diff options
Diffstat (limited to 'servers/rendering/renderer_rd')
46 files changed, 12060 insertions, 9319 deletions
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..0fdd864d47 --- /dev/null +++ b/servers/rendering/renderer_rd/cluster_builder_rd.cpp @@ -0,0 +1,555 @@ +/*************************************************************************/ +/* cluster_builder_rd.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2021 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2021 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); + copymem(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); + copymem(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); + copymem(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); + copymem(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); + copymem(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); + copymem(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 = (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.ids.push_back(state_uniform); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 2; + u.ids.push_back(element_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 3; + u.ids.push_back(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.ids.push_back(cluster_render_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 2; + u.ids.push_back(cluster_buffer); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 3; + u.ids.push_back(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.ids.push_back(cluster_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 2; + u.ids.push_back(p_color_buffer); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 3; + u.ids.push_back(p_depth_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 4; + u.ids.push_back(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 Transform &p_view_transform, const CameraMatrix &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); + } + + CameraMatrix 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 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; + + RendererStorageRD::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 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; +#if 0 + for (uint32_t j = i+1; j < element_count; j++) { + if (elements[i].type!=elements[j].type) { + break; + } + instances++; + } +#endif + 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 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 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..dc1707b534 --- /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) 2007-2021 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2021 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/renderer_storage_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" + +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; + + Transform view_xform; + CameraMatrix adjusted_projection; + CameraMatrix 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 Transform &p_view_transform, const CameraMatrix &p_cam_projection, bool p_flip_y); + + _FORCE_INLINE_ void add_light(LightType p_type, const Transform &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]; + + Transform 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]; + + RendererStorageRD::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::deg2rad(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.origin, -xform.basis.get_axis(Vector3::AXIS_Z)); + float dist = base_plane.distance_to(Vector3()); + if (dist >= 0 && dist < radius) { + //inside, check angle + float angle = Math::rad2deg(Math::acos((-xform.origin.normalized()).dot(-xform.basis.get_axis(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 + + RendererStorageRD::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 Transform &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]; + Transform 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.elements[i].length(); + scale[i] *= s; + xform.basis.elements[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]; + + RendererStorageRD::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_H diff --git a/servers/rendering/renderer_rd/effects_rd.cpp b/servers/rendering/renderer_rd/effects_rd.cpp index 6e1d61ff94..bc304aedd8 100644 --- a/servers/rendering/renderer_rd/effects_rd.cpp +++ b/servers/rendering/renderer_rd/effects_rd.cpp @@ -299,15 +299,12 @@ void EffectsRD::copy_to_rect(RID p_source_rd_texture, RID p_dest_texture, const copy.push_constant.target[0] = p_rect.position.x; copy.push_constant.target[1] = p_rect.position.y; - int32_t x_groups = (p_rect.size.width - 1) / 8 + 1; - int32_t y_groups = (p_rect.size.height - 1) / 8 + 1; - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[p_8_bit_dst ? COPY_MODE_SIMPLY_COPY_8BIT : COPY_MODE_SIMPLY_COPY]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_rd_texture), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_texture), 3); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_rect.size.width, p_rect.size.height, 1); RD::get_singleton()->compute_list_end(); } @@ -322,15 +319,12 @@ void EffectsRD::copy_cubemap_to_panorama(RID p_source_cube, RID p_dest_panorama, copy.push_constant.target[1] = 0; copy.push_constant.camera_z_far = p_lod; - int32_t x_groups = (p_panorama_size.width - 1) / 8 + 1; - int32_t y_groups = (p_panorama_size.height - 1) / 8 + 1; - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[p_is_array ? COPY_MODE_CUBE_ARRAY_TO_PANORAMA : COPY_MODE_CUBE_TO_PANORAMA]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_cube), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_panorama), 3); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_panorama_size.width, p_panorama_size.height, 1); RD::get_singleton()->compute_list_end(); } @@ -349,15 +343,12 @@ void EffectsRD::copy_depth_to_rect_and_linearize(RID p_source_rd_texture, RID p_ copy.push_constant.camera_z_far = p_z_far; copy.push_constant.camera_z_near = p_z_near; - int32_t x_groups = (p_rect.size.width - 1) / 8 + 1; - int32_t y_groups = (p_rect.size.height - 1) / 8 + 1; - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[COPY_MODE_LINEARIZE_DEPTH]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_rd_texture), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_texture), 3); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_rect.size.width, p_rect.size.height, 1); RD::get_singleton()->compute_list_end(); } @@ -374,15 +365,12 @@ void EffectsRD::copy_depth_to_rect(RID p_source_rd_texture, RID p_dest_texture, copy.push_constant.target[0] = p_rect.position.x; copy.push_constant.target[1] = p_rect.position.y; - int32_t x_groups = (p_rect.size.width - 1) / 8 + 1; - int32_t y_groups = (p_rect.size.height - 1) / 8 + 1; - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[COPY_MODE_SIMPLY_COPY_DEPTH]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_rd_texture), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_texture), 3); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_rect.size.width, p_rect.size.height, 1); RD::get_singleton()->compute_list_end(); } @@ -400,14 +388,11 @@ void EffectsRD::set_color(RID p_dest_texture, const Color &p_color, const Rect2i copy.push_constant.set_color[2] = p_color.b; copy.push_constant.set_color[3] = p_color.a; - int32_t x_groups = (p_region.size.width - 1) / 8 + 1; - int32_t y_groups = (p_region.size.height - 1) / 8 + 1; - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[p_8bit_dst ? COPY_MODE_SET_COLOR_8BIT : COPY_MODE_SET_COLOR]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_texture), 3); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_region.size.width, p_region.size.height, 1); RD::get_singleton()->compute_list_end(); } @@ -420,8 +405,6 @@ void EffectsRD::gaussian_blur(RID p_source_rd_texture, RID p_texture, RID p_back copy.push_constant.section[2] = p_region.size.width; copy.push_constant.section[3] = p_region.size.height; - int32_t x_groups = (p_region.size.width - 1) / 8 + 1; - int32_t y_groups = (p_region.size.height - 1) / 8 + 1; //HORIZONTAL RD::DrawListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[p_8bit_dst ? COPY_MODE_GAUSSIAN_COPY_8BIT : COPY_MODE_GAUSSIAN_COPY]); @@ -431,7 +414,7 @@ void EffectsRD::gaussian_blur(RID p_source_rd_texture, RID p_texture, RID p_back copy.push_constant.flags = base_flags | COPY_FLAG_HORIZONTAL; RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_region.size.width, p_region.size.height, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); @@ -442,7 +425,7 @@ void EffectsRD::gaussian_blur(RID p_source_rd_texture, RID p_texture, RID p_back copy.push_constant.flags = base_flags; RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_region.size.width, p_region.size.height, 1); RD::get_singleton()->compute_list_end(); } @@ -452,9 +435,6 @@ void EffectsRD::gaussian_glow(RID p_source_rd_texture, RID p_back_texture, const 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; - int32_t x_groups = (p_size.width + 7) / 8; - int32_t y_groups = (p_size.height + 7) / 8; - copy.push_constant.section[2] = p_size.x; copy.push_constant.section[3] = p_size.y; @@ -479,16 +459,13 @@ void EffectsRD::gaussian_glow(RID p_source_rd_texture, RID p_back_texture, const 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(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_size.width, p_size.height, 1); RD::get_singleton()->compute_list_end(); } void EffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal_roughness, RenderingServer::EnvironmentSSRRoughnessQuality p_roughness_quality, RID p_blur_radius, RID p_blur_radius2, RID p_metallic, const Color &p_metallic_mask, RID p_depth, RID p_scale_depth, RID p_scale_normal, RID p_output, RID p_output_blur, const Size2i &p_screen_size, int p_max_steps, float p_fade_in, float p_fade_out, float p_tolerance, const CameraMatrix &p_camera) { RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - int32_t x_groups = (p_screen_size.width - 1) / 8 + 1; - int32_t y_groups = (p_screen_size.height - 1) / 8 + 1; - { //scale color and depth to half ssr_scale.push_constant.camera_z_far = p_camera.get_z_far(); ssr_scale.push_constant.camera_z_near = p_camera.get_z_near(); @@ -506,7 +483,7 @@ void EffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal_roughness, R RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssr_scale.push_constant, sizeof(ScreenSpaceReflectionScalePushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + 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); } @@ -547,7 +524,7 @@ void EffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal_roughness, R } RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_scale_normal), 2); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); } if (p_roughness_quality != RS::ENV_SSR_ROUGNESS_QUALITY_DISABLED) { @@ -585,7 +562,7 @@ void EffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal_roughness, R RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssr_filter.push_constant, sizeof(ScreenSpaceReflectionFilterPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + 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); @@ -600,7 +577,7 @@ void EffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal_roughness, R RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssr_filter.push_constant, sizeof(ScreenSpaceReflectionFilterPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); } RD::get_singleton()->compute_list_end(); @@ -609,9 +586,6 @@ void EffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal_roughness, R void EffectsRD::sub_surface_scattering(RID p_diffuse, RID p_diffuse2, RID p_depth, const CameraMatrix &p_camera, const Size2i &p_screen_size, float p_scale, float p_depth_scale, RenderingServer::SubSurfaceScatteringQuality p_quality) { RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - int32_t x_groups = (p_screen_size.width - 1) / 8 + 1; - int32_t y_groups = (p_screen_size.height - 1) / 8 + 1; - Plane p = p_camera.xform4(Plane(1, 0, -1, 1)); p.normal /= p.d; float unit_size = p.normal.x; @@ -635,7 +609,7 @@ void EffectsRD::sub_surface_scattering(RID p_diffuse, RID p_diffuse2, RID p_dept RD::get_singleton()->compute_list_set_push_constant(compute_list, &sss.push_constant, sizeof(SubSurfaceScatteringPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + 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); @@ -646,7 +620,7 @@ void EffectsRD::sub_surface_scattering(RID p_diffuse, RID p_diffuse2, RID p_dept 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(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); RD::get_singleton()->compute_list_end(); } @@ -690,39 +664,33 @@ void EffectsRD::make_mipmap(RID p_source_rd_texture, RID p_dest_texture, const S copy.push_constant.section[2] = p_size.width; copy.push_constant.section[3] = p_size.height; - int32_t x_groups = (p_size.width - 1) / 8 + 1; - int32_t y_groups = (p_size.height - 1) / 8 + 1; - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[COPY_MODE_MIPMAP]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_rd_texture), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_texture), 3); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_size.width, p_size.height, 1); RD::get_singleton()->compute_list_end(); } -void EffectsRD::copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dest_texture, const Rect2i &p_rect, float p_z_near, float p_z_far, float p_bias, bool p_dp_flip) { +void EffectsRD::copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dst_framebuffer, const Rect2 &p_rect, float p_z_near, float p_z_far, bool p_dp_flip) { CopyToDPPushConstant push_constant; - push_constant.screen_size[0] = p_rect.size.x; - push_constant.screen_size[1] = p_rect.size.y; - push_constant.dest_offset[0] = p_rect.position.x; - push_constant.dest_offset[1] = p_rect.position.y; - push_constant.bias = p_bias; + 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.z_flip = p_dp_flip; - int32_t x_groups = (p_rect.size.width - 1) / 8 + 1; - int32_t y_groups = (p_rect.size.height - 1) / 8 + 1; + 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, _get_uniform_set_from_texture(p_source_rd_texture), 0); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, cube_to_dp.pipeline); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_rd_texture), 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_texture), 1); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(CopyToDPPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); - RD::get_singleton()->compute_list_end(); + 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 EffectsRD::tonemapper(RID p_source_color, RID p_dst_framebuffer, const TonemapSettings &p_settings) { @@ -807,10 +775,7 @@ void EffectsRD::luminance_reduction(RID p_source_texture, const Size2i p_source_ RD::get_singleton()->compute_list_set_push_constant(compute_list, &luminance_reduce.push_constant, sizeof(LuminanceReducePushConstant)); - int32_t x_groups = (luminance_reduce.push_constant.source_size[0] - 1) / 8 + 1; - int32_t y_groups = (luminance_reduce.push_constant.source_size[1] - 1) / 8 + 1; - - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + 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); @@ -851,14 +816,12 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_base_texture), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_depth_texture), 1); - int32_t x_groups = (p_base_texture_size.x - 1) / 8 + 1; - int32_t y_groups = (p_base_texture_size.y - 1) / 8 + 1; bokeh.push_constant.size[0] = p_base_texture_size.x; bokeh.push_constant.size[1] = p_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(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_base_texture_size.x, p_base_texture_size.y, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); if (p_bokeh_shape == RS::DOF_BOKEH_BOX || p_bokeh_shape == RS::DOF_BOKEH_HEXAGON) { @@ -875,8 +838,6 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_halfsize_texture1), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_base_texture), 1); - x_groups = ((p_base_texture_size.x >> 1) - 1) / 8 + 1; - y_groups = ((p_base_texture_size.y >> 1) - 1) / 8 + 1; bokeh.push_constant.size[0] = p_base_texture_size.x >> 1; bokeh.push_constant.size[1] = p_base_texture_size.y >> 1; bokeh.push_constant.half_size = true; @@ -890,7 +851,7 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + 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 @@ -906,7 +867,7 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + 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 (p_quality == RS::DOF_BLUR_QUALITY_VERY_LOW || p_quality == RS::DOF_BLUR_QUALITY_LOW) { @@ -917,8 +878,6 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_base_texture), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_halfsize_texture2), 1); - x_groups = (p_base_texture_size.x - 1) / 8 + 1; - y_groups = (p_base_texture_size.y - 1) / 8 + 1; bokeh.push_constant.size[0] = p_base_texture_size.x; bokeh.push_constant.size[1] = p_base_texture_size.y; bokeh.push_constant.half_size = false; @@ -926,7 +885,7 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_base_texture_size.x, p_base_texture_size.y, 1); } } else { //circle @@ -944,15 +903,13 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_halfsize_texture1), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_base_texture), 1); - x_groups = ((p_base_texture_size.x >> 1) - 1) / 8 + 1; - y_groups = ((p_base_texture_size.y >> 1) - 1) / 8 + 1; bokeh.push_constant.size[0] = p_base_texture_size.x >> 1; bokeh.push_constant.size[1] = p_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(compute_list, x_groups, y_groups, 1); + 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 @@ -964,8 +921,6 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_base_texture), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_halfsize_texture1), 1); - x_groups = (p_base_texture_size.x - 1) / 8 + 1; - y_groups = (p_base_texture_size.y - 1) / 8 + 1; bokeh.push_constant.size[0] = p_base_texture_size.x; bokeh.push_constant.size[1] = p_base_texture_size.y; bokeh.push_constant.half_size = false; @@ -973,7 +928,7 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_base_texture_size.x, p_base_texture_size.y, 1); } RD::get_singleton()->compute_list_end(); @@ -998,20 +953,20 @@ void EffectsRD::gather_ssao(RD::ComputeListID p_compute_list, const Vector<RID> RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, _get_uniform_set_from_image(p_ao_slices[i]), 2); RD::get_singleton()->compute_list_set_push_constant(p_compute_list, &ssao.gather_push_constant, sizeof(SSAOGatherPushConstant)); - int x_groups = ((p_settings.full_screen_size.x >> (p_settings.half_size ? 2 : 1)) - 1) / 8 + 1; - int y_groups = ((p_settings.full_screen_size.y >> (p_settings.half_size ? 2 : 1)) - 1) / 8 + 1; + Size2i size = Size2i(p_settings.full_screen_size.x >> (p_settings.half_size ? 2 : 1), p_settings.full_screen_size.y >> (p_settings.half_size ? 