shadow_atlas->quadrants[i].subdivision); + } + + //erase shadow atlas reference from lights + for (const KeyValue<RID, uint32_t> &E : shadow_atlas->shadow_owners) { + LightInstance *li = light_instance_owner.get_or_null(E.key); + ERR_CONTINUE(!li); + li->shadow_atlases.erase(p_atlas); + } + + //clear owners + shadow_atlas->shadow_owners.clear(); + + shadow_atlas->size = p_size; + shadow_atlas->use_16_bits = p_16_bits; +} + +void LightStorage::shadow_atlas_set_quadrant_subdivision(RID p_atlas, int p_quadrant, int p_subdivision) { + ShadowAtlas *shadow_atlas = shadow_atlas_owner.get_or_null(p_atlas); + ERR_FAIL_COND(!shadow_atlas); + ERR_FAIL_INDEX(p_quadrant, 4); + ERR_FAIL_INDEX(p_subdivision, 16384); + + uint32_t subdiv = next_power_of_2(p_subdivision); + if (subdiv & 0xaaaaaaaa) { //sqrt(subdiv) must be integer + subdiv <<= 1; + } + + subdiv = int(Math::sqrt((float)subdiv)); + + //obtain the number that will be x*x + + if (shadow_atlas->quadrants[p_quadrant].subdivision == subdiv) { + return; + } + + //erase all data from quadrant + for (int i = 0; i < shadow_atlas->quadrants[p_quadrant].shadows.size(); i++) { + if (shadow_atlas->quadrants[p_quadrant].shadows[i].owner.is_valid()) { + shadow_atlas->shadow_owners.erase(shadow_atlas->quadrants[p_quadrant].shadows[i].owner); + LightInstance *li = light_instance_owner.get_or_null(shadow_atlas->quadrants[p_quadrant].shadows[i].owner); + ERR_CONTINUE(!li); + li->shadow_atlases.erase(p_atlas); + } + } + + shadow_atlas->quadrants[p_quadrant].shadows.clear(); + shadow_atlas->quadrants[p_quadrant].shadows.resize(subdiv * subdiv); + shadow_atlas->quadrants[p_quadrant].subdivision = subdiv; + + //cache the smallest subdiv (for faster allocation in light update) + + shadow_atlas->smallest_subdiv = 1 << 30; + + for (int i = 0; i < 4; i++) { + if (shadow_atlas->quadrants[i].subdivision) { + shadow_atlas->smallest_subdiv = MIN(shadow_atlas->smallest_subdiv, shadow_atlas->quadrants[i].subdivision); + } + } + + if (shadow_atlas->smallest_subdiv == 1 << 30) { + shadow_atlas->smallest_subdiv = 0; + } + + //resort the size orders, simple bublesort for 4 elements.. + + int swaps = 0; + do { + swaps = 0; + + for (int i = 0; i < 3; i++) { + if (shadow_atlas->quadrants[shadow_atlas->size_order[i]].subdivision < shadow_atlas->quadrants[shadow_atlas->size_order[i + 1]].subdivision) { + SWAP(shadow_atlas->size_order[i], shadow_atlas->size_order[i + 1]); + swaps++; + } + } + } while (swaps > 0); +} + +bool LightStorage::_shadow_atlas_find_shadow(ShadowAtlas *shadow_atlas, int *p_in_quadrants, int p_quadrant_count, int p_current_subdiv, uint64_t p_tick, int &r_quadrant, int &r_shadow) { + for (int i = p_quadrant_count - 1; i >= 0; i--) { + int qidx = p_in_quadrants[i]; + + if (shadow_atlas->quadrants[qidx].subdivision == (uint32_t)p_current_subdiv) { + return false; + } + + //look for an empty space + int sc = shadow_atlas->quadrants[qidx].shadows.size(); + const ShadowAtlas::Quadrant::Shadow *sarr = shadow_atlas->quadrants[qidx].shadows.ptr(); + + int found_free_idx = -1; //found a free one + int found_used_idx = -1; //found existing one, must steal it + uint64_t min_pass = 0; // pass of the existing one, try to use the least recently used one (LRU fashion) + + for (int j = 0; j < sc; j++) { + if (!sarr[j].owner.is_valid()) { + found_free_idx = j; + break; + } + + LightInstance *sli = light_instance_owner.get_or_null(sarr[j].owner); + ERR_CONTINUE(!sli); + + if (sli->last_scene_pass != RendererSceneRenderRD::get_singleton()->get_scene_pass()) { + //was just allocated, don't kill it so soon, wait a bit.. + if (p_tick - sarr[j].alloc_tick < shadow_atlas_realloc_tolerance_msec) { + continue; + } + + if (found_used_idx == -1 || sli->last_scene_pass < min_pass) { + found_used_idx = j; + min_pass = sli->last_scene_pass; + } + } + } + + if (found_free_idx == -1 && found_used_idx == -1) { + continue; //nothing found + } + + if (found_free_idx == -1 && found_used_idx != -1) { + found_free_idx = found_used_idx; + } + + r_quadrant = qidx; + r_shadow = found_free_idx; + + return true; + } + + return false; +} + +bool LightStorage::_shadow_atlas_find_omni_shadows(ShadowAtlas *shadow_atlas, int *p_in_quadrants, int p_quadrant_count, int p_current_subdiv, uint64_t p_tick, int &r_quadrant, int &r_shadow) { + for (int i = p_quadrant_count - 1; i >= 0; i--) { + int qidx = p_in_quadrants[i]; + + if (shadow_atlas->quadrants[qidx].subdivision == (uint32_t)p_current_subdiv) { + return false; + } + + //look for an empty space + int sc = shadow_atlas->quadrants[qidx].shadows.size(); + const ShadowAtlas::Quadrant::Shadow *sarr = shadow_atlas->quadrants[qidx].shadows.ptr(); + + int found_idx = -1; + uint64_t min_pass = 0; // sum of currently selected spots, try to get the least recently used pair + + for (int j = 0; j < sc - 1; j++) { + uint64_t pass = 0; + + if (sarr[j].owner.is_valid()) { + LightInstance *sli = light_instance_owner.get_or_null(sarr[j].owner); + ERR_CONTINUE(!sli); + + if (sli->last_scene_pass == RendererSceneRenderRD::get_singleton()->get_scene_pass()) { + continue; + } + + //was just allocated, don't kill it so soon, wait a bit.. + if (p_tick - sarr[j].alloc_tick < shadow_atlas_realloc_tolerance_msec) { + continue; + } + pass += sli->last_scene_pass; + } + + if (sarr[j + 1].owner.is_valid()) { + LightInstance *sli = light_instance_owner.get_or_null(sarr[j + 1].owner); + ERR_CONTINUE(!sli); + + if (sli->last_scene_pass == RendererSceneRenderRD::get_singleton()->get_scene_pass()) { + continue; + } + + //was just allocated, don't kill it so soon, wait a bit.. + if (p_tick - sarr[j + 1].alloc_tick < shadow_atlas_realloc_tolerance_msec) { + continue; + } + pass += sli->last_scene_pass; + } + + if (found_idx == -1 || pass < min_pass) { + found_idx = j; + min_pass = pass; + + // we found two empty spots, no need to check the rest + if (pass == 0) { + break; + } + } + } + + if (found_idx == -1) { + continue; //nothing found + } + + r_quadrant = qidx; + r_shadow = found_idx; + + return true; + } + + return false; +} + +bool LightStorage::shadow_atlas_update_light(RID p_atlas, RID p_light_instance, float p_coverage, uint64_t p_light_version) { + ShadowAtlas *shadow_atlas = shadow_atlas_owner.get_or_null(p_atlas); + ERR_FAIL_COND_V(!shadow_atlas, false); + + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + ERR_FAIL_COND_V(!li, false); + + if (shadow_atlas->size == 0 || shadow_atlas->smallest_subdiv == 0) { + return false; + } + + uint32_t quad_size = shadow_atlas->size >> 1; + int desired_fit = MIN(quad_size / shadow_atlas->smallest_subdiv, next_power_of_2(quad_size * p_coverage)); + + int valid_quadrants[4]; + int valid_quadrant_count = 0; + int best_size = -1; //best size found + int best_subdiv = -1; //subdiv for the best size + + //find the quadrants this fits into, and the best possible size it can fit into + for (int i = 0; i < 4; i++) { + int q = shadow_atlas->size_order[i]; + int sd = shadow_atlas->quadrants[q].subdivision; + if (sd == 0) { + continue; //unused + } + + int max_fit = quad_size / sd; + + if (best_size != -1 && max_fit > best_size) { + break; //too large + } + + valid_quadrants[valid_quadrant_count++] = q; + best_subdiv = sd; + + if (max_fit >= desired_fit) { + best_size = max_fit; + } + } + + ERR_FAIL_COND_V(valid_quadrant_count == 0, false); + + uint64_t tick = OS::get_singleton()->get_ticks_msec(); + + uint32_t old_key = SHADOW_INVALID; + uint32_t old_quadrant = SHADOW_INVALID; + uint32_t old_shadow = SHADOW_INVALID; + int old_subdivision = -1; + + bool should_realloc = false; + bool should_redraw = false; + + if (shadow_atlas->shadow_owners.has(p_light_instance)) { + old_key = shadow_atlas->shadow_owners[p_light_instance]; + old_quadrant = (old_key >> QUADRANT_SHIFT) & 0x3; + old_shadow = old_key & SHADOW_INDEX_MASK; + + should_realloc = shadow_atlas->quadrants[old_quadrant].subdivision != (uint32_t)best_subdiv && (shadow_atlas->quadrants[old_quadrant].shadows[old_shadow].alloc_tick - tick > shadow_atlas_realloc_tolerance_msec); + should_redraw = shadow_atlas->quadrants[old_quadrant].shadows[old_shadow].version != p_light_version; + + if (!should_realloc) { + shadow_atlas->quadrants[old_quadrant].shadows.write[old_shadow].version = p_light_version; + //already existing, see if it should redraw or it's just OK + return should_redraw; + } + + old_subdivision = shadow_atlas->quadrants[old_quadrant].subdivision; + } + + bool is_omni = li->light_type == RS::LIGHT_OMNI; + bool found_shadow = false; + int new_quadrant = -1; + int new_shadow = -1; + + if (is_omni) { + found_shadow = _shadow_atlas_find_omni_shadows(shadow_atlas, valid_quadrants, valid_quadrant_count, old_subdivision, tick, new_quadrant, new_shadow); + } else { + found_shadow = _shadow_atlas_find_shadow(shadow_atlas, valid_quadrants, valid_quadrant_count, old_subdivision, tick, new_quadrant, new_shadow); + } + + if (found_shadow) { + if (old_quadrant != SHADOW_INVALID) { + shadow_atlas->quadrants[old_quadrant].shadows.write[old_shadow].version = 0; + shadow_atlas->quadrants[old_quadrant].shadows.write[old_shadow].owner = RID(); + + if (old_key & OMNI_LIGHT_FLAG) { + shadow_atlas->quadrants[old_quadrant].shadows.write[old_shadow + 1].version = 0; + shadow_atlas->quadrants[old_quadrant].shadows.write[old_shadow + 1].owner = RID(); + } + } + + uint32_t new_key = new_quadrant << QUADRANT_SHIFT; + new_key |= new_shadow; + + ShadowAtlas::Quadrant::Shadow *sh = &shadow_atlas->quadrants[new_quadrant].shadows.write[new_shadow]; + _shadow_atlas_invalidate_shadow(sh, p_atlas, shadow_atlas, new_quadrant, new_shadow); + + sh->owner = p_light_instance; + sh->alloc_tick = tick; + sh->version = p_light_version; + + if (is_omni) { + new_key |= OMNI_LIGHT_FLAG; + + int new_omni_shadow = new_shadow + 1; + ShadowAtlas::Quadrant::Shadow *extra_sh = &shadow_atlas->quadrants[new_quadrant].shadows.write[new_omni_shadow]; + _shadow_atlas_invalidate_shadow(extra_sh, p_atlas, shadow_atlas, new_quadrant, new_omni_shadow); + + extra_sh->owner = p_light_instance; + extra_sh->alloc_tick = tick; + extra_sh->version = p_light_version; + } + + li->shadow_atlases.insert(p_atlas); + + //update it in map + shadow_atlas->shadow_owners[p_light_instance] = new_key; + //make it dirty, as it should redraw anyway + return true; + } + + return should_redraw; +} + +void LightStorage::_shadow_atlas_invalidate_shadow(ShadowAtlas::Quadrant::Shadow *p_shadow, RID p_atlas, ShadowAtlas *p_shadow_atlas, uint32_t p_quadrant, uint32_t p_shadow_idx) { + if (p_shadow->owner.is_valid()) { + LightInstance *sli = light_instance_owner.get_or_null(p_shadow->owner); + uint32_t old_key = p_shadow_atlas->shadow_owners[p_shadow->owner]; + + if (old_key & OMNI_LIGHT_FLAG) { + uint32_t s = old_key & SHADOW_INDEX_MASK; + uint32_t omni_shadow_idx = p_shadow_idx + (s == (uint32_t)p_shadow_idx ? 1 : -1); + ShadowAtlas::Quadrant::Shadow *omni_shadow = &p_shadow_atlas->quadrants[p_quadrant].shadows.write[omni_shadow_idx]; + omni_shadow->version = 0; + omni_shadow->owner = RID(); + } + + p_shadow_atlas->shadow_owners.erase(p_shadow->owner); + p_shadow->version = 0; + p_shadow->owner = RID(); + sli->shadow_atlases.erase(p_atlas); + } +} + +void LightStorage::shadow_atlas_update(RID p_atlas) { + ShadowAtlas *shadow_atlas = shadow_atlas_owner.get_or_null(p_atlas); + ERR_FAIL_COND(!shadow_atlas); + + _update_shadow_atlas(shadow_atlas); +} + +/* DIRECTIONAL SHADOW */ + +void LightStorage::update_directional_shadow_atlas() { + if (directional_shadow.depth.is_null() && directional_shadow.size > 0) { + RD::TextureFormat tf; + tf.format = directional_shadow.use_16_bits ? RD::DATA_FORMAT_D16_UNORM : RD::DATA_FORMAT_D32_SFLOAT; + tf.width = directional_shadow.size; + tf.height = directional_shadow.size; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT; + + directional_shadow.depth = RD::get_singleton()->texture_create(tf, RD::TextureView()); + Vector<RID> fb_tex; + fb_tex.push_back(directional_shadow.depth); + directional_shadow.fb = RD::get_singleton()->framebuffer_create(fb_tex); + } +} +void LightStorage::directional_shadow_atlas_set_size(int p_size, bool p_16_bits) { + p_size = nearest_power_of_2_templated(p_size); + + if (directional_shadow.size == p_size && directional_shadow.use_16_bits == p_16_bits) { + return; + } + + directional_shadow.size = p_size; + directional_shadow.use_16_bits = p_16_bits; + + if (directional_shadow.depth.is_valid()) { + RD::get_singleton()->free(directional_shadow.depth); + directional_shadow.depth = RID(); + RendererSceneRenderRD::get_singleton()->base_uniforms_changed(); + } +} + +void LightStorage::set_directional_shadow_count(int p_count) { + directional_shadow.light_count = p_count; + directional_shadow.current_light = 0; +} + +static Rect2i _get_directional_shadow_rect(int p_size, int p_shadow_count, int p_shadow_index) { + int split_h = 1; + int split_v = 1; + + while (split_h * split_v < p_shadow_count) { + if (split_h == split_v) { + split_h <<= 1; + } else { + split_v <<= 1; + } + } + + Rect2i rect(0, 0, p_size, p_size); + rect.size.width /= split_h; + rect.size.height /= split_v; + + rect.position.x = rect.size.width * (p_shadow_index % split_h); + rect.position.y = rect.size.height * (p_shadow_index / split_h); + + return rect; +} + +Rect2i LightStorage::get_directional_shadow_rect() { + return _get_directional_shadow_rect(directional_shadow.size, directional_shadow.light_count, directional_shadow.current_light); +} + +int LightStorage::get_directional_light_shadow_size(RID p_light_intance) { + ERR_FAIL_COND_V(directional_shadow.light_count == 0, 0); + + Rect2i r = _get_directional_shadow_rect(directional_shadow.size, directional_shadow.light_count, 0); + + LightInstance *light_instance = light_instance_owner.get_or_null(p_light_intance); + ERR_FAIL_COND_V(!light_instance, 0); + + switch (light_directional_get_shadow_mode(light_instance->light)) { + case RS::LIGHT_DIRECTIONAL_SHADOW_ORTHOGONAL: + break; //none + case RS::LIGHT_DIRECTIONAL_SHADOW_PARALLEL_2_SPLITS: + r.size.height /= 2; + break; + case RS::LIGHT_DIRECTIONAL_SHADOW_PARALLEL_4_SPLITS: + r.size /= 2; + break; + } + + return MAX(r.size.width, r.size.height); +} + +/* SHADOW CUBEMAPS */ + +LightStorage::ShadowCubemap *LightStorage::_get_shadow_cubemap(int p_size) { + if (!shadow_cubemaps.has(p_size)) { + ShadowCubemap sc; + { + RD::TextureFormat tf; + tf.format = RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_D32_SFLOAT, RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) ? RD::DATA_FORMAT_D32_SFLOAT : RD::DATA_FORMAT_X8_D24_UNORM_PACK32; + tf.width = p_size; + tf.height = p_size; + tf.texture_type = RD::TEXTURE_TYPE_CUBE; + tf.array_layers = 6; + tf.usage_bits = RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; + sc.cubemap = RD::get_singleton()->texture_create(tf, RD::TextureView()); + } + + for (int i = 0; i < 6; i++) { + RID side_texture = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), sc.cubemap, i, 0); + Vector<RID> fbtex; + fbtex.push_back(side_texture); + sc.side_fb[i] = RD::get_singleton()->framebuffer_create(fbtex); + } + + shadow_cubemaps[p_size] = sc; + } + + return &shadow_cubemaps[p_size]; +} + +RID LightStorage::get_cubemap(int p_size) { + ShadowCubemap *cubemap = _get_shadow_cubemap(p_size); + + return cubemap->cubemap; +} + +RID LightStorage::get_cubemap_fb(int p_size, int p_pass) { + ShadowCubemap *cubemap = _get_shadow_cubemap(p_size); + + return cubemap->side_fb[p_pass]; +} diff --git a/servers/rendering/renderer_rd/storage_rd/light_storage.h b/servers/rendering/renderer_rd/storage_rd/light_storage.h new file mode 100644 index 0000000000..79006ad982 --- /dev/null +++ b/servers/rendering/renderer_rd/storage_rd/light_storage.h @@ -0,0 +1,1081 @@ +/*************************************************************************/ +/* light_storage.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef LIGHT_STORAGE_RD_H +#define LIGHT_STORAGE_RD_H + +#include "core/templates/local_vector.h" +#include "core/templates/paged_array.h" +#include "core/templates/rid_owner.h" +#include "core/templates/self_list.h" +#include "servers/rendering/renderer_rd/cluster_builder_rd.h" +#include "servers/rendering/renderer_rd/environment/sky.h" +#include "servers/rendering/renderer_rd/storage_rd/forward_id_storage.h" +#include "servers/rendering/renderer_rd/storage_rd/texture_storage.h" +#include "servers/rendering/storage/light_storage.h" +#include "servers/rendering/storage/utilities.h" + +struct RenderDataRD; + +namespace RendererRD { + +class LightStorage : public RendererLightStorage { +public: + enum ShadowAtlastQuadrant { + QUADRANT_SHIFT = 27, + OMNI_LIGHT_FLAG = 1 << 26, + SHADOW_INDEX_MASK = OMNI_LIGHT_FLAG - 1, + SHADOW_INVALID = 0xFFFFFFFF + }; + +private: + static LightStorage *singleton; + uint32_t max_cluster_elements = 512; + + /* LIGHT */ + struct Light { + RS::LightType type; + float param[RS::LIGHT_PARAM_MAX]; + Color color = Color(1, 1, 1, 1); + RID projector; + bool shadow = false; + bool negative = false; + bool reverse_cull = false; + RS::LightBakeMode bake_mode = RS::LIGHT_BAKE_DYNAMIC; + uint32_t max_sdfgi_cascade = 2; + uint32_t cull_mask = 0xFFFFFFFF; + bool distance_fade = false; + real_t distance_fade_begin = 40.0; + real_t distance_fade_shadow = 50.0; + real_t distance_fade_length = 10.0; + RS::LightOmniShadowMode omni_shadow_mode = RS::LIGHT_OMNI_SHADOW_DUAL_PARABOLOID; + RS::LightDirectionalShadowMode directional_shadow_mode = RS::LIGHT_DIRECTIONAL_SHADOW_ORTHOGONAL; + bool directional_blend_splits = false; + RS::LightDirectionalSkyMode directional_sky_mode = RS::LIGHT_DIRECTIONAL_SKY_MODE_LIGHT_AND_SKY; + uint64_t version = 0; + + Dependency dependency; + }; + + mutable RID_Owner<Light, true> light_owner; + + /* LIGHT INSTANCE */ + + struct LightInstance { + struct ShadowTransform { + Projection camera; + Transform3D transform; + float farplane; + float split; + float bias_scale; + float shadow_texel_size; + float range_begin; + Rect2 atlas_rect; + Vector2 uv_scale; + }; + + RS::LightType light_type = RS::LIGHT_DIRECTIONAL; + + ShadowTransform shadow_transform[6]; + + AABB aabb; + RID self; + RID light; + Transform3D transform; + + Vector3 light_vector; + Vector3 spot_vector; + float linear_att = 0.0; + + uint64_t shadow_pass = 0; + uint64_t last_scene_pass = 0; + uint64_t last_scene_shadow_pass = 0; + uint64_t last_pass = 0; + uint32_t cull_mask = 0; + uint32_t light_directional_index = 0; + + Rect2 directional_rect; + + HashSet<RID> shadow_atlases; //shadow atlases where this light is registered + + ForwardID forward_id = -1; + + LightInstance() {} + }; + + mutable RID_Owner<LightInstance> light_instance_owner; + + /* OMNI/SPOT LIGHT DATA */ + + struct LightData { + float position[3]; + float inv_radius; + float direction[3]; // in omni, x and y are used for dual paraboloid offset + float size; + + float color[3]; + float attenuation; + + float inv_spot_attenuation; + float cos_spot_angle; + float specular_amount; + float shadow_opacity; + + float atlas_rect[4]; // in omni, used for atlas uv, in spot, used for projector uv + float shadow_matrix[16]; + float shadow_bias; + float shadow_normal_bias; + float transmittance_bias; + float soft_shadow_size; + float soft_shadow_scale; + uint32_t mask; + float volumetric_fog_energy; + uint32_t bake_mode; + float projector_rect[4]; + }; + + struct LightInstanceDepthSort { + float depth; + LightInstance *light_instance; + Light *light; + bool operator<(const LightInstanceDepthSort &p_sort) const { + return depth < p_sort.depth; + } + }; + + uint32_t max_lights; + uint32_t omni_light_count = 0; + uint32_t spot_light_count = 0; + LightData *omni_lights = nullptr; + LightData *spot_lights = nullptr; + LightInstanceDepthSort *omni_light_sort = nullptr; + LightInstanceDepthSort *spot_light_sort = nullptr; + RID omni_light_buffer; + RID spot_light_buffer; + + /* DIRECTIONAL LIGHT DATA */ + + struct DirectionalLightData { + float direction[3]; + float energy; + float color[3]; + float size; + float specular; + uint32_t mask; + float softshadow_angle; + float soft_shadow_scale; + uint32_t blend_splits; + float shadow_opacity; + float fade_from; + float fade_to; + uint32_t pad[2]; + uint32_t bake_mode; + float volumetric_fog_energy; + float shadow_bias[4]; + float shadow_normal_bias[4]; + float shadow_transmittance_bias[4]; + float shadow_z_range[4]; + float shadow_range_begin[4]; + float shadow_split_offsets[4]; + float shadow_matrices[4][16]; + float uv_scale1[2]; + float uv_scale2[2]; + float uv_scale3[2]; + float uv_scale4[2]; + }; + + uint32_t max_directional_lights; + DirectionalLightData *directional_lights = nullptr; + RID directional_light_buffer; + + /* REFLECTION PROBE */ + + struct ReflectionProbe { + RS::ReflectionProbeUpdateMode update_mode = RS::REFLECTION_PROBE_UPDATE_ONCE; + int resolution = 256; + float intensity = 1.0; + RS::ReflectionProbeAmbientMode ambient_mode = RS::REFLECTION_PROBE_AMBIENT_ENVIRONMENT; + Color ambient_color; + float ambient_color_energy = 1.0; + float max_distance = 0; + Vector3 extents = Vector3(1, 1, 1); + Vector3 origin_offset; + bool interior = false; + bool box_projection = false; + bool enable_shadows = false; + uint32_t cull_mask = (1 << 20) - 1; + float mesh_lod_threshold = 0.01; + float baked_exposure = 1.0; + + Dependency dependency; + }; + mutable RID_Owner<ReflectionProbe, true> reflection_probe_owner; + + /* REFLECTION ATLAS */ + + struct ReflectionAtlas { + int count = 0; + int size = 0; + + RID reflection; + RID depth_buffer; + RID depth_fb; + + struct Reflection { + RID owner; + RendererRD::SkyRD::ReflectionData data; + RID fbs[6]; + }; + + Vector<Reflection> reflections; + + ClusterBuilderRD *cluster_builder = nullptr; // only used if cluster builder is supported by the renderer. + }; + + mutable RID_Owner<ReflectionAtlas> reflection_atlas_owner; + + /* REFLECTION PROBE INSTANCE */ + + struct ReflectionProbeInstance { + RID probe; + int atlas_index = -1; + RID atlas; + + bool dirty = true; + bool rendering = false; + int processing_layer = 1; + int processing_side = 0; + + uint32_t render_step = 0; + uint64_t last_pass = 0; + uint32_t cull_mask = 0; + + RendererRD::ForwardID forward_id = -1; + + Transform3D transform; + }; + + mutable RID_Owner<ReflectionProbeInstance> reflection_probe_instance_owner; + + /* REFLECTION DATA */ + + enum { + REFLECTION_AMBIENT_DISABLED = 0, + REFLECTION_AMBIENT_ENVIRONMENT = 1, + REFLECTION_AMBIENT_COLOR = 2, + }; + + struct ReflectionData { + float box_extents[3]; + float index; + float box_offset[3]; + uint32_t mask; + float ambient[3]; // ambient color, + float intensity; + uint32_t exterior; + uint32_t box_project; + uint32_t ambient_mode; + float exposure_normalization; + float local_matrix[16]; // up to here for spot and omni, rest is for directional + }; + + struct ReflectionProbeInstanceSort { + float depth; + ReflectionProbeInstance *probe_instance; + bool operator<(const ReflectionProbeInstanceSort &p_sort) const { + return depth < p_sort.depth; + } + }; + + uint32_t max_reflections; + uint32_t reflection_count = 0; + // uint32_t max_reflection_probes_per_instance = 0; // seems unused + ReflectionData *reflections = nullptr; + ReflectionProbeInstanceSort *reflection_sort = nullptr; + RID reflection_buffer; + + /* LIGHTMAP */ + + struct Lightmap { + RID light_texture; + bool uses_spherical_harmonics = false; + bool interior = false; + AABB bounds = AABB(Vector3(), Vector3(1, 1, 1)); + float baked_exposure = 1.0; + int32_t array_index = -1; //unassigned + PackedVector3Array points; + PackedColorArray point_sh; + PackedInt32Array tetrahedra; + PackedInt32Array bsp_tree; + + struct BSP { + static const int32_t EMPTY_LEAF = INT32_MIN; + float plane[4]; + int32_t over = EMPTY_LEAF, under = EMPTY_LEAF; + }; + + Dependency dependency; + }; + + bool using_lightmap_array; + Vector<RID> lightmap_textures; + uint64_t lightmap_array_version = 0; + float lightmap_probe_capture_update_speed = 4; + + mutable RID_Owner<Lightmap, true> lightmap_owner; + + /* LIGHTMAP INSTANCE */ + + struct LightmapInstance { + RID lightmap; + Transform3D transform; + }; + + mutable RID_Owner<LightmapInstance> lightmap_instance_owner; + + /* SHADOW ATLAS */ + + uint64_t shadow_atlas_realloc_tolerance_msec = 500; + + struct ShadowShrinkStage { + RID texture; + RID filter_texture; + uint32_t size = 0; + }; + + struct ShadowAtlas { + struct Quadrant { + uint32_t subdivision = 0; + + struct Shadow { + RID owner; + uint64_t version = 0; + uint64_t fog_version = 0; // used for fog + uint64_t alloc_tick = 0; + + Shadow() {} + }; + + Vector<Shadow> shadows; + + Quadrant() {} + } quadrants[4]; + + int size_order[4] = { 0, 1, 2, 3 }; + uint32_t smallest_subdiv = 0; + + int size = 0; + bool use_16_bits = true; + + RID depth; + RID fb; //for copying + + HashMap<RID, uint32_t> shadow_owners; + }; + + RID_Owner<ShadowAtlas> shadow_atlas_owner; + + void _update_shadow_atlas(ShadowAtlas *shadow_atlas); + + void _shadow_atlas_invalidate_shadow(ShadowAtlas::Quadrant::Shadow *p_shadow, RID p_atlas, ShadowAtlas *p_shadow_atlas, uint32_t p_quadrant, uint32_t p_shadow_idx); + bool _shadow_atlas_find_shadow(ShadowAtlas *shadow_atlas, int *p_in_quadrants, int p_quadrant_count, int p_current_subdiv, uint64_t p_tick, int &r_quadrant, int &r_shadow); + bool _shadow_atlas_find_omni_shadows(ShadowAtlas *shadow_atlas, int *p_in_quadrants, int p_quadrant_count, int p_current_subdiv, uint64_t p_tick, int &r_quadrant, int &r_shadow); + + /* DIRECTIONAL SHADOW */ + + struct DirectionalShadow { + RID depth; + RID fb; //when renderign direct + + int light_count = 0; + int size = 0; + bool use_16_bits = true; + int current_light = 0; + } directional_shadow; + + /* SHADOW CUBEMAPS */ + + struct ShadowCubemap { + RID cubemap; + RID side_fb[6]; + }; + + HashMap<int, ShadowCubemap> shadow_cubemaps; + ShadowCubemap *_get_shadow_cubemap(int p_size); + +public: + static LightStorage *get_singleton(); + + LightStorage(); + virtual ~LightStorage(); + + bool free(RID p_rid); + + /* Settings */ + void set_max_cluster_elements(const uint32_t p_max_cluster_elements) { + max_cluster_elements = p_max_cluster_elements; + set_max_reflection_probes(p_max_cluster_elements); + set_max_lights(p_max_cluster_elements); + } + uint32_t get_max_cluster_elements() const { return max_cluster_elements; } + + /* LIGHT */ + + bool owns_light(RID p_rid) { return light_owner.owns(p_rid); }; + + void _light_initialize(RID p_rid, RS::LightType p_type); + + virtual RID directional_light_allocate() override; + virtual void directional_light_initialize(RID p_light) override; + + virtual RID omni_light_allocate() override; + virtual void omni_light_initialize(RID p_light) override; + + virtual RID spot_light_allocate() override; + virtual void spot_light_initialize(RID p_light) override; + + virtual void light_free(RID p_rid) override; + + virtual void light_set_color(RID p_light, const Color &p_color) override; + virtual void light_set_param(RID p_light, RS::LightParam p_param, float p_value) override; + virtual void light_set_shadow(RID p_light, bool p_enabled) override; + virtual void light_set_projector(RID p_light, RID p_texture) override; + virtual void light_set_negative(RID p_light, bool p_enable) override; + virtual void light_set_cull_mask(RID p_light, uint32_t p_mask) override; + virtual void light_set_distance_fade(RID p_light, bool p_enabled, float p_begin, float p_shadow, float p_length) override; + virtual void light_set_reverse_cull_face_mode(RID p_light, bool p_enabled) override; + virtual void light_set_bake_mode(RID p_light, RS::LightBakeMode p_bake_mode) override; + virtual void light_set_max_sdfgi_cascade(RID p_light, uint32_t p_cascade) override; + + virtual void light_omni_set_shadow_mode(RID p_light, RS::LightOmniShadowMode p_mode) override; + + virtual void light_directional_set_shadow_mode(RID p_light, RS::LightDirectionalShadowMode p_mode) override; + virtual void light_directional_set_blend_splits(RID p_light, bool p_enable) override; + virtual bool light_directional_get_blend_splits(RID p_light) const override; + virtual void light_directional_set_sky_mode(RID p_light, RS::LightDirectionalSkyMode p_mode) override; + virtual RS::LightDirectionalSkyMode light_directional_get_sky_mode(RID p_light) const override; + + virtual RS::LightDirectionalShadowMode light_directional_get_shadow_mode(RID p_light) override; + virtual RS::LightOmniShadowMode light_omni_get_shadow_mode(RID p_light) override; + + virtual RS::LightType light_get_type(RID p_light) const override { + const Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND_V(!light, RS::LIGHT_DIRECTIONAL); + + return light->type; + } + virtual AABB light_get_aabb(RID p_light) const override; + + virtual float light_get_param(RID p_light, RS::LightParam p_param) override { + const Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND_V(!light, 0); + + return light->param[p_param]; + } + + _FORCE_INLINE_ RID light_get_projector(RID p_light) { + const Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND_V(!light, RID()); + + return light->projector; + } + + virtual Color light_get_color(RID p_light) override { + const Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND_V(!light, Color()); + + return light->color; + } + + _FORCE_INLINE_ uint32_t light_get_cull_mask(RID p_light) { + const Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND_V(!light, 0); + + return light->cull_mask; + } + + _FORCE_INLINE_ bool light_is_distance_fade_enabled(RID p_light) { + const Light *light = light_owner.get_or_null(p_light); + return light->distance_fade; + } + + _FORCE_INLINE_ float light_get_distance_fade_begin(RID p_light) { + const Light *light = light_owner.get_or_null(p_light); + return light->distance_fade_begin; + } + + _FORCE_INLINE_ float light_get_distance_fade_shadow(RID p_light) { + const Light *light = light_owner.get_or_null(p_light); + return light->distance_fade_shadow; + } + + _FORCE_INLINE_ float light_get_distance_fade_length(RID p_light) { + const Light *light = light_owner.get_or_null(p_light); + return light->distance_fade_length; + } + + virtual bool light_has_shadow(RID p_light) const override { + const Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND_V(!light, RS::LIGHT_DIRECTIONAL); + + return light->shadow; + } + + virtual bool light_has_projector(RID p_light) const override { + const Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND_V(!light, RS::LIGHT_DIRECTIONAL); + + return TextureStorage::get_singleton()->owns_texture(light->projector); + } + + _FORCE_INLINE_ bool light_is_negative(RID p_light) const { + const Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND_V(!light, RS::LIGHT_DIRECTIONAL); + + return light->negative; + } + + _FORCE_INLINE_ float light_get_transmittance_bias(RID p_light) const { + const Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND_V(!light, 0.0); + + return light->param[RS::LIGHT_PARAM_TRANSMITTANCE_BIAS]; + } + + virtual RS::LightBakeMode light_get_bake_mode(RID p_light) override; + virtual uint32_t light_get_max_sdfgi_cascade(RID p_light) override; + virtual uint64_t light_get_version(RID p_light) const override; + + Dependency *light_get_dependency(RID p_light) const; + + /* LIGHT INSTANCE API */ + + bool owns_light_instance(RID p_rid) { return light_instance_owner.owns(p_rid); }; + + virtual RID light_instance_create(RID p_light) override; + virtual void light_instance_free(RID p_light) override; + virtual void light_instance_set_transform(RID p_light_instance, const Transform3D &p_transform) override; + virtual void light_instance_set_aabb(RID p_light_instance, const AABB &p_aabb) override; + virtual void light_instance_set_shadow_transform(RID p_light_instance, const Projection &p_projection, const Transform3D &p_transform, float p_far, float p_split, int p_pass, float p_shadow_texel_size, float p_bias_scale = 1.0, float p_range_begin = 0, const Vector2 &p_uv_scale = Vector2()) override; + virtual void light_instance_mark_visible(RID p_light_instance) override; + + _FORCE_INLINE_ RID light_instance_get_base_light(RID p_light_instance) { + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + return li->light; + } + + _FORCE_INLINE_ Transform3D light_instance_get_base_transform(RID p_light_instance) { + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + return li->transform; + } + + _FORCE_INLINE_ AABB light_instance_get_base_aabb(RID p_light_instance) { + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + return li->aabb; + } + + _FORCE_INLINE_ void light_instance_set_cull_mask(RID p_light_instance, uint32_t p_cull_mask) { + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + li->cull_mask = p_cull_mask; + } + + _FORCE_INLINE_ uint32_t light_instance_get_cull_mask(RID p_light_instance) { + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + return li->cull_mask; + } + + _FORCE_INLINE_ Rect2 light_instance_get_shadow_atlas_rect(RID p_light_instance, RID p_shadow_atlas, Vector2i &r_omni_offset) { + ShadowAtlas *shadow_atlas = shadow_atlas_owner.get_or_null(p_shadow_atlas); + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + uint32_t key = shadow_atlas->shadow_owners[li->self]; + + uint32_t quadrant = (key >> QUADRANT_SHIFT) & 0x3; + uint32_t shadow = key & SHADOW_INDEX_MASK; + + ERR_FAIL_COND_V(shadow >= (uint32_t)shadow_atlas->quadrants[quadrant].shadows.size(), Rect2()); + + uint32_t atlas_size = shadow_atlas->size; + uint32_t quadrant_size = atlas_size >> 1; + + uint32_t x = (quadrant & 1) * quadrant_size; + uint32_t y = (quadrant >> 1) * quadrant_size; + + uint32_t shadow_size = (quadrant_size / shadow_atlas->quadrants[quadrant].subdivision); + x += (shadow % shadow_atlas->quadrants[quadrant].subdivision) * shadow_size; + y += (shadow / shadow_atlas->quadrants[quadrant].subdivision) * shadow_size; + + if (key & OMNI_LIGHT_FLAG) { + if (((shadow + 1) % shadow_atlas->quadrants[quadrant].subdivision) == 0) { + r_omni_offset.x = 1 - int(shadow_atlas->quadrants[quadrant].subdivision); + r_omni_offset.y = 1; + } else { + r_omni_offset.x = 1; + r_omni_offset.y = 0; + } + } + + uint32_t width = shadow_size; + uint32_t height = shadow_size; + + return Rect2(x / float(shadow_atlas->size), y / float(shadow_atlas->size), width / float(shadow_atlas->size), height / float(shadow_atlas->size)); + } + + _FORCE_INLINE_ float light_instance_get_shadow_texel_size(RID p_light_instance, RID p_shadow_atlas) { +#ifdef DEBUG_ENABLED + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + ERR_FAIL_COND_V(!li->shadow_atlases.has(p_shadow_atlas), 0); +#endif + ShadowAtlas *shadow_atlas = shadow_atlas_owner.get_or_null(p_shadow_atlas); + ERR_FAIL_COND_V(!shadow_atlas, 0); +#ifdef DEBUG_ENABLED + ERR_FAIL_COND_V(!shadow_atlas->shadow_owners.has(p_light_instance), 0); +#endif + uint32_t key = shadow_atlas->shadow_owners[p_light_instance]; + + uint32_t quadrant = (key >> QUADRANT_SHIFT) & 0x3; + + uint32_t quadrant_size = shadow_atlas->size >> 1; + + uint32_t shadow_size = (quadrant_size / shadow_atlas->quadrants[quadrant].subdivision); + + return float(1.0) / shadow_size; + } + + _FORCE_INLINE_ Projection light_instance_get_shadow_camera(RID p_light_instance, int p_index) { + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + return li->shadow_transform[p_index].camera; + } + + _FORCE_INLINE_ Transform3D + light_instance_get_shadow_transform(RID p_light_instance, int p_index) { + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + return li->shadow_transform[p_index].transform; + } + _FORCE_INLINE_ float light_instance_get_shadow_bias_scale(RID p_light_instance, int p_index) { + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + return li->shadow_transform[p_index].bias_scale; + } + _FORCE_INLINE_ float light_instance_get_shadow_range(RID p_light_instance, int p_index) { + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + return li->shadow_transform[p_index].farplane; + } + _FORCE_INLINE_ float light_instance_get_shadow_range_begin(RID p_light_instance, int p_index) { + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + return li->shadow_transform[p_index].range_begin; + } + + _FORCE_INLINE_ Vector2 light_instance_get_shadow_uv_scale(RID p_light_instance, int p_index) { + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + return li->shadow_transform[p_index].uv_scale; + } + + _FORCE_INLINE_ void light_instance_set_directional_shadow_atlas_rect(RID p_light_instance, int p_index, const Rect2 p_atlas_rect) { + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + li->shadow_transform[p_index].atlas_rect = p_atlas_rect; + } + + _FORCE_INLINE_ Rect2 light_instance_get_directional_shadow_atlas_rect(RID p_light_instance, int p_index) { + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + return li->shadow_transform[p_index].atlas_rect; + } + + _FORCE_INLINE_ float light_instance_get_directional_shadow_split(RID p_light_instance, int p_index) { + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + return li->shadow_transform[p_index].split; + } + + _FORCE_INLINE_ float light_instance_get_directional_shadow_texel_size(RID p_light_instance, int p_index) { + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + return li->shadow_transform[p_index].shadow_texel_size; + } + + _FORCE_INLINE_ void light_instance_set_render_pass(RID p_light_instance, uint64_t p_pass) { + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + li->last_pass = p_pass; + } + + _FORCE_INLINE_ uint64_t light_instance_get_render_pass(RID p_light_instance) { + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + return li->last_pass; + } + + _FORCE_INLINE_ void light_instance_set_shadow_pass(RID p_light_instance, uint64_t p_pass) { + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + li->last_scene_shadow_pass = p_pass; + } + + _FORCE_INLINE_ uint64_t light_instance_get_shadow_pass(RID p_light_instance) { + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + return li->last_scene_shadow_pass; + } + + _FORCE_INLINE_ ForwardID light_instance_get_forward_id(RID p_light_instance) { + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + return li->forward_id; + } + + _FORCE_INLINE_ RS::LightType light_instance_get_type(RID p_light_instance) { + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + return li->light_type; + } + + _FORCE_INLINE_ void light_instance_set_directional_rect(RID p_light_instance, const Rect2 &p_directional_rect) { + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + li->directional_rect = p_directional_rect; + } + + _FORCE_INLINE_ Rect2 light_instance_get_directional_rect(RID p_light_instance) { + LightInstance *li = light_instance_owner.get_or_null(p_light_instance); + return li->directional_rect; + } + + /* LIGHT DATA */ + + void free_light_data(); + void set_max_lights(const uint32_t p_max_lights); + RID get_omni_light_buffer() { return omni_light_buffer; } + RID get_spot_light_buffer() { return spot_light_buffer; } + RID get_directional_light_buffer() { return directional_light_buffer; } + uint32_t get_max_directional_lights() { return max_directional_lights; } + bool has_directional_shadows(const uint32_t p_directional_light_count) { + for (uint32_t i = 0; i < p_directional_light_count; i++) { + if (directional_lights[i].shadow_opacity > 0.001) { + return true; + } + } + return false; + } + void update_light_buffers(RenderDataRD *p_render_data, const PagedArray<RID> &p_lights, const Transform3D &p_camera_transform, RID p_shadow_atlas, bool p_using_shadows, uint32_t &r_directional_light_count, uint32_t &r_positional_light_count, bool &r_directional_light_soft_shadows); + + /* REFLECTION PROBE */ + + bool owns_reflection_probe(RID p_rid) { return reflection_probe_owner.owns(p_rid); }; + + virtual RID reflection_probe_allocate() override; + virtual void reflection_probe_initialize(RID p_reflection_probe) override; + virtual void reflection_probe_free(RID p_rid) override; + + virtual void reflection_probe_set_update_mode(RID p_probe, RS::ReflectionProbeUpdateMode p_mode) override; + virtual void reflection_probe_set_intensity(RID p_probe, float p_intensity) override; + virtual void reflection_probe_set_ambient_mode(RID p_probe, RS::ReflectionProbeAmbientMode p_mode) override; + virtual void reflection_probe_set_ambient_color(RID p_probe, const Color &p_color) override; + virtual void reflection_probe_set_ambient_energy(RID p_probe, float p_energy) override; + virtual void reflection_probe_set_max_distance(RID p_probe, float p_distance) override; + virtual void reflection_probe_set_extents(RID p_probe, const Vector3 &p_extents) override; + virtual void reflection_probe_set_origin_offset(RID p_probe, const Vector3 &p_offset) override; + virtual void reflection_probe_set_as_interior(RID p_probe, bool p_enable) override; + virtual void reflection_probe_set_enable_box_projection(RID p_probe, bool p_enable) override; + virtual void reflection_probe_set_enable_shadows(RID p_probe, bool p_enable) override; + virtual void reflection_probe_set_cull_mask(RID p_probe, uint32_t p_layers) override; + virtual void reflection_probe_set_resolution(RID p_probe, int p_resolution) override; + virtual void reflection_probe_set_mesh_lod_threshold(RID p_probe, float p_ratio) override; + + void reflection_probe_set_baked_exposure(RID p_probe, float p_exposure); + + virtual AABB reflection_probe_get_aabb(RID p_probe) const override; + virtual RS::ReflectionProbeUpdateMode reflection_probe_get_update_mode(RID p_probe) const override; + virtual uint32_t reflection_probe_get_cull_mask(RID p_probe) const override; + virtual Vector3 reflection_probe_get_extents(RID p_probe) const override; + virtual Vector3 reflection_probe_get_origin_offset(RID p_probe) const override; + virtual float reflection_probe_get_origin_max_distance(RID p_probe) const override; + virtual float reflection_probe_get_mesh_lod_threshold(RID p_probe) const override; + + int reflection_probe_get_resolution(RID p_probe) const; + float reflection_probe_get_baked_exposure(RID p_probe) const; + virtual bool reflection_probe_renders_shadows(RID p_probe) const override; + + float reflection_probe_get_intensity(RID p_probe) const; + bool reflection_probe_is_interior(RID p_probe) const; + bool reflection_probe_is_box_projection(RID p_probe) const; + RS::ReflectionProbeAmbientMode reflection_probe_get_ambient_mode(RID p_probe) const; + Color reflection_probe_get_ambient_color(RID p_probe) const; + float reflection_probe_get_ambient_color_energy(RID p_probe) const; + + Dependency *reflection_probe_get_dependency(RID p_probe) const; + + /* REFLECTION ATLAS */ + + bool owns_reflection_atlas(RID p_rid) { return reflection_atlas_owner.owns(p_rid); } + + virtual RID reflection_atlas_create() override; + virtual void reflection_atlas_free(RID p_ref_atlas) override; + virtual void reflection_atlas_set_size(RID p_ref_atlas, int p_reflection_size, int p_reflection_count) override; + virtual int reflection_atlas_get_size(RID p_ref_atlas) const override; + + _FORCE_INLINE_ RID reflection_atlas_get_texture(RID p_ref_atlas) { + ReflectionAtlas *atlas = reflection_atlas_owner.get_or_null(p_ref_atlas); + ERR_FAIL_COND_V(!atlas, RID()); + return atlas->reflection; + } + + /* REFLECTION PROBE INSTANCE */ + + bool owns_reflection_probe_instance(RID p_rid) { return reflection_probe_instance_owner.owns(p_rid); } + + virtual RID reflection_probe_instance_create(RID p_probe) override; + virtual void reflection_probe_instance_free(RID p_instance) override; + virtual void reflection_probe_instance_set_transform(RID p_instance, const Transform3D &p_transform) override; + virtual void reflection_probe_release_atlas_index(RID p_instance) override; + virtual bool reflection_probe_instance_needs_redraw(RID p_instance) override; + virtual bool reflection_probe_instance_has_reflection(RID p_instance) override; + virtual bool reflection_probe_instance_begin_render(RID p_instance, RID p_reflection_atlas) override; + virtual bool reflection_probe_instance_postprocess_step(RID p_instance) override; + + uint32_t reflection_probe_instance_get_resolution(RID p_instance); + RID reflection_probe_instance_get_framebuffer(RID p_instance, int p_index); + RID reflection_probe_instance_get_depth_framebuffer(RID p_instance, int p_index); + + _FORCE_INLINE_ RID reflection_probe_instance_get_probe(RID p_instance) { + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.get_or_null(p_instance); + ERR_FAIL_COND_V(!rpi, RID()); + + return rpi->probe; + } + + _FORCE_INLINE_ RendererRD::ForwardID reflection_probe_instance_get_forward_id(RID p_instance) { + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.get_or_null(p_instance); + ERR_FAIL_COND_V(!rpi, 0); + + return rpi->forward_id; + } + + _FORCE_INLINE_ void reflection_probe_instance_set_cull_mask(RID p_instance, uint32_t p_render_pass) { + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.get_or_null(p_instance); + ERR_FAIL_COND(!rpi); + rpi->cull_mask = p_render_pass; + } + + _FORCE_INLINE_ void reflection_probe_instance_set_render_pass(RID p_instance, uint32_t p_render_pass) { + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.get_or_null(p_instance); + ERR_FAIL_COND(!rpi); + rpi->last_pass = p_render_pass; + } + + _FORCE_INLINE_ uint32_t reflection_probe_instance_get_render_pass(RID p_instance) { + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.get_or_null(p_instance); + ERR_FAIL_COND_V(!rpi, 0); + + return rpi->last_pass; + } + + _FORCE_INLINE_ Transform3D reflection_probe_instance_get_transform(RID p_instance) { + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.get_or_null(p_instance); + ERR_FAIL_COND_V(!rpi, Transform3D()); + + return rpi->transform; + } + + _FORCE_INLINE_ int reflection_probe_instance_get_atlas_index(RID p_instance) { + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.get_or_null(p_instance); + ERR_FAIL_COND_V(!rpi, -1); + + return rpi->atlas_index; + } + + ClusterBuilderRD *reflection_probe_instance_get_cluster_builder(RID p_instance, ClusterBuilderSharedDataRD *p_cluster_builder_shared); + + /* REFLECTION DATA */ + + void free_reflection_data(); + void set_max_reflection_probes(const uint32_t p_max_reflection_probes); + RID get_reflection_probe_buffer() { return reflection_buffer; } + void update_reflection_probe_buffer(RenderDataRD *p_render_data, const PagedArray<RID> &p_reflections, const Transform3D &p_camera_inverse_transform, RID p_environment); + + /* LIGHTMAP */ + + bool owns_lightmap(RID p_rid) { return lightmap_owner.owns(p_rid); }; + + virtual RID lightmap_allocate() override; + virtual void lightmap_initialize(RID p_lightmap) override; + virtual void lightmap_free(RID p_rid) override; + + virtual void lightmap_set_textures(RID p_lightmap, RID p_light, bool p_uses_spherical_haromics) override; + virtual void lightmap_set_probe_bounds(RID p_lightmap, const AABB &p_bounds) override; + virtual void lightmap_set_probe_interior(RID p_lightmap, bool p_interior) override; + virtual void lightmap_set_probe_capture_data(RID p_lightmap, const PackedVector3Array &p_points, const PackedColorArray &p_point_sh, const PackedInt32Array &p_tetrahedra, const PackedInt32Array &p_bsp_tree) override; + virtual void lightmap_set_baked_exposure_normalization(RID p_lightmap, float p_exposure) override; + virtual PackedVector3Array lightmap_get_probe_capture_points(RID p_lightmap) const override; + virtual PackedColorArray lightmap_get_probe_capture_sh(RID p_lightmap) const override; + virtual PackedInt32Array lightmap_get_probe_capture_tetrahedra(RID p_lightmap) const override; + virtual PackedInt32Array lightmap_get_probe_capture_bsp_tree(RID p_lightmap) const override; + virtual AABB lightmap_get_aabb(RID p_lightmap) const override; + virtual bool lightmap_is_interior(RID p_lightmap) const override; + virtual void lightmap_tap_sh_light(RID p_lightmap, const Vector3 &p_point, Color *r_sh) override; + virtual void lightmap_set_probe_capture_update_speed(float p_speed) override; + + Dependency *lightmap_get_dependency(RID p_lightmap) const; + + virtual float lightmap_get_probe_capture_update_speed() const override { + return lightmap_probe_capture_update_speed; + } + _FORCE_INLINE_ RID lightmap_get_texture(RID p_lightmap) const { + const Lightmap *lm = lightmap_owner.get_or_null(p_lightmap); + ERR_FAIL_COND_V(!lm, RID()); + return lm->light_texture; + } + _FORCE_INLINE_ float lightmap_get_baked_exposure_normalization(RID p_lightmap) const { + const Lightmap *lm = lightmap_owner.get_or_null(p_lightmap); + ERR_FAIL_COND_V(!lm, 1.0); + return lm->baked_exposure; + } + _FORCE_INLINE_ int32_t lightmap_get_array_index(RID p_lightmap) const { + ERR_FAIL_COND_V(!using_lightmap_array, -1); //only for arrays + const Lightmap *lm = lightmap_owner.get_or_null(p_lightmap); + return lm->array_index; + } + _FORCE_INLINE_ bool lightmap_uses_spherical_harmonics(RID p_lightmap) const { + ERR_FAIL_COND_V(!using_lightmap_array, false); //only for arrays + const Lightmap *lm = lightmap_owner.get_or_null(p_lightmap); + return lm->uses_spherical_harmonics; + } + _FORCE_INLINE_ uint64_t lightmap_array_get_version() const { + ERR_FAIL_COND_V(!using_lightmap_array, 0); //only for arrays + return lightmap_array_version; + } + + _FORCE_INLINE_ int lightmap_array_get_size() const { + ERR_FAIL_COND_V(!using_lightmap_array, 0); //only for arrays + return lightmap_textures.size(); + } + + _FORCE_INLINE_ const Vector<RID> &lightmap_array_get_textures() const { + ERR_FAIL_COND_V(!using_lightmap_array, lightmap_textures); //only for arrays + return lightmap_textures; + } + + /* LIGHTMAP INSTANCE */ + + bool owns_lightmap_instance(RID p_rid) { return lightmap_instance_owner.owns(p_rid); }; + + virtual RID lightmap_instance_create(RID p_lightmap) override; + virtual void lightmap_instance_free(RID p_lightmap) override; + virtual void lightmap_instance_set_transform(RID p_lightmap, const Transform3D &p_transform) override; + _FORCE_INLINE_ bool lightmap_instance_is_valid(RID p_lightmap_instance) { + return lightmap_instance_owner.get_or_null(p_lightmap_instance) != nullptr; + } + + _FORCE_INLINE_ RID lightmap_instance_get_lightmap(RID p_lightmap_instance) { + LightmapInstance *li = lightmap_instance_owner.get_or_null(p_lightmap_instance); + return li->lightmap; + } + _FORCE_INLINE_ Transform3D lightmap_instance_get_transform(RID p_lightmap_instance) { + LightmapInstance *li = lightmap_instance_owner.get_or_null(p_lightmap_instance); + return li->transform; + } + + /* SHADOW ATLAS API */ + + bool owns_shadow_atlas(RID p_rid) { return shadow_atlas_owner.owns(p_rid); }; + + virtual RID shadow_atlas_create() override; + virtual void shadow_atlas_free(RID p_atlas) override; + + virtual void shadow_atlas_set_size(RID p_atlas, int p_size, bool p_16_bits = true) override; + virtual void shadow_atlas_set_quadrant_subdivision(RID p_atlas, int p_quadrant, int p_subdivision) override; + virtual bool shadow_atlas_update_light(RID p_atlas, RID p_light_instance, float p_coverage, uint64_t p_light_version) override; + _FORCE_INLINE_ bool shadow_atlas_owns_light_instance(RID p_atlas, RID p_light_intance) { + ShadowAtlas *atlas = shadow_atlas_owner.get_or_null(p_atlas); + ERR_FAIL_COND_V(!atlas, false); + return atlas->shadow_owners.has(p_light_intance); + } + _FORCE_INLINE_ uint32_t shadow_atlas_get_light_instance_key(RID p_atlas, RID p_light_intance) { + ShadowAtlas *atlas = shadow_atlas_owner.get_or_null(p_atlas); + ERR_FAIL_COND_V(!atlas, -1); + return atlas->shadow_owners[p_light_intance]; + } + + _FORCE_INLINE_ RID shadow_atlas_get_texture(RID p_atlas) { + ShadowAtlas *atlas = shadow_atlas_owner.get_or_null(p_atlas); + ERR_FAIL_COND_V(!atlas, RID()); + return atlas->depth; + } + + _FORCE_INLINE_ int shadow_atlas_get_size(RID p_atlas) { + ShadowAtlas *atlas = shadow_atlas_owner.get_or_null(p_atlas); + ERR_FAIL_COND_V(!atlas, 0); + return atlas->size; + } + + _FORCE_INLINE_ int shadow_atlas_get_quadrant_shadow_size(RID p_atlas, uint32_t p_quadrant) { + ShadowAtlas *atlas = shadow_atlas_owner.get_or_null(p_atlas); + ERR_FAIL_COND_V(!atlas, 0); + ERR_FAIL_UNSIGNED_INDEX_V(p_quadrant, 4, 0); + return atlas->quadrants[p_quadrant].shadows.size(); + } + + _FORCE_INLINE_ uint32_t shadow_atlas_get_quadrant_subdivision(RID p_atlas, uint32_t p_quadrant) { + ShadowAtlas *atlas = shadow_atlas_owner.get_or_null(p_atlas); + ERR_FAIL_COND_V(!atlas, 0); + ERR_FAIL_UNSIGNED_INDEX_V(p_quadrant, 4, 0); + return atlas->quadrants[p_quadrant].subdivision; + } + + _FORCE_INLINE_ RID shadow_atlas_get_fb(RID p_atlas) { + ShadowAtlas *atlas = shadow_atlas_owner.get_or_null(p_atlas); + ERR_FAIL_COND_V(!atlas, RID()); + return atlas->fb; + } + + virtual void shadow_atlas_update(RID p_atlas) override; + + /* DIRECTIONAL SHADOW */ + + virtual void directional_shadow_atlas_set_size(int p_size, bool p_16_bits = true) override; + virtual int get_directional_light_shadow_size(RID p_light_intance) override; + virtual void set_directional_shadow_count(int p_count) override; + + Rect2i get_directional_shadow_rect(); + void update_directional_shadow_atlas(); + + _FORCE_INLINE_ RID directional_shadow_get_texture() { + return directional_shadow.depth; + } + + _FORCE_INLINE_ int directional_shadow_get_size() { + return directional_shadow.size; + } + + _FORCE_INLINE_ RID direction_shadow_get_fb() { + return directional_shadow.fb; + } + + _FORCE_INLINE_ void directional_shadow_increase_current_light() { + directional_shadow.current_light++; + } + + /* SHADOW CUBEMAPS */ + + RID get_cubemap(int p_size); + RID get_cubemap_fb(int p_size, int p_pass); +}; + +} // namespace RendererRD + +#endif // LIGHT_STORAGE_RD_H diff --git a/servers/rendering/renderer_rd/storage_rd/material_storage.cpp b/servers/rendering/renderer_rd/storage_rd/material_storage.cpp new file mode 100644 index 0000000000..6703d8e7d0 --- /dev/null +++ b/servers/rendering/renderer_rd/storage_rd/material_storage.cpp @@ -0,0 +1,2746 @@ +/*************************************************************************/ +/* material_storage.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "material_storage.h" +#include "core/config/engine.h" +#include "core/config/project_settings.h" +#include "core/io/resource_loader.h" +#include "texture_storage.h" + +using namespace RendererRD; + +/////////////////////////////////////////////////////////////////////////// +// UBI helper functions + +_FORCE_INLINE_ static void _fill_std140_variant_ubo_value(ShaderLanguage::DataType type, int p_array_size, const Variant &value, uint8_t *data, bool p_linear_color) { + switch (type) { + case ShaderLanguage::TYPE_BOOL: { + uint32_t *gui = (uint32_t *)data; + + if (p_array_size > 0) { + const PackedInt32Array &ba = value; + int s = ba.size(); + const int *r = ba.ptr(); + + for (int i = 0, j = 0; i < p_array_size; i++, j += 4) { + if (i < s) { + gui[j] = (r[i] != 0) ? 1 : 0; + } else { + gui[j] = 0; + } + gui[j + 1] = 0; // ignored + gui[j + 2] = 0; // ignored + gui[j + 3] = 0; // ignored + } + } else { + bool v = value; + gui[0] = v ? 1 : 0; + } + } break; + case ShaderLanguage::TYPE_BVEC2: { + uint32_t *gui = (uint32_t *)data; + + if (p_array_size > 0) { + const PackedInt32Array &ba = value; + int s = ba.size(); + const int *r = ba.ptr(); + int count = 2 * p_array_size; + + for (int i = 0, j = 0; i < count; i += 2, j += 4) { + if (i < s) { + gui[j] = r[i] ? 1 : 0; + gui[j + 1] = r[i + 1] ? 1 : 0; + } else { + gui[j] = 0; + gui[j + 1] = 0; + } + gui[j + 2] = 0; // ignored + gui[j + 3] = 0; // ignored + } + } else { + uint32_t v = value; + gui[0] = v & 1 ? 1 : 0; + gui[1] = v & 2 ? 1 : 0; + } + } break; + case ShaderLanguage::TYPE_BVEC3: { + uint32_t *gui = (uint32_t *)data; + + if (p_array_size > 0) { + const PackedInt32Array &ba = value; + int s = ba.size(); + const int *r = ba.ptr(); + int count = 3 * p_array_size; + + for (int i = 0, j = 0; i < count; i += 3, j += 4) { + if (i < s) { + gui[j] = r[i] ? 1 : 0; + gui[j + 1] = r[i + 1] ? 1 : 0; + gui[j + 2] = r[i + 2] ? 1 : 0; + } else { + gui[j] = 0; + gui[j + 1] = 0; + gui[j + 2] = 0; + } + gui[j + 3] = 0; // ignored + } + } else { + uint32_t v = value; + gui[0] = (v & 1) ? 1 : 0; + gui[1] = (v & 2) ? 1 : 0; + gui[2] = (v & 4) ? 1 : 0; + } + } break; + case ShaderLanguage::TYPE_BVEC4: { + uint32_t *gui = (uint32_t *)data; + + if (p_array_size > 0) { + const PackedInt32Array &ba = value; + int s = ba.size(); + const int *r = ba.ptr(); + int count = 4 * p_array_size; + + for (int i = 0; i < count; i += 4) { + if (i < s) { + gui[i] = r[i] ? 1 : 0; + gui[i + 1] = r[i + 1] ? 1 : 0; + gui[i + 2] = r[i + 2] ? 1 : 0; + gui[i + 3] = r[i + 3] ? 1 : 0; + } else { + gui[i] = 0; + gui[i + 1] = 0; + gui[i + 2] = 0; + gui[i + 3] = 0; + } + } + } else { + uint32_t v = value; + gui[0] = (v & 1) ? 1 : 0; + gui[1] = (v & 2) ? 1 : 0; + gui[2] = (v & 4) ? 1 : 0; + gui[3] = (v & 8) ? 1 : 0; + } + } break; + case ShaderLanguage::TYPE_INT: { + int32_t *gui = (int32_t *)data; + + if (p_array_size > 0) { + Vector<int> iv = value; + int s = iv.size(); + const int *r = iv.ptr(); + + for (int i = 0, j = 0; i < p_array_size; i++, j += 4) { + if (i < s) { + gui[j] = r[i]; + } else { + gui[j] = 0; + } + gui[j + 1] = 0; // ignored + gui[j + 2] = 0; // ignored + gui[j + 3] = 0; // ignored + } + } else { + int v = value; + gui[0] = v; + } + } break; + case ShaderLanguage::TYPE_IVEC2: { + int32_t *gui = (int32_t *)data; + if (p_array_size > 0) { + Vector<int> iv = value; + int s = iv.size(); + int count = 2 * p_array_size; + + const int *r = iv.ptr(); + for (int i = 0, j = 0; i < count; i += 2, j += 4) { + if (i < s) { + gui[j] = r[i]; + gui[j + 1] = r[i + 1]; + } else { + gui[j] = 0; + gui[j + 1] = 0; + } + gui[j + 2] = 0; // ignored + gui[j + 3] = 0; // ignored + } + } else { + Vector2i v = value; + gui[0] = v.x; + gui[1] = v.y; + } + } break; + case ShaderLanguage::TYPE_IVEC3: { + int32_t *gui = (int32_t *)data; + if (p_array_size > 0) { + Vector<int> iv = value; + int s = iv.size(); + int count = 3 * p_array_size; + + const int *r = iv.ptr(); + for (int i = 0, j = 0; i < count; i += 3, j += 4) { + if (i < s) { + gui[j] = r[i]; + gui[j + 1] = r[i + 1]; + gui[j + 2] = r[i + 2]; + } else { + gui[j] = 0; + gui[j + 1] = 0; + gui[j + 2] = 0; + } + gui[j + 3] = 0; // ignored + } + } else { + Vector3i v = value; + gui[0] = v.x; + gui[1] = v.y; + gui[2] = v.z; + } + } break; + case ShaderLanguage::TYPE_IVEC4: { + int32_t *gui = (int32_t *)data; + if (p_array_size > 0) { + Vector<int> iv = value; + int s = iv.size(); + int count = 4 * p_array_size; + + const int *r = iv.ptr(); + for (int i = 0; i < count; i += 4) { + if (i < s) { + gui[i] = r[i]; + gui[i + 1] = r[i + 1]; + gui[i + 2] = r[i + 2]; + gui[i + 3] = r[i + 3]; + } else { + gui[i] = 0; + gui[i + 1] = 0; + gui[i + 2] = 0; + gui[i + 3] = 0; + } + } + } else { + Vector4i v = value; + gui[0] = v.x; + gui[1] = v.y; + gui[2] = v.z; + gui[3] = v.w; + } + } break; + case ShaderLanguage::TYPE_UINT: { + uint32_t *gui = (uint32_t *)data; + + if (p_array_size > 0) { + Vector<int> iv = value; + int s = iv.size(); + const int *r = iv.ptr(); + + for (int i = 0, j = 0; i < p_array_size; i++, j += 4) { + if (i < s) { + gui[j] = r[i]; + } else { + gui[j] = 0; + } + gui[j + 1] = 0; // ignored + gui[j + 2] = 0; // ignored + gui[j + 3] = 0; // ignored + } + } else { + int v = value; + gui[0] = v; + } + } break; + case ShaderLanguage::TYPE_UVEC2: { + uint32_t *gui = (uint32_t *)data; + if (p_array_size > 0) { + Vector<int> iv = value; + int s = iv.size(); + int count = 2 * p_array_size; + + const int *r = iv.ptr(); + for (int i = 0, j = 0; i < count; i += 2, j += 4) { + if (i < s) { + gui[j] = r[i]; + gui[j + 1] = r[i + 1]; + } else { + gui[j] = 0; + gui[j + 1] = 0; + } + gui[j + 2] = 0; // ignored + gui[j + 3] = 0; // ignored + } + } else { + Vector2i v = value; + gui[0] = v.x; + gui[1] = v.y; + } + } break; + case ShaderLanguage::TYPE_UVEC3: { + uint32_t *gui = (uint32_t *)data; + if (p_array_size > 0) { + Vector<int> iv = value; + int s = iv.size(); + int count = 3 * p_array_size; + + const int *r = iv.ptr(); + for (int i = 0, j = 0; i < count; i += 3, j += 4) { + if (i < s) { + gui[j] = r[i]; + gui[j + 1] = r[i + 1]; + gui[j + 2] = r[i + 2]; + } else { + gui[j] = 0; + gui[j + 1] = 0; + gui[j + 2] = 0; + } + gui[j + 3] = 0; // ignored + } + } else { + Vector3i v = value; + gui[0] = v.x; + gui[1] = v.y; + gui[2] = v.z; + } + } break; + case ShaderLanguage::TYPE_UVEC4: { + uint32_t *gui = (uint32_t *)data; + if (p_array_size > 0) { + Vector<int> iv = value; + int s = iv.size(); + int count = 4 * p_array_size; + + const int *r = iv.ptr(); + for (int i = 0; i < count; i++) { + if (i < s) { + gui[i] = r[i]; + gui[i + 1] = r[i + 1]; + gui[i + 2] = r[i + 2]; + gui[i + 3] = r[i + 3]; + } else { + gui[i] = 0; + gui[i + 1] = 0; + gui[i + 2] = 0; + gui[i + 3] = 0; + } + } + } else { + Vector4i v = value; + gui[0] = v.x; + gui[1] = v.y; + gui[2] = v.z; + gui[3] = v.w; + } + } break; + case ShaderLanguage::TYPE_FLOAT: { + float *gui = reinterpret_cast<float *>(data); + + if (p_array_size > 0) { + const PackedFloat32Array &a = value; + int s = a.size(); + + for (int i = 0, j = 0; i < p_array_size; i++, j += 4) { + if (i < s) { + gui[j] = a[i]; + } else { + gui[j] = 0; + } + gui[j + 1] = 0; // ignored + gui[j + 2] = 0; // ignored + gui[j + 3] = 0; // ignored + } + } else { + float v = value; + gui[0] = v; + } + } break; + case ShaderLanguage::TYPE_VEC2: { + float *gui = reinterpret_cast<float *>(data); + + if (p_array_size > 0) { + const PackedVector2Array &a = value; + int s = a.size(); + + for (int i = 0, j = 0; i < p_array_size; i++, j += 4) { + if (i < s) { + gui[j] = a[i].x; + gui[j + 1] = a[i].y; + } else { + gui[j] = 0; + gui[j + 1] = 0; + } + gui[j + 2] = 0; // ignored + gui[j + 3] = 0; // ignored + } + } else { + Vector2 v = value; + gui[0] = v.x; + gui[1] = v.y; + } + } break; + case ShaderLanguage::TYPE_VEC3: { + float *gui = reinterpret_cast<float *>(data); + + if (p_array_size > 0) { + if (value.get_type() == Variant::PACKED_COLOR_ARRAY) { + const PackedColorArray &a = value; + int s = a.size(); + + for (int i = 0, j = 0; i < p_array_size; i++, j += 4) { + if (i < s) { + Color color = a[i]; + if (p_linear_color) { + color = color.srgb_to_linear(); + } + gui[j] = color.r; + gui[j + 1] = color.g; + gui[j + 2] = color.b; + } else { + gui[j] = 0; + gui[j + 1] = 0; + gui[j + 2] = 0; + } + gui[j + 3] = 0; // ignored + } + } else { + const PackedVector3Array &a = value; + int s = a.size(); + + for (int i = 0, j = 0; i < p_array_size; i++, j += 4) { + if (i < s) { + gui[j] = a[i].x; + gui[j + 1] = a[i].y; + gui[j + 2] = a[i].z; + } else { + gui[j] = 0; + gui[j + 1] = 0; + gui[j + 2] = 0; + } + gui[j + 3] = 0; // ignored + } + } + } else { + if (value.get_type() == Variant::COLOR) { + Color v = value; + + if (p_linear_color) { + v = v.srgb_to_linear(); + } + + gui[0] = v.r; + gui[1] = v.g; + gui[2] = v.b; + } else { + Vector3 v = value; + gui[0] = v.x; + gui[1] = v.y; + gui[2] = v.z; + } + } + } break; + case ShaderLanguage::TYPE_VEC4: { + float *gui = reinterpret_cast<float *>(data); + + if (p_array_size > 0) { + if (value.get_type() == Variant::PACKED_COLOR_ARRAY) { + const PackedColorArray &a = value; + int s = a.size(); + + for (int i = 0, j = 0; i < p_array_size; i++, j += 4) { + if (i < s) { + Color color = a[i]; + if (p_linear_color) { + color = color.srgb_to_linear(); + } + gui[j] = color.r; + gui[j + 1] = color.g; + gui[j + 2] = color.b; + gui[j + 3] = color.a; + } else { + gui[j] = 0; + gui[j + 1] = 0; + gui[j + 2] = 0; + gui[j + 3] = 0; + } + } + } else { + const PackedFloat32Array &a = value; + int s = a.size(); + int count = 4 * p_array_size; + + for (int i = 0; i < count; i += 4) { + if (i + 3 < s) { + gui[i] = a[i]; + gui[i + 1] = a[i + 1]; + gui[i + 2] = a[i + 2]; + gui[i + 3] = a[i + 3]; + } else { + gui[i] = 0; + gui[i + 1] = 0; + gui[i + 2] = 0; + gui[i + 3] = 0; + } + } + } + } else { + if (value.get_type() == Variant::COLOR) { + Color v = value; + + if (p_linear_color) { + v = v.srgb_to_linear(); + } + + gui[0] = v.r; + gui[1] = v.g; + gui[2] = v.b; + gui[3] = v.a; + } else if (value.get_type() == Variant::RECT2) { + Rect2 v = value; + + gui[0] = v.position.x; + gui[1] = v.position.y; + gui[2] = v.size.x; + gui[3] = v.size.y; + } else if (value.get_type() == Variant::QUATERNION) { + Quaternion v = value; + + gui[0] = v.x; + gui[1] = v.y; + gui[2] = v.z; + gui[3] = v.w; + } else if (value.get_type() == Variant::PLANE) { + Plane v = value; + + gui[0] = v.normal.x; + gui[1] = v.normal.y; + gui[2] = v.normal.z; + gui[3] = v.d; + } else { + Vector4 v = value; + + gui[0] = v.x; + gui[1] = v.y; + gui[2] = v.z; + gui[3] = v.w; + } + } + } break; + case ShaderLanguage::TYPE_MAT2: { + float *gui = reinterpret_cast<float *>(data); + + if (p_array_size > 0) { + const PackedFloat32Array &a = value; + int s = a.size(); + + for (int i = 0, j = 0; i < p_array_size * 4; i += 4, j += 8) { + if (i + 3 < s) { + gui[j] = a[i]; + gui[j + 1] = a[i + 1]; + + gui[j + 4] = a[i + 2]; + gui[j + 5] = a[i + 3]; + } else { + gui[j] = 1; + gui[j + 1] = 0; + + gui[j + 4] = 0; + gui[j + 5] = 1; + } + gui[j + 2] = 0; // ignored + gui[j + 3] = 0; // ignored + gui[j + 6] = 0; // ignored + gui[j + 7] = 0; // ignored + } + } else { + Transform2D v = value; + + //in std140 members of mat2 are treated as vec4s + gui[0] = v.columns[0][0]; + gui[1] = v.columns[0][1]; + gui[2] = 0; // ignored + gui[3] = 0; // ignored + + gui[4] = v.columns[1][0]; + gui[5] = v.columns[1][1]; + gui[6] = 0; // ignored + gui[7] = 0; // ignored + } + } break; + case ShaderLanguage::TYPE_MAT3: { + float *gui = reinterpret_cast<float *>(data); + + if (p_array_size > 0) { + const PackedFloat32Array &a = value; + int s = a.size(); + + for (int i = 0, j = 0; i < p_array_size * 9; i += 9, j += 12) { + if (i + 8 < s) { + gui[j] = a[i]; + gui[j + 1] = a[i + 1]; + gui[j + 2] = a[i + 2]; + + gui[j + 4] = a[i + 3]; + gui[j + 5] = a[i + 4]; + gui[j + 6] = a[i + 5]; + + gui[j + 8] = a[i + 6]; + gui[j + 9] = a[i + 7]; + gui[j + 10] = a[i + 8]; + } else { + gui[j] = 1; + gui[j + 1] = 0; + gui[j + 2] = 0; + + gui[j + 4] = 0; + gui[j + 5] = 1; + gui[j + 6] = 0; + + gui[j + 8] = 0; + gui[j + 9] = 0; + gui[j + 10] = 1; + } + gui[j + 3] = 0; // ignored + gui[j + 7] = 0; // ignored + gui[j + 11] = 0; // ignored + } + } else { + Basis v = value; + gui[0] = v.rows[0][0]; + gui[1] = v.rows[1][0]; + gui[2] = v.rows[2][0]; + gui[3] = 0; // ignored + + gui[4] = v.rows[0][1]; + gui[5] = v.rows[1][1]; + gui[6] = v.rows[2][1]; + gui[7] = 0; // ignored + + gui[8] = v.rows[0][2]; + gui[9] = v.rows[1][2]; + gui[10] = v.rows[2][2]; + gui[11] = 0; // ignored + } + } break; + case ShaderLanguage::TYPE_MAT4: { + float *gui = reinterpret_cast<float *>(data); + + if (p_array_size > 0) { + const PackedFloat32Array &a = value; + int s = a.size(); + + for (int i = 0; i < p_array_size * 16; i += 16) { + if (i + 15 < s) { + gui[i] = a[i]; + gui[i + 1] = a[i + 1]; + gui[i + 2] = a[i + 2]; + gui[i + 3] = a[i + 3]; + + gui[i + 4] = a[i + 4]; + gui[i + 5] = a[i + 5]; + gui[i + 6] = a[i + 6]; + gui[i + 7] = a[i + 7]; + + gui[i + 8] = a[i + 8]; + gui[i + 9] = a[i + 9]; + gui[i + 10] = a[i + 10]; + gui[i + 11] = a[i + 11]; + + gui[i + 12] = a[i + 12]; + gui[i + 13] = a[i + 13]; + gui[i + 14] = a[i + 14]; + gui[i + 15] = a[i + 15]; + } else { + gui[i] = 1; + gui[i + 1] = 0; + gui[i + 2] = 0; + gui[i + 3] = 0; + + gui[i + 4] = 0; + gui[i + 5] = 1; + gui[i + 6] = 0; + gui[i + 7] = 0; + + gui[i + 8] = 0; + gui[i + 9] = 0; + gui[i + 10] = 1; + gui[i + 11] = 0; + + gui[i + 12] = 0; + gui[i + 13] = 0; + gui[i + 14] = 0; + gui[i + 15] = 1; + } + } + } else if (value.get_type() == Variant::TRANSFORM3D) { + Transform3D v = value; + gui[0] = v.basis.rows[0][0]; + gui[1] = v.basis.rows[1][0]; + gui[2] = v.basis.rows[2][0]; + gui[3] = 0; + + gui[4] = v.basis.rows[0][1]; + gui[5] = v.basis.rows[1][1]; + gui[6] = v.basis.rows[2][1]; + gui[7] = 0; + + gui[8] = v.basis.rows[0][2]; + gui[9] = v.basis.rows[1][2]; + gui[10] = v.basis.rows[2][2]; + gui[11] = 0; + + gui[12] = v.origin.x; + gui[13] = v.origin.y; + gui[14] = v.origin.z; + gui[15] = 1; + } else { + Projection v = value; + for (int i = 0; i < 4; i++) { + for (int j = 0; j < 4; j++) { + gui[i * 4 + j] = v.columns[i][j]; + } + } + } + } break; + default: { + } + } +} + +_FORCE_INLINE_ static void _fill_std140_ubo_value(ShaderLanguage::DataType type, const Vector<ShaderLanguage::ConstantNode::Value> &value, uint8_t *data) { + switch (type) { + case ShaderLanguage::TYPE_BOOL: { + uint32_t *gui = (uint32_t *)data; + gui[0] = value[0].boolean ? 1 : 0; + } break; + case ShaderLanguage::TYPE_BVEC2: { + uint32_t *gui = (uint32_t *)data; + gui[0] = value[0].boolean ? 1 : 0; + gui[1] = value[1].boolean ? 1 : 0; + + } break; + case ShaderLanguage::TYPE_BVEC3: { + uint32_t *gui = (uint32_t *)data; + gui[0] = value[0].boolean ? 1 : 0; + gui[1] = value[1].boolean ? 1 : 0; + gui[2] = value[2].boolean ? 1 : 0; + + } break; + case ShaderLanguage::TYPE_BVEC4: { + uint32_t *gui = (uint32_t *)data; + gui[0] = value[0].boolean ? 1 : 0; + gui[1] = value[1].boolean ? 1 : 0; + gui[2] = value[2].boolean ? 1 : 0; + gui[3] = value[3].boolean ? 1 : 0; + + } break; + case ShaderLanguage::TYPE_INT: { + int32_t *gui = (int32_t *)data; + gui[0] = value[0].sint; + + } break; + case ShaderLanguage::TYPE_IVEC2: { + int32_t *gui = (int32_t *)data; + + for (int i = 0; i < 2; i++) { + gui[i] = value[i].sint; + } + + } break; + case ShaderLanguage::TYPE_IVEC3: { + int32_t *gui = (int32_t *)data; + + for (int i = 0; i < 3; i++) { + gui[i] = value[i].sint; + } + + } break; + case ShaderLanguage::TYPE_IVEC4: { + int32_t *gui = (int32_t *)data; + + for (int i = 0; i < 4; i++) { + gui[i] = value[i].sint; + } + + } break; + case ShaderLanguage::TYPE_UINT: { + uint32_t *gui = (uint32_t *)data; + gui[0] = value[0].uint; + + } break; + case ShaderLanguage::TYPE_UVEC2: { + int32_t *gui = (int32_t *)data; + + for (int i = 0; i < 2; i++) { + gui[i] = value[i].uint; + } + } break; + case ShaderLanguage::TYPE_UVEC3: { + int32_t *gui = (int32_t *)data; + + for (int i = 0; i < 3; i++) { + gui[i] = value[i].uint; + } + + } break; + case ShaderLanguage::TYPE_UVEC4: { + int32_t *gui = (int32_t *)data; + + for (int i = 0; i < 4; i++) { + gui[i] = value[i].uint; + } + } break; + case ShaderLanguage::TYPE_FLOAT: { + float *gui = reinterpret_cast<float *>(data); + gui[0] = value[0].real; + + } break; + case ShaderLanguage::TYPE_VEC2: { + float *gui = reinterpret_cast<float *>(data); + + for (int i = 0; i < 2; i++) { + gui[i] = value[i].real; + } + + } break; + case ShaderLanguage::TYPE_VEC3: { + float *gui = reinterpret_cast<float *>(data); + + for (int i = 0; i < 3; i++) { + gui[i] = value[i].real; + } + + } break; + case ShaderLanguage::TYPE_VEC4: { + float *gui = reinterpret_cast<float *>(data); + + for (int i = 0; i < 4; i++) { + gui[i] = value[i].real; + } + } break; + case ShaderLanguage::TYPE_MAT2: { + float *gui = reinterpret_cast<float *>(data); + + //in std140 members of mat2 are treated as vec4s + gui[0] = value[0].real; + gui[1] = value[1].real; + gui[2] = 0; + gui[3] = 0; + gui[4] = value[2].real; + gui[5] = value[3].real; + gui[6] = 0; + gui[7] = 0; + } break; + case ShaderLanguage::TYPE_MAT3: { + float *gui = reinterpret_cast<float *>(data); + + gui[0] = value[0].real; + gui[1] = value[1].real; + gui[2] = value[2].real; + gui[3] = 0; + gui[4] = value[3].real; + gui[5] = value[4].real; + gui[6] = value[5].real; + gui[7] = 0; + gui[8] = value[6].real; + gui[9] = value[7].real; + gui[10] = value[8].real; + gui[11] = 0; + } break; + case ShaderLanguage::TYPE_MAT4: { + float *gui = reinterpret_cast<float *>(data); + + for (int i = 0; i < 16; i++) { + gui[i] = value[i].real; + } + } break; + default: { + } + } +} + +_FORCE_INLINE_ static void _fill_std140_ubo_empty(ShaderLanguage::DataType type, int p_array_size, uint8_t *data) { + if (p_array_size <= 0) { + p_array_size = 1; + } + + switch (type) { + case ShaderLanguage::TYPE_BOOL: + case ShaderLanguage::TYPE_INT: + case ShaderLanguage::TYPE_UINT: + case ShaderLanguage::TYPE_FLOAT: { + memset(data, 0, 4 * p_array_size); + } break; + case ShaderLanguage::TYPE_BVEC2: + case ShaderLanguage::TYPE_IVEC2: + case ShaderLanguage::TYPE_UVEC2: + case ShaderLanguage::TYPE_VEC2: { + memset(data, 0, 8 * p_array_size); + } break; + case ShaderLanguage::TYPE_BVEC3: + case ShaderLanguage::TYPE_IVEC3: + case ShaderLanguage::TYPE_UVEC3: + case ShaderLanguage::TYPE_VEC3: { + memset(data, 0, 12 * p_array_size); + } break; + case ShaderLanguage::TYPE_BVEC4: + case ShaderLanguage::TYPE_IVEC4: + case ShaderLanguage::TYPE_UVEC4: + case ShaderLanguage::TYPE_VEC4: { + memset(data, 0, 16 * p_array_size); + } break; + case ShaderLanguage::TYPE_MAT2: { + memset(data, 0, 32 * p_array_size); + } break; + case ShaderLanguage::TYPE_MAT3: { + memset(data, 0, 48 * p_array_size); + } break; + case ShaderLanguage::TYPE_MAT4: { + memset(data, 0, 64 * p_array_size); + } break; + + default: { + } + } +} + +/////////////////////////////////////////////////////////////////////////// +// MaterialStorage::MaterialData + +void MaterialStorage::MaterialData::update_uniform_buffer(const HashMap<StringName, ShaderLanguage::ShaderNode::Uniform> &p_uniforms, const uint32_t *p_uniform_offsets, const HashMap<StringName, Variant> &p_parameters, uint8_t *p_buffer, uint32_t p_buffer_size, bool p_use_linear_color) { + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + bool uses_global_buffer = false; + + for (const KeyValue<StringName, ShaderLanguage::ShaderNode::Uniform> &E : p_uniforms) { + if (E.value.order < 0) { + continue; // texture, does not go here + } + + if (E.value.scope == ShaderLanguage::ShaderNode::Uniform::SCOPE_INSTANCE) { + continue; //instance uniforms don't appear in the buffer + } + + if (E.value.hint == ShaderLanguage::ShaderNode::Uniform::HINT_SCREEN_TEXTURE || + E.value.hint == ShaderLanguage::ShaderNode::Uniform::HINT_NORMAL_ROUGHNESS_TEXTURE || + E.value.hint == ShaderLanguage::ShaderNode::Uniform::HINT_DEPTH_TEXTURE) { + continue; + } + + if (E.value.scope == ShaderLanguage::ShaderNode::Uniform::SCOPE_GLOBAL) { + //this is a global variable, get the index to it + GlobalShaderUniforms::Variable *gv = material_storage->global_shader_uniforms.variables.getptr(E.key); + uint32_t index = 0; + if (gv) { + index = gv->buffer_index; + } else { + WARN_PRINT("Shader uses global parameter '" + E.key + "', but it was removed at some point. Material will not display correctly."); + } + + uint32_t offset = p_uniform_offsets[E.value.order]; + uint32_t *intptr = (uint32_t *)&p_buffer[offset]; + *intptr = index; + uses_global_buffer = true; + continue; + } + + //regular uniform + uint32_t offset = p_uniform_offsets[E.value.order]; +#ifdef DEBUG_ENABLED + uint32_t size = 0U; + // The following code enforces a 16-byte alignment of uniform arrays. + if (E.value.array_size > 0) { + size = ShaderLanguage::get_datatype_size(E.value.type) * E.value.array_size; + int m = (16 * E.value.array_size); + if ((size % m) != 0U) { + size += m - (size % m); + } + } else { + size = ShaderLanguage::get_datatype_size(E.value.type); + } + ERR_CONTINUE(offset + size > p_buffer_size); +#endif + uint8_t *data = &p_buffer[offset]; + HashMap<StringName, Variant>::ConstIterator V = p_parameters.find(E.key); + + if (V) { + //user provided + _fill_std140_variant_ubo_value(E.value.type, E.value.array_size, V->value, data, p_use_linear_color); + + } else if (E.value.default_value.size()) { + //default value + _fill_std140_ubo_value(E.value.type, E.value.default_value, data); + //value=E.value.default_value; + } else { + //zero because it was not provided + if ((E.value.type == ShaderLanguage::TYPE_VEC3 || E.value.type == ShaderLanguage::TYPE_VEC4) && E.value.hint == ShaderLanguage::ShaderNode::Uniform::HINT_SOURCE_COLOR) { + //colors must be set as black, with alpha as 1.0 + _fill_std140_variant_ubo_value(E.value.type, E.value.array_size, Color(0, 0, 0, 1), data, p_use_linear_color); + } else { + //else just zero it out + _fill_std140_ubo_empty(E.value.type, E.value.array_size, data); + } + } + } + + if (uses_global_buffer != (global_buffer_E != nullptr)) { + if (uses_global_buffer) { + global_buffer_E = material_storage->global_shader_uniforms.materials_using_buffer.push_back(self); + } else { + material_storage->global_shader_uniforms.materials_using_buffer.erase(global_buffer_E); + global_buffer_E = nullptr; + } + } +} + +MaterialStorage::MaterialData::~MaterialData() { + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + + if (global_buffer_E) { + //unregister global buffers + material_storage->global_shader_uniforms.materials_using_buffer.erase(global_buffer_E); + } + + if (global_texture_E) { + //unregister global textures + + for (const KeyValue<StringName, uint64_t> &E : used_global_textures) { + GlobalShaderUniforms::Variable *v = material_storage->global_shader_uniforms.variables.getptr(E.key); + if (v) { + v->texture_materials.erase(self); + } + } + //unregister material from those using global textures + material_storage->global_shader_uniforms.materials_using_texture.erase(global_texture_E); + } + + if (uniform_buffer.is_valid()) { + RD::get_singleton()->free(uniform_buffer); + } +} + +void MaterialStorage::MaterialData::update_textures(const HashMap<StringName, Variant> &p_parameters, const HashMap<StringName, HashMap<int, RID>> &p_default_textures, const Vector<ShaderCompiler::GeneratedCode::Texture> &p_texture_uniforms, RID *p_textures, bool p_use_linear_color) { + TextureStorage *texture_storage = TextureStorage::get_singleton(); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + +#ifdef TOOLS_ENABLED + TextureStorage::Texture *roughness_detect_texture = nullptr; + RS::TextureDetectRoughnessChannel roughness_channel = RS::TEXTURE_DETECT_ROUGHNESS_R; + TextureStorage::Texture *normal_detect_texture = nullptr; +#endif + + bool uses_global_textures = false; + global_textures_pass++; + + for (int i = 0, k = 0; i < p_texture_uniforms.size(); i++) { + const StringName &uniform_name = p_texture_uniforms[i].name; + int uniform_array_size = p_texture_uniforms[i].array_size; + + Vector<RID> textures; + + if (p_texture_uniforms[i].hint == ShaderLanguage::ShaderNode::Uniform::HINT_SCREEN_TEXTURE || + p_texture_uniforms[i].hint == ShaderLanguage::ShaderNode::Uniform::HINT_NORMAL_ROUGHNESS_TEXTURE || + p_texture_uniforms[i].hint == ShaderLanguage::ShaderNode::Uniform::HINT_DEPTH_TEXTURE) { + continue; + } + + if (p_texture_uniforms[i].global) { + uses_global_textures = true; + + GlobalShaderUniforms::Variable *v = material_storage->global_shader_uniforms.variables.getptr(uniform_name); + if (v) { + if (v->buffer_index >= 0) { + WARN_PRINT("Shader uses global parameter texture '" + String(uniform_name) + "', but it changed type and is no longer a texture!."); + + } else { + HashMap<StringName, uint64_t>::Iterator E = used_global_textures.find(uniform_name); + if (!E) { + E = used_global_textures.insert(uniform_name, global_textures_pass); + v->texture_materials.insert(self); + } else { + E->value = global_textures_pass; + } + + textures.push_back(v->override.get_type() != Variant::NIL ? v->override : v->value); + } + + } else { + WARN_PRINT("Shader uses global parameter texture '" + String(uniform_name) + "', but it was removed at some point. Material will not display correctly."); + } + } else { + HashMap<StringName, Variant>::ConstIterator V = p_parameters.find(uniform_name); + if (V) { + if (V->value.is_array()) { + Array array = (Array)V->value; + if (uniform_array_size > 0) { + for (int j = 0; j < array.size(); j++) { + textures.push_back(array[j]); + } + } else { + if (array.size() > 0) { + textures.push_back(array[0]); + } + } + } else { + textures.push_back(V->value); + } + } + + if (uniform_array_size > 0) { + if (textures.size() < uniform_array_size) { + HashMap<StringName, HashMap<int, RID>>::ConstIterator W = p_default_textures.find(uniform_name); + for (int j = textures.size(); j < uniform_array_size; j++) { + if (W && W->value.has(j)) { + textures.push_back(W->value[j]); + } else { + textures.push_back(RID()); + } + } + } + } else if (textures.is_empty()) { + HashMap<StringName, HashMap<int, RID>>::ConstIterator W = p_default_textures.find(uniform_name); + if (W && W->value.has(0)) { + textures.push_back(W->value[0]); + } + } + } + + RID rd_texture; + + if (textures.is_empty()) { + //check default usage + switch (p_texture_uniforms[i].type) { + case ShaderLanguage::TYPE_ISAMPLER2D: + case ShaderLanguage::TYPE_USAMPLER2D: + case ShaderLanguage::TYPE_SAMPLER2D: { + switch (p_texture_uniforms[i].hint) { + case ShaderLanguage::ShaderNode::Uniform::HINT_DEFAULT_BLACK: { + rd_texture = texture_storage->texture_rd_get_default(TextureStorage::DEFAULT_RD_TEXTURE_BLACK); + } break; + case ShaderLanguage::ShaderNode::Uniform::HINT_DEFAULT_TRANSPARENT: { + rd_texture = texture_storage->texture_rd_get_default(TextureStorage::DEFAULT_RD_TEXTURE_TRANSPARENT); + } break; + case ShaderLanguage::ShaderNode::Uniform::HINT_ANISOTROPY: { + rd_texture = texture_storage->texture_rd_get_default(TextureStorage::DEFAULT_RD_TEXTURE_ANISO); + } break; + case ShaderLanguage::ShaderNode::Uniform::HINT_NORMAL: { + rd_texture = texture_storage->texture_rd_get_default(TextureStorage::DEFAULT_RD_TEXTURE_NORMAL); + } break; + case ShaderLanguage::ShaderNode::Uniform::HINT_ROUGHNESS_NORMAL: { + rd_texture = texture_storage->texture_rd_get_default(TextureStorage::DEFAULT_RD_TEXTURE_NORMAL); + } break; + default: { + rd_texture = texture_storage->texture_rd_get_default(TextureStorage::DEFAULT_RD_TEXTURE_WHITE); + } break; + } + } break; + + case ShaderLanguage::TYPE_SAMPLERCUBE: { + switch (p_texture_uniforms[i].hint) { + case ShaderLanguage::ShaderNode::Uniform::HINT_DEFAULT_BLACK: { + rd_texture = texture_storage->texture_rd_get_default(TextureStorage::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK); + } break; + default: { + rd_texture = texture_storage->texture_rd_get_default(TextureStorage::DEFAULT_RD_TEXTURE_CUBEMAP_WHITE); + } break; + } + } break; + case ShaderLanguage::TYPE_SAMPLERCUBEARRAY: { + rd_texture = texture_storage->texture_rd_get_default(TextureStorage::DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_BLACK); + } break; + + case ShaderLanguage::TYPE_ISAMPLER3D: + case ShaderLanguage::TYPE_USAMPLER3D: + case ShaderLanguage::TYPE_SAMPLER3D: { + rd_texture = texture_storage->texture_rd_get_default(TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE); + } break; + + case ShaderLanguage::TYPE_ISAMPLER2DARRAY: + case ShaderLanguage::TYPE_USAMPLER2DARRAY: + case ShaderLanguage::TYPE_SAMPLER2DARRAY: { + rd_texture = texture_storage->texture_rd_get_default(TextureStorage::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE); + } break; + + default: { + } + } +#ifdef TOOLS_ENABLED + if (roughness_detect_texture && normal_detect_texture && !normal_detect_texture->path.is_empty()) { + roughness_detect_texture->detect_roughness_callback(roughness_detect_texture->detect_roughness_callback_ud, normal_detect_texture->path, roughness_channel); + } +#endif + if (uniform_array_size > 0) { + for (int j = 0; j < uniform_array_size; j++) { + p_textures[k++] = rd_texture; + } + } else { + p_textures[k++] = rd_texture; + } + } else { + bool srgb = p_use_linear_color && p_texture_uniforms[i].use_color; + + for (int j = 0; j < textures.size(); j++) { + TextureStorage::Texture *tex = TextureStorage::get_singleton()->get_texture(textures[j]); + + if (tex) { + rd_texture = (srgb && tex->rd_texture_srgb.is_valid()) ? tex->rd_texture_srgb : tex->rd_texture; +#ifdef TOOLS_ENABLED + if (tex->detect_3d_callback && p_use_linear_color) { + tex->detect_3d_callback(tex->detect_3d_callback_ud); + } + if (tex->detect_normal_callback && (p_texture_uniforms[i].hint == ShaderLanguage::ShaderNode::Uniform::HINT_NORMAL || p_texture_uniforms[i].hint == ShaderLanguage::ShaderNode::Uniform::HINT_ROUGHNESS_NORMAL)) { + if (p_texture_uniforms[i].hint == ShaderLanguage::ShaderNode::Uniform::HINT_ROUGHNESS_NORMAL) { + normal_detect_texture = tex; + } + tex->detect_normal_callback(tex->detect_normal_callback_ud); + } + if (tex->detect_roughness_callback && (p_texture_uniforms[i].hint >= ShaderLanguage::ShaderNode::Uniform::HINT_ROUGHNESS_R || p_texture_uniforms[i].hint <= ShaderLanguage::ShaderNode::Uniform::HINT_ROUGHNESS_GRAY)) { + //find the normal texture + roughness_detect_texture = tex; + roughness_channel = RS::TextureDetectRoughnessChannel(p_texture_uniforms[i].hint - ShaderLanguage::ShaderNode::Uniform::HINT_ROUGHNESS_R); + } +#endif + } + if (rd_texture.is_null()) { + rd_texture = texture_storage->texture_rd_get_default(TextureStorage::DEFAULT_RD_TEXTURE_WHITE); + } +#ifdef TOOLS_ENABLED + if (roughness_detect_texture && normal_detect_texture && !normal_detect_texture->path.is_empty()) { + roughness_detect_texture->detect_roughness_callback(roughness_detect_texture->detect_roughness_callback_ud, normal_detect_texture->path, roughness_channel); + } +#endif + p_textures[k++] = rd_texture; + } + } + } + { + //for textures no longer used, unregister them + List<StringName> to_delete; + for (KeyValue<StringName, uint64_t> &E : used_global_textures) { + if (E.value != global_textures_pass) { + to_delete.push_back(E.key); + + GlobalShaderUniforms::Variable *v = material_storage->global_shader_uniforms.variables.getptr(E.key); + if (v) { + v->texture_materials.erase(self); + } + } + } + + while (to_delete.front()) { + used_global_textures.erase(to_delete.front()->get()); + to_delete.pop_front(); + } + //handle registering/unregistering global textures + if (uses_global_textures != (global_texture_E != nullptr)) { + if (uses_global_textures) { + global_texture_E = material_storage->global_shader_uniforms.materials_using_texture.push_back(self); + } else { + material_storage->global_shader_uniforms.materials_using_texture.erase(global_texture_E); + global_texture_E = nullptr; + } + } + } +} + +void MaterialStorage::MaterialData::free_parameters_uniform_set(RID p_uniform_set) { + if (p_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(p_uniform_set)) { + RD::get_singleton()->uniform_set_set_invalidation_callback(p_uniform_set, nullptr, nullptr); + RD::get_singleton()->free(p_uniform_set); + } +} + +bool MaterialStorage::MaterialData::update_parameters_uniform_set(const HashMap<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty, const HashMap<StringName, ShaderLanguage::ShaderNode::Uniform> &p_uniforms, const uint32_t *p_uniform_offsets, const Vector<ShaderCompiler::GeneratedCode::Texture> &p_texture_uniforms, const HashMap<StringName, HashMap<int, RID>> &p_default_texture_params, uint32_t p_ubo_size, RID &uniform_set, RID p_shader, uint32_t p_shader_uniform_set, bool p_use_linear_color, uint32_t p_barrier) { + if ((uint32_t)ubo_data.size() != p_ubo_size) { + p_uniform_dirty = true; + if (uniform_buffer.is_valid()) { + RD::get_singleton()->free(uniform_buffer); + uniform_buffer = RID(); + } + + ubo_data.resize(p_ubo_size); + if (ubo_data.size()) { + uniform_buffer = RD::get_singleton()->uniform_buffer_create(ubo_data.size()); + memset(ubo_data.ptrw(), 0, ubo_data.size()); //clear + } + + //clear previous uniform set + if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { + RD::get_singleton()->uniform_set_set_invalidation_callback(uniform_set, nullptr, nullptr); + RD::get_singleton()->free(uniform_set); + uniform_set = RID(); + } + } + + //check whether buffer changed + if (p_uniform_dirty && ubo_data.size()) { + update_uniform_buffer(p_uniforms, p_uniform_offsets, p_parameters, ubo_data.ptrw(), ubo_data.size(), p_use_linear_color); + RD::get_singleton()->buffer_update(uniform_buffer, 0, ubo_data.size(), ubo_data.ptrw(), p_barrier); + } + + uint32_t tex_uniform_count = 0U; + for (int i = 0; i < p_texture_uniforms.size(); i++) { + tex_uniform_count += uint32_t(p_texture_uniforms[i].array_size > 0 ? p_texture_uniforms[i].array_size : 1); + } + + if ((uint32_t)texture_cache.size() != tex_uniform_count || p_textures_dirty) { + texture_cache.resize(tex_uniform_count); + p_textures_dirty = true; + + //clear previous uniform set + if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { + RD::get_singleton()->uniform_set_set_invalidation_callback(uniform_set, nullptr, nullptr); + RD::get_singleton()->free(uniform_set); + uniform_set = RID(); + } + } + + if (p_textures_dirty && tex_uniform_count) { + update_textures(p_parameters, p_default_texture_params, p_texture_uniforms, texture_cache.ptrw(), true); + } + + if (p_ubo_size == 0 && (p_texture_uniforms.size() == 0)) { + // This material does not require an uniform set, so don't create it. + return false; + } + + if (!p_textures_dirty && uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { + //no reason to update uniform set, only UBO (or nothing) was needed to update + return false; + } + + Vector<RD::Uniform> uniforms; + + { + if (p_ubo_size) { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 0; + u.append_id(uniform_buffer); + uniforms.push_back(u); + } + + const RID *textures = texture_cache.ptrw(); + for (int i = 0, k = 0; i < p_texture_uniforms.size(); i++) { + const int array_size = p_texture_uniforms[i].array_size; + + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 1 + k; + if (array_size > 0) { + for (int j = 0; j < array_size; j++) { + u.append_id(textures[k++]); + } + } else { + u.append_id(textures[k++]); + } + uniforms.push_back(u); + } + } + + uniform_set = RD::get_singleton()->uniform_set_create(uniforms, p_shader, p_shader_uniform_set); + + RD::get_singleton()->uniform_set_set_invalidation_callback(uniform_set, MaterialStorage::_material_uniform_set_erased, &self); + + return true; +} + +/////////////////////////////////////////////////////////////////////////// +// MaterialStorage + +MaterialStorage *MaterialStorage::singleton = nullptr; + +MaterialStorage *MaterialStorage::get_singleton() { + return singleton; +} + +MaterialStorage::MaterialStorage() { + singleton = this; + + //default samplers + for (int i = 1; i < RS::CANVAS_ITEM_TEXTURE_FILTER_MAX; i++) { + for (int j = 1; j < RS::CANVAS_ITEM_TEXTURE_REPEAT_MAX; j++) { + RD::SamplerState sampler_state; + switch (i) { + case RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST: { + sampler_state.mag_filter = RD::SAMPLER_FILTER_NEAREST; + sampler_state.min_filter = RD::SAMPLER_FILTER_NEAREST; + sampler_state.max_lod = 0; + } break; + case RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR: { + sampler_state.mag_filter = RD::SAMPLER_FILTER_LINEAR; + sampler_state.min_filter = RD::SAMPLER_FILTER_LINEAR; + sampler_state.max_lod = 0; + } break; + case RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS: { + sampler_state.mag_filter = RD::SAMPLER_FILTER_NEAREST; + sampler_state.min_filter = RD::SAMPLER_FILTER_NEAREST; + if (GLOBAL_GET("rendering/textures/default_filters/use_nearest_mipmap_filter")) { + sampler_state.mip_filter = RD::SAMPLER_FILTER_NEAREST; + } else { + sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + } + } break; + case RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS: { + sampler_state.mag_filter = RD::SAMPLER_FILTER_LINEAR; + sampler_state.min_filter = RD::SAMPLER_FILTER_LINEAR; + if (GLOBAL_GET("rendering/textures/default_filters/use_nearest_mipmap_filter")) { + sampler_state.mip_filter = RD::SAMPLER_FILTER_NEAREST; + } else { + sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + } + + } break; + case RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC: { + sampler_state.mag_filter = RD::SAMPLER_FILTER_NEAREST; + sampler_state.min_filter = RD::SAMPLER_FILTER_NEAREST; + if (GLOBAL_GET("rendering/textures/default_filters/use_nearest_mipmap_filter")) { + sampler_state.mip_filter = RD::SAMPLER_FILTER_NEAREST; + } else { + sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + } + sampler_state.use_anisotropy = true; + sampler_state.anisotropy_max = 1 << int(GLOBAL_GET("rendering/textures/default_filters/anisotropic_filtering_level")); + } break; + case RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC: { + sampler_state.mag_filter = RD::SAMPLER_FILTER_LINEAR; + sampler_state.min_filter = RD::SAMPLER_FILTER_LINEAR; + if (GLOBAL_GET("rendering/textures/default_filters/use_nearest_mipmap_filter")) { + sampler_state.mip_filter = RD::SAMPLER_FILTER_NEAREST; + } else { + sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + } + sampler_state.use_anisotropy = true; + sampler_state.anisotropy_max = 1 << int(GLOBAL_GET("rendering/textures/default_filters/anisotropic_filtering_level")); + + } break; + default: { + } + } + switch (j) { + case RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED: { + sampler_state.repeat_u = RD::SAMPLER_REPEAT_MODE_CLAMP_TO_EDGE; + sampler_state.repeat_v = RD::SAMPLER_REPEAT_MODE_CLAMP_TO_EDGE; + sampler_state.repeat_w = RD::SAMPLER_REPEAT_MODE_CLAMP_TO_EDGE; + + } break; + case RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED: { + sampler_state.repeat_u = RD::SAMPLER_REPEAT_MODE_REPEAT; + sampler_state.repeat_v = RD::SAMPLER_REPEAT_MODE_REPEAT; + sampler_state.repeat_w = RD::SAMPLER_REPEAT_MODE_REPEAT; + } break; + case RS::CANVAS_ITEM_TEXTURE_REPEAT_MIRROR: { + sampler_state.repeat_u = RD::SAMPLER_REPEAT_MODE_MIRRORED_REPEAT; + sampler_state.repeat_v = RD::SAMPLER_REPEAT_MODE_MIRRORED_REPEAT; + sampler_state.repeat_w = RD::SAMPLER_REPEAT_MODE_MIRRORED_REPEAT; + } break; + default: { + } + } + + default_rd_samplers[i][j] = RD::get_singleton()->sampler_create(sampler_state); + } + } + + //custom sampler + sampler_rd_configure_custom(0.0f); + + // buffers + { //create index array for copy shaders + Vector<uint8_t> pv; + pv.resize(6 * 4); + { + uint8_t *w = pv.ptrw(); + int *p32 = (int *)w; + p32[0] = 0; + p32[1] = 1; + p32[2] = 2; + p32[3] = 0; + p32[4] = 2; + p32[5] = 3; + } + quad_index_buffer = RD::get_singleton()->index_buffer_create(6, RenderingDevice::INDEX_BUFFER_FORMAT_UINT32, pv); + quad_index_array = RD::get_singleton()->index_array_create(quad_index_buffer, 0, 6); + } + + // Shaders + for (int i = 0; i < SHADER_TYPE_MAX; i++) { + shader_data_request_func[i] = nullptr; + } + + static_assert(sizeof(GlobalShaderUniforms::Value) == 16); + + global_shader_uniforms.buffer_size = MAX(4096, (int)GLOBAL_GET("rendering/limits/global_shader_variables/buffer_size")); + global_shader_uniforms.buffer_values = memnew_arr(GlobalShaderUniforms::Value, global_shader_uniforms.buffer_size); + memset(global_shader_uniforms.buffer_values, 0, sizeof(GlobalShaderUniforms::Value) * global_shader_uniforms.buffer_size); + global_shader_uniforms.buffer_usage = memnew_arr(GlobalShaderUniforms::ValueUsage, global_shader_uniforms.buffer_size); + global_shader_uniforms.buffer_dirty_regions = memnew_arr(bool, global_shader_uniforms.buffer_size / GlobalShaderUniforms::BUFFER_DIRTY_REGION_SIZE); + memset(global_shader_uniforms.buffer_dirty_regions, 0, sizeof(bool) * global_shader_uniforms.buffer_size / GlobalShaderUniforms::BUFFER_DIRTY_REGION_SIZE); + global_shader_uniforms.buffer = RD::get_singleton()->storage_buffer_create(sizeof(GlobalShaderUniforms::Value) * global_shader_uniforms.buffer_size); +} + +MaterialStorage::~MaterialStorage() { + memdelete_arr(global_shader_uniforms.buffer_values); + memdelete_arr(global_shader_uniforms.buffer_usage); + memdelete_arr(global_shader_uniforms.buffer_dirty_regions); + RD::get_singleton()->free(global_shader_uniforms.buffer); + + // buffers + + RD::get_singleton()->free(quad_index_buffer); //array gets freed as dependency + + //def samplers + for (int i = 1; i < RS::CANVAS_ITEM_TEXTURE_FILTER_MAX; i++) { + for (int j = 1; j < RS::CANVAS_ITEM_TEXTURE_REPEAT_MAX; j++) { + RD::get_singleton()->free(default_rd_samplers[i][j]); + } + } + + //custom samplers + for (int i = 1; i < RS::CANVAS_ITEM_TEXTURE_FILTER_MAX; i++) { + for (int j = 0; j < RS::CANVAS_ITEM_TEXTURE_REPEAT_MAX; j++) { + if (custom_rd_samplers[i][j].is_valid()) { + RD::get_singleton()->free(custom_rd_samplers[i][j]); + } + } + } + + singleton = nullptr; +} + +bool MaterialStorage::free(RID p_rid) { + if (owns_shader(p_rid)) { + shader_free(p_rid); + return true; + } else if (owns_material(p_rid)) { + material_free(p_rid); + return true; + } + + return false; +} + +/* Samplers */ + +void MaterialStorage::sampler_rd_configure_custom(float p_mipmap_bias) { + for (int i = 1; i < RS::CANVAS_ITEM_TEXTURE_FILTER_MAX; i++) { + for (int j = 1; j < RS::CANVAS_ITEM_TEXTURE_REPEAT_MAX; j++) { + RD::SamplerState sampler_state; + switch (i) { + case RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST: { + sampler_state.mag_filter = RD::SAMPLER_FILTER_NEAREST; + sampler_state.min_filter = RD::SAMPLER_FILTER_NEAREST; + sampler_state.max_lod = 0; + } break; + case RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR: { + sampler_state.mag_filter = RD::SAMPLER_FILTER_LINEAR; + sampler_state.min_filter = RD::SAMPLER_FILTER_LINEAR; + sampler_state.max_lod = 0; + } break; + case RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS: { + sampler_state.mag_filter = RD::SAMPLER_FILTER_NEAREST; + sampler_state.min_filter = RD::SAMPLER_FILTER_NEAREST; + if (GLOBAL_GET("rendering/textures/default_filters/use_nearest_mipmap_filter")) { + sampler_state.mip_filter = RD::SAMPLER_FILTER_NEAREST; + } else { + sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + } + sampler_state.lod_bias = p_mipmap_bias; + } break; + case RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS: { + sampler_state.mag_filter = RD::SAMPLER_FILTER_LINEAR; + sampler_state.min_filter = RD::SAMPLER_FILTER_LINEAR; + if (GLOBAL_GET("rendering/textures/default_filters/use_nearest_mipmap_filter")) { + sampler_state.mip_filter = RD::SAMPLER_FILTER_NEAREST; + } else { + sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + } + sampler_state.lod_bias = p_mipmap_bias; + + } break; + case RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC: { + sampler_state.mag_filter = RD::SAMPLER_FILTER_NEAREST; + sampler_state.min_filter = RD::SAMPLER_FILTER_NEAREST; + if (GLOBAL_GET("rendering/textures/default_filters/use_nearest_mipmap_filter")) { + sampler_state.mip_filter = RD::SAMPLER_FILTER_NEAREST; + } else { + sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + } + sampler_state.lod_bias = p_mipmap_bias; + sampler_state.use_anisotropy = true; + sampler_state.anisotropy_max = 1 << int(GLOBAL_GET("rendering/textures/default_filters/anisotropic_filtering_level")); + } break; + case RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC: { + sampler_state.mag_filter = RD::SAMPLER_FILTER_LINEAR; + sampler_state.min_filter = RD::SAMPLER_FILTER_LINEAR; + if (GLOBAL_GET("rendering/textures/default_filters/use_nearest_mipmap_filter")) { + sampler_state.mip_filter = RD::SAMPLER_FILTER_NEAREST; + } else { + sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + } + sampler_state.lod_bias = p_mipmap_bias; + sampler_state.use_anisotropy = true; + sampler_state.anisotropy_max = 1 << int(GLOBAL_GET("rendering/textures/default_filters/anisotropic_filtering_level")); + + } break; + default: { + } + } + switch (j) { + case RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED: { + sampler_state.repeat_u = RD::SAMPLER_REPEAT_MODE_CLAMP_TO_EDGE; + sampler_state.repeat_v = RD::SAMPLER_REPEAT_MODE_CLAMP_TO_EDGE; + sampler_state.repeat_w = RD::SAMPLER_REPEAT_MODE_CLAMP_TO_EDGE; + + } break; + case RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED: { + sampler_state.repeat_u = RD::SAMPLER_REPEAT_MODE_REPEAT; + sampler_state.repeat_v = RD::SAMPLER_REPEAT_MODE_REPEAT; + sampler_state.repeat_w = RD::SAMPLER_REPEAT_MODE_REPEAT; + } break; + case RS::CANVAS_ITEM_TEXTURE_REPEAT_MIRROR: { + sampler_state.repeat_u = RD::SAMPLER_REPEAT_MODE_MIRRORED_REPEAT; + sampler_state.repeat_v = RD::SAMPLER_REPEAT_MODE_MIRRORED_REPEAT; + sampler_state.repeat_w = RD::SAMPLER_REPEAT_MODE_MIRRORED_REPEAT; + } break; + default: { + } + } + + if (custom_rd_samplers[i][j].is_valid()) { + RD::get_singleton()->free(custom_rd_samplers[i][j]); + } + + custom_rd_samplers[i][j] = RD::get_singleton()->sampler_create(sampler_state); + } + } +} + +/* GLOBAL SHADER UNIFORM API */ + +int32_t MaterialStorage::_global_shader_uniform_allocate(uint32_t p_elements) { + int32_t idx = 0; + while (idx + p_elements <= global_shader_uniforms.buffer_size) { + if (global_shader_uniforms.buffer_usage[idx].elements == 0) { + bool valid = true; + for (uint32_t i = 1; i < p_elements; i++) { + if (global_shader_uniforms.buffer_usage[idx + i].elements > 0) { + valid = false; + idx += i + global_shader_uniforms.buffer_usage[idx + i].elements; + break; + } + } + + if (!valid) { + continue; //if not valid, idx is in new position + } + + return idx; + } else { + idx += global_shader_uniforms.buffer_usage[idx].elements; + } + } + + return -1; +} + +void MaterialStorage::_global_shader_uniform_store_in_buffer(int32_t p_index, RS::GlobalShaderParameterType p_type, const Variant &p_value) { + switch (p_type) { + case RS::GLOBAL_VAR_TYPE_BOOL: { + GlobalShaderUniforms::Value &bv = global_shader_uniforms.buffer_values[p_index]; + bool b = p_value; + bv.x = b ? 1.0 : 0.0; + bv.y = 0.0; + bv.z = 0.0; + bv.w = 0.0; + + } break; + case RS::GLOBAL_VAR_TYPE_BVEC2: { + GlobalShaderUniforms::Value &bv = global_shader_uniforms.buffer_values[p_index]; + uint32_t bvec = p_value; + bv.x = (bvec & 1) ? 1.0 : 0.0; + bv.y = (bvec & 2) ? 1.0 : 0.0; + bv.z = 0.0; + bv.w = 0.0; + } break; + case RS::GLOBAL_VAR_TYPE_BVEC3: { + GlobalShaderUniforms::Value &bv = global_shader_uniforms.buffer_values[p_index]; + uint32_t bvec = p_value; + bv.x = (bvec & 1) ? 1.0 : 0.0; + bv.y = (bvec & 2) ? 1.0 : 0.0; + bv.z = (bvec & 4) ? 1.0 : 0.0; + bv.w = 0.0; + } break; + case RS::GLOBAL_VAR_TYPE_BVEC4: { + GlobalShaderUniforms::Value &bv = global_shader_uniforms.buffer_values[p_index]; + uint32_t bvec = p_value; + bv.x = (bvec & 1) ? 1.0 : 0.0; + bv.y = (bvec & 2) ? 1.0 : 0.0; + bv.z = (bvec & 4) ? 1.0 : 0.0; + bv.w = (bvec & 8) ? 1.0 : 0.0; + } break; + case RS::GLOBAL_VAR_TYPE_INT: { + GlobalShaderUniforms::ValueInt &bv = *(GlobalShaderUniforms::ValueInt *)&global_shader_uniforms.buffer_values[p_index]; + int32_t v = p_value; + bv.x = v; + bv.y = 0; + bv.z = 0; + bv.w = 0; + } break; + case RS::GLOBAL_VAR_TYPE_IVEC2: { + GlobalShaderUniforms::ValueInt &bv = *(GlobalShaderUniforms::ValueInt *)&global_shader_uniforms.buffer_values[p_index]; + Vector2i v = p_value; + bv.x = v.x; + bv.y = v.y; + bv.z = 0; + bv.w = 0; + } break; + case RS::GLOBAL_VAR_TYPE_IVEC3: { + GlobalShaderUniforms::ValueInt &bv = *(GlobalShaderUniforms::ValueInt *)&global_shader_uniforms.buffer_values[p_index]; + Vector3i v = p_value; + bv.x = v.x; + bv.y = v.y; + bv.z = v.z; + bv.w = 0; + } break; + case RS::GLOBAL_VAR_TYPE_IVEC4: { + GlobalShaderUniforms::ValueInt &bv = *(GlobalShaderUniforms::ValueInt *)&global_shader_uniforms.buffer_values[p_index]; + Vector<int32_t> v = p_value; + bv.x = v.size() >= 1 ? v[0] : 0; + bv.y = v.size() >= 2 ? v[1] : 0; + bv.z = v.size() >= 3 ? v[2] : 0; + bv.w = v.size() >= 4 ? v[3] : 0; + } break; + case RS::GLOBAL_VAR_TYPE_RECT2I: { + GlobalShaderUniforms::ValueInt &bv = *(GlobalShaderUniforms::ValueInt *)&global_shader_uniforms.buffer_values[p_index]; + Rect2i v = p_value; + bv.x = v.position.x; + bv.y = v.position.y; + bv.z = v.size.x; + bv.w = v.size.y; + } break; + case RS::GLOBAL_VAR_TYPE_UINT: { + GlobalShaderUniforms::ValueUInt &bv = *(GlobalShaderUniforms::ValueUInt *)&global_shader_uniforms.buffer_values[p_index]; + uint32_t v = p_value; + bv.x = v; + bv.y = 0; + bv.z = 0; + bv.w = 0; + } break; + case RS::GLOBAL_VAR_TYPE_UVEC2: { + GlobalShaderUniforms::ValueUInt &bv = *(GlobalShaderUniforms::ValueUInt *)&global_shader_uniforms.buffer_values[p_index]; + Vector2i v = p_value; + bv.x = v.x; + bv.y = v.y; + bv.z = 0; + bv.w = 0; + } break; + case RS::GLOBAL_VAR_TYPE_UVEC3: { + GlobalShaderUniforms::ValueUInt &bv = *(GlobalShaderUniforms::ValueUInt *)&global_shader_uniforms.buffer_values[p_index]; + Vector3i v = p_value; + bv.x = v.x; + bv.y = v.y; + bv.z = v.z; + bv.w = 0; + } break; + case RS::GLOBAL_VAR_TYPE_UVEC4: { + GlobalShaderUniforms::ValueUInt &bv = *(GlobalShaderUniforms::ValueUInt *)&global_shader_uniforms.buffer_values[p_index]; + Vector<int32_t> v = p_value; + bv.x = v.size() >= 1 ? v[0] : 0; + bv.y = v.size() >= 2 ? v[1] : 0; + bv.z = v.size() >= 3 ? v[2] : 0; + bv.w = v.size() >= 4 ? v[3] : 0; + } break; + case RS::GLOBAL_VAR_TYPE_FLOAT: { + GlobalShaderUniforms::Value &bv = global_shader_uniforms.buffer_values[p_index]; + float v = p_value; + bv.x = v; + bv.y = 0; + bv.z = 0; + bv.w = 0; + } break; + case RS::GLOBAL_VAR_TYPE_VEC2: { + GlobalShaderUniforms::Value &bv = global_shader_uniforms.buffer_values[p_index]; + Vector2 v = p_value; + bv.x = v.x; + bv.y = v.y; + bv.z = 0; + bv.w = 0; + } break; + case RS::GLOBAL_VAR_TYPE_VEC3: { + GlobalShaderUniforms::Value &bv = global_shader_uniforms.buffer_values[p_index]; + Vector3 v = p_value; + bv.x = v.x; + bv.y = v.y; + bv.z = v.z; + bv.w = 0; + } break; + case RS::GLOBAL_VAR_TYPE_VEC4: { + GlobalShaderUniforms::Value &bv = global_shader_uniforms.buffer_values[p_index]; + Plane v = p_value; + bv.x = v.normal.x; + bv.y = v.normal.y; + bv.z = v.normal.z; + bv.w = v.d; + } break; + case RS::GLOBAL_VAR_TYPE_COLOR: { + GlobalShaderUniforms::Value &bv = global_shader_uniforms.buffer_values[p_index]; + Color v = p_value; + bv.x = v.r; + bv.y = v.g; + bv.z = v.b; + bv.w = v.a; + + GlobalShaderUniforms::Value &bv_linear = global_shader_uniforms.buffer_values[p_index + 1]; + v = v.srgb_to_linear(); + bv_linear.x = v.r; + bv_linear.y = v.g; + bv_linear.z = v.b; + bv_linear.w = v.a; + + } break; + case RS::GLOBAL_VAR_TYPE_RECT2: { + GlobalShaderUniforms::Value &bv = global_shader_uniforms.buffer_values[p_index]; + Rect2 v = p_value; + bv.x = v.position.x; + bv.y = v.position.y; + bv.z = v.size.x; + bv.w = v.size.y; + } break; + case RS::GLOBAL_VAR_TYPE_MAT2: { + GlobalShaderUniforms::Value *bv = &global_shader_uniforms.buffer_values[p_index]; + Vector<float> m2 = p_value; + if (m2.size() < 4) { + m2.resize(4); + } + bv[0].x = m2[0]; + bv[0].y = m2[1]; + bv[0].z = 0; + bv[0].w = 0; + + bv[1].x = m2[2]; + bv[1].y = m2[3]; + bv[1].z = 0; + bv[1].w = 0; + + } break; + case RS::GLOBAL_VAR_TYPE_MAT3: { + GlobalShaderUniforms::Value *bv = &global_shader_uniforms.buffer_values[p_index]; + Basis v = p_value; + bv[0].x = v.rows[0][0]; + bv[0].y = v.rows[1][0]; + bv[0].z = v.rows[2][0]; + bv[0].w = 0; + + bv[1].x = v.rows[0][1]; + bv[1].y = v.rows[1][1]; + bv[1].z = v.rows[2][1]; + bv[1].w = 0; + + bv[2].x = v.rows[0][2]; + bv[2].y = v.rows[1][2]; + bv[2].z = v.rows[2][2]; + bv[2].w = 0; + + } break; + case RS::GLOBAL_VAR_TYPE_MAT4: { + GlobalShaderUniforms::Value *bv = &global_shader_uniforms.buffer_values[p_index]; + + Vector<float> m2 = p_value; + if (m2.size() < 16) { + m2.resize(16); + } + + bv[0].x = m2[0]; + bv[0].y = m2[1]; + bv[0].z = m2[2]; + bv[0].w = m2[3]; + + bv[1].x = m2[4]; + bv[1].y = m2[5]; + bv[1].z = m2[6]; + bv[1].w = m2[7]; + + bv[2].x = m2[8]; + bv[2].y = m2[9]; + bv[2].z = m2[10]; + bv[2].w = m2[11]; + + bv[3].x = m2[12]; + bv[3].y = m2[13]; + bv[3].z = m2[14]; + bv[3].w = m2[15]; + + } break; + case RS::GLOBAL_VAR_TYPE_TRANSFORM_2D: { + GlobalShaderUniforms::Value *bv = &global_shader_uniforms.buffer_values[p_index]; + Transform2D v = p_value; + bv[0].x = v.columns[0][0]; + bv[0].y = v.columns[0][1]; + bv[0].z = 0; + bv[0].w = 0; + + bv[1].x = v.columns[1][0]; + bv[1].y = v.columns[1][1]; + bv[1].z = 0; + bv[1].w = 0; + + bv[2].x = v.columns[2][0]; + bv[2].y = v.columns[2][1]; + bv[2].z = 1; + bv[2].w = 0; + + } break; + case RS::GLOBAL_VAR_TYPE_TRANSFORM: { + GlobalShaderUniforms::Value *bv = &global_shader_uniforms.buffer_values[p_index]; + Transform3D v = p_value; + bv[0].x = v.basis.rows[0][0]; + bv[0].y = v.basis.rows[1][0]; + bv[0].z = v.basis.rows[2][0]; + bv[0].w = 0; + + bv[1].x = v.basis.rows[0][1]; + bv[1].y = v.basis.rows[1][1]; + bv[1].z = v.basis.rows[2][1]; + bv[1].w = 0; + + bv[2].x = v.basis.rows[0][2]; + bv[2].y = v.basis.rows[1][2]; + bv[2].z = v.basis.rows[2][2]; + bv[2].w = 0; + + bv[3].x = v.origin.x; + bv[3].y = v.origin.y; + bv[3].z = v.origin.z; + bv[3].w = 1; + + } break; + default: { + ERR_FAIL(); + } + } +} + +void MaterialStorage::_global_shader_uniform_mark_buffer_dirty(int32_t p_index, int32_t p_elements) { + int32_t prev_chunk = -1; + + for (int32_t i = 0; i < p_elements; i++) { + int32_t chunk = (p_index + i) / GlobalShaderUniforms::BUFFER_DIRTY_REGION_SIZE; + if (chunk != prev_chunk) { + if (!global_shader_uniforms.buffer_dirty_regions[chunk]) { + global_shader_uniforms.buffer_dirty_regions[chunk] = true; + global_shader_uniforms.buffer_dirty_region_count++; + } + } + + prev_chunk = chunk; + } +} + +void MaterialStorage::global_shader_parameter_add(const StringName &p_name, RS::GlobalShaderParameterType p_type, const Variant &p_value) { + ERR_FAIL_COND(global_shader_uniforms.variables.has(p_name)); + GlobalShaderUniforms::Variable gv; + gv.type = p_type; + gv.value = p_value; + gv.buffer_index = -1; + + if (p_type >= RS::GLOBAL_VAR_TYPE_SAMPLER2D) { + //is texture + global_shader_uniforms.must_update_texture_materials = true; //normally there are none + } else { + gv.buffer_elements = 1; + if (p_type == RS::GLOBAL_VAR_TYPE_COLOR || p_type == RS::GLOBAL_VAR_TYPE_MAT2) { + //color needs to elements to store srgb and linear + gv.buffer_elements = 2; + } + if (p_type == RS::GLOBAL_VAR_TYPE_MAT3 || p_type == RS::GLOBAL_VAR_TYPE_TRANSFORM_2D) { + //color needs to elements to store srgb and linear + gv.buffer_elements = 3; + } + if (p_type == RS::GLOBAL_VAR_TYPE_MAT4 || p_type == RS::GLOBAL_VAR_TYPE_TRANSFORM) { + //color needs to elements to store srgb and linear + gv.buffer_elements = 4; + } + + //is vector, allocate in buffer and update index + gv.buffer_index = _global_shader_uniform_allocate(gv.buffer_elements); + ERR_FAIL_COND_MSG(gv.buffer_index < 0, vformat("Failed allocating global variable '%s' out of buffer memory. Consider increasing it in the Project Settings.", String(p_name))); + global_shader_uniforms.buffer_usage[gv.buffer_index].elements = gv.buffer_elements; + _global_shader_uniform_store_in_buffer(gv.buffer_index, gv.type, gv.value); + _global_shader_uniform_mark_buffer_dirty(gv.buffer_index, gv.buffer_elements); + + global_shader_uniforms.must_update_buffer_materials = true; //normally there are none + } + + global_shader_uniforms.variables[p_name] = gv; +} + +void MaterialStorage::global_shader_parameter_remove(const StringName &p_name) { + if (!global_shader_uniforms.variables.has(p_name)) { + return; + } + const GlobalShaderUniforms::Variable &gv = global_shader_uniforms.variables[p_name]; + + if (gv.buffer_index >= 0) { + global_shader_uniforms.buffer_usage[gv.buffer_index].elements = 0; + global_shader_uniforms.must_update_buffer_materials = true; + } else { + global_shader_uniforms.must_update_texture_materials = true; + } + + global_shader_uniforms.variables.erase(p_name); +} + +Vector<StringName> MaterialStorage::global_shader_parameter_get_list() const { + if (!Engine::get_singleton()->is_editor_hint()) { + ERR_FAIL_V_MSG(Vector<StringName>(), "This function should never be used outside the editor, it can severely damage performance."); + } + + Vector<StringName> names; + for (const KeyValue<StringName, GlobalShaderUniforms::Variable> &E : global_shader_uniforms.variables) { + names.push_back(E.key); + } + names.sort_custom<StringName::AlphCompare>(); + return names; +} + +void MaterialStorage::global_shader_parameter_set(const StringName &p_name, const Variant &p_value) { + ERR_FAIL_COND(!global_shader_uniforms.variables.has(p_name)); + GlobalShaderUniforms::Variable &gv = global_shader_uniforms.variables[p_name]; + gv.value = p_value; + if (gv.override.get_type() == Variant::NIL) { + if (gv.buffer_index >= 0) { + //buffer + _global_shader_uniform_store_in_buffer(gv.buffer_index, gv.type, gv.value); + _global_shader_uniform_mark_buffer_dirty(gv.buffer_index, gv.buffer_elements); + } else { + //texture + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + for (const RID &E : gv.texture_materials) { + Material *material = material_storage->get_material(E); + ERR_CONTINUE(!material); + material_storage->_material_queue_update(material, false, true); + } + } + } +} + +void MaterialStorage::global_shader_parameter_set_override(const StringName &p_name, const Variant &p_value) { + if (!global_shader_uniforms.variables.has(p_name)) { + return; //variable may not exist + } + + ERR_FAIL_COND(p_value.get_type() == Variant::OBJECT); + + GlobalShaderUniforms::Variable &gv = global_shader_uniforms.variables[p_name]; + + gv.override = p_value; + + if (gv.buffer_index >= 0) { + //buffer + if (gv.override.get_type() == Variant::NIL) { + _global_shader_uniform_store_in_buffer(gv.buffer_index, gv.type, gv.value); + } else { + _global_shader_uniform_store_in_buffer(gv.buffer_index, gv.type, gv.override); + } + + _global_shader_uniform_mark_buffer_dirty(gv.buffer_index, gv.buffer_elements); + } else { + //texture + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + for (const RID &E : gv.texture_materials) { + Material *material = material_storage->get_material(E); + ERR_CONTINUE(!material); + material_storage->_material_queue_update(material, false, true); + } + } +} + +Variant MaterialStorage::global_shader_parameter_get(const StringName &p_name) const { + if (!Engine::get_singleton()->is_editor_hint()) { + ERR_FAIL_V_MSG(Variant(), "This function should never be used outside the editor, it can severely damage performance."); + } + + if (!global_shader_uniforms.variables.has(p_name)) { + return Variant(); + } + + return global_shader_uniforms.variables[p_name].value; +} + +RS::GlobalShaderParameterType MaterialStorage::global_shader_parameter_get_type_internal(const StringName &p_name) const { + if (!global_shader_uniforms.variables.has(p_name)) { + return RS::GLOBAL_VAR_TYPE_MAX; + } + + return global_shader_uniforms.variables[p_name].type; +} + +RS::GlobalShaderParameterType MaterialStorage::global_shader_parameter_get_type(const StringName &p_name) const { + if (!Engine::get_singleton()->is_editor_hint()) { + ERR_FAIL_V_MSG(RS::GLOBAL_VAR_TYPE_MAX, "This function should never be used outside the editor, it can severely damage performance."); + } + + return global_shader_parameter_get_type_internal(p_name); +} + +void MaterialStorage::global_shader_parameters_load_settings(bool p_load_textures) { + List<PropertyInfo> settings; + ProjectSettings::get_singleton()->get_property_list(&settings); + + for (const PropertyInfo &E : settings) { + if (E.name.begins_with("shader_globals/")) { + StringName name = E.name.get_slice("/", 1); + Dictionary d = GLOBAL_GET(E.name); + + ERR_CONTINUE(!d.has("type")); + ERR_CONTINUE(!d.has("value")); + + String type = d["type"]; + + static const char *global_var_type_names[RS::GLOBAL_VAR_TYPE_MAX] = { + "bool", + "bvec2", + "bvec3", + "bvec4", + "int", + "ivec2", + "ivec3", + "ivec4", + "rect2i", + "uint", + "uvec2", + "uvec3", + "uvec4", + "float", + "vec2", + "vec3", + "vec4", + "color", + "rect2", + "mat2", + "mat3", + "mat4", + "transform_2d", + "transform", + "sampler2D", + "sampler2DArray", + "sampler3D", + "samplerCube", + }; + + RS::GlobalShaderParameterType gvtype = RS::GLOBAL_VAR_TYPE_MAX; + + for (int i = 0; i < RS::GLOBAL_VAR_TYPE_MAX; i++) { + if (global_var_type_names[i] == type) { + gvtype = RS::GlobalShaderParameterType(i); + break; + } + } + + ERR_CONTINUE(gvtype == RS::GLOBAL_VAR_TYPE_MAX); //type invalid + + Variant value = d["value"]; + + if (gvtype >= RS::GLOBAL_VAR_TYPE_SAMPLER2D) { + //textire + if (!p_load_textures) { + value = RID(); + continue; + } + + String path = value; + Ref<Resource> resource = ResourceLoader::load(path); + ERR_CONTINUE(resource.is_null()); + value = resource; + } + + if (global_shader_uniforms.variables.has(name)) { + //has it, update it + global_shader_parameter_set(name, value); + } else { + global_shader_parameter_add(name, gvtype, value); + } + } + } +} + +void MaterialStorage::global_shader_parameters_clear() { + global_shader_uniforms.variables.clear(); //not right but for now enough +} + +RID MaterialStorage::global_shader_uniforms_get_storage_buffer() const { + return global_shader_uniforms.buffer; +} + +int32_t MaterialStorage::global_shader_parameters_instance_allocate(RID p_instance) { + ERR_FAIL_COND_V(global_shader_uniforms.instance_buffer_pos.has(p_instance), -1); + int32_t pos = _global_shader_uniform_allocate(ShaderLanguage::MAX_INSTANCE_UNIFORM_INDICES); + global_shader_uniforms.instance_buffer_pos[p_instance] = pos; //save anyway + ERR_FAIL_COND_V_MSG(pos < 0, -1, "Too many instances using shader instance variables. Increase buffer size in Project Settings."); + global_shader_uniforms.buffer_usage[pos].elements = ShaderLanguage::MAX_INSTANCE_UNIFORM_INDICES; + return pos; +} + +void MaterialStorage::global_shader_parameters_instance_free(RID p_instance) { + ERR_FAIL_COND(!global_shader_uniforms.instance_buffer_pos.has(p_instance)); + int32_t pos = global_shader_uniforms.instance_buffer_pos[p_instance]; + if (pos >= 0) { + global_shader_uniforms.buffer_usage[pos].elements = 0; + } + global_shader_uniforms.instance_buffer_pos.erase(p_instance); +} + +void MaterialStorage::global_shader_parameters_instance_update(RID p_instance, int p_index, const Variant &p_value, int p_flags_count) { + if (!global_shader_uniforms.instance_buffer_pos.has(p_instance)) { + return; //just not allocated, ignore + } + int32_t pos = global_shader_uniforms.instance_buffer_pos[p_instance]; + + if (pos < 0) { + return; //again, not allocated, ignore + } + ERR_FAIL_INDEX(p_index, ShaderLanguage::MAX_INSTANCE_UNIFORM_INDICES); + + Variant::Type value_type = p_value.get_type(); + ERR_FAIL_COND_MSG(p_value.get_type() > Variant::COLOR, "Unsupported variant type for instance parameter: " + Variant::get_type_name(value_type)); //anything greater not supported + + const ShaderLanguage::DataType datatype_from_value[Variant::COLOR + 1] = { + ShaderLanguage::TYPE_MAX, //nil + ShaderLanguage::TYPE_BOOL, //bool + ShaderLanguage::TYPE_INT, //int + ShaderLanguage::TYPE_FLOAT, //float + ShaderLanguage::TYPE_MAX, //string + ShaderLanguage::TYPE_VEC2, //vec2 + ShaderLanguage::TYPE_IVEC2, //vec2i + ShaderLanguage::TYPE_VEC4, //rect2 + ShaderLanguage::TYPE_IVEC4, //rect2i + ShaderLanguage::TYPE_VEC3, // vec3 + ShaderLanguage::TYPE_IVEC3, //vec3i + ShaderLanguage::TYPE_MAX, //xform2d not supported here + ShaderLanguage::TYPE_VEC4, //vec4 + ShaderLanguage::TYPE_IVEC4, //vec4i + ShaderLanguage::TYPE_VEC4, //plane + ShaderLanguage::TYPE_VEC4, //quat + ShaderLanguage::TYPE_MAX, //aabb not supported here + ShaderLanguage::TYPE_MAX, //basis not supported here + ShaderLanguage::TYPE_MAX, //xform not supported here + ShaderLanguage::TYPE_MAX, //projection not supported here + ShaderLanguage::TYPE_VEC4 //color + }; + + ShaderLanguage::DataType datatype = ShaderLanguage::TYPE_MAX; + if (value_type == Variant::INT && p_flags_count > 0) { + switch (p_flags_count) { + case 1: + datatype = ShaderLanguage::TYPE_BVEC2; + break; + case 2: + datatype = ShaderLanguage::TYPE_BVEC3; + break; + case 3: + datatype = ShaderLanguage::TYPE_BVEC4; + break; + } + } else { + datatype = datatype_from_value[value_type]; + } + ERR_FAIL_COND_MSG(datatype == ShaderLanguage::TYPE_MAX, "Unsupported variant type for instance parameter: " + Variant::get_type_name(value_type)); //anything greater not supported + + pos += p_index; + + _fill_std140_variant_ubo_value(datatype, 0, p_value, (uint8_t *)&global_shader_uniforms.buffer_values[pos], true); //instances always use linear color in this renderer + _global_shader_uniform_mark_buffer_dirty(pos, 1); +} + +void MaterialStorage::_update_global_shader_uniforms() { + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + if (global_shader_uniforms.buffer_dirty_region_count > 0) { + uint32_t total_regions = global_shader_uniforms.buffer_size / GlobalShaderUniforms::BUFFER_DIRTY_REGION_SIZE; + if (total_regions / global_shader_uniforms.buffer_dirty_region_count <= 4) { + // 25% of regions dirty, just update all buffer + RD::get_singleton()->buffer_update(global_shader_uniforms.buffer, 0, sizeof(GlobalShaderUniforms::Value) * global_shader_uniforms.buffer_size, global_shader_uniforms.buffer_values); + memset(global_shader_uniforms.buffer_dirty_regions, 0, sizeof(bool) * total_regions); + } else { + uint32_t region_byte_size = sizeof(GlobalShaderUniforms::Value) * GlobalShaderUniforms::BUFFER_DIRTY_REGION_SIZE; + + for (uint32_t i = 0; i < total_regions; i++) { + if (global_shader_uniforms.buffer_dirty_regions[i]) { + RD::get_singleton()->buffer_update(global_shader_uniforms.buffer, i * region_byte_size, region_byte_size, &global_shader_uniforms.buffer_values[i * GlobalShaderUniforms::BUFFER_DIRTY_REGION_SIZE]); + + global_shader_uniforms.buffer_dirty_regions[i] = false; + } + } + } + + global_shader_uniforms.buffer_dirty_region_count = 0; + } + + if (global_shader_uniforms.must_update_buffer_materials) { + // only happens in the case of a buffer variable added or removed, + // so not often. + for (const RID &E : global_shader_uniforms.materials_using_buffer) { + Material *material = material_storage->get_material(E); + ERR_CONTINUE(!material); //wtf + + material_storage->_material_queue_update(material, true, false); + } + + global_shader_uniforms.must_update_buffer_materials = false; + } + + if (global_shader_uniforms.must_update_texture_materials) { + // only happens in the case of a buffer variable added or removed, + // so not often. + for (const RID &E : global_shader_uniforms.materials_using_texture) { + Material *material = material_storage->get_material(E); + ERR_CONTINUE(!material); //wtf + + material_storage->_material_queue_update(material, false, true); + } + + global_shader_uniforms.must_update_texture_materials = false; + } +} + +/* SHADER API */ + +RID MaterialStorage::shader_allocate() { + return shader_owner.allocate_rid(); +} + +void MaterialStorage::shader_initialize(RID p_rid) { + Shader shader; + shader.data = nullptr; + shader.type = SHADER_TYPE_MAX; + + shader_owner.initialize_rid(p_rid, shader); +} + +void MaterialStorage::shader_free(RID p_rid) { + Shader *shader = shader_owner.get_or_null(p_rid); + ERR_FAIL_COND(!shader); + + //make material unreference this + while (shader->owners.size()) { + material_set_shader((*shader->owners.begin())->self, RID()); + } + + //clear data if exists + if (shader->data) { + memdelete(shader->data); + } + shader_owner.free(p_rid); +} + +void MaterialStorage::shader_set_code(RID p_shader, const String &p_code) { + Shader *shader = shader_owner.get_or_null(p_shader); + ERR_FAIL_COND(!shader); + + shader->code = p_code; + String mode_string = ShaderLanguage::get_shader_type(p_code); + + ShaderType new_type; + if (mode_string == "canvas_item") { + new_type = SHADER_TYPE_2D; + } else if (mode_string == "particles") { + new_type = SHADER_TYPE_PARTICLES; + } else if (mode_string == "spatial") { + new_type = SHADER_TYPE_3D; + } else if (mode_string == "sky") { + new_type = SHADER_TYPE_SKY; + } else if (mode_string == "fog") { + new_type = SHADER_TYPE_FOG; + } else { + new_type = SHADER_TYPE_MAX; + } + + if (new_type != shader->type) { + if (shader->data) { + memdelete(shader->data); + shader->data = nullptr; + } + + for (Material *E : shader->owners) { + Material *material = E; + material->shader_type = new_type; + if (material->data) { + memdelete(material->data); + material->data = nullptr; + } + } + + shader->type = new_type; + + if (new_type < SHADER_TYPE_MAX && shader_data_request_func[new_type]) { + shader->data = shader_data_request_func[new_type](); + } else { + shader->type = SHADER_TYPE_MAX; //invalid + } + + for (Material *E : shader->owners) { + Material *material = E; + if (shader->data) { + material->data = material_get_data_request_function(new_type)(shader->data); + material->data->self = material->self; + material->data->set_next_pass(material->next_pass); + material->data->set_render_priority(material->priority); + } + material->shader_type = new_type; + } + + if (shader->data) { + for (const KeyValue<StringName, HashMap<int, RID>> &E : shader->default_texture_parameter) { + for (const KeyValue<int, RID> &E2 : E.value) { + shader->data->set_default_texture_parameter(E.key, E2.value, E2.key); + } + } + } + } + + if (shader->data) { + shader->data->set_path_hint(shader->path_hint); + shader->data->set_code(p_code); + } + + for (Material *E : shader->owners) { + Material *material = E; + material->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_MATERIAL); + _material_queue_update(material, true, true); + } +} + +void MaterialStorage::shader_set_path_hint(RID p_shader, const String &p_path) { + Shader *shader = shader_owner.get_or_null(p_shader); + ERR_FAIL_COND(!shader); + + shader->path_hint = p_path; + if (shader->data) { + shader->data->set_path_hint(p_path); + } +} + +String MaterialStorage::shader_get_code(RID p_shader) const { + Shader *shader = shader_owner.get_or_null(p_shader); + ERR_FAIL_COND_V(!shader, String()); + return shader->code; +} + +void MaterialStorage::get_shader_parameter_list(RID p_shader, List<PropertyInfo> *p_param_list) const { + Shader *shader = shader_owner.get_or_null(p_shader); + ERR_FAIL_COND(!shader); + if (shader->data) { + return shader->data->get_shader_uniform_list(p_param_list); + } +} + +void MaterialStorage::shader_set_default_texture_parameter(RID p_shader, const StringName &p_name, RID p_texture, int p_index) { + Shader *shader = shader_owner.get_or_null(p_shader); + ERR_FAIL_COND(!shader); + + if (p_texture.is_valid() && TextureStorage::get_singleton()->owns_texture(p_texture)) { + if (!shader->default_texture_parameter.has(p_name)) { + shader->default_texture_parameter[p_name] = HashMap<int, RID>(); + } + shader->default_texture_parameter[p_name][p_index] = p_texture; + } else { + if (shader->default_texture_parameter.has(p_name) && shader->default_texture_parameter[p_name].has(p_index)) { + shader->default_texture_parameter[p_name].erase(p_index); + + if (shader->default_texture_parameter[p_name].is_empty()) { + shader->default_texture_parameter.erase(p_name); + } + } + } + if (shader->data) { + shader->data->set_default_texture_parameter(p_name, p_texture, p_index); + } + for (Material *E : shader->owners) { + Material *material = E; + _material_queue_update(material, false, true); + } +} + +RID MaterialStorage::shader_get_default_texture_parameter(RID p_shader, const StringName &p_name, int p_index) const { + Shader *shader = shader_owner.get_or_null(p_shader); + ERR_FAIL_COND_V(!shader, RID()); + if (shader->default_texture_parameter.has(p_name) && shader->default_texture_parameter[p_name].has(p_index)) { + return shader->default_texture_parameter[p_name][p_index]; + } + + return RID(); +} + +Variant MaterialStorage::shader_get_parameter_default(RID p_shader, const StringName &p_param) const { + Shader *shader = shader_owner.get_or_null(p_shader); + ERR_FAIL_COND_V(!shader, Variant()); + if (shader->data) { + return shader->data->get_default_parameter(p_param); + } + return Variant(); +} + +void MaterialStorage::shader_set_data_request_function(ShaderType p_shader_type, ShaderDataRequestFunction p_function) { + ERR_FAIL_INDEX(p_shader_type, SHADER_TYPE_MAX); + shader_data_request_func[p_shader_type] = p_function; +} + +RS::ShaderNativeSourceCode MaterialStorage::shader_get_native_source_code(RID p_shader) const { + Shader *shader = shader_owner.get_or_null(p_shader); + ERR_FAIL_COND_V(!shader, RS::ShaderNativeSourceCode()); + if (shader->data) { + return shader->data->get_native_source_code(); + } + return RS::ShaderNativeSourceCode(); +} + +/* MATERIAL API */ + +void MaterialStorage::_material_uniform_set_erased(void *p_material) { + RID rid = *(RID *)p_material; + Material *material = MaterialStorage::get_singleton()->get_material(rid); + if (material) { + if (material->data) { + // Uniform set may be gone because a dependency was erased. This happens + // if a texture is deleted, so re-create it. + MaterialStorage::get_singleton()->_material_queue_update(material, false, true); + } + material->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_MATERIAL); + } +} + +void MaterialStorage::_material_queue_update(Material *material, bool p_uniform, bool p_texture) { + material->uniform_dirty = material->uniform_dirty || p_uniform; + material->texture_dirty = material->texture_dirty || p_texture; + + if (material->update_element.in_list()) { + return; + } + + material_update_list.add(&material->update_element); +} + +void MaterialStorage::_update_queued_materials() { + while (material_update_list.first()) { + Material *material = material_update_list.first()->self(); + bool uniforms_changed = false; + + if (material->data) { + uniforms_changed = material->data->update_parameters(material->params, material->uniform_dirty, material->texture_dirty); + } + material->texture_dirty = false; + material->uniform_dirty = false; + + material_update_list.remove(&material->update_element); + + if (uniforms_changed) { + //some implementations such as 3D renderer cache the material uniform set, so update is required + material->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_MATERIAL); + } + } +} + +RID MaterialStorage::material_allocate() { + return material_owner.allocate_rid(); +} + +void MaterialStorage::material_initialize(RID p_rid) { + material_owner.initialize_rid(p_rid); + Material *material = material_owner.get_or_null(p_rid); + material->self = p_rid; +} + +void MaterialStorage::material_free(RID p_rid) { + Material *material = material_owner.get_or_null(p_rid); + ERR_FAIL_COND(!material); + + material_set_shader(p_rid, RID()); //clean up shader + material->dependency.deleted_notify(p_rid); + + material_owner.free(p_rid); +} + +void MaterialStorage::material_set_shader(RID p_material, RID p_shader) { + Material *material = material_owner.get_or_null(p_material); + ERR_FAIL_COND(!material); + + if (material->data) { + memdelete(material->data); + material->data = nullptr; + } + + if (material->shader) { + material->shader->owners.erase(material); + material->shader = nullptr; + material->shader_type = SHADER_TYPE_MAX; + } + + if (p_shader.is_null()) { + material->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_MATERIAL); + material->shader_id = 0; + return; + } + + Shader *shader = get_shader(p_shader); + ERR_FAIL_COND(!shader); + material->shader = shader; + material->shader_type = shader->type; + material->shader_id = p_shader.get_local_index(); + shader->owners.insert(material); + + if (shader->type == SHADER_TYPE_MAX) { + return; + } + + ERR_FAIL_COND(shader->data == nullptr); + + material->data = material_data_request_func[shader->type](shader->data); + material->data->self = p_material; + material->data->set_next_pass(material->next_pass); + material->data->set_render_priority(material->priority); + //updating happens later + material->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_MATERIAL); + _material_queue_update(material, true, true); +} + +MaterialStorage::ShaderData *MaterialStorage::material_get_shader_data(RID p_material) { + const MaterialStorage::Material *material = MaterialStorage::get_singleton()->get_material(p_material); + if (material && material->shader && material->shader->data) { + return material->shader->data; + } + + return nullptr; +} + +void MaterialStorage::material_set_param(RID p_material, const StringName &p_param, const Variant &p_value) { + Material *material = material_owner.get_or_null(p_material); + ERR_FAIL_COND(!material); + + if (p_value.get_type() == Variant::NIL) { + material->params.erase(p_param); + } else { + ERR_FAIL_COND(p_value.get_type() == Variant::OBJECT); //object not allowed + material->params[p_param] = p_value; + } + + if (material->shader && material->shader->data) { //shader is valid + bool is_texture = material->shader->data->is_parameter_texture(p_param); + _material_queue_update(material, !is_texture, is_texture); + } else { + _material_queue_update(material, true, true); + } +} + +Variant MaterialStorage::material_get_param(RID p_material, const StringName &p_param) const { + Material *material = material_owner.get_or_null(p_material); + ERR_FAIL_COND_V(!material, Variant()); + if (material->params.has(p_param)) { + return material->params[p_param]; + } else { + return Variant(); + } +} + +void MaterialStorage::material_set_next_pass(RID p_material, RID p_next_material) { + Material *material = material_owner.get_or_null(p_material); + ERR_FAIL_COND(!material); + + if (material->next_pass == p_next_material) { + return; + } + + material->next_pass = p_next_material; + if (material->data) { + material->data->set_next_pass(p_next_material); + } + + material->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_MATERIAL); +} + +void MaterialStorage::material_set_render_priority(RID p_material, int priority) { + Material *material = material_owner.get_or_null(p_material); + ERR_FAIL_COND(!material); + material->priority = priority; + if (material->data) { + material->data->set_render_priority(priority); + } +} + +bool MaterialStorage::material_is_animated(RID p_material) { + Material *material = material_owner.get_or_null(p_material); + ERR_FAIL_COND_V(!material, false); + if (material->shader && material->shader->data) { + if (material->shader->data->is_animated()) { + return true; + } else if (material->next_pass.is_valid()) { + return material_is_animated(material->next_pass); + } + } + return false; //by default nothing is animated +} + +bool MaterialStorage::material_casts_shadows(RID p_material) { + Material *material = material_owner.get_or_null(p_material); + ERR_FAIL_COND_V(!material, true); + if (material->shader && material->shader->data) { + if (material->shader->data->casts_shadows()) { + return true; + } else if (material->next_pass.is_valid()) { + return material_casts_shadows(material->next_pass); + } + } + return true; //by default everything casts shadows +} + +void MaterialStorage::material_get_instance_shader_parameters(RID p_material, List<InstanceShaderParam> *r_parameters) { + Material *material = material_owner.get_or_null(p_material); + ERR_FAIL_COND(!material); + if (material->shader && material->shader->data) { + material->shader->data->get_instance_param_list(r_parameters); + + if (material->next_pass.is_valid()) { + material_get_instance_shader_parameters(material->next_pass, r_parameters); + } + } +} + +void MaterialStorage::material_update_dependency(RID p_material, DependencyTracker *p_instance) { + Material *material = material_owner.get_or_null(p_material); + ERR_FAIL_COND(!material); + p_instance->update_dependency(&material->dependency); + if (material->next_pass.is_valid()) { + material_update_dependency(material->next_pass, p_instance); + } +} + +void MaterialStorage::material_set_data_request_function(ShaderType p_shader_type, MaterialStorage::MaterialDataRequestFunction p_function) { + ERR_FAIL_INDEX(p_shader_type, SHADER_TYPE_MAX); + material_data_request_func[p_shader_type] = p_function; +} + +MaterialStorage::MaterialDataRequestFunction MaterialStorage::material_get_data_request_function(ShaderType p_shader_type) { + ERR_FAIL_INDEX_V(p_shader_type, SHADER_TYPE_MAX, nullptr); + return material_data_request_func[p_shader_type]; +} diff --git a/servers/rendering/renderer_rd/storage_rd/material_storage.h b/servers/rendering/renderer_rd/storage_rd/material_storage.h new file mode 100644 index 0000000000..fb25e56ed3 --- /dev/null +++ b/servers/rendering/renderer_rd/storage_rd/material_storage.h @@ -0,0 +1,435 @@ +/*************************************************************************/ +/* material_storage.