2 : 1)); - RD::get_singleton()->compute_list_dispatch(p_compute_list, x_groups, y_groups, 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 EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_depth_mipmaps_texture, const Vector<RID> &p_depth_mipmaps, RID p_ao, const Vector<RID> p_ao_slices, RID p_ao_pong, const Vector<RID> p_ao_pong_slices, RID p_upscale_buffer, RID p_importance_map, RID p_importance_map_pong, const CameraMatrix &p_projection, const SSAOSettings &p_settings, bool p_invalidate_uniform_sets) { RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - + RD::get_singleton()->draw_command_begin_label("SSAO"); /* FIRST PASS */ // Downsample and deinterleave the depth buffer. { + RD::get_singleton()->draw_command_begin_label("Downsample Depth"); if (p_invalidate_uniform_sets) { Vector<RD::Uniform> uniforms; { @@ -1074,16 +1029,17 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep } RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.downsample_push_constant, sizeof(SSAODownsamplePushConstant)); - int x_groups = (MAX(1, p_settings.full_screen_size.x >> (p_settings.half_size ? 2 : 1)) - 1) / 8 + 1; - int y_groups = (MAX(1, p_settings.full_screen_size.y >> (p_settings.half_size ? 2 : 1)) - 1) / 8 + 1; + Size2i size(MAX(1, p_settings.full_screen_size.x >> (p_settings.half_size ? 2 : 1)), MAX(1, p_settings.full_screen_size.y >> (p_settings.half_size ? 2 : 1))); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 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(); // Downsample SSAO } /* 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; @@ -1184,6 +1140,7 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep } if (p_settings.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_settings.half_screen_size.x; ssao.importance_map_push_constant.half_screen_pixel_size[1] = 1.0 / p_settings.half_screen_size.y; ssao.importance_map_push_constant.intensity = p_settings.intensity; @@ -1192,21 +1149,19 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GATHER_BASE]); gather_ssao(compute_list, p_ao_pong_slices, p_settings, true); //generate importance map - int x_groups = (p_settings.quarter_screen_size.x - 1) / 8 + 1; - int y_groups = (p_settings.quarter_screen_size.y - 1) / 8 + 1; RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GENERATE_IMPORTANCE_MAP]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_ao_pong), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_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(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_settings.quarter_screen_size.x, p_settings.quarter_screen_size.y, 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::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_importance_map), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_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(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_settings.quarter_screen_size.x, p_settings.quarter_screen_size.y, 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]); @@ -1214,21 +1169,24 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_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(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_settings.quarter_screen_size.x, p_settings.quarter_screen_size.y, 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_ao_slices, p_settings, false); + 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_settings.half_screen_size.x; ssao.blur_push_constant.half_screen_pixel_size[1] = 1.0 / p_settings.half_screen_size.y; @@ -1268,22 +1226,22 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep } RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.blur_push_constant, sizeof(SSAOBlurPushConstant)); - int x_groups = ((p_settings.full_screen_size.x >> (p_settings.half_size ? 2 : 1)) - 1) / 8 + 1; - int y_groups = ((p_settings.full_screen_size.y >> (p_settings.half_size ? 2 : 1)) - 1) / 8 + 1; - - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + Size2i size(p_settings.full_screen_size.x >> (p_settings.half_size ? 2 : 1), p_settings.full_screen_size.y >> (p_settings.half_size ? 2 : 1)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, size.x, size.y, 1); } if (p_settings.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; @@ -1307,17 +1265,15 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.interleave_push_constant, sizeof(SSAOInterleavePushConstant)); - int x_groups = (p_settings.full_screen_size.x - 1) / 8 + 1; - int y_groups = (p_settings.full_screen_size.y - 1) / 8 + 1; - - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + 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()->compute_list_end(); + RD::get_singleton()->draw_command_end_label(); //SSAO + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_TRANSFER); //wait for upcoming transfer int zero[1] = { 0 }; - RD::get_singleton()->buffer_update(ssao.importance_map_load_counter, 0, sizeof(uint32_t), &zero, false); + RD::get_singleton()->buffer_update(ssao.importance_map_load_counter, 0, sizeof(uint32_t), &zero, 0); //no barrier } void EffectsRD::roughness_limit(RID p_source_normal, RID p_roughness, const Size2i &p_size, float p_curve) { @@ -1330,12 +1286,9 @@ void EffectsRD::roughness_limit(RID p_source_normal, RID p_roughness, const Size 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); - 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, &roughness_limiter.push_constant, sizeof(RoughnessLimiterPushConstant)); //not used but set anyway - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_size.x, p_size.y, 1); RD::get_singleton()->compute_list_end(); } @@ -1448,7 +1401,7 @@ void EffectsRD::render_sky(RD::DrawListID p_list, float p_time, RID p_fb, RID p_ RD::get_singleton()->draw_list_draw(draw_list, true); } -void EffectsRD::resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RID p_source_giprobe, RID p_dest_depth, RID p_dest_normal_roughness, RID p_dest_giprobe, Vector2i p_screen_size, int p_samples) { +void EffectsRD::resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RID p_source_giprobe, RID p_dest_depth, RID p_dest_normal_roughness, RID p_dest_giprobe, Vector2i p_screen_size, int p_samples, uint32_t p_barrier) { ResolvePushConstant push_constant; push_constant.screen_size[0] = p_screen_size.x; push_constant.screen_size[1] = p_screen_size.y; @@ -1465,54 +1418,9 @@ void EffectsRD::resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RI 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, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.x, p_screen_size.y, 1); - RD::get_singleton()->compute_list_end(); -} - -void EffectsRD::reduce_shadow(RID p_source_shadow, RID p_dest_shadow, const Size2i &p_source_size, const Rect2i &p_source_rect, int p_shrink_limit, RD::ComputeListID compute_list) { - uint32_t push_constant[8] = { (uint32_t)p_source_size.x, (uint32_t)p_source_size.y, (uint32_t)p_source_rect.position.x, (uint32_t)p_source_rect.position.y, (uint32_t)p_shrink_limit, 0, 0, 0 }; - - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, shadow_reduce.pipelines[SHADOW_REDUCE_REDUCE]); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_image_pair(p_source_shadow, p_dest_shadow), 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(uint32_t) * 8); - - RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_source_rect.size.width, p_source_rect.size.height, 1, 8, 8, 1); -} -void EffectsRD::filter_shadow(RID p_shadow, RID p_backing_shadow, const Size2i &p_source_size, const Rect2i &p_source_rect, RenderingServer::EnvVolumetricFogShadowFilter p_filter, RD::ComputeListID compute_list, bool p_vertical, bool p_horizontal) { - uint32_t push_constant[8] = { (uint32_t)p_source_size.x, (uint32_t)p_source_size.y, (uint32_t)p_source_rect.position.x, (uint32_t)p_source_rect.position.y, 0, 0, 0, 0 }; - - switch (p_filter) { - case RS::ENV_VOLUMETRIC_FOG_SHADOW_FILTER_DISABLED: - case RS::ENV_VOLUMETRIC_FOG_SHADOW_FILTER_LOW: { - push_constant[5] = 0; - } break; - case RS::ENV_VOLUMETRIC_FOG_SHADOW_FILTER_MEDIUM: { - push_constant[5] = 9; - } break; - case RS::ENV_VOLUMETRIC_FOG_SHADOW_FILTER_HIGH: { - push_constant[5] = 18; - } break; - } - - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, shadow_reduce.pipelines[SHADOW_REDUCE_FILTER]); - if (p_vertical) { - push_constant[6] = 1; - push_constant[7] = 0; - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_image_pair(p_shadow, p_backing_shadow), 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(uint32_t) * 8); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_source_rect.size.width, p_source_rect.size.height, 1, 8, 8, 1); - } - if (p_vertical && p_horizontal) { - RD::get_singleton()->compute_list_add_barrier(compute_list); - } - if (p_horizontal) { - push_constant[6] = 0; - push_constant[7] = 1; - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_image_pair(p_backing_shadow, p_shadow), 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(uint32_t) * 8); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_source_rect.size.width, p_source_rect.size.height, 1, 8, 8, 1); - } + RD::get_singleton()->compute_list_end(p_barrier); } void EffectsRD::sort_buffer(RID p_uniform_set, int p_size) { @@ -1678,8 +1586,12 @@ EffectsRD::EffectsRD() { cube_to_dp.shader.initialize(copy_modes); cube_to_dp.shader_version = cube_to_dp.shader.version_create(); - - cube_to_dp.pipeline = RD::get_singleton()->compute_pipeline_create(cube_to_dp.shader.version_get_shader(cube_to_dp.shader_version, 0)); + 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); } { @@ -1776,7 +1688,7 @@ EffectsRD::EffectsRD() { } } - RD::get_singleton()->buffer_update(ssao.gather_constants_buffer, 0, sizeof(SSAOGatherConstants), &gather_constants, false); + RD::get_singleton()->buffer_update(ssao.gather_constants_buffer, 0, sizeof(SSAOGatherConstants), &gather_constants); } { Vector<String> ssao_modes; @@ -1795,7 +1707,8 @@ EffectsRD::EffectsRD() { } 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, false); + 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; { @@ -1806,6 +1719,7 @@ EffectsRD::EffectsRD() { 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; @@ -1834,7 +1748,7 @@ EffectsRD::EffectsRD() { 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++; } } @@ -1867,7 +1781,7 @@ EffectsRD::EffectsRD() { { // Initialize cubemap filter - filter.use_high_quality = GLOBAL_GET("rendering/quality/reflections/fast_filter_high_quality"); + 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"); @@ -1883,10 +1797,10 @@ EffectsRD::EffectsRD() { 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], false); + 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], false); + RD::get_singleton()->buffer_update(filter.coefficient_buffer, 0, sizeof(low_quality_coeffs), &low_quality_coeffs[0]); } Vector<RD::Uniform> uniforms; @@ -2005,20 +1919,6 @@ EffectsRD::EffectsRD() { } { - Vector<String> shadow_reduce_modes; - shadow_reduce_modes.push_back("\n#define MODE_REDUCE\n"); - shadow_reduce_modes.push_back("\n#define MODE_FILTER\n"); - - shadow_reduce.shader.initialize(shadow_reduce_modes); - - shadow_reduce.shader_version = shadow_reduce.shader.version_create(); - - for (int i = 0; i < SHADOW_REDUCE_MAX; i++) { - shadow_reduce.pipelines[i] = RD::get_singleton()->compute_pipeline_create(shadow_reduce.shader.version_get_shader(shadow_reduce.shader_version, i)); - } - } - - { Vector<String> sort_modes; sort_modes.push_back("\n#define MODE_SORT_BLOCK\n"); sort_modes.push_back("\n#define MODE_SORT_STEP\n"); @@ -2039,12 +1939,14 @@ EffectsRD::EffectsRD() { 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; @@ -2104,5 +2006,4 @@ EffectsRD::~EffectsRD() { ssr_scale.shader.version_free(ssr_scale.shader_version); sss.shader.version_free(sss.shader_version); tonemap.shader.version_free(tonemap.shader_version); - shadow_reduce.shader.version_free(shadow_reduce.shader_version); } diff --git a/servers/rendering/renderer_rd/effects_rd.h b/servers/rendering/renderer_rd/effects_rd.h index e2cdd0c3d8..1ba25e301b 100644 --- a/servers/rendering/renderer_rd/effects_rd.h +++ b/servers/rendering/renderer_rd/effects_rd.h @@ -46,7 +46,6 @@ #include "servers/rendering/renderer_rd/shaders/screen_space_reflection.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/screen_space_reflection_filter.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/screen_space_reflection_scale.glsl.gen.h" -#include "servers/rendering/renderer_rd/shaders/shadow_reduce.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/sort.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/specular_merge.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/ssao.glsl.gen.h" @@ -234,18 +233,17 @@ class EffectsRD { } luminance_reduce; struct CopyToDPPushConstant { - int32_t screen_size[2]; - int32_t dest_offset[2]; - float bias; float z_far; float z_near; uint32_t z_flip; + uint32_t pad; + float screen_rect[4]; }; struct CoptToDP { CubeToDpShaderRD shader; RID shader_version; - RID pipeline; + PipelineCacheRD pipeline; } cube_to_dp; struct BokehPushConstant { @@ -598,18 +596,6 @@ class EffectsRD { RID pipelines[RESOLVE_MODE_MAX]; //3 quality levels } resolve; - enum ShadowReduceMode { - SHADOW_REDUCE_REDUCE, - SHADOW_REDUCE_FILTER, - SHADOW_REDUCE_MAX - }; - - struct ShadowReduce { - ShadowReduceShaderRD shader; - RID shader_version; - RID pipelines[SHADOW_REDUCE_MAX]; - } shadow_reduce; - enum SortMode { SORT_MODE_BLOCK, SORT_MODE_STEP, @@ -687,7 +673,7 @@ public: void cubemap_roughness(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 make_mipmap(RID p_source_rd_texture, RID p_dest_texture, const Size2i &p_size); - void copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dest_texture, const Rect2i &p_rect, float p_z_near, float p_z_far, float p_bias, bool p_dp_flip); + void copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dest_texture, const Rect2 &p_rect, float p_z_near, float p_z_far, bool p_dp_flip); 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 bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i &p_base_texture_size, RID p_secondary_texture, RID p_bokeh_texture1, RID p_bokeh_texture2, bool p_dof_far, float p_dof_far_begin, float p_dof_far_size, bool p_dof_near, float p_dof_near_begin, float p_dof_near_size, float p_bokeh_size, RS::DOFBokehShape p_bokeh_shape, RS::DOFBlurQuality p_quality, bool p_use_jitter, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal); @@ -764,10 +750,7 @@ public: void merge_specular(RID p_dest_framebuffer, RID p_specular, RID p_base, RID p_reflection); void sub_surface_scattering(RID p_diffuse, RID p_diffuse2, RID p_depth, const CameraMatrix &p_camera, const Size2i &p_screen_size, float p_scale, float p_depth_scale, RS::SubSurfaceScatteringQuality p_quality); - void resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RID p_source_giprobe, RID p_dest_depth, RID p_dest_normal_roughness, RID p_dest_giprobe, Vector2i p_screen_size, int p_samples); - - void reduce_shadow(RID p_source_shadow, RID p_dest_shadow, const Size2i &p_source_size, const Rect2i &p_source_rect, int p_shrink_limit, RenderingDevice::ComputeListID compute_list); - void filter_shadow(RID p_shadow, RID p_backing_shadow, const Size2i &p_source_size, const Rect2i &p_source_rect, RS::EnvVolumetricFogShadowFilter p_filter, RenderingDevice::ComputeListID compute_list, bool p_vertical = true, bool p_horizontal = true); + void resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RID p_source_giprobe, RID p_dest_depth, RID p_dest_normal_roughness, RID p_dest_giprobe, Vector2i p_screen_size, int p_samples, uint32_t p_barrier = RD::BARRIER_MASK_ALL); void sort_buffer(RID p_uniform_set, int p_size); diff --git a/servers/rendering/renderer_rd/light_cluster_builder.cpp b/servers/rendering/renderer_rd/light_cluster_builder.cpp deleted file mode 100644 index bb807ca4ca..0000000000 --- a/servers/rendering/renderer_rd/light_cluster_builder.cpp +++ /dev/null @@ -1,252 +0,0 @@ -/*************************************************************************/ -/* light_cluster_builder.cpp */ -/*************************************************************************/ -/* This file is part of: */ -/* GODOT ENGINE */ -/* https://godotengine.org */ -/*************************************************************************/ -/* Copyright (c) 2007-2021 Juan Linietsky, Ariel Manzur. */ -/* Copyright (c) 2014-2021 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_cluster_builder.h" - -void LightClusterBuilder::begin(const Transform &p_view_transform, const CameraMatrix &p_cam_projection) { - view_xform = p_view_transform; - projection = p_cam_projection; - z_near = -projection.get_z_near(); - z_far = -projection.get_z_far(); - - //reset counts - light_count = 0; - refprobe_count = 0; - decal_count = 0; - item_count = 0; - sort_id_count = 0; -} - -void LightClusterBuilder::bake_cluster() { - float slice_depth = (z_near - z_far) / depth; - - uint8_t *cluster_dataw = cluster_data.ptrw(); - Cell *cluster_data_ptr = (Cell *)cluster_dataw; - //clear the cluster - zeromem(cluster_data_ptr, (width * height * depth * sizeof(Cell))); - - /* Step 1, create cell positions and count them */ - - for (uint32_t i = 0; i < item_count; i++) { - const Item &item = items[i]; - - int from_slice = Math::floor((z_near - (item.aabb.position.z + item.aabb.size.z)) / slice_depth); - int to_slice = Math::floor((z_near - item.aabb.position.z) / slice_depth); - - if (from_slice >= (int)depth || to_slice < 0) { - continue; //sorry no go - } - - from_slice = MAX(0, from_slice); - to_slice = MIN((int)depth - 1, to_slice); - - for (int j = from_slice; j <= to_slice; j++) { - Vector3 min = item.aabb.position; - Vector3 max = item.aabb.position + item.aabb.size; - - float limit_near = MIN((z_near - slice_depth * j), max.z); - float limit_far = MAX((z_near - slice_depth * (j + 1)), min.z); - - max.z = limit_near; - min.z = limit_near; - - Vector3 proj_min = projection.xform(min); - Vector3 proj_max = projection.xform(max); - - int near_from_x = int(Math::floor((proj_min.x * 0.5 + 0.5) * width)); - int near_from_y = int(Math::floor((-proj_max.y * 0.5 + 0.5) * height)); - int near_to_x = int(Math::floor((proj_max.x * 0.5 + 0.5) * width)); - int near_to_y = int(Math::floor((-proj_min.y * 0.5 + 0.5) * height)); - - max.z = limit_far; - min.z = limit_far; - - proj_min = projection.xform(min); - proj_max = projection.xform(max); - - int far_from_x = int(Math::floor((proj_min.x * 0.5 + 0.5) * width)); - int far_from_y = int(Math::floor((-proj_max.y * 0.5 + 0.5) * height)); - int far_to_x = int(Math::floor((proj_max.x * 0.5 + 0.5) * width)); - int far_to_y = int(Math::floor((-proj_min.y * 0.5 + 0.5) * height)); - - //print_line(itos(j) + " near - " + Vector2i(near_from_x, near_from_y) + " -> " + Vector2i(near_to_x, near_to_y)); - //print_line(itos(j) + " far - " + Vector2i(far_from_x, far_from_y) + " -> " + Vector2i(far_to_x, far_to_y)); - - int from_x = MIN(near_from_x, far_from_x); - int from_y = MIN(near_from_y, far_from_y); - int to_x = MAX(near_to_x, far_to_x); - int to_y = MAX(near_to_y, far_to_y); - - if (from_x >= (int)width || to_x < 0 || from_y >= (int)height || to_y < 0) { - continue; - } - - int sx = MAX(0, from_x); - int sy = MAX(0, from_y); - int dx = MIN((int)width - 1, to_x); - int dy = MIN((int)height - 1, to_y); - - //print_line(itos(j) + " - " + Vector2i(sx, sy) + " -> " + Vector2i(dx, dy)); - - for (int x = sx; x <= dx; x++) { - for (int y = sy; y <= dy; y++) { - uint32_t offset = j * (width * height) + y * width + x; - - if (unlikely(sort_id_count == sort_id_max)) { - sort_id_max = nearest_power_of_2_templated(sort_id_max + 1); - sort_ids = (SortID *)memrealloc(sort_ids, sizeof(SortID) * sort_id_max); - if (ids.size()) { - ids.resize(sort_id_max); - RD::get_singleton()->free(items_buffer); - items_buffer = RD::get_singleton()->storage_buffer_create(sizeof(uint32_t) * sort_id_max); - } - } - - sort_ids[sort_id_count].cell_index = offset; - sort_ids[sort_id_count].item_index = item.index; - sort_ids[sort_id_count].item_type = item.type; - - sort_id_count++; - - //for now, only count - cluster_data_ptr[offset].item_pointers[item.type]++; - //print_line("at offset " + itos(offset) + " value: " + itos(cluster_data_ptr[offset].item_pointers[item.type])); - } - } - } - } - - /* Step 2, Assign pointers (and reset counters) */ - - uint32_t offset = 0; - for (uint32_t i = 0; i < (width * height * depth); i++) { - for (int j = 0; j < ITEM_TYPE_MAX; j++) { - uint32_t count = cluster_data_ptr[i].item_pointers[j]; //save count - cluster_data_ptr[i].item_pointers[j] = offset; //replace count by pointer - offset += count; //increase offset by count; - } - } - - //print_line("offset: " + itos(offset)); - /* Step 3, Place item lists */ - - uint32_t *ids_ptr = ids.ptrw(); - - for (uint32_t i = 0; i < sort_id_count; i++) { - const SortID &id = sort_ids[i]; - Cell &cell = cluster_data_ptr[id.cell_index]; - uint32_t pointer = cell.item_pointers[id.item_type] & POINTER_MASK; - uint32_t counter = cell.item_pointers[id.item_type] >> COUNTER_SHIFT; - ids_ptr[pointer + counter] = id.item_index; - - cell.item_pointers[id.item_type] = pointer | ((counter + 1) << COUNTER_SHIFT); - } - - RD::get_singleton()->texture_update(cluster_texture, 0, cluster_data, true); - RD::get_singleton()->buffer_update(items_buffer, 0, offset * sizeof(uint32_t), ids_ptr, true); -} - -void LightClusterBuilder::setup(uint32_t p_width, uint32_t p_height, uint32_t p_depth) { - if (width == p_width && height == p_height && depth == p_depth) { - return; - } - if (cluster_texture.is_valid()) { - RD::get_singleton()->free(cluster_texture); - } - - width = p_width; - height = p_height; - depth = p_depth; - - cluster_data.resize(width * height * depth * sizeof(Cell)); - - { - RD::TextureFormat tf; - tf.format = RD::DATA_FORMAT_R32G32B32A32_UINT; - tf.texture_type = RD::TEXTURE_TYPE_3D; - tf.width = width; - tf.height = height; - tf.depth = depth; - tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT; - - cluster_texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); - } -} - -RID LightClusterBuilder::get_cluster_texture() const { - return cluster_texture; -} - -RID LightClusterBuilder::get_cluster_indices_buffer() const { - return items_buffer; -} - -LightClusterBuilder::LightClusterBuilder() { - //initialize accumulators to something - lights = (LightData *)memalloc(sizeof(LightData) * 1024); - light_max = 1024; - - refprobes = (OrientedBoxData *)memalloc(sizeof(OrientedBoxData) * 1024); - refprobe_max = 1024; - - decals = (OrientedBoxData *)memalloc(sizeof(OrientedBoxData) * 1024); - decal_max = 1024; - - items = (Item *)memalloc(sizeof(Item) * 1024); - item_max = 1024; - - sort_ids = (SortID *)memalloc(sizeof(SortID) * 1024); - ids.resize(2014); - items_buffer = RD::get_singleton()->storage_buffer_create(sizeof(uint32_t) * 1024); - item_max = 1024; -} - -LightClusterBuilder::~LightClusterBuilder() { - if (cluster_data.size()) { - RD::get_singleton()->free(cluster_texture); - } - - if (lights) { - memfree(lights); - } - if (refprobes) { - memfree(refprobes); - } - if (decals) { - memfree(decals); - } - if (items) { - memfree(items); - } - if (sort_ids) { - memfree(sort_ids); - RD::get_singleton()->free(items_buffer); - } -} diff --git a/servers/rendering/renderer_rd/light_cluster_builder.h b/servers/rendering/renderer_rd/light_cluster_builder.h deleted file mode 100644 index 8f77ece6f5..0000000000 --- a/servers/rendering/renderer_rd/light_cluster_builder.h +++ /dev/null @@ -1,290 +0,0 @@ -/*************************************************************************/ -/* light_cluster_builder.h */ -/*************************************************************************/ -/* This file is part of: */ -/* GODOT ENGINE */ -/* https://godotengine.org */ -/*************************************************************************/ -/* Copyright (c) 2007-2021 Juan Linietsky, Ariel Manzur. */ -/* Copyright (c) 2014-2021 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_CLUSTER_BUILDER_H -#define LIGHT_CLUSTER_BUILDER_H - -#include "servers/rendering/renderer_rd/renderer_storage_rd.h" - -class LightClusterBuilder { -public: - enum LightType { - LIGHT_TYPE_OMNI, - LIGHT_TYPE_SPOT - }; - - enum ItemType { - ITEM_TYPE_OMNI_LIGHT, - ITEM_TYPE_SPOT_LIGHT, - ITEM_TYPE_REFLECTION_PROBE, - ITEM_TYPE_DECAL, - ITEM_TYPE_MAX //should always be 4 - }; - - enum { - COUNTER_SHIFT = 20, //one million total ids - POINTER_MASK = (1 << COUNTER_SHIFT) - 1, - COUNTER_MASK = 0xfff // 4096 items per cell - }; - -private: - struct LightData { - float position[3]; - uint32_t type; - float radius; - float spot_aperture; - uint32_t pad[2]; - }; - - uint32_t light_count = 0; - uint32_t light_max = 0; - LightData *lights = nullptr; - - struct OrientedBoxData { - float position[3]; - uint32_t pad; - float x_axis[3]; - uint32_t pad2; - float y_axis[3]; - uint32_t pad3; - float z_axis[3]; - uint32_t pad4; - }; - - uint32_t refprobe_count = 0; - uint32_t refprobe_max = 0; - OrientedBoxData *refprobes = nullptr; - - uint32_t decal_count = 0; - uint32_t decal_max = 0; - OrientedBoxData *decals = nullptr; - - struct Item { - AABB aabb; - ItemType type; - uint32_t index; - }; - - Item *items = nullptr; - uint32_t item_count = 0; - uint32_t item_max = 0; - - uint32_t width = 0; - uint32_t height = 0; - uint32_t depth = 0; - - struct Cell { - uint32_t item_pointers[ITEM_TYPE_MAX]; - }; - - Vector<uint8_t> cluster_data; - RID cluster_texture; - - struct SortID { - uint32_t cell_index; - uint32_t item_index; - ItemType item_type; - }; - - SortID *sort_ids = nullptr; - Vector<uint32_t> ids; - uint32_t sort_id_count = 0; - uint32_t sort_id_max = 0; - RID items_buffer; - - Transform view_xform; - CameraMatrix projection; - float z_far = 0; - float z_near = 0; - - _FORCE_INLINE_ void _add_item(const AABB &p_aabb, ItemType p_type, uint32_t p_index) { - if (unlikely(item_count == item_max)) { - item_max = nearest_power_of_2_templated(item_max + 1); - items = (Item *)memrealloc(items, sizeof(Item) * item_max); - } - - Item &item = items[item_count]; - item.aabb = p_aabb; - item.index = p_index; - item.type = p_type; - item_count++; - } - -public: - void begin(const Transform &p_view_transform, const CameraMatrix &p_cam_projection); - - _FORCE_INLINE_ void add_light(LightType p_type, const Transform &p_transform, float p_radius, float p_spot_aperture) { - if (unlikely(light_count == light_max)) { - light_max = nearest_power_of_2_templated(light_max + 1); - lights = (LightData *)memrealloc(lights, sizeof(LightData) * light_max); - } - - LightData &ld = lights[light_count]; - ld.type = p_type; - ld.position[0] = p_transform.origin.x; - ld.position[1] = p_transform.origin.y; - ld.position[2] = p_transform.origin.z; - ld.radius = p_radius; - ld.spot_aperture = p_spot_aperture; - - Transform xform = view_xform * p_transform; - - ld.radius *= xform.basis.get_uniform_scale(); - - AABB aabb; - - switch (p_type) { - case LIGHT_TYPE_OMNI: { - aabb.position = xform.origin; - aabb.size = Vector3(ld.radius, ld.radius, ld.radius); - aabb.position -= aabb.size; - aabb.size *= 2.0; - - _add_item(aabb, ITEM_TYPE_OMNI_LIGHT, light_count); - } break; - case LIGHT_TYPE_SPOT: { - float r = ld.radius; - real_t len = Math::tan(Math::deg2rad(ld.spot_aperture)) * r; - - aabb.position = xform.origin; - aabb.expand_to(xform.xform(Vector3(len, len, -r))); - aabb.expand_to(xform.xform(Vector3(-len, len, -r))); - aabb.expand_to(xform.xform(Vector3(-len, -len, -r))); - aabb.expand_to(xform.xform(Vector3(len, -len, -r))); - _add_item(aabb, ITEM_TYPE_SPOT_LIGHT, light_count); - } break; - } - - light_count++; - } - - _FORCE_INLINE_ void add_reflection_probe(const Transform &p_transform, const Vector3 &p_half_extents) { - if (unlikely(refprobe_count == refprobe_max)) { - refprobe_max = nearest_power_of_2_templated(refprobe_max + 1); - refprobes = (OrientedBoxData *)memrealloc(refprobes, sizeof(OrientedBoxData) * refprobe_max); - } - - Transform xform = view_xform * p_transform; - - OrientedBoxData &rp = refprobes[refprobe_count]; - Vector3 origin = xform.origin; - rp.position[0] = origin.x; - rp.position[1] = origin.y; - rp.position[2] = origin.z; - - Vector3 x_axis = xform.basis.get_axis(0) * p_half_extents.x; - rp.x_axis[0] = x_axis.x; - rp.x_axis[1] = x_axis.y; - rp.x_axis[2] = x_axis.z; - - Vector3 y_axis = xform.basis.get_axis(1) * p_half_extents.y; - rp.y_axis[0] = y_axis.x; - rp.y_axis[1] = y_axis.y; - rp.y_axis[2] = y_axis.z; - - Vector3 z_axis = xform.basis.get_axis(2) * p_half_extents.z; - rp.z_axis[0] = z_axis.x; - rp.z_axis[1] = z_axis.y; - rp.z_axis[2] = z_axis.z; - - AABB aabb; - - aabb.position = origin + x_axis + y_axis + z_axis; - aabb.expand_to(origin + x_axis + y_axis - z_axis); - aabb.expand_to(origin + x_axis - y_axis + z_axis); - aabb.expand_to(origin + x_axis - y_axis - z_axis); - aabb.expand_to(origin - x_axis + y_axis + z_axis); - aabb.expand_to(origin - x_axis + y_axis - z_axis); - aabb.expand_to(origin - x_axis - y_axis + z_axis); - aabb.expand_to(origin - x_axis - y_axis - z_axis); - - _add_item(aabb, ITEM_TYPE_REFLECTION_PROBE, refprobe_count); - - refprobe_count++; - } - - _FORCE_INLINE_ void add_decal(const Transform &p_transform, const Vector3 &p_half_extents) { - if (unlikely(decal_count == decal_max)) { - decal_max = nearest_power_of_2_templated(decal_max + 1); - decals = (OrientedBoxData *)memrealloc(decals, sizeof(OrientedBoxData) * decal_max); - } - - Transform xform = view_xform * p_transform; - - OrientedBoxData &dc = decals[decal_count]; - - Vector3 origin = xform.origin; - dc.position[0] = origin.x; - dc.position[1] = origin.y; - dc.position[2] = origin.z; - - Vector3 x_axis = xform.basis.get_axis(0) * p_half_extents.x; - dc.x_axis[0] = x_axis.x; - dc.x_axis[1] = x_axis.y; - dc.x_axis[2] = x_axis.z; - - Vector3 y_axis = xform.basis.get_axis(1) * p_half_extents.y; - dc.y_axis[0] = y_axis.x; - dc.y_axis[1] = y_axis.y; - dc.y_axis[2] = y_axis.z; - - Vector3 z_axis = xform.basis.get_axis(2) * p_half_extents.z; - dc.z_axis[0] = z_axis.x; - dc.z_axis[1] = z_axis.y; - dc.z_axis[2] = z_axis.z; - - AABB aabb; - - aabb.position = origin + x_axis + y_axis + z_axis; - aabb.expand_to(origin + x_axis + y_axis - z_axis); - aabb.expand_to(origin + x_axis - y_axis + z_axis); - aabb.expand_to(origin + x_axis - y_axis - z_axis); - aabb.expand_to(origin - x_axis + y_axis + z_axis); - aabb.expand_to(origin - x_axis + y_axis - z_axis); - aabb.expand_to(origin - x_axis - y_axis + z_axis); - aabb.expand_to(origin - x_axis - y_axis - z_axis); - - _add_item(aabb, ITEM_TYPE_DECAL, decal_count); - - decal_count++; - } - - void bake_cluster(); - - void setup(uint32_t p_width, uint32_t p_height, uint32_t p_depth); - - RID get_cluster_texture() const; - RID get_cluster_indices_buffer() const; - - LightClusterBuilder(); - ~LightClusterBuilder(); -}; - -#endif // LIGHT_CLUSTER_BUILDER_H diff --git a/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp b/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp index 05ffc0086d..7d6e2fa8e4 100644 --- a/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp @@ -1367,7 +1367,7 @@ void RendererCanvasRenderRD::canvas_render_items(RID p_to_render_target, Item *p } if (light_count > 0) { - RD::get_singleton()->buffer_update(state.lights_uniform_buffer, 0, sizeof(LightUniform) * light_count, &state.light_uniforms[0], true); + RD::get_singleton()->buffer_update(state.lights_uniform_buffer, 0, sizeof(LightUniform) * light_count, &state.light_uniforms[0]); } { @@ -1421,7 +1421,7 @@ void RendererCanvasRenderRD::canvas_render_items(RID p_to_render_target, Item *p //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, true); + RD::get_singleton()->buffer_update(state.canvas_state_buffer, 0, sizeof(State::Buffer), &state_buffer); } { //default filter/repeat @@ -1622,7 +1622,7 @@ void RendererCanvasRenderRD::light_update_shadow(RID p_rid, int p_shadow_index, projection.set_frustum(xmin, xmax, ymin, ymax, nearp, farp); } - Vector3 cam_target = Basis(Vector3(0, 0, Math_PI * 2 * ((i + 3) / 4.0))).xform(Vector3(0, 1, 0)); + Vector3 cam_target = Basis(Vector3(0, 0, Math_TAU * ((i + 3) / 4.0))).xform(Vector3(0, 1, 0)); projection = projection * CameraMatrix(Transform().looking_at(cam_target, Vector3(0, 0, -1)).affine_inverse()); ShadowRenderPushConstant push_constant; @@ -2239,6 +2239,11 @@ Variant RendererCanvasRenderRD::ShaderData::get_default_parameter(const StringNa return Variant(); } +RS::ShaderNativeSourceCode RendererCanvasRenderRD::ShaderData::get_native_source_code() const { + RendererCanvasRenderRD *canvas_singleton = (RendererCanvasRenderRD *)RendererCanvasRender::singleton; + return canvas_singleton->shader.canvas_shader.version_get_native_source_code(version); +} + RendererCanvasRenderRD::ShaderData::ShaderData() { valid = false; uses_screen_texture = false; @@ -2501,7 +2506,7 @@ RendererCanvasRenderRD::RendererCanvasRenderRD(RendererStorageRD *p_storage) { actions.renames["FRAGCOORD"] = "gl_FragCoord"; actions.renames["POINT_COORD"] = "gl_PointCoord"; - actions.renames["LIGHT_POSITION"] = "light_pos"; + actions.renames["LIGHT_POSITION"] = "light_position"; actions.renames["LIGHT_COLOR"] = "light_color"; actions.renames["LIGHT_ENERGY"] = "light_energy"; actions.renames["LIGHT"] = "light"; @@ -2690,9 +2695,10 @@ RendererCanvasRenderRD::RendererCanvasRenderRD(RendererStorageRD *p_storage) { state.default_transforms_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, shader.default_version_rd_shader, TRANSFORMS_UNIFORM_SET); } - default_canvas_texture = storage->canvas_texture_create(); + default_canvas_texture = storage->canvas_texture_allocate(); + storage->canvas_texture_initialize(default_canvas_texture); - state.shadow_texture_size = GLOBAL_GET("rendering/quality/2d_shadow_atlas/size"); + state.shadow_texture_size = GLOBAL_GET("rendering/2d/shadow_atlas/size"); //create functions for shader and material storage->shader_set_data_request_function(RendererStorageRD::SHADER_TYPE_2D, _create_shader_funcs); @@ -2701,9 +2707,14 @@ RendererCanvasRenderRD::RendererCanvasRenderRD(RendererStorageRD *p_storage) { state.time = 0; { - default_canvas_group_shader = storage->shader_create(); + default_canvas_group_shader = storage->shader_allocate(); + storage->shader_initialize(default_canvas_group_shader); + storage->shader_set_code(default_canvas_group_shader, "shader_type canvas_item; \nvoid fragment() {\n\tvec4 c = textureLod(SCREEN_TEXTURE,SCREEN_UV,0.0); if (c.a > 0.0001) c.rgb/=c.a; COLOR *= c; \n}\n"); - default_canvas_group_material = storage->material_create(); + + default_canvas_group_material = storage->material_allocate(); + storage->material_initialize(default_canvas_group_material); + storage->material_set_shader(default_canvas_group_material, default_canvas_group_shader); } diff --git a/servers/rendering/renderer_rd/renderer_canvas_render_rd.h b/servers/rendering/renderer_rd/renderer_canvas_render_rd.h index 545eeaa106..cb947d7180 100644 --- a/servers/rendering/renderer_rd/renderer_canvas_render_rd.h +++ b/servers/rendering/renderer_rd/renderer_canvas_render_rd.h @@ -188,6 +188,8 @@ class RendererCanvasRenderRD : public RendererCanvasRender { 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; + ShaderData(); virtual ~ShaderData(); }; diff --git a/servers/rendering/renderer_rd/renderer_compositor_rd.cpp b/servers/rendering/renderer_rd/renderer_compositor_rd.cpp index be2552bd32..d5ac05d1d1 100644 --- a/servers/rendering/renderer_rd/renderer_compositor_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_compositor_rd.cpp @@ -175,5 +175,5 @@ RendererCompositorRD::RendererCompositorRD() { storage = memnew(RendererStorageRD); canvas = memnew(RendererCanvasRenderRD(storage)); - scene = memnew(RendererSceneRenderForward(storage)); + scene = memnew(RendererSceneRenderForwardClustered(storage)); } diff --git a/servers/rendering/renderer_rd/renderer_compositor_rd.h b/servers/rendering/renderer_rd/renderer_compositor_rd.h index cb85fc79e0..67a843452b 100644 --- a/servers/rendering/renderer_rd/renderer_compositor_rd.h +++ b/servers/rendering/renderer_rd/renderer_compositor_rd.h @@ -35,14 +35,14 @@ #include "core/templates/thread_work_pool.h" #include "servers/rendering/renderer_compositor.h" #include "servers/rendering/renderer_rd/renderer_canvas_render_rd.h" -#include "servers/rendering/renderer_rd/renderer_scene_render_forward.h" +#include "servers/rendering/renderer_rd/renderer_scene_render_forward_clustered.h" #include "servers/rendering/renderer_rd/renderer_storage_rd.h" class RendererCompositorRD : public RendererCompositor { protected: RendererCanvasRenderRD *canvas; RendererStorageRD *storage; - RendererSceneRenderForward *scene; + RendererSceneRenderRD *scene; RID copy_viewports_rd_shader; RID copy_viewports_rd_pipeline; diff --git a/servers/rendering/renderer_rd/renderer_scene_environment_rd.cpp b/servers/rendering/renderer_rd/renderer_scene_environment_rd.cpp new file mode 100644 index 0000000000..d631cb4bac --- /dev/null +++ b/servers/rendering/renderer_rd/renderer_scene_environment_rd.cpp @@ -0,0 +1,126 @@ +/*************************************************************************/ +/* renderer_scene_environment_rd.