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef MATERIAL_STORAGE_RD_H +#define MATERIAL_STORAGE_RD_H + +#include "core/math/projection.h" +#include "core/templates/local_vector.h" +#include "core/templates/rid_owner.h" +#include "core/templates/self_list.h" +#include "servers/rendering/shader_compiler.h" +#include "servers/rendering/shader_language.h" +#include "servers/rendering/storage/material_storage.h" +#include "servers/rendering/storage/utilities.h" + +namespace RendererRD { + +class MaterialStorage : public RendererMaterialStorage { +public: + enum ShaderType { + SHADER_TYPE_2D, + SHADER_TYPE_3D, + SHADER_TYPE_PARTICLES, + SHADER_TYPE_SKY, + SHADER_TYPE_FOG, + SHADER_TYPE_MAX + }; + + struct ShaderData { + virtual void set_code(const String &p_Code) = 0; + virtual void set_path_hint(const String &p_hint) = 0; + virtual void set_default_texture_parameter(const StringName &p_name, RID p_texture, int p_index) = 0; + virtual void get_shader_uniform_list(List<PropertyInfo> *p_param_list) const = 0; + + virtual void get_instance_param_list(List<RendererMaterialStorage::InstanceShaderParam> *p_param_list) const = 0; + virtual bool is_parameter_texture(const StringName &p_param) const = 0; + virtual bool is_animated() const = 0; + virtual bool casts_shadows() const = 0; + virtual Variant get_default_parameter(const StringName &p_parameter) const = 0; + virtual RS::ShaderNativeSourceCode get_native_source_code() const { return RS::ShaderNativeSourceCode(); } + + virtual ~ShaderData() {} + }; + + struct MaterialData { + void update_uniform_buffer(const HashMap<StringName, ShaderLanguage::ShaderNode::Uniform> &p_uniforms, const uint32_t *p_uniform_offsets, const HashMap<StringName, Variant> &p_parameters, uint8_t *p_buffer, uint32_t p_buffer_size, bool p_use_linear_color); + void update_textures(const HashMap<StringName, Variant> &p_parameters, const HashMap<StringName, HashMap<int, RID>> &p_default_textures, const Vector<ShaderCompiler::GeneratedCode::Texture> &p_texture_uniforms, RID *p_textures, bool p_use_linear_color); + + virtual void set_render_priority(int p_priority) = 0; + virtual void set_next_pass(RID p_pass) = 0; + virtual bool update_parameters(const HashMap<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) = 0; + virtual ~MaterialData(); + + //to be used internally by update_parameters, in the most common configuration of material parameters + bool update_parameters_uniform_set(const HashMap<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty, const HashMap<StringName, ShaderLanguage::ShaderNode::Uniform> &p_uniforms, const uint32_t *p_uniform_offsets, const Vector<ShaderCompiler::GeneratedCode::Texture> &p_texture_uniforms, const HashMap<StringName, HashMap<int, RID>> &p_default_texture_params, uint32_t p_ubo_size, RID &uniform_set, RID p_shader, uint32_t p_shader_uniform_set, bool p_use_linear_color, uint32_t p_barrier = RD::BARRIER_MASK_ALL); + void free_parameters_uniform_set(RID p_uniform_set); + + private: + friend class MaterialStorage; + + RID self; + List<RID>::Element *global_buffer_E = nullptr; + List<RID>::Element *global_texture_E = nullptr; + uint64_t global_textures_pass = 0; + HashMap<StringName, uint64_t> used_global_textures; + + //internally by update_parameters_uniform_set + Vector<uint8_t> ubo_data; + RID uniform_buffer; + Vector<RID> texture_cache; + }; + +private: + static MaterialStorage *singleton; + + /* Samplers */ + + RID default_rd_samplers[RS::CANVAS_ITEM_TEXTURE_FILTER_MAX][RS::CANVAS_ITEM_TEXTURE_REPEAT_MAX]; + RID custom_rd_samplers[RS::CANVAS_ITEM_TEXTURE_FILTER_MAX][RS::CANVAS_ITEM_TEXTURE_REPEAT_MAX]; + + /* Buffers */ + + RID quad_index_buffer; + RID quad_index_array; + + /* GLOBAL SHADER UNIFORM API */ + + struct GlobalShaderUniforms { + enum { + BUFFER_DIRTY_REGION_SIZE = 1024 + }; + struct Variable { + HashSet<RID> texture_materials; // materials using this + + RS::GlobalShaderParameterType type; + Variant value; + Variant override; + int32_t buffer_index; //for vectors + int32_t buffer_elements; //for vectors + }; + + HashMap<StringName, Variable> variables; + + struct Value { + float x; + float y; + float z; + float w; + }; + + struct ValueInt { + int32_t x; + int32_t y; + int32_t z; + int32_t w; + }; + + struct ValueUInt { + uint32_t x; + uint32_t y; + uint32_t z; + uint32_t w; + }; + + struct ValueUsage { + uint32_t elements = 0; + }; + + List<RID> materials_using_buffer; + List<RID> materials_using_texture; + + RID buffer; + Value *buffer_values = nullptr; + ValueUsage *buffer_usage = nullptr; + bool *buffer_dirty_regions = nullptr; + uint32_t buffer_dirty_region_count = 0; + + uint32_t buffer_size; + + bool must_update_texture_materials = false; + bool must_update_buffer_materials = false; + + HashMap<RID, int32_t> instance_buffer_pos; + } global_shader_uniforms; + + int32_t _global_shader_uniform_allocate(uint32_t p_elements); + void _global_shader_uniform_store_in_buffer(int32_t p_index, RS::GlobalShaderParameterType p_type, const Variant &p_value); + void _global_shader_uniform_mark_buffer_dirty(int32_t p_index, int32_t p_elements); + + /* SHADER API */ + + struct Material; + + struct Shader { + ShaderData *data = nullptr; + String code; + String path_hint; + ShaderType type; + HashMap<StringName, HashMap<int, RID>> default_texture_parameter; + HashSet<Material *> owners; + }; + + typedef ShaderData *(*ShaderDataRequestFunction)(); + ShaderDataRequestFunction shader_data_request_func[SHADER_TYPE_MAX]; + + mutable RID_Owner<Shader, true> shader_owner; + Shader *get_shader(RID p_rid) { return shader_owner.get_or_null(p_rid); } + + /* MATERIAL API */ + + typedef MaterialData *(*MaterialDataRequestFunction)(ShaderData *); + + struct Material { + RID self; + MaterialData *data = nullptr; + Shader *shader = nullptr; + //shortcut to shader data and type + ShaderType shader_type = SHADER_TYPE_MAX; + uint32_t shader_id = 0; + bool uniform_dirty = false; + bool texture_dirty = false; + HashMap<StringName, Variant> params; + int32_t priority = 0; + RID next_pass; + SelfList<Material> update_element; + + Dependency dependency; + + Material() : + update_element(this) {} + }; + + MaterialDataRequestFunction material_data_request_func[SHADER_TYPE_MAX]; + mutable RID_Owner<Material, true> material_owner; + Material *get_material(RID p_rid) { return material_owner.get_or_null(p_rid); }; + + SelfList<Material>::List material_update_list; + + static void _material_uniform_set_erased(void *p_material); + +public: + static MaterialStorage *get_singleton(); + + MaterialStorage(); + virtual ~MaterialStorage(); + + bool free(RID p_rid); + + /* Helpers */ + + static _FORCE_INLINE_ void store_transform(const Transform3D &p_mtx, float *p_array) { + p_array[0] = p_mtx.basis.rows[0][0]; + p_array[1] = p_mtx.basis.rows[1][0]; + p_array[2] = p_mtx.basis.rows[2][0]; + p_array[3] = 0; + p_array[4] = p_mtx.basis.rows[0][1]; + p_array[5] = p_mtx.basis.rows[1][1]; + p_array[6] = p_mtx.basis.rows[2][1]; + p_array[7] = 0; + p_array[8] = p_mtx.basis.rows[0][2]; + p_array[9] = p_mtx.basis.rows[1][2]; + p_array[10] = p_mtx.basis.rows[2][2]; + p_array[11] = 0; + p_array[12] = p_mtx.origin.x; + p_array[13] = p_mtx.origin.y; + p_array[14] = p_mtx.origin.z; + p_array[15] = 1; + } + + static _FORCE_INLINE_ void store_basis_3x4(const Basis &p_mtx, float *p_array) { + p_array[0] = p_mtx.rows[0][0]; + p_array[1] = p_mtx.rows[1][0]; + p_array[2] = p_mtx.rows[2][0]; + p_array[3] = 0; + p_array[4] = p_mtx.rows[0][1]; + p_array[5] = p_mtx.rows[1][1]; + p_array[6] = p_mtx.rows[2][1]; + p_array[7] = 0; + p_array[8] = p_mtx.rows[0][2]; + p_array[9] = p_mtx.rows[1][2]; + p_array[10] = p_mtx.rows[2][2]; + p_array[11] = 0; + } + + static _FORCE_INLINE_ void store_transform_3x3(const Basis &p_mtx, float *p_array) { + p_array[0] = p_mtx.rows[0][0]; + p_array[1] = p_mtx.rows[1][0]; + p_array[2] = p_mtx.rows[2][0]; + p_array[3] = 0; + p_array[4] = p_mtx.rows[0][1]; + p_array[5] = p_mtx.rows[1][1]; + p_array[6] = p_mtx.rows[2][1]; + p_array[7] = 0; + p_array[8] = p_mtx.rows[0][2]; + p_array[9] = p_mtx.rows[1][2]; + p_array[10] = p_mtx.rows[2][2]; + p_array[11] = 0; + } + + static _FORCE_INLINE_ void store_transform_transposed_3x4(const Transform3D &p_mtx, float *p_array) { + p_array[0] = p_mtx.basis.rows[0][0]; + p_array[1] = p_mtx.basis.rows[0][1]; + p_array[2] = p_mtx.basis.rows[0][2]; + p_array[3] = p_mtx.origin.x; + p_array[4] = p_mtx.basis.rows[1][0]; + p_array[5] = p_mtx.basis.rows[1][1]; + p_array[6] = p_mtx.basis.rows[1][2]; + p_array[7] = p_mtx.origin.y; + p_array[8] = p_mtx.basis.rows[2][0]; + p_array[9] = p_mtx.basis.rows[2][1]; + p_array[10] = p_mtx.basis.rows[2][2]; + p_array[11] = p_mtx.origin.z; + } + + static _FORCE_INLINE_ void store_camera(const Projection &p_mtx, float *p_array) { + for (int i = 0; i < 4; i++) { + for (int j = 0; j < 4; j++) { + p_array[i * 4 + j] = p_mtx.columns[i][j]; + } + } + } + + static _FORCE_INLINE_ void store_soft_shadow_kernel(const float *p_kernel, float *p_array) { + for (int i = 0; i < 128; i++) { + p_array[i] = p_kernel[i]; + } + } + + // http://andrewthall.org/papers/df64_qf128.pdf +#ifdef REAL_T_IS_DOUBLE + static _FORCE_INLINE_ void split_double(double a, float *ahi, float *alo) { + const double SPLITTER = (1 << 29) + 1; + double t = a * SPLITTER; + double thi = t - (t - a); + double tlo = a - thi; + *ahi = (float)thi; + *alo = (float)tlo; + } +#endif + + /* Samplers */ + + _FORCE_INLINE_ RID sampler_rd_get_default(RS::CanvasItemTextureFilter p_filter, RS::CanvasItemTextureRepeat p_repeat) { + return default_rd_samplers[p_filter][p_repeat]; + } + _FORCE_INLINE_ RID sampler_rd_get_custom(RS::CanvasItemTextureFilter p_filter, RS::CanvasItemTextureRepeat p_repeat) { + return custom_rd_samplers[p_filter][p_repeat]; + } + + void sampler_rd_configure_custom(float mipmap_bias); + + // void sampler_rd_set_default(float p_mipmap_bias); + + /* Buffers */ + + RID get_quad_index_array() { return quad_index_array; } + + /* GLOBAL SHADER UNIFORM API */ + + void _update_global_shader_uniforms(); + + virtual void global_shader_parameter_add(const StringName &p_name, RS::GlobalShaderParameterType p_type, const Variant &p_value) override; + virtual void global_shader_parameter_remove(const StringName &p_name) override; + virtual Vector<StringName> global_shader_parameter_get_list() const override; + + virtual void global_shader_parameter_set(const StringName &p_name, const Variant &p_value) override; + virtual void global_shader_parameter_set_override(const StringName &p_name, const Variant &p_value) override; + virtual Variant global_shader_parameter_get(const StringName &p_name) const override; + virtual RS::GlobalShaderParameterType global_shader_parameter_get_type(const StringName &p_name) const override; + RS::GlobalShaderParameterType global_shader_parameter_get_type_internal(const StringName &p_name) const; + + virtual void global_shader_parameters_load_settings(bool p_load_textures = true) override; + virtual void global_shader_parameters_clear() override; + + virtual int32_t global_shader_parameters_instance_allocate(RID p_instance) override; + virtual void global_shader_parameters_instance_free(RID p_instance) override; + virtual void global_shader_parameters_instance_update(RID p_instance, int p_index, const Variant &p_value, int p_flags_count = 0) override; + + RID global_shader_uniforms_get_storage_buffer() const; + + /* SHADER API */ + + bool owns_shader(RID p_rid) { return shader_owner.owns(p_rid); }; + + virtual RID shader_allocate() override; + virtual void shader_initialize(RID p_shader) override; + virtual void shader_free(RID p_rid) override; + + virtual void shader_set_code(RID p_shader, const String &p_code) override; + virtual void shader_set_path_hint(RID p_shader, const String &p_path) override; + virtual String shader_get_code(RID p_shader) const override; + virtual void get_shader_parameter_list(RID p_shader, List<PropertyInfo> *p_param_list) const override; + + virtual void shader_set_default_texture_parameter(RID p_shader, const StringName &p_name, RID p_texture, int p_index) override; + virtual RID shader_get_default_texture_parameter(RID p_shader, const StringName &p_name, int p_index) const override; + virtual Variant shader_get_parameter_default(RID p_shader, const StringName &p_param) const override; + void shader_set_data_request_function(ShaderType p_shader_type, ShaderDataRequestFunction p_function); + + virtual RS::ShaderNativeSourceCode shader_get_native_source_code(RID p_shader) const override; + + /* MATERIAL API */ + + bool owns_material(RID p_rid) { return material_owner.owns(p_rid); }; + + void _material_queue_update(Material *material, bool p_uniform, bool p_texture); + void _update_queued_materials(); + + virtual RID material_allocate() override; + virtual void material_initialize(RID p_material) override; + virtual void material_free(RID p_rid) override; + + virtual void material_set_shader(RID p_material, RID p_shader) override; + ShaderData *material_get_shader_data(RID p_material); + + virtual void material_set_param(RID p_material, const StringName &p_param, const Variant &p_value) override; + virtual Variant material_get_param(RID p_material, const StringName &p_param) const override; + + virtual void material_set_next_pass(RID p_material, RID p_next_material) override; + virtual void material_set_render_priority(RID p_material, int priority) override; + + virtual bool material_is_animated(RID p_material) override; + virtual bool material_casts_shadows(RID p_material) override; + + virtual void material_get_instance_shader_parameters(RID p_material, List<InstanceShaderParam> *r_parameters) override; + + virtual void material_update_dependency(RID p_material, DependencyTracker *p_instance) override; + + void material_set_data_request_function(ShaderType p_shader_type, MaterialDataRequestFunction p_function); + MaterialDataRequestFunction material_get_data_request_function(ShaderType p_shader_type); + + _FORCE_INLINE_ uint32_t material_get_shader_id(RID p_material) { + Material *material = material_owner.get_or_null(p_material); + return material->shader_id; + } + + _FORCE_INLINE_ MaterialData *material_get_data(RID p_material, ShaderType p_shader_type) { + Material *material = material_owner.get_or_null(p_material); + if (!material || material->shader_type != p_shader_type) { + return nullptr; + } else { + return material->data; + } + } +}; + +} // namespace RendererRD + +#endif // MATERIAL_STORAGE_RD_H diff --git a/servers/rendering/renderer_rd/storage_rd/mesh_storage.cpp b/servers/rendering/renderer_rd/storage_rd/mesh_storage.cpp new file mode 100644 index 0000000000..1e74d31383 --- /dev/null +++ b/servers/rendering/renderer_rd/storage_rd/mesh_storage.cpp @@ -0,0 +1,2074 @@ +/*************************************************************************/ +/* mesh_storage.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "mesh_storage.h" +#include "../../rendering_server_globals.h" + +using namespace RendererRD; + +MeshStorage *MeshStorage::singleton = nullptr; + +MeshStorage *MeshStorage::get_singleton() { + return singleton; +} + +MeshStorage::MeshStorage() { + singleton = this; + + default_rd_storage_buffer = RD::get_singleton()->storage_buffer_create(sizeof(uint32_t) * 4); + + //default rd buffers + { + Vector<uint8_t> buffer; + { + buffer.resize(sizeof(float) * 3); + { + uint8_t *w = buffer.ptrw(); + float *fptr = reinterpret_cast<float *>(w); + fptr[0] = 0.0; + fptr[1] = 0.0; + fptr[2] = 0.0; + } + mesh_default_rd_buffers[DEFAULT_RD_BUFFER_VERTEX] = RD::get_singleton()->vertex_buffer_create(buffer.size(), buffer); + } + + { //normal + buffer.resize(sizeof(float) * 3); + { + uint8_t *w = buffer.ptrw(); + float *fptr = reinterpret_cast<float *>(w); + fptr[0] = 1.0; + fptr[1] = 0.0; + fptr[2] = 0.0; + } + mesh_default_rd_buffers[DEFAULT_RD_BUFFER_NORMAL] = RD::get_singleton()->vertex_buffer_create(buffer.size(), buffer); + } + + { //tangent + buffer.resize(sizeof(float) * 4); + { + uint8_t *w = buffer.ptrw(); + float *fptr = reinterpret_cast<float *>(w); + fptr[0] = 1.0; + fptr[1] = 0.0; + fptr[2] = 0.0; + fptr[3] = 0.0; + } + mesh_default_rd_buffers[DEFAULT_RD_BUFFER_TANGENT] = RD::get_singleton()->vertex_buffer_create(buffer.size(), buffer); + } + + { //color + buffer.resize(sizeof(float) * 4); + { + uint8_t *w = buffer.ptrw(); + float *fptr = reinterpret_cast<float *>(w); + fptr[0] = 1.0; + fptr[1] = 1.0; + fptr[2] = 1.0; + fptr[3] = 1.0; + } + mesh_default_rd_buffers[DEFAULT_RD_BUFFER_COLOR] = RD::get_singleton()->vertex_buffer_create(buffer.size(), buffer); + } + + { //tex uv 1 + buffer.resize(sizeof(float) * 2); + { + uint8_t *w = buffer.ptrw(); + float *fptr = reinterpret_cast<float *>(w); + fptr[0] = 0.0; + fptr[1] = 0.0; + } + mesh_default_rd_buffers[DEFAULT_RD_BUFFER_TEX_UV] = RD::get_singleton()->vertex_buffer_create(buffer.size(), buffer); + } + { //tex uv 2 + buffer.resize(sizeof(float) * 2); + { + uint8_t *w = buffer.ptrw(); + float *fptr = reinterpret_cast<float *>(w); + fptr[0] = 0.0; + fptr[1] = 0.0; + } + mesh_default_rd_buffers[DEFAULT_RD_BUFFER_TEX_UV2] = RD::get_singleton()->vertex_buffer_create(buffer.size(), buffer); + } + + for (int i = 0; i < RS::ARRAY_CUSTOM_COUNT; i++) { + buffer.resize(sizeof(float) * 4); + { + uint8_t *w = buffer.ptrw(); + float *fptr = reinterpret_cast<float *>(w); + fptr[0] = 0.0; + fptr[1] = 0.0; + fptr[2] = 0.0; + fptr[3] = 0.0; + } + mesh_default_rd_buffers[DEFAULT_RD_BUFFER_CUSTOM0 + i] = RD::get_singleton()->vertex_buffer_create(buffer.size(), buffer); + } + + { //bones + buffer.resize(sizeof(uint32_t) * 4); + { + uint8_t *w = buffer.ptrw(); + uint32_t *fptr = reinterpret_cast<uint32_t *>(w); + fptr[0] = 0; + fptr[1] = 0; + fptr[2] = 0; + fptr[3] = 0; + } + mesh_default_rd_buffers[DEFAULT_RD_BUFFER_BONES] = RD::get_singleton()->vertex_buffer_create(buffer.size(), buffer); + } + + { //weights + buffer.resize(sizeof(float) * 4); + { + uint8_t *w = buffer.ptrw(); + float *fptr = reinterpret_cast<float *>(w); + fptr[0] = 0.0; + fptr[1] = 0.0; + fptr[2] = 0.0; + fptr[3] = 0.0; + } + mesh_default_rd_buffers[DEFAULT_RD_BUFFER_WEIGHTS] = RD::get_singleton()->vertex_buffer_create(buffer.size(), buffer); + } + } + + { + Vector<String> skeleton_modes; + skeleton_modes.push_back("\n#define MODE_2D\n"); + skeleton_modes.push_back(""); + + skeleton_shader.shader.initialize(skeleton_modes); + skeleton_shader.version = skeleton_shader.shader.version_create(); + for (int i = 0; i < SkeletonShader::SHADER_MODE_MAX; i++) { + skeleton_shader.version_shader[i] = skeleton_shader.shader.version_get_shader(skeleton_shader.version, i); + skeleton_shader.pipeline[i] = RD::get_singleton()->compute_pipeline_create(skeleton_shader.version_shader[i]); + } + + { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.binding = 0; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.append_id(default_rd_storage_buffer); + uniforms.push_back(u); + } + skeleton_shader.default_skeleton_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, skeleton_shader.version_shader[0], SkeletonShader::UNIFORM_SET_SKELETON); + } + } +} + +MeshStorage::~MeshStorage() { + //def buffers + for (int i = 0; i < DEFAULT_RD_BUFFER_MAX; i++) { + RD::get_singleton()->free(mesh_default_rd_buffers[i]); + } + + skeleton_shader.shader.version_free(skeleton_shader.version); + + RD::get_singleton()->free(default_rd_storage_buffer); + + singleton = nullptr; +} + +bool MeshStorage::free(RID p_rid) { + if (owns_mesh(p_rid)) { + mesh_free(p_rid); + return true; + } else if (owns_mesh_instance(p_rid)) { + mesh_instance_free(p_rid); + return true; + } else if (owns_multimesh(p_rid)) { + multimesh_free(p_rid); + return true; + } else if (owns_skeleton(p_rid)) { + skeleton_free(p_rid); + return true; + } + + return false; +} + +/* MESH API */ + +RID MeshStorage::mesh_allocate() { + return mesh_owner.allocate_rid(); +} + +void MeshStorage::mesh_initialize(RID p_rid) { + mesh_owner.initialize_rid(p_rid, Mesh()); +} + +void MeshStorage::mesh_free(RID p_rid) { + mesh_clear(p_rid); + mesh_set_shadow_mesh(p_rid, RID()); + Mesh *mesh = mesh_owner.get_or_null(p_rid); + ERR_FAIL_COND(!mesh); + + mesh->dependency.deleted_notify(p_rid); + if (mesh->instances.size()) { + ERR_PRINT("deleting mesh with active instances"); + } + if (mesh->shadow_owners.size()) { + for (Mesh *E : mesh->shadow_owners) { + Mesh *shadow_owner = E; + shadow_owner->shadow_mesh = RID(); + shadow_owner->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_MESH); + } + } + mesh_owner.free(p_rid); +} + +void MeshStorage::mesh_set_blend_shape_count(RID p_mesh, int p_blend_shape_count) { + ERR_FAIL_COND(p_blend_shape_count < 0); + + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND(!mesh); + + ERR_FAIL_COND(mesh->surface_count > 0); //surfaces already exist + + mesh->blend_shape_count = p_blend_shape_count; +} + +/// Returns stride +void MeshStorage::mesh_add_surface(RID p_mesh, const RS::SurfaceData &p_surface) { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND(!mesh); + + ERR_FAIL_COND(mesh->surface_count == RS::MAX_MESH_SURFACES); + +#ifdef DEBUG_ENABLED + //do a validation, to catch errors first + { + uint32_t stride = 0; + uint32_t attrib_stride = 0; + uint32_t skin_stride = 0; + + for (int i = 0; i < RS::ARRAY_WEIGHTS; i++) { + if ((p_surface.format & (1 << i))) { + switch (i) { + case RS::ARRAY_VERTEX: { + if (p_surface.format & RS::ARRAY_FLAG_USE_2D_VERTICES) { + stride += sizeof(float) * 2; + } else { + stride += sizeof(float) * 3; + } + + } break; + case RS::ARRAY_NORMAL: { + stride += sizeof(int32_t); + + } break; + case RS::ARRAY_TANGENT: { + stride += sizeof(int32_t); + + } break; + case RS::ARRAY_COLOR: { + attrib_stride += sizeof(uint32_t); + } break; + case RS::ARRAY_TEX_UV: { + attrib_stride += sizeof(float) * 2; + + } break; + case RS::ARRAY_TEX_UV2: { + attrib_stride += sizeof(float) * 2; + + } break; + case RS::ARRAY_CUSTOM0: + case RS::ARRAY_CUSTOM1: + case RS::ARRAY_CUSTOM2: + case RS::ARRAY_CUSTOM3: { + int idx = i - RS::ARRAY_CUSTOM0; + const uint32_t fmt_shift[RS::ARRAY_CUSTOM_COUNT] = { RS::ARRAY_FORMAT_CUSTOM0_SHIFT, RS::ARRAY_FORMAT_CUSTOM1_SHIFT, RS::ARRAY_FORMAT_CUSTOM2_SHIFT, RS::ARRAY_FORMAT_CUSTOM3_SHIFT }; + uint32_t fmt = (p_surface.format >> fmt_shift[idx]) & RS::ARRAY_FORMAT_CUSTOM_MASK; + const uint32_t fmtsize[RS::ARRAY_CUSTOM_MAX] = { 4, 4, 4, 8, 4, 8, 12, 16 }; + attrib_stride += fmtsize[fmt]; + + } break; + case RS::ARRAY_WEIGHTS: + case RS::ARRAY_BONES: { + //uses a separate array + bool use_8 = p_surface.format & RS::ARRAY_FLAG_USE_8_BONE_WEIGHTS; + skin_stride += sizeof(int16_t) * (use_8 ? 16 : 8); + } break; + } + } + } + + int expected_size = stride * p_surface.vertex_count; + ERR_FAIL_COND_MSG(expected_size != p_surface.vertex_data.size(), "Size of vertex data provided (" + itos(p_surface.vertex_data.size()) + ") does not match expected (" + itos(expected_size) + ")"); + + int bs_expected_size = expected_size * mesh->blend_shape_count; + + ERR_FAIL_COND_MSG(bs_expected_size != p_surface.blend_shape_data.size(), "Size of blend shape data provided (" + itos(p_surface.blend_shape_data.size()) + ") does not match expected (" + itos(bs_expected_size) + ")"); + + int expected_attrib_size = attrib_stride * p_surface.vertex_count; + ERR_FAIL_COND_MSG(expected_attrib_size != p_surface.attribute_data.size(), "Size of attribute data provided (" + itos(p_surface.attribute_data.size()) + ") does not match expected (" + itos(expected_attrib_size) + ")"); + + if ((p_surface.format & RS::ARRAY_FORMAT_WEIGHTS) && (p_surface.format & RS::ARRAY_FORMAT_BONES)) { + expected_size = skin_stride * p_surface.vertex_count; + ERR_FAIL_COND_MSG(expected_size != p_surface.skin_data.size(), "Size of skin data provided (" + itos(p_surface.skin_data.size()) + ") does not match expected (" + itos(expected_size) + ")"); + } + } + +#endif + + Mesh::Surface *s = memnew(Mesh::Surface); + + s->format = p_surface.format; + s->primitive = p_surface.primitive; + + bool use_as_storage = (p_surface.skin_data.size() || mesh->blend_shape_count > 0); + + if (p_surface.vertex_data.size()) { + s->vertex_buffer = RD::get_singleton()->vertex_buffer_create(p_surface.vertex_data.size(), p_surface.vertex_data, use_as_storage); + s->vertex_buffer_size = p_surface.vertex_data.size(); + } + + if (p_surface.attribute_data.size()) { + s->attribute_buffer = RD::get_singleton()->vertex_buffer_create(p_surface.attribute_data.size(), p_surface.attribute_data); + } + if (p_surface.skin_data.size()) { + s->skin_buffer = RD::get_singleton()->vertex_buffer_create(p_surface.skin_data.size(), p_surface.skin_data, use_as_storage); + s->skin_buffer_size = p_surface.skin_data.size(); + } + + s->vertex_count = p_surface.vertex_count; + + if (p_surface.format & RS::ARRAY_FORMAT_BONES) { + mesh->has_bone_weights = true; + } + + if (p_surface.index_count) { + bool is_index_16 = p_surface.vertex_count <= 65536 && p_surface.vertex_count > 0; + + s->index_buffer = RD::get_singleton()->index_buffer_create(p_surface.index_count, is_index_16 ? RD::INDEX_BUFFER_FORMAT_UINT16 : RD::INDEX_BUFFER_FORMAT_UINT32, p_surface.index_data, false); + s->index_count = p_surface.index_count; + s->index_array = RD::get_singleton()->index_array_create(s->index_buffer, 0, s->index_count); + if (p_surface.lods.size()) { + s->lods = memnew_arr(Mesh::Surface::LOD, p_surface.lods.size()); + s->lod_count = p_surface.lods.size(); + + for (int i = 0; i < p_surface.lods.size(); i++) { + uint32_t indices = p_surface.lods[i].index_data.size() / (is_index_16 ? 2 : 4); + s->lods[i].index_buffer = RD::get_singleton()->index_buffer_create(indices, is_index_16 ? RD::INDEX_BUFFER_FORMAT_UINT16 : RD::INDEX_BUFFER_FORMAT_UINT32, p_surface.lods[i].index_data); + s->lods[i].index_array = RD::get_singleton()->index_array_create(s->lods[i].index_buffer, 0, indices); + s->lods[i].edge_length = p_surface.lods[i].edge_length; + s->lods[i].index_count = indices; + } + } + } + + ERR_FAIL_COND_MSG(!p_surface.index_count && !p_surface.vertex_count, "Meshes must contain a vertex array, an index array, or both"); + + s->aabb = p_surface.aabb; + s->bone_aabbs = p_surface.bone_aabbs; //only really useful for returning them. + + if (mesh->blend_shape_count > 0) { + s->blend_shape_buffer = RD::get_singleton()->storage_buffer_create(p_surface.blend_shape_data.size(), p_surface.blend_shape_data); + } + + if (use_as_storage) { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.binding = 0; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + if (s->vertex_buffer.is_valid()) { + u.append_id(s->vertex_buffer); + } else { + u.append_id(default_rd_storage_buffer); + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 1; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + if (s->skin_buffer.is_valid()) { + u.append_id(s->skin_buffer); + } else { + u.append_id(default_rd_storage_buffer); + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 2; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + if (s->blend_shape_buffer.is_valid()) { + u.append_id(s->blend_shape_buffer); + } else { + u.append_id(default_rd_storage_buffer); + } + uniforms.push_back(u); + } + + s->uniform_set = RD::get_singleton()->uniform_set_create(uniforms, skeleton_shader.version_shader[0], SkeletonShader::UNIFORM_SET_SURFACE); + } + + if (mesh->surface_count == 0) { + mesh->bone_aabbs = p_surface.bone_aabbs; + mesh->aabb = p_surface.aabb; + } else { + if (mesh->bone_aabbs.size() < p_surface.bone_aabbs.size()) { + // ArrayMesh::_surface_set_data only allocates bone_aabbs up to max_bone + // Each surface may affect different numbers of bones. + mesh->bone_aabbs.resize(p_surface.bone_aabbs.size()); + } + for (int i = 0; i < p_surface.bone_aabbs.size(); i++) { + const AABB &bone = p_surface.bone_aabbs[i]; + if (bone.has_volume()) { + mesh->bone_aabbs.write[i].merge_with(bone); + } + } + mesh->aabb.merge_with(p_surface.aabb); + } + + s->material = p_surface.material; + + mesh->surfaces = (Mesh::Surface **)memrealloc(mesh->surfaces, sizeof(Mesh::Surface *) * (mesh->surface_count + 1)); + mesh->surfaces[mesh->surface_count] = s; + mesh->surface_count++; + + for (MeshInstance *mi : mesh->instances) { + _mesh_instance_add_surface(mi, mesh, mesh->surface_count - 1); + } + + mesh->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_MESH); + + for (Mesh *E : mesh->shadow_owners) { + Mesh *shadow_owner = E; + shadow_owner->shadow_mesh = RID(); + shadow_owner->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_MESH); + } + + mesh->material_cache.clear(); +} + +int MeshStorage::mesh_get_blend_shape_count(RID p_mesh) const { + const Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND_V(!mesh, -1); + return mesh->blend_shape_count; +} + +void MeshStorage::mesh_set_blend_shape_mode(RID p_mesh, RS::BlendShapeMode p_mode) { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND(!mesh); + ERR_FAIL_INDEX((int)p_mode, 2); + + mesh->blend_shape_mode = p_mode; +} + +RS::BlendShapeMode MeshStorage::mesh_get_blend_shape_mode(RID p_mesh) const { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND_V(!mesh, RS::BLEND_SHAPE_MODE_NORMALIZED); + return mesh->blend_shape_mode; +} + +void MeshStorage::mesh_surface_update_vertex_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data) { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND(!mesh); + ERR_FAIL_UNSIGNED_INDEX((uint32_t)p_surface, mesh->surface_count); + ERR_FAIL_COND(p_data.size() == 0); + ERR_FAIL_COND(mesh->surfaces[p_surface]->vertex_buffer.is_null()); + uint64_t data_size = p_data.size(); + const uint8_t *r = p_data.ptr(); + + RD::get_singleton()->buffer_update(mesh->surfaces[p_surface]->vertex_buffer, p_offset, data_size, r); +} + +void MeshStorage::mesh_surface_update_attribute_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data) { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND(!mesh); + ERR_FAIL_UNSIGNED_INDEX((uint32_t)p_surface, mesh->surface_count); + ERR_FAIL_COND(p_data.size() == 0); + ERR_FAIL_COND(mesh->surfaces[p_surface]->attribute_buffer.is_null()); + uint64_t data_size = p_data.size(); + const uint8_t *r = p_data.ptr(); + + RD::get_singleton()->buffer_update(mesh->surfaces[p_surface]->attribute_buffer, p_offset, data_size, r); +} + +void MeshStorage::mesh_surface_update_skin_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data) { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND(!mesh); + ERR_FAIL_UNSIGNED_INDEX((uint32_t)p_surface, mesh->surface_count); + ERR_FAIL_COND(p_data.size() == 0); + ERR_FAIL_COND(mesh->surfaces[p_surface]->skin_buffer.is_null()); + uint64_t data_size = p_data.size(); + const uint8_t *r = p_data.ptr(); + + RD::get_singleton()->buffer_update(mesh->surfaces[p_surface]->skin_buffer, p_offset, data_size, r); +} + +void MeshStorage::mesh_surface_set_material(RID p_mesh, int p_surface, RID p_material) { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND(!mesh); + ERR_FAIL_UNSIGNED_INDEX((uint32_t)p_surface, mesh->surface_count); + mesh->surfaces[p_surface]->material = p_material; + + mesh->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_MATERIAL); + mesh->material_cache.clear(); +} + +RID MeshStorage::mesh_surface_get_material(RID p_mesh, int p_surface) const { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND_V(!mesh, RID()); + ERR_FAIL_UNSIGNED_INDEX_V((uint32_t)p_surface, mesh->surface_count, RID()); + + return mesh->surfaces[p_surface]->material; +} + +RS::SurfaceData MeshStorage::mesh_get_surface(RID p_mesh, int p_surface) const { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND_V(!mesh, RS::SurfaceData()); + ERR_FAIL_UNSIGNED_INDEX_V((uint32_t)p_surface, mesh->surface_count, RS::SurfaceData()); + + Mesh::Surface &s = *mesh->surfaces[p_surface]; + + RS::SurfaceData sd; + sd.format = s.format; + if (s.vertex_buffer.is_valid()) { + sd.vertex_data = RD::get_singleton()->buffer_get_data(s.vertex_buffer); + } + if (s.attribute_buffer.is_valid()) { + sd.attribute_data = RD::get_singleton()->buffer_get_data(s.attribute_buffer); + } + if (s.skin_buffer.is_valid()) { + sd.skin_data = RD::get_singleton()->buffer_get_data(s.skin_buffer); + } + sd.vertex_count = s.vertex_count; + sd.index_count = s.index_count; + sd.primitive = s.primitive; + + if (sd.index_count) { + sd.index_data = RD::get_singleton()->buffer_get_data(s.index_buffer); + } + sd.aabb = s.aabb; + for (uint32_t i = 0; i < s.lod_count; i++) { + RS::SurfaceData::LOD lod; + lod.edge_length = s.lods[i].edge_length; + lod.index_data = RD::get_singleton()->buffer_get_data(s.lods[i].index_buffer); + sd.lods.push_back(lod); + } + + sd.bone_aabbs = s.bone_aabbs; + + if (s.blend_shape_buffer.is_valid()) { + sd.blend_shape_data = RD::get_singleton()->buffer_get_data(s.blend_shape_buffer); + } + + return sd; +} + +int MeshStorage::mesh_get_surface_count(RID p_mesh) const { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND_V(!mesh, 0); + return mesh->surface_count; +} + +void MeshStorage::mesh_set_custom_aabb(RID p_mesh, const AABB &p_aabb) { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND(!mesh); + mesh->custom_aabb = p_aabb; +} + +AABB MeshStorage::mesh_get_custom_aabb(RID p_mesh) const { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND_V(!mesh, AABB()); + return mesh->custom_aabb; +} + +AABB MeshStorage::mesh_get_aabb(RID p_mesh, RID p_skeleton) { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND_V(!mesh, AABB()); + + if (mesh->custom_aabb != AABB()) { + return mesh->custom_aabb; + } + + Skeleton *skeleton = skeleton_owner.get_or_null(p_skeleton); + + if (!skeleton || skeleton->size == 0) { + return mesh->aabb; + } + + AABB aabb; + + for (uint32_t i = 0; i < mesh->surface_count; i++) { + AABB laabb; + if ((mesh->surfaces[i]->format & RS::ARRAY_FORMAT_BONES) && mesh->surfaces[i]->bone_aabbs.size()) { + int bs = mesh->surfaces[i]->bone_aabbs.size(); + const AABB *skbones = mesh->surfaces[i]->bone_aabbs.ptr(); + + int sbs = skeleton->size; + ERR_CONTINUE(bs > sbs); + const float *baseptr = skeleton->data.ptr(); + + bool first = true; + + if (skeleton->use_2d) { + for (int j = 0; j < bs; j++) { + if (skbones[0].size == Vector3()) { + continue; //bone is unused + } + + const float *dataptr = baseptr + j * 8; + + Transform3D mtx; + + mtx.basis.rows[0].x = dataptr[0]; + mtx.basis.rows[1].x = dataptr[1]; + mtx.origin.x = dataptr[3]; + + mtx.basis.rows[0].y = dataptr[4]; + mtx.basis.rows[1].y = dataptr[5]; + mtx.origin.y = dataptr[7]; + + AABB baabb = mtx.xform(skbones[j]); + + if (first) { + laabb = baabb; + first = false; + } else { + laabb.merge_with(baabb); + } + } + } else { + for (int j = 0; j < bs; j++) { + if (skbones[0].size == Vector3()) { + continue; //bone is unused + } + + const float *dataptr = baseptr + j * 12; + + Transform3D mtx; + + mtx.basis.rows[0][0] = dataptr[0]; + mtx.basis.rows[0][1] = dataptr[1]; + mtx.basis.rows[0][2] = dataptr[2]; + mtx.origin.x = dataptr[3]; + mtx.basis.rows[1][0] = dataptr[4]; + mtx.basis.rows[1][1] = dataptr[5]; + mtx.basis.rows[1][2] = dataptr[6]; + mtx.origin.y = dataptr[7]; + mtx.basis.rows[2][0] = dataptr[8]; + mtx.basis.rows[2][1] = dataptr[9]; + mtx.basis.rows[2][2] = dataptr[10]; + mtx.origin.z = dataptr[11]; + + AABB baabb = mtx.xform(skbones[j]); + if (first) { + laabb = baabb; + first = false; + } else { + laabb.merge_with(baabb); + } + } + } + + if (laabb.size == Vector3()) { + laabb = mesh->surfaces[i]->aabb; + } + } else { + laabb = mesh->surfaces[i]->aabb; + } + + if (i == 0) { + aabb = laabb; + } else { + aabb.merge_with(laabb); + } + } + + return aabb; +} + +void MeshStorage::mesh_set_shadow_mesh(RID p_mesh, RID p_shadow_mesh) { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND(!mesh); + + Mesh *shadow_mesh = mesh_owner.get_or_null(mesh->shadow_mesh); + if (shadow_mesh) { + shadow_mesh->shadow_owners.erase(mesh); + } + mesh->shadow_mesh = p_shadow_mesh; + + shadow_mesh = mesh_owner.get_or_null(mesh->shadow_mesh); + + if (shadow_mesh) { + shadow_mesh->shadow_owners.insert(mesh); + } + + mesh->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_MESH); +} + +void MeshStorage::mesh_clear(RID p_mesh) { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND(!mesh); + for (uint32_t i = 0; i < mesh->surface_count; i++) { + Mesh::Surface &s = *mesh->surfaces[i]; + if (s.vertex_buffer.is_valid()) { + RD::get_singleton()->free(s.vertex_buffer); //clears arrays as dependency automatically, including all versions + } + if (s.attribute_buffer.is_valid()) { + RD::get_singleton()->free(s.attribute_buffer); + } + if (s.skin_buffer.is_valid()) { + RD::get_singleton()->free(s.skin_buffer); + } + if (s.versions) { + memfree(s.versions); //reallocs, so free with memfree. + } + + if (s.index_buffer.is_valid()) { + RD::get_singleton()->free(s.index_buffer); + } + + if (s.lod_count) { + for (uint32_t j = 0; j < s.lod_count; j++) { + RD::get_singleton()->free(s.lods[j].index_buffer); + } + memdelete_arr(s.lods); + } + + if (s.blend_shape_buffer.is_valid()) { + RD::get_singleton()->free(s.blend_shape_buffer); + } + + memdelete(mesh->surfaces[i]); + } + if (mesh->surfaces) { + memfree(mesh->surfaces); + } + + mesh->surfaces = nullptr; + mesh->surface_count = 0; + mesh->material_cache.clear(); + //clear instance data + for (MeshInstance *mi : mesh->instances) { + _mesh_instance_clear(mi); + } + mesh->has_bone_weights = false; + mesh->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_MESH); + + for (Mesh *E : mesh->shadow_owners) { + Mesh *shadow_owner = E; + shadow_owner->shadow_mesh = RID(); + shadow_owner->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_MESH); + } +} + +bool MeshStorage::mesh_needs_instance(RID p_mesh, bool p_has_skeleton) { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND_V(!mesh, false); + + return mesh->blend_shape_count > 0 || (mesh->has_bone_weights && p_has_skeleton); +} + +Dependency *MeshStorage::mesh_get_dependency(RID p_mesh) const { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND_V(!mesh, nullptr); + + return &mesh->dependency; +} + +/* MESH INSTANCE */ + +RID MeshStorage::mesh_instance_create(RID p_base) { + Mesh *mesh = mesh_owner.get_or_null(p_base); + ERR_FAIL_COND_V(!mesh, RID()); + + RID rid = mesh_instance_owner.make_rid(); + MeshInstance *mi = mesh_instance_owner.get_or_null(rid); + + mi->mesh = mesh; + + for (uint32_t i = 0; i < mesh->surface_count; i++) { + _mesh_instance_add_surface(mi, mesh, i); + } + + mi->I = mesh->instances.push_back(mi); + + mi->dirty = true; + + return rid; +} + +void MeshStorage::mesh_instance_free(RID p_rid) { + MeshInstance *mi = mesh_instance_owner.get_or_null(p_rid); + _mesh_instance_clear(mi); + mi->mesh->instances.erase(mi->I); + mi->I = nullptr; + + mesh_instance_owner.free(p_rid); +} + +void MeshStorage::mesh_instance_set_skeleton(RID p_mesh_instance, RID p_skeleton) { + MeshInstance *mi = mesh_instance_owner.get_or_null(p_mesh_instance); + if (mi->skeleton == p_skeleton) { + return; + } + mi->skeleton = p_skeleton; + mi->skeleton_version = 0; + mi->dirty = true; +} + +void MeshStorage::mesh_instance_set_blend_shape_weight(RID p_mesh_instance, int p_shape, float p_weight) { + MeshInstance *mi = mesh_instance_owner.get_or_null(p_mesh_instance); + ERR_FAIL_COND(!mi); + ERR_FAIL_INDEX(p_shape, (int)mi->blend_weights.size()); + mi->blend_weights[p_shape] = p_weight; + mi->weights_dirty = true; + //will be eventually updated +} + +void MeshStorage::_mesh_instance_clear(MeshInstance *mi) { + for (uint32_t i = 0; i < mi->surfaces.size(); i++) { + if (mi->surfaces[i].versions) { + for (uint32_t j = 0; j < mi->surfaces[i].version_count; j++) { + RD::get_singleton()->free(mi->surfaces[i].versions[j].vertex_array); + } + memfree(mi->surfaces[i].versions); + } + if (mi->surfaces[i].vertex_buffer.is_valid()) { + RD::get_singleton()->free(mi->surfaces[i].vertex_buffer); + } + } + mi->surfaces.clear(); + + if (mi->blend_weights_buffer.is_valid()) { + RD::get_singleton()->free(mi->blend_weights_buffer); + } + mi->blend_weights.clear(); + mi->weights_dirty = false; + mi->skeleton_version = 0; +} + +void MeshStorage::_mesh_instance_add_surface(MeshInstance *mi, Mesh *mesh, uint32_t p_surface) { + if (mesh->blend_shape_count > 0 && mi->blend_weights_buffer.is_null()) { + mi->blend_weights.resize(mesh->blend_shape_count); + for (uint32_t i = 0; i < mi->blend_weights.size(); i++) { + mi->blend_weights[i] = 0; + } + mi->blend_weights_buffer = RD::get_singleton()->storage_buffer_create(sizeof(float) * mi->blend_weights.size(), mi->blend_weights.to_byte_array()); + mi->weights_dirty = true; + } + + MeshInstance::Surface s; + if ((mesh->blend_shape_count > 0 || (mesh->surfaces[p_surface]->format & RS::ARRAY_FORMAT_BONES)) && mesh->surfaces[p_surface]->vertex_buffer_size > 0) { + //surface warrants transform + s.vertex_buffer = RD::get_singleton()->vertex_buffer_create(mesh->surfaces[p_surface]->vertex_buffer_size, Vector<uint8_t>(), true); + + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.binding = 1; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.append_id(s.vertex_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 2; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + if (mi->blend_weights_buffer.is_valid()) { + u.append_id(mi->blend_weights_buffer); + } else { + u.append_id(default_rd_storage_buffer); + } + uniforms.push_back(u); + } + s.uniform_set = RD::get_singleton()->uniform_set_create(uniforms, skeleton_shader.version_shader[0], SkeletonShader::UNIFORM_SET_INSTANCE); + } + + mi->surfaces.push_back(s); + mi->dirty = true; +} + +void MeshStorage::mesh_instance_check_for_update(RID p_mesh_instance) { + MeshInstance *mi = mesh_instance_owner.get_or_null(p_mesh_instance); + + bool needs_update = mi->dirty; + + if (mi->weights_dirty && !mi->weight_update_list.in_list()) { + dirty_mesh_instance_weights.add(&mi->weight_update_list); + needs_update = true; + } + + if (mi->array_update_list.in_list()) { + return; + } + + if (!needs_update && mi->skeleton.is_valid()) { + Skeleton *sk = skeleton_owner.get_or_null(mi->skeleton); + if (sk && sk->version != mi->skeleton_version) { + needs_update = true; + } + } + + if (needs_update) { + dirty_mesh_instance_arrays.add(&mi->array_update_list); + } +} + +void MeshStorage::update_mesh_instances() { + while (dirty_mesh_instance_weights.first()) { + MeshInstance *mi = dirty_mesh_instance_weights.first()->self(); + + if (mi->blend_weights_buffer.is_valid()) { + RD::get_singleton()->buffer_update(mi->blend_weights_buffer, 0, mi->blend_weights.size() * sizeof(float), mi->blend_weights.ptr()); + } + dirty_mesh_instance_weights.remove(&mi->weight_update_list); + mi->weights_dirty = false; + } + if (dirty_mesh_instance_arrays.first() == nullptr) { + return; //nothing to do + } + + //process skeletons and blend shapes + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + + while (dirty_mesh_instance_arrays.first()) { + MeshInstance *mi = dirty_mesh_instance_arrays.first()->self(); + + Skeleton *sk = skeleton_owner.get_or_null(mi->skeleton); + + for (uint32_t i = 0; i < mi->surfaces.size(); i++) { + if (mi->surfaces[i].uniform_set == RID() || mi->mesh->surfaces[i]->uniform_set == RID()) { + continue; + } + + bool array_is_2d = mi->mesh->surfaces[i]->format & RS::ARRAY_FLAG_USE_2D_VERTICES; + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, skeleton_shader.pipeline[array_is_2d ? SkeletonShader::SHADER_MODE_2D : SkeletonShader::SHADER_MODE_3D]); + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, mi->surfaces[i].uniform_set, SkeletonShader::UNIFORM_SET_INSTANCE); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, mi->mesh->surfaces[i]->uniform_set, SkeletonShader::UNIFORM_SET_SURFACE); + if (sk && sk->uniform_set_mi.is_valid()) { + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, sk->uniform_set_mi, SkeletonShader::UNIFORM_SET_SKELETON); + } else { + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, skeleton_shader.default_skeleton_uniform_set, SkeletonShader::UNIFORM_SET_SKELETON); + } + + SkeletonShader::PushConstant push_constant; + + push_constant.has_normal = mi->mesh->surfaces[i]->format & RS::ARRAY_FORMAT_NORMAL; + push_constant.has_tangent = mi->mesh->surfaces[i]->format & RS::ARRAY_FORMAT_TANGENT; + push_constant.has_skeleton = sk != nullptr && sk->use_2d == array_is_2d && (mi->mesh->surfaces[i]->format & RS::ARRAY_FORMAT_BONES); + push_constant.has_blend_shape = mi->mesh->blend_shape_count > 0; + + push_constant.vertex_count = mi->mesh->surfaces[i]->vertex_count; + push_constant.vertex_stride = (mi->mesh->surfaces[i]->vertex_buffer_size / mi->mesh->surfaces[i]->vertex_count) / 4; + push_constant.skin_stride = (mi->mesh->surfaces[i]->skin_buffer_size / mi->mesh->surfaces[i]->vertex_count) / 4; + push_constant.skin_weight_offset = (mi->mesh->surfaces[i]->format & RS::ARRAY_FLAG_USE_8_BONE_WEIGHTS) ? 4 : 2; + + push_constant.blend_shape_count = mi->mesh->blend_shape_count; + push_constant.normalized_blend_shapes = mi->mesh->blend_shape_mode == RS::BLEND_SHAPE_MODE_NORMALIZED; + push_constant.pad0 = 0; + push_constant.pad1 = 0; + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SkeletonShader::PushConstant)); + + //dispatch without barrier, so all is done at the same time + RD::get_singleton()->compute_list_dispatch_threads(compute_list, push_constant.vertex_count, 1, 1); + } + + mi->dirty = false; + if (sk) { + mi->skeleton_version = sk->version; + } + dirty_mesh_instance_arrays.remove(&mi->array_update_list); + } + + RD::get_singleton()->compute_list_end(); +} + +void MeshStorage::_mesh_surface_generate_version_for_input_mask(Mesh::Surface::Version &v, Mesh::Surface *s, uint32_t p_input_mask, MeshInstance::Surface *mis) { + Vector<RD::VertexAttribute> attributes; + Vector<RID> buffers; + + uint32_t stride = 0; + uint32_t attribute_stride = 0; + uint32_t skin_stride = 0; + + for (int i = 0; i < RS::ARRAY_INDEX; i++) { + RD::VertexAttribute vd; + RID buffer; + vd.location = i; + + if (!(s->format & (1 << i))) { + // Not supplied by surface, use default value + buffer = mesh_default_rd_buffers[i]; + vd.stride = 0; + switch (i) { + case RS::ARRAY_VERTEX: { + vd.format = RD::DATA_FORMAT_R32G32B32_SFLOAT; + + } break; + case RS::ARRAY_NORMAL: { + vd.format = RD::DATA_FORMAT_R32G32B32_SFLOAT; + } break; + case RS::ARRAY_TANGENT: { + vd.format = RD::DATA_FORMAT_R32G32B32A32_SFLOAT; + } break; + case RS::ARRAY_COLOR: { + vd.format = RD::DATA_FORMAT_R32G32B32A32_SFLOAT; + + } break; + case RS::ARRAY_TEX_UV: { + vd.format = RD::DATA_FORMAT_R32G32_SFLOAT; + + } break; + case RS::ARRAY_TEX_UV2: { + vd.format = RD::DATA_FORMAT_R32G32_SFLOAT; + } break; + case RS::ARRAY_CUSTOM0: + case RS::ARRAY_CUSTOM1: + case RS::ARRAY_CUSTOM2: + case RS::ARRAY_CUSTOM3: { + //assumed weights too + vd.format = RD::DATA_FORMAT_R32G32B32A32_SFLOAT; + } break; + case RS::ARRAY_BONES: { + //assumed weights too + vd.format = RD::DATA_FORMAT_R32G32B32A32_UINT; + } break; + case RS::ARRAY_WEIGHTS: { + //assumed weights too + vd.format = RD::DATA_FORMAT_R32G32B32A32_SFLOAT; + } break; + } + } else { + //Supplied, use it + + vd.stride = 1; //mark that it needs a stride set (default uses 0) + + switch (i) { + case RS::ARRAY_VERTEX: { + vd.offset = stride; + + if (s->format & RS::ARRAY_FLAG_USE_2D_VERTICES) { + vd.format = RD::DATA_FORMAT_R32G32_SFLOAT; + stride += sizeof(float) * 2; + } else { + vd.format = RD::DATA_FORMAT_R32G32B32_SFLOAT; + stride += sizeof(float) * 3; + } + + if (mis) { + buffer = mis->vertex_buffer; + } else { + buffer = s->vertex_buffer; + } + + } break; + case RS::ARRAY_NORMAL: { + vd.offset = stride; + vd.format = RD::DATA_FORMAT_R16G16_UNORM; + stride += sizeof(uint16_t) * 2; + + if (mis) { + buffer = mis->vertex_buffer; + } else { + buffer = s->vertex_buffer; + } + } break; + case RS::ARRAY_TANGENT: { + vd.offset = stride; + vd.format = RD::DATA_FORMAT_R16G16_UNORM; + stride += sizeof(uint16_t) * 2; + + if (mis) { + buffer = mis->vertex_buffer; + } else { + buffer = s->vertex_buffer; + } + } break; + case RS::ARRAY_COLOR: { + vd.offset = attribute_stride; + + vd.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + attribute_stride += sizeof(int8_t) * 4; + buffer = s->attribute_buffer; + } break; + case RS::ARRAY_TEX_UV: { + vd.offset = attribute_stride; + + vd.format = RD::DATA_FORMAT_R32G32_SFLOAT; + attribute_stride += sizeof(float) * 2; + buffer = s->attribute_buffer; + + } break; + case RS::ARRAY_TEX_UV2: { + vd.offset = attribute_stride; + + vd.format = RD::DATA_FORMAT_R32G32_SFLOAT; + attribute_stride += sizeof(float) * 2; + buffer = s->attribute_buffer; + } break; + case RS::ARRAY_CUSTOM0: + case RS::ARRAY_CUSTOM1: + case RS::ARRAY_CUSTOM2: + case RS::ARRAY_CUSTOM3: { + vd.offset = attribute_stride; + + int idx = i - RS::ARRAY_CUSTOM0; + const uint32_t fmt_shift[RS::ARRAY_CUSTOM_COUNT] = { RS::ARRAY_FORMAT_CUSTOM0_SHIFT, RS::ARRAY_FORMAT_CUSTOM1_SHIFT, RS::ARRAY_FORMAT_CUSTOM2_SHIFT, RS::ARRAY_FORMAT_CUSTOM3_SHIFT }; + uint32_t fmt = (s->format >> fmt_shift[idx]) & RS::ARRAY_FORMAT_CUSTOM_MASK; + const uint32_t fmtsize[RS::ARRAY_CUSTOM_MAX] = { 4, 4, 4, 8, 4, 8, 12, 16 }; + const RD::DataFormat fmtrd[RS::ARRAY_CUSTOM_MAX] = { RD::DATA_FORMAT_R8G8B8A8_UNORM, RD::DATA_FORMAT_R8G8B8A8_SNORM, RD::DATA_FORMAT_R16G16_SFLOAT, RD::DATA_FORMAT_R16G16B16A16_SFLOAT, RD::DATA_FORMAT_R32_SFLOAT, RD::DATA_FORMAT_R32G32_SFLOAT, RD::DATA_FORMAT_R32G32B32_SFLOAT, RD::DATA_FORMAT_R32G32B32A32_SFLOAT }; + vd.format = fmtrd[fmt]; + attribute_stride += fmtsize[fmt]; + buffer = s->attribute_buffer; + } break; + case RS::ARRAY_BONES: { + vd.offset = skin_stride; + + vd.format = RD::DATA_FORMAT_R16G16B16A16_UINT; + skin_stride += sizeof(int16_t) * 4; + buffer = s->skin_buffer; + } break; + case RS::ARRAY_WEIGHTS: { + vd.offset = skin_stride; + + vd.format = RD::DATA_FORMAT_R16G16B16A16_UNORM; + skin_stride += sizeof(int16_t) * 4; + buffer = s->skin_buffer; + } break; + } + } + + if (!(p_input_mask & (1 << i))) { + continue; // Shader does not need this, skip it (but computing stride was important anyway) + } + + attributes.push_back(vd); + buffers.push_back(buffer); + } + + //update final stride + for (int i = 0; i < attributes.size(); i++) { + if (attributes[i].stride == 0) { + continue; //default location + } + int loc = attributes[i].location; + + if (loc < RS::ARRAY_COLOR) { + attributes.write[i].stride = stride; + } else if (loc < RS::ARRAY_BONES) { + attributes.write[i].stride = attribute_stride; + } else { + attributes.write[i].stride = skin_stride; + } + } + + v.input_mask = p_input_mask; + v.vertex_format = RD::get_singleton()->vertex_format_create(attributes); + v.vertex_array = RD::get_singleton()->vertex_array_create(s->vertex_count, v.vertex_format, buffers); +} + +////////////////// MULTIMESH + +RID MeshStorage::multimesh_allocate() { + return multimesh_owner.allocate_rid(); +} +void MeshStorage::multimesh_initialize(RID p_rid) { + multimesh_owner.initialize_rid(p_rid, MultiMesh()); +} + +void MeshStorage::multimesh_free(RID p_rid) { + _update_dirty_multimeshes(); + multimesh_allocate_data(p_rid, 0, RS::MULTIMESH_TRANSFORM_2D); + MultiMesh *multimesh = multimesh_owner.get_or_null(p_rid); + multimesh->dependency.deleted_notify(p_rid); + multimesh_owner.free(p_rid); +} + +void MeshStorage::multimesh_allocate_data(RID p_multimesh, int p_instances, RS::MultimeshTransformFormat p_transform_format, bool p_use_colors, bool p_use_custom_data) { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + ERR_FAIL_COND(!multimesh); + + if (multimesh->instances == p_instances && multimesh->xform_format == p_transform_format && multimesh->uses_colors == p_use_colors && multimesh->uses_custom_data == p_use_custom_data) { + return; + } + + if (multimesh->buffer.is_valid()) { + RD::get_singleton()->free(multimesh->buffer); + multimesh->buffer = RID(); + multimesh->uniform_set_2d = RID(); //cleared by dependency + multimesh->uniform_set_3d = RID(); //cleared by dependency + } + + if (multimesh->data_cache_dirty_regions) { + memdelete_arr(multimesh->data_cache_dirty_regions); + multimesh->data_cache_dirty_regions = nullptr; + multimesh->data_cache_dirty_region_count = 0; + } + + if (multimesh->previous_data_cache_dirty_regions) { + memdelete_arr(multimesh->previous_data_cache_dirty_regions); + multimesh->previous_data_cache_dirty_regions = nullptr; + multimesh->previous_data_cache_dirty_region_count = 0; + } + + multimesh->instances = p_instances; + multimesh->xform_format = p_transform_format; + multimesh->uses_colors = p_use_colors; + multimesh->color_offset_cache = p_transform_format == RS::MULTIMESH_TRANSFORM_2D ? 8 : 12; + multimesh->uses_custom_data = p_use_custom_data; + multimesh->custom_data_offset_cache = multimesh->color_offset_cache + (p_use_colors ? 4 : 0); + multimesh->stride_cache = multimesh->custom_data_offset_cache + (p_use_custom_data ? 4 : 0); + multimesh->buffer_set = false; + + //print_line("allocate, elements: " + itos(p_instances) + " 2D: " + itos(p_transform_format == RS::MULTIMESH_TRANSFORM_2D) + " colors " + itos(multimesh->uses_colors) + " data " + itos(multimesh->uses_custom_data) + " stride " + itos(multimesh->stride_cache) + " total size " + itos(multimesh->stride_cache * multimesh->instances)); + multimesh->data_cache = Vector<float>(); + multimesh->aabb = AABB(); + multimesh->aabb_dirty = false; + multimesh->visible_instances = MIN(multimesh->visible_instances, multimesh->instances); + multimesh->motion_vectors_current_offset = 0; + multimesh->motion_vectors_previous_offset = 0; + multimesh->motion_vectors_last_change = -1; + + if (multimesh->instances) { + uint32_t buffer_size = multimesh->instances * multimesh->stride_cache * sizeof(float); + if (multimesh->motion_vectors_enabled) { + buffer_size *= 2; + } + multimesh->buffer = RD::get_singleton()->storage_buffer_create(buffer_size); + } + + multimesh->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_MULTIMESH); +} + +bool MeshStorage::_multimesh_enable_motion_vectors(RID p_multimesh) { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + ERR_FAIL_COND_V(!multimesh, false); + + if (multimesh->motion_vectors_enabled) { + return false; + } + + multimesh->motion_vectors_enabled = true; + + multimesh->motion_vectors_current_offset = 0; + multimesh->motion_vectors_previous_offset = 0; + multimesh->motion_vectors_last_change = -1; + + if (!multimesh->data_cache.is_empty()) { + multimesh->data_cache.append_array(multimesh->data_cache); + } + + if (multimesh->buffer_set) { + RD::get_singleton()->barrier(); + Vector<uint8_t> buffer_data = RD::get_singleton()->buffer_get_data(multimesh->buffer); + if (!multimesh->data_cache.is_empty()) { + memcpy(buffer_data.ptrw(), multimesh->data_cache.ptr(), buffer_data.size()); + } + + RD::get_singleton()->free(multimesh->buffer); + uint32_t buffer_size = multimesh->instances * multimesh->stride_cache * sizeof(float) * 2; + multimesh->buffer = RD::get_singleton()->storage_buffer_create(buffer_size); + RD::get_singleton()->buffer_update(multimesh->buffer, 0, buffer_data.size(), buffer_data.ptr(), RD::BARRIER_MASK_NO_BARRIER); + RD::get_singleton()->buffer_update(multimesh->buffer, buffer_data.size(), buffer_data.size(), buffer_data.ptr()); + multimesh->uniform_set_3d = RID(); // Cleared by dependency + return true; + } + return false; // Update the transforms uniform set cache +} + +void MeshStorage::_multimesh_get_motion_vectors_offsets(RID p_multimesh, uint32_t &r_current_offset, uint32_t &r_prev_offset) { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + ERR_FAIL_COND(!multimesh); + r_current_offset = multimesh->motion_vectors_current_offset; + if (RSG::rasterizer->get_frame_number() - multimesh->motion_vectors_last_change >= 2) { + multimesh->motion_vectors_previous_offset = multimesh->motion_vectors_current_offset; + } + r_prev_offset = multimesh->motion_vectors_previous_offset; +} + +int MeshStorage::multimesh_get_instance_count(RID p_multimesh) const { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + ERR_FAIL_COND_V(!multimesh, 0); + return multimesh->instances; +} + +void MeshStorage::multimesh_set_mesh(RID p_multimesh, RID p_mesh) { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + ERR_FAIL_COND(!multimesh); + if (multimesh->mesh == p_mesh) { + return; + } + multimesh->mesh = p_mesh; + + if (multimesh->instances == 0) { + return; + } + + if (multimesh->data_cache.size()) { + //we have a data cache, just mark it dirt + _multimesh_mark_all_dirty(multimesh, false, true); + } else if (multimesh->instances) { + //need to re-create AABB unfortunately, calling this has a penalty + if (multimesh->buffer_set) { + Vector<uint8_t> buffer = RD::get_singleton()->buffer_get_data(multimesh->buffer); + const uint8_t *r = buffer.ptr() + multimesh->motion_vectors_current_offset * multimesh->stride_cache * sizeof(float); + const float *data = reinterpret_cast<const float *>(r); + _multimesh_re_create_aabb(multimesh, data, multimesh->instances); + } + } + + multimesh->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_MESH); +} + +#define MULTIMESH_DIRTY_REGION_SIZE 512 + +void MeshStorage::_multimesh_make_local(MultiMesh *multimesh) const { + if (multimesh->data_cache.size() > 0) { + return; //already local + } + + // this means that the user wants to load/save individual elements, + // for this, the data must reside on CPU, so just copy it there. + uint32_t buffer_size = multimesh->instances * multimesh->stride_cache; + if (multimesh->motion_vectors_enabled) { + buffer_size *= 2; + } + multimesh->data_cache.resize(buffer_size); + { + float *w = multimesh->data_cache.ptrw(); + + if (multimesh->buffer_set) { + Vector<uint8_t> buffer = RD::get_singleton()->buffer_get_data(multimesh->buffer); + { + const uint8_t *r = buffer.ptr(); + memcpy(w, r, buffer.size()); + } + } else { + memset(w, 0, buffer_size * sizeof(float)); + } + } + uint32_t data_cache_dirty_region_count = (multimesh->instances - 1) / MULTIMESH_DIRTY_REGION_SIZE + 1; + multimesh->data_cache_dirty_regions = memnew_arr(bool, data_cache_dirty_region_count); + memset(multimesh->data_cache_dirty_regions, 0, data_cache_dirty_region_count * sizeof(bool)); + multimesh->data_cache_dirty_region_count = 0; + + multimesh->previous_data_cache_dirty_regions = memnew_arr(bool, data_cache_dirty_region_count); + memset(multimesh->previous_data_cache_dirty_regions, 0, data_cache_dirty_region_count * sizeof(bool)); + multimesh->previous_data_cache_dirty_region_count = 0; +} + +void MeshStorage::_multimesh_update_motion_vectors_data_cache(MultiMesh *multimesh) { + ERR_FAIL_COND(multimesh->data_cache.is_empty()); + + if (!multimesh->motion_vectors_enabled) { + return; + } + + uint32_t frame = RSG::rasterizer->get_frame_number(); + if (multimesh->motion_vectors_last_change != frame) { + multimesh->motion_vectors_previous_offset = multimesh->motion_vectors_current_offset; + multimesh->motion_vectors_current_offset = multimesh->instances - multimesh->motion_vectors_current_offset; + multimesh->motion_vectors_last_change = frame; + + if (multimesh->previous_data_cache_dirty_region_count > 0) { + uint8_t *data = (uint8_t *)multimesh->data_cache.ptrw(); + uint32_t current_ofs = multimesh->motion_vectors_current_offset * multimesh->stride_cache * sizeof(float); + uint32_t previous_ofs = multimesh->motion_vectors_previous_offset * multimesh->stride_cache * sizeof(float); + uint32_t visible_instances = multimesh->visible_instances >= 0 ? multimesh->visible_instances : multimesh->instances; + uint32_t visible_region_count = visible_instances == 0 ? 