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2021 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2021 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 "servers/rendering/renderer_rd/renderer_scene_environment_rd.h" + +uint64_t RendererSceneEnvironmentRD::auto_exposure_counter = 2; + +void RendererSceneEnvironmentRD::set_ambient_light(const Color &p_color, RS::EnvironmentAmbientSource p_ambient, float p_energy, float p_sky_contribution, RS::EnvironmentReflectionSource p_reflection_source, const Color &p_ao_color) { + ambient_light = p_color; + ambient_source = p_ambient; + ambient_light_energy = p_energy; + ambient_sky_contribution = p_sky_contribution; + reflection_source = p_reflection_source; + ao_color = p_ao_color; +} + +void RendererSceneEnvironmentRD::set_tonemap(RS::EnvironmentToneMapper p_tone_mapper, float p_exposure, float p_white, bool p_auto_exposure, float p_min_luminance, float p_max_luminance, float p_auto_exp_speed, float p_auto_exp_scale) { + exposure = p_exposure; + tone_mapper = p_tone_mapper; + if (!auto_exposure && p_auto_exposure) { + auto_exposure_version = ++auto_exposure_counter; + } + auto_exposure = p_auto_exposure; + white = p_white; + min_luminance = p_min_luminance; + max_luminance = p_max_luminance; + auto_exp_speed = p_auto_exp_speed; + auto_exp_scale = p_auto_exp_scale; +} + +void RendererSceneEnvironmentRD::set_glow(bool p_enable, Vector<float> p_levels, float p_intensity, float p_strength, float p_mix, float p_bloom_threshold, RS::EnvironmentGlowBlendMode p_blend_mode, float p_hdr_bleed_threshold, float p_hdr_bleed_scale, float p_hdr_luminance_cap) { + ERR_FAIL_COND_MSG(p_levels.size() != 7, "Size of array of glow levels must be 7"); + glow_enabled = p_enable; + glow_levels = p_levels; + glow_intensity = p_intensity; + glow_strength = p_strength; + glow_mix = p_mix; + glow_bloom = p_bloom_threshold; + glow_blend_mode = p_blend_mode; + glow_hdr_bleed_threshold = p_hdr_bleed_threshold; + glow_hdr_bleed_scale = p_hdr_bleed_scale; + glow_hdr_luminance_cap = p_hdr_luminance_cap; +} + +void RendererSceneEnvironmentRD::set_sdfgi(bool p_enable, RS::EnvironmentSDFGICascades p_cascades, float p_min_cell_size, RS::EnvironmentSDFGIYScale p_y_scale, bool p_use_occlusion, float p_bounce_feedback, bool p_read_sky, float p_energy, float p_normal_bias, float p_probe_bias) { + sdfgi_enabled = p_enable; + sdfgi_cascades = p_cascades; + sdfgi_min_cell_size = p_min_cell_size; + sdfgi_use_occlusion = p_use_occlusion; + sdfgi_bounce_feedback = p_bounce_feedback; + sdfgi_read_sky_light = p_read_sky; + sdfgi_energy = p_energy; + sdfgi_normal_bias = p_normal_bias; + sdfgi_probe_bias = p_probe_bias; + sdfgi_y_scale = p_y_scale; +} + +void RendererSceneEnvironmentRD::set_fog(bool p_enable, const Color &p_light_color, float p_light_energy, float p_sun_scatter, float p_density, float p_height, float p_height_density, float p_fog_aerial_perspective) { + fog_enabled = p_enable; + fog_light_color = p_light_color; + fog_light_energy = p_light_energy; + fog_sun_scatter = p_sun_scatter; + fog_density = p_density; + fog_height = p_height; + fog_height_density = p_height_density; + fog_aerial_perspective = p_fog_aerial_perspective; +} + +void RendererSceneEnvironmentRD::set_volumetric_fog(bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, bool p_temporal_reprojection, float p_temporal_reprojection_amount) { + volumetric_fog_enabled = p_enable; + volumetric_fog_density = p_density; + volumetric_fog_light = p_light; + volumetric_fog_light_energy = p_light_energy; + volumetric_fog_length = p_length; + volumetric_fog_detail_spread = p_detail_spread; + volumetric_fog_gi_inject = p_gi_inject; + volumetric_fog_temporal_reprojection = p_temporal_reprojection; + volumetric_fog_temporal_reprojection_amount = p_temporal_reprojection_amount; +} + +void RendererSceneEnvironmentRD::set_ssr(bool p_enable, int p_max_steps, float p_fade_int, float p_fade_out, float p_depth_tolerance) { + ssr_enabled = p_enable; + ssr_max_steps = p_max_steps; + ssr_fade_in = p_fade_int; + ssr_fade_out = p_fade_out; + ssr_depth_tolerance = p_depth_tolerance; +} + +void RendererSceneEnvironmentRD::set_ssao(bool p_enable, float p_radius, float p_intensity, float p_power, float p_detail, float p_horizon, float p_sharpness, float p_light_affect, float p_ao_channel_affect) { + ssao_enabled = p_enable; + ssao_radius = p_radius; + ssao_intensity = p_intensity; + ssao_power = p_power; + ssao_detail = p_detail; + ssao_horizon = p_horizon; + ssao_sharpness = p_sharpness; + ssao_direct_light_affect = p_light_affect; + ssao_ao_channel_affect = p_ao_channel_affect; +} diff --git a/servers/rendering/renderer_rd/renderer_scene_environment_rd.h b/servers/rendering/renderer_rd/renderer_scene_environment_rd.h new file mode 100644 index 0000000000..992c4bf471 --- /dev/null +++ b/servers/rendering/renderer_rd/renderer_scene_environment_rd.h @@ -0,0 +1,155 @@ +/*************************************************************************/ +/* renderer_scene_environment_rd.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2021 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2021 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 RENDERING_SERVER_SCENE_ENVIRONMENT_RD_H +#define RENDERING_SERVER_SCENE_ENVIRONMENT_RD_H + +#include "servers/rendering/renderer_scene_render.h" +#include "servers/rendering/rendering_device.h" + +class RendererSceneEnvironmentRD { +private: + static uint64_t auto_exposure_counter; + +public: + // BG + RS::EnvironmentBG background = RS::ENV_BG_CLEAR_COLOR; + RID sky; + float sky_custom_fov = 0.0; + Basis sky_orientation; + Color bg_color; + float bg_energy = 1.0; + int canvas_max_layer = 0; + RS::EnvironmentAmbientSource ambient_source = RS::ENV_AMBIENT_SOURCE_BG; + Color ambient_light; + float ambient_light_energy = 1.0; + float ambient_sky_contribution = 1.0; + RS::EnvironmentReflectionSource reflection_source = RS::ENV_REFLECTION_SOURCE_BG; + Color ao_color; + + /// Tonemap + + RS::EnvironmentToneMapper tone_mapper; + float exposure = 1.0; + float white = 1.0; + bool auto_exposure = false; + float min_luminance = 0.2; + float max_luminance = 8.0; + float auto_exp_speed = 0.2; + float auto_exp_scale = 0.5; + uint64_t auto_exposure_version = 0; + + // Fog + bool fog_enabled = false; + Color fog_light_color = Color(0.5, 0.6, 0.7); + float fog_light_energy = 1.0; + float fog_sun_scatter = 0.0; + float fog_density = 0.001; + float fog_height = 0.0; + float fog_height_density = 0.0; //can be negative to invert effect + float fog_aerial_perspective = 0.0; + + /// Volumetric Fog + /// + bool volumetric_fog_enabled = false; + float volumetric_fog_density = 0.01; + Color volumetric_fog_light = Color(0, 0, 0); + float volumetric_fog_light_energy = 0.0; + float volumetric_fog_length = 64.0; + float volumetric_fog_detail_spread = 2.0; + float volumetric_fog_gi_inject = 0.0; + bool volumetric_fog_temporal_reprojection = true; + float volumetric_fog_temporal_reprojection_amount = 0.9; + + /// Glow + + bool glow_enabled = false; + Vector<float> glow_levels; + float glow_intensity = 0.8; + float glow_strength = 1.0; + float glow_bloom = 0.0; + float glow_mix = 0.01; + RS::EnvironmentGlowBlendMode glow_blend_mode = RS::ENV_GLOW_BLEND_MODE_SOFTLIGHT; + float glow_hdr_bleed_threshold = 1.0; + float glow_hdr_luminance_cap = 12.0; + float glow_hdr_bleed_scale = 2.0; + + /// SSAO + + bool ssao_enabled = false; + float ssao_radius = 1.0; + float ssao_intensity = 2.0; + float ssao_power = 1.5; + float ssao_detail = 0.5; + float ssao_horizon = 0.06; + float ssao_sharpness = 0.98; + float ssao_direct_light_affect = 0.0; + float ssao_ao_channel_affect = 0.0; + + /// SSR + /// + bool ssr_enabled = false; + int ssr_max_steps = 64; + float ssr_fade_in = 0.15; + float ssr_fade_out = 2.0; + float ssr_depth_tolerance = 0.2; + + /// SDFGI + bool sdfgi_enabled = false; + RS::EnvironmentSDFGICascades sdfgi_cascades; + float sdfgi_min_cell_size = 0.2; + bool sdfgi_use_occlusion = false; + float sdfgi_bounce_feedback = 0.0; + bool sdfgi_read_sky_light = false; + float sdfgi_energy = 1.0; + float sdfgi_normal_bias = 1.1; + float sdfgi_probe_bias = 1.1; + RS::EnvironmentSDFGIYScale sdfgi_y_scale = RS::ENV_SDFGI_Y_SCALE_DISABLED; + + /// Adjustments + + bool adjustments_enabled = false; + float adjustments_brightness = 1.0f; + float adjustments_contrast = 1.0f; + float adjustments_saturation = 1.0f; + bool use_1d_color_correction = false; + RID color_correction = RID(); + + void set_ambient_light(const Color &p_color, RS::EnvironmentAmbientSource p_ambient, float p_energy, float p_sky_contribution, RS::EnvironmentReflectionSource p_reflection_source, const Color &p_ao_color); + void set_tonemap(RS::EnvironmentToneMapper p_tone_mapper, float p_exposure, float p_white, bool p_auto_exposure, float p_min_luminance, float p_max_luminance, float p_auto_exp_speed, float p_auto_exp_scale); + void set_glow(bool p_enable, Vector<float> p_levels, float p_intensity, float p_strength, float p_mix, float p_bloom_threshold, RS::EnvironmentGlowBlendMode p_blend_mode, float p_hdr_bleed_threshold, float p_hdr_bleed_scale, float p_hdr_luminance_cap); + void set_sdfgi(bool p_enable, RS::EnvironmentSDFGICascades p_cascades, float p_min_cell_size, RS::EnvironmentSDFGIYScale p_y_scale, bool p_use_occlusion, float p_bounce_feedback, bool p_read_sky, float p_energy, float p_normal_bias, float p_probe_bias); + void set_fog(bool p_enable, const Color &p_light_color, float p_light_energy, float p_sun_scatter, float p_density, float p_height, float p_height_density, float p_fog_aerial_perspective); + void set_volumetric_fog(bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, bool p_temporal_reprojection, float p_temporal_reprojection_amount); + void set_ssr(bool p_enable, int p_max_steps, float p_fade_int, float p_fade_out, float p_depth_tolerance); + void set_ssao(bool p_enable, float p_radius, float p_intensity, float p_power, float p_detail, float p_horizon, float p_sharpness, float p_light_affect, float p_ao_channel_affect); +}; + +#endif /* !RENDERING_SERVER_SCENE_ENVIRONMENT_RD_H */ diff --git a/servers/rendering/renderer_rd/renderer_scene_gi_rd.cpp b/servers/rendering/renderer_rd/renderer_scene_gi_rd.cpp new file mode 100644 index 0000000000..4e4e553605 --- /dev/null +++ b/servers/rendering/renderer_rd/renderer_scene_gi_rd.cpp @@ -0,0 +1,3403 @@ +/*************************************************************************/ +/* renderer_scene_gi_rd.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2021 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2021 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_gi_rd.h" + +#include "core/config/project_settings.h" +#include "servers/rendering/renderer_rd/renderer_scene_render_rd.h" +#include "servers/rendering/rendering_server_default.h" + +const Vector3i RendererSceneGIRD::SDFGI::Cascade::DIRTY_ALL = Vector3i(0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF); + +//////////////////////////////////////////////////////////////////////////////// +// SDFGI + +void RendererSceneGIRD::SDFGI::create(RendererSceneEnvironmentRD *p_env, const Vector3 &p_world_position, uint32_t p_requested_history_size, RendererSceneGIRD *p_gi) { + storage = p_gi->storage; + gi = p_gi; + cascade_mode = p_env->sdfgi_cascades; + min_cell_size = p_env->sdfgi_min_cell_size; + uses_occlusion = p_env->sdfgi_use_occlusion; + y_scale_mode = p_env->sdfgi_y_scale; + static const float y_scale[3] = { 1.0, 1.5, 2.0 }; + y_mult = y_scale[y_scale_mode]; + static const int cascasde_size[3] = { 4, 6, 8 }; + cascades.resize(cascasde_size[cascade_mode]); + 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()); + tf_render.format = RD::DATA_FORMAT_R32_UINT; + render_emission = RD::get_singleton()->texture_create(tf_render, RD::TextureView()); + render_emission_aniso = RD::get_singleton()->texture_create(tf_render, RD::TextureView()); + + 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()); + } + + tf_render.format = RD::DATA_FORMAT_R32_UINT; + render_geom_facing = RD::get_singleton()->texture_create(tf_render, RD::TextureView()); + + tf_render.format = RD::DATA_FORMAT_R8G8B8A8_UINT; + render_sdf[0] = RD::get_singleton()->texture_create(tf_render, RD::TextureView()); + render_sdf[1] = RD::get_singleton()->texture_create(tf_render, RD::TextureView()); + + 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()); + render_sdf_half[1] = RD::get_singleton()->texture_create(tf_render, RD::TextureView()); + } + + 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()); + lightprobe_average_scroll = RD::get_singleton()->texture_create(tf_probe_average, RD::TextureView()); + + { + //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::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()); + } + + 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::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()); + + cascade.light_data = RD::get_singleton()->texture_create(tf_light, RD::TextureView()); + + cascade.light_aniso_0_tex = RD::get_singleton()->texture_create(tf_aniso0, RD::TextureView()); + cascade.light_aniso_1_tex = RD::get_singleton()->texture_create(tf_aniso1, RD::TextureView()); + + { + 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()->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()->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.ids.push_back(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.ids.push_back(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.ids.push_back(render_occlusion[j]); + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 4; + u.ids.push_back(render_emission); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 5; + u.ids.push_back(render_emission_aniso); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 6; + u.ids.push_back(render_geom_facing); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 7; + u.ids.push_back(cascade.sdf_tex); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 8; + u.ids.push_back(occlusion_data); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 10; + u.ids.push_back(cascade.solid_cell_dispatch_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 11; + u.ids.push_back(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.ids.push_back(render_albedo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 2; + u.ids.push_back(render_geom_facing); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 3; + u.ids.push_back(render_emission); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 4; + u.ids.push_back(render_emission_aniso); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 5; + u.ids.push_back(cascade.solid_cell_dispatch_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 6; + u.ids.push_back(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.ids.push_back(render_occlusion[j]); + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 2; + u.ids.push_back(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.ids.push_back(cascades[j].sdf_tex); + } else { + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE)); + } + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 2; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.ids.push_back(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.ids.push_back(cascade.solid_cell_dispatch_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 4; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.ids.push_back(cascade.solid_cell_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 5; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.ids.push_back(cascade.light_data); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 6; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.ids.push_back(cascade.light_aniso_0_tex); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 7; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.ids.push_back(cascade.light_aniso_1_tex); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 8; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.ids.push_back(cascades_ubo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 9; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.ids.push_back(cascade.lights_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 10; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.ids.push_back(lightprobe_texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 11; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.ids.push_back(occlusion_texture); + uniforms.push_back(u); + } + + cascade.sdf_direct_light_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->sdfgi_shader.direct_light.version_get_shader(gi->sdfgi_shader.direct_light_shader, 0), 0); + } + + //preprocess initialize uniform set + { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 1; + u.ids.push_back(render_albedo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 2; + u.ids.push_back(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.ids.push_back(render_albedo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 2; + u.ids.push_back(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.ids.push_back(render_sdf[0]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 2; + u.ids.push_back(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); + SWAP(uniforms.write[0].ids.write[0], uniforms.write[1].ids.write[0]); + 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.ids.push_back(render_sdf_half[0]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 2; + u.ids.push_back(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); + SWAP(uniforms.write[0].ids.write[0], uniforms.write[1].ids.write[0]); + 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.ids.push_back(render_albedo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 2; + u.ids.push_back(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.ids.push_back(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.ids.push_back(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.ids.push_back(render_occlusion[i]); + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 3; + u.ids.push_back(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.ids.push_back(cascades[j].sdf_tex); + } else { + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::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.ids.push_back(cascades[j].light_tex); + } else { + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::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.ids.push_back(cascades[j].light_aniso_0_tex); + } else { + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::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.ids.push_back(cascades[j].light_aniso_1_tex); + } else { + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE)); + } + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 6; + u.ids.push_back(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.ids.push_back(cascades_ubo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 8; + u.ids.push_back(lightprobe_data); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 9; + u.ids.push_back(cascades[i].lightprobe_history_tex); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 10; + u.ids.push_back(cascades[i].lightprobe_average_tex); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 11; + u.ids.push_back(lightprobe_history_scroll); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 12; + u.ids.push_back(lightprobe_average_scroll); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 13; + RID parent_average; + 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.ids.push_back(parent_average); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 14; + u.ids.push_back(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 = p_env->sdfgi_bounce_feedback; + energy = p_env->sdfgi_energy; + normal_bias = p_env->sdfgi_normal_bias; + probe_bias = p_env->sdfgi_probe_bias; + reads_sky = p_env->sdfgi_read_sky_light; +} + +void RendererSceneGIRD::SDFGI::erase() { + 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(occlusion_data); + RD::get_singleton()->free(ambient_texture); + + RD::get_singleton()->free(cascades_ubo); +} + +void RendererSceneGIRD::SDFGI::update(RendererSceneEnvironmentRD *p_env, const Vector3 &p_world_position) { + bounce_feedback = p_env->sdfgi_bounce_feedback; + energy = p_env->sdfgi_energy; + normal_bias = p_env->sdfgi_normal_bias; + probe_bias = p_env->sdfgi_probe_bias; + reads_sky = p_env->sdfgi_read_sky_light; + + 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 RendererSceneGIRD::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 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_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 RendererSceneGIRD::SDFGI::update_probes(RendererSceneEnvironmentRD *p_env, RendererSceneSkyRD::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 = p_env->volumetric_fog_enabled; + + 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) { + push_constant.sky_energy = p_env->bg_energy; + + if (p_env->background == RS::ENV_BG_CLEAR_COLOR) { + push_constant.sky_mode = SDFGIShader::IntegratePushConstant::SKY_MODE_COLOR; + Color c = storage->get_default_clear_color().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 (p_env->background == RS::ENV_BG_COLOR) { + push_constant.sky_mode = SDFGIShader::IntegratePushConstant::SKY_MODE_COLOR; + Color c = p_env->bg_color; + push_constant.sky_color[0] = c.r; + push_constant.sky_color[1] = c.g; + push_constant.sky_color[2] = c.b; + + } else if (p_env->background == 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.ids.push_back(p_sky->radiance); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 1; + u.ids.push_back(storage->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 RendererSceneGIRD::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 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 RendererSceneGIRD::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 RendererSceneGIRD::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 RendererSceneGIRD::SDFGI::debug_draw(const CameraMatrix &p_projection, const Transform &p_transform, int p_width, int p_height, RID p_render_target, RID p_texture) { + if (!debug_uniform_set.is_valid() || !RD::get_singleton()->uniform_set_is_valid(debug_uniform_set)) { + 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.ids.push_back(cascades[i].sdf_tex); + } else { + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::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.ids.push_back(cascades[i].light_tex); + } else { + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::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.ids.push_back(cascades[i].light_aniso_0_tex); + } else { + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::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.ids.push_back(cascades[i].light_aniso_1_tex); + } else { + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE)); + } + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 5; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.ids.push_back(occlusion_texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 8; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.ids.push_back(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.ids.push_back(cascades_ubo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 10; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.ids.push_back(p_texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 11; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.ids.push_back(lightprobe_texture); + uniforms.push_back(u); + } + debug_uniform_set = 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, 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.probe_axis_size = probe_axis_count; + push_constant.use_occlusion = uses_occlusion; + push_constant.y_mult = y_mult; + + Vector2 vp_half = p_projection.get_viewport_half_extents(); + push_constant.cam_extent[0] = vp_half.x; + push_constant.cam_extent[1] = vp_half.y; + push_constant.cam_extent[2] = -p_projection.get_z_near(); + + push_constant.cam_transform[0] = p_transform.basis.elements[0][0]; + push_constant.cam_transform[1] = p_transform.basis.elements[1][0]; + push_constant.cam_transform[2] = p_transform.basis.elements[2][0]; + push_constant.cam_transform[3] = 0; + push_constant.cam_transform[4] = p_transform.basis.elements[0][1]; + push_constant.cam_transform[5] = p_transform.basis.elements[1][1]; + push_constant.cam_transform[6] = p_transform.basis.elements[2][1]; + push_constant.cam_transform[7] = 0; + push_constant.cam_transform[8] = p_transform.basis.elements[0][2]; + push_constant.cam_transform[9] = p_transform.basis.elements[1][2]; + push_constant.cam_transform[10] = p_transform.basis.elements[2][2]; + push_constant.cam_transform[11] = 0; + push_constant.cam_transform[12] = p_transform.origin.x; + push_constant.cam_transform[13] = p_transform.origin.y; + push_constant.cam_transform[14] = p_transform.origin.z; + push_constant.cam_transform[15] = 1; + + 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(); + + Size2 rtsize = storage->render_target_get_size(p_render_target); + storage->get_effects()->copy_to_fb_rect(p_texture, storage->render_target_get_rd_framebuffer(p_render_target), Rect2(Vector2(), rtsize), true); +} + +void RendererSceneGIRD::SDFGI::debug_probes(RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform) { + SDFGIShader::DebugProbesPushConstant push_constant; + + for (int i = 0; i < 4; i++) { + for (int j = 0; j < 4; j++) { + push_constant.projection[i * 4 + j] = p_camera_with_transform.matrix[i][j]; + } + } + + //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.ids.push_back(cascades_ubo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 2; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.ids.push_back(lightprobe_texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 3; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.ids.push_back(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.ids.push_back(occlusion_texture); + 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); + } + + RD::get_singleton()->draw_list_bind_render_pipeline(p_draw_list, gi->sdfgi_shader.debug_probes_pipeline[SDFGIShader::PROBE_DEBUG_PROBES].get_render_pipeline(RD::INVALID_FORMAT_ID, RD::get_singleton()->framebuffer_get_format(p_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(p_draw_list, debug_probes_uniform_set, 0); + RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(SDFGIShader::DebugProbesPushConstant)); + RD::get_singleton()->draw_list_draw(p_draw_list, false, total_probes, total_points); + + if (gi->sdfgi_debug_probe_dir != Vector3()) { + print_line("CLICK DEBUG ME?"); + 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); + } + } + } + } + } + + if (gi->sdfgi_debug_probe_enabled) { + print_line("found: " + gi->sdfgi_debug_probe_index); + } else { + print_line("no found"); + } + 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(p_draw_list, gi->sdfgi_shader.debug_probes_pipeline[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(p_draw_list, debug_probes_uniform_set, 0); + RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(SDFGIShader::DebugProbesPushConstant)); + RD::get_singleton()->draw_list_draw(p_draw_list, false, cell_count, total_points); + } +} + +void RendererSceneGIRD::SDFGI::pre_process_gi(const Transform &p_transform, RendererSceneRenderRD *p_scene_render) { + /* 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; + } + + 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_scene_render->render_state.sdfgi_update_data->directional_lights->size(); j++) { + if (idx == SDFGI::MAX_DYNAMIC_LIGHTS) { + break; + } + + RendererSceneRenderRD::LightInstance *li = p_scene_render->light_instance_owner.getornull(p_scene_render->render_state.sdfgi_update_data->directional_lights->get(j)); + ERR_CONTINUE(!li); + + if (storage->light_directional_is_sky_only(li->light)) { + continue; + } + + Vector3 dir = -li->transform.basis.get_axis(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 = storage->light_get_color(li->light); + color = color.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 = storage->light_get_param(li->light, RS::LIGHT_PARAM_ENERGY); + lights[idx].has_shadow = storage->light_has_shadow(li->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_scene_render->render_state.sdfgi_update_data->positional_light_count; j++) { + if (idx == SDFGI::MAX_DYNAMIC_LIGHTS) { + break; + } + + RendererSceneRenderRD::LightInstance *li = p_scene_render->light_instance_owner.getornull(p_scene_render->render_state.sdfgi_update_data->positional_light_instances[j]); + ERR_CONTINUE(!li); + + uint32_t max_sdfgi_cascade = storage->light_get_max_sdfgi_cascade(li->light); + if (i > max_sdfgi_cascade) { + continue; + } + + if (!cascade_aabb.intersects(li->aabb)) { + continue; + } + + Vector3 dir = -li->transform.basis.get_axis(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 = li->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 = storage->light_get_color(li->light); + color = color.to_linear(); + lights[idx].color[0] = color.r; + lights[idx].color[1] = color.g; + lights[idx].color[2] = color.b; + lights[idx].type = storage->light_get_type(li->light); + lights[idx].energy = storage->light_get_param(li->light, RS::LIGHT_PARAM_ENERGY); + lights[idx].has_shadow = storage->light_has_shadow(li->light); + lights[idx].attenuation = storage->light_get_param(li->light, RS::LIGHT_PARAM_ATTENUATION); + lights[idx].radius = storage->light_get_param(li->light, RS::LIGHT_PARAM_RANGE); + lights[idx].cos_spot_angle = Math::cos(Math::deg2rad(storage->light_get_param(li->light, RS::LIGHT_PARAM_SPOT_ANGLE))); + lights[idx].inv_spot_attenuation = 1.0f / storage->light_get_param(li->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 RendererSceneGIRD::SDFGI::render_region(RID p_render_buffers, int p_region, const PagedArray<RendererSceneRender::GeometryInstance *> &p_instances, RendererSceneRenderRD *p_scene_render) { + //print_line("rendering region " + itos(p_region)); + RendererSceneRenderRD::RenderBuffers *rb = p_scene_render->render_buffers_owner.getornull(p_render_buffers); + ERR_FAIL_COND(!rb); // 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)); + p_scene_render->_render_sdfgi(p_render_buffers, from, size, bounds, p_instances, render_albedo, render_emission, render_emission_aniso, render_geom_facing); + + 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; + zeromem(&push_constant, sizeof(SDFGIShader::PreprocessPushConstant)); + + RENDER_TIMESTAMP("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; + + 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"); + + 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"); + + //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.instance(); + for (uint32_t i = 0; i < cascade_size; i++) { + Vector<uint8_t> subarr = data.subarray(128 * 128 * i, 128 * 128 * (i + 1) - 1); + img->create(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.instance(); + for (uint32_t i = 0; i < cascade_size; i++) { + Vector<uint8_t> subarr = data.subarray(128 * 128 * i * 2, 128 * 128 * (i + 1) * 2 - 1); + 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 RendererSceneGIRD::SDFGI::render_static_lights(RID p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const PagedArray<RID> *p_positional_light_cull_result, RendererSceneRenderRD *p_scene_render) { + RendererSceneRenderRD::RenderBuffers *rb = p_scene_render->render_buffers_owner.getornull(p_render_buffers); + ERR_FAIL_COND(!rb); // we wouldn't be here if this failed but... + + RD::get_singleton()->draw_command_begin_label("SDFGI Render Static Lighs"); + + update_cascades(); + ; //need cascades updated for this + + 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; + } + + RendererSceneRenderRD::LightInstance *li = p_scene_render->light_instance_owner.getornull(p_positional_light_cull_result[i][j]); + ERR_CONTINUE(!li); + + uint32_t max_sdfgi_cascade = storage->light_get_max_sdfgi_cascade(li->light); + if (p_cascade_indices[i] > max_sdfgi_cascade) { + continue; + } + + if (!cascade_aabb.intersects(li->aabb)) { + continue; + } + + lights[idx].type = storage->light_get_type(li->light); + + Vector3 dir = -li->transform.basis.get_axis(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 = li->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 = storage->light_get_color(li->light); + color = color.to_linear(); + lights[idx].color[0] = color.r; + lights[idx].color[1] = color.g; + lights[idx].color[2] = color.b; + lights[idx].energy = storage->light_get_param(li->light, RS::LIGHT_PARAM_ENERGY); + lights[idx].has_shadow = storage->light_has_shadow(li->light); + lights[idx].attenuation = storage->light_get_param(li->light, RS::LIGHT_PARAM_ATTENUATION); + lights[idx].radius = storage->light_get_param(li->light, RS::LIGHT_PARAM_RANGE); + lights[idx].cos_spot_angle = Math::cos(Math::deg2rad(storage->light_get_param(li->light, RS::LIGHT_PARAM_SPOT_ANGLE))); + lights[idx].inv_spot_attenuation = 1.0f / storage->light_get_param(li->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_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(); +} + +//////////////////////////////////////////////////////////////////////////////// +// GIProbeInstance + +void RendererSceneGIRD::GIProbeInstance::update(bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RendererSceneRender::GeometryInstance *> &p_dynamic_objects, RendererSceneRenderRD *p_scene_render) { + uint32_t data_version = storage->gi_probe_get_data_version(probe); + + // (RE)CREATE IF NEEDED + + if (last_probe_data_version != data_version) { + //need to re-create everything + if (texture.is_valid()) { + RD::get_singleton()->free(texture); + RD::get_singleton()->free(write_buffer); + 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); + } + + dynamic_maps.clear(); + + Vector3i octree_size = storage->gi_probe_get_octree_size(probe); + + if (octree_size != Vector3i()) { + //can create a 3D texture + Vector<int> levels = storage->gi_probe_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()->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++) { + GIProbeInstance::Mipmap mipmap; + mipmap.texture = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), texture, 0, i, 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.ids.push_back(storage->gi_probe_get_octree_buffer(probe)); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 2; + u.ids.push_back(storage->gi_probe_get_data_buffer(probe)); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 4; + u.ids.push_back(write_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 9; + u.ids.push_back(storage->gi_probe_get_sdf_texture(probe)); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 10; + u.ids.push_back(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.ids.push_back(gi->gi_probe_lights_uniform); + copy_uniforms.push_back(u); + } + + mipmap.uniform_set = RD::get_singleton()->uniform_set_create(copy_uniforms, gi->giprobe_lighting_shader_version_shaders[GI_PROBE_SHADER_VERSION_COMPUTE_LIGHT], 0); + + copy_uniforms = uniforms; //restore + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 5; + u.ids.push_back(texture); + copy_uniforms.push_back(u); + } + mipmap.second_bounce_uniform_set = RD::get_singleton()->uniform_set_create(copy_uniforms, gi->giprobe_lighting_shader_version_shaders[GI_PROBE_SHADER_VERSION_COMPUTE_SECOND_BOUNCE], 0); + } else { + mipmap.uniform_set = RD::get_singleton()->uniform_set_create(copy_uniforms, gi->giprobe_lighting_shader_version_shaders[GI_PROBE_SHADER_VERSION_COMPUTE_MIPMAP], 0); + } + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 5; + u.ids.push_back(mipmap.texture); + uniforms.push_back(u); + } + + mipmap.write_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->giprobe_lighting_shader_version_shaders[GI_PROBE_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()) { + GIProbeInstance::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()); + + if (dynamic_maps.size() == 0) { + //render depth for first one + dtf.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; + dtf.usage_bits = RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT; + dmap.fb_depth = RD::get_singleton()->texture_create(dtf, RD::TextureView()); + } + + //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()); + + 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()); + dmap.normal = RD::get_singleton()->texture_create(dtf, RD::TextureView()); + dmap.orm = RD::get_singleton()->texture_create(dtf, RD::TextureView()); + + 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.ids.push_back(gi->gi_probe_lights_uniform); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 5; + u.ids.push_back(dmap.albedo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 6; + u.ids.push_back(dmap.normal); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 7; + u.ids.push_back(dmap.orm); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 8; + u.ids.push_back(dmap.fb_depth); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 9; + u.ids.push_back(storage->gi_probe_get_sdf_texture(probe)); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 10; + u.ids.push_back(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.ids.push_back(dmap.texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 12; + u.ids.push_back(dmap.depth); + uniforms.push_back(u); + } + + dmap.uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->giprobe_lighting_shader_version_shaders[GI_PROBE_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.ids.push_back(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.ids.push_back(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.ids.push_back(dmap.texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 8; + u.ids.push_back(dmap.depth); + uniforms.push_back(u); + } + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 9; + u.ids.push_back(storage->gi_probe_get_sdf_texture(probe)); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 10; + u.ids.push_back(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.ids.push_back(mipmaps[dmap.mipmap].texture); + uniforms.push_back(u); + } + } + + dmap.uniform_set = RD::get_singleton()->uniform_set_create( + uniforms, + gi->giprobe_lighting_shader_version_shaders[(write && plot) ? GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE_PLOT : (write ? GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE : GI_PROBE_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 + + p_scene_render->_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->gi_probe_max_lights, (uint32_t)p_light_instances.size()); + + { + Transform to_cell = storage->gi_probe_get_to_cell_xform(probe); + Transform to_probe_xform = (transform * to_cell.affine_inverse()).affine_inverse(); + //update lights + + for (uint32_t i = 0; i < light_count; i++) { + GIProbeLight &l = gi->gi_probe_lights[i]; + RID light_instance = p_light_instances[i]; + RID light = p_scene_render->light_instance_get_base_light(light_instance); + + l.type = storage->light_get_type(light); + if (l.type == RS::LIGHT_DIRECTIONAL && storage->light_directional_is_sky_only(light)) { + light_count--; + continue; + } + + l.attenuation = storage->light_get_param(light, RS::LIGHT_PARAM_ATTENUATION); + l.energy = storage->light_get_param(light, RS::LIGHT_PARAM_ENERGY) * storage->light_get_param(light, RS::LIGHT_PARAM_INDIRECT_ENERGY); + l.radius = to_cell.basis.xform(Vector3(storage->light_get_param(light, RS::LIGHT_PARAM_RANGE), 0, 0)).length(); + Color color = storage->light_get_color(light).to_linear(); + l.color[0] = color.r; + l.color[1] = color.g; + l.color[2] = color.b; + + l.cos_spot_angle = Math::cos(Math::deg2rad(storage->light_get_param(light, RS::LIGHT_PARAM_SPOT_ANGLE))); + l.inv_spot_attenuation = 1.0f / storage->light_get_param(light, RS::LIGHT_PARAM_SPOT_ATTENUATION); + + Transform xform = p_scene_render->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_axis(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 = storage->light_has_shadow(light); + } + + RD::get_singleton()->buffer_update(gi->gi_probe_lights_uniform, 0, sizeof(GIProbeLight) * light_count, gi->gi_probe_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 = storage->gi_probe_get_octree_size(probe); + + GIProbePushConstant 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 = storage->gi_probe_get_propagation(probe); + push_constant.dynamic_range = storage->gi_probe_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 = storage->gi_probe_is_using_two_bounces(probe) ? 