0 : (visible_instances - 1) / MULTIMESH_DIRTY_REGION_SIZE + 1; + uint32_t region_size = multimesh->stride_cache * MULTIMESH_DIRTY_REGION_SIZE * sizeof(float); + uint32_t size = multimesh->stride_cache * (uint32_t)multimesh->instances * (uint32_t)sizeof(float); + for (uint32_t i = 0; i < visible_region_count; i++) { + if (multimesh->previous_data_cache_dirty_regions[i]) { + uint32_t offset = i * region_size; + memcpy(data + current_ofs + offset, data + previous_ofs + offset, MIN(region_size, size - offset)); + } + } + } + } +} + +void MeshStorage::_multimesh_mark_dirty(MultiMesh *multimesh, int p_index, bool p_aabb) { + uint32_t region_index = p_index / MULTIMESH_DIRTY_REGION_SIZE; +#ifdef DEBUG_ENABLED + uint32_t data_cache_dirty_region_count = (multimesh->instances - 1) / MULTIMESH_DIRTY_REGION_SIZE + 1; + ERR_FAIL_UNSIGNED_INDEX(region_index, data_cache_dirty_region_count); //bug +#endif + if (!multimesh->data_cache_dirty_regions[region_index]) { + multimesh->data_cache_dirty_regions[region_index] = true; + multimesh->data_cache_dirty_region_count++; + } + + if (p_aabb) { + multimesh->aabb_dirty = true; + } + + if (!multimesh->dirty) { + multimesh->dirty_list = multimesh_dirty_list; + multimesh_dirty_list = multimesh; + multimesh->dirty = true; + } +} + +void MeshStorage::_multimesh_mark_all_dirty(MultiMesh *multimesh, bool p_data, bool p_aabb) { + if (p_data) { + uint32_t data_cache_dirty_region_count = (multimesh->instances - 1) / MULTIMESH_DIRTY_REGION_SIZE + 1; + + for (uint32_t i = 0; i < data_cache_dirty_region_count; i++) { + if (!multimesh->data_cache_dirty_regions[i]) { + multimesh->data_cache_dirty_regions[i] = true; + multimesh->data_cache_dirty_region_count++; + } + } + } + + if (p_aabb) { + multimesh->aabb_dirty = true; + } + + if (!multimesh->dirty) { + multimesh->dirty_list = multimesh_dirty_list; + multimesh_dirty_list = multimesh; + multimesh->dirty = true; + } +} + +void MeshStorage::_multimesh_re_create_aabb(MultiMesh *multimesh, const float *p_data, int p_instances) { + ERR_FAIL_COND(multimesh->mesh.is_null()); + AABB aabb; + AABB mesh_aabb = mesh_get_aabb(multimesh->mesh); + for (int i = 0; i < p_instances; i++) { + const float *data = p_data + multimesh->stride_cache * i; + Transform3D t; + + if (multimesh->xform_format == RS::MULTIMESH_TRANSFORM_3D) { + t.basis.rows[0][0] = data[0]; + t.basis.rows[0][1] = data[1]; + t.basis.rows[0][2] = data[2]; + t.origin.x = data[3]; + t.basis.rows[1][0] = data[4]; + t.basis.rows[1][1] = data[5]; + t.basis.rows[1][2] = data[6]; + t.origin.y = data[7]; + t.basis.rows[2][0] = data[8]; + t.basis.rows[2][1] = data[9]; + t.basis.rows[2][2] = data[10]; + t.origin.z = data[11]; + + } else { + t.basis.rows[0].x = data[0]; + t.basis.rows[1].x = data[1]; + t.origin.x = data[3]; + + t.basis.rows[0].y = data[4]; + t.basis.rows[1].y = data[5]; + t.origin.y = data[7]; + } + + if (i == 0) { + aabb = t.xform(mesh_aabb); + } else { + aabb.merge_with(t.xform(mesh_aabb)); + } + } + + multimesh->aabb = aabb; +} + +void MeshStorage::multimesh_instance_set_transform(RID p_multimesh, int p_index, const Transform3D &p_transform) { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + ERR_FAIL_COND(!multimesh); + ERR_FAIL_INDEX(p_index, multimesh->instances); + ERR_FAIL_COND(multimesh->xform_format != RS::MULTIMESH_TRANSFORM_3D); + + _multimesh_make_local(multimesh); + _multimesh_update_motion_vectors_data_cache(multimesh); + + { + float *w = multimesh->data_cache.ptrw(); + + float *dataptr = w + (multimesh->motion_vectors_current_offset + p_index) * multimesh->stride_cache; + + dataptr[0] = p_transform.basis.rows[0][0]; + dataptr[1] = p_transform.basis.rows[0][1]; + dataptr[2] = p_transform.basis.rows[0][2]; + dataptr[3] = p_transform.origin.x; + dataptr[4] = p_transform.basis.rows[1][0]; + dataptr[5] = p_transform.basis.rows[1][1]; + dataptr[6] = p_transform.basis.rows[1][2]; + dataptr[7] = p_transform.origin.y; + dataptr[8] = p_transform.basis.rows[2][0]; + dataptr[9] = p_transform.basis.rows[2][1]; + dataptr[10] = p_transform.basis.rows[2][2]; + dataptr[11] = p_transform.origin.z; + } + + _multimesh_mark_dirty(multimesh, p_index, true); +} + +void MeshStorage::multimesh_instance_set_transform_2d(RID p_multimesh, int p_index, const Transform2D &p_transform) { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + ERR_FAIL_COND(!multimesh); + ERR_FAIL_INDEX(p_index, multimesh->instances); + ERR_FAIL_COND(multimesh->xform_format != RS::MULTIMESH_TRANSFORM_2D); + + _multimesh_make_local(multimesh); + _multimesh_update_motion_vectors_data_cache(multimesh); + + { + float *w = multimesh->data_cache.ptrw(); + + float *dataptr = w + (multimesh->motion_vectors_current_offset + p_index) * multimesh->stride_cache; + + dataptr[0] = p_transform.columns[0][0]; + dataptr[1] = p_transform.columns[1][0]; + dataptr[2] = 0; + dataptr[3] = p_transform.columns[2][0]; + dataptr[4] = p_transform.columns[0][1]; + dataptr[5] = p_transform.columns[1][1]; + dataptr[6] = 0; + dataptr[7] = p_transform.columns[2][1]; + } + + _multimesh_mark_dirty(multimesh, p_index, true); +} + +void MeshStorage::multimesh_instance_set_color(RID p_multimesh, int p_index, const Color &p_color) { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + ERR_FAIL_COND(!multimesh); + ERR_FAIL_INDEX(p_index, multimesh->instances); + ERR_FAIL_COND(!multimesh->uses_colors); + + _multimesh_make_local(multimesh); + _multimesh_update_motion_vectors_data_cache(multimesh); + + { + float *w = multimesh->data_cache.ptrw(); + + float *dataptr = w + (multimesh->motion_vectors_current_offset + p_index) * multimesh->stride_cache + multimesh->color_offset_cache; + + dataptr[0] = p_color.r; + dataptr[1] = p_color.g; + dataptr[2] = p_color.b; + dataptr[3] = p_color.a; + } + + _multimesh_mark_dirty(multimesh, p_index, false); +} + +void MeshStorage::multimesh_instance_set_custom_data(RID p_multimesh, int p_index, const Color &p_color) { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + ERR_FAIL_COND(!multimesh); + ERR_FAIL_INDEX(p_index, multimesh->instances); + ERR_FAIL_COND(!multimesh->uses_custom_data); + + _multimesh_make_local(multimesh); + _multimesh_update_motion_vectors_data_cache(multimesh); + + { + float *w = multimesh->data_cache.ptrw(); + + float *dataptr = w + (multimesh->motion_vectors_current_offset + p_index) * multimesh->stride_cache + multimesh->custom_data_offset_cache; + + dataptr[0] = p_color.r; + dataptr[1] = p_color.g; + dataptr[2] = p_color.b; + dataptr[3] = p_color.a; + } + + _multimesh_mark_dirty(multimesh, p_index, false); +} + +RID MeshStorage::multimesh_get_mesh(RID p_multimesh) const { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + ERR_FAIL_COND_V(!multimesh, RID()); + + return multimesh->mesh; +} + +Dependency *MeshStorage::multimesh_get_dependency(RID p_multimesh) const { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + ERR_FAIL_COND_V(!multimesh, nullptr); + + return &multimesh->dependency; +} + +Transform3D MeshStorage::multimesh_instance_get_transform(RID p_multimesh, int p_index) const { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + ERR_FAIL_COND_V(!multimesh, Transform3D()); + ERR_FAIL_INDEX_V(p_index, multimesh->instances, Transform3D()); + ERR_FAIL_COND_V(multimesh->xform_format != RS::MULTIMESH_TRANSFORM_3D, Transform3D()); + + _multimesh_make_local(multimesh); + + Transform3D t; + { + const float *r = multimesh->data_cache.ptr(); + + const float *dataptr = r + (multimesh->motion_vectors_current_offset + p_index) * multimesh->stride_cache; + + t.basis.rows[0][0] = dataptr[0]; + t.basis.rows[0][1] = dataptr[1]; + t.basis.rows[0][2] = dataptr[2]; + t.origin.x = dataptr[3]; + t.basis.rows[1][0] = dataptr[4]; + t.basis.rows[1][1] = dataptr[5]; + t.basis.rows[1][2] = dataptr[6]; + t.origin.y = dataptr[7]; + t.basis.rows[2][0] = dataptr[8]; + t.basis.rows[2][1] = dataptr[9]; + t.basis.rows[2][2] = dataptr[10]; + t.origin.z = dataptr[11]; + } + + return t; +} + +Transform2D MeshStorage::multimesh_instance_get_transform_2d(RID p_multimesh, int p_index) const { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + ERR_FAIL_COND_V(!multimesh, Transform2D()); + ERR_FAIL_INDEX_V(p_index, multimesh->instances, Transform2D()); + ERR_FAIL_COND_V(multimesh->xform_format != RS::MULTIMESH_TRANSFORM_2D, Transform2D()); + + _multimesh_make_local(multimesh); + + Transform2D t; + { + const float *r = multimesh->data_cache.ptr(); + + const float *dataptr = r + (multimesh->motion_vectors_current_offset + p_index) * multimesh->stride_cache; + + t.columns[0][0] = dataptr[0]; + t.columns[1][0] = dataptr[1]; + t.columns[2][0] = dataptr[3]; + t.columns[0][1] = dataptr[4]; + t.columns[1][1] = dataptr[5]; + t.columns[2][1] = dataptr[7]; + } + + return t; +} + +Color MeshStorage::multimesh_instance_get_color(RID p_multimesh, int p_index) const { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + ERR_FAIL_COND_V(!multimesh, Color()); + ERR_FAIL_INDEX_V(p_index, multimesh->instances, Color()); + ERR_FAIL_COND_V(!multimesh->uses_colors, Color()); + + _multimesh_make_local(multimesh); + + Color c; + { + const float *r = multimesh->data_cache.ptr(); + + const float *dataptr = r + (multimesh->motion_vectors_current_offset + p_index) * multimesh->stride_cache + multimesh->color_offset_cache; + + c.r = dataptr[0]; + c.g = dataptr[1]; + c.b = dataptr[2]; + c.a = dataptr[3]; + } + + return c; +} + +Color MeshStorage::multimesh_instance_get_custom_data(RID p_multimesh, int p_index) const { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + ERR_FAIL_COND_V(!multimesh, Color()); + ERR_FAIL_INDEX_V(p_index, multimesh->instances, Color()); + ERR_FAIL_COND_V(!multimesh->uses_custom_data, Color()); + + _multimesh_make_local(multimesh); + + Color c; + { + const float *r = multimesh->data_cache.ptr(); + + const float *dataptr = r + (multimesh->motion_vectors_current_offset + p_index) * multimesh->stride_cache + multimesh->custom_data_offset_cache; + + c.r = dataptr[0]; + c.g = dataptr[1]; + c.b = dataptr[2]; + c.a = dataptr[3]; + } + + return c; +} + +void MeshStorage::multimesh_set_buffer(RID p_multimesh, const Vector<float> &p_buffer) { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + ERR_FAIL_COND(!multimesh); + ERR_FAIL_COND(p_buffer.size() != (multimesh->instances * (int)multimesh->stride_cache)); + + if (multimesh->motion_vectors_enabled) { + uint32_t frame = RSG::rasterizer->get_frame_number(); + + if (multimesh->motion_vectors_last_change != frame) { + multimesh->motion_vectors_previous_offset = multimesh->motion_vectors_current_offset; + multimesh->motion_vectors_current_offset = multimesh->instances - multimesh->motion_vectors_current_offset; + multimesh->motion_vectors_last_change = frame; + } + } + + { + const float *r = p_buffer.ptr(); + RD::get_singleton()->buffer_update(multimesh->buffer, multimesh->motion_vectors_current_offset * multimesh->stride_cache * sizeof(float), p_buffer.size() * sizeof(float), r); + multimesh->buffer_set = true; + } + + if (multimesh->data_cache.size()) { + float *cache_data = multimesh->data_cache.ptrw(); + memcpy(cache_data + (multimesh->motion_vectors_current_offset * multimesh->stride_cache), p_buffer.ptr(), p_buffer.size() * sizeof(float)); + _multimesh_mark_all_dirty(multimesh, true, true); //update AABB + } else if (multimesh->mesh.is_valid()) { + //if we have a mesh set, we need to re-generate the AABB from the new data + const float *data = p_buffer.ptr(); + + _multimesh_re_create_aabb(multimesh, data, multimesh->instances); + multimesh->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); + } +} + +Vector<float> MeshStorage::multimesh_get_buffer(RID p_multimesh) const { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + ERR_FAIL_COND_V(!multimesh, Vector<float>()); + if (multimesh->buffer.is_null()) { + return Vector<float>(); + } else { + Vector<float> ret; + ret.resize(multimesh->instances * multimesh->stride_cache); + float *w = ret.ptrw(); + + if (multimesh->data_cache.size()) { + const uint8_t *r = (uint8_t *)multimesh->data_cache.ptr() + multimesh->motion_vectors_current_offset * multimesh->stride_cache * sizeof(float); + memcpy(w, r, ret.size() * sizeof(float)); + } else { + Vector<uint8_t> buffer = RD::get_singleton()->buffer_get_data(multimesh->buffer); + const uint8_t *r = buffer.ptr() + multimesh->motion_vectors_current_offset * multimesh->stride_cache * sizeof(float); + memcpy(w, r, ret.size() * sizeof(float)); + } + return ret; + } +} + +void MeshStorage::multimesh_set_visible_instances(RID p_multimesh, int p_visible) { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + ERR_FAIL_COND(!multimesh); + ERR_FAIL_COND(p_visible < -1 || p_visible > multimesh->instances); + if (multimesh->visible_instances == p_visible) { + return; + } + + if (multimesh->data_cache.size()) { + //there is a data cache.. + _multimesh_mark_all_dirty(multimesh, false, true); + } + + multimesh->visible_instances = p_visible; + + multimesh->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_MULTIMESH_VISIBLE_INSTANCES); +} + +int MeshStorage::multimesh_get_visible_instances(RID p_multimesh) const { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + ERR_FAIL_COND_V(!multimesh, 0); + return multimesh->visible_instances; +} + +AABB MeshStorage::multimesh_get_aabb(RID p_multimesh) const { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + ERR_FAIL_COND_V(!multimesh, AABB()); + if (multimesh->aabb_dirty) { + const_cast<MeshStorage *>(this)->_update_dirty_multimeshes(); + } + return multimesh->aabb; +} + +void MeshStorage::_update_dirty_multimeshes() { + while (multimesh_dirty_list) { + MultiMesh *multimesh = multimesh_dirty_list; + + if (multimesh->data_cache.size()) { //may have been cleared, so only process if it exists + + uint32_t visible_instances = multimesh->visible_instances >= 0 ? multimesh->visible_instances : multimesh->instances; + uint32_t buffer_offset = multimesh->motion_vectors_current_offset * multimesh->stride_cache; + const float *data = multimesh->data_cache.ptr() + buffer_offset; + + uint32_t total_dirty_regions = multimesh->data_cache_dirty_region_count + multimesh->previous_data_cache_dirty_region_count; + if (total_dirty_regions != 0) { + uint32_t data_cache_dirty_region_count = (multimesh->instances - 1) / MULTIMESH_DIRTY_REGION_SIZE + 1; + uint32_t visible_region_count = visible_instances == 0 ? 0 : (visible_instances - 1) / MULTIMESH_DIRTY_REGION_SIZE + 1; + + uint32_t region_size = multimesh->stride_cache * MULTIMESH_DIRTY_REGION_SIZE * sizeof(float); + if (total_dirty_regions > 32 || total_dirty_regions > visible_region_count / 2) { + //if there too many dirty regions, or represent the majority of regions, just copy all, else transfer cost piles up too much + RD::get_singleton()->buffer_update(multimesh->buffer, buffer_offset * sizeof(float), MIN(visible_region_count * region_size, multimesh->instances * (uint32_t)multimesh->stride_cache * (uint32_t)sizeof(float)), data); + } else { + //not that many regions? update them all + for (uint32_t i = 0; i < visible_region_count; i++) { + if (multimesh->data_cache_dirty_regions[i] || multimesh->previous_data_cache_dirty_regions[i]) { + uint32_t offset = i * region_size; + uint32_t size = multimesh->stride_cache * (uint32_t)multimesh->instances * (uint32_t)sizeof(float); + uint32_t region_start_index = multimesh->stride_cache * MULTIMESH_DIRTY_REGION_SIZE * i; + RD::get_singleton()->buffer_update(multimesh->buffer, buffer_offset * sizeof(float) + offset, MIN(region_size, size - offset), &data[region_start_index], RD::BARRIER_MASK_NO_BARRIER); + } + } + RD::get_singleton()->barrier(RD::BARRIER_MASK_NO_BARRIER, RD::BARRIER_MASK_ALL); + } + + memcpy(multimesh->previous_data_cache_dirty_regions, multimesh->data_cache_dirty_regions, data_cache_dirty_region_count * sizeof(bool)); + memset(multimesh->data_cache_dirty_regions, 0, data_cache_dirty_region_count * sizeof(bool)); + + multimesh->previous_data_cache_dirty_region_count = multimesh->data_cache_dirty_region_count; + multimesh->data_cache_dirty_region_count = 0; + } + + if (multimesh->aabb_dirty) { + //aabb is dirty.. + _multimesh_re_create_aabb(multimesh, data, visible_instances); + multimesh->aabb_dirty = false; + multimesh->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); + } + } + + multimesh_dirty_list = multimesh->dirty_list; + + multimesh->dirty_list = nullptr; + multimesh->dirty = false; + } + + multimesh_dirty_list = nullptr; +} + +/* SKELETON API */ + +RID MeshStorage::skeleton_allocate() { + return skeleton_owner.allocate_rid(); +} +void MeshStorage::skeleton_initialize(RID p_rid) { + skeleton_owner.initialize_rid(p_rid, Skeleton()); +} + +void MeshStorage::skeleton_free(RID p_rid) { + _update_dirty_skeletons(); + skeleton_allocate_data(p_rid, 0); + Skeleton *skeleton = skeleton_owner.get_or_null(p_rid); + skeleton->dependency.deleted_notify(p_rid); + skeleton_owner.free(p_rid); +} + +void MeshStorage::_skeleton_make_dirty(Skeleton *skeleton) { + if (!skeleton->dirty) { + skeleton->dirty = true; + skeleton->dirty_list = skeleton_dirty_list; + skeleton_dirty_list = skeleton; + } +} + +void MeshStorage::skeleton_allocate_data(RID p_skeleton, int p_bones, bool p_2d_skeleton) { + Skeleton *skeleton = skeleton_owner.get_or_null(p_skeleton); + ERR_FAIL_COND(!skeleton); + ERR_FAIL_COND(p_bones < 0); + + if (skeleton->size == p_bones && skeleton->use_2d == p_2d_skeleton) { + return; + } + + skeleton->size = p_bones; + skeleton->use_2d = p_2d_skeleton; + skeleton->uniform_set_3d = RID(); + + if (skeleton->buffer.is_valid()) { + RD::get_singleton()->free(skeleton->buffer); + skeleton->buffer = RID(); + skeleton->data.clear(); + skeleton->uniform_set_mi = RID(); + } + + if (skeleton->size) { + skeleton->data.resize(skeleton->size * (skeleton->use_2d ? 8 : 12)); + skeleton->buffer = RD::get_singleton()->storage_buffer_create(skeleton->data.size() * sizeof(float)); + memset(skeleton->data.ptrw(), 0, skeleton->data.size() * sizeof(float)); + + _skeleton_make_dirty(skeleton); + + { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.binding = 0; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.append_id(skeleton->buffer); + uniforms.push_back(u); + } + skeleton->uniform_set_mi = RD::get_singleton()->uniform_set_create(uniforms, skeleton_shader.version_shader[0], SkeletonShader::UNIFORM_SET_SKELETON); + } + } + + skeleton->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_SKELETON_DATA); +} + +int MeshStorage::skeleton_get_bone_count(RID p_skeleton) const { + Skeleton *skeleton = skeleton_owner.get_or_null(p_skeleton); + ERR_FAIL_COND_V(!skeleton, 0); + + return skeleton->size; +} + +void MeshStorage::skeleton_bone_set_transform(RID p_skeleton, int p_bone, const Transform3D &p_transform) { + Skeleton *skeleton = skeleton_owner.get_or_null(p_skeleton); + + ERR_FAIL_COND(!skeleton); + ERR_FAIL_INDEX(p_bone, skeleton->size); + ERR_FAIL_COND(skeleton->use_2d); + + float *dataptr = skeleton->data.ptrw() + p_bone * 12; + + dataptr[0] = p_transform.basis.rows[0][0]; + dataptr[1] = p_transform.basis.rows[0][1]; + dataptr[2] = p_transform.basis.rows[0][2]; + dataptr[3] = p_transform.origin.x; + dataptr[4] = p_transform.basis.rows[1][0]; + dataptr[5] = p_transform.basis.rows[1][1]; + dataptr[6] = p_transform.basis.rows[1][2]; + dataptr[7] = p_transform.origin.y; + dataptr[8] = p_transform.basis.rows[2][0]; + dataptr[9] = p_transform.basis.rows[2][1]; + dataptr[10] = p_transform.basis.rows[2][2]; + dataptr[11] = p_transform.origin.z; + + _skeleton_make_dirty(skeleton); +} + +Transform3D MeshStorage::skeleton_bone_get_transform(RID p_skeleton, int p_bone) const { + Skeleton *skeleton = skeleton_owner.get_or_null(p_skeleton); + + ERR_FAIL_COND_V(!skeleton, Transform3D()); + ERR_FAIL_INDEX_V(p_bone, skeleton->size, Transform3D()); + ERR_FAIL_COND_V(skeleton->use_2d, Transform3D()); + + const float *dataptr = skeleton->data.ptr() + p_bone * 12; + + Transform3D t; + + t.basis.rows[0][0] = dataptr[0]; + t.basis.rows[0][1] = dataptr[1]; + t.basis.rows[0][2] = dataptr[2]; + t.origin.x = dataptr[3]; + t.basis.rows[1][0] = dataptr[4]; + t.basis.rows[1][1] = dataptr[5]; + t.basis.rows[1][2] = dataptr[6]; + t.origin.y = dataptr[7]; + t.basis.rows[2][0] = dataptr[8]; + t.basis.rows[2][1] = dataptr[9]; + t.basis.rows[2][2] = dataptr[10]; + t.origin.z = dataptr[11]; + + return t; +} + +void MeshStorage::skeleton_bone_set_transform_2d(RID p_skeleton, int p_bone, const Transform2D &p_transform) { + Skeleton *skeleton = skeleton_owner.get_or_null(p_skeleton); + + ERR_FAIL_COND(!skeleton); + ERR_FAIL_INDEX(p_bone, skeleton->size); + ERR_FAIL_COND(!skeleton->use_2d); + + float *dataptr = skeleton->data.ptrw() + p_bone * 8; + + dataptr[0] = p_transform.columns[0][0]; + dataptr[1] = p_transform.columns[1][0]; + dataptr[2] = 0; + dataptr[3] = p_transform.columns[2][0]; + dataptr[4] = p_transform.columns[0][1]; + dataptr[5] = p_transform.columns[1][1]; + dataptr[6] = 0; + dataptr[7] = p_transform.columns[2][1]; + + _skeleton_make_dirty(skeleton); +} + +Transform2D MeshStorage::skeleton_bone_get_transform_2d(RID p_skeleton, int p_bone) const { + Skeleton *skeleton = skeleton_owner.get_or_null(p_skeleton); + + ERR_FAIL_COND_V(!skeleton, Transform2D()); + ERR_FAIL_INDEX_V(p_bone, skeleton->size, Transform2D()); + ERR_FAIL_COND_V(!skeleton->use_2d, Transform2D()); + + const float *dataptr = skeleton->data.ptr() + p_bone * 8; + + Transform2D t; + t.columns[0][0] = dataptr[0]; + t.columns[1][0] = dataptr[1]; + t.columns[2][0] = dataptr[3]; + t.columns[0][1] = dataptr[4]; + t.columns[1][1] = dataptr[5]; + t.columns[2][1] = dataptr[7]; + + return t; +} + +void MeshStorage::skeleton_set_base_transform_2d(RID p_skeleton, const Transform2D &p_base_transform) { + Skeleton *skeleton = skeleton_owner.get_or_null(p_skeleton); + + ERR_FAIL_NULL(skeleton); + ERR_FAIL_COND(!skeleton->use_2d); + + skeleton->base_transform_2d = p_base_transform; +} + +void MeshStorage::_update_dirty_skeletons() { + while (skeleton_dirty_list) { + Skeleton *skeleton = skeleton_dirty_list; + + if (skeleton->size) { + RD::get_singleton()->buffer_update(skeleton->buffer, 0, skeleton->data.size() * sizeof(float), skeleton->data.ptr()); + } + + skeleton_dirty_list = skeleton->dirty_list; + + skeleton->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_SKELETON_BONES); + + skeleton->version++; + + skeleton->dirty = false; + skeleton->dirty_list = nullptr; + } + + skeleton_dirty_list = nullptr; +} + +void MeshStorage::skeleton_update_dependency(RID p_skeleton, DependencyTracker *p_instance) { + Skeleton *skeleton = skeleton_owner.get_or_null(p_skeleton); + ERR_FAIL_COND(!skeleton); + + p_instance->update_dependency(&skeleton->dependency); +} diff --git a/servers/rendering/renderer_rd/storage_rd/mesh_storage.h b/servers/rendering/renderer_rd/storage_rd/mesh_storage.h new file mode 100644 index 0000000000..6a7400631d --- /dev/null +++ b/servers/rendering/renderer_rd/storage_rd/mesh_storage.h @@ -0,0 +1,705 @@ +/*************************************************************************/ +/* mesh_storage.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef MESH_STORAGE_RD_H +#define MESH_STORAGE_RD_H + +#include "core/templates/local_vector.h" +#include "core/templates/rid_owner.h" +#include "core/templates/self_list.h" +#include "servers/rendering/renderer_rd/shaders/skeleton.glsl.gen.h" +#include "servers/rendering/storage/mesh_storage.h" +#include "servers/rendering/storage/utilities.h" + +namespace RendererRD { + +class MeshStorage : public RendererMeshStorage { +public: + enum DefaultRDBuffer { + DEFAULT_RD_BUFFER_VERTEX, + DEFAULT_RD_BUFFER_NORMAL, + DEFAULT_RD_BUFFER_TANGENT, + DEFAULT_RD_BUFFER_COLOR, + DEFAULT_RD_BUFFER_TEX_UV, + DEFAULT_RD_BUFFER_TEX_UV2, + DEFAULT_RD_BUFFER_CUSTOM0, + DEFAULT_RD_BUFFER_CUSTOM1, + DEFAULT_RD_BUFFER_CUSTOM2, + DEFAULT_RD_BUFFER_CUSTOM3, + DEFAULT_RD_BUFFER_BONES, + DEFAULT_RD_BUFFER_WEIGHTS, + DEFAULT_RD_BUFFER_MAX, + }; + +private: + static MeshStorage *singleton; + + RID default_rd_storage_buffer; + + /* Mesh */ + + RID mesh_default_rd_buffers[DEFAULT_RD_BUFFER_MAX]; + + struct MeshInstance; + + struct Mesh { + struct Surface { + RS::PrimitiveType primitive = RS::PRIMITIVE_POINTS; + uint32_t format = 0; + + RID vertex_buffer; + RID attribute_buffer; + RID skin_buffer; + uint32_t vertex_count = 0; + uint32_t vertex_buffer_size = 0; + uint32_t skin_buffer_size = 0; + + // A different pipeline needs to be allocated + // depending on the inputs available in the + // material. + // There are never that many geometry/material + // combinations, so a simple array is the most + // cache-efficient structure. + + struct Version { + uint32_t input_mask = 0; + RD::VertexFormatID vertex_format = 0; + RID vertex_array; + }; + + SpinLock version_lock; //needed to access versions + Version *versions = nullptr; //allocated on demand + uint32_t version_count = 0; + + RID index_buffer; + RID index_array; + uint32_t index_count = 0; + + struct LOD { + float edge_length = 0.0; + uint32_t index_count = 0; + RID index_buffer; + RID index_array; + }; + + LOD *lods = nullptr; + uint32_t lod_count = 0; + + AABB aabb; + + Vector<AABB> bone_aabbs; + + RID blend_shape_buffer; + + RID material; + + uint32_t render_index = 0; + uint64_t render_pass = 0; + + uint32_t multimesh_render_index = 0; + uint64_t multimesh_render_pass = 0; + + uint32_t particles_render_index = 0; + uint64_t particles_render_pass = 0; + + RID uniform_set; + }; + + uint32_t blend_shape_count = 0; + RS::BlendShapeMode blend_shape_mode = RS::BLEND_SHAPE_MODE_NORMALIZED; + + Surface **surfaces = nullptr; + uint32_t surface_count = 0; + + Vector<AABB> bone_aabbs; + + bool has_bone_weights = false; + + AABB aabb; + AABB custom_aabb; + + Vector<RID> material_cache; + + List<MeshInstance *> instances; + + RID shadow_mesh; + HashSet<Mesh *> shadow_owners; + + Dependency dependency; + }; + + mutable RID_Owner<Mesh, true> mesh_owner; + + /* Mesh Instance API */ + + struct MeshInstance { + Mesh *mesh = nullptr; + RID skeleton; + struct Surface { + RID vertex_buffer; + RID uniform_set; + + Mesh::Surface::Version *versions = nullptr; //allocated on demand + uint32_t version_count = 0; + }; + LocalVector<Surface> surfaces; + LocalVector<float> blend_weights; + + RID blend_weights_buffer; + List<MeshInstance *>::Element *I = nullptr; //used to erase itself + uint64_t skeleton_version = 0; + bool dirty = false; + bool weights_dirty = false; + SelfList<MeshInstance> weight_update_list; + SelfList<MeshInstance> array_update_list; + MeshInstance() : + weight_update_list(this), array_update_list(this) {} + }; + + void _mesh_surface_generate_version_for_input_mask(Mesh::Surface::Version &v, Mesh::Surface *s, uint32_t p_input_mask, MeshInstance::Surface *mis = nullptr); + + void _mesh_instance_clear(MeshInstance *mi); + void _mesh_instance_add_surface(MeshInstance *mi, Mesh *mesh, uint32_t p_surface); + + mutable RID_Owner<MeshInstance> mesh_instance_owner; + + SelfList<MeshInstance>::List dirty_mesh_instance_weights; + SelfList<MeshInstance>::List dirty_mesh_instance_arrays; + + /* MultiMesh */ + + struct MultiMesh { + RID mesh; + int instances = 0; + RS::MultimeshTransformFormat xform_format = RS::MULTIMESH_TRANSFORM_3D; + bool uses_colors = false; + bool uses_custom_data = false; + int visible_instances = -1; + AABB aabb; + bool aabb_dirty = false; + bool buffer_set = false; + bool motion_vectors_enabled = false; + uint32_t motion_vectors_current_offset = 0; + uint32_t motion_vectors_previous_offset = 0; + uint64_t motion_vectors_last_change = -1; + uint32_t stride_cache = 0; + uint32_t color_offset_cache = 0; + uint32_t custom_data_offset_cache = 0; + + Vector<float> data_cache; //used if individual setting is used + bool *data_cache_dirty_regions = nullptr; + uint32_t data_cache_dirty_region_count = 0; + bool *previous_data_cache_dirty_regions = nullptr; + uint32_t previous_data_cache_dirty_region_count = 0; + + RID buffer; //storage buffer + RID uniform_set_3d; + RID uniform_set_2d; + + bool dirty = false; + MultiMesh *dirty_list = nullptr; + + Dependency dependency; + }; + + mutable RID_Owner<MultiMesh, true> multimesh_owner; + + MultiMesh *multimesh_dirty_list = nullptr; + + _FORCE_INLINE_ void _multimesh_make_local(MultiMesh *multimesh) const; + _FORCE_INLINE_ void _multimesh_update_motion_vectors_data_cache(MultiMesh *multimesh); + _FORCE_INLINE_ void _multimesh_mark_dirty(MultiMesh *multimesh, int p_index, bool p_aabb); + _FORCE_INLINE_ void _multimesh_mark_all_dirty(MultiMesh *multimesh, bool p_data, bool p_aabb); + _FORCE_INLINE_ void _multimesh_re_create_aabb(MultiMesh *multimesh, const float *p_data, int p_instances); + + /* Skeleton */ + + struct SkeletonShader { + struct PushConstant { + uint32_t has_normal; + uint32_t has_tangent; + uint32_t has_skeleton; + uint32_t has_blend_shape; + + uint32_t vertex_count; + uint32_t vertex_stride; + uint32_t skin_stride; + uint32_t skin_weight_offset; + + uint32_t blend_shape_count; + uint32_t normalized_blend_shapes; + uint32_t pad0; + uint32_t pad1; + }; + + enum { + UNIFORM_SET_INSTANCE = 0, + UNIFORM_SET_SURFACE = 1, + UNIFORM_SET_SKELETON = 2, + }; + enum { + SHADER_MODE_2D, + SHADER_MODE_3D, + SHADER_MODE_MAX + }; + + SkeletonShaderRD shader; + RID version; + RID version_shader[SHADER_MODE_MAX]; + RID pipeline[SHADER_MODE_MAX]; + + RID default_skeleton_uniform_set; + } skeleton_shader; + + struct Skeleton { + bool use_2d = false; + int size = 0; + Vector<float> data; + RID buffer; + + bool dirty = false; + Skeleton *dirty_list = nullptr; + Transform2D base_transform_2d; + + RID uniform_set_3d; + RID uniform_set_mi; + + uint64_t version = 1; + + Dependency dependency; + }; + + mutable RID_Owner<Skeleton, true> skeleton_owner; + + _FORCE_INLINE_ void _skeleton_make_dirty(Skeleton *skeleton); + + Skeleton *skeleton_dirty_list = nullptr; + +public: + static MeshStorage *get_singleton(); + + MeshStorage(); + virtual ~MeshStorage(); + + bool free(RID p_rid); + + RID get_default_rd_storage_buffer() const { return default_rd_storage_buffer; } + + /* MESH API */ + + bool owns_mesh(RID p_rid) { return mesh_owner.owns(p_rid); }; + + virtual RID mesh_allocate() override; + virtual void mesh_initialize(RID p_mesh) override; + virtual void mesh_free(RID p_rid) override; + + virtual void mesh_set_blend_shape_count(RID p_mesh, int p_blend_shape_count) override; + + /// Return stride + virtual void mesh_add_surface(RID p_mesh, const RS::SurfaceData &p_surface) override; + + virtual int mesh_get_blend_shape_count(RID p_mesh) const override; + + virtual void mesh_set_blend_shape_mode(RID p_mesh, RS::BlendShapeMode p_mode) override; + virtual RS::BlendShapeMode mesh_get_blend_shape_mode(RID p_mesh) const override; + + virtual void mesh_surface_update_vertex_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data) override; + virtual void mesh_surface_update_attribute_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data) override; + virtual void mesh_surface_update_skin_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data) override; + + virtual void mesh_surface_set_material(RID p_mesh, int p_surface, RID p_material) override; + virtual RID mesh_surface_get_material(RID p_mesh, int p_surface) const override; + + virtual RS::SurfaceData mesh_get_surface(RID p_mesh, int p_surface) const override; + + virtual int mesh_get_surface_count(RID p_mesh) const override; + + virtual void mesh_set_custom_aabb(RID p_mesh, const AABB &p_aabb) override; + virtual AABB mesh_get_custom_aabb(RID p_mesh) const override; + + virtual AABB mesh_get_aabb(RID p_mesh, RID p_skeleton = RID()) override; + virtual void mesh_set_shadow_mesh(RID p_mesh, RID p_shadow_mesh) override; + + virtual void mesh_clear(RID p_mesh) override; + + virtual bool mesh_needs_instance(RID p_mesh, bool p_has_skeleton) override; + + _FORCE_INLINE_ const RID *mesh_get_surface_count_and_materials(RID p_mesh, uint32_t &r_surface_count) { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND_V(!mesh, nullptr); + r_surface_count = mesh->surface_count; + if (r_surface_count == 0) { + return nullptr; + } + if (mesh->material_cache.is_empty()) { + mesh->material_cache.resize(mesh->surface_count); + for (uint32_t i = 0; i < r_surface_count; i++) { + mesh->material_cache.write[i] = mesh->surfaces[i]->material; + } + } + + return mesh->material_cache.ptr(); + } + + _FORCE_INLINE_ void *mesh_get_surface(RID p_mesh, uint32_t p_surface_index) { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND_V(!mesh, nullptr); + ERR_FAIL_UNSIGNED_INDEX_V(p_surface_index, mesh->surface_count, nullptr); + + return mesh->surfaces[p_surface_index]; + } + + _FORCE_INLINE_ RID mesh_get_shadow_mesh(RID p_mesh) { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + ERR_FAIL_COND_V(!mesh, RID()); + + return mesh->shadow_mesh; + } + + _FORCE_INLINE_ RS::PrimitiveType mesh_surface_get_primitive(void *p_surface) { + Mesh::Surface *surface = reinterpret_cast<Mesh::Surface *>(p_surface); + return surface->primitive; + } + + _FORCE_INLINE_ bool mesh_surface_has_lod(void *p_surface) const { + Mesh::Surface *s = reinterpret_cast<Mesh::Surface *>(p_surface); + return s->lod_count > 0; + } + + _FORCE_INLINE_ uint32_t mesh_surface_get_vertices_drawn_count(void *p_surface) const { + Mesh::Surface *s = reinterpret_cast<Mesh::Surface *>(p_surface); + return s->index_count ? s->index_count : s->vertex_count; + } + + _FORCE_INLINE_ uint32_t mesh_surface_get_lod(void *p_surface, float p_model_scale, float p_distance_threshold, float p_mesh_lod_threshold, uint32_t &r_index_count) const { + Mesh::Surface *s = reinterpret_cast<Mesh::Surface *>(p_surface); + + int32_t current_lod = -1; + r_index_count = s->index_count; + for (uint32_t i = 0; i < s->lod_count; i++) { + float screen_size = s->lods[i].edge_length * p_model_scale / p_distance_threshold; + if (screen_size > p_mesh_lod_threshold) { + break; + } + current_lod = i; + } + if (current_lod == -1) { + return 0; + } else { + r_index_count = s->lods[current_lod].index_count; + return current_lod + 1; + } + } + + _FORCE_INLINE_ RID mesh_surface_get_index_array(void *p_surface, uint32_t p_lod) const { + Mesh::Surface *s = reinterpret_cast<Mesh::Surface *>(p_surface); + + if (p_lod == 0) { + return s->index_array; + } else { + return s->lods[p_lod - 1].index_array; + } + } + + _FORCE_INLINE_ void mesh_surface_get_vertex_arrays_and_format(void *p_surface, uint32_t p_input_mask, RID &r_vertex_array_rd, RD::VertexFormatID &r_vertex_format) { + Mesh::Surface *s = reinterpret_cast<Mesh::Surface *>(p_surface); + + s->version_lock.lock(); + + //there will never be more than, at much, 3 or 4 versions, so iterating is the fastest way + + for (uint32_t i = 0; i < s->version_count; i++) { + if (s->versions[i].input_mask != p_input_mask) { + continue; + } + //we have this version, hooray + r_vertex_format = s->versions[i].vertex_format; + r_vertex_array_rd = s->versions[i].vertex_array; + s->version_lock.unlock(); + return; + } + + uint32_t version = s->version_count; + s->version_count++; + s->versions = (Mesh::Surface::Version *)memrealloc(s->versions, sizeof(Mesh::Surface::Version) * s->version_count); + + _mesh_surface_generate_version_for_input_mask(s->versions[version], s, p_input_mask); + + r_vertex_format = s->versions[version].vertex_format; + r_vertex_array_rd = s->versions[version].vertex_array; + + s->version_lock.unlock(); + } + + _FORCE_INLINE_ void mesh_instance_surface_get_vertex_arrays_and_format(RID p_mesh_instance, uint32_t p_surface_index, uint32_t p_input_mask, RID &r_vertex_array_rd, RD::VertexFormatID &r_vertex_format) { + MeshInstance *mi = mesh_instance_owner.get_or_null(p_mesh_instance); + ERR_FAIL_COND(!mi); + Mesh *mesh = mi->mesh; + ERR_FAIL_UNSIGNED_INDEX(p_surface_index, mesh->surface_count); + + MeshInstance::Surface *mis = &mi->surfaces[p_surface_index]; + Mesh::Surface *s = mesh->surfaces[p_surface_index]; + + s->version_lock.lock(); + + //there will never be more than, at much, 3 or 4 versions, so iterating is the fastest way + + for (uint32_t i = 0; i < mis->version_count; i++) { + if (mis->versions[i].input_mask != p_input_mask) { + continue; + } + //we have this version, hooray + r_vertex_format = mis->versions[i].vertex_format; + r_vertex_array_rd = mis->versions[i].vertex_array; + s->version_lock.unlock(); + return; + } + + uint32_t version = mis->version_count; + mis->version_count++; + mis->versions = (Mesh::Surface::Version *)memrealloc(mis->versions, sizeof(Mesh::Surface::Version) * mis->version_count); + + _mesh_surface_generate_version_for_input_mask(mis->versions[version], s, p_input_mask, mis); + + r_vertex_format = mis->versions[version].vertex_format; + r_vertex_array_rd = mis->versions[version].vertex_array; + + s->version_lock.unlock(); + } + + _FORCE_INLINE_ RID mesh_get_default_rd_buffer(DefaultRDBuffer p_buffer) { + ERR_FAIL_INDEX_V(p_buffer, DEFAULT_RD_BUFFER_MAX, RID()); + return mesh_default_rd_buffers[p_buffer]; + } + + _FORCE_INLINE_ uint32_t mesh_surface_get_render_pass_index(RID p_mesh, uint32_t p_surface_index, uint64_t p_render_pass, uint32_t *r_index) { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + Mesh::Surface *s = mesh->surfaces[p_surface_index]; + + if (s->render_pass != p_render_pass) { + (*r_index)++; + s->render_pass = p_render_pass; + s->render_index = *r_index; + } + + return s->render_index; + } + + _FORCE_INLINE_ uint32_t mesh_surface_get_multimesh_render_pass_index(RID p_mesh, uint32_t p_surface_index, uint64_t p_render_pass, uint32_t *r_index) { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + Mesh::Surface *s = mesh->surfaces[p_surface_index]; + + if (s->multimesh_render_pass != p_render_pass) { + (*r_index)++; + s->multimesh_render_pass = p_render_pass; + s->multimesh_render_index = *r_index; + } + + return s->multimesh_render_index; + } + + _FORCE_INLINE_ uint32_t mesh_surface_get_particles_render_pass_index(RID p_mesh, uint32_t p_surface_index, uint64_t p_render_pass, uint32_t *r_index) { + Mesh *mesh = mesh_owner.get_or_null(p_mesh); + Mesh::Surface *s = mesh->surfaces[p_surface_index]; + + if (s->particles_render_pass != p_render_pass) { + (*r_index)++; + s->particles_render_pass = p_render_pass; + s->particles_render_index = *r_index; + } + + return s->particles_render_index; + } + + Dependency *mesh_get_dependency(RID p_mesh) const; + + /* MESH INSTANCE API */ + + bool owns_mesh_instance(RID p_rid) const { return mesh_instance_owner.owns(p_rid); }; + + virtual RID mesh_instance_create(RID p_base) override; + virtual void mesh_instance_free(RID p_rid) override; + virtual void mesh_instance_set_skeleton(RID p_mesh_instance, RID p_skeleton) override; + virtual void mesh_instance_set_blend_shape_weight(RID p_mesh_instance, int p_shape, float p_weight) override; + virtual void mesh_instance_check_for_update(RID p_mesh_instance) override; + virtual void update_mesh_instances() override; + + /* MULTIMESH API */ + + bool owns_multimesh(RID p_rid) { return multimesh_owner.owns(p_rid); }; + + virtual RID multimesh_allocate() override; + virtual void multimesh_initialize(RID p_multimesh) override; + virtual void multimesh_free(RID p_rid) override; + + virtual void multimesh_allocate_data(RID p_multimesh, int p_instances, RS::MultimeshTransformFormat p_transform_format, bool p_use_colors = false, bool p_use_custom_data = false) override; + virtual int multimesh_get_instance_count(RID p_multimesh) const override; + + virtual void multimesh_set_mesh(RID p_multimesh, RID p_mesh) override; + virtual void multimesh_instance_set_transform(RID p_multimesh, int p_index, const Transform3D &p_transform) override; + virtual void multimesh_instance_set_transform_2d(RID p_multimesh, int p_index, const Transform2D &p_transform) override; + virtual void multimesh_instance_set_color(RID p_multimesh, int p_index, const Color &p_color) override; + virtual void multimesh_instance_set_custom_data(RID p_multimesh, int p_index, const Color &p_color) override; + + virtual RID multimesh_get_mesh(RID p_multimesh) const override; + + virtual Transform3D multimesh_instance_get_transform(RID p_multimesh, int p_index) const override; + virtual Transform2D multimesh_instance_get_transform_2d(RID p_multimesh, int p_index) const override; + virtual Color multimesh_instance_get_color(RID p_multimesh, int p_index) const override; + virtual Color multimesh_instance_get_custom_data(RID p_multimesh, int p_index) const override; + + virtual void multimesh_set_buffer(RID p_multimesh, const Vector<float> &p_buffer) override; + virtual Vector<float> multimesh_get_buffer(RID p_multimesh) const override; + + virtual void multimesh_set_visible_instances(RID p_multimesh, int p_visible) override; + virtual int multimesh_get_visible_instances(RID p_multimesh) const override; + + virtual AABB multimesh_get_aabb(RID p_multimesh) const override; + + void _update_dirty_multimeshes(); + bool _multimesh_enable_motion_vectors(RID p_multimesh); + void _multimesh_get_motion_vectors_offsets(RID p_multimesh, uint32_t &r_current_offset, uint32_t &r_prev_offset); + + _FORCE_INLINE_ RS::MultimeshTransformFormat multimesh_get_transform_format(RID p_multimesh) const { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + return multimesh->xform_format; + } + + _FORCE_INLINE_ bool multimesh_uses_colors(RID p_multimesh) const { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + return multimesh->uses_colors; + } + + _FORCE_INLINE_ bool multimesh_uses_custom_data(RID p_multimesh) const { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + return multimesh->uses_custom_data; + } + + _FORCE_INLINE_ uint32_t multimesh_get_instances_to_draw(RID p_multimesh) const { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + if (multimesh->visible_instances >= 0) { + return multimesh->visible_instances; + } + return multimesh->instances; + } + + _FORCE_INLINE_ RID multimesh_get_3d_uniform_set(RID p_multimesh, RID p_shader, uint32_t p_set) const { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + if (multimesh == nullptr) { + return RID(); + } + if (!multimesh->uniform_set_3d.is_valid()) { + if (!multimesh->buffer.is_valid()) { + return RID(); + } + Vector<RD::Uniform> uniforms; + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 0; + u.append_id(multimesh->buffer); + uniforms.push_back(u); + multimesh->uniform_set_3d = RD::get_singleton()->uniform_set_create(uniforms, p_shader, p_set); + } + + return multimesh->uniform_set_3d; + } + + _FORCE_INLINE_ RID multimesh_get_2d_uniform_set(RID p_multimesh, RID p_shader, uint32_t p_set) const { + MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); + if (multimesh == nullptr) { + return RID(); + } + if (!multimesh->uniform_set_2d.is_valid()) { + if (!multimesh->buffer.is_valid()) { + return RID(); + } + Vector<RD::Uniform> uniforms; + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 0; + u.append_id(multimesh->buffer); + uniforms.push_back(u); + multimesh->uniform_set_2d = RD::get_singleton()->uniform_set_create(uniforms, p_shader, p_set); + } + + return multimesh->uniform_set_2d; + } + + Dependency *multimesh_get_dependency(RID p_multimesh) const; + + /* SKELETON API */ + + bool owns_skeleton(RID p_rid) const { return skeleton_owner.owns(p_rid); }; + + virtual RID skeleton_allocate() override; + virtual void skeleton_initialize(RID p_skeleton) override; + virtual void skeleton_free(RID p_rid) override; + + virtual void skeleton_allocate_data(RID p_skeleton, int p_bones, bool p_2d_skeleton = false) override; + virtual void skeleton_set_base_transform_2d(RID p_skeleton, const Transform2D &p_base_transform) override; + void skeleton_set_world_transform(RID p_skeleton, bool p_enable, const Transform3D &p_world_transform); + virtual int skeleton_get_bone_count(RID p_skeleton) const override; + virtual void skeleton_bone_set_transform(RID p_skeleton, int p_bone, const Transform3D &p_transform) override; + virtual Transform3D skeleton_bone_get_transform(RID p_skeleton, int p_bone) const override; + virtual void skeleton_bone_set_transform_2d(RID p_skeleton, int p_bone, const Transform2D &p_transform) override; + virtual Transform2D skeleton_bone_get_transform_2d(RID p_skeleton, int p_bone) const override; + + virtual void skeleton_update_dependency(RID p_skeleton, DependencyTracker *p_instance) override; + + void _update_dirty_skeletons(); + + _FORCE_INLINE_ bool skeleton_is_valid(RID p_skeleton) { + return skeleton_owner.get_or_null(p_skeleton) != nullptr; + } + + _FORCE_INLINE_ RID skeleton_get_3d_uniform_set(RID p_skeleton, RID p_shader, uint32_t p_set) const { + Skeleton *skeleton = skeleton_owner.get_or_null(p_skeleton); + ERR_FAIL_COND_V(!skeleton, RID()); + ERR_FAIL_COND_V(skeleton->size == 0, RID()); + if (skeleton->use_2d) { + return RID(); + } + if (!skeleton->uniform_set_3d.is_valid()) { + Vector<RD::Uniform> uniforms; + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 0; + u.append_id(skeleton->buffer); + uniforms.push_back(u); + skeleton->uniform_set_3d = RD::get_singleton()->uniform_set_create(uniforms, p_shader, p_set); + } + + return skeleton->uniform_set_3d; + } +}; + +} // namespace RendererRD + +#endif // MESH_STORAGE_RD_H diff --git a/servers/rendering/renderer_rd/storage_rd/particles_storage.cpp b/servers/rendering/renderer_rd/storage_rd/particles_storage.cpp new file mode 100644 index 0000000000..240f743387 --- /dev/null +++ b/servers/rendering/renderer_rd/storage_rd/particles_storage.cpp @@ -0,0 +1,1933 @@ +/*************************************************************************/ +/* particles_storage.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "particles_storage.h" + +#include "servers/rendering/renderer_rd/renderer_compositor_rd.h" +#include "servers/rendering/rendering_server_globals.h" +#include "texture_storage.h" + +using namespace RendererRD; + +ParticlesStorage *ParticlesStorage::singleton = nullptr; + +ParticlesStorage *ParticlesStorage::get_singleton() { + return singleton; +} + +ParticlesStorage::ParticlesStorage() { + singleton = this; + + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + + /* Particles */ + + { + // Initialize particles + Vector<String> particles_modes; + particles_modes.push_back(""); + particles_shader.shader.initialize(particles_modes, String()); + } + MaterialStorage::get_singleton()->shader_set_data_request_function(MaterialStorage::SHADER_TYPE_PARTICLES, _create_particles_shader_funcs); + MaterialStorage::get_singleton()->material_set_data_request_function(MaterialStorage::SHADER_TYPE_PARTICLES, _create_particles_material_funcs); + + { + ShaderCompiler::DefaultIdentifierActions actions; + + actions.renames["COLOR"] = "PARTICLE.color"; + actions.renames["VELOCITY"] = "PARTICLE.velocity"; + //actions.renames["MASS"] = "mass"; ? + actions.renames["ACTIVE"] = "particle_active"; + actions.renames["RESTART"] = "restart"; + actions.renames["CUSTOM"] = "PARTICLE.custom"; + for (int i = 0; i < ParticlesShader::MAX_USERDATAS; i++) { + String udname = "USERDATA" + itos(i + 1); + actions.renames[udname] = "PARTICLE.userdata" + itos(i + 1); + actions.usage_defines[udname] = "#define USERDATA" + itos(i + 1) + "_USED\n"; + } + actions.renames["TRANSFORM"] = "PARTICLE.xform"; + actions.renames["TIME"] = "frame_history.data[0].time"; + actions.renames["PI"] = _MKSTR(Math_PI); + actions.renames["TAU"] = _MKSTR(Math_TAU); + actions.renames["E"] = _MKSTR(Math_E); + actions.renames["LIFETIME"] = "params.lifetime"; + actions.renames["DELTA"] = "local_delta"; + actions.renames["NUMBER"] = "particle_number"; + actions.renames["INDEX"] = "index"; + //actions.renames["GRAVITY"] = "current_gravity"; + actions.renames["EMISSION_TRANSFORM"] = "FRAME.emission_transform"; + actions.renames["RANDOM_SEED"] = "FRAME.random_seed"; + actions.renames["FLAG_EMIT_POSITION"] = "EMISSION_FLAG_HAS_POSITION"; + actions.renames["FLAG_EMIT_ROT_SCALE"] = "EMISSION_FLAG_HAS_ROTATION_SCALE"; + actions.renames["FLAG_EMIT_VELOCITY"] = "EMISSION_FLAG_HAS_VELOCITY"; + actions.renames["FLAG_EMIT_COLOR"] = "EMISSION_FLAG_HAS_COLOR"; + actions.renames["FLAG_EMIT_CUSTOM"] = "EMISSION_FLAG_HAS_CUSTOM"; + actions.renames["RESTART_POSITION"] = "restart_position"; + actions.renames["RESTART_ROT_SCALE"] = "restart_rotation_scale"; + actions.renames["RESTART_VELOCITY"] = "restart_velocity"; + actions.renames["RESTART_COLOR"] = "restart_color"; + actions.renames["RESTART_CUSTOM"] = "restart_custom"; + actions.renames["emit_subparticle"] = "emit_subparticle"; + actions.renames["COLLIDED"] = "collided"; + actions.renames["COLLISION_NORMAL"] = "collision_normal"; + actions.renames["COLLISION_DEPTH"] = "collision_depth"; + actions.renames["ATTRACTOR_FORCE"] = "attractor_force"; + + actions.render_mode_defines["disable_force"] = "#define DISABLE_FORCE\n"; + actions.render_mode_defines["disable_velocity"] = "#define DISABLE_VELOCITY\n"; + actions.render_mode_defines["keep_data"] = "#define ENABLE_KEEP_DATA\n"; + actions.render_mode_defines["collision_use_scale"] = "#define USE_COLLISION_SCALE\n"; + + actions.sampler_array_name = "material_samplers"; + actions.base_texture_binding_index = 1; + actions.texture_layout_set = 3; + actions.base_uniform_string = "material."; + actions.base_varying_index = 10; + + actions.default_filter = ShaderLanguage::FILTER_LINEAR_MIPMAP; + actions.default_repeat = ShaderLanguage::REPEAT_ENABLE; + actions.global_buffer_array_variable = "global_shader_uniforms.data"; + + particles_shader.compiler.initialize(actions); + } + + { + // default material and shader for particles shader + particles_shader.default_shader = material_storage->shader_allocate(); + material_storage->shader_initialize(particles_shader.default_shader); + material_storage->shader_set_code(particles_shader.default_shader, R"( +// Default particles shader. + +shader_type particles; + +void process() { + COLOR = vec4(1.0); +} +)"); + particles_shader.default_material = material_storage->material_allocate(); + material_storage->material_initialize(particles_shader.default_material); + material_storage->material_set_shader(particles_shader.default_material, particles_shader.default_shader); + + ParticleProcessMaterialData *md = static_cast<ParticleProcessMaterialData *>(material_storage->material_get_data(particles_shader.default_material, MaterialStorage::SHADER_TYPE_PARTICLES)); + particles_shader.default_shader_rd = particles_shader.shader.version_get_shader(md->shader_data->version, 0); + + Vector<RD::Uniform> uniforms; + + { + Vector<RID> ids; + ids.resize(12); + RID *ids_ptr = ids.ptrw(); + ids_ptr[0] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[1] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[2] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[3] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[4] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[5] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[6] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[7] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[8] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[9] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[10] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[11] = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + + RD::Uniform u(RD::UNIFORM_TYPE_SAMPLER, 1, ids); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 2; + u.append_id(material_storage->global_shader_uniforms_get_storage_buffer()); + uniforms.push_back(u); + } + + particles_shader.base_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, particles_shader.default_shader_rd, 0); + } + + { + Vector<String> copy_modes; + for (int i = 0; i <= ParticlesShader::MAX_USERDATAS; i++) { + if (i == 0) { + copy_modes.push_back("\n#define MODE_FILL_INSTANCES\n"); + copy_modes.push_back("\n#define MODE_FILL_SORT_BUFFER\n#define USE_SORT_BUFFER\n"); + copy_modes.push_back("\n#define MODE_FILL_INSTANCES\n#define USE_SORT_BUFFER\n"); + } else { + copy_modes.push_back("\n#define MODE_FILL_INSTANCES\n#define USERDATA_COUNT " + itos(i) + "\n"); + copy_modes.push_back("\n#define MODE_FILL_SORT_BUFFER\n#define USE_SORT_BUFFER\n#define USERDATA_COUNT " + itos(i) + "\n"); + copy_modes.push_back("\n#define MODE_FILL_INSTANCES\n#define USE_SORT_BUFFER\n#define USERDATA_COUNT " + itos(i) + "\n"); + } + } + + particles_shader.copy_shader.initialize(copy_modes); + + particles_shader.copy_shader_version = particles_shader.copy_shader.version_create(); + + for (int i = 0; i <= ParticlesShader::MAX_USERDATAS; i++) { + for (int j = 0; j < ParticlesShader::COPY_MODE_MAX; j++) { + particles_shader.copy_pipelines[i * ParticlesShader::COPY_MODE_MAX + j] = RD::get_singleton()->compute_pipeline_create(particles_shader.copy_shader.version_get_shader(particles_shader.copy_shader_version, i * ParticlesShader::COPY_MODE_MAX + j)); + } + } + } +} + +ParticlesStorage::~ParticlesStorage() { + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + + particles_shader.copy_shader.version_free(particles_shader.copy_shader_version); + + material_storage->material_free(particles_shader.default_material); + material_storage->shader_free(particles_shader.default_shader); + + singleton = nullptr; +} + +bool ParticlesStorage::free(RID p_rid) { + if (owns_particles(p_rid)) { + particles_free(p_rid); + return true; + } else if (owns_particles_collision(p_rid)) { + particles_collision_free(p_rid); + return true; + } else if (owns_particles_collision_instance(p_rid)) { + particles_collision_instance_free(p_rid); + return true; + } + + return false; +} + +/* PARTICLES */ + +RID ParticlesStorage::particles_allocate() { + return particles_owner.allocate_rid(); +} + +void ParticlesStorage::particles_initialize(RID p_rid) { + particles_owner.initialize_rid(p_rid, Particles()); +} + +void ParticlesStorage::particles_free(RID p_rid) { + update_particles(); + Particles *particles = particles_owner.get_or_null(p_rid); + particles->dependency.deleted_notify(p_rid); + _particles_free_data(particles); + particles_owner.free(p_rid); +} + +void ParticlesStorage::particles_set_mode(RID p_particles, RS::ParticlesMode p_mode) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + if (particles->mode == p_mode) { + return; + } + + _particles_free_data(particles); + + particles->mode = p_mode; +} + +void ParticlesStorage::particles_set_emitting(RID p_particles, bool p_emitting) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + particles->emitting = p_emitting; +} + +bool ParticlesStorage::particles_get_emitting(RID p_particles) { + ERR_FAIL_COND_V_MSG(RSG::threaded, false, "This function should never be used with threaded rendering, as it stalls the renderer."); + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND_V(!particles, false); + + return particles->emitting; +} + +void ParticlesStorage::_particles_free_data(Particles *particles) { + if (particles->particle_buffer.is_valid()) { + RD::get_singleton()->free(particles->particle_buffer); + particles->particle_buffer = RID(); + RD::get_singleton()->free(particles->particle_instance_buffer); + particles->particle_instance_buffer = RID(); + } + + particles->userdata_count = 0; + + if (particles->frame_params_buffer.is_valid()) { + RD::get_singleton()->free(particles->frame_params_buffer); + particles->frame_params_buffer = RID(); + } + particles->particles_transforms_buffer_uniform_set = RID(); + + if (RD::get_singleton()->uniform_set_is_valid(particles->trail_bind_pose_uniform_set)) { + RD::get_singleton()->free(particles->trail_bind_pose_uniform_set); + } + particles->trail_bind_pose_uniform_set = RID(); + + if (particles->trail_bind_pose_buffer.is_valid()) { + RD::get_singleton()->free(particles->trail_bind_pose_buffer); + particles->trail_bind_pose_buffer = RID(); + } + if (RD::get_singleton()->uniform_set_is_valid(particles->collision_textures_uniform_set)) { + RD::get_singleton()->free(particles->collision_textures_uniform_set); + } + particles->collision_textures_uniform_set = RID(); + + if (particles->particles_sort_buffer.is_valid()) { + RD::get_singleton()->free(particles->particles_sort_buffer); + particles->particles_sort_buffer = RID(); + particles->particles_sort_uniform_set = RID(); + } + + if (particles->emission_buffer != nullptr) { + particles->emission_buffer = nullptr; + particles->emission_buffer_data.clear(); + RD::get_singleton()->free(particles->emission_storage_buffer); + particles->emission_storage_buffer = RID(); + } + + if (RD::get_singleton()->uniform_set_is_valid(particles->particles_material_uniform_set)) { + //will need to be re-created + RD::get_singleton()->free(particles->particles_material_uniform_set); + } + particles->particles_material_uniform_set = RID(); +} + +void ParticlesStorage::particles_set_amount(RID p_particles, int p_amount) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + if (particles->amount == p_amount) { + return; + } + + _particles_free_data(particles); + + particles->amount = p_amount; + + particles->prev_ticks = 0; + particles->phase = 0; + particles->prev_phase = 0; + particles->clear = true; + + particles->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_PARTICLES); +} + +void ParticlesStorage::particles_set_lifetime(RID p_particles, double p_lifetime) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + particles->lifetime = p_lifetime; +} + +void ParticlesStorage::particles_set_one_shot(RID p_particles, bool p_one_shot) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + particles->one_shot = p_one_shot; +} + +void ParticlesStorage::particles_set_pre_process_time(RID p_particles, double p_time) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + particles->pre_process_time = p_time; +} +void ParticlesStorage::particles_set_explosiveness_ratio(RID p_particles, real_t p_ratio) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + particles->explosiveness = p_ratio; +} +void ParticlesStorage::particles_set_randomness_ratio(RID p_particles, real_t p_ratio) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + particles->randomness = p_ratio; +} + +void ParticlesStorage::particles_set_custom_aabb(RID p_particles, const AABB &p_aabb) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + particles->custom_aabb = p_aabb; + particles->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); +} + +void ParticlesStorage::particles_set_speed_scale(RID p_particles, double p_scale) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + particles->speed_scale = p_scale; +} +void ParticlesStorage::particles_set_use_local_coordinates(RID p_particles, bool p_enable) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + particles->use_local_coords = p_enable; + particles->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_PARTICLES); +} + +void ParticlesStorage::particles_set_fixed_fps(RID p_particles, int p_fps) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + particles->fixed_fps = p_fps; + + _particles_free_data(particles); + + particles->prev_ticks = 0; + particles->phase = 0; + particles->prev_phase = 0; + particles->clear = true; + + particles->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_PARTICLES); +} + +void ParticlesStorage::particles_set_interpolate(RID p_particles, bool p_enable) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + particles->interpolate = p_enable; +} + +void ParticlesStorage::particles_set_fractional_delta(RID p_particles, bool p_enable) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + particles->fractional_delta = p_enable; +} + +void ParticlesStorage::particles_set_trails(RID p_particles, bool p_enable, double p_length) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + ERR_FAIL_COND(p_length < 0.1); + p_length = MIN(10.0, p_length); + + particles->trails_enabled = p_enable; + particles->trail_length = p_length; + + _particles_free_data(particles); + + particles->prev_ticks = 0; + particles->phase = 0; + particles->prev_phase = 0; + particles->clear = true; + + particles->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_PARTICLES); +} + +void ParticlesStorage::particles_set_trail_bind_poses(RID p_particles, const Vector<Transform3D> &p_bind_poses) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + if (particles->trail_bind_pose_buffer.is_valid() && particles->trail_bind_poses.size() != p_bind_poses.size()) { + _particles_free_data(particles); + + particles->prev_ticks = 0; + particles->phase = 0; + particles->prev_phase = 0; + particles->clear = true; + } + particles->trail_bind_poses = p_bind_poses; + particles->trail_bind_poses_dirty = true; + + particles->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_PARTICLES); +} + +void ParticlesStorage::particles_set_collision_base_size(RID p_particles, real_t p_size) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + particles->collision_base_size = p_size; +} + +void ParticlesStorage::particles_set_transform_align(RID p_particles, RS::ParticlesTransformAlign p_transform_align) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + particles->transform_align = p_transform_align; +} + +void ParticlesStorage::particles_set_process_material(RID p_particles, RID p_material) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + particles->process_material = p_material; + particles->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_PARTICLES); //the instance buffer may have changed +} + +RID ParticlesStorage::particles_get_process_material(RID p_particles) const { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND_V(!particles, RID()); + + return particles->process_material; +} + +void ParticlesStorage::particles_set_draw_order(RID p_particles, RS::ParticlesDrawOrder p_order) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + particles->draw_order = p_order; +} + +void ParticlesStorage::particles_set_draw_passes(RID p_particles, int p_passes) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + particles->draw_passes.resize(p_passes); +} + +void ParticlesStorage::particles_set_draw_pass_mesh(RID p_particles, int p_pass, RID p_mesh) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + ERR_FAIL_INDEX(p_pass, particles->draw_passes.size()); + particles->draw_passes.write[p_pass] = p_mesh; +} + +void ParticlesStorage::particles_restart(RID p_particles) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + particles->restart_request = true; +} + +void ParticlesStorage::_particles_allocate_emission_buffer(Particles *particles) { + ERR_FAIL_COND(particles->emission_buffer != nullptr); + + particles->emission_buffer_data.resize(sizeof(ParticleEmissionBuffer::Data) * particles->amount + sizeof(uint32_t) * 4); + memset(particles->emission_buffer_data.ptrw(), 0, particles->emission_buffer_data.size()); + particles->emission_buffer = reinterpret_cast<ParticleEmissionBuffer *>(particles->emission_buffer_data.ptrw()); + particles->emission_buffer->particle_max = particles->amount; + + particles->emission_storage_buffer = RD::get_singleton()->storage_buffer_create(particles->emission_buffer_data.size(), particles->emission_buffer_data); + + if (RD::get_singleton()->uniform_set_is_valid(particles->particles_material_uniform_set)) { + //will need to be re-created + RD::get_singleton()->free(particles->particles_material_uniform_set); + particles->particles_material_uniform_set = RID(); + } +} + +void ParticlesStorage::particles_set_subemitter(RID p_particles, RID p_subemitter_particles) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + ERR_FAIL_COND(p_particles == p_subemitter_particles); + + particles->sub_emitter = p_subemitter_particles; + + if (RD::get_singleton()->uniform_set_is_valid(particles->particles_material_uniform_set)) { + RD::get_singleton()->free(particles->particles_material_uniform_set); + particles->particles_material_uniform_set = RID(); //clear and force to re create sub emitting + } +} + +void ParticlesStorage::particles_emit(RID p_particles, const Transform3D &p_transform, const Vector3 &p_velocity, const Color &p_color, const Color &p_custom, uint32_t p_emit_flags) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + ERR_FAIL_COND(particles->amount == 0); + + if (particles->emitting) { + particles->clear = true; + particles->emitting = false; + } + + if (particles->emission_buffer == nullptr) { + _particles_allocate_emission_buffer(particles); + } + + particles->inactive = false; + particles->inactive_time = 0; + + int32_t idx = particles->emission_buffer->particle_count; + if (idx < particles->emission_buffer->particle_max) { + RendererRD::MaterialStorage::store_transform(p_transform, particles->emission_buffer->data[idx].xform); + + particles->emission_buffer->data[idx].velocity[0] = p_velocity.x; + particles->emission_buffer->data[idx].velocity[1] = p_velocity.y; + particles->emission_buffer->data[idx].velocity[2] = p_velocity.z; + + particles->emission_buffer->data[idx].custom[0] = p_custom.r; + particles->emission_buffer->data[idx].custom[1] = p_custom.g; + particles->emission_buffer->data[idx].custom[2] = p_custom.b; + particles->emission_buffer->data[idx].custom[3] = p_custom.a; + + particles->emission_buffer->data[idx].color[0] = p_color.r; + particles->emission_buffer->data[idx].color[1] = p_color.g; + particles->emission_buffer->data[idx].color[2] = p_color.b; + particles->emission_buffer->data[idx].color[3] = p_color.a; + + particles->emission_buffer->data[idx].flags = p_emit_flags; + particles->emission_buffer->particle_count++; + } +} + +void ParticlesStorage::particles_request_process(RID p_particles) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + if (!particles->dirty) { + particles->dirty = true; + particles->update_list = particle_update_list; + particle_update_list = particles; + } +} + +AABB ParticlesStorage::particles_get_current_aabb(RID p_particles) { + if (RSG::threaded) { + WARN_PRINT_ONCE("Calling this function with threaded rendering enabled stalls the renderer, use with care."); + } + + const Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND_V(!particles, AABB()); + + int total_amount = particles->amount; + if (particles->trails_enabled && particles->trail_bind_poses.size() > 1) { + total_amount *= particles->trail_bind_poses.size(); + } + + Vector<uint8_t> buffer = RD::get_singleton()->buffer_get_data(particles->particle_buffer); + ERR_FAIL_COND_V(buffer.size() != (int)(total_amount * sizeof(ParticleData)), AABB()); + + Transform3D inv = particles->emission_transform.affine_inverse(); + + AABB aabb; + if (buffer.size()) { + bool first = true; + + const uint8_t *data_ptr = (const uint8_t *)buffer.ptr(); + uint32_t particle_data_size = sizeof(ParticleData) + sizeof(float) * particles->userdata_count; + + for (int i = 0; i < total_amount; i++) { + const ParticleData &particle_data = *(const ParticleData *)&data_ptr[particle_data_size * i]; + if (particle_data.active) { + Vector3 pos = Vector3(particle_data.xform[12], particle_data.xform[13], particle_data.xform[14]); + if (!particles->use_local_coords) { + pos = inv.xform(pos); + } + if (first) { + aabb.position = pos; + first = false; + } else { + aabb.expand_to(pos); + } + } + } + } + + float longest_axis_size = 0; + for (int i = 0; i < particles->draw_passes.size(); i++) { + if (particles->draw_passes[i].is_valid()) { + AABB maabb = MeshStorage::get_singleton()->mesh_get_aabb(particles->draw_passes[i], RID()); + longest_axis_size = MAX(maabb.get_longest_axis_size(), longest_axis_size); + } + } + + aabb.grow_by(longest_axis_size); + + return aabb; +} + +AABB ParticlesStorage::particles_get_aabb(RID p_particles) const { + const Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND_V(!particles, AABB()); + + return particles->custom_aabb; +} + +void ParticlesStorage::particles_set_emission_transform(RID p_particles, const Transform3D &p_transform) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + particles->emission_transform = p_transform; +} + +int ParticlesStorage::particles_get_draw_passes(RID p_particles) const { + const Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND_V(!particles, 0); + + return particles->draw_passes.size(); +} + +RID ParticlesStorage::particles_get_draw_pass_mesh(RID p_particles, int p_pass) const { + const Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND_V(!particles, RID()); + ERR_FAIL_INDEX_V(p_pass, particles->draw_passes.size(), RID()); + + return particles->draw_passes[p_pass]; +} + +void ParticlesStorage::particles_add_collision(RID p_particles, RID p_particles_collision_instance) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + particles->collisions.insert(p_particles_collision_instance); +} + +void ParticlesStorage::particles_remove_collision(RID p_particles, RID p_particles_collision_instance) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + particles->collisions.erase(p_particles_collision_instance); +} + +void ParticlesStorage::particles_set_canvas_sdf_collision(RID p_particles, bool p_enable, const Transform2D &p_xform, const Rect2 &p_to_screen, RID p_texture) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + particles->has_sdf_collision = p_enable; + particles->sdf_collision_transform = p_xform; + particles->sdf_collision_to_screen = p_to_screen; + particles->sdf_collision_texture = p_texture; +} + +void ParticlesStorage::_particles_process(Particles *p_particles, double p_delta) { + TextureStorage *texture_storage = TextureStorage::get_singleton(); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + + if (p_particles->particles_material_uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(p_particles->particles_material_uniform_set)) { + Vector<RD::Uniform> uniforms; + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 0; + u.append_id(p_particles->frame_params_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 1; + u.append_id(p_particles->particle_buffer); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 2; + if (p_particles->emission_storage_buffer.is_valid()) { + u.append_id(p_particles->emission_storage_buffer); + } else { + u.append_id(MeshStorage::get_singleton()->get_default_rd_storage_buffer()); + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 3; + Particles *sub_emitter = particles_owner.get_or_null(p_particles->sub_emitter); + if (sub_emitter) { + if (sub_emitter->emission_buffer == nullptr) { //no emission buffer, allocate emission buffer + _particles_allocate_emission_buffer(sub_emitter); + } + u.append_id(sub_emitter->emission_storage_buffer); + } else { + u.append_id(MeshStorage::get_singleton()->get_default_rd_storage_buffer()); + } + uniforms.push_back(u); + } + + p_particles->particles_material_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, particles_shader.default_shader_rd, 1); + } + + double new_phase = Math::fmod((double)p_particles->phase + (p_delta / p_particles->lifetime) * p_particles->speed_scale, 1.0); + + //move back history (if there is any) + for (uint32_t i = p_particles->frame_history.size() - 1; i > 0; i--) { + p_particles->frame_history[i] = p_particles->frame_history[i - 1]; + } + //update current frame + ParticlesFrameParams &frame_params = p_particles->frame_history[0]; + + if (p_particles->clear) { + p_particles->cycle_number = 0; + p_particles->random_seed = Math::rand(); + } else if (new_phase < p_particles->phase) { + if (p_particles->one_shot) { + p_particles->emitting = false; + } + p_particles->cycle_number++; + } + + frame_params.emitting = p_particles->emitting; + frame_params.system_phase = new_phase; + frame_params.prev_system_phase = p_particles->phase; + + p_particles->phase = new_phase; + + frame_params.time = RendererCompositorRD::singleton->get_total_time(); + frame_params.delta = p_delta * p_particles->speed_scale; + frame_params.random_seed = p_particles->random_seed; + frame_params.explosiveness = p_particles->explosiveness; + frame_params.randomness = p_particles->randomness; + + if (p_particles->use_local_coords) { + RendererRD::MaterialStorage::store_transform(Transform3D(), frame_params.emission_transform); + } else { + RendererRD::MaterialStorage::store_transform(p_particles->emission_transform, frame_params.emission_transform); + } + + frame_params.cycle = p_particles->cycle_number; + frame_params.frame = p_particles->frame_counter++; + frame_params.pad0 = 0; + frame_params.pad1 = 0; + frame_params.pad2 = 0; + + { //collision and attractors + + frame_params.collider_count = 0; + frame_params.attractor_count = 0; + frame_params.particle_size = p_particles->collision_base_size; + + RID collision_3d_textures[ParticlesFrameParams::MAX_3D_TEXTURES]; + RID collision_heightmap_texture; + + Transform3D to_particles; + if (p_particles->use_local_coords) { + to_particles = p_particles->emission_transform.affine_inverse(); + } + + if (p_particles->has_sdf_collision && RD::get_singleton()->texture_is_valid(p_particles->sdf_collision_texture)) { + //2D collision + + Transform2D xform = p_particles->sdf_collision_transform; //will use dotproduct manually so invert beforehand + Transform2D revert = xform.affine_inverse(); + frame_params.collider_count = 1; + frame_params.colliders[0].transform[0] = xform.columns[0][0]; + frame_params.colliders[0].transform[1] = xform.columns[0][1]; + frame_params.colliders[0].transform[2] = 0; + frame_params.colliders[0].transform[3] = xform.columns[2][0]; + + frame_params.colliders[0].transform[4] = xform.columns[1][0]; + frame_params.colliders[0].transform[5] = xform.columns[1][1]; + frame_params.colliders[0].transform[6] = 0; + frame_params.colliders[0].transform[7] = xform.columns[2][1]; + + frame_params.colliders[0].transform[8] = revert.columns[0][0]; + frame_params.colliders[0].transform[9] = revert.columns[0][1]; + frame_params.colliders[0].transform[10] = 0; + frame_params.colliders[0].transform[11] = revert.columns[2][0]; + + frame_params.colliders[0].transform[12] = revert.columns[1][0]; + frame_params.colliders[0].transform[13] = revert.columns[1][1]; + frame_params.colliders[0].transform[14] = 0; + frame_params.colliders[0].transform[15] = revert.columns[2][1]; + + frame_params.colliders[0].extents[0] = p_particles->sdf_collision_to_screen.size.x; + frame_params.colliders[0].extents[1] = p_particles->sdf_collision_to_screen.size.y; + frame_params.colliders[0].extents[2] = p_particles->sdf_collision_to_screen.position.x; + frame_params.colliders[0].scale = p_particles->sdf_collision_to_screen.position.y; + frame_params.colliders[0].texture_index = 0; + frame_params.colliders[0].type = ParticlesFrameParams::COLLISION_TYPE_2D_SDF; + + collision_heightmap_texture = p_particles->sdf_collision_texture; + + //replace in all other history frames where used because parameters are no longer valid if screen moves + for (uint32_t i = 1; i < p_particles->frame_history.size(); i++) { + if (p_particles->frame_history[i].collider_count > 0 && p_particles->frame_history[i].colliders[0].type == ParticlesFrameParams::COLLISION_TYPE_2D_SDF) { + p_particles->frame_history[i].colliders[0] = frame_params.colliders[0]; + } + } + } + + uint32_t collision_3d_textures_used = 0; + for (const RID &E : p_particles->collisions) { + ParticlesCollisionInstance *pci = particles_collision_instance_owner.get_or_null(E); + if (!pci || !pci->active) { + continue; + } + ParticlesCollision *pc = particles_collision_owner.get_or_null(pci->collision); + ERR_CONTINUE(!pc); + + Transform3D to_collider = pci->transform; + if (p_particles->use_local_coords) { + to_collider = to_particles * to_collider; + } + Vector3 scale = to_collider.basis.get_scale(); + to_collider.basis.orthonormalize(); + + if (pc->type <= RS::PARTICLES_COLLISION_TYPE_VECTOR_FIELD_ATTRACT) { + //attractor + if (frame_params.attractor_count >= ParticlesFrameParams::MAX_ATTRACTORS) { + continue; + } + + ParticlesFrameParams::Attractor &attr = frame_params.attractors[frame_params.attractor_count]; + + RendererRD::MaterialStorage::store_transform(to_collider, attr.transform); + attr.strength = pc->attractor_strength; + attr.attenuation = pc->attractor_attenuation; + attr.directionality = pc->attractor_directionality; + + switch (pc->type) { + case RS::PARTICLES_COLLISION_TYPE_SPHERE_ATTRACT: { + attr.type = ParticlesFrameParams::ATTRACTOR_TYPE_SPHERE; + float radius = pc->radius; + radius *= (scale.x + scale.y + scale.z) / 3.0; + attr.extents[0] = radius; + attr.extents[1] = radius; + attr.extents[2] = radius; + } break; + case RS::PARTICLES_COLLISION_TYPE_BOX_ATTRACT: { + attr.type = ParticlesFrameParams::ATTRACTOR_TYPE_BOX; + Vector3 extents = pc->extents * scale; + attr.extents[0] = extents.x; + attr.extents[1] = extents.y; + attr.extents[2] = extents.z; + } break; + case RS::PARTICLES_COLLISION_TYPE_VECTOR_FIELD_ATTRACT: { + if (collision_3d_textures_used >= ParticlesFrameParams::MAX_3D_TEXTURES) { + continue; + } + attr.type = ParticlesFrameParams::ATTRACTOR_TYPE_VECTOR_FIELD; + Vector3 extents = pc->extents * scale; + attr.extents[0] = extents.x; + attr.extents[1] = extents.y; + attr.extents[2] = extents.z; + attr.texture_index = collision_3d_textures_used; + + collision_3d_textures[collision_3d_textures_used] = pc->field_texture; + collision_3d_textures_used++; + } break; + default: { + } + } + + frame_params.attractor_count++; + } else { + //collider + if (frame_params.collider_count >= ParticlesFrameParams::MAX_COLLIDERS) { + continue; + } + + ParticlesFrameParams::Collider &col = frame_params.colliders[frame_params.collider_count]; + + RendererRD::MaterialStorage::store_transform(to_collider, col.transform); + switch (pc->type) { + case RS::PARTICLES_COLLISION_TYPE_SPHERE_COLLIDE: { + col.type = ParticlesFrameParams::COLLISION_TYPE_SPHERE; + float radius = pc->radius; + radius *= (scale.x + scale.y + scale.z) / 3.0; + col.extents[0] = radius; + col.extents[1] = radius; + col.extents[2] = radius; + } break; + case RS::PARTICLES_COLLISION_TYPE_BOX_COLLIDE: { + col.type = ParticlesFrameParams::COLLISION_TYPE_BOX; + Vector3 extents = pc->extents * scale; + col.extents[0] = extents.x; + col.extents[1] = extents.y; + col.extents[2] = extents.z; + } break; + case RS::PARTICLES_COLLISION_TYPE_SDF_COLLIDE: { + if (collision_3d_textures_used >= ParticlesFrameParams::MAX_3D_TEXTURES) { + continue; + } + col.type = ParticlesFrameParams::COLLISION_TYPE_SDF; + Vector3 extents = pc->extents * scale; + col.extents[0] = extents.x; + col.extents[1] = extents.y; + col.extents[2] = extents.z; + col.texture_index = collision_3d_textures_used; + col.scale = (scale.x + scale.y + scale.z) * 0.333333333333; //non uniform scale non supported + + collision_3d_textures[collision_3d_textures_used] = pc->field_texture; + collision_3d_textures_used++; + } break; + case RS::PARTICLES_COLLISION_TYPE_HEIGHTFIELD_COLLIDE: { + if (collision_heightmap_texture != RID()) { //already taken + continue; + } + + col.type = ParticlesFrameParams::COLLISION_TYPE_HEIGHT_FIELD; + Vector3 extents = pc->extents * scale; + col.extents[0] = extents.x; + col.extents[1] = extents.y; + col.extents[2] = extents.z; + collision_heightmap_texture = pc->heightfield_texture; + } break; + default: { + } + } + + frame_params.collider_count++; + } + } + + bool different = false; + if (collision_3d_textures_used == p_particles->collision_3d_textures_used) { + for (int i = 0; i < ParticlesFrameParams::MAX_3D_TEXTURES; i++) { + if (p_particles->collision_3d_textures[i] != collision_3d_textures[i]) { + different = true; + break; + } + } + } + + if (collision_heightmap_texture != p_particles->collision_heightmap_texture) { + different = true; + } + + bool uniform_set_valid = RD::get_singleton()->uniform_set_is_valid(p_particles->collision_textures_uniform_set); + + if (different || !uniform_set_valid) { + if (uniform_set_valid) { + RD::get_singleton()->free(p_particles->collision_textures_uniform_set); + } + + Vector<RD::Uniform> uniforms; + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 0; + for (uint32_t i = 0; i < ParticlesFrameParams::MAX_3D_TEXTURES; i++) { + RID rd_tex; + if (i < collision_3d_textures_used) { + if (TextureStorage::get_singleton()->texture_get_type(collision_3d_textures[i]) == TextureStorage::TYPE_3D) { + rd_tex = TextureStorage::get_singleton()->texture_get_rd_texture(collision_3d_textures[i]); + } + } + + if (rd_tex == RID()) { + rd_tex = texture_storage->texture_rd_get_default(TextureStorage::DEFAULT_RD_TEXTURE_3D_WHITE); + } + u.append_id(rd_tex); + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 1; + if (collision_heightmap_texture.is_valid()) { + u.append_id(collision_heightmap_texture); + } else { + u.append_id(texture_storage->texture_rd_get_default(TextureStorage::DEFAULT_RD_TEXTURE_BLACK)); + } + uniforms.push_back(u); + } + p_particles->collision_textures_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, particles_shader.default_shader_rd, 2); + } + } + + ParticlesShader::PushConstant push_constant; + + int process_amount = p_particles->amount; + + if (p_particles->trails_enabled && p_particles->trail_bind_poses.size() > 1) { + process_amount *= p_particles->trail_bind_poses.size(); + } + push_constant.clear = p_particles->clear; + push_constant.total_particles = p_particles->amount; + push_constant.lifetime = p_particles->lifetime; + push_constant.trail_size = p_particles->trail_params.size(); + push_constant.use_fractional_delta = p_particles->fractional_delta; + push_constant.sub_emitter_mode = !p_particles->emitting && p_particles->emission_buffer && (p_particles->emission_buffer->particle_count > 0 || p_particles->force_sub_emit); + push_constant.trail_pass = false; + + p_particles->force_sub_emit = false; //reset + + Particles *sub_emitter = particles_owner.get_or_null(p_particles->sub_emitter); + + if (sub_emitter && sub_emitter->emission_storage_buffer.is_valid()) { + // print_line("updating subemitter buffer"); + int32_t zero[4] = { 0, sub_emitter->amount, 0, 0 }; + RD::get_singleton()->buffer_update(sub_emitter->emission_storage_buffer, 0, sizeof(uint32_t) * 4, zero); + push_constant.can_emit = true; + + if (sub_emitter->emitting) { + sub_emitter->emitting = false; + sub_emitter->clear = true; //will need to clear if it was emitting, sorry + } + //make sure the sub emitter processes particles too + sub_emitter->inactive = false; + sub_emitter->inactive_time = 0; + + sub_emitter->force_sub_emit = true; + + } else { + push_constant.can_emit = false; + } + + if (p_particles->emission_buffer && p_particles->emission_buffer->particle_count) { + RD::get_singleton()->buffer_update(p_particles->emission_storage_buffer, 0, sizeof(uint32_t) * 4 + sizeof(ParticleEmissionBuffer::Data) * p_particles->emission_buffer->particle_count, p_particles->emission_buffer); + p_particles->emission_buffer->particle_count = 0; + } + + p_particles->clear = false; + + if (p_particles->trail_params.size() > 1) { + //fill the trail params + for (uint32_t i = 0; i < p_particles->trail_params.size(); i++) { + uint32_t src_idx = i * p_particles->frame_history.size() / p_particles->trail_params.size(); + p_particles->trail_params[i] = p_particles->frame_history[src_idx]; + } + } else { + p_particles->trail_params[0] = p_particles->frame_history[0]; + } + + RD::get_singleton()->buffer_update(p_particles->frame_params_buffer, 0, sizeof(ParticlesFrameParams) * p_particles->trail_params.size(), p_particles->trail_params.ptr()); + + ParticleProcessMaterialData *m = static_cast<ParticleProcessMaterialData *>(material_storage->material_get_data(p_particles->process_material, MaterialStorage::SHADER_TYPE_PARTICLES)); + if (!m) { + m = static_cast<ParticleProcessMaterialData *>(material_storage->material_get_data(particles_shader.default_material, MaterialStorage::SHADER_TYPE_PARTICLES)); + } + + ERR_FAIL_COND(!m); + + p_particles->has_collision_cache = m->shader_data->uses_collision; + + //todo should maybe compute all particle systems together? + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, m->shader_data->pipeline); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles_shader.base_uniform_set, 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, p_particles->particles_material_uniform_set, 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, p_particles->collision_textures_uniform_set, 2); + + if (m->uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(m->uniform_set)) { + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, m->uniform_set, 3); + } + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ParticlesShader::PushConstant)); + + if (p_particles->trails_enabled && p_particles->trail_bind_poses.size() > 1) { + //trails requires two passes in order to catch particle starts + RD::get_singleton()->compute_list_dispatch_threads(compute_list, process_amount / p_particles->trail_bind_poses.size(), 1, 1); + + RD::get_singleton()->compute_list_add_barrier(compute_list); + + push_constant.trail_pass = true; + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ParticlesShader::PushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, process_amount - p_particles->amount, 1, 1); + } else { + RD::get_singleton()->compute_list_dispatch_threads(compute_list, process_amount, 1, 1); + } + + RD::get_singleton()->compute_list_end(); +} + +void ParticlesStorage::particles_set_view_axis(RID p_particles, const Vector3 &p_axis, const Vector3 &p_up_axis) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND(!particles); + + if (particles->draw_order != RS::PARTICLES_DRAW_ORDER_VIEW_DEPTH && particles->transform_align != RS::PARTICLES_TRANSFORM_ALIGN_Z_BILLBOARD && particles->transform_align != RS::PARTICLES_TRANSFORM_ALIGN_Z_BILLBOARD_Y_TO_VELOCITY) { + return; + } + + if (particles->particle_buffer.is_null()) { + return; //particles have not processed yet + } + + bool do_sort = particles->draw_order == RS::PARTICLES_DRAW_ORDER_VIEW_DEPTH; + + //copy to sort buffer + if (do_sort && particles->particles_sort_buffer == RID()) { + uint32_t size = particles->amount; + if (size & 1) { + size++; //make multiple of 16 + } + size *= sizeof(float) * 2; + particles->particles_sort_buffer = RD::get_singleton()->storage_buffer_create(size); + + { + Vector<RD::Uniform> uniforms; + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 0; + u.append_id(particles->particles_sort_buffer); + uniforms.push_back(u); + } + + particles->particles_sort_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, particles_shader.copy_shader.version_get_shader(particles_shader.copy_shader_version, ParticlesShader::COPY_MODE_FILL_SORT_BUFFER), 1); + } + } + + ParticlesShader::CopyPushConstant copy_push_constant; + + if (particles->trails_enabled && particles->trail_bind_poses.size() > 1) { + int fixed_fps = 60.0; + if (particles->fixed_fps > 0) { + fixed_fps = particles->fixed_fps; + } + + copy_push_constant.trail_size = particles->trail_bind_poses.size(); + copy_push_constant.trail_total = particles->frame_history.size(); + copy_push_constant.frame_delta = 1.0 / fixed_fps; + } else { + copy_push_constant.trail_size = 1; + copy_push_constant.trail_total = 1; + copy_push_constant.frame_delta = 0.0; + } + + copy_push_constant.order_by_lifetime = (particles->draw_order == RS::PARTICLES_DRAW_ORDER_LIFETIME || particles->draw_order == RS::PARTICLES_DRAW_ORDER_REVERSE_LIFETIME); + copy_push_constant.lifetime_split = MIN(particles->amount * particles->phase, particles->amount - 1); + copy_push_constant.lifetime_reverse = particles->draw_order == RS::PARTICLES_DRAW_ORDER_REVERSE_LIFETIME; + + copy_push_constant.frame_remainder = particles->interpolate ? particles->frame_remainder : 0.0; + copy_push_constant.total_particles = particles->amount; + copy_push_constant.copy_mode_2d = false; + + Vector3 axis = -p_axis; // cameras look to z negative + + if (particles->use_local_coords) { + axis = particles->emission_transform.basis.xform_inv(axis).normalized(); + } + + copy_push_constant.sort_direction[0] = axis.x; + copy_push_constant.sort_direction[1] = axis.y; + copy_push_constant.sort_direction[2] = axis.z; + + copy_push_constant.align_up[0] = p_up_axis.x; + copy_push_constant.align_up[1] = p_up_axis.y; + copy_push_constant.align_up[2] = p_up_axis.z; + + copy_push_constant.align_mode = particles->transform_align; + + if (do_sort) { + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, particles_shader.copy_pipelines[ParticlesShader::COPY_MODE_FILL_SORT_BUFFER + particles->userdata_count * ParticlesShader::COPY_MODE_MAX]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->particles_copy_uniform_set, 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->particles_sort_uniform_set, 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->trail_bind_pose_uniform_set, 2); + RD::get_singleton()->compute_list_set_push_constant(compute_list, ©_push_constant, sizeof(ParticlesShader::CopyPushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, particles->amount, 1, 1); + + RD::get_singleton()->compute_list_end(); + RendererCompositorRD::singleton->get_effects()->sort_buffer(particles->particles_sort_uniform_set, particles->amount); + } + + copy_push_constant.total_particles *= copy_push_constant.total_particles; + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + uint32_t copy_pipeline = do_sort ? ParticlesShader::COPY_MODE_FILL_INSTANCES_WITH_SORT_BUFFER : ParticlesShader::COPY_MODE_FILL_INSTANCES; + copy_pipeline += particles->userdata_count * ParticlesShader::COPY_MODE_MAX; + copy_push_constant.copy_mode_2d = particles->mode == RS::PARTICLES_MODE_2D ? 1 : 0; + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, particles_shader.copy_pipelines[copy_pipeline]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->particles_copy_uniform_set, 0); + if (do_sort) { + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->particles_sort_uniform_set, 1); + } + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->trail_bind_pose_uniform_set, 2); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, ©_push_constant, sizeof(ParticlesShader::CopyPushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, copy_push_constant.total_particles, 1, 1); + + RD::get_singleton()->compute_list_end(); +} + +void ParticlesStorage::_particles_update_buffers(Particles *particles) { + uint32_t userdata_count = 0; + + MaterialStorage::ShaderData *shader_data = MaterialStorage::get_singleton()->material_get_shader_data(particles->process_material); + if (shader_data) { + const ParticlesShaderData *particle_shader_data = static_cast<const ParticlesShaderData *>(shader_data); + userdata_count = particle_shader_data->userdata_count; + } + + if (userdata_count != particles->userdata_count) { + // Mismatch userdata, re-create buffers. + _particles_free_data(particles); + } + + if (particles->amount > 0 && particles->particle_buffer.is_null()) { + int total_amount = particles->amount; + if (particles->trails_enabled && particles->trail_bind_poses.size() > 1) { + total_amount *= particles->trail_bind_poses.size(); + } + + uint32_t xform_size = particles->mode == RS::PARTICLES_MODE_2D ? 2 : 3; + + particles->particle_buffer = RD::get_singleton()->storage_buffer_create((sizeof(ParticleData) + userdata_count * sizeof(float) * 4) * total_amount); + + particles->userdata_count = userdata_count; + + particles->particle_instance_buffer = RD::get_singleton()->storage_buffer_create(sizeof(float) * 4 * (xform_size + 1 + 1) * total_amount); + //needs to clear it + + { + Vector<RD::Uniform> uniforms; + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 1; + u.append_id(particles->particle_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 2; + u.append_id(particles->particle_instance_buffer); + uniforms.push_back(u); + } + + particles->particles_copy_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, particles_shader.copy_shader.version_get_shader(particles_shader.copy_shader_version, 0), 0); + } + } +} +void ParticlesStorage::update_particles() { + while (particle_update_list) { + //use transform feedback to process particles + + Particles *particles = particle_update_list; + + //take and remove + particle_update_list = particles->update_list; + particles->update_list = nullptr; + particles->dirty = false; + + _particles_update_buffers(particles); + + if (particles->restart_request) { + particles->prev_ticks = 0; + particles->phase = 0; + particles->prev_phase = 0; + particles->clear = true; + particles->restart_request = false; + } + + if (particles->inactive && !particles->emitting) { + //go next + continue; + } + + if (particles->emitting) { + if (particles->inactive) { + //restart system from scratch + particles->prev_ticks = 0; + particles->phase = 0; + particles->prev_phase = 0; + particles->clear = true; + } + particles->inactive = false; + particles->inactive_time = 0; + } else { + particles->inactive_time += particles->speed_scale * RendererCompositorRD::singleton->get_frame_delta_time(); + if (particles->inactive_time > particles->lifetime * 1.2) { + particles->inactive = true; + continue; + } + } + + // TODO: Should use display refresh rate for all this. + float screen_hz = 60; + + int fixed_fps = 0; + if (particles->fixed_fps > 0) { + fixed_fps = particles->fixed_fps; + } else if (particles->trails_enabled && particles->trail_bind_poses.size() > 1) { + fixed_fps = screen_hz; + } + { + //update trails + int history_size = 1; + int trail_steps = 1; + if (particles->trails_enabled && particles->trail_bind_poses.size() > 1) { + history_size = MAX(1, int(particles->trail_length * fixed_fps)); + trail_steps = particles->trail_bind_poses.size(); + } + + if (uint32_t(history_size) != particles->frame_history.size()) { + particles->frame_history.resize(history_size); + memset(particles->frame_history.ptr(), 0, sizeof(ParticlesFrameParams) * history_size); + } + + if (uint32_t(trail_steps) != particles->trail_params.size() || particles->frame_params_buffer.is_null()) { + particles->trail_params.resize(trail_steps); + if (particles->frame_params_buffer.is_valid()) { + RD::get_singleton()->free(particles->frame_params_buffer); + } + particles->frame_params_buffer = RD::get_singleton()->storage_buffer_create(sizeof(ParticlesFrameParams) * trail_steps); + } + + if (particles->trail_bind_poses.size() > 1 && particles->trail_bind_pose_buffer.is_null()) { + particles->trail_bind_pose_buffer = RD::get_singleton()->storage_buffer_create(sizeof(float) * 16 * particles->trail_bind_poses.size()); + particles->trail_bind_poses_dirty = true; + } + + if (particles->trail_bind_pose_uniform_set.is_null()) { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 0; + if (particles->trail_bind_pose_buffer.is_valid()) { + u.append_id(particles->trail_bind_pose_buffer); + } else { + u.append_id(MeshStorage::get_singleton()->get_default_rd_storage_buffer()); + } + uniforms.push_back(u); + } + + particles->trail_bind_pose_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, particles_shader.copy_shader.version_get_shader(particles_shader.copy_shader_version, 0), 2); + } + + if (particles->trail_bind_pose_buffer.is_valid() && particles->trail_bind_poses_dirty) { + if (particles_shader.pose_update_buffer.size() < uint32_t(particles->trail_bind_poses.size()) * 16) { + particles_shader.pose_update_buffer.resize(particles->trail_bind_poses.size() * 16); + } + + for (int i = 0; i < particles->trail_bind_poses.size(); i++) { + RendererRD::MaterialStorage::store_transform(particles->trail_bind_poses[i], &particles_shader.pose_update_buffer[i * 16]); + } + + RD::get_singleton()->buffer_update(particles->trail_bind_pose_buffer, 0, particles->trail_bind_poses.size() * 16 * sizeof(float), particles_shader.pose_update_buffer.ptr()); + } + } + + bool zero_time_scale = Engine::get_singleton()->get_time_scale() <= 0.0; + + if (particles->clear && particles->pre_process_time > 0.0) { + double frame_time; + if (fixed_fps > 0) { + frame_time = 1.0 / fixed_fps; + } else { + frame_time = 1.0 / 30.0; + } + + double todo = particles->pre_process_time; + + while (todo >= 0) { + _particles_process(particles, frame_time); + todo -= frame_time; + } + } + + if (fixed_fps > 0) { + double frame_time; + double decr; + if (zero_time_scale) { + frame_time = 0.0; + decr = 1.0 / fixed_fps; + } else { + frame_time = 1.0 / fixed_fps; + decr = frame_time; + } + double delta = RendererCompositorRD::singleton->get_frame_delta_time(); + if (delta > 0.1) { //avoid recursive stalls if fps goes below 10 + delta = 0.1; + } else if (delta <= 0.0) { //unlikely but.. + delta = 0.001; + } + double todo = particles->frame_remainder + delta; + + while (todo >= frame_time) { + _particles_process(particles, frame_time); + todo -= decr; + } + + particles->frame_remainder = todo; + + } else { + if (zero_time_scale) { + _particles_process(particles, 0.0); + } else { + _particles_process(particles, RendererCompositorRD::singleton->get_frame_delta_time()); + } + } + + //copy particles to instance buffer + + if (particles->draw_order != RS::PARTICLES_DRAW_ORDER_VIEW_DEPTH && particles->transform_align != RS::PARTICLES_TRANSFORM_ALIGN_Z_BILLBOARD && particles->transform_align != RS::PARTICLES_TRANSFORM_ALIGN_Z_BILLBOARD_Y_TO_VELOCITY) { + //does not need view dependent operation, do copy here + ParticlesShader::CopyPushConstant copy_push_constant; + + int total_amount = particles->amount; + if (particles->trails_enabled && particles->trail_bind_poses.size() > 1) { + total_amount *= particles->trail_bind_poses.size(); + } + + // Affect 2D only. + if (particles->use_local_coords) { + // In local mode, particle positions are calculated locally (relative to the node position) + // and they're also drawn locally. + // It works as expected, so we just pass an identity transform. + RendererRD::MaterialStorage::store_transform(Transform3D(), copy_push_constant.inv_emission_transform); + } else { + // In global mode, particle positions are calculated globally (relative to the canvas origin) + // but they're drawn locally. + // So, we need to pass the inverse of the emission transform to bring the + // particles to local coordinates before drawing. + Transform3D inv = particles->emission_transform.affine_inverse(); + RendererRD::MaterialStorage::store_transform(inv, copy_push_constant.inv_emission_transform); + } + + copy_push_constant.total_particles = total_amount; + copy_push_constant.frame_remainder = particles->interpolate ? particles->frame_remainder : 0.0; + copy_push_constant.align_mode = particles->transform_align; + copy_push_constant.align_up[0] = 0; + copy_push_constant.align_up[1] = 0; + copy_push_constant.align_up[2] = 0; + + if (particles->trails_enabled && particles->trail_bind_poses.size() > 1) { + copy_push_constant.trail_size = particles->trail_bind_poses.size(); + copy_push_constant.trail_total = particles->frame_history.size(); + copy_push_constant.frame_delta = 1.0 / fixed_fps; + } else { + copy_push_constant.trail_size = 1; + copy_push_constant.trail_total = 1; + copy_push_constant.frame_delta = 0.0; + } + + copy_push_constant.order_by_lifetime = (particles->draw_order == RS::PARTICLES_DRAW_ORDER_LIFETIME || particles->draw_order == RS::PARTICLES_DRAW_ORDER_REVERSE_LIFETIME); + copy_push_constant.lifetime_split = MIN(particles->amount * particles->phase, particles->amount - 1); + copy_push_constant.lifetime_reverse = particles->draw_order == RS::PARTICLES_DRAW_ORDER_REVERSE_LIFETIME; + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + copy_push_constant.copy_mode_2d = particles->mode == RS::PARTICLES_MODE_2D ? 1 : 0; + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, particles_shader.copy_pipelines[ParticlesShader::COPY_MODE_FILL_INSTANCES + particles->userdata_count * ParticlesShader::COPY_MODE_MAX]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->particles_copy_uniform_set, 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->trail_bind_pose_uniform_set, 2); + RD::get_singleton()->compute_list_set_push_constant(compute_list, ©_push_constant, sizeof(ParticlesShader::CopyPushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, total_amount, 1, 1); + + RD::get_singleton()->compute_list_end(); + } + + particles->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); + } +} + +Dependency *ParticlesStorage::particles_get_dependency(RID p_particles) const { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_NULL_V(particles, nullptr); + + return &particles->dependency; +} + +bool ParticlesStorage::particles_is_inactive(RID p_particles) const { + ERR_FAIL_COND_V_MSG(RSG::threaded, false, "This function should never be used with threaded rendering, as it stalls the renderer."); + const Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND_V(!particles, false); + return !particles->emitting && particles->inactive; +} + +/* Particles SHADER */ + +void ParticlesStorage::ParticlesShaderData::set_path_hint(const String &p_path) { + path = p_path; +} +void ParticlesStorage::ParticlesShaderData::set_code(const String &p_code) { + ParticlesStorage *particles_storage = ParticlesStorage::get_singleton(); + //compile + + code = p_code; + valid = false; + ubo_size = 0; + uniforms.clear(); + uses_collision = false; + + if (code.is_empty()) { + return; //just invalid, but no error + } + + ShaderCompiler::GeneratedCode gen_code; + ShaderCompiler::IdentifierActions actions; + actions.entry_point_stages["start"] = ShaderCompiler::STAGE_COMPUTE; + actions.entry_point_stages["process"] = ShaderCompiler::STAGE_COMPUTE; + + /* + uses_time = false; + + actions.render_mode_flags["use_half_res_pass"] = &uses_half_res; + actions.render_mode_flags["use_quarter_res_pass"] = &uses_quarter_res; + + actions.usage_flag_pointers["TIME"] = &uses_time; +*/ + + actions.usage_flag_pointers["COLLIDED"] = &uses_collision; + + userdata_count = 0; + for (uint32_t i = 0; i < ParticlesShader::MAX_USERDATAS; i++) { + userdatas_used[i] = false; + actions.usage_flag_pointers["USERDATA" + itos(i + 1)] = &userdatas_used[i]; + } + + actions.uniforms = &uniforms; + + Error err = particles_storage->particles_shader.compiler.compile(RS::SHADER_PARTICLES, code, &actions, path, gen_code); + ERR_FAIL_COND_MSG(err != OK, "Shader compilation failed."); + + if (version.is_null()) { + version = particles_storage->particles_shader.shader.version_create(); + } + + for (uint32_t i = 0; i < ParticlesShader::MAX_USERDATAS; i++) { + if (userdatas_used[i]) { + userdata_count++; + } + } + + particles_storage->particles_shader.shader.version_set_compute_code(version, gen_code.code, gen_code.uniforms, gen_code.stage_globals[ShaderCompiler::STAGE_COMPUTE], gen_code.defines); + ERR_FAIL_COND(!particles_storage->particles_shader.shader.version_is_valid(version)); + + ubo_size = gen_code.uniform_total_size; + ubo_offsets = gen_code.uniform_offsets; + texture_uniforms = gen_code.texture_uniforms; + + //update pipelines + + pipeline = RD::get_singleton()->compute_pipeline_create(particles_storage->particles_shader.shader.version_get_shader(version, 0)); + + valid = true; +} + +void ParticlesStorage::ParticlesShaderData::set_default_texture_parameter(const StringName &p_name, RID p_texture, int p_index) { + if (!p_texture.is_valid()) { + if (default_texture_params.has(p_name) && default_texture_params[p_name].has(p_index)) { + default_texture_params[p_name].erase(p_index); + + if (default_texture_params[p_name].is_empty()) { + default_texture_params.erase(p_name); + } + } + } else { + if (!default_texture_params.has(p_name)) { + default_texture_params[p_name] = HashMap<int, RID>(); + } + default_texture_params[p_name][p_index] = p_texture; + } +} + +void ParticlesStorage::ParticlesShaderData::get_shader_uniform_list(List<PropertyInfo> *p_param_list) const { + HashMap<int, StringName> order; + + for (const KeyValue<StringName, ShaderLanguage::ShaderNode::Uniform> &E : uniforms) { + if (E.value.scope == ShaderLanguage::ShaderNode::Uniform::SCOPE_GLOBAL || E.value.scope == ShaderLanguage::ShaderNode::Uniform::SCOPE_INSTANCE) { + continue; + } + + if (E.value.texture_order >= 0) { + order[E.value.texture_order + 100000] = E.key; + } else { + order[E.value.order] = E.key; + } + } + + String last_group; + for (const KeyValue<int, StringName> &E : order) { + String group = uniforms[E.value].group; + if (!uniforms[E.value].subgroup.is_empty()) { + group += "::" + uniforms[E.value].subgroup; + } + + if (group != last_group) { + PropertyInfo pi; + pi.usage = PROPERTY_USAGE_GROUP; + pi.name = group; + p_param_list->push_back(pi); + + last_group = group; + } + + PropertyInfo pi = ShaderLanguage::uniform_to_property_info(uniforms[E.value]); + pi.name = E.value; + p_param_list->push_back(pi); + } +} + +void ParticlesStorage::ParticlesShaderData::get_instance_param_list(List<RendererMaterialStorage::InstanceShaderParam> *p_param_list) const { + for (const KeyValue<StringName, ShaderLanguage::ShaderNode::Uniform> &E : uniforms) { + if (E.value.scope != ShaderLanguage::ShaderNode::Uniform::SCOPE_INSTANCE) { + continue; + } + + RendererMaterialStorage::InstanceShaderParam p; + p.info = ShaderLanguage::uniform_to_property_info(E.value); + p.info.name = E.key; //supply name + p.index = E.value.instance_index; + p.default_value = ShaderLanguage::constant_value_to_variant(E.value.default_value, E.value.type, E.value.array_size, E.value.hint); + p_param_list->push_back(p); + } +} + +bool ParticlesStorage::ParticlesShaderData::is_parameter_texture(const StringName &p_param) const { + if (!uniforms.has(p_param)) { + return false; + } + + return uniforms[p_param].texture_order >= 0; +} + +bool ParticlesStorage::ParticlesShaderData::is_animated() const { + return false; +} + +bool ParticlesStorage::ParticlesShaderData::casts_shadows() const { + return false; +} + +Variant ParticlesStorage::ParticlesShaderData::get_default_parameter(const StringName &p_parameter) const { + if (uniforms.has(p_parameter)) { + ShaderLanguage::ShaderNode::Uniform uniform = uniforms[p_parameter]; + Vector<ShaderLanguage::ConstantNode::Value> default_value = uniform.default_value; + return ShaderLanguage::constant_value_to_variant(default_value, uniform.type, uniform.array_size, uniform.hint); + } + return Variant(); +} + +RS::ShaderNativeSourceCode ParticlesStorage::ParticlesShaderData::get_native_source_code() const { + return ParticlesStorage::get_singleton()->particles_shader.shader.version_get_native_source_code(version); +} + +ParticlesStorage::ParticlesShaderData::~ParticlesShaderData() { + //pipeline variants will clear themselves if shader is gone + if (version.is_valid()) { + ParticlesStorage::get_singleton()->particles_shader.shader.version_free(version); + } +} + +MaterialStorage::ShaderData *ParticlesStorage::_create_particles_shader_func() { + ParticlesShaderData *shader_data = memnew(ParticlesShaderData); + return shader_data; +} + +bool ParticlesStorage::ParticleProcessMaterialData::update_parameters(const HashMap<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { + return update_parameters_uniform_set(p_parameters, p_uniform_dirty, p_textures_dirty, shader_data->uniforms, shader_data->ubo_offsets.ptr(), shader_data->texture_uniforms, shader_data->default_texture_params, shader_data->ubo_size, uniform_set, ParticlesStorage::get_singleton()->particles_shader.shader.version_get_shader(shader_data->version, 0), 3, true); +} + +ParticlesStorage::ParticleProcessMaterialData::~ParticleProcessMaterialData() { + free_parameters_uniform_set(uniform_set); +} + +MaterialStorage::MaterialData *ParticlesStorage::_create_particles_material_func(ParticlesShaderData *p_shader) { + ParticleProcessMaterialData *material_data = memnew(ParticleProcessMaterialData); + material_data->shader_data = p_shader; + //update will happen later anyway so do nothing. + return material_data; +} +//////// + +/* PARTICLES COLLISION API */ + +RID ParticlesStorage::particles_collision_allocate() { + return particles_collision_owner.allocate_rid(); +} +void ParticlesStorage::particles_collision_initialize(RID p_rid) { + particles_collision_owner.initialize_rid(p_rid, ParticlesCollision()); +} + +void ParticlesStorage::particles_collision_free(RID p_rid) { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_rid); + + if (particles_collision->heightfield_texture.is_valid()) { + RD::get_singleton()->free(particles_collision->heightfield_texture); + } + particles_collision->dependency.deleted_notify(p_rid); + particles_collision_owner.free(p_rid); +} + +RID ParticlesStorage::particles_collision_get_heightfield_framebuffer(RID p_particles_collision) const { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND_V(!particles_collision, RID()); + ERR_FAIL_COND_V(particles_collision->type != RS::PARTICLES_COLLISION_TYPE_HEIGHTFIELD_COLLIDE, RID()); + + if (particles_collision->heightfield_texture == RID()) { + //create + const int resolutions[RS::PARTICLES_COLLISION_HEIGHTFIELD_RESOLUTION_MAX] = { 256, 512, 1024, 2048, 4096, 8192 }; + Size2i size; + if (particles_collision->extents.x > particles_collision->extents.z) { + size.x = resolutions[particles_collision->heightfield_resolution]; + size.y = int32_t(particles_collision->extents.z / particles_collision->extents.x * size.x); + } else { + size.y = resolutions[particles_collision->heightfield_resolution]; + size.x = int32_t(particles_collision->extents.x / particles_collision->extents.z * size.y); + } + + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_D32_SFLOAT; + tf.width = size.x; + tf.height = size.y; + tf.texture_type = RD::TEXTURE_TYPE_2D; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT; + + particles_collision->heightfield_texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); + + Vector<RID> fb_tex; + fb_tex.push_back(particles_collision->heightfield_texture); + particles_collision->heightfield_fb = RD::get_singleton()->framebuffer_create(fb_tex); + particles_collision->heightfield_fb_size = size; + } + + return particles_collision->heightfield_fb; +} + +void ParticlesStorage::particles_collision_set_collision_type(RID p_particles_collision, RS::ParticlesCollisionType p_type) { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND(!particles_collision); + + if (p_type == particles_collision->type) { + return; + } + + if (particles_collision->heightfield_texture.is_valid()) { + RD::get_singleton()->free(particles_collision->heightfield_texture); + particles_collision->heightfield_texture = RID(); + } + particles_collision->type = p_type; + particles_collision->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); +} + +void ParticlesStorage::particles_collision_set_cull_mask(RID p_particles_collision, uint32_t p_cull_mask) { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND(!particles_collision); + particles_collision->cull_mask = p_cull_mask; +} + +void ParticlesStorage::particles_collision_set_sphere_radius(RID p_particles_collision, real_t p_radius) { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND(!particles_collision); + + particles_collision->radius = p_radius; + particles_collision->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); +} + +void ParticlesStorage::particles_collision_set_box_extents(RID p_particles_collision, const Vector3 &p_extents) { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND(!particles_collision); + + particles_collision->extents = p_extents; + particles_collision->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); +} + +void ParticlesStorage::particles_collision_set_attractor_strength(RID p_particles_collision, real_t p_strength) { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND(!particles_collision); + + particles_collision->attractor_strength = p_strength; +} + +void ParticlesStorage::particles_collision_set_attractor_directionality(RID p_particles_collision, real_t p_directionality) { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND(!particles_collision); + + particles_collision->attractor_directionality = p_directionality; +} + +void ParticlesStorage::particles_collision_set_attractor_attenuation(RID p_particles_collision, real_t p_curve) { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND(!particles_collision); + + particles_collision->attractor_attenuation = p_curve; +} + +void ParticlesStorage::particles_collision_set_field_texture(RID p_particles_collision, RID p_texture) { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND(!particles_collision); + + particles_collision->field_texture = p_texture; +} + +void ParticlesStorage::particles_collision_height_field_update(RID p_particles_collision) { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND(!particles_collision); + particles_collision->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); +} + +void ParticlesStorage::particles_collision_set_height_field_resolution(RID p_particles_collision, RS::ParticlesCollisionHeightfieldResolution p_resolution) { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND(!particles_collision); + ERR_FAIL_INDEX(p_resolution, RS::PARTICLES_COLLISION_HEIGHTFIELD_RESOLUTION_MAX); + + if (particles_collision->heightfield_resolution == p_resolution) { + return; + } + + particles_collision->heightfield_resolution = p_resolution; + + if (particles_collision->heightfield_texture.is_valid()) { + RD::get_singleton()->free(particles_collision->heightfield_texture); + particles_collision->heightfield_texture = RID(); + } +} + +AABB ParticlesStorage::particles_collision_get_aabb(RID p_particles_collision) const { + ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND_V(!particles_collision, AABB()); + + switch (particles_collision->type) { + case RS::PARTICLES_COLLISION_TYPE_SPHERE_ATTRACT: + case RS::PARTICLES_COLLISION_TYPE_SPHERE_COLLIDE: { + AABB aabb; + aabb.position = -Vector3(1, 1, 1) * particles_collision->radius; + aabb.size = Vector3(2, 2, 2) * particles_collision->radius; + return aabb; + } + default: { + AABB aabb; + aabb.position = -particles_collision->extents; + aabb.size = particles_collision->extents * 2; + return aabb; + } + } +} + +Vector3 ParticlesStorage::particles_collision_get_extents(RID p_particles_collision) const { + const ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND_V(!particles_collision, Vector3()); + return particles_collision->extents; +} + +bool ParticlesStorage::particles_collision_is_heightfield(RID p_particles_collision) const { + const ParticlesCollision *particles_collision = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_COND_V(!particles_collision, false); + return particles_collision->type == RS::PARTICLES_COLLISION_TYPE_HEIGHTFIELD_COLLIDE; +} + +Dependency *ParticlesStorage::particles_collision_get_dependency(RID p_particles_collision) const { + ParticlesCollision *pc = particles_collision_owner.get_or_null(p_particles_collision); + ERR_FAIL_NULL_V(pc, nullptr); + + return &pc->dependency; +} + +/* Particles collision instance */ + +RID ParticlesStorage::particles_collision_instance_create(RID p_collision) { + ParticlesCollisionInstance pci; + pci.collision = p_collision; + return particles_collision_instance_owner.make_rid(pci); +} + +void ParticlesStorage::particles_collision_instance_free(RID p_rid) { + particles_collision_instance_owner.free(p_rid); +} + +void ParticlesStorage::particles_collision_instance_set_transform(RID p_collision_instance, const Transform3D &p_transform) { + ParticlesCollisionInstance *pci = particles_collision_instance_owner.get_or_null(p_collision_instance); + ERR_FAIL_COND(!pci); + pci->transform = p_transform; +} + +void ParticlesStorage::particles_collision_instance_set_active(RID p_collision_instance, bool p_active) { + ParticlesCollisionInstance *pci = particles_collision_instance_owner.get_or_null(p_collision_instance); + ERR_FAIL_COND(!pci); + pci->active = p_active; +} diff --git a/servers/rendering/renderer_rd/storage_rd/particles_storage.h b/servers/rendering/renderer_rd/storage_rd/particles_storage.h new file mode 100644 index 0000000000..017844626f --- /dev/null +++ b/servers/rendering/renderer_rd/storage_rd/particles_storage.h @@ -0,0 +1,564 @@ +/*************************************************************************/ +/* particles_storage.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef PARTICLES_STORAGE_RD_H +#define PARTICLES_STORAGE_RD_H + +#include "core/templates/local_vector.h" +#include "core/templates/rid_owner.h" +#include "core/templates/self_list.h" +#include "servers/rendering/renderer_rd/shaders/particles.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/particles_copy.glsl.gen.h" +#include "servers/rendering/renderer_rd/storage_rd/material_storage.h" +#include "servers/rendering/shader_compiler.h" +#include "servers/rendering/storage/particles_storage.h" +#include "servers/rendering/storage/utilities.h" + +namespace RendererRD { + +class ParticlesStorage : public RendererParticlesStorage { +private: + static ParticlesStorage *singleton; + + /* PARTICLES */ + + struct ParticleData { + float xform[16]; + float velocity[3]; + uint32_t active; + float color[4]; + float custom[3]; + float lifetime; + }; + + struct ParticlesFrameParams { + enum { + MAX_ATTRACTORS = 32, + MAX_COLLIDERS = 32, + MAX_3D_TEXTURES = 7 + }; + + enum AttractorType { + ATTRACTOR_TYPE_SPHERE, + ATTRACTOR_TYPE_BOX, + ATTRACTOR_TYPE_VECTOR_FIELD, + }; + + struct Attractor { + float transform[16]; + float extents[3]; //exents or radius + uint32_t type; + + uint32_t texture_index; //texture index for vector field + float strength; + float attenuation; + float directionality; + }; + + enum CollisionType { + COLLISION_TYPE_SPHERE, + COLLISION_TYPE_BOX, + COLLISION_TYPE_SDF, + COLLISION_TYPE_HEIGHT_FIELD, + COLLISION_TYPE_2D_SDF, + + }; + + struct Collider { + float transform[16]; + float extents[3]; //exents or radius + uint32_t type; + + uint32_t texture_index; //texture index for vector field + real_t scale; + uint32_t pad[2]; + }; + + uint32_t emitting; + float system_phase; + float prev_system_phase; + uint32_t cycle; + + real_t explosiveness; + real_t randomness; + float time; + float delta; + + uint32_t frame; + uint32_t pad0; + uint32_t pad1; + uint32_t pad2; + + uint32_t random_seed; + uint32_t attractor_count; + uint32_t collider_count; + float particle_size; + + float emission_transform[16]; + + Attractor attractors[MAX_ATTRACTORS]; + Collider colliders[MAX_COLLIDERS]; + }; + + struct ParticleEmissionBufferData { + }; + + struct ParticleEmissionBuffer { + struct Data { + float xform[16]; + float velocity[3]; + uint32_t flags; + float color[4]; + float custom[4]; + }; + + int32_t particle_count; + int32_t particle_max; + uint32_t pad1; + uint32_t pad2; + Data data[1]; //its 2020 and empty arrays are still non standard in C++ + }; + + struct Particles { + RS::ParticlesMode mode = RS::PARTICLES_MODE_3D; + bool inactive = true; + double inactive_time = 0.0; + bool emitting = false; + bool one_shot = false; + int amount = 0; + double lifetime = 1.0; + double pre_process_time = 0.0; + real_t explosiveness = 0.0; + real_t randomness = 0.0; + bool restart_request = false; + AABB custom_aabb = AABB(Vector3(-4, -4, -4), Vector3(8, 8, 8)); + bool use_local_coords = false; + bool has_collision_cache = false; + + bool has_sdf_collision = false; + Transform2D sdf_collision_transform; + Rect2 sdf_collision_to_screen; + RID sdf_collision_texture; + + RID process_material; + uint32_t frame_counter = 0; + RS::ParticlesTransformAlign transform_align = RS::PARTICLES_TRANSFORM_ALIGN_DISABLED; + + RS::ParticlesDrawOrder draw_order = RS::PARTICLES_DRAW_ORDER_INDEX; + + Vector<RID> draw_passes; + Vector<Transform3D> trail_bind_poses; + bool trail_bind_poses_dirty = false; + RID trail_bind_pose_buffer; + RID trail_bind_pose_uniform_set; + + RID particle_buffer; + RID particle_instance_buffer; + RID frame_params_buffer; + + uint32_t userdata_count = 0; + + RID particles_material_uniform_set; + RID particles_copy_uniform_set; + RID particles_transforms_buffer_uniform_set; + RID collision_textures_uniform_set; + + RID collision_3d_textures[ParticlesFrameParams::MAX_3D_TEXTURES]; + uint32_t collision_3d_textures_used = 0; + RID collision_heightmap_texture; + + RID particles_sort_buffer; + RID particles_sort_uniform_set; + + bool dirty = false; + Particles *update_list = nullptr; + + RID sub_emitter; + + double phase = 0.0; + double prev_phase = 0.0; + uint64_t prev_ticks = 0; + uint32_t random_seed = 0; + + uint32_t cycle_number = 0; + + double speed_scale = 1.0; + + int fixed_fps = 30; + bool interpolate = true; + bool fractional_delta = false; + double frame_remainder = 0; + real_t collision_base_size = 0.01; + + bool clear = true; + + bool force_sub_emit = false; + + Transform3D emission_transform; + + Vector<uint8_t> emission_buffer_data; + + ParticleEmissionBuffer *emission_buffer = nullptr; + RID emission_storage_buffer; + + HashSet<RID> collisions; + + Dependency dependency; + + double trail_length = 1.0; + bool trails_enabled = false; + LocalVector<ParticlesFrameParams> frame_history; + LocalVector<ParticlesFrameParams> trail_params; + + Particles() { + } + }; + + void _particles_process(Particles *p_particles, double p_delta); + void _particles_allocate_emission_buffer(Particles *particles); + void _particles_free_data(Particles *particles); + void _particles_update_buffers(Particles *particles); + + struct ParticlesShader { + struct PushConstant { + float lifetime; + uint32_t clear; + uint32_t total_particles; + uint32_t trail_size; + + uint32_t use_fractional_delta; + uint32_t sub_emitter_mode; + uint32_t can_emit; + uint32_t trail_pass; + }; + + ParticlesShaderRD shader; + ShaderCompiler compiler; + + RID default_shader; + RID default_material; + RID default_shader_rd; + + RID base_uniform_set; + + struct CopyPushConstant { + float sort_direction[3]; + uint32_t total_particles; + + uint32_t trail_size; + uint32_t trail_total; + float frame_delta; + float frame_remainder; + + float align_up[3]; + uint32_t align_mode; + + uint32_t order_by_lifetime; + uint32_t lifetime_split; + uint32_t lifetime_reverse; + uint32_t copy_mode_2d; + + float inv_emission_transform[16]; + }; + + enum { + MAX_USERDATAS = 6 + }; + enum { + COPY_MODE_FILL_INSTANCES, + COPY_MODE_FILL_SORT_BUFFER, + COPY_MODE_FILL_INSTANCES_WITH_SORT_BUFFER, + COPY_MODE_MAX, + }; + + ParticlesCopyShaderRD copy_shader; + RID copy_shader_version; + RID copy_pipelines[COPY_MODE_MAX * (MAX_USERDATAS + 1)]; + + LocalVector<float> pose_update_buffer; + + } particles_shader; + + Particles *particle_update_list = nullptr; + + mutable RID_Owner<Particles, true> particles_owner; + + /* Particle Shader */ + + struct ParticlesShaderData : public MaterialStorage::ShaderData { + bool valid = false; + RID version; + bool uses_collision = false; + + HashMap<StringName, ShaderLanguage::ShaderNode::Uniform> uniforms; + Vector<ShaderCompiler::GeneratedCode::Texture> texture_uniforms; + + Vector<uint32_t> ubo_offsets; + uint32_t ubo_size = 0; + + String path; + String code; + HashMap<StringName, HashMap<int, RID>> default_texture_params; + + RID pipeline; + + bool uses_time = false; + + bool userdatas_used[ParticlesShader::MAX_USERDATAS] = {}; + uint32_t userdata_count = 0; + + virtual void set_code(const String &p_Code); + virtual void set_path_hint(const String &p_hint); + virtual void set_default_texture_parameter(const StringName &p_name, RID p_texture, int p_index); + virtual void get_shader_uniform_list(List<PropertyInfo> *p_param_list) const; + virtual void get_instance_param_list(List<RendererMaterialStorage::InstanceShaderParam> *p_param_list) const; + virtual bool is_parameter_texture(const StringName &p_param) const; + virtual bool is_animated() const; + virtual bool casts_shadows() const; + virtual Variant get_default_parameter(const StringName &p_parameter) const; + virtual RS::ShaderNativeSourceCode get_native_source_code() const; + + ParticlesShaderData() {} + virtual ~ParticlesShaderData(); + }; + + MaterialStorage::ShaderData *_create_particles_shader_func(); + static MaterialStorage::ShaderData *_create_particles_shader_funcs() { + return ParticlesStorage::get_singleton()->_create_particles_shader_func(); + } + + struct ParticleProcessMaterialData : public MaterialStorage::MaterialData { + ParticlesShaderData *shader_data = nullptr; + RID uniform_set; + + virtual void set_render_priority(int p_priority) {} + virtual void set_next_pass(RID p_pass) {} + virtual bool update_parameters(const HashMap<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); + virtual ~ParticleProcessMaterialData(); + }; + + MaterialStorage::MaterialData *_create_particles_material_func(ParticlesShaderData *p_shader); + static MaterialStorage::MaterialData *_create_particles_material_funcs(MaterialStorage::ShaderData *p_shader) { + return ParticlesStorage::get_singleton()->_create_particles_material_func(static_cast<ParticlesShaderData *>(p_shader)); + } + + /* Particles Collision */ + + struct ParticlesCollision { + RS::ParticlesCollisionType type = RS::PARTICLES_COLLISION_TYPE_SPHERE_ATTRACT; + uint32_t cull_mask = 0xFFFFFFFF; + float radius = 1.0; + Vector3 extents = Vector3(1, 1, 1); + float attractor_strength = 1.0; + float attractor_attenuation = 1.0; + float attractor_directionality = 0.0; + RID field_texture; + RID heightfield_texture; + RID heightfield_fb; + Size2i heightfield_fb_size; + + RS::ParticlesCollisionHeightfieldResolution heightfield_resolution = RS::PARTICLES_COLLISION_HEIGHTFIELD_RESOLUTION_1024; + + Dependency dependency; + }; + + struct ParticlesCollisionInstance { + RID collision; + Transform3D transform; + bool active = false; + }; + + mutable RID_Owner<ParticlesCollision, true> particles_collision_owner; + + mutable RID_Owner<ParticlesCollisionInstance> particles_collision_instance_owner; + +public: + static ParticlesStorage *get_singleton(); + + ParticlesStorage(); + virtual ~ParticlesStorage(); + + bool free(RID p_rid); + + /* PARTICLES */ + + bool owns_particles(RID p_rid) { return particles_owner.owns(p_rid); } + + virtual RID particles_allocate() override; + virtual void particles_initialize(RID p_particles_collision) override; + virtual void particles_free(RID p_rid) override; + + virtual void particles_set_mode(RID p_particles, RS::ParticlesMode p_mode) override; + virtual void particles_set_emitting(RID p_particles, bool p_emitting) override; + virtual void particles_set_amount(RID p_particles, int p_amount) override; + virtual void particles_set_lifetime(RID p_particles, double p_lifetime) override; + virtual void particles_set_one_shot(RID p_particles, bool p_one_shot) override; + virtual void particles_set_pre_process_time(RID p_particles, double p_time) override; + virtual void particles_set_explosiveness_ratio(RID p_particles, real_t p_ratio) override; + virtual void particles_set_randomness_ratio(RID p_particles, real_t p_ratio) override; + virtual void particles_set_custom_aabb(RID p_particles, const AABB &p_aabb) override; + virtual void particles_set_speed_scale(RID p_particles, double p_scale) override; + virtual void particles_set_use_local_coordinates(RID p_particles, bool p_enable) override; + virtual void particles_set_process_material(RID p_particles, RID p_material) override; + virtual RID particles_get_process_material(RID p_particles) const override; + + virtual void particles_set_fixed_fps(RID p_particles, int p_fps) override; + virtual void particles_set_interpolate(RID p_particles, bool p_enable) override; + virtual void particles_set_fractional_delta(RID p_particles, bool p_enable) override; + virtual void particles_set_collision_base_size(RID p_particles, real_t p_size) override; + virtual void particles_set_transform_align(RID p_particles, RS::ParticlesTransformAlign p_transform_align) override; + + virtual void particles_set_trails(RID p_particles, bool p_enable, double p_length) override; + virtual void particles_set_trail_bind_poses(RID p_particles, const Vector<Transform3D> &p_bind_poses) override; + + virtual void particles_restart(RID p_particles) override; + virtual void particles_emit(RID p_particles, const Transform3D &p_transform, const Vector3 &p_velocity, const Color &p_color, const Color &p_custom, uint32_t p_emit_flags) override; + + virtual void particles_set_subemitter(RID p_particles, RID p_subemitter_particles) override; + + virtual void particles_set_draw_order(RID p_particles, RS::ParticlesDrawOrder p_order) override; + + virtual void particles_set_draw_passes(RID p_particles, int p_count) override; + virtual void particles_set_draw_pass_mesh(RID p_particles, int p_pass, RID p_mesh) override; + + virtual void particles_request_process(RID p_particles) override; + virtual AABB particles_get_current_aabb(RID p_particles) override; + virtual AABB particles_get_aabb(RID p_particles) const override; + + virtual void particles_set_emission_transform(RID p_particles, const Transform3D &p_transform) override; + + virtual bool particles_get_emitting(RID p_particles) override; + virtual int particles_get_draw_passes(RID p_particles) const override; + virtual RID particles_get_draw_pass_mesh(RID p_particles, int p_pass) const override; + + virtual void particles_set_view_axis(RID p_particles, const Vector3 &p_axis, const Vector3 &p_up_axis) override; + + virtual bool particles_is_inactive(RID p_particles) const override; + + _FORCE_INLINE_ RS::ParticlesMode particles_get_mode(RID p_particles) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND_V(!particles, RS::PARTICLES_MODE_2D); + return particles->mode; + } + + _FORCE_INLINE_ uint32_t particles_get_amount(RID p_particles, uint32_t &r_trail_divisor) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND_V(!particles, 0); + + if (particles->trails_enabled && particles->trail_bind_poses.size() > 1) { + r_trail_divisor = particles->trail_bind_poses.size(); + } else { + r_trail_divisor = 1; + } + + return particles->amount * r_trail_divisor; + } + + _FORCE_INLINE_ bool particles_has_collision(RID p_particles) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND_V(!particles, 0); + + return particles->has_collision_cache; + } + + _FORCE_INLINE_ uint32_t particles_is_using_local_coords(RID p_particles) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND_V(!particles, false); + + return particles->use_local_coords; + } + + _FORCE_INLINE_ RID particles_get_instance_buffer_uniform_set(RID p_particles, RID p_shader, uint32_t p_set) { + Particles *particles = particles_owner.get_or_null(p_particles); + ERR_FAIL_COND_V(!particles, RID()); + if (particles->particles_transforms_buffer_uniform_set.is_null()) { + _particles_update_buffers(particles); + + Vector<RD::Uniform> uniforms; + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 0; + u.append_id(particles->particle_instance_buffer); + uniforms.push_back(u); + } + + particles->particles_transforms_buffer_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, p_shader, p_set); + } + + return particles->particles_transforms_buffer_uniform_set; + } + + virtual void particles_add_collision(RID p_particles, RID p_particles_collision_instance) override; + virtual void particles_remove_collision(RID p_particles, RID p_particles_collision_instance) override; + virtual void particles_set_canvas_sdf_collision(RID p_particles, bool p_enable, const Transform2D &p_xform, const Rect2 &p_to_screen, RID p_texture) override; + + virtual void update_particles() override; + + Dependency *particles_get_dependency(RID p_particles) const; + + /* Particles Collision */ + + bool owns_particles_collision(RID p_rid) { return particles_collision_owner.owns(p_rid); } + + virtual RID particles_collision_allocate() override; + virtual void particles_collision_initialize(RID p_particles_collision) override; + virtual void particles_collision_free(RID p_rid) override; + + virtual void particles_collision_set_collision_type(RID p_particles_collision, RS::ParticlesCollisionType p_type) override; + virtual void particles_collision_set_cull_mask(RID p_particles_collision, uint32_t p_cull_mask) override; + virtual void particles_collision_set_sphere_radius(RID p_particles_collision, real_t p_radius) override; //for spheres + virtual void particles_collision_set_box_extents(RID p_particles_collision, const Vector3 &p_extents) override; //for non-spheres + virtual void particles_collision_set_attractor_strength(RID p_particles_collision, real_t p_strength) override; + virtual void particles_collision_set_attractor_directionality(RID p_particles_collision, real_t p_directionality) override; + virtual void particles_collision_set_attractor_attenuation(RID p_particles_collision, real_t p_curve) override; + virtual void particles_collision_set_field_texture(RID p_particles_collision, RID p_texture) override; //for SDF and vector field, heightfield is dynamic + virtual void particles_collision_height_field_update(RID p_particles_collision) override; //for SDF and vector field + virtual void particles_collision_set_height_field_resolution(RID p_particles_collision, RS::ParticlesCollisionHeightfieldResolution p_resolution) override; //for SDF and vector field + virtual AABB particles_collision_get_aabb(RID p_particles_collision) const override; + Vector3 particles_collision_get_extents(RID p_particles_collision) const; + virtual bool particles_collision_is_heightfield(RID p_particles_collision) const override; + virtual RID particles_collision_get_heightfield_framebuffer(RID p_particles_collision) const override; + + Dependency *particles_collision_get_dependency(RID p_particles) const; + + //used from 2D and 3D + bool owns_particles_collision_instance(RID p_rid) { return particles_collision_instance_owner.owns(p_rid); } + + virtual RID particles_collision_instance_create(RID p_collision) override; + virtual void particles_collision_instance_free(RID p_rid) override; + virtual void particles_collision_instance_set_transform(RID p_collision_instance, const Transform3D &p_transform) override; + virtual void particles_collision_instance_set_active(RID p_collision_instance, bool p_active) override; +}; + +} // namespace RendererRD + +#endif // PARTICLES_STORAGE_RD_H diff --git a/servers/rendering/renderer_rd/storage_rd/render_buffer_custom_data_rd.h b/servers/rendering/renderer_rd/storage_rd/render_buffer_custom_data_rd.h new file mode 100644 index 0000000000..d904012914 --- /dev/null +++ b/servers/rendering/renderer_rd/storage_rd/render_buffer_custom_data_rd.h @@ -0,0 +1,48 @@ +/*************************************************************************/ +/* render_buffer_custom_data_rd.