2 : 1; + } else { + passes = 1; //only re-blitting is necessary + } + int wg_size = 64; + int wg_limit_x = 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->giprobe_lighting_shader_version_pipelines[pass == 0 ? GI_PROBE_SHADER_VERSION_COMPUTE_LIGHT : GI_PROBE_SHADER_VERSION_COMPUTE_SECOND_BOUNCE]); + } else if (i == 1) { + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->giprobe_lighting_shader_version_pipelines[GI_PROBE_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; + + int wg_todo = (mipmaps[i].cell_count - 1) / wg_size + 1; + while (wg_todo) { + int wg_count = MIN(wg_todo, wg_limit_x); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(GIProbePushConstant)); + 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->giprobe_lighting_shader_version_pipelines[GI_PROBE_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; + + int wg_todo = (mipmaps[i].cell_count - 1) / wg_size + 1; + while (wg_todo) { + int wg_count = MIN(wg_todo, wg_limit_x); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(GIProbePushConstant)); + 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 = storage->gi_probe_get_octree_size(probe); + int multiplier = dynamic_maps[0].size / MAX(MAX(octree_size.x, octree_size.y), octree_size.z); + + Transform oversample_scale; + oversample_scale.basis.scale(Vector3(multiplier, multiplier, multiplier)); + + Transform to_cell = oversample_scale * storage->gi_probe_get_to_cell_xform(probe); + Transform to_world_xform = transform * to_cell.affine_inverse(); + Transform 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++) { + RendererSceneRender::GeometryInstance *instance = p_dynamic_objects[i]; + + //transform aabb to giprobe + AABB aabb = (to_probe_xform * p_scene_render->geometry_instance_get_transform(instance)).xform(p_scene_render->geometry_instance_get_aabb(instance)); + + //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.position + aabb.size * 0.5; + Transform xform; + xform.set_look_at(center - aabb.size * 0.5 * render_dir, center, up_dir); + + Vector3 x_dir = xform.basis.get_axis(0).abs(); + int x_axis = int(Vector3(0, 1, 2).dot(x_dir)); + Vector3 y_dir = xform.basis.get_axis(1).abs(); + int y_axis = int(Vector3(0, 1, 2).dot(y_dir)); + Vector3 z_dir = -xform.basis.get_axis(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_axis(0)) < 0); + bool y_flip = bool(Vector3(1, 1, 1).dot(xform.basis.get_axis(1)) < 0); + bool z_flip = bool(Vector3(1, 1, 1).dot(xform.basis.get_axis(2)) > 0); + + CameraMatrix 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 (p_scene_render->cull_argument.size() == 0) { + p_scene_render->cull_argument.push_back(nullptr); + } + p_scene_render->cull_argument[0] = instance; + + p_scene_render->_render_material(to_world_xform * xform, cm, true, p_scene_render->cull_argument, dynamic_maps[0].fb, Rect2i(Vector2i(), rect.size)); + + GIProbeDynamicPushConstant push_constant; + zeromem(&push_constant, sizeof(GIProbeDynamicPushConstant)); + 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 = storage->gi_probe_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 = storage->gi_probe_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->giprobe_lighting_shader_version_pipelines[GI_PROBE_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(GIProbeDynamicPushConstant)); + 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->giprobe_lighting_shader_version_pipelines[GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE]); + } else if (k < dynamic_maps.size() - 1) { + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->giprobe_lighting_shader_version_pipelines[GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE_PLOT]); + } else { + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi->giprobe_lighting_shader_version_pipelines[GI_PROBE_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(GIProbeDynamicPushConstant)); + 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 = storage->gi_probe_get_version(probe); +} + +void RendererSceneGIRD::GIProbeInstance::debug(RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform, bool p_lighting, bool p_emission, float p_alpha) { + if (mipmaps.size() == 0) { + return; + } + + CameraMatrix cam_transform = (p_camera_with_transform * CameraMatrix(transform)) * CameraMatrix(storage->gi_probe_get_to_cell_xform(probe).affine_inverse()); + + int level = 0; + Vector3i octree_size = storage->gi_probe_get_octree_size(probe); + + GIProbeDebugPushConstant push_constant; + push_constant.alpha = p_alpha; + push_constant.dynamic_range = storage->gi_probe_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.matrix[i][j]; + } + } + + if (gi->giprobe_debug_uniform_set.is_valid()) { + RD::get_singleton()->free(gi->giprobe_debug_uniform_set); + } + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 1; + u.ids.push_back(storage->gi_probe_get_data_buffer(probe)); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 2; + u.ids.push_back(texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 3; + u.ids.push_back(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->giprobe_debug_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi->giprobe_debug_shader_version_shaders[0], 0); + + int giprobe_debug_pipeline = GI_PROBE_DEBUG_COLOR; + if (p_emission) { + giprobe_debug_pipeline = GI_PROBE_DEBUG_EMISSION; + } else if (p_lighting) { + giprobe_debug_pipeline = has_dynamic_object_data ? GI_PROBE_DEBUG_LIGHT_FULL : GI_PROBE_DEBUG_LIGHT; + } + RD::get_singleton()->draw_list_bind_render_pipeline( + p_draw_list, + gi->giprobe_debug_shader_version_pipelines[giprobe_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->giprobe_debug_uniform_set, 0); + RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(GIProbeDebugPushConstant)); + RD::get_singleton()->draw_list_draw(p_draw_list, false, cell_count, 36); +} + +//////////////////////////////////////////////////////////////////////////////// +// GIRD + +RendererSceneGIRD::RendererSceneGIRD() { + 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))); +} + +RendererSceneGIRD::~RendererSceneGIRD() { +} + +void RendererSceneGIRD::init(RendererStorageRD *p_storage, RendererSceneSkyRD *p_sky) { + storage = p_storage; + + /* GI */ + + { + //kinda complicated to compute the amount of slots, we try to use as many as we can + + gi_probe_max_lights = 32; + + gi_probe_lights = memnew_arr(GIProbeLight, gi_probe_max_lights); + gi_probe_lights_uniform = RD::get_singleton()->uniform_buffer_create(gi_probe_max_lights * sizeof(GIProbeLight)); + gi_probe_quality = RS::GIProbeQuality(CLAMP(int(GLOBAL_GET("rendering/global_illumination/gi_probes/quality")), 0, 1)); + + String defines = "\n#define MAX_LIGHTS " + itos(gi_probe_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"); + + giprobe_shader.initialize(versions, defines); + giprobe_lighting_shader_version = giprobe_shader.version_create(); + for (int i = 0; i < GI_PROBE_SHADER_VERSION_MAX; i++) { + giprobe_lighting_shader_version_shaders[i] = giprobe_shader.version_get_shader(giprobe_lighting_shader_version, i); + giprobe_lighting_shader_version_pipelines[i] = RD::get_singleton()->compute_pipeline_create(giprobe_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"); + + giprobe_debug_shader.initialize(versions, defines); + giprobe_debug_shader_version = giprobe_debug_shader.version_create(); + for (int i = 0; i < GI_PROBE_DEBUG_MAX; i++) { + giprobe_debug_shader_version_shaders[i] = giprobe_debug_shader.version_get_shader(giprobe_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; + + giprobe_debug_shader_version_pipelines[i].setup(giprobe_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; + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_WHITE)); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 1; + u.ids.push_back(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_GIPROBES\n"); + gi_modes.push_back("\n#define USE_SDFGI\n"); + gi_modes.push_back("\n#define USE_SDFGI\n\n#define USE_GIPROBES\n"); + gi_modes.push_back("\n#define MODE_HALF_RES\n#define USE_GIPROBES\n"); + gi_modes.push_back("\n#define MODE_HALF_RES\n#define USE_SDFGI\n"); + gi_modes.push_back("\n#define MODE_HALF_RES\n#define USE_SDFGI\n\n#define USE_GIPROBES\n"); + + shader.initialize(gi_modes, defines); + shader_version = shader.version_create(); + for (int i = 0; i < MODE_MAX; i++) { + pipelines[i] = RD::get_singleton()->compute_pipeline_create(shader.version_get_shader(shader_version, i)); + } + + 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_VISIBILITY\n"); + + sdfgi_shader.debug_probes.initialize(versions, defines); + 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++) { + 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_giprobe_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(GIProbeData) * MAX_GIPROBES); +} + +void RendererSceneGIRD::free() { + RD::get_singleton()->free(default_giprobe_buffer); + RD::get_singleton()->free(gi_probe_lights_uniform); + RD::get_singleton()->free(sdfgi_ubo); + + giprobe_debug_shader.version_free(giprobe_debug_shader_version); + giprobe_shader.version_free(giprobe_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); + + memdelete_arr(gi_probe_lights); +} + +RendererSceneGIRD::SDFGI *RendererSceneGIRD::create_sdfgi(RendererSceneEnvironmentRD *p_env, const Vector3 &p_world_position, uint32_t p_requested_history_size) { + SDFGI *sdfgi = memnew(SDFGI); + + sdfgi->create(p_env, p_world_position, p_requested_history_size, this); + + return sdfgi; +} + +void RendererSceneGIRD::setup_giprobes(RID p_render_buffers, const Transform &p_transform, const PagedArray<RID> &p_gi_probes, uint32_t &r_gi_probes_used, RendererSceneRenderRD *p_scene_render) { + r_gi_probes_used = 0; + + // feels a little dirty to use our container this way but.... + RendererSceneRenderRD::RenderBuffers *rb = p_scene_render->render_buffers_owner.getornull(p_render_buffers); + ERR_FAIL_COND(rb == nullptr); + + RID gi_probe_buffer = p_scene_render->render_buffers_get_gi_probe_buffer(p_render_buffers); + + RD::get_singleton()->draw_command_begin_label("GIProbes Setup"); + + GIProbeData gi_probe_data[MAX_GIPROBES]; + + bool giprobes_changed = false; + + Transform to_camera; + to_camera.origin = p_transform.origin; //only translation, make local + + for (int i = 0; i < MAX_GIPROBES; i++) { + RID texture; + if (i < (int)p_gi_probes.size()) { + GIProbeInstance *gipi = get_probe_instance(p_gi_probes[i]); + + if (gipi) { + texture = gipi->texture; + GIProbeData &gipd = gi_probe_data[i]; + + RID base_probe = gipi->probe; + + Transform to_cell = storage->gi_probe_get_to_cell_xform(gipi->probe) * gipi->transform.affine_inverse() * to_camera; + + gipd.xform[0] = to_cell.basis.elements[0][0]; + gipd.xform[1] = to_cell.basis.elements[1][0]; + gipd.xform[2] = to_cell.basis.elements[2][0]; + gipd.xform[3] = 0; + gipd.xform[4] = to_cell.basis.elements[0][1]; + gipd.xform[5] = to_cell.basis.elements[1][1]; + gipd.xform[6] = to_cell.basis.elements[2][1]; + gipd.xform[7] = 0; + gipd.xform[8] = to_cell.basis.elements[0][2]; + gipd.xform[9] = to_cell.basis.elements[1][2]; + gipd.xform[10] = to_cell.basis.elements[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 = storage->gi_probe_get_octree_size(base_probe); + + gipd.bounds[0] = bounds.x; + gipd.bounds[1] = bounds.y; + gipd.bounds[2] = bounds.z; + + gipd.dynamic_range = storage->gi_probe_get_dynamic_range(base_probe) * storage->gi_probe_get_energy(base_probe); + gipd.bias = storage->gi_probe_get_bias(base_probe); + gipd.normal_bias = storage->gi_probe_get_normal_bias(base_probe); + gipd.blend_ambient = !storage->gi_probe_is_interior(base_probe); + gipd.anisotropy_strength = 0; + gipd.ao = storage->gi_probe_get_ao(base_probe); + gipd.ao_size = Math::pow(storage->gi_probe_get_ao_size(base_probe), 4.0f); + gipd.mipmaps = gipi->mipmaps.size(); + } + + r_gi_probes_used++; + } + + if (texture == RID()) { + texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE); + } + + if (texture != rb->gi.giprobe_textures[i]) { + giprobes_changed = true; + rb->gi.giprobe_textures[i] = texture; + } + } + + if (giprobes_changed) { + if (RD::get_singleton()->uniform_set_is_valid(rb->gi_uniform_set)) { + RD::get_singleton()->free(rb->gi_uniform_set); + } + rb->gi_uniform_set = RID(); + if (rb->volumetric_fog) { + if (RD::get_singleton()->uniform_set_is_valid(rb->volumetric_fog->uniform_set)) { + RD::get_singleton()->free(rb->volumetric_fog->uniform_set); + RD::get_singleton()->free(rb->volumetric_fog->uniform_set2); + } + rb->volumetric_fog->uniform_set = RID(); + rb->volumetric_fog->uniform_set2 = RID(); + } + } + + if (p_gi_probes.size() > 0) { + RD::get_singleton()->buffer_update(gi_probe_buffer, 0, sizeof(GIProbeData) * MIN((uint64_t)MAX_GIPROBES, p_gi_probes.size()), gi_probe_data, RD::BARRIER_MASK_COMPUTE); + } + + RD::get_singleton()->draw_command_end_label(); +} + +void RendererSceneGIRD::process_gi(RID p_render_buffers, RID p_normal_roughness_buffer, RID p_gi_probe_buffer, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform, const PagedArray<RID> &p_gi_probes, RendererSceneRenderRD *p_scene_render) { + RD::get_singleton()->draw_command_begin_label("GI Render"); + + RendererSceneRenderRD::RenderBuffers *rb = p_scene_render->render_buffers_owner.getornull(p_render_buffers); + ERR_FAIL_COND(rb == nullptr); + RendererSceneEnvironmentRD *env = p_scene_render->environment_owner.getornull(p_environment); + + if (rb->ambient_buffer.is_null() || rb->using_half_size_gi != half_resolution) { + if (rb->ambient_buffer.is_valid()) { + RD::get_singleton()->free(rb->ambient_buffer); + RD::get_singleton()->free(rb->reflection_buffer); + } + + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.width = rb->width; + tf.height = rb->height; + if (half_resolution) { + tf.width >>= 1; + tf.height >>= 1; + } + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + rb->reflection_buffer = RD::get_singleton()->texture_create(tf, RD::TextureView()); + rb->ambient_buffer = RD::get_singleton()->texture_create(tf, RD::TextureView()); + rb->using_half_size_gi = half_resolution; + + p_scene_render->_render_buffers_uniform_set_changed(p_render_buffers); + } + + PushConstant push_constant; + + push_constant.screen_size[0] = rb->width; + push_constant.screen_size[1] = rb->height; + push_constant.z_near = p_projection.get_z_near(); + push_constant.z_far = p_projection.get_z_far(); + push_constant.orthogonal = p_projection.is_orthogonal(); + push_constant.proj_info[0] = -2.0f / (rb->width * p_projection.matrix[0][0]); + push_constant.proj_info[1] = -2.0f / (rb->height * p_projection.matrix[1][1]); + push_constant.proj_info[2] = (1.0f - p_projection.matrix[0][2]) / p_projection.matrix[0][0]; + push_constant.proj_info[3] = (1.0f + p_projection.matrix[1][2]) / p_projection.matrix[1][1]; + push_constant.max_giprobes = MIN((uint64_t)MAX_GIPROBES, p_gi_probes.size()); + push_constant.high_quality_vct = gi_probe_quality == RS::GI_PROBE_QUALITY_HIGH; + + bool use_sdfgi = rb->sdfgi != nullptr; + bool use_giprobes = push_constant.max_giprobes > 0; + + if (env) { + push_constant.ao_color[0] = env->ao_color.r; + push_constant.ao_color[1] = env->ao_color.g; + push_constant.ao_color[2] = env->ao_color.b; + } else { + push_constant.ao_color[0] = 0; + push_constant.ao_color[1] = 0; + push_constant.ao_color[2] = 0; + } + + push_constant.cam_rotation[0] = p_transform.basis[0][0]; + push_constant.cam_rotation[1] = p_transform.basis[1][0]; + push_constant.cam_rotation[2] = p_transform.basis[2][0]; + push_constant.cam_rotation[3] = 0; + push_constant.cam_rotation[4] = p_transform.basis[0][1]; + push_constant.cam_rotation[5] = p_transform.basis[1][1]; + push_constant.cam_rotation[6] = p_transform.basis[2][1]; + push_constant.cam_rotation[7] = 0; + push_constant.cam_rotation[8] = p_transform.basis[0][2]; + push_constant.cam_rotation[9] = p_transform.basis[1][2]; + push_constant.cam_rotation[10] = p_transform.basis[2][2]; + push_constant.cam_rotation[11] = 0; + + if (rb->gi_uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(rb->gi_uniform_set)) { + 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 (rb->sdfgi && j < rb->sdfgi->cascades.size()) { + u.ids.push_back(rb->sdfgi->cascades[j].sdf_tex); + } else { + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::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 (rb->sdfgi && j < rb->sdfgi->cascades.size()) { + u.ids.push_back(rb->sdfgi->cascades[j].light_tex); + } else { + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::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 (rb->sdfgi && j < rb->sdfgi->cascades.size()) { + u.ids.push_back(rb->sdfgi->cascades[j].light_aniso_0_tex); + } else { + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::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 (rb->sdfgi && j < rb->sdfgi->cascades.size()) { + u.ids.push_back(rb->sdfgi->cascades[j].light_aniso_1_tex); + } else { + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE)); + } + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 5; + if (rb->sdfgi) { + u.ids.push_back(rb->sdfgi->occlusion_texture); + } else { + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE)); + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 6; + u.ids.push_back(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.ids.push_back(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.ids.push_back(rb->ambient_buffer); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 10; + u.ids.push_back(rb->reflection_buffer); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 11; + if (rb->sdfgi) { + u.ids.push_back(rb->sdfgi->lightprobe_texture); + } else { + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE)); + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 12; + u.ids.push_back(rb->depth_texture); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 13; + u.ids.push_back(p_normal_roughness_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 14; + RID buffer = p_gi_probe_buffer.is_valid() ? p_gi_probe_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); + u.ids.push_back(buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 15; + u.ids.push_back(sdfgi_ubo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 16; + u.ids.push_back(rb->gi.giprobe_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 17; + for (int i = 0; i < MAX_GIPROBES; i++) { + u.ids.push_back(rb->gi.giprobe_textures[i]); + } + uniforms.push_back(u); + } + + rb->gi_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, shader.version_get_shader(shader_version, 0), 0); + } + + Mode mode; + + if (rb->using_half_size_gi) { + mode = (use_sdfgi && use_giprobes) ? MODE_HALF_RES_COMBINED : (use_sdfgi ? MODE_HALF_RES_SDFGI : MODE_HALF_RES_GIPROBE); + } else { + mode = (use_sdfgi && use_giprobes) ? MODE_COMBINED : (use_sdfgi ? MODE_SDFGI : MODE_GIPROBE); + } + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(true); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, pipelines[mode]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->gi_uniform_set, 0); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(PushConstant)); + + if (rb->using_half_size_gi) { + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->width >> 1, rb->height >> 1, 1); + } else { + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->width, rb->height, 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 RendererSceneGIRD::gi_probe_instance_create(RID p_base) { + GIProbeInstance gi_probe; + gi_probe.gi = this; + gi_probe.storage = storage; + gi_probe.probe = p_base; + RID rid = gi_probe_instance_owner.make_rid(gi_probe); + return rid; +} + +void RendererSceneGIRD::gi_probe_instance_set_transform_to_data(RID p_probe, const Transform &p_xform) { + GIProbeInstance *gi_probe = get_probe_instance(p_probe); + ERR_FAIL_COND(!gi_probe); + + gi_probe->transform = p_xform; +} + +bool RendererSceneGIRD::gi_probe_needs_update(RID p_probe) const { + GIProbeInstance *gi_probe = get_probe_instance(p_probe); + ERR_FAIL_COND_V(!gi_probe, false); + + return gi_probe->last_probe_version != storage->gi_probe_get_version(gi_probe->probe); +} + +void RendererSceneGIRD::gi_probe_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RendererSceneRender::GeometryInstance *> &p_dynamic_objects, RendererSceneRenderRD *p_scene_render) { + GIProbeInstance *gi_probe = get_probe_instance(p_probe); + ERR_FAIL_COND(!gi_probe); + + gi_probe->update(p_update_light_instances, p_light_instances, p_dynamic_objects, p_scene_render); +} + +void RendererSceneGIRD::debug_giprobe(RID p_gi_probe, RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform, bool p_lighting, bool p_emission, float p_alpha) { + GIProbeInstance *gi_probe = gi_probe_instance_owner.getornull(p_gi_probe); + ERR_FAIL_COND(!gi_probe); + + gi_probe->debug(p_draw_list, p_framebuffer, p_camera_with_transform, p_lighting, p_emission, p_alpha); +} diff --git a/servers/rendering/renderer_rd/renderer_scene_gi_rd.h b/servers/rendering/renderer_rd/renderer_scene_gi_rd.h new file mode 100644 index 0000000000..c0f3318538 --- /dev/null +++ b/servers/rendering/renderer_rd/renderer_scene_gi_rd.h @@ -0,0 +1,668 @@ +/*************************************************************************/ +/* renderer_scene_gi_rd.