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef RENDER_BUFFER_CUSTOM_DATA_RD_H +#define RENDER_BUFFER_CUSTOM_DATA_RD_H + +#include "core/object/ref_counted.h" + +class RenderSceneBuffersRD; + +class RenderBufferCustomDataRD : public RefCounted { + GDCLASS(RenderBufferCustomDataRD, RefCounted); + +public: + virtual void configure(RenderSceneBuffersRD *p_render_buffers) = 0; + virtual void free_data() = 0; // called on cleanup + +private: +}; + +#endif // RENDER_BUFFER_CUSTOM_DATA_RD_H diff --git a/servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.cpp b/servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.cpp new file mode 100644 index 0000000000..0c2092f03e --- /dev/null +++ b/servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.cpp @@ -0,0 +1,587 @@ +/*************************************************************************/ +/* render_scene_buffers_rd.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "render_scene_buffers_rd.h" +#include "servers/rendering/renderer_rd/renderer_scene_render_rd.h" +#include "servers/rendering/renderer_rd/storage_rd/material_storage.h" +#include "servers/rendering/renderer_rd/storage_rd/texture_storage.h" + +RenderSceneBuffersRD::RenderSceneBuffersRD() { +} + +RenderSceneBuffersRD::~RenderSceneBuffersRD() { + cleanup(); + + data_buffers.clear(); +} + +void RenderSceneBuffersRD::_bind_methods() { + ClassDB::bind_method(D_METHOD("has_texture", "context", "name"), &RenderSceneBuffersRD::has_texture); + // FIXME we can't pass RD::DataFormat, RD::TextureSamples and RD::TextureView in ClassDB, need to solve views differently... + // ClassDB::bind_method(D_METHOD("create_texture", "context", "name", "data_format", "usage_bits", "texture_samples", "size", "layers", "mipmaps", "unique"), &RenderSceneBuffersRD::create_texture); + // ClassDB::bind_method(D_METHOD("create_texture_from_format", "context", "name", "format", "view", "unique"), &RenderSceneBuffersRD::create_texture_from_format); + // ClassDB::bind_method(D_METHOD("create_texture_view", "context", "name", "view_name", "view"), &RenderSceneBuffersRD::has_texture); + ClassDB::bind_method(D_METHOD("get_texture", "context", "name"), &RenderSceneBuffersRD::get_texture); + // ClassDB::bind_method(D_METHOD("get_texture_format", "context", "name"), &RenderSceneBuffersRD::get_texture_format); + ClassDB::bind_method(D_METHOD("get_texture_slice", "context", "name", "layer", "mipmap"), &RenderSceneBuffersRD::get_texture_slice); + ClassDB::bind_method(D_METHOD("get_texture_slice_size", "context", "name", "layer", "mipmap"), &RenderSceneBuffersRD::get_texture_slice_size); + ClassDB::bind_method(D_METHOD("clear_context", "context"), &RenderSceneBuffersRD::clear_context); +} + +void RenderSceneBuffersRD::update_sizes(NamedTexture &p_named_texture) { + ERR_FAIL_COND(p_named_texture.texture.is_null()); + + uint32_t size = p_named_texture.format.array_layers * p_named_texture.format.mipmaps; + p_named_texture.sizes.resize(size); + + Size2i mipmap_size = Size2i(p_named_texture.format.width, p_named_texture.format.height); + + for (uint32_t mipmap = 0; mipmap < p_named_texture.format.mipmaps; mipmap++) { + for (uint32_t layer = 0; layer < p_named_texture.format.array_layers; layer++) { + uint32_t index = layer * p_named_texture.format.mipmaps + mipmap; + + p_named_texture.sizes.ptrw()[index] = mipmap_size; + } + + mipmap_size.width = MAX(1, mipmap_size.width >> 1); + mipmap_size.height = MAX(1, mipmap_size.height >> 1); + } +} + +void RenderSceneBuffersRD::free_named_texture(NamedTexture &p_named_texture) { + if (p_named_texture.texture.is_valid()) { + RD::get_singleton()->free(p_named_texture.texture); + } + p_named_texture.texture = RID(); + p_named_texture.slices.clear(); // slices should be freed automatically as dependents... +} + +void RenderSceneBuffersRD::cleanup() { + // Free our data buffers (but don't destroy them) + for (KeyValue<StringName, Ref<RenderBufferCustomDataRD>> &E : data_buffers) { + E.value->free_data(); + } + + // Clear our named textures + for (KeyValue<NTKey, NamedTexture> &E : named_textures) { + free_named_texture(E.value); + } + named_textures.clear(); + + // old stuff, to be re-evaluated... + + for (int i = 0; i < luminance.fb.size(); i++) { + RD::get_singleton()->free(luminance.fb[i]); + } + luminance.fb.clear(); + + for (int i = 0; i < luminance.reduce.size(); i++) { + RD::get_singleton()->free(luminance.reduce[i]); + } + luminance.reduce.clear(); + + if (luminance.current_fb.is_valid()) { + RD::get_singleton()->free(luminance.current_fb); + luminance.current_fb = RID(); + } + + if (luminance.current.is_valid()) { + RD::get_singleton()->free(luminance.current); + luminance.current = RID(); + } +} + +void RenderSceneBuffersRD::configure(RID p_render_target, const Size2i p_internal_size, const Size2i p_target_size, float p_fsr_sharpness, float p_texture_mipmap_bias, RS::ViewportMSAA p_msaa_3d, RenderingServer::ViewportScreenSpaceAA p_screen_space_aa, bool p_use_taa, bool p_use_debanding, uint32_t p_view_count) { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + + ERR_FAIL_COND_MSG(p_view_count == 0, "Must have at least 1 view"); + + target_size = p_target_size; + internal_size = p_internal_size; + + // FIXME, right now we do this because only our clustered renderer supports FSR upscale + // this does mean that with linear upscale if we use subpasses, we could get into trouble. + if (!can_be_storage) { + internal_size = target_size; + } + + if (p_use_taa) { + // Use negative mipmap LOD bias when TAA is enabled to compensate for loss of sharpness. + // This restores sharpness in still images to be roughly at the same level as without TAA, + // but moving scenes will still be blurrier. + p_texture_mipmap_bias -= 0.5; + } + + if (p_screen_space_aa == RS::VIEWPORT_SCREEN_SPACE_AA_FXAA) { + // Use negative mipmap LOD bias when FXAA is enabled to compensate for loss of sharpness. + // If both TAA and FXAA are enabled, combine their negative LOD biases together. + p_texture_mipmap_bias -= 0.25; + } + + material_storage->sampler_rd_configure_custom(p_texture_mipmap_bias); + + // need to check if we really need to do this here.. + RendererSceneRenderRD::get_singleton()->update_uniform_sets(); + + render_target = p_render_target; + fsr_sharpness = p_fsr_sharpness; + msaa_3d = p_msaa_3d; + screen_space_aa = p_screen_space_aa; + use_taa = p_use_taa; + use_debanding = p_use_debanding; + view_count = p_view_count; + + // cleanout any old buffers we had. + cleanup(); + + // create our 3D render buffers + { + // Create our color buffer(s) + uint32_t usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | (can_be_storage ? RD::TEXTURE_USAGE_STORAGE_BIT : 0) | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + usage_bits |= RD::TEXTURE_USAGE_INPUT_ATTACHMENT_BIT; // only needed when using subpasses in the mobile renderer + + // our internal texture should have MSAA support if applicable + if (msaa_3d != RS::VIEWPORT_MSAA_DISABLED) { + usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; + } + + create_texture(RB_SCOPE_BUFFERS, RB_TEX_COLOR, base_data_format, usage_bits); + } + + // Create our depth buffer + { + // TODO Lazy create this in case we've got an external depth buffer + + RD::DataFormat format; + uint32_t usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT; + + if (msaa_3d == RS::VIEWPORT_MSAA_DISABLED) { + format = RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_D24_UNORM_S8_UINT, (RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT)) ? RD::DATA_FORMAT_D24_UNORM_S8_UINT : RD::DATA_FORMAT_D32_SFLOAT_S8_UINT; + usage_bits |= RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT; + } else { + format = RD::DATA_FORMAT_R32_SFLOAT; + usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + } + + create_texture(RB_SCOPE_BUFFERS, RB_TEX_DEPTH, format, usage_bits); + } + + // VRS (note, our vrs object will only be set if VRS is supported) + RID vrs_texture; + RS::ViewportVRSMode vrs_mode = texture_storage->render_target_get_vrs_mode(p_render_target); + if (vrs && vrs_mode != RS::VIEWPORT_VRS_DISABLED) { + uint32_t usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_VRS_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + vrs_texture = create_texture(RB_SCOPE_VRS, RB_TEXTURE, RD::DATA_FORMAT_R8_UINT, usage_bits, RD::TEXTURE_SAMPLES_1, vrs->get_vrs_texture_size(internal_size)); + } + + for (KeyValue<StringName, Ref<RenderBufferCustomDataRD>> &E : data_buffers) { + E.value->configure(this); + } +} + +void RenderSceneBuffersRD::set_fsr_sharpness(float p_fsr_sharpness) { + fsr_sharpness = p_fsr_sharpness; +} + +void RenderSceneBuffersRD::set_texture_mipmap_bias(float p_texture_mipmap_bias) { + RendererRD::MaterialStorage *material_storage = RendererRD::MaterialStorage::get_singleton(); + material_storage->sampler_rd_configure_custom(p_texture_mipmap_bias); +} + +void RenderSceneBuffersRD::set_use_debanding(bool p_use_debanding) { + use_debanding = p_use_debanding; +} + +// Named textures + +bool RenderSceneBuffersRD::has_texture(const StringName &p_context, const StringName &p_texture_name) const { + NTKey key(p_context, p_texture_name); + + return named_textures.has(key); +} + +RID RenderSceneBuffersRD::create_texture(const StringName &p_context, const StringName &p_texture_name, const RD::DataFormat p_data_format, const uint32_t p_usage_bits, const RD::TextureSamples p_texture_samples, const Size2i p_size, const uint32_t p_layers, const uint32_t p_mipmaps, bool p_unique) { + // Keep some useful data, we use default values when these are 0. + Size2i size = p_size == Size2i(0, 0) ? internal_size : p_size; + uint32_t layers = p_layers == 0 ? view_count : p_layers; + uint32_t mipmaps = p_mipmaps == 0 ? 1 : p_mipmaps; + + // Create our texture + RD::TextureFormat tf; + tf.format = p_data_format; + if (layers > 1) { + tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; + } + + tf.width = size.x; + tf.height = size.y; + tf.depth = 1; + tf.array_layers = layers; + tf.mipmaps = mipmaps; + tf.usage_bits = p_usage_bits; + tf.samples = p_texture_samples; + + return create_texture_from_format(p_context, p_texture_name, tf, RD::TextureView(), p_unique); +} + +RID RenderSceneBuffersRD::create_texture_from_format(const StringName &p_context, const StringName &p_texture_name, const RD::TextureFormat &p_texture_format, RD::TextureView p_view, bool p_unique) { + // TODO p_unique, if p_unique is true, this is a texture that can be shared. This will be implemented later as an optimisation. + + NTKey key(p_context, p_texture_name); + + // check if this is a known texture + if (named_textures.has(key)) { + return named_textures[key].texture; + } + + // Add a new entry.. + NamedTexture &named_texture = named_textures[key]; + named_texture.format = p_texture_format; + named_texture.is_unique = p_unique; + named_texture.texture = RD::get_singleton()->texture_create(p_texture_format, p_view); + + Array arr; + arr.push_back(p_context); + arr.push_back(p_texture_name); + RD::get_singleton()->set_resource_name(named_texture.texture, String("RenderBuffer {0}/{1}").format(arr)); + + update_sizes(named_texture); + + // The rest is lazy created.. + + return named_texture.texture; +} + +RID RenderSceneBuffersRD::create_texture_view(const StringName &p_context, const StringName &p_texture_name, const StringName p_view_name, RD::TextureView p_view) { + NTKey view_key(p_context, p_view_name); + + // check if this is a known texture + if (named_textures.has(view_key)) { + return named_textures[view_key].texture; + } + + NTKey key(p_context, p_texture_name); + + ERR_FAIL_COND_V(!named_textures.has(key), RID()); + + NamedTexture &named_texture = named_textures[key]; + NamedTexture &view_texture = named_textures[view_key]; + + view_texture.format = named_texture.format; + view_texture.is_unique = named_texture.is_unique; + + view_texture.texture = RD::get_singleton()->texture_create_shared(p_view, named_texture.texture); + + Array arr; + arr.push_back(p_context); + arr.push_back(p_view_name); + RD::get_singleton()->set_resource_name(view_texture.texture, String("RenderBuffer View {0}/{1}").format(arr)); + + update_sizes(named_texture); + + return view_texture.texture; +} + +RID RenderSceneBuffersRD::get_texture(const StringName &p_context, const StringName &p_texture_name) const { + NTKey key(p_context, p_texture_name); + + ERR_FAIL_COND_V(!named_textures.has(key), RID()); + + return named_textures[key].texture; +} + +const RD::TextureFormat RenderSceneBuffersRD::get_texture_format(const StringName &p_context, const StringName &p_texture_name) const { + NTKey key(p_context, p_texture_name); + + ERR_FAIL_COND_V(!named_textures.has(key), RD::TextureFormat()); + + return named_textures[key].format; +} + +RID RenderSceneBuffersRD::get_texture_slice(const StringName &p_context, const StringName &p_texture_name, const uint32_t p_layer, const uint32_t p_mipmap) { + NTKey key(p_context, p_texture_name); + + // check if this is a known texture + ERR_FAIL_COND_V(!named_textures.has(key), RID()); + NamedTexture &named_texture = named_textures[key]; + ERR_FAIL_COND_V(named_texture.texture.is_null(), RID()); + + // check if we're in bounds + ERR_FAIL_UNSIGNED_INDEX_V(p_layer, named_texture.format.array_layers, RID()); + ERR_FAIL_UNSIGNED_INDEX_V(p_mipmap, named_texture.format.mipmaps, RID()); + + // if we don't have multiple layers or mipmaps, we can just return our texture as is + if (named_texture.format.array_layers == 1 && named_texture.format.mipmaps == 1) { + return named_texture.texture; + } + + // get our index and make sure we have enough entries in our slices vector + uint32_t index = p_layer * named_texture.format.mipmaps + p_mipmap; + while (named_texture.slices.size() <= int(index)) { + named_texture.slices.push_back(RID()); + } + + // create our slice if we don't have it already + if (named_texture.slices[index].is_null()) { + named_texture.slices.ptrw()[index] = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), named_texture.texture, p_layer, p_mipmap); + + Array arr; + arr.push_back(p_context); + arr.push_back(p_texture_name); + arr.push_back(itos(p_layer)); + arr.push_back(itos(p_mipmap)); + RD::get_singleton()->set_resource_name(named_texture.slices[index], String("RenderBuffer {0}/{1} slice {2}/{3}").format(arr)); + } + + // and return our slice + return named_texture.slices[index]; +} + +Size2i RenderSceneBuffersRD::get_texture_slice_size(const StringName &p_context, const StringName &p_texture_name, const uint32_t p_layer, const uint32_t p_mipmap) { + NTKey key(p_context, p_texture_name); + + // check if this is a known texture + ERR_FAIL_COND_V(!named_textures.has(key), Size2i()); + NamedTexture &named_texture = named_textures[key]; + ERR_FAIL_COND_V(named_texture.texture.is_null(), Size2i()); + + // check if we're in bounds + ERR_FAIL_UNSIGNED_INDEX_V(p_layer, named_texture.format.array_layers, Size2i()); + ERR_FAIL_UNSIGNED_INDEX_V(p_mipmap, named_texture.format.mipmaps, Size2i()); + + // get our index + uint32_t index = p_layer * named_texture.format.mipmaps + p_mipmap; + + // and return our size + return named_texture.sizes[index]; +} + +void RenderSceneBuffersRD::clear_context(const StringName &p_context) { + Vector<NTKey> to_free; // free these + + // Find all entries for our context, we don't want to free them yet or our loop fails. + for (KeyValue<NTKey, NamedTexture> &E : named_textures) { + if (E.key.context == p_context) { + to_free.push_back(E.key); + } + } + + // Now free these and remove them from our textures + for (NTKey &key : to_free) { + free_named_texture(named_textures[key]); + named_textures.erase(key); + } +} + +// Allocate shared buffers +void RenderSceneBuffersRD::allocate_blur_textures() { + if (has_texture(RB_SCOPE_BUFFERS, RB_TEX_BLUR_0)) { + // already allocated... + return; + } + + uint32_t mipmaps_required = Image::get_image_required_mipmaps(internal_size.x, internal_size.y, Image::FORMAT_RGBAH); + + uint32_t usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; + if (can_be_storage) { + usage_bits += RD::TEXTURE_USAGE_STORAGE_BIT; + } else { + usage_bits += RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + } + + create_texture(RB_SCOPE_BUFFERS, RB_TEX_BLUR_0, base_data_format, usage_bits, RD::TEXTURE_SAMPLES_1, internal_size, view_count, mipmaps_required); + create_texture(RB_SCOPE_BUFFERS, RB_TEX_BLUR_1, base_data_format, usage_bits, RD::TEXTURE_SAMPLES_1, Size2i(internal_size.x >> 1, internal_size.y >> 1), view_count, mipmaps_required - 1); + + // if !can_be_storage we need a half width version + if (!can_be_storage) { + create_texture(RB_SCOPE_BUFFERS, RB_TEX_HALF_BLUR, base_data_format, usage_bits, RD::TEXTURE_SAMPLES_1, Size2i(internal_size.x >> 1, internal_size.y), 1, mipmaps_required); + } + + // TODO redo this: + if (!can_be_storage) { + // create 4 weight textures, 2 full size, 2 half size + + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R16_SFLOAT; // We could probably use DATA_FORMAT_R8_SNORM if we don't pre-multiply by blur_size but that depends on whether we can remove DEPTH_GAP + tf.width = internal_size.x; + tf.height = internal_size.y; + tf.texture_type = RD::TEXTURE_TYPE_2D; + tf.array_layers = 1; // Our DOF effect handles one eye per turn + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + tf.mipmaps = 1; + for (uint32_t i = 0; i < 4; i++) { + // associated blur texture + RID texture; + if (i == 1) { + texture = get_texture_slice(RB_SCOPE_BUFFERS, RB_TEX_BLUR_0, 0, 0); + } else if (i == 2) { + texture = get_texture_slice(RB_SCOPE_BUFFERS, RB_TEX_BLUR_1, 0, 0); + } else if (i == 3) { + texture = get_texture_slice(RB_SCOPE_BUFFERS, RB_TEX_BLUR_0, 0, 1); + } + + // create weight texture + weight_buffers[i].weight = RD::get_singleton()->texture_create(tf, RD::TextureView()); + + // create frame buffer + Vector<RID> fb; + if (i != 0) { + fb.push_back(texture); + } + fb.push_back(weight_buffers[i].weight); + weight_buffers[i].fb = RD::get_singleton()->framebuffer_create(fb); + + if (i == 1) { + // next 2 are half size + tf.width = MAX(1u, tf.width >> 1); + tf.height = MAX(1u, tf.height >> 1); + } + } + } +} + +// Data buffers + +bool RenderSceneBuffersRD::has_custom_data(const StringName &p_name) { + return data_buffers.has(p_name); +} + +void RenderSceneBuffersRD::set_custom_data(const StringName &p_name, Ref<RenderBufferCustomDataRD> p_data) { + if (p_data.is_valid()) { + data_buffers[p_name] = p_data; + } else if (has_custom_data(p_name)) { + data_buffers.erase(p_name); + } +} + +Ref<RenderBufferCustomDataRD> RenderSceneBuffersRD::get_custom_data(const StringName &p_name) const { + ERR_FAIL_COND_V(!data_buffers.has(p_name), Ref<RenderBufferCustomDataRD>()); + + Ref<RenderBufferCustomDataRD> ret = data_buffers[p_name]; + + return ret; +} + +// Depth texture + +RID RenderSceneBuffersRD::get_depth_texture() { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + RID depth = texture_storage->render_target_get_override_depth(render_target); + if (depth.is_valid()) { + return depth; + } else { + return get_texture(RB_SCOPE_BUFFERS, RB_TEX_DEPTH); + } +} + +RID RenderSceneBuffersRD::get_depth_texture(const uint32_t p_layer) { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + RID depth_slice = texture_storage->render_target_get_override_depth_slice(render_target, p_layer); + if (depth_slice.is_valid()) { + return depth_slice; + } else { + return get_texture_slice(RB_SCOPE_BUFFERS, RB_TEX_DEPTH, p_layer, 0); + } +} + +// Velocity texture. + +void RenderSceneBuffersRD::ensure_velocity() { + if (!has_texture(RB_SCOPE_BUFFERS, RB_TEX_VELOCITY)) { + uint32_t usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + + if (msaa_3d != RS::VIEWPORT_MSAA_DISABLED) { + uint32_t msaa_usage_bits = usage_bits | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT; + usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; + + const RD::TextureSamples ts[RS::VIEWPORT_MSAA_MAX] = { + RD::TEXTURE_SAMPLES_1, + RD::TEXTURE_SAMPLES_2, + RD::TEXTURE_SAMPLES_4, + RD::TEXTURE_SAMPLES_8, + }; + + RD::TextureSamples texture_samples = ts[msaa_3d]; + + create_texture(RB_SCOPE_BUFFERS, RB_TEX_VELOCITY_MSAA, RD::DATA_FORMAT_R16G16_SFLOAT, msaa_usage_bits, texture_samples); + } + + create_texture(RB_SCOPE_BUFFERS, RB_TEX_VELOCITY, RD::DATA_FORMAT_R16G16_SFLOAT, usage_bits); + } +} + +bool RenderSceneBuffersRD::has_velocity_buffer(bool p_has_msaa) { + if (p_has_msaa) { + return has_texture(RB_SCOPE_BUFFERS, RB_TEX_VELOCITY_MSAA); + } else { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + RID velocity = texture_storage->render_target_get_override_velocity(render_target); + if (velocity.is_valid()) { + return true; + } else { + return has_texture(RB_SCOPE_BUFFERS, RB_TEX_VELOCITY); + } + } +} + +RID RenderSceneBuffersRD::get_velocity_buffer(bool p_get_msaa) { + if (p_get_msaa) { + if (!has_texture(RB_SCOPE_BUFFERS, RB_TEX_VELOCITY_MSAA)) { + return RID(); + } else { + return get_texture(RB_SCOPE_BUFFERS, RB_TEX_VELOCITY_MSAA); + } + } else { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + RID velocity = texture_storage->render_target_get_override_velocity(render_target); + if (velocity.is_valid()) { + return velocity; + } else if (!has_texture(RB_SCOPE_BUFFERS, RB_TEX_VELOCITY)) { + return RID(); + } else { + return get_texture(RB_SCOPE_BUFFERS, RB_TEX_VELOCITY); + } + } +} + +RID RenderSceneBuffersRD::get_velocity_buffer(bool p_get_msaa, uint32_t p_layer) { + if (p_get_msaa) { + return get_texture_slice(RB_SCOPE_BUFFERS, RB_TEX_VELOCITY_MSAA, p_layer, 0); + } else { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + RID velocity_slice = texture_storage->render_target_get_override_velocity_slice(render_target, p_layer); + if (velocity_slice.is_valid()) { + return velocity_slice; + } else { + return get_texture_slice(RB_SCOPE_BUFFERS, RB_TEX_VELOCITY, p_layer, 0); + } + } +} diff --git a/servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.h b/servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.h new file mode 100644 index 0000000000..6907f69b93 --- /dev/null +++ b/servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.h @@ -0,0 +1,226 @@ +/*************************************************************************/ +/* render_scene_buffers_rd.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef RENDER_SCENE_BUFFERS_RD_H +#define RENDER_SCENE_BUFFERS_RD_H + +#include "../effects/vrs.h" +#include "../framebuffer_cache_rd.h" +#include "core/templates/hash_map.h" +#include "render_buffer_custom_data_rd.h" +#include "servers/rendering/rendering_device.h" +#include "servers/rendering/rendering_method.h" +#include "servers/rendering/storage/render_scene_buffers.h" + +#define RB_SCOPE_BUFFERS SNAME("render_buffers") +#define RB_SCOPE_VRS SNAME("VRS") + +#define RB_TEXTURE SNAME("texture") +#define RB_TEX_COLOR SNAME("color") +#define RB_TEX_COLOR_MSAA SNAME("color_msaa") +#define RB_TEX_DEPTH SNAME("depth") +#define RB_TEX_DEPTH_MSAA SNAME("depth_msaa") +#define RB_TEX_VELOCITY SNAME("velocity") +#define RB_TEX_VELOCITY_MSAA SNAME("velocity_msaa") + +#define RB_TEX_BLUR_0 SNAME("blur_0") +#define RB_TEX_BLUR_1 SNAME("blur_1") +#define RB_TEX_HALF_BLUR SNAME("half_blur") // only for raster! + +#define RB_TEX_BACK_DEPTH SNAME("back_depth") + +class RenderSceneBuffersRD : public RenderSceneBuffers { + GDCLASS(RenderSceneBuffersRD, RenderSceneBuffers); + +private: + bool can_be_storage = true; + uint32_t max_cluster_elements = 512; + RD::DataFormat base_data_format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + RendererRD::VRS *vrs = nullptr; + uint64_t auto_exposure_version = 1; + + // Our render target represents our final destination that we display on screen. + RID render_target; + Size2i target_size = Size2i(0, 0); + uint32_t view_count = 1; + + // The internal size of the textures we render 3D to in case we render at a lower resolution and upscale + Size2i internal_size = Size2i(0, 0); + float fsr_sharpness = 0.2f; + + // Aliassing settings + RS::ViewportMSAA msaa_3d = RS::VIEWPORT_MSAA_DISABLED; + RS::ViewportScreenSpaceAA screen_space_aa = RS::VIEWPORT_SCREEN_SPACE_AA_DISABLED; + bool use_taa = false; + bool use_debanding = false; + + // Named Textures + + struct NTKey { + StringName context; + StringName buffer_name; + + bool operator==(const NTKey &p_val) const { + return (context == p_val.context) && (buffer_name == p_val.buffer_name); + } + + static uint32_t hash(const NTKey &p_val) { + // FIXME, properly hash two stringnames together + uint32_t h = p_val.context.hash(); + h = hash_murmur3_one_32(p_val.buffer_name.hash(), h); + return hash_fmix32(h); + } + + NTKey() {} + NTKey(const StringName p_context, const StringName p_texture_name) { + context = p_context; + buffer_name = p_texture_name; + } + }; + + struct NamedTexture { + // Cache the data used to create our texture + RD::TextureFormat format; + bool is_unique; // If marked as unique, we return it into our pool + + // Our texture objects, slices are lazy (i.e. only created when requested). + RID texture; + Vector<RID> slices; + Vector<Size2i> sizes; + }; + + mutable HashMap<NTKey, NamedTexture, NTKey> named_textures; + void update_sizes(NamedTexture &p_named_texture); + void free_named_texture(NamedTexture &p_named_texture); + + // Data buffers + mutable HashMap<StringName, Ref<RenderBufferCustomDataRD>> data_buffers; + +protected: + static void _bind_methods(); + +public: + RenderSceneBuffersRD(); + virtual ~RenderSceneBuffersRD(); + + // info from our renderer + void set_can_be_storage(const bool p_can_be_storage) { can_be_storage = p_can_be_storage; } + void set_max_cluster_elements(const uint32_t p_max_elements) { max_cluster_elements = p_max_elements; } + uint32_t get_max_cluster_elements() { return max_cluster_elements; } + void set_base_data_format(const RD::DataFormat p_base_data_format) { base_data_format = p_base_data_format; } + RD::DataFormat get_base_data_format() const { return base_data_format; } + void set_vrs(RendererRD::VRS *p_vrs) { vrs = p_vrs; } + + void cleanup(); + virtual void configure(RID p_render_target, const Size2i p_internal_size, const Size2i p_target_size, float p_fsr_sharpness, float p_texture_mipmap_bias, RS::ViewportMSAA p_msaa_3d, RenderingServer::ViewportScreenSpaceAA p_screen_space_aa, bool p_use_taa, bool p_use_debanding, uint32_t p_view_count) override; + virtual void set_fsr_sharpness(float p_fsr_sharpness) override; + virtual void set_texture_mipmap_bias(float p_texture_mipmap_bias) override; + virtual void set_use_debanding(bool p_use_debanding) override; + + // Named Textures + + bool has_texture(const StringName &p_context, const StringName &p_texture_name) const; + RID create_texture(const StringName &p_context, const StringName &p_texture_name, const RD::DataFormat p_data_format, const uint32_t p_usage_bits, const RD::TextureSamples p_texture_samples = RD::TEXTURE_SAMPLES_1, const Size2i p_size = Size2i(0, 0), const uint32_t p_layers = 0, const uint32_t p_mipmaps = 1, bool p_unique = true); + RID create_texture_from_format(const StringName &p_context, const StringName &p_texture_name, const RD::TextureFormat &p_texture_format, RD::TextureView p_view = RD::TextureView(), bool p_unique = true); + RID create_texture_view(const StringName &p_context, const StringName &p_texture_name, const StringName p_view_name, RD::TextureView p_view = RD::TextureView()); + RID get_texture(const StringName &p_context, const StringName &p_texture_name) const; + const RD::TextureFormat get_texture_format(const StringName &p_context, const StringName &p_texture_name) const; + RID get_texture_slice(const StringName &p_context, const StringName &p_texture_name, const uint32_t p_layer, const uint32_t p_mipmap); + Size2i get_texture_slice_size(const StringName &p_context, const StringName &p_texture_name, const uint32_t p_layer, const uint32_t p_mipmap); + + void clear_context(const StringName &p_context); + + // Allocate shared buffers + void allocate_blur_textures(); + + // Custom data + bool has_custom_data(const StringName &p_name); + void set_custom_data(const StringName &p_name, Ref<RenderBufferCustomDataRD> p_data); + Ref<RenderBufferCustomDataRD> get_custom_data(const StringName &p_name) const; + + // Getters + + _FORCE_INLINE_ RID get_render_target() const { return render_target; } + _FORCE_INLINE_ uint32_t get_view_count() const { return view_count; } + _FORCE_INLINE_ Size2i get_internal_size() const { return internal_size; } + _FORCE_INLINE_ Size2i get_target_size() const { return target_size; } + _FORCE_INLINE_ float get_fsr_sharpness() const { return fsr_sharpness; } + _FORCE_INLINE_ RS::ViewportMSAA get_msaa_3d() const { return msaa_3d; } + _FORCE_INLINE_ RS::ViewportScreenSpaceAA get_screen_space_aa() const { return screen_space_aa; } + _FORCE_INLINE_ bool get_use_taa() const { return use_taa; } + _FORCE_INLINE_ bool get_use_debanding() const { return use_debanding; } + + uint64_t get_auto_exposure_version() const { return auto_exposure_version; } + void set_auto_exposure_version(const uint64_t p_auto_exposure_version) { auto_exposure_version = p_auto_exposure_version; } + + // For our internal textures we provide some easy access methods. + + _FORCE_INLINE_ RID get_internal_texture() const { + return get_texture(RB_SCOPE_BUFFERS, RB_TEX_COLOR); + } + _FORCE_INLINE_ RID get_internal_texture(const uint32_t p_layer) { + return get_texture_slice(RB_SCOPE_BUFFERS, RB_TEX_COLOR, p_layer, 0); + } + + RID get_depth_texture(); + RID get_depth_texture(const uint32_t p_layer); + + // back buffer (color) + RID get_back_buffer_texture() const { return has_texture(RB_SCOPE_BUFFERS, RB_TEX_BLUR_0) ? get_texture(RB_SCOPE_BUFFERS, RB_TEX_BLUR_0) : RID(); } // We (re)use our blur texture here. + + // Velocity, currently only used by TAA (Clustered) but we'll be using this in other places soon too. + + void ensure_velocity(); + bool has_velocity_buffer(bool p_has_msaa); + RID get_velocity_buffer(bool p_get_msaa); + RID get_velocity_buffer(bool p_get_msaa, uint32_t p_layer); + + //////////////////////////////////////////////////////////////////////////////////////////////////////////// + // Everything after this needs to be re-evaluated, this is all old implementation + + struct WeightBuffers { + RID weight; + RID fb; // FB with both texture and weight writing into one level lower + }; + + // 2 full size, 2 half size + WeightBuffers weight_buffers[4]; // Only used in raster + + struct Luminance { + Vector<RID> reduce; + RID current; + + // used only on mobile renderer + Vector<RID> fb; + RID current_fb; + } luminance; +}; + +#endif // RENDER_SCENE_BUFFERS_RD_H diff --git a/servers/rendering/renderer_rd/storage_rd/render_scene_data_rd.cpp b/servers/rendering/renderer_rd/storage_rd/render_scene_data_rd.cpp new file mode 100644 index 0000000000..7dd790d1da --- /dev/null +++ b/servers/rendering/renderer_rd/storage_rd/render_scene_data_rd.cpp @@ -0,0 +1,252 @@ +/*************************************************************************/ +/* render_scene_data_rd.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "render_scene_data_rd.h" +#include "servers/rendering/renderer_rd/renderer_scene_render_rd.h" +#include "servers/rendering/renderer_rd/storage_rd/light_storage.h" +#include "servers/rendering/renderer_rd/storage_rd/texture_storage.h" +#include "servers/rendering/rendering_server_default.h" + +RID RenderSceneDataRD::create_uniform_buffer() { + return RD::get_singleton()->uniform_buffer_create(sizeof(UBODATA)); +} + +void RenderSceneDataRD::update_ubo(RID p_uniform_buffer, RS::ViewportDebugDraw p_debug_mode, RID p_env, RID p_reflection_probe_instance, RID p_camera_attributes, bool p_flip_y, bool p_pancake_shadows, const Size2i &p_screen_size, const Color &p_default_bg_color, float p_luminance_multiplier, bool p_opaque_render_buffers) { + RendererSceneRenderRD *render_scene_render = RendererSceneRenderRD::get_singleton(); + + UBODATA ubo_data; + memset(&ubo_data, 0, sizeof(UBODATA)); + + // just for easy access.. + UBO &ubo = ubo_data.ubo; + UBO &prev_ubo = ubo_data.prev_ubo; + + Projection correction; + correction.set_depth_correction(p_flip_y); + correction.add_jitter_offset(taa_jitter); + Projection projection = correction * cam_projection; + + //store camera into ubo + RendererRD::MaterialStorage::store_camera(projection, ubo.projection_matrix); + RendererRD::MaterialStorage::store_camera(projection.inverse(), ubo.inv_projection_matrix); + RendererRD::MaterialStorage::store_transform(cam_transform, ubo.inv_view_matrix); + RendererRD::MaterialStorage::store_transform(cam_transform.affine_inverse(), ubo.view_matrix); + +#ifdef REAL_T_IS_DOUBLE + RendererRD::MaterialStorage::split_double(-cam_transform.origin.x, &ubo.inv_view_matrix[12], &ubo.inv_view_matrix[3]); + RendererRD::MaterialStorage::split_double(-cam_transform.origin.y, &ubo.inv_view_matrix[13], &ubo.inv_view_matrix[7]); + RendererRD::MaterialStorage::split_double(-cam_transform.origin.z, &ubo.inv_view_matrix[14], &ubo.inv_view_matrix[11]); +#endif + + for (uint32_t v = 0; v < view_count; v++) { + projection = correction * view_projection[v]; + RendererRD::MaterialStorage::store_camera(projection, ubo.projection_matrix_view[v]); + RendererRD::MaterialStorage::store_camera(projection.inverse(), ubo.inv_projection_matrix_view[v]); + + ubo.eye_offset[v][0] = view_eye_offset[v].x; + ubo.eye_offset[v][1] = view_eye_offset[v].y; + ubo.eye_offset[v][2] = view_eye_offset[v].z; + ubo.eye_offset[v][3] = 0.0; + } + + ubo.taa_jitter[0] = taa_jitter.x; + ubo.taa_jitter[1] = taa_jitter.y; + + ubo.z_far = z_far; + ubo.z_near = z_near; + + ubo.pancake_shadows = p_pancake_shadows; + + RendererRD::MaterialStorage::store_soft_shadow_kernel(render_scene_render->directional_penumbra_shadow_kernel_get(), ubo.directional_penumbra_shadow_kernel); + RendererRD::MaterialStorage::store_soft_shadow_kernel(render_scene_render->directional_soft_shadow_kernel_get(), ubo.directional_soft_shadow_kernel); + RendererRD::MaterialStorage::store_soft_shadow_kernel(render_scene_render->penumbra_shadow_kernel_get(), ubo.penumbra_shadow_kernel); + RendererRD::MaterialStorage::store_soft_shadow_kernel(render_scene_render->soft_shadow_kernel_get(), ubo.soft_shadow_kernel); + + ubo.viewport_size[0] = p_screen_size.x; + ubo.viewport_size[1] = p_screen_size.y; + + Size2 screen_pixel_size = Vector2(1.0, 1.0) / Size2(p_screen_size); + ubo.screen_pixel_size[0] = screen_pixel_size.x; + ubo.screen_pixel_size[1] = screen_pixel_size.y; + + ubo.shadow_atlas_pixel_size[0] = shadow_atlas_pixel_size.x; + ubo.shadow_atlas_pixel_size[1] = shadow_atlas_pixel_size.y; + + ubo.directional_shadow_pixel_size[0] = directional_shadow_pixel_size.x; + ubo.directional_shadow_pixel_size[1] = directional_shadow_pixel_size.y; + + ubo.time = time; + + ubo.directional_light_count = directional_light_count; + ubo.dual_paraboloid_side = dual_paraboloid_side; + ubo.opaque_prepass_threshold = opaque_prepass_threshold; + ubo.material_uv2_mode = material_uv2_mode; + + ubo.fog_enabled = false; + + if (p_debug_mode == RS::VIEWPORT_DEBUG_DRAW_UNSHADED) { + ubo.use_ambient_light = true; + ubo.ambient_light_color_energy[0] = 1; + ubo.ambient_light_color_energy[1] = 1; + ubo.ambient_light_color_energy[2] = 1; + ubo.ambient_light_color_energy[3] = 1.0; + ubo.use_ambient_cubemap = false; + ubo.use_reflection_cubemap = false; + } else if (p_env.is_valid()) { + RS::EnvironmentBG env_bg = render_scene_render->environment_get_background(p_env); + RS::EnvironmentAmbientSource ambient_src = render_scene_render->environment_get_ambient_source(p_env); + + float bg_energy_multiplier = render_scene_render->environment_get_bg_energy_multiplier(p_env); + + ubo.ambient_light_color_energy[3] = bg_energy_multiplier; + + ubo.ambient_color_sky_mix = render_scene_render->environment_get_ambient_sky_contribution(p_env); + + //ambient + if (ambient_src == RS::ENV_AMBIENT_SOURCE_BG && (env_bg == RS::ENV_BG_CLEAR_COLOR || env_bg == RS::ENV_BG_COLOR)) { + Color color = env_bg == RS::ENV_BG_CLEAR_COLOR ? p_default_bg_color : render_scene_render->environment_get_bg_color(p_env); + color = color.srgb_to_linear(); + + ubo.ambient_light_color_energy[0] = color.r * bg_energy_multiplier; + ubo.ambient_light_color_energy[1] = color.g * bg_energy_multiplier; + ubo.ambient_light_color_energy[2] = color.b * bg_energy_multiplier; + ubo.use_ambient_light = true; + ubo.use_ambient_cubemap = false; + } else { + float energy = render_scene_render->environment_get_ambient_light_energy(p_env); + Color color = render_scene_render->environment_get_ambient_light(p_env); + color = color.srgb_to_linear(); + ubo.ambient_light_color_energy[0] = color.r * energy; + ubo.ambient_light_color_energy[1] = color.g * energy; + ubo.ambient_light_color_energy[2] = color.b * energy; + + Basis sky_transform = render_scene_render->environment_get_sky_orientation(p_env); + sky_transform = sky_transform.inverse() * cam_transform.basis; + RendererRD::MaterialStorage::store_transform_3x3(sky_transform, ubo.radiance_inverse_xform); + + ubo.use_ambient_cubemap = (ambient_src == RS::ENV_AMBIENT_SOURCE_BG && env_bg == RS::ENV_BG_SKY) || ambient_src == RS::ENV_AMBIENT_SOURCE_SKY; + ubo.use_ambient_light = ubo.use_ambient_cubemap || ambient_src == RS::ENV_AMBIENT_SOURCE_COLOR; + } + + //specular + RS::EnvironmentReflectionSource ref_src = render_scene_render->environment_get_reflection_source(p_env); + if ((ref_src == RS::ENV_REFLECTION_SOURCE_BG && env_bg == RS::ENV_BG_SKY) || ref_src == RS::ENV_REFLECTION_SOURCE_SKY) { + ubo.use_reflection_cubemap = true; + } else { + ubo.use_reflection_cubemap = false; + } + + ubo.fog_enabled = render_scene_render->environment_get_fog_enabled(p_env); + ubo.fog_density = render_scene_render->environment_get_fog_density(p_env); + ubo.fog_height = render_scene_render->environment_get_fog_height(p_env); + ubo.fog_height_density = render_scene_render->environment_get_fog_height_density(p_env); + ubo.fog_aerial_perspective = render_scene_render->environment_get_fog_aerial_perspective(p_env); + + Color fog_color = render_scene_render->environment_get_fog_light_color(p_env).srgb_to_linear(); + float fog_energy = render_scene_render->environment_get_fog_light_energy(p_env); + + ubo.fog_light_color[0] = fog_color.r * fog_energy; + ubo.fog_light_color[1] = fog_color.g * fog_energy; + ubo.fog_light_color[2] = fog_color.b * fog_energy; + + ubo.fog_sun_scatter = render_scene_render->environment_get_fog_sun_scatter(p_env); + } else { + if (p_reflection_probe_instance.is_valid() && RendererRD::LightStorage::get_singleton()->reflection_probe_is_interior(p_reflection_probe_instance)) { + ubo.use_ambient_light = false; + } else { + ubo.use_ambient_light = true; + Color clear_color = p_default_bg_color; + clear_color = clear_color.srgb_to_linear(); + ubo.ambient_light_color_energy[0] = clear_color.r; + ubo.ambient_light_color_energy[1] = clear_color.g; + ubo.ambient_light_color_energy[2] = clear_color.b; + ubo.ambient_light_color_energy[3] = 1.0; + } + + ubo.use_ambient_cubemap = false; + ubo.use_reflection_cubemap = false; + } + + if (p_camera_attributes.is_valid()) { + ubo.emissive_exposure_normalization = RSG::camera_attributes->camera_attributes_get_exposure_normalization_factor(p_camera_attributes); + ubo.IBL_exposure_normalization = 1.0; + if (p_env.is_valid()) { + RID sky_rid = render_scene_render->environment_get_sky(p_env); + if (sky_rid.is_valid()) { + float current_exposure = RSG::camera_attributes->camera_attributes_get_exposure_normalization_factor(p_camera_attributes) * render_scene_render->environment_get_bg_intensity(p_env) / p_luminance_multiplier; + ubo.IBL_exposure_normalization = current_exposure / MAX(0.001, render_scene_render->get_sky()->sky_get_baked_exposure(sky_rid)); + } + } + } else if (emissive_exposure_normalization > 0.0) { + // This branch is triggered when using render_material(). + // Emissive is set outside the function. + ubo.emissive_exposure_normalization = emissive_exposure_normalization; + // IBL isn't used don't set it. + } else { + ubo.emissive_exposure_normalization = 1.0; + ubo.IBL_exposure_normalization = 1.0; + } + + ubo.roughness_limiter_enabled = p_opaque_render_buffers && render_scene_render->screen_space_roughness_limiter_is_active(); + ubo.roughness_limiter_amount = render_scene_render->screen_space_roughness_limiter_get_amount(); + ubo.roughness_limiter_limit = render_scene_render->screen_space_roughness_limiter_get_limit(); + + if (calculate_motion_vectors) { + // Q : Should we make a complete copy or should we define a separate UBO with just the components we need? + memcpy(&prev_ubo, &ubo, sizeof(UBO)); + + Projection prev_correction; + prev_correction.set_depth_correction(true); + prev_correction.add_jitter_offset(prev_taa_jitter); + Projection prev_projection = prev_correction * prev_cam_projection; + + //store camera into ubo + RendererRD::MaterialStorage::store_camera(prev_projection, prev_ubo.projection_matrix); + RendererRD::MaterialStorage::store_camera(prev_projection.inverse(), prev_ubo.inv_projection_matrix); + RendererRD::MaterialStorage::store_transform(prev_cam_transform, prev_ubo.inv_view_matrix); + RendererRD::MaterialStorage::store_transform(prev_cam_transform.affine_inverse(), prev_ubo.view_matrix); + + for (uint32_t v = 0; v < view_count; v++) { + prev_projection = prev_correction * view_projection[v]; + RendererRD::MaterialStorage::store_camera(prev_projection, prev_ubo.projection_matrix_view[v]); + RendererRD::MaterialStorage::store_camera(prev_projection.inverse(), prev_ubo.inv_projection_matrix_view[v]); + } + prev_ubo.taa_jitter[0] = prev_taa_jitter.x; + prev_ubo.taa_jitter[1] = prev_taa_jitter.y; + prev_ubo.time -= time_step; + } + + uniform_buffer = p_uniform_buffer; + RD::get_singleton()->buffer_update(uniform_buffer, 0, sizeof(UBODATA), &ubo, RD::BARRIER_MASK_RASTER); +} + +RID RenderSceneDataRD::get_uniform_buffer() { + return uniform_buffer; +} diff --git a/servers/rendering/renderer_rd/storage_rd/render_scene_data_rd.h b/servers/rendering/renderer_rd/storage_rd/render_scene_data_rd.h new file mode 100644 index 0000000000..9c031acc1e --- /dev/null +++ b/servers/rendering/renderer_rd/storage_rd/render_scene_data_rd.h @@ -0,0 +1,156 @@ +/*************************************************************************/ +/* render_scene_data_rd.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef RENDER_SCENE_DATA_RD_H +#define RENDER_SCENE_DATA_RD_H + +#include "render_scene_buffers_rd.h" +#include "servers/rendering/renderer_scene_render.h" +#include "servers/rendering/rendering_device.h" + +// This is a container for data related to rendering a single frame of a viewport where we load this data into a UBO +// that can be used by the main scene shader but also by various effects. + +class RenderSceneDataRD { +public: + bool calculate_motion_vectors = false; + + Transform3D cam_transform; + Projection cam_projection; + Vector2 taa_jitter; + bool cam_orthogonal = false; + + // For stereo rendering + uint32_t view_count = 1; + Vector3 view_eye_offset[RendererSceneRender::MAX_RENDER_VIEWS]; + Projection view_projection[RendererSceneRender::MAX_RENDER_VIEWS]; + + Transform3D prev_cam_transform; + Projection prev_cam_projection; + Vector2 prev_taa_jitter; + Projection prev_view_projection[RendererSceneRender::MAX_RENDER_VIEWS]; + + float z_near = 0.0; + float z_far = 0.0; + + float lod_distance_multiplier = 0.0; + float screen_mesh_lod_threshold = 0.0; + + uint32_t directional_light_count = 0; + float dual_paraboloid_side = 0.0; + float opaque_prepass_threshold = 0.0; + bool material_uv2_mode = false; + float emissive_exposure_normalization = 0.0; + + Size2 shadow_atlas_pixel_size; + Size2 directional_shadow_pixel_size; + + float time; + float time_step; + + RID create_uniform_buffer(); + void update_ubo(RID p_uniform_buffer, RS::ViewportDebugDraw p_debug_mode, RID p_env, RID p_reflection_probe_instance, RID p_camera_attributes, bool p_flip_y, bool p_pancake_shadows, const Size2i &p_screen_size, const Color &p_default_bg_color, float p_luminance_multiplier, bool p_opaque_render_buffers); + RID get_uniform_buffer(); + +private: + RID uniform_buffer; // loaded into this uniform buffer (supplied externally) + + // This struct is loaded into Set 1 - Binding 0, populated at start of rendering a frame, must match with shader code + struct UBO { + float projection_matrix[16]; + float inv_projection_matrix[16]; + float inv_view_matrix[16]; + float view_matrix[16]; + + float projection_matrix_view[RendererSceneRender::MAX_RENDER_VIEWS][16]; + float inv_projection_matrix_view[RendererSceneRender::MAX_RENDER_VIEWS][16]; + float eye_offset[RendererSceneRender::MAX_RENDER_VIEWS][4]; + + float viewport_size[2]; + float screen_pixel_size[2]; + + float directional_penumbra_shadow_kernel[128]; //32 vec4s + float directional_soft_shadow_kernel[128]; + float penumbra_shadow_kernel[128]; + float soft_shadow_kernel[128]; + + float radiance_inverse_xform[12]; + + float ambient_light_color_energy[4]; + + float ambient_color_sky_mix; + uint32_t use_ambient_light; + uint32_t use_ambient_cubemap; + uint32_t use_reflection_cubemap; + + float shadow_atlas_pixel_size[2]; + float directional_shadow_pixel_size[2]; + + uint32_t directional_light_count; + float dual_paraboloid_side; + float z_far; + float z_near; + + uint32_t roughness_limiter_enabled; + float roughness_limiter_amount; + float roughness_limiter_limit; + float opaque_prepass_threshold; + + // Fog + uint32_t fog_enabled; + float fog_density; + float fog_height; + float fog_height_density; + + float fog_light_color[3]; + float fog_sun_scatter; + + float fog_aerial_perspective; + float time; + float reflection_multiplier; + uint32_t material_uv2_mode; + + float taa_jitter[2]; + float emissive_exposure_normalization; // Needed to normalize emissive when using physical units. + float IBL_exposure_normalization; // Adjusts for baked exposure. + + uint32_t pancake_shadows; + uint32_t pad1; + uint32_t pad2; + uint32_t pad3; + }; + + struct UBODATA { + UBO ubo; + UBO prev_ubo; + }; +}; + +#endif // RENDER_SCENE_DATA_RD_H diff --git a/servers/rendering/renderer_rd/storage_rd/texture_storage.cpp b/servers/rendering/renderer_rd/storage_rd/texture_storage.cpp new file mode 100644 index 0000000000..bc70c57b69 --- /dev/null +++ b/servers/rendering/renderer_rd/storage_rd/texture_storage.cpp @@ -0,0 +1,3262 @@ +/*************************************************************************/ +/* texture_storage.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "texture_storage.h" + +#include "../effects/copy_effects.h" +#include "../framebuffer_cache_rd.h" +#include "material_storage.h" +#include "servers/rendering/renderer_rd/renderer_scene_render_rd.h" + +using namespace RendererRD; + +/////////////////////////////////////////////////////////////////////////// +// TextureStorage::CanvasTexture + +void TextureStorage::CanvasTexture::clear_sets() { + if (cleared_cache) { + return; + } + for (int i = 1; i < RS::CANVAS_ITEM_TEXTURE_FILTER_MAX; i++) { + for (int j = 1; j < RS::CANVAS_ITEM_TEXTURE_REPEAT_MAX; j++) { + if (RD::get_singleton()->uniform_set_is_valid(uniform_sets[i][j])) { + RD::get_singleton()->free(uniform_sets[i][j]); + uniform_sets[i][j] = RID(); + } + } + } + cleared_cache = true; +} + +TextureStorage::CanvasTexture::~CanvasTexture() { + clear_sets(); +} + +/////////////////////////////////////////////////////////////////////////// +// TextureStorage::Texture + +void TextureStorage::Texture::cleanup() { + if (RD::get_singleton()->texture_is_valid(rd_texture_srgb)) { + //erase this first, as it's a dependency of the one below + RD::get_singleton()->free(rd_texture_srgb); + } + if (RD::get_singleton()->texture_is_valid(rd_texture)) { + RD::get_singleton()->free(rd_texture); + } + if (canvas_texture) { + memdelete(canvas_texture); + } +} + +/////////////////////////////////////////////////////////////////////////// +// TextureStorage + +TextureStorage *TextureStorage::singleton = nullptr; + +TextureStorage *TextureStorage::get_singleton() { + return singleton; +} + +TextureStorage::TextureStorage() { + singleton = this; + + { //create default textures + + RD::TextureFormat tformat; + tformat.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + tformat.width = 4; + tformat.height = 4; + tformat.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT; + tformat.texture_type = RD::TEXTURE_TYPE_2D; + + Vector<uint8_t> pv; + pv.resize(16 * 4); + for (int i = 0; i < 16; i++) { + // Opaque white. + pv.set(i * 4 + 0, 255); + pv.set(i * 4 + 1, 255); + pv.set(i * 4 + 2, 255); + pv.set(i * 4 + 3, 255); + } + + { + Vector<Vector<uint8_t>> vpv; + vpv.push_back(pv); + default_rd_textures[DEFAULT_RD_TEXTURE_WHITE] = RD::get_singleton()->texture_create(tformat, RD::TextureView(), vpv); + } + + for (int i = 0; i < 16; i++) { + // Opaque black. + pv.set(i * 4 + 0, 0); + pv.set(i * 4 + 1, 0); + pv.set(i * 4 + 2, 0); + pv.set(i * 4 + 3, 255); + } + + { + Vector<Vector<uint8_t>> vpv; + vpv.push_back(pv); + default_rd_textures[DEFAULT_RD_TEXTURE_BLACK] = RD::get_singleton()->texture_create(tformat, RD::TextureView(), vpv); + } + + for (int i = 0; i < 16; i++) { + // Transparent black. + pv.set(i * 4 + 0, 0); + pv.set(i * 4 + 1, 0); + pv.set(i * 4 + 2, 0); + pv.set(i * 4 + 3, 0); + } + + { + Vector<Vector<uint8_t>> vpv; + vpv.push_back(pv); + default_rd_textures[DEFAULT_RD_TEXTURE_TRANSPARENT] = RD::get_singleton()->texture_create(tformat, RD::TextureView(), vpv); + } + + for (int i = 0; i < 16; i++) { + // Opaque normal map "flat" color. + pv.set(i * 4 + 0, 128); + pv.set(i * 4 + 1, 128); + pv.set(i * 4 + 2, 255); + pv.set(i * 4 + 3, 255); + } + + { + Vector<Vector<uint8_t>> vpv; + vpv.push_back(pv); + default_rd_textures[DEFAULT_RD_TEXTURE_NORMAL] = RD::get_singleton()->texture_create(tformat, RD::TextureView(), vpv); + } + + for (int i = 0; i < 16; i++) { + // Opaque flowmap "flat" color. + pv.set(i * 4 + 0, 255); + pv.set(i * 4 + 1, 128); + pv.set(i * 4 + 2, 255); + pv.set(i * 4 + 3, 255); + } + + { + Vector<Vector<uint8_t>> vpv; + vpv.push_back(pv); + default_rd_textures[DEFAULT_RD_TEXTURE_ANISO] = RD::get_singleton()->texture_create(tformat, RD::TextureView(), vpv); + } + + { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_D16_UNORM; + tf.width = 4; + tf.height = 4; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT | RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT; + tf.texture_type = RD::TEXTURE_TYPE_2D; + + Vector<uint8_t> sv; + sv.resize(16 * 2); + uint16_t *ptr = (uint16_t *)sv.ptrw(); + for (int i = 0; i < 16; i++) { + ptr[i] = Math::make_half_float(1.0f); + } + + Vector<Vector<uint8_t>> vpv; + vpv.push_back(sv); + default_rd_textures[DEFAULT_RD_TEXTURE_DEPTH] = RD::get_singleton()->texture_create(tf, RD::TextureView(), vpv); + } + + for (int i = 0; i < 16; i++) { + pv.set(i * 4 + 0, 0); + pv.set(i * 4 + 1, 0); + pv.set(i * 4 + 2, 0); + pv.set(i * 4 + 3, 0); + } + + default_rd_textures[DEFAULT_RD_TEXTURE_MULTIMESH_BUFFER] = RD::get_singleton()->texture_buffer_create(16, RD::DATA_FORMAT_R8G8B8A8_UNORM, pv); + + for (int i = 0; i < 16; i++) { + pv.set(i * 4 + 0, 0); + pv.set(i * 4 + 1, 0); + pv.set(i * 4 + 2, 0); + pv.set(i * 4 + 3, 0); + } + + { + tformat.format = RD::DATA_FORMAT_R8G8B8A8_UINT; + Vector<Vector<uint8_t>> vpv; + vpv.push_back(pv); + default_rd_textures[DEFAULT_RD_TEXTURE_2D_UINT] = RD::get_singleton()->texture_create(tformat, RD::TextureView(), vpv); + } + } + + { //create default black cubemap array + + RD::TextureFormat tformat; + tformat.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + tformat.width = 4; + tformat.height = 4; + tformat.array_layers = 6; + tformat.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT; + tformat.texture_type = RD::TEXTURE_TYPE_CUBE_ARRAY; + + Vector<uint8_t> pv; + pv.resize(16 * 4); + for (int i = 0; i < 16; i++) { + pv.set(i * 4 + 0, 0); + pv.set(i * 4 + 1, 0); + pv.set(i * 4 + 2, 0); + pv.set(i * 4 + 3, 0); + } + + { + Vector<Vector<uint8_t>> vpv; + for (int i = 0; i < 6; i++) { + vpv.push_back(pv); + } + default_rd_textures[DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_BLACK] = RD::get_singleton()->texture_create(tformat, RD::TextureView(), vpv); + } + } + + { //create default white cubemap array + + RD::TextureFormat tformat; + tformat.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + tformat.width = 4; + tformat.height = 4; + tformat.array_layers = 6; + tformat.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT; + tformat.texture_type = RD::TEXTURE_TYPE_CUBE_ARRAY; + + Vector<uint8_t> pv; + pv.resize(16 * 4); + for (int i = 0; i < 16; i++) { + pv.set(i * 4 + 0, 255); + pv.set(i * 4 + 1, 255); + pv.set(i * 4 + 2, 255); + pv.set(i * 4 + 3, 255); + } + + { + Vector<Vector<uint8_t>> vpv; + for (int i = 0; i < 6; i++) { + vpv.push_back(pv); + } + default_rd_textures[DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_WHITE] = RD::get_singleton()->texture_create(tformat, RD::TextureView(), vpv); + } + } + + { //create default black cubemap + + RD::TextureFormat tformat; + tformat.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + tformat.width = 4; + tformat.height = 4; + tformat.array_layers = 6; + tformat.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT; + tformat.texture_type = RD::TEXTURE_TYPE_CUBE; + + Vector<uint8_t> pv; + pv.resize(16 * 4); + for (int i = 0; i < 16; i++) { + pv.set(i * 4 + 0, 0); + pv.set(i * 4 + 1, 0); + pv.set(i * 4 + 2, 0); + pv.set(i * 4 + 3, 0); + } + + { + Vector<Vector<uint8_t>> vpv; + for (int i = 0; i < 6; i++) { + vpv.push_back(pv); + } + default_rd_textures[DEFAULT_RD_TEXTURE_CUBEMAP_BLACK] = RD::get_singleton()->texture_create(tformat, RD::TextureView(), vpv); + } + } + + { //create default white cubemap + + RD::TextureFormat tformat; + tformat.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + tformat.width = 4; + tformat.height = 4; + tformat.array_layers = 6; + tformat.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT; + tformat.texture_type = RD::TEXTURE_TYPE_CUBE; + + Vector<uint8_t> pv; + pv.resize(16 * 4); + for (int i = 0; i < 16; i++) { + pv.set(i * 4 + 0, 255); + pv.set(i * 4 + 1, 255); + pv.set(i * 4 + 2, 255); + pv.set(i * 4 + 3, 255); + } + + { + Vector<Vector<uint8_t>> vpv; + for (int i = 0; i < 6; i++) { + vpv.push_back(pv); + } + default_rd_textures[DEFAULT_RD_TEXTURE_CUBEMAP_WHITE] = RD::get_singleton()->texture_create(tformat, RD::TextureView(), vpv); + } + } + + { //create default 3D + + RD::TextureFormat tformat; + tformat.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + tformat.width = 4; + tformat.height = 4; + tformat.depth = 4; + tformat.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT; + tformat.texture_type = RD::TEXTURE_TYPE_3D; + + Vector<uint8_t> pv; + pv.resize(64 * 4); + for (int i = 0; i < 64; i++) { + pv.set(i * 4 + 0, 0); + pv.set(i * 4 + 1, 0); + pv.set(i * 4 + 2, 0); + pv.set(i * 4 + 3, 0); + } + + { + Vector<Vector<uint8_t>> vpv; + vpv.push_back(pv); + default_rd_textures[DEFAULT_RD_TEXTURE_3D_BLACK] = RD::get_singleton()->texture_create(tformat, RD::TextureView(), vpv); + } + for (int i = 0; i < 64; i++) { + pv.set(i * 4 + 0, 255); + pv.set(i * 4 + 1, 255); + pv.set(i * 4 + 2, 255); + pv.set(i * 4 + 3, 255); + } + + { + Vector<Vector<uint8_t>> vpv; + vpv.push_back(pv); + default_rd_textures[DEFAULT_RD_TEXTURE_3D_WHITE] = RD::get_singleton()->texture_create(tformat, RD::TextureView(), vpv); + } + } + + { //create default array + + RD::TextureFormat tformat; + tformat.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + tformat.width = 4; + tformat.height = 4; + tformat.array_layers = 1; + tformat.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT; + tformat.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; + + Vector<uint8_t> pv; + pv.resize(16 * 4); + for (int i = 0; i < 16; i++) { + pv.set(i * 4 + 0, 255); + pv.set(i * 4 + 1, 255); + pv.set(i * 4 + 2, 255); + pv.set(i * 4 + 3, 255); + } + + { + Vector<Vector<uint8_t>> vpv; + vpv.push_back(pv); + default_rd_textures[DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE] = RD::get_singleton()->texture_create(tformat, RD::TextureView(), vpv); + } + } + + { // default atlas texture + RD::TextureFormat tformat; + tformat.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + tformat.width = 4; + tformat.height = 4; + tformat.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT; + tformat.texture_type = RD::TEXTURE_TYPE_2D; + + Vector<uint8_t> pv; + pv.resize(16 * 4); + for (int i = 0; i < 16; i++) { + pv.set(i * 4 + 0, 0); + pv.set(i * 4 + 1, 0); + pv.set(i * 4 + 2, 0); + pv.set(i * 4 + 3, 255); + } + + { + Vector<Vector<uint8_t>> vpv; + vpv.push_back(pv); + decal_atlas.texture = RD::get_singleton()->texture_create(tformat, RD::TextureView(), vpv); + decal_atlas.texture_srgb = decal_atlas.texture; + } + } + + { //create default VRS + + RD::TextureFormat tformat; + tformat.format = RD::DATA_FORMAT_R8_UINT; + tformat.width = 4; + tformat.height = 4; + tformat.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT; + if (RD::get_singleton()->has_feature(RD::SUPPORTS_ATTACHMENT_VRS)) { + tformat.usage_bits |= RD::TEXTURE_USAGE_VRS_ATTACHMENT_BIT; + } + tformat.texture_type = RD::TEXTURE_TYPE_2D; + + Vector<uint8_t> pv; + pv.resize(4 * 4); + for (int i = 0; i < 4 * 4; i++) { + pv.set(i, 0); + } + + { + Vector<Vector<uint8_t>> vpv; + vpv.push_back(pv); + default_rd_textures[DEFAULT_RD_TEXTURE_VRS] = RD::get_singleton()->texture_create(tformat, RD::TextureView(), vpv); + } + } + + { + Vector<String> sdf_modes; + sdf_modes.push_back("\n#define MODE_LOAD\n"); + sdf_modes.push_back("\n#define MODE_LOAD_SHRINK\n"); + sdf_modes.push_back("\n#define MODE_PROCESS\n"); + sdf_modes.push_back("\n#define MODE_PROCESS_OPTIMIZED\n"); + sdf_modes.push_back("\n#define MODE_STORE\n"); + sdf_modes.push_back("\n#define MODE_STORE_SHRINK\n"); + + rt_sdf.shader.initialize(sdf_modes); + + rt_sdf.shader_version = rt_sdf.shader.version_create(); + + for (int i = 0; i < RenderTargetSDF::SHADER_MAX; i++) { + rt_sdf.pipelines[i] = RD::get_singleton()->compute_pipeline_create(rt_sdf.shader.version_get_shader(rt_sdf.shader_version, i)); + } + } +} + +TextureStorage::~TextureStorage() { + rt_sdf.shader.version_free(rt_sdf.shader_version); + + free_decal_data(); + + if (decal_atlas.textures.size()) { + ERR_PRINT("Decal Atlas: " + itos(decal_atlas.textures.size()) + " textures were not removed from the atlas."); + } + + if (decal_atlas.texture.is_valid()) { + RD::get_singleton()->free(decal_atlas.texture); + } + + //def textures + for (int i = 0; i < DEFAULT_RD_TEXTURE_MAX; i++) { + if (default_rd_textures[i].is_valid()) { + RD::get_singleton()->free(default_rd_textures[i]); + } + } + + singleton = nullptr; +} + +bool TextureStorage::free(RID p_rid) { + if (owns_texture(p_rid)) { + texture_free(p_rid); + return true; + } else if (owns_canvas_texture(p_rid)) { + canvas_texture_free(p_rid); + return true; + } else if (owns_decal(p_rid)) { + decal_free(p_rid); + return true; + } else if (owns_decal_instance(p_rid)) { + decal_instance_free(p_rid); + return true; + } else if (owns_render_target(p_rid)) { + render_target_free(p_rid); + return true; + } + + return false; +} + +bool TextureStorage::can_create_resources_async() const { + return true; +} + +/* Canvas Texture API */ + +RID TextureStorage::canvas_texture_allocate() { + return canvas_texture_owner.allocate_rid(); +} + +void TextureStorage::canvas_texture_initialize(RID p_rid) { + canvas_texture_owner.initialize_rid(p_rid); +} + +void TextureStorage::canvas_texture_free(RID p_rid) { + canvas_texture_owner.free(p_rid); +} + +void TextureStorage::canvas_texture_set_channel(RID p_canvas_texture, RS::CanvasTextureChannel p_channel, RID p_texture) { + CanvasTexture *ct = canvas_texture_owner.get_or_null(p_canvas_texture); + ERR_FAIL_NULL(ct); + + switch (p_channel) { + case RS::CANVAS_TEXTURE_CHANNEL_DIFFUSE: { + ct->diffuse = p_texture; + } break; + case RS::CANVAS_TEXTURE_CHANNEL_NORMAL: { + ct->normal_map = p_texture; + } break; + case RS::CANVAS_TEXTURE_CHANNEL_SPECULAR: { + ct->specular = p_texture; + } break; + } + + ct->clear_sets(); +} + +void TextureStorage::canvas_texture_set_shading_parameters(RID p_canvas_texture, const Color &p_specular_color, float p_shininess) { + CanvasTexture *ct = canvas_texture_owner.get_or_null(p_canvas_texture); + ERR_FAIL_NULL(ct); + + ct->specular_color.r = p_specular_color.r; + ct->specular_color.g = p_specular_color.g; + ct->specular_color.b = p_specular_color.b; + ct->specular_color.a = p_shininess; + ct->clear_sets(); +} + +void TextureStorage::canvas_texture_set_texture_filter(RID p_canvas_texture, RS::CanvasItemTextureFilter p_filter) { + CanvasTexture *ct = canvas_texture_owner.get_or_null(p_canvas_texture); + ERR_FAIL_NULL(ct); + + ct->texture_filter = p_filter; + ct->clear_sets(); +} + +void TextureStorage::canvas_texture_set_texture_repeat(RID p_canvas_texture, RS::CanvasItemTextureRepeat p_repeat) { + CanvasTexture *ct = canvas_texture_owner.get_or_null(p_canvas_texture); + ERR_FAIL_NULL(ct); + ct->texture_repeat = p_repeat; + ct->clear_sets(); +} + +bool TextureStorage::canvas_texture_get_uniform_set(RID p_texture, RS::CanvasItemTextureFilter p_base_filter, RS::CanvasItemTextureRepeat p_base_repeat, RID p_base_shader, int p_base_set, RID &r_uniform_set, Size2i &r_size, Color &r_specular_shininess, bool &r_use_normal, bool &r_use_specular) { + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + + CanvasTexture *ct = nullptr; + Texture *t = get_texture(p_texture); + + // TODO once we have our texture storage split off we'll look into moving this code into canvas_texture + + if (t) { + //regular texture + if (!t->canvas_texture) { + t->canvas_texture = memnew(CanvasTexture); + t->canvas_texture->diffuse = p_texture; + } + + ct = t->canvas_texture; + } else { + ct = canvas_texture_owner.get_or_null(p_texture); + } + + if (!ct) { + return false; //invalid texture RID + } + + RS::CanvasItemTextureFilter filter = ct->texture_filter != RS::CANVAS_ITEM_TEXTURE_FILTER_DEFAULT ? ct->texture_filter : p_base_filter; + ERR_FAIL_COND_V(filter == RS::CANVAS_ITEM_TEXTURE_FILTER_DEFAULT, false); + + RS::CanvasItemTextureRepeat repeat = ct->texture_repeat != RS::CANVAS_ITEM_TEXTURE_REPEAT_DEFAULT ? ct->texture_repeat : p_base_repeat; + ERR_FAIL_COND_V(repeat == RS::CANVAS_ITEM_TEXTURE_REPEAT_DEFAULT, false); + + RID uniform_set = ct->uniform_sets[filter][repeat]; + if (!RD::get_singleton()->uniform_set_is_valid(uniform_set)) { + //create and update + Vector<RD::Uniform> uniforms; + { //diffuse + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 0; + + t = get_texture(ct->diffuse); + if (!t) { + u.append_id(texture_rd_get_default(DEFAULT_RD_TEXTURE_WHITE)); + ct->size_cache = Size2i(1, 1); + } else { + u.append_id(t->rd_texture); + ct->size_cache = Size2i(t->width_2d, t->height_2d); + } + uniforms.push_back(u); + } + { //normal + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 1; + + t = get_texture(ct->normal_map); + if (!t) { + u.append_id(texture_rd_get_default(DEFAULT_RD_TEXTURE_NORMAL)); + ct->use_normal_cache = false; + } else { + u.append_id(t->rd_texture); + ct->use_normal_cache = true; + } + uniforms.push_back(u); + } + { //specular + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 2; + + t = get_texture(ct->specular); + if (!t) { + u.append_id(texture_rd_get_default(DEFAULT_RD_TEXTURE_WHITE)); + ct->use_specular_cache = false; + } else { + u.append_id(t->rd_texture); + ct->use_specular_cache = true; + } + uniforms.push_back(u); + } + { //sampler + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 3; + u.append_id(material_storage->sampler_rd_get_default(filter, repeat)); + uniforms.push_back(u); + } + + uniform_set = RD::get_singleton()->uniform_set_create(uniforms, p_base_shader, p_base_set); + ct->uniform_sets[filter][repeat] = uniform_set; + ct->cleared_cache = false; + } + + r_uniform_set = uniform_set; + r_size = ct->size_cache; + r_specular_shininess = ct->specular_color; + r_use_normal = ct->use_normal_cache; + r_use_specular = ct->use_specular_cache; + + return true; +} + +/* Texture API */ + +RID TextureStorage::texture_allocate() { + return texture_owner.allocate_rid(); +} + +void TextureStorage::texture_free(RID p_texture) { + Texture *t = texture_owner.get_or_null(p_texture); + ERR_FAIL_COND(!t); + ERR_FAIL_COND(t->is_render_target); + + t->cleanup(); + + if (t->is_proxy && t->proxy_to.is_valid()) { + Texture *proxy_to = texture_owner.get_or_null(t->proxy_to); + if (proxy_to) { + proxy_to->proxies.erase(p_texture); + } + } + + decal_atlas_remove_texture(p_texture); + + for (int i = 0; i < t->proxies.size(); i++) { + Texture *p = texture_owner.get_or_null(t->proxies[i]); + ERR_CONTINUE(!p); + p->proxy_to = RID(); + p->rd_texture = RID(); + p->rd_texture_srgb = RID(); + } + + texture_owner.free(p_texture); +} + +void TextureStorage::texture_2d_initialize(RID p_texture, const Ref<Image> &p_image) { + ERR_FAIL_COND(p_image.is_null()); + + TextureToRDFormat ret_format; + Ref<Image> image = _validate_texture_format(p_image, ret_format); + + Texture texture; + + texture.type = TextureStorage::TYPE_2D; + + texture.width = p_image->get_width(); + texture.height = p_image->get_height(); + texture.layers = 1; + texture.mipmaps = p_image->get_mipmap_count() + 1; + texture.depth = 1; + texture.format = p_image->get_format(); + texture.validated_format = image->get_format(); + + texture.rd_type = RD::TEXTURE_TYPE_2D; + texture.rd_format = ret_format.format; + texture.rd_format_srgb = ret_format.format_srgb; + + RD::TextureFormat rd_format; + RD::TextureView rd_view; + { //attempt register + rd_format.format = texture.rd_format; + rd_format.width = texture.width; + rd_format.height = texture.height; + rd_format.depth = 1; + rd_format.array_layers = 1; + rd_format.mipmaps = texture.mipmaps; + rd_format.texture_type = texture.rd_type; + rd_format.samples = RD::TEXTURE_SAMPLES_1; + rd_format.