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2021 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2021 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 RENDERING_SERVER_SCENE_GI_RD_H +#define RENDERING_SERVER_SCENE_GI_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/renderer_scene_environment_rd.h" +#include "servers/rendering/renderer_rd/renderer_scene_sky_rd.h" +#include "servers/rendering/renderer_rd/renderer_storage_rd.h" +#include "servers/rendering/renderer_rd/shaders/gi.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/giprobe.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/giprobe_debug.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/sdfgi_debug_probes.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/sdfgi_preprocess.glsl.gen.h" +#include "servers/rendering/renderer_scene_render.h" +#include "servers/rendering/rendering_device.h" + +// Forward declare RendererSceneRenderRD so we can pass it into some of our methods, these classes are pretty tightly bound +class RendererSceneRenderRD; + +class RendererSceneGIRD { +private: + RendererStorageRD *storage; + + /* GIPROBE INSTANCE */ + + struct GIProbeLight { + 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 GIProbePushConstant { + 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 GIProbeDynamicPushConstant { + 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]; + }; + + GIProbeLight *gi_probe_lights; + uint32_t gi_probe_max_lights; + RID gi_probe_lights_uniform; + + enum { + GI_PROBE_SHADER_VERSION_COMPUTE_LIGHT, + GI_PROBE_SHADER_VERSION_COMPUTE_SECOND_BOUNCE, + GI_PROBE_SHADER_VERSION_COMPUTE_MIPMAP, + GI_PROBE_SHADER_VERSION_WRITE_TEXTURE, + GI_PROBE_SHADER_VERSION_DYNAMIC_OBJECT_LIGHTING, + GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE, + GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_PLOT, + GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE_PLOT, + GI_PROBE_SHADER_VERSION_MAX + }; + + GiprobeShaderRD giprobe_shader; + RID giprobe_lighting_shader_version; + RID giprobe_lighting_shader_version_shaders[GI_PROBE_SHADER_VERSION_MAX]; + RID giprobe_lighting_shader_version_pipelines[GI_PROBE_SHADER_VERSION_MAX]; + + enum { + GI_PROBE_DEBUG_COLOR, + GI_PROBE_DEBUG_LIGHT, + GI_PROBE_DEBUG_EMISSION, + GI_PROBE_DEBUG_LIGHT_FULL, + GI_PROBE_DEBUG_MAX + }; + + struct GIProbeDebugPushConstant { + float projection[16]; + uint32_t cell_offset; + float dynamic_range; + float alpha; + uint32_t level; + int32_t bounds[3]; + uint32_t pad; + }; + + GiprobeDebugShaderRD giprobe_debug_shader; + RID giprobe_debug_shader_version; + RID giprobe_debug_shader_version_shaders[GI_PROBE_DEBUG_MAX]; + PipelineCacheRD giprobe_debug_shader_version_pipelines[GI_PROBE_DEBUG_MAX]; + RID giprobe_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]; + uint32_t use_occlusion; + float y_mult; + + float cam_extent[3]; + uint32_t probe_axis_size; + + float cam_transform[16]; + }; + + SdfgiDebugShaderRD debug; + RID debug_shader; + RID debug_shader_version; + RID debug_pipeline; + + enum ProbeDebugMode { + PROBE_DEBUG_PROBES, + PROBE_DEBUG_VISIBILITY, + PROBE_DEBUG_MAX + }; + + struct DebugProbesPushConstant { + float projection[16]; + + 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; + + float shadow_color[4]; + }; + + 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: + /* GIPROBE INSTANCE */ + + //@TODO GIProbeInstance is still directly used in the render code, we'll address this when we refactor the render code itself. + + struct GIProbeInstance { + // access to our containers + RendererStorageRD *storage; + RendererSceneGIRD *gi; + + 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; + + Transform transform; + + void update(bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RendererSceneRender::GeometryInstance *> &p_dynamic_objects, RendererSceneRenderRD *p_scene_render); + void debug(RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform, bool p_lighting, bool p_emission, float p_alpha); + }; + + mutable RID_Owner<GIProbeInstance> gi_probe_instance_owner; + + _FORCE_INLINE_ GIProbeInstance *get_probe_instance(RID p_probe) const { + return gi_probe_instance_owner.getornull(p_probe); + }; + + _FORCE_INLINE_ RID gi_probe_instance_get_texture(RID p_probe) { + GIProbeInstance *gi_probe = get_probe_instance(p_probe); + ERR_FAIL_COND_V(!gi_probe, RID()); + return gi_probe->texture; + }; + + RS::GIProbeQuality gi_probe_quality = RS::GI_PROBE_QUALITY_HIGH; + + /* SDFGI */ + + struct SDFGI { + 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; + }; + + //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_uniform_set; + RID scroll_uniform_set; + RID scroll_occlusion_uniform_set; + RID integrate_uniform_set; + RID lights_buffer; + + bool all_dynamic_lights_dirty = true; + }; + + // access to our containers + RendererStorageRD *storage; + RendererSceneGIRD *gi; + + // 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; + + RS::EnvironmentSDFGICascades cascade_mode; + 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; + RID debug_probes_uniform_set; + RID cascades_ubo; + + bool uses_occlusion = false; + float bounce_feedback = 0.0; + bool reads_sky = false; + float energy = 1.0; + float normal_bias = 1.1; + float probe_bias = 1.1; + RS::EnvironmentSDFGIYScale y_scale_mode = RS::ENV_SDFGI_Y_SCALE_DISABLED; + + 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; + + void create(RendererSceneEnvironmentRD *p_env, const Vector3 &p_world_position, uint32_t p_requested_history_size, RendererSceneGIRD *p_gi); + void erase(); + void update(RendererSceneEnvironmentRD *p_env, const Vector3 &p_world_position); + void update_light(); + void update_probes(RendererSceneEnvironmentRD *p_env, RendererSceneSkyRD::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(const CameraMatrix &p_projection, const Transform &p_transform, int p_width, int p_height, RID p_render_target, RID p_texture); + void debug_probes(RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform); + + void pre_process_gi(const Transform &p_transform, RendererSceneRenderRD *p_scene_render); + void render_region(RID p_render_buffers, int p_region, const PagedArray<RendererSceneRender::GeometryInstance *> &p_instances, RendererSceneRenderRD *p_scene_render); + void render_static_lights(RID p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const PagedArray<RID> *p_positional_light_cull_result, RendererSceneRenderRD *p_scene_render); + }; + + RS::EnvironmentSDFGIRayCount sdfgi_ray_count = RS::ENV_SDFGI_RAY_COUNT_16; + RS::EnvironmentSDFGIFramesToConverge sdfgi_frames_to_converge = RS::ENV_SDFGI_CONVERGE_IN_10_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; } + + /* GI */ + enum { + MAX_GIPROBES = 8 + }; + + // Struct for use in render buffer + struct RenderBuffersGI { + RID giprobe_textures[MAX_GIPROBES]; + RID giprobe_buffer; + + RID full_buffer; + RID full_dispatch; + RID full_mask; + }; + + 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 + }; + + ProbeCascadeData cascades[SDFGI::MAX_CASCADES]; + }; + + struct GIProbeData { + float xform[16]; + float bounds[3]; + float dynamic_range; + + float bias; + float normal_bias; + uint32_t blend_ambient; + uint32_t texture_slot; + + float anisotropy_strength; + float ao; + float ao_size; + uint32_t mipmaps; + }; + + struct PushConstant { + int32_t screen_size[2]; + float z_near; + float z_far; + + float proj_info[4]; + float ao_color[3]; + uint32_t max_giprobes; + + uint32_t high_quality_vct; + uint32_t orthogonal; + uint32_t pad[2]; + + float cam_rotation[12]; + }; + + RID sdfgi_ubo; + enum Mode { + MODE_GIPROBE, + MODE_SDFGI, + MODE_COMBINED, + MODE_HALF_RES_GIPROBE, + MODE_HALF_RES_SDFGI, + MODE_HALF_RES_COMBINED, + MODE_MAX + }; + + RID default_giprobe_buffer; + + bool half_resolution = false; + GiShaderRD shader; + RID shader_version; + RID pipelines[MODE_MAX]; + + RendererSceneGIRD(); + ~RendererSceneGIRD(); + + void init(RendererStorageRD *p_storage, RendererSceneSkyRD *p_sky); + void free(); + + SDFGI *create_sdfgi(RendererSceneEnvironmentRD *p_env, const Vector3 &p_world_position, uint32_t p_requested_history_size); + + void setup_giprobes(RID p_render_buffers, const Transform &p_transform, const PagedArray<RID> &p_gi_probes, uint32_t &r_gi_probes_used, RendererSceneRenderRD *p_scene_render); + void process_gi(RID p_render_buffers, RID p_normal_roughness_buffer, RID p_gi_probe_buffer, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform, const PagedArray<RID> &p_gi_probes, RendererSceneRenderRD *p_scene_render); + + RID gi_probe_instance_create(RID p_base); + void gi_probe_instance_set_transform_to_data(RID p_probe, const Transform &p_xform); + bool gi_probe_needs_update(RID p_probe) const; + void gi_probe_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RendererSceneRender::GeometryInstance *> &p_dynamic_objects, RendererSceneRenderRD *p_scene_render); + void debug_giprobe(RID p_gi_probe, RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform, bool p_lighting, bool p_emission, float p_alpha); +}; + +#endif /* !RENDERING_SERVER_SCENE_GI_RD_H */ diff --git a/servers/rendering/renderer_rd/renderer_scene_render_forward.cpp b/servers/rendering/renderer_rd/renderer_scene_render_forward_clustered.cpp index c0939f23ef..7a19495f48 100644 --- a/servers/rendering/renderer_rd/renderer_scene_render_forward.cpp +++ b/servers/rendering/renderer_rd/renderer_scene_render_forward_clustered.cpp @@ -1,5 +1,5 @@ /*************************************************************************/ -/* renderer_scene_render_forward.cpp */ +/* renderer_scene_render_forward_clustered.cpp */ /*************************************************************************/ /* This file is part of: */ /* GODOT ENGINE */ @@ -28,13 +28,13 @@ /* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ /*************************************************************************/ -#include "renderer_scene_render_forward.h" +#include "renderer_scene_render_forward_clustered.h" #include "core/config/project_settings.h" #include "servers/rendering/rendering_device.h" #include "servers/rendering/rendering_server_default.h" /* SCENE SHADER */ -void RendererSceneRenderForward::ShaderData::set_code(const String &p_code) { +void RendererSceneRenderForwardClustered::ShaderData::set_code(const String &p_code) { //compile code = p_code; @@ -123,7 +123,7 @@ void RendererSceneRenderForward::ShaderData::set_code(const String &p_code) { actions.uniforms = &uniforms; - RendererSceneRenderForward *scene_singleton = (RendererSceneRenderForward *)RendererSceneRenderForward::singleton; + RendererSceneRenderForwardClustered *scene_singleton = (RendererSceneRenderForwardClustered *)RendererSceneRenderForwardClustered::singleton; Error err = scene_singleton->shader.compiler.compile(RS::SHADER_SPATIAL, code, &actions, path, gen_code); @@ -257,7 +257,7 @@ void RendererSceneRenderForward::ShaderData::set_code(const String &p_code) { 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<RendererSceneRenderForward *>(singleton)->shader.scene_shader.is_variant_enabled(k)) { + if (!static_cast<RendererSceneRenderForwardClustered *>(singleton)->shader.scene_shader.is_variant_enabled(k)) { continue; } RD::PipelineRasterizationState raster_state; @@ -324,7 +324,7 @@ void RendererSceneRenderForward::ShaderData::set_code(const String &p_code) { valid = true; } -void RendererSceneRenderForward::ShaderData::set_default_texture_param(const StringName &p_name, RID p_texture) { +void RendererSceneRenderForwardClustered::ShaderData::set_default_texture_param(const StringName &p_name, RID p_texture) { if (!p_texture.is_valid()) { default_texture_params.erase(p_name); } else { @@ -332,7 +332,7 @@ void RendererSceneRenderForward::ShaderData::set_default_texture_param(const Str } } -void RendererSceneRenderForward::ShaderData::get_param_list(List<PropertyInfo> *p_param_list) const { +void RendererSceneRenderForwardClustered::ShaderData::get_param_list(List<PropertyInfo> *p_param_list) const { Map<int, StringName> order; for (Map<StringName, ShaderLanguage::ShaderNode::Uniform>::Element *E = uniforms.front(); E; E = E->next()) { @@ -354,7 +354,7 @@ void RendererSceneRenderForward::ShaderData::get_param_list(List<PropertyInfo> * } } -void RendererSceneRenderForward::ShaderData::get_instance_param_list(List<RendererStorage::InstanceShaderParam> *p_param_list) const { +void RendererSceneRenderForwardClustered::ShaderData::get_instance_param_list(List<RendererStorage::InstanceShaderParam> *p_param_list) const { for (Map<StringName, ShaderLanguage::ShaderNode::Uniform>::Element *E = uniforms.front(); E; E = E->next()) { if (E->get().scope != ShaderLanguage::ShaderNode::Uniform::SCOPE_INSTANCE) { continue; @@ -369,7 +369,7 @@ void RendererSceneRenderForward::ShaderData::get_instance_param_list(List<Render } } -bool RendererSceneRenderForward::ShaderData::is_param_texture(const StringName &p_param) const { +bool RendererSceneRenderForwardClustered::ShaderData::is_param_texture(const StringName &p_param) const { if (!uniforms.has(p_param)) { return false; } @@ -377,15 +377,15 @@ bool RendererSceneRenderForward::ShaderData::is_param_texture(const StringName & return uniforms[p_param].texture_order >= 0; } -bool RendererSceneRenderForward::ShaderData::is_animated() const { +bool RendererSceneRenderForwardClustered::ShaderData::is_animated() const { return false; } -bool RendererSceneRenderForward::ShaderData::casts_shadows() const { +bool RendererSceneRenderForwardClustered::ShaderData::casts_shadows() const { return false; } -Variant RendererSceneRenderForward::ShaderData::get_default_parameter(const StringName &p_parameter) const { +Variant RendererSceneRenderForwardClustered::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; @@ -394,13 +394,19 @@ Variant RendererSceneRenderForward::ShaderData::get_default_parameter(const Stri return Variant(); } -RendererSceneRenderForward::ShaderData::ShaderData() { +RS::ShaderNativeSourceCode RendererSceneRenderForwardClustered::ShaderData::get_native_source_code() const { + RendererSceneRenderForwardClustered *scene_singleton = (RendererSceneRenderForwardClustered *)RendererSceneRenderForwardClustered::singleton; + + return scene_singleton->shader.scene_shader.version_get_native_source_code(version); +} + +RendererSceneRenderForwardClustered::ShaderData::ShaderData() { valid = false; uses_screen_texture = false; } -RendererSceneRenderForward::ShaderData::~ShaderData() { - RendererSceneRenderForward *scene_singleton = (RendererSceneRenderForward *)RendererSceneRenderForward::singleton; +RendererSceneRenderForwardClustered::ShaderData::~ShaderData() { + RendererSceneRenderForwardClustered *scene_singleton = (RendererSceneRenderForwardClustered *)RendererSceneRenderForwardClustered::singleton; ERR_FAIL_COND(!scene_singleton); //pipeline variants will clear themselves if shader is gone if (version.is_valid()) { @@ -408,21 +414,21 @@ RendererSceneRenderForward::ShaderData::~ShaderData() { } } -RendererStorageRD::ShaderData *RendererSceneRenderForward::_create_shader_func() { +RendererStorageRD::ShaderData *RendererSceneRenderForwardClustered::_create_shader_func() { ShaderData *shader_data = memnew(ShaderData); return shader_data; } -void RendererSceneRenderForward::MaterialData::set_render_priority(int p_priority) { +void RendererSceneRenderForwardClustered::MaterialData::set_render_priority(int p_priority) { priority = p_priority - RS::MATERIAL_RENDER_PRIORITY_MIN; //8 bits } -void RendererSceneRenderForward::MaterialData::set_next_pass(RID p_pass) { +void RendererSceneRenderForwardClustered::MaterialData::set_next_pass(RID p_pass) { next_pass = p_pass; } -void RendererSceneRenderForward::MaterialData::update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { - RendererSceneRenderForward *scene_singleton = (RendererSceneRenderForward *)RendererSceneRenderForward::singleton; +void RendererSceneRenderForwardClustered::MaterialData::update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { + RendererSceneRenderForwardClustered *scene_singleton = (RendererSceneRenderForwardClustered *)RendererSceneRenderForwardClustered::singleton; if ((uint32_t)ubo_data.size() != shader_data->ubo_size) { p_uniform_dirty = true; @@ -447,7 +453,7 @@ void RendererSceneRenderForward::MaterialData::update_parameters(const Map<Strin //check whether buffer changed if (p_uniform_dirty && ubo_data.size()) { update_uniform_buffer(shader_data->uniforms, shader_data->ubo_offsets.ptr(), p_parameters, ubo_data.ptrw(), ubo_data.size(), false); - RD::get_singleton()->buffer_update(uniform_buffer, 0, ubo_data.size(), ubo_data.ptrw()); + RD::get_singleton()->buffer_update(uniform_buffer, 0, ubo_data.size(), ubo_data.ptrw(), RD::BARRIER_MASK_RASTER); } uint32_t tex_uniform_count = shader_data->texture_uniforms.size(); @@ -501,7 +507,7 @@ void RendererSceneRenderForward::MaterialData::update_parameters(const Map<Strin uniform_set = RD::get_singleton()->uniform_set_create(uniforms, scene_singleton->shader.scene_shader.version_get_shader(shader_data->version, 0), MATERIAL_UNIFORM_SET); } -RendererSceneRenderForward::MaterialData::~MaterialData() { +RendererSceneRenderForwardClustered::MaterialData::~MaterialData() { if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { RD::get_singleton()->free(uniform_set); } @@ -511,7 +517,7 @@ RendererSceneRenderForward::MaterialData::~MaterialData() { } } -RendererStorageRD::MaterialData *RendererSceneRenderForward::_create_material_func(ShaderData *p_shader) { +RendererStorageRD::MaterialData *RendererSceneRenderForwardClustered::_create_material_func(ShaderData *p_shader) { MaterialData *material_data = memnew(MaterialData); material_data->shader_data = p_shader; material_data->last_frame = false; @@ -519,11 +525,11 @@ RendererStorageRD::MaterialData *RendererSceneRenderForward::_create_material_fu return