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT; + if (texture.rd_format_srgb != RD::DATA_FORMAT_MAX) { + rd_format.shareable_formats.push_back(texture.rd_format); + rd_format.shareable_formats.push_back(texture.rd_format_srgb); + } + } + { + rd_view.swizzle_r = ret_format.swizzle_r; + rd_view.swizzle_g = ret_format.swizzle_g; + rd_view.swizzle_b = ret_format.swizzle_b; + rd_view.swizzle_a = ret_format.swizzle_a; + } + Vector<uint8_t> data = image->get_data(); //use image data + Vector<Vector<uint8_t>> data_slices; + data_slices.push_back(data); + texture.rd_texture = RD::get_singleton()->texture_create(rd_format, rd_view, data_slices); + ERR_FAIL_COND(texture.rd_texture.is_null()); + if (texture.rd_format_srgb != RD::DATA_FORMAT_MAX) { + rd_view.format_override = texture.rd_format_srgb; + texture.rd_texture_srgb = RD::get_singleton()->texture_create_shared(rd_view, texture.rd_texture); + if (texture.rd_texture_srgb.is_null()) { + RD::get_singleton()->free(texture.rd_texture); + ERR_FAIL_COND(texture.rd_texture_srgb.is_null()); + } + } + + //used for 2D, overridable + texture.width_2d = texture.width; + texture.height_2d = texture.height; + texture.is_render_target = false; + texture.rd_view = rd_view; + texture.is_proxy = false; + + texture_owner.initialize_rid(p_texture, texture); +} + +void TextureStorage::texture_2d_layered_initialize(RID p_texture, const Vector<Ref<Image>> &p_layers, RS::TextureLayeredType p_layered_type) { + ERR_FAIL_COND(p_layers.size() == 0); + + ERR_FAIL_COND(p_layered_type == RS::TEXTURE_LAYERED_CUBEMAP && p_layers.size() != 6); + ERR_FAIL_COND(p_layered_type == RS::TEXTURE_LAYERED_CUBEMAP_ARRAY && (p_layers.size() < 6 || (p_layers.size() % 6) != 0)); + + TextureToRDFormat ret_format; + Vector<Ref<Image>> images; + { + int valid_width = 0; + int valid_height = 0; + bool valid_mipmaps = false; + Image::Format valid_format = Image::FORMAT_MAX; + + for (int i = 0; i < p_layers.size(); i++) { + ERR_FAIL_COND(p_layers[i]->is_empty()); + + if (i == 0) { + valid_width = p_layers[i]->get_width(); + valid_height = p_layers[i]->get_height(); + valid_format = p_layers[i]->get_format(); + valid_mipmaps = p_layers[i]->has_mipmaps(); + } else { + ERR_FAIL_COND(p_layers[i]->get_width() != valid_width); + ERR_FAIL_COND(p_layers[i]->get_height() != valid_height); + ERR_FAIL_COND(p_layers[i]->get_format() != valid_format); + ERR_FAIL_COND(p_layers[i]->has_mipmaps() != valid_mipmaps); + } + + images.push_back(_validate_texture_format(p_layers[i], ret_format)); + } + } + + Texture texture; + + texture.type = TextureStorage::TYPE_LAYERED; + texture.layered_type = p_layered_type; + + texture.width = p_layers[0]->get_width(); + texture.height = p_layers[0]->get_height(); + texture.layers = p_layers.size(); + texture.mipmaps = p_layers[0]->get_mipmap_count() + 1; + texture.depth = 1; + texture.format = p_layers[0]->get_format(); + texture.validated_format = images[0]->get_format(); + + switch (p_layered_type) { + case RS::TEXTURE_LAYERED_2D_ARRAY: { + texture.rd_type = RD::TEXTURE_TYPE_2D_ARRAY; + } break; + case RS::TEXTURE_LAYERED_CUBEMAP: { + texture.rd_type = RD::TEXTURE_TYPE_CUBE; + } break; + case RS::TEXTURE_LAYERED_CUBEMAP_ARRAY: { + texture.rd_type = RD::TEXTURE_TYPE_CUBE_ARRAY; + } break; + default: + ERR_FAIL(); // Shouldn't happen, silence warnings. + } + + texture.rd_format = ret_format.format; + texture.rd_format_srgb = ret_format.format_srgb; + + RD::TextureFormat rd_format; + RD::TextureView rd_view; + { //attempt register + rd_format.format = texture.rd_format; + rd_format.width = texture.width; + rd_format.height = texture.height; + rd_format.depth = 1; + rd_format.array_layers = texture.layers; + rd_format.mipmaps = texture.mipmaps; + rd_format.texture_type = texture.rd_type; + rd_format.samples = RD::TEXTURE_SAMPLES_1; + rd_format.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT; + if (texture.rd_format_srgb != RD::DATA_FORMAT_MAX) { + rd_format.shareable_formats.push_back(texture.rd_format); + rd_format.shareable_formats.push_back(texture.rd_format_srgb); + } + } + { + rd_view.swizzle_r = ret_format.swizzle_r; + rd_view.swizzle_g = ret_format.swizzle_g; + rd_view.swizzle_b = ret_format.swizzle_b; + rd_view.swizzle_a = ret_format.swizzle_a; + } + Vector<Vector<uint8_t>> data_slices; + for (int i = 0; i < images.size(); i++) { + Vector<uint8_t> data = images[i]->get_data(); //use image data + data_slices.push_back(data); + } + texture.rd_texture = RD::get_singleton()->texture_create(rd_format, rd_view, data_slices); + ERR_FAIL_COND(texture.rd_texture.is_null()); + if (texture.rd_format_srgb != RD::DATA_FORMAT_MAX) { + rd_view.format_override = texture.rd_format_srgb; + texture.rd_texture_srgb = RD::get_singleton()->texture_create_shared(rd_view, texture.rd_texture); + if (texture.rd_texture_srgb.is_null()) { + RD::get_singleton()->free(texture.rd_texture); + ERR_FAIL_COND(texture.rd_texture_srgb.is_null()); + } + } + + //used for 2D, overridable + texture.width_2d = texture.width; + texture.height_2d = texture.height; + texture.is_render_target = false; + texture.rd_view = rd_view; + texture.is_proxy = false; + + texture_owner.initialize_rid(p_texture, texture); +} + +void TextureStorage::texture_3d_initialize(RID p_texture, Image::Format p_format, int p_width, int p_height, int p_depth, bool p_mipmaps, const Vector<Ref<Image>> &p_data) { + ERR_FAIL_COND(p_data.size() == 0); + + Image::Image3DValidateError verr = Image::validate_3d_image(p_format, p_width, p_height, p_depth, p_mipmaps, p_data); + if (verr != Image::VALIDATE_3D_OK) { + ERR_FAIL_MSG(Image::get_3d_image_validation_error_text(verr)); + } + + TextureToRDFormat ret_format; + Image::Format validated_format = Image::FORMAT_MAX; + Vector<uint8_t> all_data; + uint32_t mipmap_count = 0; + Vector<Texture::BufferSlice3D> slices; + { + Vector<Ref<Image>> images; + uint32_t all_data_size = 0; + images.resize(p_data.size()); + for (int i = 0; i < p_data.size(); i++) { + TextureToRDFormat f; + images.write[i] = _validate_texture_format(p_data[i], f); + if (i == 0) { + ret_format = f; + validated_format = images[0]->get_format(); + } + + all_data_size += images[i]->get_data().size(); + } + + all_data.resize(all_data_size); //consolidate all data here + uint32_t offset = 0; + Size2i prev_size; + for (int i = 0; i < p_data.size(); i++) { + uint32_t s = images[i]->get_data().size(); + + memcpy(&all_data.write[offset], images[i]->get_data().ptr(), s); + { + Texture::BufferSlice3D slice; + slice.size.width = images[i]->get_width(); + slice.size.height = images[i]->get_height(); + slice.offset = offset; + slice.buffer_size = s; + slices.push_back(slice); + } + offset += s; + + Size2i img_size(images[i]->get_width(), images[i]->get_height()); + if (img_size != prev_size) { + mipmap_count++; + } + prev_size = img_size; + } + } + + Texture texture; + + texture.type = TextureStorage::TYPE_3D; + texture.width = p_width; + texture.height = p_height; + texture.depth = p_depth; + texture.mipmaps = mipmap_count; + texture.format = p_data[0]->get_format(); + texture.validated_format = validated_format; + + texture.buffer_size_3d = all_data.size(); + texture.buffer_slices_3d = slices; + + texture.rd_type = RD::TEXTURE_TYPE_3D; + texture.rd_format = ret_format.format; + texture.rd_format_srgb = ret_format.format_srgb; + + RD::TextureFormat rd_format; + RD::TextureView rd_view; + { //attempt register + rd_format.format = texture.rd_format; + rd_format.width = texture.width; + rd_format.height = texture.height; + rd_format.depth = texture.depth; + rd_format.array_layers = 1; + rd_format.mipmaps = texture.mipmaps; + rd_format.texture_type = texture.rd_type; + rd_format.samples = RD::TEXTURE_SAMPLES_1; + rd_format.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT; + if (texture.rd_format_srgb != RD::DATA_FORMAT_MAX) { + rd_format.shareable_formats.push_back(texture.rd_format); + rd_format.shareable_formats.push_back(texture.rd_format_srgb); + } + } + { + rd_view.swizzle_r = ret_format.swizzle_r; + rd_view.swizzle_g = ret_format.swizzle_g; + rd_view.swizzle_b = ret_format.swizzle_b; + rd_view.swizzle_a = ret_format.swizzle_a; + } + Vector<Vector<uint8_t>> data_slices; + data_slices.push_back(all_data); //one slice + + texture.rd_texture = RD::get_singleton()->texture_create(rd_format, rd_view, data_slices); + ERR_FAIL_COND(texture.rd_texture.is_null()); + if (texture.rd_format_srgb != RD::DATA_FORMAT_MAX) { + rd_view.format_override = texture.rd_format_srgb; + texture.rd_texture_srgb = RD::get_singleton()->texture_create_shared(rd_view, texture.rd_texture); + if (texture.rd_texture_srgb.is_null()) { + RD::get_singleton()->free(texture.rd_texture); + ERR_FAIL_COND(texture.rd_texture_srgb.is_null()); + } + } + + //used for 2D, overridable + texture.width_2d = texture.width; + texture.height_2d = texture.height; + texture.is_render_target = false; + texture.rd_view = rd_view; + texture.is_proxy = false; + + texture_owner.initialize_rid(p_texture, texture); +} + +void TextureStorage::texture_proxy_initialize(RID p_texture, RID p_base) { + Texture *tex = texture_owner.get_or_null(p_base); + ERR_FAIL_COND(!tex); + Texture proxy_tex = *tex; + + proxy_tex.rd_view.format_override = tex->rd_format; + proxy_tex.rd_texture = RD::get_singleton()->texture_create_shared(proxy_tex.rd_view, tex->rd_texture); + if (proxy_tex.rd_texture_srgb.is_valid()) { + proxy_tex.rd_view.format_override = tex->rd_format_srgb; + proxy_tex.rd_texture_srgb = RD::get_singleton()->texture_create_shared(proxy_tex.rd_view, tex->rd_texture); + } + proxy_tex.proxy_to = p_base; + proxy_tex.is_render_target = false; + proxy_tex.is_proxy = true; + proxy_tex.proxies.clear(); + + texture_owner.initialize_rid(p_texture, proxy_tex); + + tex->proxies.push_back(p_texture); +} + +void TextureStorage::_texture_2d_update(RID p_texture, const Ref<Image> &p_image, int p_layer, bool p_immediate) { + ERR_FAIL_COND(p_image.is_null() || p_image->is_empty()); + + Texture *tex = texture_owner.get_or_null(p_texture); + ERR_FAIL_COND(!tex); + ERR_FAIL_COND(tex->is_render_target); + ERR_FAIL_COND(p_image->get_width() != tex->width || p_image->get_height() != tex->height); + ERR_FAIL_COND(p_image->get_format() != tex->format); + + if (tex->type == TextureStorage::TYPE_LAYERED) { + ERR_FAIL_INDEX(p_layer, tex->layers); + } + +#ifdef TOOLS_ENABLED + tex->image_cache_2d.unref(); +#endif + TextureToRDFormat f; + Ref<Image> validated = _validate_texture_format(p_image, f); + + RD::get_singleton()->texture_update(tex->rd_texture, p_layer, validated->get_data()); +} + +void TextureStorage::texture_2d_update(RID p_texture, const Ref<Image> &p_image, int p_layer) { + _texture_2d_update(p_texture, p_image, p_layer, false); +} + +void TextureStorage::texture_3d_update(RID p_texture, const Vector<Ref<Image>> &p_data) { + Texture *tex = texture_owner.get_or_null(p_texture); + ERR_FAIL_COND(!tex); + ERR_FAIL_COND(tex->type != TextureStorage::TYPE_3D); + + Image::Image3DValidateError verr = Image::validate_3d_image(tex->format, tex->width, tex->height, tex->depth, tex->mipmaps > 1, p_data); + if (verr != Image::VALIDATE_3D_OK) { + ERR_FAIL_MSG(Image::get_3d_image_validation_error_text(verr)); + } + + Vector<uint8_t> all_data; + { + Vector<Ref<Image>> images; + uint32_t all_data_size = 0; + images.resize(p_data.size()); + for (int i = 0; i < p_data.size(); i++) { + Ref<Image> image = p_data[i]; + if (image->get_format() != tex->validated_format) { + image = image->duplicate(); + image->convert(tex->validated_format); + } + all_data_size += images[i]->get_data().size(); + images.push_back(image); + } + + all_data.resize(all_data_size); //consolidate all data here + uint32_t offset = 0; + + for (int i = 0; i < p_data.size(); i++) { + uint32_t s = images[i]->get_data().size(); + memcpy(&all_data.write[offset], images[i]->get_data().ptr(), s); + offset += s; + } + } + + RD::get_singleton()->texture_update(tex->rd_texture, 0, all_data); +} + +void TextureStorage::texture_proxy_update(RID p_texture, RID p_proxy_to) { + Texture *tex = texture_owner.get_or_null(p_texture); + ERR_FAIL_COND(!tex); + ERR_FAIL_COND(!tex->is_proxy); + Texture *proxy_to = texture_owner.get_or_null(p_proxy_to); + ERR_FAIL_COND(!proxy_to); + ERR_FAIL_COND(proxy_to->is_proxy); + + if (tex->proxy_to.is_valid()) { + //unlink proxy + if (RD::get_singleton()->texture_is_valid(tex->rd_texture)) { + RD::get_singleton()->free(tex->rd_texture); + tex->rd_texture = RID(); + } + if (RD::get_singleton()->texture_is_valid(tex->rd_texture_srgb)) { + RD::get_singleton()->free(tex->rd_texture_srgb); + tex->rd_texture_srgb = RID(); + } + Texture *prev_tex = texture_owner.get_or_null(tex->proxy_to); + ERR_FAIL_COND(!prev_tex); + prev_tex->proxies.erase(p_texture); + } + + *tex = *proxy_to; + + tex->proxy_to = p_proxy_to; + tex->is_render_target = false; + tex->is_proxy = true; + tex->proxies.clear(); + proxy_to->proxies.push_back(p_texture); + + tex->rd_view.format_override = tex->rd_format; + tex->rd_texture = RD::get_singleton()->texture_create_shared(tex->rd_view, proxy_to->rd_texture); + if (tex->rd_texture_srgb.is_valid()) { + tex->rd_view.format_override = tex->rd_format_srgb; + tex->rd_texture_srgb = RD::get_singleton()->texture_create_shared(tex->rd_view, proxy_to->rd_texture); + } +} + +//these two APIs can be used together or in combination with the others. +void TextureStorage::texture_2d_placeholder_initialize(RID p_texture) { + //this could be better optimized to reuse an existing image , done this way + //for now to get it working + Ref<Image> image = Image::create_empty(4, 4, false, Image::FORMAT_RGBA8); + image->fill(Color(1, 0, 1, 1)); + + texture_2d_initialize(p_texture, image); +} + +void TextureStorage::texture_2d_layered_placeholder_initialize(RID p_texture, RS::TextureLayeredType p_layered_type) { + //this could be better optimized to reuse an existing image , done this way + //for now to get it working + Ref<Image> image = Image::create_empty(4, 4, false, Image::FORMAT_RGBA8); + image->fill(Color(1, 0, 1, 1)); + + Vector<Ref<Image>> images; + if (p_layered_type == RS::TEXTURE_LAYERED_2D_ARRAY) { + images.push_back(image); + } else { + //cube + for (int i = 0; i < 6; i++) { + images.push_back(image); + } + } + + texture_2d_layered_initialize(p_texture, images, p_layered_type); +} + +void TextureStorage::texture_3d_placeholder_initialize(RID p_texture) { + //this could be better optimized to reuse an existing image , done this way + //for now to get it working + Ref<Image> image = Image::create_empty(4, 4, false, Image::FORMAT_RGBA8); + image->fill(Color(1, 0, 1, 1)); + + Vector<Ref<Image>> images; + //cube + for (int i = 0; i < 4; i++) { + images.push_back(image); + } + + texture_3d_initialize(p_texture, Image::FORMAT_RGBA8, 4, 4, 4, false, images); +} + +Ref<Image> TextureStorage::texture_2d_get(RID p_texture) const { + Texture *tex = texture_owner.get_or_null(p_texture); + ERR_FAIL_COND_V(!tex, Ref<Image>()); + +#ifdef TOOLS_ENABLED + if (tex->image_cache_2d.is_valid() && !tex->is_render_target) { + return tex->image_cache_2d; + } +#endif + Vector<uint8_t> data = RD::get_singleton()->texture_get_data(tex->rd_texture, 0); + ERR_FAIL_COND_V(data.size() == 0, Ref<Image>()); + Ref<Image> image = Image::create_from_data(tex->width, tex->height, tex->mipmaps > 1, tex->validated_format, data); + ERR_FAIL_COND_V(image->is_empty(), Ref<Image>()); + if (tex->format != tex->validated_format) { + image->convert(tex->format); + } + +#ifdef TOOLS_ENABLED + if (Engine::get_singleton()->is_editor_hint() && !tex->is_render_target) { + tex->image_cache_2d = image; + } +#endif + + return image; +} + +Ref<Image> TextureStorage::texture_2d_layer_get(RID p_texture, int p_layer) const { + Texture *tex = texture_owner.get_or_null(p_texture); + ERR_FAIL_COND_V(!tex, Ref<Image>()); + + Vector<uint8_t> data = RD::get_singleton()->texture_get_data(tex->rd_texture, p_layer); + ERR_FAIL_COND_V(data.size() == 0, Ref<Image>()); + Ref<Image> image = Image::create_from_data(tex->width, tex->height, tex->mipmaps > 1, tex->validated_format, data); + ERR_FAIL_COND_V(image->is_empty(), Ref<Image>()); + if (tex->format != tex->validated_format) { + image->convert(tex->format); + } + + return image; +} + +Vector<Ref<Image>> TextureStorage::texture_3d_get(RID p_texture) const { + Texture *tex = texture_owner.get_or_null(p_texture); + ERR_FAIL_COND_V(!tex, Vector<Ref<Image>>()); + ERR_FAIL_COND_V(tex->type != TextureStorage::TYPE_3D, Vector<Ref<Image>>()); + + Vector<uint8_t> all_data = RD::get_singleton()->texture_get_data(tex->rd_texture, 0); + + ERR_FAIL_COND_V(all_data.size() != (int)tex->buffer_size_3d, Vector<Ref<Image>>()); + + Vector<Ref<Image>> ret; + + for (int i = 0; i < tex->buffer_slices_3d.size(); i++) { + const Texture::BufferSlice3D &bs = tex->buffer_slices_3d[i]; + ERR_FAIL_COND_V(bs.offset >= (uint32_t)all_data.size(), Vector<Ref<Image>>()); + ERR_FAIL_COND_V(bs.offset + bs.buffer_size > (uint32_t)all_data.size(), Vector<Ref<Image>>()); + Vector<uint8_t> sub_region = all_data.slice(bs.offset, bs.offset + bs.buffer_size); + + Ref<Image> img = Image::create_from_data(bs.size.width, bs.size.height, false, tex->validated_format, sub_region); + ERR_FAIL_COND_V(img->is_empty(), Vector<Ref<Image>>()); + if (tex->format != tex->validated_format) { + img->convert(tex->format); + } + + ret.push_back(img); + } + + return ret; +} + +void TextureStorage::texture_replace(RID p_texture, RID p_by_texture) { + Texture *tex = texture_owner.get_or_null(p_texture); + ERR_FAIL_COND(!tex); + ERR_FAIL_COND(tex->proxy_to.is_valid()); //can't replace proxy + Texture *by_tex = texture_owner.get_or_null(p_by_texture); + ERR_FAIL_COND(!by_tex); + ERR_FAIL_COND(by_tex->proxy_to.is_valid()); //can't replace proxy + + if (tex == by_tex) { + return; + } + + if (tex->rd_texture_srgb.is_valid()) { + RD::get_singleton()->free(tex->rd_texture_srgb); + } + RD::get_singleton()->free(tex->rd_texture); + + if (tex->canvas_texture) { + memdelete(tex->canvas_texture); + tex->canvas_texture = nullptr; + } + + Vector<RID> proxies_to_update = tex->proxies; + Vector<RID> proxies_to_redirect = by_tex->proxies; + + *tex = *by_tex; + + tex->proxies = proxies_to_update; //restore proxies, so they can be updated + + if (tex->canvas_texture) { + tex->canvas_texture->diffuse = p_texture; //update + } + + for (int i = 0; i < proxies_to_update.size(); i++) { + texture_proxy_update(proxies_to_update[i], p_texture); + } + for (int i = 0; i < proxies_to_redirect.size(); i++) { + texture_proxy_update(proxies_to_redirect[i], p_texture); + } + //delete last, so proxies can be updated + texture_owner.free(p_by_texture); + + decal_atlas_mark_dirty_on_texture(p_texture); +} + +void TextureStorage::texture_set_size_override(RID p_texture, int p_width, int p_height) { + Texture *tex = texture_owner.get_or_null(p_texture); + ERR_FAIL_COND(!tex); + ERR_FAIL_COND(tex->type != TextureStorage::TYPE_2D); + + tex->width_2d = p_width; + tex->height_2d = p_height; +} + +void TextureStorage::texture_set_path(RID p_texture, const String &p_path) { + Texture *tex = texture_owner.get_or_null(p_texture); + ERR_FAIL_COND(!tex); + + tex->path = p_path; +} + +String TextureStorage::texture_get_path(RID p_texture) const { + Texture *tex = texture_owner.get_or_null(p_texture); + ERR_FAIL_COND_V(!tex, String()); + + return tex->path; +} + +void TextureStorage::texture_set_detect_3d_callback(RID p_texture, RS::TextureDetectCallback p_callback, void *p_userdata) { + Texture *tex = texture_owner.get_or_null(p_texture); + ERR_FAIL_COND(!tex); + + tex->detect_3d_callback_ud = p_userdata; + tex->detect_3d_callback = p_callback; +} + +void TextureStorage::texture_set_detect_normal_callback(RID p_texture, RS::TextureDetectCallback p_callback, void *p_userdata) { + Texture *tex = texture_owner.get_or_null(p_texture); + ERR_FAIL_COND(!tex); + + tex->detect_normal_callback_ud = p_userdata; + tex->detect_normal_callback = p_callback; +} + +void TextureStorage::texture_set_detect_roughness_callback(RID p_texture, RS::TextureDetectRoughnessCallback p_callback, void *p_userdata) { + Texture *tex = texture_owner.get_or_null(p_texture); + ERR_FAIL_COND(!tex); + + tex->detect_roughness_callback_ud = p_userdata; + tex->detect_roughness_callback = p_callback; +} + +void TextureStorage::texture_debug_usage(List<RS::TextureInfo> *r_info) { +} + +void TextureStorage::texture_set_force_redraw_if_visible(RID p_texture, bool p_enable) { +} + +Size2 TextureStorage::texture_size_with_proxy(RID p_proxy) { + return texture_2d_get_size(p_proxy); +} + +Ref<Image> TextureStorage::_validate_texture_format(const Ref<Image> &p_image, TextureToRDFormat &r_format) { + Ref<Image> image = p_image->duplicate(); + + switch (p_image->get_format()) { + case Image::FORMAT_L8: { + r_format.format = RD::DATA_FORMAT_R8_UNORM; + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_ONE; + } break; //luminance + case Image::FORMAT_LA8: { + r_format.format = RD::DATA_FORMAT_R8G8_UNORM; + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_G; + } break; //luminance-alpha + case Image::FORMAT_R8: { + r_format.format = RD::DATA_FORMAT_R8_UNORM; + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_ZERO; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_ZERO; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_ONE; + } break; + case Image::FORMAT_RG8: { + r_format.format = RD::DATA_FORMAT_R8G8_UNORM; + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_G; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_ZERO; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_ONE; + } break; + case Image::FORMAT_RGB8: { + //this format is not mandatory for specification, check if supported first + if (false && RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_R8G8B8_UNORM, RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT) && RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_R8G8B8_SRGB, RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT)) { + r_format.format = RD::DATA_FORMAT_R8G8B8_UNORM; + r_format.format_srgb = RD::DATA_FORMAT_R8G8B8_SRGB; + } else { + //not supported, reconvert + r_format.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + r_format.format_srgb = RD::DATA_FORMAT_R8G8B8A8_SRGB; + image->convert(Image::FORMAT_RGBA8); + } + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_G; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_B; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_ONE; + + } break; + case Image::FORMAT_RGBA8: { + r_format.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + r_format.format_srgb = RD::DATA_FORMAT_R8G8B8A8_SRGB; + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_G; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_B; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_A; + } break; + case Image::FORMAT_RGBA4444: { + r_format.format = RD::DATA_FORMAT_B4G4R4A4_UNORM_PACK16; + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_B; //needs swizzle + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_G; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_A; + } break; + case Image::FORMAT_RGB565: { + r_format.format = RD::DATA_FORMAT_B5G6R5_UNORM_PACK16; + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_B; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_G; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_A; + } break; + case Image::FORMAT_RF: { + r_format.format = RD::DATA_FORMAT_R32_SFLOAT; + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_ZERO; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_ZERO; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_ONE; + } break; //float + case Image::FORMAT_RGF: { + r_format.format = RD::DATA_FORMAT_R32G32_SFLOAT; + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_G; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_ZERO; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_ONE; + } break; + case Image::FORMAT_RGBF: { + //this format is not mandatory for specification, check if supported first + if (RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_R32G32B32_SFLOAT, RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT)) { + r_format.format = RD::DATA_FORMAT_R32G32B32_SFLOAT; + } else { + //not supported, reconvert + r_format.format = RD::DATA_FORMAT_R32G32B32A32_SFLOAT; + image->convert(Image::FORMAT_RGBAF); + } + + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_G; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_B; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_ONE; + } break; + case Image::FORMAT_RGBAF: { + r_format.format = RD::DATA_FORMAT_R32G32B32A32_SFLOAT; + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_G; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_B; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_A; + + } break; + case Image::FORMAT_RH: { + r_format.format = RD::DATA_FORMAT_R16_SFLOAT; + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_ZERO; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_ZERO; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_ONE; + + } break; //half float + case Image::FORMAT_RGH: { + r_format.format = RD::DATA_FORMAT_R16G16_SFLOAT; + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_G; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_ZERO; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_ONE; + + } break; + case Image::FORMAT_RGBH: { + //this format is not mandatory for specification, check if supported first + if (RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_R16G16B16_SFLOAT, RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT)) { + r_format.format = RD::DATA_FORMAT_R16G16B16_SFLOAT; + } else { + //not supported, reconvert + r_format.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + image->convert(Image::FORMAT_RGBAH); + } + + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_G; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_B; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_ONE; + } break; + case Image::FORMAT_RGBAH: { + r_format.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_G; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_B; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_A; + + } break; + case Image::FORMAT_RGBE9995: { + r_format.format = RD::DATA_FORMAT_E5B9G9R9_UFLOAT_PACK32; + // TODO: Need to make a function in Image to swap bits for this. + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_IDENTITY; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_IDENTITY; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_IDENTITY; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_IDENTITY; + } break; + case Image::FORMAT_DXT1: { + if (RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_BC1_RGB_UNORM_BLOCK, RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT)) { + r_format.format = RD::DATA_FORMAT_BC1_RGB_UNORM_BLOCK; + r_format.format_srgb = RD::DATA_FORMAT_BC1_RGB_SRGB_BLOCK; + } else { + //not supported, reconvert + r_format.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + r_format.format_srgb = RD::DATA_FORMAT_R8G8B8A8_SRGB; + image->decompress(); + image->convert(Image::FORMAT_RGBA8); + } + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_G; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_B; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_ONE; + + } break; //s3tc bc1 + case Image::FORMAT_DXT3: { + if (RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_BC2_UNORM_BLOCK, RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT)) { + r_format.format = RD::DATA_FORMAT_BC2_UNORM_BLOCK; + r_format.format_srgb = RD::DATA_FORMAT_BC2_SRGB_BLOCK; + } else { + //not supported, reconvert + r_format.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + r_format.format_srgb = RD::DATA_FORMAT_R8G8B8A8_SRGB; + image->decompress(); + image->convert(Image::FORMAT_RGBA8); + } + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_G; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_B; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_A; + + } break; //bc2 + case Image::FORMAT_DXT5: { + if (RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_BC3_UNORM_BLOCK, RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT)) { + r_format.format = RD::DATA_FORMAT_BC3_UNORM_BLOCK; + r_format.format_srgb = RD::DATA_FORMAT_BC3_SRGB_BLOCK; + } else { + //not supported, reconvert + r_format.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + r_format.format_srgb = RD::DATA_FORMAT_R8G8B8A8_SRGB; + image->decompress(); + image->convert(Image::FORMAT_RGBA8); + } + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_G; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_B; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_A; + } break; //bc3 + case Image::FORMAT_RGTC_R: { + if (RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_BC4_UNORM_BLOCK, RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT)) { + r_format.format = RD::DATA_FORMAT_BC4_UNORM_BLOCK; + } else { + //not supported, reconvert + r_format.format = RD::DATA_FORMAT_R8_UNORM; + image->decompress(); + image->convert(Image::FORMAT_R8); + } + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_ZERO; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_ZERO; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_ONE; + + } break; + case Image::FORMAT_RGTC_RG: { + if (RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_BC5_UNORM_BLOCK, RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT)) { + r_format.format = RD::DATA_FORMAT_BC5_UNORM_BLOCK; + } else { + //not supported, reconvert + r_format.format = RD::DATA_FORMAT_R8G8_UNORM; + image->decompress(); + image->convert(Image::FORMAT_RG8); + } + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_G; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_ZERO; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_ONE; + + } break; + case Image::FORMAT_BPTC_RGBA: { + if (RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_BC7_UNORM_BLOCK, RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT)) { + r_format.format = RD::DATA_FORMAT_BC7_UNORM_BLOCK; + r_format.format_srgb = RD::DATA_FORMAT_BC7_SRGB_BLOCK; + } else { + //not supported, reconvert + r_format.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + r_format.format_srgb = RD::DATA_FORMAT_R8G8B8A8_SRGB; + image->decompress(); + image->convert(Image::FORMAT_RGBA8); + } + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_G; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_B; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_A; + + } break; //btpc bc7 + case Image::FORMAT_BPTC_RGBF: { + if (RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_BC6H_SFLOAT_BLOCK, RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT)) { + r_format.format = RD::DATA_FORMAT_BC6H_SFLOAT_BLOCK; + } else { + //not supported, reconvert + r_format.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + image->decompress(); + image->convert(Image::FORMAT_RGBAH); + } + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_G; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_B; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_ONE; + } break; //float bc6h + case Image::FORMAT_BPTC_RGBFU: { + if (RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_BC6H_UFLOAT_BLOCK, RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT)) { + r_format.format = RD::DATA_FORMAT_BC6H_UFLOAT_BLOCK; + } else { + //not supported, reconvert + r_format.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + image->decompress(); + image->convert(Image::FORMAT_RGBAH); + } + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_G; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_B; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_ONE; + } break; //unsigned float bc6hu + case Image::FORMAT_ETC2_R11: { + if (RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_EAC_R11_UNORM_BLOCK, RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT)) { + r_format.format = RD::DATA_FORMAT_EAC_R11_UNORM_BLOCK; + } else { + //not supported, reconvert + r_format.format = RD::DATA_FORMAT_R8_UNORM; + image->decompress(); + image->convert(Image::FORMAT_R8); + } + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_ZERO; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_ZERO; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_ONE; + + } break; //etc2 + case Image::FORMAT_ETC2_R11S: { + if (RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_EAC_R11_SNORM_BLOCK, RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT)) { + r_format.format = RD::DATA_FORMAT_EAC_R11_SNORM_BLOCK; + } else { + //not supported, reconvert + r_format.format = RD::DATA_FORMAT_R8_SNORM; + image->decompress(); + image->convert(Image::FORMAT_R8); + } + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_ZERO; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_ZERO; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_ONE; + } break; //signed: {} break; NOT srgb. + case Image::FORMAT_ETC2_RG11: { + if (RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_EAC_R11G11_UNORM_BLOCK, RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT)) { + r_format.format = RD::DATA_FORMAT_EAC_R11G11_UNORM_BLOCK; + } else { + //not supported, reconvert + r_format.format = RD::DATA_FORMAT_R8G8_UNORM; + image->decompress(); + image->convert(Image::FORMAT_RG8); + } + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_G; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_ZERO; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_ONE; + } break; + case Image::FORMAT_ETC2_RG11S: { + if (RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_EAC_R11G11_SNORM_BLOCK, RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT)) { + r_format.format = RD::DATA_FORMAT_EAC_R11G11_SNORM_BLOCK; + } else { + //not supported, reconvert + r_format.format = RD::DATA_FORMAT_R8G8_SNORM; + image->decompress(); + image->convert(Image::FORMAT_RG8); + } + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_G; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_ZERO; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_ONE; + } break; + case Image::FORMAT_ETC: + case Image::FORMAT_ETC2_RGB8: { + //ETC2 is backwards compatible with ETC1, and all modern platforms support it + if (RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_ETC2_R8G8B8_UNORM_BLOCK, RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT)) { + r_format.format = RD::DATA_FORMAT_ETC2_R8G8B8_UNORM_BLOCK; + r_format.format_srgb = RD::DATA_FORMAT_ETC2_R8G8B8_SRGB_BLOCK; + } else { + //not supported, reconvert + r_format.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + r_format.format_srgb = RD::DATA_FORMAT_R8G8B8A8_SRGB; + image->decompress(); + image->convert(Image::FORMAT_RGBA8); + } + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_G; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_B; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_ONE; + + } break; + case Image::FORMAT_ETC2_RGBA8: { + if (RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_ETC2_R8G8B8A8_UNORM_BLOCK, RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT)) { + r_format.format = RD::DATA_FORMAT_ETC2_R8G8B8A8_UNORM_BLOCK; + r_format.format_srgb = RD::DATA_FORMAT_ETC2_R8G8B8A8_SRGB_BLOCK; + } else { + //not supported, reconvert + r_format.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + r_format.format_srgb = RD::DATA_FORMAT_R8G8B8A8_SRGB; + image->decompress(); + image->convert(Image::FORMAT_RGBA8); + } + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_G; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_B; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_A; + } break; + case Image::FORMAT_ETC2_RGB8A1: { + if (RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_ETC2_R8G8B8A1_UNORM_BLOCK, RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT)) { + r_format.format = RD::DATA_FORMAT_ETC2_R8G8B8A1_UNORM_BLOCK; + r_format.format_srgb = RD::DATA_FORMAT_ETC2_R8G8B8A1_SRGB_BLOCK; + } else { + //not supported, reconvert + r_format.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + r_format.format_srgb = RD::DATA_FORMAT_R8G8B8A8_SRGB; + image->decompress(); + image->convert(Image::FORMAT_RGBA8); + } + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_G; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_B; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_A; + } break; + case Image::FORMAT_ETC2_RA_AS_RG: { + if (RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_ETC2_R8G8B8A8_UNORM_BLOCK, RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT)) { + r_format.format = RD::DATA_FORMAT_ETC2_R8G8B8A8_UNORM_BLOCK; + r_format.format_srgb = RD::DATA_FORMAT_ETC2_R8G8B8A8_SRGB_BLOCK; + } else { + //not supported, reconvert + r_format.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + r_format.format_srgb = RD::DATA_FORMAT_R8G8B8A8_SRGB; + image->decompress(); + image->convert(Image::FORMAT_RGBA8); + } + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_A; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_ZERO; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_ONE; + } break; + case Image::FORMAT_DXT5_RA_AS_RG: { + if (RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_BC3_UNORM_BLOCK, RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT)) { + r_format.format = RD::DATA_FORMAT_BC3_UNORM_BLOCK; + r_format.format_srgb = RD::DATA_FORMAT_BC3_SRGB_BLOCK; + } else { + //not supported, reconvert + r_format.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + r_format.format_srgb = RD::DATA_FORMAT_R8G8B8A8_SRGB; + image->decompress(); + image->convert(Image::FORMAT_RGBA8); + } + r_format.swizzle_r = RD::TEXTURE_SWIZZLE_R; + r_format.swizzle_g = RD::TEXTURE_SWIZZLE_A; + r_format.swizzle_b = RD::TEXTURE_SWIZZLE_ZERO; + r_format.swizzle_a = RD::TEXTURE_SWIZZLE_ONE; + } break; + + default: { + } + } + + return image; +} + +/* DECAL API */ + +RID TextureStorage::decal_atlas_get_texture() const { + return decal_atlas.texture; +} + +RID TextureStorage::decal_atlas_get_texture_srgb() const { + return decal_atlas.texture_srgb; +} + +RID TextureStorage::decal_allocate() { + return decal_owner.allocate_rid(); +} + +void TextureStorage::decal_initialize(RID p_decal) { + decal_owner.initialize_rid(p_decal, Decal()); +} + +void TextureStorage::decal_free(RID p_rid) { + Decal *decal = decal_owner.get_or_null(p_rid); + for (int i = 0; i < RS::DECAL_TEXTURE_MAX; i++) { + if (decal->textures[i].is_valid() && owns_texture(decal->textures[i])) { + texture_remove_from_decal_atlas(decal->textures[i]); + } + } + decal->dependency.deleted_notify(p_rid); + decal_owner.free(p_rid); +} + +void TextureStorage::decal_set_extents(RID p_decal, const Vector3 &p_extents) { + Decal *decal = decal_owner.get_or_null(p_decal); + ERR_FAIL_COND(!decal); + decal->extents = p_extents; + decal->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); +} + +void TextureStorage::decal_set_texture(RID p_decal, RS::DecalTexture p_type, RID p_texture) { + Decal *decal = decal_owner.get_or_null(p_decal); + ERR_FAIL_COND(!decal); + ERR_FAIL_INDEX(p_type, RS::DECAL_TEXTURE_MAX); + + if (decal->textures[p_type] == p_texture) { + return; + } + + ERR_FAIL_COND(p_texture.is_valid() && !owns_texture(p_texture)); + + if (decal->textures[p_type].is_valid() && owns_texture(decal->textures[p_type])) { + texture_remove_from_decal_atlas(decal->textures[p_type]); + } + + decal->textures[p_type] = p_texture; + + if (decal->textures[p_type].is_valid()) { + texture_add_to_decal_atlas(decal->textures[p_type]); + } + + decal->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_DECAL); +} + +void TextureStorage::decal_set_emission_energy(RID p_decal, float p_energy) { + Decal *decal = decal_owner.get_or_null(p_decal); + ERR_FAIL_COND(!decal); + decal->emission_energy = p_energy; +} + +void TextureStorage::decal_set_albedo_mix(RID p_decal, float p_mix) { + Decal *decal = decal_owner.get_or_null(p_decal); + ERR_FAIL_COND(!decal); + decal->albedo_mix = p_mix; +} + +void TextureStorage::decal_set_modulate(RID p_decal, const Color &p_modulate) { + Decal *decal = decal_owner.get_or_null(p_decal); + ERR_FAIL_COND(!decal); + decal->modulate = p_modulate; +} + +void TextureStorage::decal_set_cull_mask(RID p_decal, uint32_t p_layers) { + Decal *decal = decal_owner.get_or_null(p_decal); + ERR_FAIL_COND(!decal); + decal->cull_mask = p_layers; + decal->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); +} + +void TextureStorage::decal_set_distance_fade(RID p_decal, bool p_enabled, float p_begin, float p_length) { + Decal *decal = decal_owner.get_or_null(p_decal); + ERR_FAIL_COND(!decal); + decal->distance_fade = p_enabled; + decal->distance_fade_begin = p_begin; + decal->distance_fade_length = p_length; +} + +void TextureStorage::decal_set_fade(RID p_decal, float p_above, float p_below) { + Decal *decal = decal_owner.get_or_null(p_decal); + ERR_FAIL_COND(!decal); + decal->upper_fade = p_above; + decal->lower_fade = p_below; +} + +void TextureStorage::decal_set_normal_fade(RID p_decal, float p_fade) { + Decal *decal = decal_owner.get_or_null(p_decal); + ERR_FAIL_COND(!decal); + decal->normal_fade = p_fade; +} + +void TextureStorage::decal_atlas_mark_dirty_on_texture(RID p_texture) { + if (decal_atlas.textures.has(p_texture)) { + //belongs to decal atlas.. + + decal_atlas.dirty = true; //mark it dirty since it was most likely modified + } +} + +void TextureStorage::decal_atlas_remove_texture(RID p_texture) { + if (decal_atlas.textures.has(p_texture)) { + decal_atlas.textures.erase(p_texture); + //there is not much a point of making it dirty, just let it be. + } +} + +AABB TextureStorage::decal_get_aabb(RID p_decal) const { + Decal *decal = decal_owner.get_or_null(p_decal); + ERR_FAIL_COND_V(!decal, AABB()); + + return AABB(-decal->extents, decal->extents * 2.0); +} + +Dependency *TextureStorage::decal_get_dependency(RID p_decal) { + Decal *decal = decal_owner.get_or_null(p_decal); + ERR_FAIL_COND_V(!decal, nullptr); + + return &decal->dependency; +} + +void TextureStorage::update_decal_atlas() { + CopyEffects *copy_effects = CopyEffects::get_singleton(); + ERR_FAIL_NULL(copy_effects); + + if (!decal_atlas.dirty) { + return; //nothing to do + } + + decal_atlas.dirty = false; + + if (decal_atlas.texture.is_valid()) { + RD::get_singleton()->free(decal_atlas.texture); + decal_atlas.texture = RID(); + decal_atlas.texture_srgb = RID(); + decal_atlas.texture_mipmaps.clear(); + } + + int border = 1 << decal_atlas.mipmaps; + + if (decal_atlas.textures.size()) { + //generate atlas + Vector<DecalAtlas::SortItem> itemsv; + itemsv.resize(decal_atlas.textures.size()); + int base_size = 8; + + int idx = 0; + + for (const KeyValue<RID, DecalAtlas::Texture> &E : decal_atlas.textures) { + DecalAtlas::SortItem &si = itemsv.write[idx]; + + Texture *src_tex = get_texture(E.key); + + si.size.width = (src_tex->width / border) + 1; + si.size.height = (src_tex->height / border) + 1; + si.pixel_size = Size2i(src_tex->width, src_tex->height); + + if (base_size < si.size.width) { + base_size = nearest_power_of_2_templated(si.size.width); + } + + si.texture = E.key; + idx++; + } + + //sort items by size + itemsv.sort(); + + //attempt to create atlas + int item_count = itemsv.size(); + DecalAtlas::SortItem *items = itemsv.ptrw(); + + int atlas_height = 0; + + while (true) { + Vector<int> v_offsetsv; + v_offsetsv.resize(base_size); + + int *v_offsets = v_offsetsv.ptrw(); + memset(v_offsets, 0, sizeof(int) * base_size); + + int max_height = 0; + + for (int i = 0; i < item_count; i++) { + //best fit + DecalAtlas::SortItem &si = items[i]; + int best_idx = -1; + int best_height = 0x7FFFFFFF; + for (int j = 0; j <= base_size - si.size.width; j++) { + int height = 0; + for (int k = 0; k < si.size.width; k++) { + int h = v_offsets[k + j]; + if (h > height) { + height = h; + if (height > best_height) { + break; //already bad + } + } + } + + if (height < best_height) { + best_height = height; + best_idx = j; + } + } + + //update + for (int k = 0; k < si.size.width; k++) { + v_offsets[k + best_idx] = best_height + si.size.height; + } + + si.pos.x = best_idx; + si.pos.y = best_height; + + if (si.pos.y + si.size.height > max_height) { + max_height = si.pos.y + si.size.height; + } + } + + if (max_height <= base_size * 2) { + atlas_height = max_height; + break; //good ratio, break; + } + + base_size *= 2; + } + + decal_atlas.size.width = base_size * border; + decal_atlas.size.height = nearest_power_of_2_templated(atlas_height * border); + + for (int i = 0; i < item_count; i++) { + DecalAtlas::Texture *t = decal_atlas.textures.getptr(items[i].texture); + t->uv_rect.position = items[i].pos * border + Vector2i(border / 2, border / 2); + t->uv_rect.size = items[i].pixel_size; + + t->uv_rect.position /= Size2(decal_atlas.size); + t->uv_rect.size /= Size2(decal_atlas.size); + } + } else { + //use border as size, so it at least has enough mipmaps + decal_atlas.size.width = border; + decal_atlas.size.height = border; + } + + //blit textures + + RD::TextureFormat tformat; + tformat.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + tformat.width = decal_atlas.size.width; + tformat.height = decal_atlas.size.height; + tformat.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; + tformat.texture_type = RD::TEXTURE_TYPE_2D; + tformat.mipmaps = decal_atlas.mipmaps; + tformat.shareable_formats.push_back(RD::DATA_FORMAT_R8G8B8A8_UNORM); + tformat.shareable_formats.push_back(RD::DATA_FORMAT_R8G8B8A8_SRGB); + + decal_atlas.texture = RD::get_singleton()->texture_create(tformat, RD::TextureView()); + RD::get_singleton()->texture_clear(decal_atlas.texture, Color(0, 0, 0, 0), 0, decal_atlas.mipmaps, 0, 1); + + { + //create the framebuffer + + Size2i s = decal_atlas.size; + + for (int i = 0; i < decal_atlas.mipmaps; i++) { + DecalAtlas::MipMap mm; + mm.texture = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), decal_atlas.texture, 0, i); + Vector<RID> fb; + fb.push_back(mm.texture); + mm.fb = RD::get_singleton()->framebuffer_create(fb); + mm.size = s; + decal_atlas.texture_mipmaps.push_back(mm); + + s.width = MAX(1, s.width >> 1); + s.height = MAX(1, s.height >> 1); + } + { + //create the SRGB variant + RD::TextureView rd_view; + rd_view.format_override = RD::DATA_FORMAT_R8G8B8A8_SRGB; + decal_atlas.texture_srgb = RD::get_singleton()->texture_create_shared(rd_view, decal_atlas.texture); + } + } + + RID prev_texture; + for (int i = 0; i < decal_atlas.texture_mipmaps.size(); i++) { + const DecalAtlas::MipMap &mm = decal_atlas.texture_mipmaps[i]; + + Color clear_color(0, 0, 0, 0); + + if (decal_atlas.textures.size()) { + if (i == 0) { + Vector<Color> cc; + cc.push_back(clear_color); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(mm.fb, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, cc); + + for (const KeyValue<RID, DecalAtlas::Texture> &E : decal_atlas.textures) { + DecalAtlas::Texture *t = decal_atlas.textures.getptr(E.key); + Texture *src_tex = get_texture(E.key); + copy_effects->copy_to_atlas_fb(src_tex->rd_texture, mm.fb, t->uv_rect, draw_list, false, t->panorama_to_dp_users > 0); + } + + RD::get_singleton()->draw_list_end(); + + prev_texture = mm.texture; + } else { + copy_effects->copy_to_fb_rect(prev_texture, mm.fb, Rect2i(Point2i(), mm.size)); + prev_texture = mm.texture; + } + } else { + RD::get_singleton()->texture_clear(mm.texture, clear_color, 0, 1, 0, 1); + } + } +} + +void TextureStorage::texture_add_to_decal_atlas(RID p_texture, bool p_panorama_to_dp) { + if (!decal_atlas.textures.has(p_texture)) { + DecalAtlas::Texture t; + t.users = 1; + t.panorama_to_dp_users = p_panorama_to_dp ? 1 : 0; + decal_atlas.textures[p_texture] = t; + decal_atlas.dirty = true; + } else { + DecalAtlas::Texture *t = decal_atlas.textures.getptr(p_texture); + t->users++; + if (p_panorama_to_dp) { + t->panorama_to_dp_users++; + } + } +} + +void TextureStorage::texture_remove_from_decal_atlas(RID p_texture, bool p_panorama_to_dp) { + DecalAtlas::Texture *t = decal_atlas.textures.getptr(p_texture); + ERR_FAIL_COND(!t); + t->users--; + if (p_panorama_to_dp) { + ERR_FAIL_COND(t->panorama_to_dp_users == 0); + t->panorama_to_dp_users--; + } + if (t->users == 0) { + decal_atlas.textures.erase(p_texture); + //do not mark it dirty, there is no need to since it remains working + } +} + +/* DECAL INSTANCE API */ + +RID TextureStorage::decal_instance_create(RID p_decal) { + DecalInstance di; + di.decal = p_decal; + di.forward_id = ForwardIDStorage::get_singleton()->allocate_forward_id(FORWARD_ID_TYPE_DECAL); + return decal_instance_owner.make_rid(di); +} + +void TextureStorage::decal_instance_free(RID p_decal_instance) { + DecalInstance *di = decal_instance_owner.get_or_null(p_decal_instance); + ForwardIDStorage::get_singleton()->free_forward_id(FORWARD_ID_TYPE_DECAL, di->forward_id); + decal_instance_owner.free(p_decal_instance); +} + +void TextureStorage::decal_instance_set_transform(RID p_decal_instance, const Transform3D &p_transform) { + DecalInstance *di = decal_instance_owner.get_or_null(p_decal_instance); + ERR_FAIL_COND(!di); + di->transform = p_transform; +} + +/* DECAL DATA API */ + +void TextureStorage::free_decal_data() { + if (decal_buffer.is_valid()) { + RD::get_singleton()->free(decal_buffer); + decal_buffer = RID(); + } + + if (decals != nullptr) { + memdelete_arr(decals); + decals = nullptr; + } + + if (decal_sort != nullptr) { + memdelete_arr(decal_sort); + decal_sort = nullptr; + } +} + +void TextureStorage::set_max_decals(const uint32_t p_max_decals) { + max_decals = p_max_decals; + uint32_t decal_buffer_size = max_decals * sizeof(DecalData); + decals = memnew_arr(DecalData, max_decals); + decal_sort = memnew_arr(DecalInstanceSort, max_decals); + decal_buffer = RD::get_singleton()->storage_buffer_create(decal_buffer_size); +} + +void TextureStorage::update_decal_buffer(const PagedArray<RID> &p_decals, const Transform3D &p_camera_inverse_xform) { + ForwardIDStorage *forward_id_storage = ForwardIDStorage::get_singleton(); + + Transform3D uv_xform; + uv_xform.basis.scale(Vector3(2.0, 1.0, 2.0)); + uv_xform.origin = Vector3(-1.0, 0.0, -1.0); + + uint32_t decals_size = p_decals.size(); + + decal_count = 0; + + for (uint32_t i = 0; i < decals_size; i++) { + if (decal_count == max_decals) { + break; + } + + DecalInstance *decal_instance = decal_instance_owner.get_or_null(p_decals[i]); + if (!decal_instance) { + continue; + } + Decal *decal = decal_owner.get_or_null(decal_instance->decal); + + Transform3D xform = decal_instance->transform; + + real_t distance = -p_camera_inverse_xform.xform(xform.origin).z; + + if (decal->distance_fade) { + float fade_begin = decal->distance_fade_begin; + float fade_length = decal->distance_fade_length; + + if (distance > fade_begin) { + if (distance > fade_begin + fade_length) { + continue; // do not use this decal, its invisible + } + } + } + + decal_sort[decal_count].decal_instance = decal_instance; + decal_sort[decal_count].decal = decal; + decal_sort[decal_count].depth = distance; + decal_count++; + } + + if (decal_count > 0) { + SortArray<DecalInstanceSort> sort_array; + sort_array.sort(decal_sort, decal_count); + } + + bool using_forward_ids = forward_id_storage->uses_forward_ids(); + for (uint32_t i = 0; i < decal_count; i++) { + DecalInstance *decal_instance = decal_sort[i].decal_instance; + Decal *decal = decal_sort[i].decal; + + if (using_forward_ids) { + forward_id_storage->map_forward_id(FORWARD_ID_TYPE_DECAL, decal_instance->forward_id, i); + } + + decal_instance->cull_mask = decal->cull_mask; + + Transform3D xform = decal_instance->transform; + float fade = 1.0; + + if (decal->distance_fade) { + const real_t distance = -p_camera_inverse_xform.xform(xform.origin).z; + const float fade_begin = decal->distance_fade_begin; + const float fade_length = decal->distance_fade_length; + + if (distance > fade_begin) { + // Use `smoothstep()` to make opacity changes more gradual and less noticeable to the player. + fade = Math::smoothstep(0.0f, 1.0f, 1.0f - float(distance - fade_begin) / fade_length); + } + } + + DecalData &dd = decals[i]; + + Vector3 decal_extents = decal->extents; + + Transform3D scale_xform; + scale_xform.basis.scale(decal_extents); + Transform3D to_decal_xform = (p_camera_inverse_xform * xform * scale_xform * uv_xform).affine_inverse(); + MaterialStorage::store_transform(to_decal_xform, dd.xform); + + Vector3 normal = xform.basis.get_column(Vector3::AXIS_Y).normalized(); + normal = p_camera_inverse_xform.basis.xform(normal); //camera is normalized, so fine + + dd.normal[0] = normal.x; + dd.normal[1] = normal.y; + dd.normal[2] = normal.z; + dd.normal_fade = decal->normal_fade; + + RID albedo_tex = decal->textures[RS::DECAL_TEXTURE_ALBEDO]; + RID emission_tex = decal->textures[RS::DECAL_TEXTURE_EMISSION]; + if (albedo_tex.is_valid()) { + Rect2 rect = decal_atlas_get_texture_rect(albedo_tex); + dd.albedo_rect[0] = rect.position.x; + dd.albedo_rect[1] = rect.position.y; + dd.albedo_rect[2] = rect.size.x; + dd.albedo_rect[3] = rect.size.y; + } else { + if (!emission_tex.is_valid()) { + continue; //no albedo, no emission, no decal. + } + dd.albedo_rect[0] = 0; + dd.albedo_rect[1] = 0; + dd.albedo_rect[2] = 0; + dd.albedo_rect[3] = 0; + } + + RID normal_tex = decal->textures[RS::DECAL_TEXTURE_NORMAL]; + + if (normal_tex.is_valid()) { + Rect2 rect = decal_atlas_get_texture_rect(normal_tex); + dd.normal_rect[0] = rect.position.x; + dd.normal_rect[1] = rect.position.y; + dd.normal_rect[2] = rect.size.x; + dd.normal_rect[3] = rect.size.y; + + Basis normal_xform = p_camera_inverse_xform.basis * xform.basis.orthonormalized(); + MaterialStorage::store_basis_3x4(normal_xform, dd.normal_xform); + } else { + dd.normal_rect[0] = 0; + dd.normal_rect[1] = 0; + dd.normal_rect[2] = 0; + dd.normal_rect[3] = 0; + } + + RID orm_tex = decal->textures[RS::DECAL_TEXTURE_ORM]; + if (orm_tex.is_valid()) { + Rect2 rect = decal_atlas_get_texture_rect(orm_tex); + dd.orm_rect[0] = rect.position.x; + dd.orm_rect[1] = rect.position.y; + dd.orm_rect[2] = rect.size.x; + dd.orm_rect[3] = rect.size.y; + } else { + dd.orm_rect[0] = 0; + dd.orm_rect[1] = 0; + dd.orm_rect[2] = 0; + dd.orm_rect[3] = 0; + } + + if (emission_tex.is_valid()) { + Rect2 rect = decal_atlas_get_texture_rect(emission_tex); + dd.emission_rect[0] = rect.position.x; + dd.emission_rect[1] = rect.position.y; + dd.emission_rect[2] = rect.size.x; + dd.emission_rect[3] = rect.size.y; + } else { + dd.emission_rect[0] = 0; + dd.emission_rect[1] = 0; + dd.emission_rect[2] = 0; + dd.emission_rect[3] = 0; + } + + Color modulate = decal->modulate; + dd.modulate[0] = modulate.r; + dd.modulate[1] = modulate.g; + dd.modulate[2] = modulate.b; + dd.modulate[3] = modulate.a * fade; + dd.emission_energy = decal->emission_energy * fade; + dd.albedo_mix = decal->albedo_mix; + dd.mask = decal->cull_mask; + dd.upper_fade = decal->upper_fade; + dd.lower_fade = decal->lower_fade; + + // hook for subclass to do further processing. + RendererSceneRenderRD::get_singleton()->setup_added_decal(xform, decal_extents); + } + + if (decal_count > 0) { + RD::get_singleton()->buffer_update(decal_buffer, 0, sizeof(DecalData) * decal_count, decals, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); + } +} + +/* RENDER TARGET API */ + +RID TextureStorage::RenderTarget::get_framebuffer() { + // Note that if we're using an overridden color buffer, we're likely cycling through a texture chain. + // this is where our framebuffer cache comes in clutch.. + + if (msaa != RS::VIEWPORT_MSAA_DISABLED) { + return FramebufferCacheRD::get_singleton()->get_cache_multiview(view_count, color_multisample, overridden.color.is_valid() ? overridden.color : color); + } else { + return FramebufferCacheRD::get_singleton()->get_cache_multiview(view_count, overridden.color.is_valid() ? overridden.color : color); + } +} + +void TextureStorage::_clear_render_target(RenderTarget *rt) { + // clear overrides, we assume these are freed by the object that created them + rt->overridden.color = RID(); + rt->overridden.depth = RID(); + rt->overridden.velocity = RID(); + rt->overridden.cached_slices.clear(); // these are automatically freed when their parent textures are freed so just clear + + // free in reverse dependency order + if (rt->framebuffer_uniform_set.is_valid()) { + rt->framebuffer_uniform_set = RID(); //chain deleted + } + + if (rt->color.is_valid()) { + RD::get_singleton()->free(rt->color); + } + rt->color_slices.clear(); // these are automatically freed. + + if (rt->color_multisample.is_valid()) { + RD::get_singleton()->free(rt->color_multisample); + } + + if (rt->backbuffer.is_valid()) { + RD::get_singleton()->free(rt->backbuffer); + rt->backbuffer = RID(); + rt->backbuffer_mipmaps.clear(); + rt->backbuffer_uniform_set = RID(); //chain deleted + } + + _render_target_clear_sdf(rt); + + rt->color = RID(); + rt->color_multisample = RID(); +} + +void TextureStorage::_update_render_target(RenderTarget *rt) { + if (rt->texture.is_null()) { + //create a placeholder until updated + rt->texture = texture_allocate(); + texture_2d_placeholder_initialize(rt->texture); + Texture *tex = get_texture(rt->texture); + tex->is_render_target = true; + } + + _clear_render_target(rt); + + if (rt->size.width == 0 || rt->size.height == 0) { + return; + } + //until we implement support for HDR monitors (and render target is attached to screen), this is enough. + rt->color_format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + rt->color_format_srgb = RD::DATA_FORMAT_R8G8B8A8_SRGB; + rt->image_format = rt->is_transparent ? Image::FORMAT_RGBA8 : Image::FORMAT_RGB8; + + RD::TextureFormat rd_color_attachment_format; + RD::TextureView rd_view; + { //attempt register + rd_color_attachment_format.format = rt->color_format; + rd_color_attachment_format.width = rt->size.width; + rd_color_attachment_format.height = rt->size.height; + rd_color_attachment_format.depth = 1; + rd_color_attachment_format.array_layers = rt->view_count; // for stereo we create two (or more) layers, need to see if we can make fallback work like this too if we don't have multiview + rd_color_attachment_format.mipmaps = 1; + if (rd_color_attachment_format.array_layers > 1) { // why are we not using rt->texture_type ?? + rd_color_attachment_format.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; + } else { + rd_color_attachment_format.texture_type = RD::TEXTURE_TYPE_2D; + } + rd_color_attachment_format.samples = RD::TEXTURE_SAMPLES_1; + rd_color_attachment_format.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT; + rd_color_attachment_format.usage_bits |= RD::TEXTURE_USAGE_STORAGE_BIT; // FIXME we need this only when FSR is enabled + rd_color_attachment_format.shareable_formats.push_back(rt->color_format); + rd_color_attachment_format.shareable_formats.push_back(rt->color_format_srgb); + if (rt->msaa != RS::VIEWPORT_MSAA_DISABLED) { + rd_color_attachment_format.is_resolve_buffer = true; + } + } + + // TODO see if we can lazy create this once we actually use it as we may not need to create this if we have an overridden color buffer... + rt->color = RD::get_singleton()->texture_create(rd_color_attachment_format, rd_view); + ERR_FAIL_COND(rt->color.is_null()); + + if (rt->msaa != RS::VIEWPORT_MSAA_DISABLED) { + // Use the texture format of the color attachment for the multisample color attachment. + RD::TextureFormat rd_color_multisample_format = rd_color_attachment_format; + const RD::TextureSamples texture_samples[RS::VIEWPORT_MSAA_MAX] = { + RD::TEXTURE_SAMPLES_1, + RD::TEXTURE_SAMPLES_2, + RD::TEXTURE_SAMPLES_4, + RD::TEXTURE_SAMPLES_8, + }; + rd_color_multisample_format.samples = texture_samples[rt->msaa]; + rd_color_multisample_format.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + RD::TextureView rd_view_multisample; + rd_color_multisample_format.is_resolve_buffer = false; + rt->color_multisample = RD::get_singleton()->texture_create(rd_color_multisample_format, rd_view_multisample); + ERR_FAIL_COND(rt->color_multisample.is_null()); + } + + { //update texture + + Texture *tex = get_texture(rt->texture); + + //free existing textures + if (RD::get_singleton()->texture_is_valid(tex->rd_texture)) { + RD::get_singleton()->free(tex->rd_texture); + } + if (RD::get_singleton()->texture_is_valid(tex->rd_texture_srgb)) { + RD::get_singleton()->free(tex->rd_texture_srgb); + } + + tex->rd_texture = RID(); + tex->rd_texture_srgb = RID(); + + //create shared textures to the color buffer, + //so transparent can be supported + RD::TextureView view; + view.format_override = rt->color_format; + if (!rt->is_transparent) { + view.swizzle_a = RD::TEXTURE_SWIZZLE_ONE; + } + tex->rd_texture = RD::get_singleton()->texture_create_shared(view, rt->color); + if (rt->color_format_srgb != RD::DATA_FORMAT_MAX) { + view.format_override = rt->color_format_srgb; + tex->rd_texture_srgb = RD::get_singleton()->texture_create_shared(view, rt->color); + } + tex->rd_view = view; + tex->width = rt->size.width; + tex->height = rt->size.height; + tex->width_2d = rt->size.width; + tex->height_2d = rt->size.height; + tex->rd_format = rt->color_format; + tex->rd_format_srgb = rt->color_format_srgb; + tex->format = rt->image_format; + + Vector<RID> proxies = tex->proxies; //make a copy, since update may change it + for (int i = 0; i < proxies.size(); i++) { + texture_proxy_update(proxies[i], rt->texture); + } + } +} + +void TextureStorage::_create_render_target_backbuffer(RenderTarget *rt) { + ERR_FAIL_COND(rt->backbuffer.is_valid()); + + uint32_t mipmaps_required = Image::get_image_required_mipmaps(rt->size.width, rt->size.height, Image::FORMAT_RGBA8); + RD::TextureFormat tf; + tf.format = rt->color_format; + tf.width = rt->size.width; + tf.height = rt->size.height; + tf.texture_type = RD::TEXTURE_TYPE_2D; + tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; + tf.mipmaps = mipmaps_required; + + rt->backbuffer = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(rt->backbuffer, "Render Target Back Buffer"); + rt->backbuffer_mipmap0 = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), rt->backbuffer, 0, 0); + RD::get_singleton()->set_resource_name(rt->backbuffer_mipmap0, "Back Buffer slice mipmap 0"); + + { + Vector<RID> fb_tex; + fb_tex.push_back(rt->backbuffer_mipmap0); + rt->backbuffer_fb = RD::get_singleton()->framebuffer_create(fb_tex); + } + + if (rt->framebuffer_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(rt->framebuffer_uniform_set)) { + //the new one will require the backbuffer. + RD::get_singleton()->free(rt->framebuffer_uniform_set); + rt->framebuffer_uniform_set = RID(); + } + //create mipmaps + for (uint32_t i = 1; i < mipmaps_required; i++) { + RID mipmap = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), rt->backbuffer, 0, i); + RD::get_singleton()->set_resource_name(mipmap, "Back Buffer slice mip: " + itos(i)); + + rt->backbuffer_mipmaps.push_back(mipmap); + } +} + +RID TextureStorage::render_target_create() { + RenderTarget render_target; + + render_target.was_used = false; + render_target.clear_requested = false; + + _update_render_target(&render_target); + return render_target_owner.make_rid(render_target); +} + +void TextureStorage::render_target_free(RID p_rid) { + RenderTarget *rt = render_target_owner.get_or_null(p_rid); + + _clear_render_target(rt); + + if (rt->texture.is_valid()) { + Texture *tex = get_texture(rt->texture); + tex->is_render_target = false; + texture_free(rt->texture); + } + + render_target_owner.free(p_rid); +} + +void TextureStorage::render_target_set_position(RID p_render_target, int p_x, int p_y) { + //unused for this render target +} + +Point2i TextureStorage::render_target_get_position(RID p_render_target) const { + //unused for this render target + return Point2i(); +} + +void TextureStorage::render_target_set_size(RID p_render_target, int p_width, int p_height, uint32_t p_view_count) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND(!rt); + if (rt->size.x != p_width || rt->size.y != p_height || rt->view_count != p_view_count) { + rt->size.x = p_width; + rt->size.y = p_height; + rt->view_count = p_view_count; + _update_render_target(rt); + } +} + +Size2i TextureStorage::render_target_get_size(RID p_render_target) const { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, Size2i()); + + return rt->size; +} + +RID TextureStorage::render_target_get_texture(RID p_render_target) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, RID()); + + return rt->texture; +} + +void TextureStorage::render_target_set_override_color(RID p_render_target, RID p_texture) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND(!rt); + + rt->overridden.color = p_texture; +} + +RID TextureStorage::render_target_get_override_color(RID p_render_target) const { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, RID()); + + return rt->overridden.color; +} + +void TextureStorage::render_target_set_override_depth(RID p_render_target, RID p_texture) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND(!rt); + + rt->overridden.depth = p_texture; +} + +RID TextureStorage::render_target_get_override_depth(RID p_render_target) const { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, RID()); + + return rt->overridden.depth; +} + +RID TextureStorage::render_target_get_override_depth_slice(RID p_render_target, const uint32_t p_layer) const { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, RID()); + + if (rt->overridden.depth.is_null()) { + return RID(); + } else if (rt->view_count == 1) { + return rt->overridden.depth; + } else { + RenderTarget::RTOverridden::SliceKey key(rt->overridden.depth, p_layer); + + if (!rt->overridden.cached_slices.has(key)) { + rt->overridden.cached_slices[key] = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), rt->overridden.depth, p_layer, 0); + } + + return rt->overridden.cached_slices[key]; + } +} + +void TextureStorage::render_target_set_override_velocity(RID p_render_target, RID p_texture) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND(!rt); + + rt->overridden.velocity = p_texture; +} + +RID TextureStorage::render_target_get_override_velocity(RID p_render_target) const { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, RID()); + + return rt->overridden.velocity; +} + +RID TextureStorage::render_target_get_override_velocity_slice(RID p_render_target, const uint32_t p_layer) const { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, RID()); + + if (rt->overridden.velocity.is_null()) { + return RID(); + } else if (rt->view_count == 1) { + return rt->overridden.velocity; + } else { + RenderTarget::RTOverridden::SliceKey key(rt->overridden.velocity, p_layer); + + if (!rt->overridden.cached_slices.has(key)) { + rt->overridden.cached_slices[key] = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), rt->overridden.velocity, p_layer, 0); + } + + return rt->overridden.cached_slices[key]; + } +} + +void TextureStorage::render_target_set_transparent(RID p_render_target, bool p_is_transparent) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND(!rt); + rt->is_transparent = p_is_transparent; + _update_render_target(rt); +} + +bool TextureStorage::render_target_get_transparent(RID p_render_target) const { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, false); + + return rt->is_transparent; +} + +void TextureStorage::render_target_set_direct_to_screen(RID p_render_target, bool p_value) { +} + +bool TextureStorage::render_target_get_direct_to_screen(RID p_render_target) const { + return false; +} + +bool TextureStorage::render_target_was_used(RID p_render_target) const { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, false); + return rt->was_used; +} + +void TextureStorage::render_target_set_as_unused(RID p_render_target) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND(!rt); + rt->was_used = false; +} + +void TextureStorage::render_target_set_msaa(RID p_render_target, RS::ViewportMSAA p_msaa) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND(!rt); + if (p_msaa == rt->msaa) { + return; + } + + rt->msaa = p_msaa; + _update_render_target(rt); +} + +RS::ViewportMSAA TextureStorage::render_target_get_msaa(RID p_render_target) const { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, RS::VIEWPORT_MSAA_DISABLED); + + return rt->msaa; +} + +RID TextureStorage::render_target_get_rd_framebuffer(RID p_render_target) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, RID()); + + return rt->get_framebuffer(); +} + +RID TextureStorage::render_target_get_rd_texture(RID p_render_target) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, RID()); + + if (rt->overridden.color.is_valid()) { + return rt->overridden.color; + } else { + return rt->color; + } +} + +RID TextureStorage::render_target_get_rd_texture_slice(RID p_render_target, uint32_t p_layer) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, RID()); + + if (rt->view_count == 1) { + return rt->color; + } else { + ERR_FAIL_UNSIGNED_INDEX_V(p_layer, rt->view_count, RID()); + if (rt->color_slices.size() == 0) { + for (uint32_t v = 0; v < rt->view_count; v++) { + RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), rt->color, v, 0); + rt->color_slices.push_back(slice); + } + } + return rt->color_slices[p_layer]; + } +} + +RID TextureStorage::render_target_get_rd_backbuffer(RID p_render_target) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, RID()); + return rt->backbuffer; +} + +RID TextureStorage::render_target_get_rd_backbuffer_framebuffer(RID p_render_target) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, RID()); + + if (!rt->backbuffer.is_valid()) { + _create_render_target_backbuffer(rt); + } + + return rt->backbuffer_fb; +} + +void TextureStorage::render_target_request_clear(RID p_render_target, const Color &p_clear_color) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND(!rt); + rt->clear_requested = true; + rt->clear_color = p_clear_color; +} + +bool TextureStorage::render_target_is_clear_requested(RID p_render_target) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, false); + return rt->clear_requested; +} + +Color TextureStorage::render_target_get_clear_request_color(RID p_render_target) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, Color()); + return rt->clear_color; +} + +void TextureStorage::render_target_disable_clear_request(RID p_render_target) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND(!rt); + rt->clear_requested = false; +} + +void TextureStorage::render_target_do_clear_request(RID p_render_target) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND(!rt); + if (!rt->clear_requested) { + return; + } + Vector<Color> clear_colors; + clear_colors.push_back(rt->clear_color); + RD::get_singleton()->draw_list_begin(rt->get_framebuffer(), RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD, clear_colors); + RD::get_singleton()->draw_list_end(); + rt->clear_requested = false; +} + +void TextureStorage::render_target_set_sdf_size_and_scale(RID p_render_target, RS::ViewportSDFOversize p_size, RS::ViewportSDFScale p_scale) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND(!rt); + if (rt->sdf_oversize == p_size && rt->sdf_scale == p_scale) { + return; + } + + rt->sdf_oversize = p_size; + rt->sdf_scale = p_scale; + + _render_target_clear_sdf(rt); +} + +Rect2i TextureStorage::_render_target_get_sdf_rect(const RenderTarget *rt) const { + Size2i margin; + int scale; + switch (rt->sdf_oversize) { + case RS::VIEWPORT_SDF_OVERSIZE_100_PERCENT: { + scale = 100; + } break; + case RS::VIEWPORT_SDF_OVERSIZE_120_PERCENT: { + scale = 120; + } break; + case RS::VIEWPORT_SDF_OVERSIZE_150_PERCENT: { + scale = 150; + } break; + case RS::VIEWPORT_SDF_OVERSIZE_200_PERCENT: { + scale = 200; + } break; + default: { + } + } + + margin = (rt->size * scale / 100) - rt->size; + + Rect2i r(Vector2i(), rt->size); + r.position -= margin; + r.size += margin * 2; + + return r; +} + +Rect2i TextureStorage::render_target_get_sdf_rect(RID p_render_target) const { + const RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, Rect2i()); + + return _render_target_get_sdf_rect(rt); +} + +void TextureStorage::render_target_mark_sdf_enabled(RID p_render_target, bool p_enabled) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND(!rt); + + rt->sdf_enabled = p_enabled; +} + +bool TextureStorage::render_target_is_sdf_enabled(RID p_render_target) const { + const RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, false); + + return rt->sdf_enabled; +} + +RID TextureStorage::render_target_get_sdf_texture(RID p_render_target) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, RID()); + if (rt->sdf_buffer_read.is_null()) { + // no texture, create a dummy one for the 2D uniform set + RD::TextureFormat tformat; + tformat.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + tformat.width = 4; + tformat.height = 4; + tformat.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT; + tformat.texture_type = RD::TEXTURE_TYPE_2D; + + Vector<uint8_t> pv; + pv.resize(16 * 4); + memset(pv.ptrw(), 0, 16 * 4); + Vector<Vector<uint8_t>> vpv; + + rt->sdf_buffer_read = RD::get_singleton()->texture_create(tformat, RD::TextureView(), vpv); + } + + return rt->sdf_buffer_read; +} + +void TextureStorage::_render_target_allocate_sdf(RenderTarget *rt) { + ERR_FAIL_COND(rt->sdf_buffer_write_fb.is_valid()); + if (rt->sdf_buffer_read.is_valid()) { + RD::get_singleton()->free(rt->sdf_buffer_read); + rt->sdf_buffer_read = RID(); + } + + Size2i size = _render_target_get_sdf_rect(rt).size; + + RD::TextureFormat tformat; + tformat.format = RD::DATA_FORMAT_R8_UNORM; + tformat.width = size.width; + tformat.height = size.height; + tformat.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + tformat.texture_type = RD::TEXTURE_TYPE_2D; + + rt->sdf_buffer_write = RD::get_singleton()->texture_create(tformat, RD::TextureView()); + + { + Vector<RID> write_fb; + write_fb.push_back(rt->sdf_buffer_write); + rt->sdf_buffer_write_fb = RD::get_singleton()->framebuffer_create(write_fb); + } + + int scale; + switch (rt->sdf_scale) { + case RS::VIEWPORT_SDF_SCALE_100_PERCENT: { + scale = 100; + } break; + case RS::VIEWPORT_SDF_SCALE_50_PERCENT: { + scale = 50; + } break; + case RS::VIEWPORT_SDF_SCALE_25_PERCENT: { + scale = 25; + } break; + default: { + scale = 100; + } break; + } + + rt->process_size = size * scale / 100; + rt->process_size.x = MAX(rt->process_size.x, 1); + rt->process_size.y = MAX(rt->process_size.y, 1); + + tformat.format = RD::DATA_FORMAT_R16G16_SINT; + tformat.width = rt->process_size.width; + tformat.height = rt->process_size.height; + tformat.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT; + + rt->sdf_buffer_process[0] = RD::get_singleton()->texture_create(tformat, RD::TextureView()); + rt->sdf_buffer_process[1] = RD::get_singleton()->texture_create(tformat, RD::TextureView()); + + tformat.format = RD::DATA_FORMAT_R16_SNORM; + tformat.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + + rt->sdf_buffer_read = RD::get_singleton()->texture_create(tformat, RD::TextureView()); + + { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 1; + u.append_id(rt->sdf_buffer_write); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 2; + u.append_id(rt->sdf_buffer_read); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 3; + u.append_id(rt->sdf_buffer_process[0]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 4; + u.append_id(rt->sdf_buffer_process[1]); + uniforms.push_back(u); + } + + rt->sdf_buffer_process_uniform_sets[0] = RD::get_singleton()->uniform_set_create(uniforms, rt_sdf.shader.version_get_shader(rt_sdf.shader_version, 0), 0); + RID aux2 = uniforms.write[2].get_id(0); + RID aux3 = uniforms.write[3].get_id(0); + uniforms.write[2].set_id(0, aux3); + uniforms.write[3].set_id(0, aux2); + rt->sdf_buffer_process_uniform_sets[1] = RD::get_singleton()->uniform_set_create(uniforms, rt_sdf.shader.version_get_shader(rt_sdf.shader_version, 0), 0); + } +} + +void TextureStorage::_render_target_clear_sdf(RenderTarget *rt) { + if (rt->sdf_buffer_read.is_valid()) { + RD::get_singleton()->free(rt->sdf_buffer_read); + rt->sdf_buffer_read = RID(); + } + if (rt->sdf_buffer_write_fb.is_valid()) { + RD::get_singleton()->free(rt->sdf_buffer_write); + RD::get_singleton()->free(rt->sdf_buffer_process[0]); + RD::get_singleton()->free(rt->sdf_buffer_process[1]); + rt->sdf_buffer_write = RID(); + rt->sdf_buffer_write_fb = RID(); + rt->sdf_buffer_process[0] = RID(); + rt->sdf_buffer_process[1] = RID(); + rt->sdf_buffer_process_uniform_sets[0] = RID(); + rt->sdf_buffer_process_uniform_sets[1] = RID(); + } +} + +RID TextureStorage::render_target_get_sdf_framebuffer(RID p_render_target) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, RID()); + + if (rt->sdf_buffer_write_fb.is_null()) { + _render_target_allocate_sdf(rt); + } + + return rt->sdf_buffer_write_fb; +} +void TextureStorage::render_target_sdf_process(RID p_render_target) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND(!rt); + ERR_FAIL_COND(rt->sdf_buffer_write_fb.is_null()); + + RenderTargetSDF::PushConstant push_constant; + + Rect2i r = _render_target_get_sdf_rect(rt); + + push_constant.size[0] = r.size.width; + push_constant.size[1] = r.size.height; + push_constant.stride = 0; + push_constant.shift = 0; + push_constant.base_size[0] = r.size.width; + push_constant.base_size[1] = r.size.height; + + bool shrink = false; + + switch (rt->sdf_scale) { + case RS::VIEWPORT_SDF_SCALE_50_PERCENT: { + push_constant.size[0] >>= 1; + push_constant.size[1] >>= 1; + push_constant.shift = 1; + shrink = true; + } break; + case RS::VIEWPORT_SDF_SCALE_25_PERCENT: { + push_constant.size[0] >>= 2; + push_constant.size[1] >>= 2; + push_constant.shift = 2; + shrink = true; + } break; + default: { + }; + } + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + + /* Load */ + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, rt_sdf.pipelines[shrink ? RenderTargetSDF::SHADER_LOAD_SHRINK : RenderTargetSDF::SHADER_LOAD]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rt->sdf_buffer_process_uniform_sets[1], 0); //fill [0] + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(RenderTargetSDF::PushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, push_constant.size[0], push_constant.size[1], 1); + + /* Process */ + + int stride = nearest_power_of_2_templated(MAX(push_constant.size[0], push_constant.size[1]) / 2); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, rt_sdf.pipelines[RenderTargetSDF::SHADER_PROCESS]); + + RD::get_singleton()->compute_list_add_barrier(compute_list); + bool swap = false; + + //jumpflood + while (stride > 0) { + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rt->sdf_buffer_process_uniform_sets[swap ? 1 : 0], 0); + push_constant.stride = stride; + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(RenderTargetSDF::PushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, push_constant.size[0], push_constant.size[1], 1); + stride /= 2; + swap = !swap; + RD::get_singleton()->compute_list_add_barrier(compute_list); + } + + /* Store */ + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, rt_sdf.pipelines[shrink ? RenderTargetSDF::SHADER_STORE_SHRINK : RenderTargetSDF::SHADER_STORE]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rt->sdf_buffer_process_uniform_sets[swap ? 1 : 0], 0); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(RenderTargetSDF::PushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, push_constant.size[0], push_constant.size[1], 1); + + RD::get_singleton()->compute_list_end(); +} + +void TextureStorage::render_target_copy_to_back_buffer(RID p_render_target, const Rect2i &p_region, bool p_gen_mipmaps) { + CopyEffects *copy_effects = CopyEffects::get_singleton(); + ERR_FAIL_NULL(copy_effects); + + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND(!rt); + if (!rt->backbuffer.is_valid()) { + _create_render_target_backbuffer(rt); + } + + Rect2i region; + if (p_region == Rect2i()) { + region.size = rt->size; + } else { + region = Rect2i(Size2i(), rt->size).intersection(p_region); + if (region.size == Size2i()) { + return; //nothing to do + } + } + + // TODO figure out stereo support here + + //single texture copy for backbuffer + //RD::get_singleton()->texture_copy(rt->color, rt->backbuffer_mipmap0, Vector3(region.position.x, region.position.y, 0), Vector3(region.position.x, region.position.y, 0), Vector3(region.size.x, region.size.y, 1), 0, 0, 0, 0, true); + copy_effects->copy_to_rect(rt->color, rt->backbuffer_mipmap0, region, false, false, false, true, true); + + if (!p_gen_mipmaps) { + return; + } + RD::get_singleton()->draw_command_begin_label("Gaussian Blur Mipmaps"); + //then mipmap blur + RID prev_texture = rt->color; //use color, not backbuffer, as bb has mipmaps. + + for (int i = 0; i < rt->backbuffer_mipmaps.size(); i++) { + region.position.x >>= 1; + region.position.y >>= 1; + region.size.x = MAX(1, region.size.x >> 1); + region.size.y = MAX(1, region.size.y >> 1); + + RID mipmap = rt->backbuffer_mipmaps[i]; + copy_effects->gaussian_blur(prev_texture, mipmap, region, true); + prev_texture = mipmap; + } + RD::get_singleton()->draw_command_end_label(); +} + +void TextureStorage::render_target_clear_back_buffer(RID p_render_target, const Rect2i &p_region, const Color &p_color) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND(!rt); + + CopyEffects *copy_effects = CopyEffects::get_singleton(); + ERR_FAIL_NULL(copy_effects); + + if (!rt->backbuffer.is_valid()) { + _create_render_target_backbuffer(rt); + } + + Rect2i region; + if (p_region == Rect2i()) { + region.size = rt->size; + } else { + region = Rect2i(Size2i(), rt->size).intersection(p_region); + if (region.size == Size2i()) { + return; //nothing to do + } + } + + //single texture copy for backbuffer + copy_effects->set_color(rt->backbuffer_mipmap0, p_color, region, true); +} + +void TextureStorage::render_target_gen_back_buffer_mipmaps(RID p_render_target, const Rect2i &p_region) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND(!rt); + + CopyEffects *copy_effects = CopyEffects::get_singleton(); + ERR_FAIL_NULL(copy_effects); + + if (!rt->backbuffer.is_valid()) { + _create_render_target_backbuffer(rt); + } + + Rect2i region; + if (p_region == Rect2i()) { + region.size = rt->size; + } else { + region = Rect2i(Size2i(), rt->size).intersection(p_region); + if (region.size == Size2i()) { + return; //nothing to do + } + } + RD::get_singleton()->draw_command_begin_label("Gaussian Blur Mipmaps2"); + //then mipmap blur + RID prev_texture = rt->backbuffer_mipmap0; + + for (int i = 0; i < rt->backbuffer_mipmaps.size(); i++) { + region.position.x >>= 1; + region.position.y >>= 1; + region.size.x = MAX(1, region.size.x >> 1); + region.size.y = MAX(1, region.size.y >> 1); + + RID mipmap = rt->backbuffer_mipmaps[i]; + copy_effects->gaussian_blur(prev_texture, mipmap, region, true); + prev_texture = mipmap; + } + RD::get_singleton()->draw_command_end_label(); +} + +RID TextureStorage::render_target_get_framebuffer_uniform_set(RID p_render_target) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, RID()); + return rt->framebuffer_uniform_set; +} +RID TextureStorage::render_target_get_backbuffer_uniform_set(RID p_render_target) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, RID()); + return rt->backbuffer_uniform_set; +} + +void TextureStorage::render_target_set_framebuffer_uniform_set(RID p_render_target, RID p_uniform_set) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND(!rt); + rt->framebuffer_uniform_set = p_uniform_set; +} + +void TextureStorage::render_target_set_backbuffer_uniform_set(RID p_render_target, RID p_uniform_set) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND(!rt); + rt->backbuffer_uniform_set = p_uniform_set; +} + +void TextureStorage::render_target_set_vrs_mode(RID p_render_target, RS::ViewportVRSMode p_mode) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND(!rt); + + rt->vrs_mode = p_mode; +} + +RS::ViewportVRSMode TextureStorage::render_target_get_vrs_mode(RID p_render_target) const { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, RS::VIEWPORT_VRS_DISABLED); + + return rt->vrs_mode; +} + +void TextureStorage::render_target_set_vrs_texture(RID p_render_target, RID p_texture) { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND(!rt); + + rt->vrs_texture = p_texture; +} + +RID TextureStorage::render_target_get_vrs_texture(RID p_render_target) const { + RenderTarget *rt = render_target_owner.get_or_null(p_render_target); + ERR_FAIL_COND_V(!rt, RID()); + + return rt->vrs_texture; +} diff --git a/servers/rendering/renderer_rd/storage_rd/texture_storage.h b/servers/rendering/renderer_rd/storage_rd/texture_storage.h new file mode 100644 index 0000000000..00b4e50737 --- /dev/null +++ b/servers/rendering/renderer_rd/storage_rd/texture_storage.h @@ -0,0 +1,747 @@ +/*************************************************************************/ +/* texture_storage.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef TEXTURE_STORAGE_RD_H +#define TEXTURE_STORAGE_RD_H + +#include "core/templates/local_vector.h" +#include "core/templates/paged_array.h" +#include "core/templates/rid_owner.h" +#include "servers/rendering/renderer_rd/shaders/canvas_sdf.glsl.gen.h" +#include "servers/rendering/renderer_rd/storage_rd/forward_id_storage.h" +#include "servers/rendering/rendering_server_default.h" +#include "servers/rendering/storage/texture_storage.h" +#include "servers/rendering/storage/utilities.h" + +namespace RendererRD { + +class LightStorage; +class MaterialStorage; + +class TextureStorage : public RendererTextureStorage { +public: + enum DefaultRDTexture { + DEFAULT_RD_TEXTURE_WHITE, + DEFAULT_RD_TEXTURE_BLACK, + DEFAULT_RD_TEXTURE_TRANSPARENT, + DEFAULT_RD_TEXTURE_NORMAL, + DEFAULT_RD_TEXTURE_ANISO, + DEFAULT_RD_TEXTURE_DEPTH, + DEFAULT_RD_TEXTURE_MULTIMESH_BUFFER, + DEFAULT_RD_TEXTURE_CUBEMAP_BLACK, + DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_BLACK, + DEFAULT_RD_TEXTURE_CUBEMAP_WHITE, + DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_WHITE, + DEFAULT_RD_TEXTURE_3D_WHITE, + DEFAULT_RD_TEXTURE_3D_BLACK, + DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE, + DEFAULT_RD_TEXTURE_2D_UINT, + DEFAULT_RD_TEXTURE_VRS, + DEFAULT_RD_TEXTURE_MAX + }; + + enum TextureType { + TYPE_2D, + TYPE_LAYERED, + TYPE_3D + }; + +private: + friend class LightStorage; + friend class MaterialStorage; + + static TextureStorage *singleton; + + RID default_rd_textures[DEFAULT_RD_TEXTURE_MAX]; + + /* Canvas Texture API */ + + class CanvasTexture { + public: + RID diffuse; + RID normal_map; + RID specular; + Color specular_color = Color(1, 1, 1, 1); + float shininess = 1.0; + + RS::CanvasItemTextureFilter texture_filter = RS::CANVAS_ITEM_TEXTURE_FILTER_DEFAULT; + RS::CanvasItemTextureRepeat texture_repeat = RS::CANVAS_ITEM_TEXTURE_REPEAT_DEFAULT; + RID uniform_sets[RS::CANVAS_ITEM_TEXTURE_FILTER_MAX][RS::CANVAS_ITEM_TEXTURE_REPEAT_MAX]; + + Size2i size_cache = Size2i(1, 1); + bool use_normal_cache = false; + bool use_specular_cache = false; + bool cleared_cache = true; + + void clear_sets(); + ~CanvasTexture(); + }; + + RID_Owner<CanvasTexture, true> canvas_texture_owner; + + /* Texture API */ + + class Texture { + public: + TextureType type; + RS::TextureLayeredType layered_type = RS::TEXTURE_LAYERED_2D_ARRAY; + + RenderingDevice::TextureType rd_type; + RID rd_texture; + RID rd_texture_srgb; + RenderingDevice::DataFormat rd_format; + RenderingDevice::DataFormat rd_format_srgb; + + RD::TextureView rd_view; + + Image::Format format; + Image::Format validated_format; + + int width; + int height; + int depth; + int layers; + int mipmaps; + + int height_2d; + int width_2d; + + struct BufferSlice3D { + Size2i size; + uint32_t offset = 0; + uint32_t buffer_size = 0; + }; + Vector<BufferSlice3D> buffer_slices_3d; + uint32_t buffer_size_3d = 0; + + bool is_render_target; + bool is_proxy; + + Ref<Image> image_cache_2d; + String path; + + RID proxy_to; + Vector<RID> proxies; + + HashSet<RID> lightmap_users; + + RS::TextureDetectCallback detect_3d_callback = nullptr; + void *detect_3d_callback_ud = nullptr; + + RS::TextureDetectCallback detect_normal_callback = nullptr; + void *detect_normal_callback_ud = nullptr; + + RS::TextureDetectRoughnessCallback detect_roughness_callback = nullptr; + void *detect_roughness_callback_ud = nullptr; + + CanvasTexture *canvas_texture = nullptr; + + void cleanup(); + }; + + //textures can be created from threads, so this RID_Owner is thread safe + mutable RID_Owner<Texture, true> texture_owner; + Texture *get_texture(RID p_rid) { return texture_owner.get_or_null(p_rid); }; + + struct TextureToRDFormat { + RD::DataFormat format; + RD::DataFormat format_srgb; + RD::TextureSwizzle swizzle_r; + RD::TextureSwizzle swizzle_g; + RD::TextureSwizzle swizzle_b; + RD::TextureSwizzle swizzle_a; + TextureToRDFormat() { + format = RD::DATA_FORMAT_MAX; + format_srgb = RD::DATA_FORMAT_MAX; + swizzle_r = RD::TEXTURE_SWIZZLE_R; + swizzle_g = RD::TEXTURE_SWIZZLE_G; + swizzle_b = RD::TEXTURE_SWIZZLE_B; + swizzle_a = RD::TEXTURE_SWIZZLE_A; + } + }; + + Ref<Image> _validate_texture_format(const Ref<Image> &p_image, TextureToRDFormat &r_format); + void _texture_2d_update(RID p_texture, const Ref<Image> &p_image, int p_layer = 0, bool p_immediate = false); + + /* DECAL API */ + + struct DecalAtlas { + struct Texture { + int panorama_to_dp_users; + int users; + Rect2 uv_rect; + }; + + struct SortItem { + RID texture; + Size2i pixel_size; + Size2i size; + Point2i pos; + + bool operator<(const SortItem &p_item) const { + //sort larger to smaller + if (size.height == p_item.size.height) { + return size.width > p_item.size.width; + } else { + return size.height > p_item.size.height; + } + } + }; + + HashMap<RID, Texture> textures; + bool dirty = true; + int mipmaps = 5; + + RID texture; + RID texture_srgb; + struct MipMap { + RID fb; + RID texture; + Size2i size; + }; + Vector<MipMap> texture_mipmaps; + + Size2i size; + } decal_atlas; + + struct Decal { + Vector3 extents = Vector3(1, 1, 1); + RID textures[RS::DECAL_TEXTURE_MAX]; + float emission_energy = 1.0; + float albedo_mix = 1.0; + Color modulate = Color(1, 1, 1, 1); + uint32_t cull_mask = (1 << 20) - 1; + float upper_fade = 0.3; + float lower_fade = 0.3; + bool distance_fade = false; + float distance_fade_begin = 40.0; + float distance_fade_length = 10.0; + float normal_fade = 0.0; + + Dependency dependency; + }; + + mutable RID_Owner<Decal, true> decal_owner; + + /* DECAL INSTANCE */ + + struct DecalInstance { + RID decal; + Transform3D transform; + uint32_t cull_mask = 0; + RendererRD::ForwardID forward_id = -1; + }; + + mutable RID_Owner<DecalInstance> decal_instance_owner; + + /* DECAL DATA (UBO) */ + + struct DecalData { + float xform[16]; + float inv_extents[3]; + float albedo_mix; + float albedo_rect[4]; + float normal_rect[4]; + float orm_rect[4]; + float emission_rect[4]; + float modulate[4]; + float emission_energy; + uint32_t mask; + float upper_fade; + float lower_fade; + float normal_xform[12]; + float normal[3]; + float normal_fade; + }; + + struct DecalInstanceSort { + float depth; + DecalInstance *decal_instance; + Decal *decal; + bool operator<(const DecalInstanceSort &p_sort) const { + return depth < p_sort.depth; + } + }; + + uint32_t max_decals = 0; + uint32_t decal_count = 0; + DecalData *decals = nullptr; + DecalInstanceSort *decal_sort = nullptr; + RID decal_buffer; + + /* RENDER TARGET API */ + + struct RenderTarget { + Size2i size; + uint32_t view_count; + RID color; + Vector<RID> color_slices; + RID color_multisample; // Needed when MSAA is enabled. + + RS::ViewportMSAA msaa = RS::VIEWPORT_MSAA_DISABLED; + + //used for retrieving from CPU + RD::DataFormat color_format = RD::DATA_FORMAT_R4G4_UNORM_PACK8; + RD::DataFormat color_format_srgb = RD::DATA_FORMAT_R4G4_UNORM_PACK8; + Image::Format image_format = Image::FORMAT_L8; + + bool is_transparent = false; + + bool sdf_enabled = false; + + RID backbuffer; //used for effects + RID backbuffer_fb; + RID backbuffer_mipmap0; + + Vector<RID> backbuffer_mipmaps; + + RID framebuffer_uniform_set; + RID backbuffer_uniform_set; + + RID sdf_buffer_write; + RID sdf_buffer_write_fb; + RID sdf_buffer_process[2]; + RID sdf_buffer_read; + RID sdf_buffer_process_uniform_sets[2]; + RS::ViewportSDFOversize sdf_oversize = RS::VIEWPORT_SDF_OVERSIZE_120_PERCENT; + RS::ViewportSDFScale sdf_scale = RS::VIEWPORT_SDF_SCALE_50_PERCENT; + Size2i process_size; + + // VRS + RS::ViewportVRSMode vrs_mode = RS::VIEWPORT_VRS_DISABLED; + RID vrs_texture; + + // overridden textures + struct RTOverridden { + RID color; + RID depth; + RID velocity; + + // In a multiview scenario, which is the most likely where we + // override our destination textures, we need to obtain slices + // for each layer of these textures. + // These are likely changing every frame as we loop through + // texture chains hence we add a cache to manage these slices. + // For this we define a key using the RID of the texture and + // the layer for which we create a slice. + struct SliceKey { + RID rid; + uint32_t layer = 0; + + bool operator==(const SliceKey &p_val) const { + return (rid == p_val.rid) && (layer == p_val.layer); + } + + static uint32_t hash(const SliceKey &p_val) { + uint32_t h = hash_one_uint64(p_val.rid.get_id()); + h = hash_murmur3_one_32(p_val.layer, h); + return hash_fmix32(h); + } + + SliceKey() {} + SliceKey(RID p_rid, uint32_t p_layer) { + rid = p_rid; + layer = p_layer; + } + }; + + mutable HashMap<SliceKey, RID, SliceKey> cached_slices; + } overridden; + + //texture generated for this owner (nor RD). + RID texture; + bool was_used; + + //clear request + bool clear_requested; + Color clear_color; + + RID get_framebuffer(); + }; + + mutable RID_Owner<RenderTarget> render_target_owner; + RenderTarget *get_render_target(RID p_rid) const { return render_target_owner.get_or_null(p_rid); }; + + void _clear_render_target(RenderTarget *rt); + void _update_render_target(RenderTarget *rt); + void _create_render_target_backbuffer(RenderTarget *rt); + void _render_target_allocate_sdf(RenderTarget *rt); + void _render_target_clear_sdf(RenderTarget *rt); + Rect2i _render_target_get_sdf_rect(const RenderTarget *rt) const; + + struct RenderTargetSDF { + enum { + SHADER_LOAD, + SHADER_LOAD_SHRINK, + SHADER_PROCESS, + SHADER_PROCESS_OPTIMIZED, + SHADER_STORE, + SHADER_STORE_SHRINK, + SHADER_MAX + }; + + struct PushConstant { + int32_t size[2]; + int32_t stride; + int32_t shift; + int32_t base_size[2]; + int32_t pad[2]; + }; + + CanvasSdfShaderRD shader; + RID shader_version; + RID pipelines[SHADER_MAX]; + } rt_sdf; + +public: + static TextureStorage *get_singleton(); + + _FORCE_INLINE_ RID texture_rd_get_default(DefaultRDTexture p_texture) { + return default_rd_textures[p_texture]; + } + + TextureStorage(); + virtual ~TextureStorage(); + + bool free(RID p_rid); + + /* Canvas Texture API */ + + bool owns_canvas_texture(RID p_rid) { return canvas_texture_owner.owns(p_rid); }; + + virtual RID canvas_texture_allocate() override; + virtual void canvas_texture_initialize(RID p_rid) override; + virtual void canvas_texture_free(RID p_rid) override; + + virtual void canvas_texture_set_channel(RID p_canvas_texture, RS::CanvasTextureChannel p_channel, RID p_texture) override; + virtual void canvas_texture_set_shading_parameters(RID p_canvas_texture, const Color &p_base_color, float p_shininess) override; + + virtual void canvas_texture_set_texture_filter(RID p_item, RS::CanvasItemTextureFilter p_filter) override; + virtual void canvas_texture_set_texture_repeat(RID p_item, RS::CanvasItemTextureRepeat p_repeat) override; + + bool canvas_texture_get_uniform_set(RID p_texture, RS::CanvasItemTextureFilter p_base_filter, RS::CanvasItemTextureRepeat p_base_repeat, RID p_base_shader, int p_base_set, RID &r_uniform_set, Size2i &r_size, Color &r_specular_shininess, bool &r_use_normal, bool &r_use_specular); + + /* Texture API */ + + bool owns_texture(RID p_rid) const { return texture_owner.owns(p_rid); }; + + virtual bool can_create_resources_async() const override; + + virtual RID texture_allocate() override; + virtual void texture_free(RID p_rid) override; + + virtual void texture_2d_initialize(RID p_texture, const Ref<Image> &p_image) override; + virtual void texture_2d_layered_initialize(RID p_texture, const Vector<Ref<Image>> &p_layers, RS::TextureLayeredType p_layered_type) override; + virtual void texture_3d_initialize(RID p_texture, Image::Format, int p_width, int p_height, int p_depth, bool p_mipmaps, const Vector<Ref<Image>> &p_data) override; + virtual void texture_proxy_initialize(RID p_texture, RID p_base) override; //all slices, then all the mipmaps, must be coherent + + virtual void texture_2d_update(RID p_texture, const Ref<Image> &p_image, int p_layer = 0) override; + virtual void texture_3d_update(RID p_texture, const Vector<Ref<Image>> &p_data) override; + virtual void texture_proxy_update(RID p_proxy, RID p_base) override; + + //these two APIs can be used together or in combination with the others. + virtual void texture_2d_placeholder_initialize(RID p_texture) override; + virtual void texture_2d_layered_placeholder_initialize(RID p_texture, RenderingServer::TextureLayeredType p_layered_type) override; + virtual void texture_3d_placeholder_initialize(RID p_texture) override; + + virtual Ref<Image> texture_2d_get(RID p_texture) const override; + virtual Ref<Image> texture_2d_layer_get(RID p_texture, int p_layer) const override; + virtual Vector<Ref<Image>> texture_3d_get(RID p_texture) const override; + + virtual void texture_replace(RID p_texture, RID p_by_texture) override; + virtual void texture_set_size_override(RID p_texture, int p_width, int p_height) override; + + virtual void texture_set_path(RID p_texture, const String &p_path) override; + virtual String texture_get_path(RID p_texture) const override; + + virtual void texture_set_detect_3d_callback(RID p_texture, RS::TextureDetectCallback p_callback, void *p_userdata) override; + virtual void texture_set_detect_normal_callback(RID p_texture, RS::TextureDetectCallback p_callback, void *p_userdata) override; + virtual void texture_set_detect_roughness_callback(RID p_texture, RS::TextureDetectRoughnessCallback p_callback, void *p_userdata) override; + + virtual void texture_debug_usage(List<RS::TextureInfo> *r_info) override; + + virtual void texture_set_force_redraw_if_visible(RID p_texture, bool p_enable) override; + + virtual Size2 texture_size_with_proxy(RID p_proxy) override; + + //internal usage + _FORCE_INLINE_ TextureType texture_get_type(RID p_texture) { + RendererRD::TextureStorage::Texture *tex = texture_owner.get_or_null(p_texture); + if (tex == nullptr) { + return TYPE_2D; + } + + return tex->type; + } + + _FORCE_INLINE_ int texture_get_layers(RID p_texture) { + RendererRD::TextureStorage::Texture *tex = texture_owner.get_or_null(p_texture); + if (tex == nullptr) { + return 1; + } + + return tex->layers; + } + + _FORCE_INLINE_ RID texture_get_rd_texture(RID p_texture, bool p_srgb = false) { + if (p_texture.is_null()) { + return RID(); + } + RendererRD::TextureStorage::Texture *tex = texture_owner.get_or_null(p_texture); + + if (!tex) { + return RID(); + } + return (p_srgb && tex->rd_texture_srgb.is_valid()) ? tex->rd_texture_srgb : tex->rd_texture; + } + + _FORCE_INLINE_ Size2i texture_2d_get_size(RID p_texture) { + if (p_texture.is_null()) { + return Size2i(); + } + RendererRD::TextureStorage::Texture *tex = texture_owner.get_or_null(p_texture); + + if (!tex) { + return Size2i(); + } + return Size2i(tex->width_2d, tex->height_2d); + } + + /* DECAL API */ + + void update_decal_atlas(); + + bool owns_decal(RID p_rid) const { return decal_owner.owns(p_rid); }; + + RID decal_atlas_get_texture() const; + RID decal_atlas_get_texture_srgb() const; + _FORCE_INLINE_ Rect2 decal_atlas_get_texture_rect(RID p_texture) { + DecalAtlas::Texture *t = decal_atlas.textures.getptr(p_texture); + if (!t) { + return Rect2(); + } + + return t->uv_rect; + } + + virtual RID decal_allocate() override; + virtual void decal_initialize(RID p_decal) override; + virtual void decal_free(RID p_rid) override; + + virtual void decal_set_extents(RID p_decal, const Vector3 &p_extents) override; + virtual void decal_set_texture(RID p_decal, RS::DecalTexture p_type, RID p_texture) override; + virtual void decal_set_emission_energy(RID p_decal, float p_energy) override; + virtual void decal_set_albedo_mix(RID p_decal, float p_mix) override; + virtual void decal_set_modulate(RID p_decal, const Color &p_modulate) override; + virtual void decal_set_cull_mask(RID p_decal, uint32_t p_layers) override; + virtual void decal_set_distance_fade(RID p_decal, bool p_enabled, float p_begin, float p_length) override; + virtual void decal_set_fade(RID p_decal, float p_above, float p_below) override; + virtual void decal_set_normal_fade(RID p_decal, float p_fade) override; + + void decal_atlas_mark_dirty_on_texture(RID p_texture); + void decal_atlas_remove_texture(RID p_texture); + + virtual void texture_add_to_decal_atlas(RID p_texture, bool p_panorama_to_dp = false) override; + virtual void texture_remove_from_decal_atlas(RID p_texture, bool p_panorama_to_dp = false) override; + + _FORCE_INLINE_ Vector3 decal_get_extents(RID p_decal) { + const Decal *decal = decal_owner.get_or_null(p_decal); + return decal->extents; + } + + _FORCE_INLINE_ RID decal_get_texture(RID p_decal, RS::DecalTexture p_texture) { + const Decal *decal = decal_owner.get_or_null(p_decal); + return decal->textures[p_texture]; + } + + _FORCE_INLINE_ Color decal_get_modulate(RID p_decal) { + const Decal *decal = decal_owner.get_or_null(p_decal); + return decal->modulate; + } + + _FORCE_INLINE_ float decal_get_emission_energy(RID p_decal) { + const Decal *decal = decal_owner.get_or_null(p_decal); + return decal->emission_energy; + } + + _FORCE_INLINE_ float decal_get_albedo_mix(RID p_decal) { + const Decal *decal = decal_owner.get_or_null(p_decal); + return decal->albedo_mix; + } + + _FORCE_INLINE_ uint32_t decal_get_cull_mask(RID p_decal) { + const Decal *decal = decal_owner.get_or_null(p_decal); + return decal->cull_mask; + } + + _FORCE_INLINE_ float decal_get_upper_fade(RID p_decal) { + const Decal *decal = decal_owner.get_or_null(p_decal); + return decal->upper_fade; + } + + _FORCE_INLINE_ float decal_get_lower_fade(RID p_decal) { + const Decal *decal = decal_owner.get_or_null(p_decal); + return decal->lower_fade; + } + + _FORCE_INLINE_ float decal_get_normal_fade(RID p_decal) { + const Decal *decal = decal_owner.get_or_null(p_decal); + return decal->normal_fade; + } + + _FORCE_INLINE_ bool decal_is_distance_fade_enabled(RID p_decal) { + const Decal *decal = decal_owner.get_or_null(p_decal); + return decal->distance_fade; + } + + _FORCE_INLINE_ float decal_get_distance_fade_begin(RID p_decal) { + const Decal *decal = decal_owner.get_or_null(p_decal); + return decal->distance_fade_begin; + } + + _FORCE_INLINE_ float decal_get_distance_fade_length(RID p_decal) { + const Decal *decal = decal_owner.get_or_null(p_decal); + return decal->distance_fade_length; + } + + virtual AABB decal_get_aabb(RID p_decal) const override; + Dependency *decal_get_dependency(RID p_decal); + + /* DECAL INSTANCE API */ + + bool owns_decal_instance(RID p_rid) const { return decal_instance_owner.owns(p_rid); } + + virtual RID decal_instance_create(RID p_decal) override; + virtual void decal_instance_free(RID p_decal_instance) override; + virtual void decal_instance_set_transform(RID p_decal_instance, const Transform3D &p_transform) override; + + _FORCE_INLINE_ RID decal_instance_get_base(RID p_decal_instance) const { + DecalInstance *di = decal_instance_owner.get_or_null(p_decal_instance); + return di->decal; + } + + _FORCE_INLINE_ RendererRD::ForwardID decal_instance_get_forward_id(RID p_decal_instance) const { + DecalInstance *di = decal_instance_owner.get_or_null(p_decal_instance); + return di->forward_id; + } + + _FORCE_INLINE_ Transform3D decal_instance_get_transform(RID p_decal_instance) const { + DecalInstance *di = decal_instance_owner.get_or_null(p_decal_instance); + return di->transform; + } + + _FORCE_INLINE_ ForwardID decal_instance_get_forward_id(RID p_decal_instance) { + DecalInstance *di = decal_instance_owner.get_or_null(p_decal_instance); + return di->forward_id; + } + + _FORCE_INLINE_ void decal_instance_set_cullmask(RID p_decal_instance, uint32_t p_cull_mask) const { + DecalInstance *di = decal_instance_owner.get_or_null(p_decal_instance); + di->cull_mask = p_cull_mask; + } + + /* DECAL DATA API */ + + void free_decal_data(); + void set_max_decals(const uint32_t p_max_decals); + RID get_decal_buffer() { return decal_buffer; } + void update_decal_buffer(const PagedArray<RID> &p_decals, const Transform3D &p_camera_inverse_xform); + + /* RENDER TARGET API */ + + bool owns_render_target(RID p_rid) const { return render_target_owner.owns(p_rid); }; + + virtual RID render_target_create() override; + virtual void render_target_free(RID p_rid) override; + + virtual void render_target_set_position(RID p_render_target, int p_x, int p_y) override; + virtual Point2i render_target_get_position(RID p_render_target) const override; + virtual void render_target_set_size(RID p_render_target, int p_width, int p_height, uint32_t p_view_count) override; + virtual Size2i render_target_get_size(RID p_render_target) const override; + virtual void render_target_set_transparent(RID p_render_target, bool p_is_transparent) override; + virtual bool render_target_get_transparent(RID p_render_target) const override; + virtual void render_target_set_direct_to_screen(RID p_render_target, bool p_direct_to_screen) override; + virtual bool render_target_get_direct_to_screen(RID p_render_target) const override; + virtual bool render_target_was_used(RID p_render_target) const override; + virtual void render_target_set_as_unused(RID p_render_target) override; + virtual void render_target_set_msaa(RID p_render_target, RS::ViewportMSAA p_msaa) override; + virtual RS::ViewportMSAA render_target_get_msaa(RID p_render_target) const override; + + void render_target_copy_to_back_buffer(RID p_render_target, const Rect2i &p_region, bool p_gen_mipmaps); + void render_target_clear_back_buffer(RID p_render_target, const Rect2i &p_region, const Color &p_color); + void render_target_gen_back_buffer_mipmaps(RID p_render_target, const Rect2i &p_region); + RID render_target_get_back_buffer_uniform_set(RID p_render_target, RID p_base_shader); + + virtual void render_target_request_clear(RID p_render_target, const Color &p_clear_color) override; + virtual bool render_target_is_clear_requested(RID p_render_target) override; + virtual Color render_target_get_clear_request_color(RID p_render_target) override; + virtual void render_target_disable_clear_request(RID p_render_target) override; + virtual void render_target_do_clear_request(RID p_render_target) override; + + virtual void render_target_set_sdf_size_and_scale(RID p_render_target, RS::ViewportSDFOversize p_size, RS::ViewportSDFScale p_scale) override; + RID render_target_get_sdf_texture(RID p_render_target); + RID render_target_get_sdf_framebuffer(RID p_render_target); + void render_target_sdf_process(RID p_render_target); + virtual Rect2i render_target_get_sdf_rect(RID p_render_target) const override; + virtual void render_target_mark_sdf_enabled(RID p_render_target, bool p_enabled) override; + bool render_target_is_sdf_enabled(RID p_render_target) const; + + virtual void render_target_set_vrs_mode(RID p_render_target, RS::ViewportVRSMode p_mode) override; + virtual RS::ViewportVRSMode render_target_get_vrs_mode(RID p_render_target) const override; + virtual void render_target_set_vrs_texture(RID p_render_target, RID p_texture) override; + virtual RID render_target_get_vrs_texture(RID p_render_target) const override; + + virtual void render_target_set_override_color(RID p_render_target, RID p_texture) override; + virtual RID render_target_get_override_color(RID p_render_target) const override; + virtual void render_target_set_override_depth(RID p_render_target, RID p_texture) override; + virtual RID render_target_get_override_depth(RID p_render_target) const override; + RID render_target_get_override_depth_slice(RID p_render_target, const uint32_t p_layer) const; + virtual void render_target_set_override_velocity(RID p_render_target, RID p_texture) override; + virtual RID render_target_get_override_velocity(RID p_render_target) const override; + RID render_target_get_override_velocity_slice(RID p_render_target, const uint32_t p_layer) const; + + virtual RID render_target_get_texture(RID p_render_target) override; + + RID render_target_get_rd_framebuffer(RID p_render_target); + RID render_target_get_rd_texture(RID p_render_target); + RID render_target_get_rd_texture_slice(RID p_render_target, uint32_t p_layer); + RID render_target_get_rd_backbuffer(RID p_render_target); + RID render_target_get_rd_backbuffer_framebuffer(RID p_render_target); + + RID render_target_get_framebuffer_uniform_set(RID p_render_target); + RID render_target_get_backbuffer_uniform_set(RID p_render_target); + + void render_target_set_framebuffer_uniform_set(RID p_render_target, RID p_uniform_set); + void render_target_set_backbuffer_uniform_set(RID p_render_target, RID p_uniform_set); +}; + +} // namespace RendererRD + +#endif // TEXTURE_STORAGE_RD_H diff --git a/servers/rendering/renderer_rd/storage_rd/utilities.cpp b/servers/rendering/renderer_rd/storage_rd/utilities.cpp new file mode 100644 index 0000000000..4048c46d28 --- /dev/null +++ b/servers/rendering/renderer_rd/storage_rd/utilities.cpp @@ -0,0 +1,331 @@ +/*************************************************************************/ +/* utilities.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "utilities.h" +#include "../environment/fog.h" +#include "../environment/gi.h" +#include "light_storage.h" +#include "mesh_storage.h" +#include "particles_storage.h" +#include "texture_storage.h" + +using namespace RendererRD; + +Utilities *Utilities::singleton = nullptr; + +Utilities::Utilities() { + singleton = this; +} + +Utilities::~Utilities() { + singleton = nullptr; +} + +/* INSTANCES */ + +RS::InstanceType Utilities::get_base_type(RID p_rid) const { + if (RendererRD::MeshStorage::get_singleton()->owns_mesh(p_rid)) { + return RS::INSTANCE_MESH; + } + if (RendererRD::MeshStorage::get_singleton()->owns_multimesh(p_rid)) { + return RS::INSTANCE_MULTIMESH; + } + if (RendererRD::LightStorage::get_singleton()->owns_reflection_probe(p_rid)) { + return RS::INSTANCE_REFLECTION_PROBE; + } + if (RendererRD::TextureStorage::get_singleton()->owns_decal(p_rid)) { + return RS::INSTANCE_DECAL; + } + if (RendererRD::GI::get_singleton()->owns_voxel_gi(p_rid)) { + return RS::INSTANCE_VOXEL_GI; + } + if (RendererRD::LightStorage::get_singleton()->owns_light(p_rid)) { + return RS::INSTANCE_LIGHT; + } + if (RendererRD::LightStorage::get_singleton()->owns_lightmap(p_rid)) { + return RS::INSTANCE_LIGHTMAP; + } + if (RendererRD::ParticlesStorage::get_singleton()->owns_particles(p_rid)) { + return RS::INSTANCE_PARTICLES; + } + if (RendererRD::ParticlesStorage::get_singleton()->owns_particles_collision(p_rid)) { + return RS::INSTANCE_PARTICLES_COLLISION; + } + if (RendererRD::Fog::get_singleton()->owns_fog_volume(p_rid)) { + return RS::INSTANCE_FOG_VOLUME; + } + if (owns_visibility_notifier(p_rid)) { + return RS::INSTANCE_VISIBLITY_NOTIFIER; + } + + return RS::INSTANCE_NONE; +} + +bool Utilities::free(RID p_rid) { + if (RendererRD::LightStorage::get_singleton()->free(p_rid)) { + return true; + } else if (RendererRD::MaterialStorage::get_singleton()->free(p_rid)) { + return true; + } else if (RendererRD::MeshStorage::get_singleton()->free(p_rid)) { + return true; + } else if (RendererRD::ParticlesStorage::get_singleton()->free(p_rid)) { + return true; + } else if (RendererRD::TextureStorage::get_singleton()->free(p_rid)) { + return true; + } else if (RendererRD::GI::get_singleton()->owns_voxel_gi(p_rid)) { + RendererRD::GI::get_singleton()->voxel_gi_free(p_rid); + return true; + } else if (RendererRD::Fog::get_singleton()->owns_fog_volume(p_rid)) { + RendererRD::Fog::get_singleton()->fog_volume_free(p_rid); + return true; + } else if (owns_visibility_notifier(p_rid)) { + visibility_notifier_free(p_rid); + return true; + } else { + return false; + } +} + +/* DEPENDENCIES */ + +void Utilities::base_update_dependency(RID p_base, DependencyTracker *p_instance) { + if (MeshStorage::get_singleton()->owns_mesh(p_base)) { + Dependency *dependency = MeshStorage::get_singleton()->mesh_get_dependency(p_base); + p_instance->update_dependency(dependency); + } else if (MeshStorage::get_singleton()->owns_multimesh(p_base)) { + Dependency *dependency = MeshStorage::get_singleton()->multimesh_get_dependency(p_base); + p_instance->update_dependency(dependency); + + RID mesh = MeshStorage::get_singleton()->multimesh_get_mesh(p_base); + if (mesh.is_valid()) { + base_update_dependency(mesh, p_instance); + } + } else if (LightStorage::get_singleton()->owns_reflection_probe(p_base)) { + Dependency *dependency = LightStorage::get_singleton()->reflection_probe_get_dependency(p_base); + p_instance->update_dependency(dependency); + } else if (TextureStorage::get_singleton()->owns_decal(p_base)) { + Dependency *dependency = TextureStorage::get_singleton()->decal_get_dependency(p_base); + p_instance->update_dependency(dependency); + } else if (GI::get_singleton()->owns_voxel_gi(p_base)) { + Dependency *dependency = GI::get_singleton()->voxel_gi_get_dependency(p_base); + p_instance->update_dependency(dependency); + } else if (LightStorage::get_singleton()->owns_lightmap(p_base)) { + Dependency *dependency = LightStorage::get_singleton()->lightmap_get_dependency(p_base); + p_instance->update_dependency(dependency); + } else if (LightStorage::get_singleton()->owns_light(p_base)) { + Dependency *dependency = LightStorage::get_singleton()->light_get_dependency(p_base); + p_instance->update_dependency(dependency); + } else if (ParticlesStorage::get_singleton()->owns_particles(p_base)) { + Dependency *dependency = ParticlesStorage::get_singleton()->particles_get_dependency(p_base); + p_instance->update_dependency(dependency); + } else if (ParticlesStorage::get_singleton()->owns_particles_collision(p_base)) { + Dependency *dependency = ParticlesStorage::get_singleton()->particles_collision_get_dependency(p_base); + p_instance->update_dependency(dependency); + } else if (Fog::get_singleton()->owns_fog_volume(p_base)) { + Dependency *dependency = Fog::get_singleton()->fog_volume_get_dependency(p_base); + p_instance->update_dependency(dependency); + } else if (owns_visibility_notifier(p_base)) { + VisibilityNotifier *vn = get_visibility_notifier(p_base); + p_instance->update_dependency(&vn->dependency); + } +} + +/* VISIBILITY NOTIFIER */ + +RID Utilities::visibility_notifier_allocate() { + return visibility_notifier_owner.allocate_rid(); +} + +void Utilities::visibility_notifier_initialize(RID p_notifier) { + visibility_notifier_owner.initialize_rid(p_notifier, VisibilityNotifier()); +} + +void Utilities::visibility_notifier_free(RID p_notifier) { + VisibilityNotifier *vn = visibility_notifier_owner.get_or_null(p_notifier); + vn->dependency.deleted_notify(p_notifier); + visibility_notifier_owner.free(p_notifier); +} + +void Utilities::visibility_notifier_set_aabb(RID p_notifier, const AABB &p_aabb) { + VisibilityNotifier *vn = visibility_notifier_owner.get_or_null(p_notifier); + ERR_FAIL_COND(!vn); + vn->aabb = p_aabb; + vn->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); +} + +void Utilities::visibility_notifier_set_callbacks(RID p_notifier, const Callable &p_enter_callbable, const Callable &p_exit_callable) { + VisibilityNotifier *vn = visibility_notifier_owner.get_or_null(p_notifier); + ERR_FAIL_COND(!vn); + vn->enter_callback = p_enter_callbable; + vn->exit_callback = p_exit_callable; +} + +AABB Utilities::visibility_notifier_get_aabb(RID p_notifier) const { + const VisibilityNotifier *vn = visibility_notifier_owner.get_or_null(p_notifier); + ERR_FAIL_COND_V(!vn, AABB()); + return vn->aabb; +} + +void Utilities::visibility_notifier_call(RID p_notifier, bool p_enter, bool p_deferred) { + VisibilityNotifier *vn = visibility_notifier_owner.get_or_null(p_notifier); + ERR_FAIL_COND(!vn); + + if (p_enter) { + if (!vn->enter_callback.is_null()) { + if (p_deferred) { + vn->enter_callback.call_deferredp(nullptr, 0); + } else { + Variant r; + Callable::CallError ce; + vn->enter_callback.callp(nullptr, 0, r, ce); + } + } + } else { + if (!vn->exit_callback.is_null()) { + if (p_deferred) { + vn->exit_callback.call_deferredp(nullptr, 0); + } else { + Variant r; + Callable::CallError ce; + vn->exit_callback.callp(nullptr, 0, r, ce); + } + } + } +} + +/* TIMING */ + +void Utilities::capture_timestamps_begin() { + RD::get_singleton()->capture_timestamp("Frame Begin"); +} + +void Utilities::capture_timestamp(const String &p_name) { + RD::get_singleton()->capture_timestamp(p_name); +} + +uint32_t Utilities::get_captured_timestamps_count() const { + return RD::get_singleton()->get_captured_timestamps_count(); +} + +uint64_t Utilities::get_captured_timestamps_frame() const { + return RD::get_singleton()->get_captured_timestamps_frame(); +} + +uint64_t Utilities::get_captured_timestamp_gpu_time(uint32_t p_index) const { + return RD::get_singleton()->get_captured_timestamp_gpu_time(p_index); +} + +uint64_t Utilities::get_captured_timestamp_cpu_time(uint32_t p_index) const { + return RD::get_singleton()->get_captured_timestamp_cpu_time(p_index); +} + +String Utilities::get_captured_timestamp_name(uint32_t p_index) const { + return RD::get_singleton()->get_captured_timestamp_name(p_index); +} + +/* MISC */ + +void Utilities::update_dirty_resources() { + MaterialStorage::get_singleton()->_update_global_shader_uniforms(); //must do before materials, so it can queue them for update + MaterialStorage::get_singleton()->_update_queued_materials(); + MeshStorage::get_singleton()->_update_dirty_multimeshes(); + MeshStorage::get_singleton()->_update_dirty_skeletons(); + TextureStorage::get_singleton()->update_decal_atlas(); +} + +bool Utilities::has_os_feature(const String &p_feature) const { + if (!RD::get_singleton()) { + return false; + } + + if (p_feature == "rgtc" && RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_BC5_UNORM_BLOCK, RD::TEXTURE_USAGE_SAMPLING_BIT)) { + return true; + } + +#if !defined(ANDROID_ENABLED) && !defined(IOS_ENABLED) + // Some Android devices report support for S3TC but we don't expect that and don't export the textures. + // This could be fixed but so few devices support it that it doesn't seem useful (and makes bigger APKs). + // For good measure we do the same hack for iOS, just in case. + if (p_feature == "s3tc" && RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_BC1_RGB_UNORM_BLOCK, RD::TEXTURE_USAGE_SAMPLING_BIT)) { + return true; + } +#endif + + if (p_feature == "bptc" && RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_BC7_UNORM_BLOCK, RD::TEXTURE_USAGE_SAMPLING_BIT)) { + return true; + } + + if ((p_feature == "etc" || p_feature == "etc2") && RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_ETC2_R8G8B8_UNORM_BLOCK, RD::TEXTURE_USAGE_SAMPLING_BIT)) { + return true; + } + + return false; +} + +void Utilities::update_memory_info() { + texture_mem_cache = RenderingDevice::get_singleton()->get_memory_usage(RenderingDevice::MEMORY_TEXTURES); + buffer_mem_cache = RenderingDevice::get_singleton()->get_memory_usage(RenderingDevice::MEMORY_BUFFERS); + total_mem_cache = RenderingDevice::get_singleton()->get_memory_usage(RenderingDevice::MEMORY_TOTAL); +} + +uint64_t Utilities::get_rendering_info(RS::RenderingInfo p_info) { + if (p_info == RS::RENDERING_INFO_TEXTURE_MEM_USED) { + return texture_mem_cache; + } else if (p_info == RS::RENDERING_INFO_BUFFER_MEM_USED) { + return buffer_mem_cache; + } else if (p_info == RS::RENDERING_INFO_VIDEO_MEM_USED) { + return total_mem_cache; + } + return 0; +} + +String Utilities::get_video_adapter_name() const { + return RenderingDevice::get_singleton()->get_device_name(); +} + +String Utilities::get_video_adapter_vendor() const { + return RenderingDevice::get_singleton()->get_device_vendor_name(); +} + +RenderingDevice::DeviceType Utilities::get_video_adapter_type() const { + return RenderingDevice::get_singleton()->get_device_type(); +} + +String Utilities::get_video_adapter_api_version() const { + return RenderingDevice::get_singleton()->get_device_api_version(); +} + +Size2i Utilities::get_maximum_viewport_size() const { + RenderingDevice *device = RenderingDevice::get_singleton(); + + int max_x = device->limit_get(RenderingDevice::LIMIT_MAX_VIEWPORT_DIMENSIONS_X); + int max_y = device->limit_get(RenderingDevice::LIMIT_MAX_VIEWPORT_DIMENSIONS_Y); + return Size2i(max_x, max_y); +} diff --git a/servers/rendering/renderer_rd/storage_rd/utilities.h b/servers/rendering/renderer_rd/storage_rd/utilities.h new file mode 100644 index 0000000000..dda656c380 --- /dev/null +++ b/servers/rendering/renderer_rd/storage_rd/utilities.h @@ -0,0 +1,124 @@ +/*************************************************************************/ +/* utilities.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef UTILITIES_RD_H +#define UTILITIES_RD_H + +#include "core/templates/rid_owner.h" +#include "servers/rendering/storage/utilities.h" + +namespace RendererRD { + +/* VISIBILITY NOTIFIER */ + +struct VisibilityNotifier { + AABB aabb; + Callable enter_callback; + Callable exit_callback; + Dependency dependency; +}; + +class Utilities : public RendererUtilities { +private: + static Utilities *singleton; + + /* VISIBILITY NOTIFIER */ + + mutable RID_Owner<VisibilityNotifier> visibility_notifier_owner; + + /* MISC */ + + //keep cached since it can be called form any thread + uint64_t texture_mem_cache = 0; + uint64_t buffer_mem_cache = 0; + uint64_t total_mem_cache = 0; + +public: + static Utilities *get_singleton() { return singleton; } + + Utilities(); + virtual ~Utilities() override; + + /* INSTANCES */ + + virtual RS::InstanceType get_base_type(RID p_rid) const override; + virtual bool free(RID p_rid) override; + + /* DEPENDENCIES */ + + virtual void base_update_dependency(RID p_base, DependencyTracker *p_instance) override; + + /* VISIBILITY NOTIFIER */ + + VisibilityNotifier *get_visibility_notifier(RID p_rid) { return visibility_notifier_owner.get_or_null(p_rid); }; + bool owns_visibility_notifier(RID p_rid) const { return visibility_notifier_owner.owns(p_rid); }; + + virtual RID visibility_notifier_allocate() override; + virtual void visibility_notifier_initialize(RID p_notifier) override; + virtual void visibility_notifier_free(RID p_notifier) override; + + virtual void visibility_notifier_set_aabb(RID p_notifier, const AABB &p_aabb) override; + virtual void visibility_notifier_set_callbacks(RID p_notifier, const Callable &p_enter_callbable, const Callable &p_exit_callable) override; + + virtual AABB visibility_notifier_get_aabb(RID p_notifier) const override; + virtual void visibility_notifier_call(RID p_notifier, bool p_enter, bool p_deferred) override; + + /* TIMING */ + + virtual void capture_timestamps_begin() override; + virtual void capture_timestamp(const String &p_name) override; + virtual uint32_t get_captured_timestamps_count() const override; + virtual uint64_t get_captured_timestamps_frame() const override; + virtual uint64_t get_captured_timestamp_gpu_time(uint32_t p_index) const override; + virtual uint64_t get_captured_timestamp_cpu_time(uint32_t p_index) const override; + virtual String get_captured_timestamp_name(uint32_t p_index) const override; + + /* MISC */ + + virtual void update_dirty_resources() override; + virtual void set_debug_generate_wireframes(bool p_generate) override {} + + virtual bool has_os_feature(const String &p_feature) const override; + + virtual void update_memory_info() override; + + virtual uint64_t get_rendering_info(RS::RenderingInfo p_info) override; + + virtual String get_video_adapter_name() const override; + virtual String get_video_adapter_vendor() const override; + virtual RenderingDevice::DeviceType get_video_adapter_type() const override; + virtual String get_video_adapter_api_version() const override; + + virtual Size2i get_maximum_viewport_size() const override; +}; + +} // namespace RendererRD + +#endif // UTILITIES_RD_H diff --git a/servers/rendering/renderer_rd/uniform_set_cache_rd.cpp b/servers/rendering/renderer_rd/uniform_set_cache_rd.cpp new file mode 100644 index 0000000000..84529ca400 --- /dev/null +++ b/servers/rendering/renderer_rd/uniform_set_cache_rd.cpp @@ -0,0 +1,64 @@ +/*************************************************************************/ +/* uniform_set_cache_rd.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "uniform_set_cache_rd.h" + +UniformSetCacheRD *UniformSetCacheRD::singleton = nullptr; + +void UniformSetCacheRD::_invalidate(Cache *p_cache) { + if (p_cache->prev) { + p_cache->prev->next = p_cache->next; + } else { + // At beginning of table + uint32_t table_idx = p_cache->hash % HASH_TABLE_SIZE; + hash_table[table_idx] = p_cache->next; + } + + if (p_cache->next) { + p_cache->next->prev = p_cache->prev; + } + + cache_allocator.free(p_cache); + cache_instances_used--; +} +void UniformSetCacheRD::_uniform_set_invalidation_callback(void *p_userdata) { + singleton->_invalidate(reinterpret_cast<Cache *>(p_userdata)); +} + +UniformSetCacheRD::UniformSetCacheRD() { + ERR_FAIL_COND(singleton != nullptr); + singleton = this; +} + +UniformSetCacheRD::~UniformSetCacheRD() { + if (cache_instances_used > 0) { + ERR_PRINT("At exit: " + itos(cache_instances_used) + " uniform set cache instance(s) still in use."); + } +} diff --git a/servers/rendering/renderer_rd/uniform_set_cache_rd.h b/servers/rendering/renderer_rd/uniform_set_cache_rd.h new file mode 100644 index 0000000000..bca8b02178 --- /dev/null +++ b/servers/rendering/renderer_rd/uniform_set_cache_rd.h @@ -0,0 +1,223 @@ +/*************************************************************************/ +/* uniform_set_cache_rd.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef UNIFORM_SET_CACHE_RD_H +#define UNIFORM_SET_CACHE_RD_H + +#include "core/templates/local_vector.h" +#include "core/templates/paged_allocator.h" +#include "servers/rendering/rendering_device.h" + +class UniformSetCacheRD : public Object { + GDCLASS(UniformSetCacheRD, Object) + + struct Cache { + Cache *prev = nullptr; + Cache *next = nullptr; + uint32_t hash = 0; + RID shader; + uint32_t set = 0; + RID cache; + LocalVector<RD::Uniform> uniforms; + }; + + PagedAllocator<Cache> cache_allocator; + + enum { + HASH_TABLE_SIZE = 16381 // Prime + }; + + Cache *hash_table[HASH_TABLE_SIZE] = {}; + + static _FORCE_INLINE_ uint32_t _hash_uniform(const RD::Uniform &u, uint32_t h) { + h = hash_murmur3_one_32(u.uniform_type, h); + h = hash_murmur3_one_32(u.binding, h); + uint32_t rsize = u.get_id_count(); + for (uint32_t j = 0; j < rsize; j++) { + h = hash_murmur3_one_64(u.get_id(j).get_id(), h); + } + return hash_fmix32(h); + } + + static _FORCE_INLINE_ bool _compare_uniform(const RD::Uniform &a, const RD::Uniform &b) { + if (a.binding != b.binding) { + return false; + } + if (a.uniform_type != b.uniform_type) { + return false; + } + uint32_t rsize = a.get_id_count(); + if (rsize != b.get_id_count()) { + return false; + } + for (uint32_t j = 0; j < rsize; j++) { + if (a.get_id(j) != b.get_id(j)) { + return false; + } + } + return true; + } + + _FORCE_INLINE_ uint32_t _hash_args(uint32_t h, const RD::Uniform &arg) { + return _hash_uniform(arg, h); + } + + template <typename... Args> + uint32_t _hash_args(uint32_t h, const RD::Uniform &arg, Args... args) { + h = _hash_uniform(arg, h); + return _hash_args(h, args...); + } + + _FORCE_INLINE_ bool _compare_args(uint32_t idx, const LocalVector<RD::Uniform> &uniforms, const RD::Uniform &arg) { + return _compare_uniform(uniforms[idx], arg); + } + + template <typename... Args> + _FORCE_INLINE_ bool _compare_args(uint32_t idx, const LocalVector<RD::Uniform> &uniforms, const RD::Uniform &arg, Args... args) { + if (!_compare_uniform(uniforms[idx], arg)) { + return false; + } + return _compare_args(idx + 1, uniforms, args...); + } + + _FORCE_INLINE_ void _create_args(Vector<RD::Uniform> &uniforms, const RD::Uniform &arg) { + uniforms.push_back(arg); + } + + template <typename... Args> + _FORCE_INLINE_ void _create_args(Vector<RD::Uniform> &uniforms, const RD::Uniform &arg, Args... args) { + uniforms.push_back(arg); + _create_args(uniforms, args...); + } + + static UniformSetCacheRD *singleton; + + uint32_t cache_instances_used = 0; + + void _invalidate(Cache *p_cache); + static void _uniform_set_invalidation_callback(void *p_userdata); + + RID _allocate_from_uniforms(RID p_shader, uint32_t p_set, uint32_t p_hash, uint32_t p_table_idx, const Vector<RD::Uniform> &p_uniforms) { + RID rid = RD::get_singleton()->uniform_set_create(p_uniforms, p_shader, p_set); + ERR_FAIL_COND_V(rid.is_null(), rid); + + Cache *c = cache_allocator.alloc(); + c->hash = p_hash; + c->set = p_set; + c->shader = p_shader; + c->cache = rid; + c->uniforms.resize(p_uniforms.size()); + for (uint32_t i = 0; i < c->uniforms.size(); i++) { + c->uniforms[i] = p_uniforms[i]; + } + c->prev = nullptr; + c->next = hash_table[p_table_idx]; + if (hash_table[p_table_idx]) { + hash_table[p_table_idx]->prev = c; + } + hash_table[p_table_idx] = c; + + RD::get_singleton()->uniform_set_set_invalidation_callback(rid, _uniform_set_invalidation_callback, c); + + cache_instances_used++; + + return rid; + } + +public: + template <typename... Args> + RID get_cache(RID p_shader, uint32_t p_set, Args... args) { + uint32_t h = hash_murmur3_one_64(p_shader.get_id()); + h = hash_murmur3_one_32(p_set, h); + h = _hash_args(h, args...); + + uint32_t table_idx = h % HASH_TABLE_SIZE; + { + const Cache *c = hash_table[table_idx]; + + while (c) { + if (c->hash == h && c->set == p_set && c->shader == p_shader && sizeof...(Args) == c->uniforms.size() && _compare_args(0, c->uniforms, args...)) { + return c->cache; + } + c = c->next; + } + } + + // Not in cache, create: + + Vector<RD::Uniform> uniforms; + _create_args(uniforms, args...); + + return _allocate_from_uniforms(p_shader, p_set, h, table_idx, uniforms); + } + + template <typename... Args> + RID get_cache_vec(RID p_shader, uint32_t p_set, const Vector<RD::Uniform> &p_uniforms) { + uint32_t h = hash_murmur3_one_64(p_shader.get_id()); + h = hash_murmur3_one_32(p_set, h); + for (int i = 0; i < p_uniforms.size(); i++) { + h = _hash_uniform(p_uniforms[i], h); + } + + h = hash_fmix32(h); + + uint32_t table_idx = h % HASH_TABLE_SIZE; + { + const Cache *c = hash_table[table_idx]; + + while (c) { + if (c->hash == h && c->set == p_set && c->shader == p_shader && (uint32_t)p_uniforms.size() == c->uniforms.size()) { + bool all_ok = true; + for (int i = 0; i < p_uniforms.size(); i++) { + if (!_compare_uniform(p_uniforms[i], c->uniforms[i])) { + all_ok = false; + break; + } + } + + if (all_ok) { + return c->cache; + } + } + c = c->next; + } + } + + // Not in cache, create: + return _allocate_from_uniforms(p_shader, p_set, h, table_idx, p_uniforms); + } + + static UniformSetCacheRD *get_singleton() { return singleton; } + + UniformSetCacheRD(); + ~UniformSetCacheRD(); +}; + +#endif // UNIFORM_SET_CACHE_RD_H |