material_data; } -RendererSceneRenderForward::RenderBufferDataForward::~RenderBufferDataForward() { +RendererSceneRenderForwardClustered::RenderBufferDataForward::~RenderBufferDataForward() { clear(); } -void RendererSceneRenderForward::RenderBufferDataForward::ensure_specular() { +void RendererSceneRenderForwardClustered::RenderBufferDataForward::ensure_specular() { if (!specular.is_valid()) { RD::TextureFormat tf; tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; @@ -577,20 +583,7 @@ void RendererSceneRenderForward::RenderBufferDataForward::ensure_specular() { } } -void RendererSceneRenderForward::RenderBufferDataForward::ensure_gi() { - if (!reflection_buffer.is_valid()) { - RD::TextureFormat tf; - tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; - tf.width = width; - tf.height = height; - tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; - - reflection_buffer = RD::get_singleton()->texture_create(tf, RD::TextureView()); - ambient_buffer = RD::get_singleton()->texture_create(tf, RD::TextureView()); - } -} - -void RendererSceneRenderForward::RenderBufferDataForward::ensure_giprobe() { +void RendererSceneRenderForwardClustered::RenderBufferDataForward::ensure_giprobe() { if (!giprobe_buffer.is_valid()) { RD::TextureFormat tf; tf.format = RD::DATA_FORMAT_R8G8_UINT; @@ -626,17 +619,7 @@ void RendererSceneRenderForward::RenderBufferDataForward::ensure_giprobe() { } } -void RendererSceneRenderForward::RenderBufferDataForward::clear() { - if (ambient_buffer != RID() && ambient_buffer != color) { - RD::get_singleton()->free(ambient_buffer); - ambient_buffer = RID(); - } - - if (reflection_buffer != RID() && reflection_buffer != specular) { - RD::get_singleton()->free(reflection_buffer); - reflection_buffer = RID(); - } - +void RendererSceneRenderForwardClustered::RenderBufferDataForward::clear() { if (giprobe_buffer != RID()) { RD::get_singleton()->free(giprobe_buffer); giprobe_buffer = RID(); @@ -690,7 +673,7 @@ void RendererSceneRenderForward::RenderBufferDataForward::clear() { } } -void RendererSceneRenderForward::RenderBufferDataForward::configure(RID p_color_buffer, RID p_depth_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa) { +void RendererSceneRenderForwardClustered::RenderBufferDataForward::configure(RID p_color_buffer, RID p_depth_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa) { clear(); msaa = p_msaa; @@ -757,7 +740,7 @@ void RendererSceneRenderForward::RenderBufferDataForward::configure(RID p_color_ } } -void RendererSceneRenderForward::_allocate_normal_roughness_texture(RenderBufferDataForward *rb) { +void RendererSceneRenderForwardClustered::_allocate_normal_roughness_texture(RenderBufferDataForward *rb) { if (rb->normal_roughness_buffer.is_valid()) { return; } @@ -795,11 +778,11 @@ void RendererSceneRenderForward::_allocate_normal_roughness_texture(RenderBuffer _render_buffers_clear_uniform_set(rb); } -RendererSceneRenderRD::RenderBufferData *RendererSceneRenderForward::_create_render_buffer_data() { +RendererSceneRenderRD::RenderBufferData *RendererSceneRenderForwardClustered::_create_render_buffer_data() { return memnew(RenderBufferDataForward); } -bool RendererSceneRenderForward::free(RID p_rid) { +bool RendererSceneRenderForwardClustered::free(RID p_rid) { if (RendererSceneRenderRD::free(p_rid)) { return true; } @@ -808,8 +791,8 @@ bool RendererSceneRenderForward::free(RID p_rid) { /// RENDERING /// -template <RendererSceneRenderForward::PassMode p_pass_mode> -void RendererSceneRenderForward::_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) { +template <RendererSceneRenderForwardClustered::PassMode p_pass_mode> +void RendererSceneRenderForwardClustered::_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) { RD::DrawListID draw_list = p_draw_list; RD::FramebufferFormatID framebuffer_format = p_framebuffer_Format; @@ -827,16 +810,26 @@ void RendererSceneRenderForward::_render_list_template(RenderingDevice::DrawList bool shadow_pass = (p_params->pass_mode == PASS_MODE_SHADOW) || (p_params->pass_mode == PASS_MODE_SHADOW_DP); - float old_offset[2] = { 0, 0 }; + SceneState::PushConstant push_constant; + + if (p_params->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; + } 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]; + + push_constant.base_index = i + p_params->element_offset; RID material_uniform_set; ShaderData *shader; void *mesh_surface; - if (shadow_pass) { + if (shadow_pass || p_params->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; @@ -851,13 +844,6 @@ void RendererSceneRenderForward::_render_list_template(RenderingDevice::DrawList continue; } - if (p_params->pass_mode == PASS_MODE_DEPTH_MATERIAL) { - old_offset[0] = surf->owner->push_constant.lightmap_uv_scale[0]; - old_offset[1] = surf->owner->push_constant.lightmap_uv_scale[1]; - surf->owner->push_constant.lightmap_uv_scale[0] = p_params->uv_offset.x; - surf->owner->push_constant.lightmap_uv_scale[1] = p_params->uv_offset.y; - } - //find cull variant ShaderData::CullVariant cull_variant; @@ -879,16 +865,16 @@ void RendererSceneRenderForward::_render_list_template(RenderingDevice::DrawList switch (p_params->pass_mode) { case PASS_MODE_COLOR: case PASS_MODE_COLOR_TRANSPARENT: { - if (surf->sort.uses_lightmap) { + if (element_info.uses_lightmap) { shader_version = SHADER_VERSION_LIGHTMAP_COLOR_PASS; - } else if (surf->sort.uses_forward_gi) { + } else if (element_info.uses_forward_gi) { shader_version = SHADER_VERSION_COLOR_PASS_WITH_FORWARD_GI; } else { shader_version = SHADER_VERSION_COLOR_PASS; } } break; case PASS_MODE_COLOR_SPECULAR: { - if (surf->sort.uses_lightmap) { + if (element_info.uses_lightmap) { shader_version = SHADER_VERSION_LIGHTMAP_COLOR_PASS_WITH_SEPARATE_SPECULAR; } else { shader_version = SHADER_VERSION_COLOR_PASS_WITH_SEPARATE_SPECULAR; @@ -930,31 +916,7 @@ void RendererSceneRenderForward::_render_list_template(RenderingDevice::DrawList storage->mesh_surface_get_vertex_arrays_and_format(mesh_surface, pipeline->get_vertex_input_mask(), vertex_array_rd, vertex_format); } - if (p_params->screen_lod_threshold > 0.0 && storage->mesh_surface_has_lod(mesh_surface)) { - //lod - Vector3 support_min = surf->owner->transformed_aabb.get_support(-p_params->lod_plane.normal); - Vector3 support_max = surf->owner->transformed_aabb.get_support(p_params->lod_plane.normal); - - float distance_min = p_params->lod_plane.distance_to(support_min); - float distance_max = p_params->lod_plane.distance_to(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; - } - - index_array_rd = storage->mesh_surface_get_index_array_with_lod(mesh_surface, surf->owner->lod_model_scale * surf->owner->lod_bias, distance * p_params->lod_distance_multiplier, p_params->screen_lod_threshold); - - } else { - //no lod - index_array_rd = storage->mesh_surface_get_index_array(mesh_surface); - } + index_array_rd = 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); @@ -991,18 +953,15 @@ void RendererSceneRenderForward::_render_list_template(RenderingDevice::DrawList prev_material_uniform_set = material_uniform_set; } - RD::get_singleton()->draw_list_set_push_constant(draw_list, &surf->owner->push_constant, sizeof(GeometryInstanceForward::PushConstant)); - - RD::get_singleton()->draw_list_draw(draw_list, index_array_rd.is_valid(), surf->owner->instance_count); + RD::get_singleton()->draw_list_set_push_constant(draw_list, &push_constant, sizeof(SceneState::PushConstant)); - if (p_params->pass_mode == PASS_MODE_DEPTH_MATERIAL) { - surf->owner->push_constant.lightmap_uv_scale[0] = old_offset[0]; - surf->owner->push_constant.lightmap_uv_scale[1] = old_offset[1]; - } + uint32_t instance_count = surf->owner->instance_count > 1 ? surf->owner->instance_count : element_info.repeat; + RD::get_singleton()->draw_list_draw(draw_list, index_array_rd.is_valid(), instance_count); + i += element_info.repeat - 1; //skip equal elements } } -void RendererSceneRenderForward::_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) { +void RendererSceneRenderForwardClustered::_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) { @@ -1039,7 +998,7 @@ void RendererSceneRenderForward::_render_list(RenderingDevice::DrawListID p_draw } } -void RendererSceneRenderForward::_render_list_thread_function(uint32_t p_thread, RenderListParameters *p_params) { +void RendererSceneRenderForwardClustered::_render_list_thread_function(uint32_t p_thread, RenderListParameters *p_params) { uint32_t render_total = p_params->element_count; uint32_t total_threads = RendererThreadPool::singleton->thread_work_pool.get_thread_count(); uint32_t render_from = p_thread * render_total / total_threads; @@ -1047,7 +1006,7 @@ void RendererSceneRenderForward::_render_list_thread_function(uint32_t p_thread, _render_list(thread_draw_lists[p_thread], p_params->framebuffer_format, p_params, render_from, render_to); } -void RendererSceneRenderForward::_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) { +void RendererSceneRenderForwardClustered::_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; @@ -1055,17 +1014,17 @@ void RendererSceneRenderForward::_render_list_with_threads(RenderListParameters //multi threaded thread_draw_lists.resize(RendererThreadPool::singleton->thread_work_pool.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); - RendererThreadPool::singleton->thread_work_pool.do_work(thread_draw_lists.size(), this, &RendererSceneRenderForward::_render_list_thread_function, p_params); - RD::get_singleton()->draw_list_end(); + RendererThreadPool::singleton->thread_work_pool.do_work(thread_draw_lists.size(), this, &RendererSceneRenderForwardClustered::_render_list_thread_function, p_params); + 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(); + RD::get_singleton()->draw_list_end(p_params->barrier); } } -void RendererSceneRenderForward::_setup_environment(RID p_environment, RID p_render_buffers, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_reflection_probe, bool p_no_fog, const Size2 &p_screen_pixel_size, RID p_shadow_atlas, bool p_flip_y, const Color &p_default_bg_color, float p_znear, float p_zfar, bool p_opaque_render_buffers, bool p_pancake_shadows) { +void RendererSceneRenderForwardClustered::_setup_environment(RID p_environment, RID p_render_buffers, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_reflection_probe, bool p_no_fog, const Size2i &p_screen_size, uint32_t p_cluster_size, uint32_t p_max_cluster_elements, RID p_shadow_atlas, bool p_flip_y, const Color &p_default_bg_color, float p_znear, float p_zfar, bool p_opaque_render_buffers, bool p_pancake_shadows, int p_index) { //CameraMatrix projection = p_cam_projection; //projection.flip_y(); // Vulkan and modern APIs use Y-Down CameraMatrix correction; @@ -1093,8 +1052,18 @@ void RendererSceneRenderForward::_setup_environment(RID p_environment, RID p_ren scene_state.ubo.penumbra_shadow_samples = penumbra_shadow_samples_get(); scene_state.ubo.soft_shadow_samples = soft_shadow_samples_get(); - scene_state.ubo.screen_pixel_size[0] = p_screen_pixel_size.x; - scene_state.ubo.screen_pixel_size[1] = p_screen_pixel_size.y; + Size2 screen_pixel_size = Vector2(1.0, 1.0) / Size2(p_screen_size); + scene_state.ubo.screen_pixel_size[0] = screen_pixel_size.x; + scene_state.ubo.screen_pixel_size[1] = screen_pixel_size.y; + + scene_state.ubo.cluster_shift = get_shift_from_power_of_2(p_cluster_size); + scene_state.ubo.max_cluster_element_count_div_32 = p_max_cluster_elements / 32; + { + uint32_t cluster_screen_width = (p_screen_size.width - 1) / p_cluster_size + 1; + uint32_t cluster_screen_height = (p_screen_size.height - 1) / p_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; + } if (p_shadow_atlas.is_valid()) { Vector2 sas = shadow_atlas_get_size(p_shadow_atlas); @@ -1161,7 +1130,7 @@ void RendererSceneRenderForward::_setup_environment(RID p_environment, RID p_ren //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_get_lightprobe_octahedron_size(); + uint32_t oct_size = gi.sdfgi_get_lightprobe_octahedron_size(); scene_state.ubo.sdfgi_lightprobe_tex_pixel_size[0] = 1.0 / ((oct_size + 2) * scene_state.ubo.sdfgi_probe_axis_size * scene_state.ubo.sdfgi_probe_axis_size); scene_state.ubo.sdfgi_lightprobe_tex_pixel_size[1] = 1.0 / ((oct_size + 2) * scene_state.ubo.sdfgi_probe_axis_size); @@ -1294,27 +1263,125 @@ void RendererSceneRenderForward::_setup_environment(RID p_environment, RID p_ren scene_state.ubo.roughness_limiter_amount = screen_space_roughness_limiter_get_amount(); scene_state.ubo.roughness_limiter_limit = screen_space_roughness_limiter_get_limit(); - RD::get_singleton()->buffer_update(scene_state.uniform_buffer, 0, sizeof(SceneState::UBO), &scene_state.ubo, true); + 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); + render_pass_uniform_sets.resize(p_index + 1); + for (uint32_t i = from; i < scene_state.uniform_buffers.size(); i++) { + scene_state.uniform_buffers[i] = RD::get_singleton()->uniform_buffer_create(sizeof(SceneState::UBO)); + } + } + RD::get_singleton()->buffer_update(scene_state.uniform_buffers[p_index], 0, sizeof(SceneState::UBO), &scene_state.ubo, RD::BARRIER_MASK_RASTER); } -void RendererSceneRenderForward::_fill_render_list(const PagedArray<GeometryInstance *> &p_instances, PassMode p_pass_mode, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, bool p_using_sdfgi, bool p_using_opaque_gi) { - scene_state.used_sss = false; - scene_state.used_screen_texture = false; - scene_state.used_normal_texture = false; - scene_state.used_depth_texture = false; +void RendererSceneRenderForwardClustered::_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 RendererSceneRenderForwardClustered::_fill_instance_data(RenderListType p_render_list, 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); + + 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->store_transform_cache) { + RendererStorageRD::store_transform(inst->transform, instance_data.transform); + } else { + RendererStorageRD::store_transform(Transform(), instance_data.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_parameters_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; + + 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) { + //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; + } + + 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; + + 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); + } +} + +void RendererSceneRenderForwardClustered::_fill_render_list(RenderListType p_render_list, const PagedArray<GeometryInstance *> &p_instances, PassMode p_pass_mode, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, bool p_using_sdfgi, bool p_using_opaque_gi, const Plane &p_lod_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold, bool p_append) { + 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_cam_transform.origin, -p_cam_transform.basis.get_axis(Vector3::AXIS_Z)); near_plane.d += p_cam_projection.get_z_near(); float z_max = p_cam_projection.get_z_far() - p_cam_projection.get_z_near(); - uint32_t lightmap_captures_used = 0; + RenderList *rl = &render_list[p_render_list]; _update_dirty_geometry_instances(); - render_list.clear(); + + 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_instances.size(); i++) { - GeometryInstanceForward *inst = static_cast<GeometryInstanceForward *>(p_instances[i]); + GeometryInstanceForwardClustered *inst = static_cast<GeometryInstanceForwardClustered *>(p_instances[i]); Vector3 support_min = inst->transformed_aabb.get_support(-near_plane.normal); inst->depth = near_plane.distance_to(support_min); @@ -1325,7 +1392,7 @@ void RendererSceneRenderForward::_fill_render_list(const PagedArray<GeometryInst bool uses_lightmap = false; bool uses_gi = false; - if (p_pass_mode == PASS_MODE_COLOR) { + if (p_render_list == RENDER_LIST_OPAQUE) { //setup GI if (inst->lightmap_instance.is_valid()) { @@ -1337,15 +1404,15 @@ void RendererSceneRenderForward::_fill_render_list(const PagedArray<GeometryInst } } if (lightmap_cull_index >= 0) { - inst->push_constant.gi_offset &= 0xFFFF; - inst->push_constant.gi_offset |= lightmap_cull_index; + 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->push_constant.gi_offset = 0xFFFFFFFF; + inst->gi_offset_cache = 0xFFFFFFFF; } } else if (inst->lightmap_sh) { @@ -1359,7 +1426,7 @@ void RendererSceneRenderForward::_fill_render_list(const PagedArray<GeometryInst lcd.sh[j * 4 + 3] = src_capture[j].a; } flags |= INSTANCE_DATA_FLAG_USE_LIGHTMAP_CAPTURE; - inst->push_constant.gi_offset = lightmap_captures_used; + inst->gi_offset_cache = lightmap_captures_used; lightmap_captures_used++; uses_lightmap = true; } @@ -1386,18 +1453,19 @@ void RendererSceneRenderForward::_fill_render_list(const PagedArray<GeometryInst SWAP(probe0_index, probe1_index); } - inst->push_constant.gi_offset = probe0_index | (probe1_index << 16); + inst->gi_offset_cache = probe0_index | (probe1_index << 16); + flags |= INSTANCE_DATA_FLAG_USE_GIPROBE; uses_gi = true; } else { if (p_using_sdfgi && inst->can_sdfgi) { flags |= INSTANCE_DATA_FLAG_USE_SDFGI; uses_gi = true; } - inst->push_constant.gi_offset = 0xFFFFFFFF; + inst->gi_offset_cache = 0xFFFFFFFF; } } } - inst->push_constant.flags = flags; + inst->flags_cache = flags; GeometryInstanceSurfaceDataCache *surf = inst->surface_caches; @@ -1405,12 +1473,39 @@ void RendererSceneRenderForward::_fill_render_list(const PagedArray<GeometryInst surf->sort.uses_forward_gi = 0; surf->sort.uses_lightmap = 0; + // LOD + + if (p_screen_lod_threshold > 0.0 && storage->mesh_surface_has_lod(surf->surface)) { + //lod + Vector3 lod_support_min = inst->transformed_aabb.get_support(-p_lod_plane.normal); + Vector3 lod_support_max = inst->transformed_aabb.get_support(p_lod_plane.normal); + + float distance_min = p_lod_plane.distance_to(lod_support_min); + float distance_max = p_lod_plane.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; + } + + surf->sort.lod_index = storage->mesh_surface_get_lod(surf->surface, inst->lod_model_scale * inst->lod_bias, distance * p_lod_distance_multiplier, p_screen_lod_threshold); + } else { + surf->sort.lod_index = 0; + } + + // ADD Element if (p_pass_mode == PASS_MODE_COLOR) { if (surf->flags & (GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH | GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE)) { - render_list.add_element(surf); + rl->add_element(surf); } if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_PASS_ALPHA) { - render_list.add_alpha_element(surf); + render_list[RENDER_LIST_ALPHA].add_element(surf); if (uses_gi) { surf->sort.uses_forward_gi = 1; } @@ -1435,11 +1530,11 @@ void RendererSceneRenderForward::_fill_render_list(const PagedArray<GeometryInst } else if (p_pass_mode == PASS_MODE_SHADOW || p_pass_mode == PASS_MODE_SHADOW_DP) { if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_PASS_SHADOW) { - render_list.add_element(surf); + rl->add_element(surf); } } else { if (surf->flags & (GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH | GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE)) { - render_list.add_element(surf); + rl->add_element(surf); } } @@ -1449,19 +1544,19 @@ void RendererSceneRenderForward::_fill_render_list(const PagedArray<GeometryInst } } - if (lightmap_captures_used) { - RD::get_singleton()->buffer_update(scene_state.lightmap_capture_buffer, 0, sizeof(LightmapCaptureData) * lightmap_captures_used, scene_state.lightmap_captures, true); + 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 RendererSceneRenderForward::_setup_giprobes(const PagedArray<RID> &p_giprobes) { +void RendererSceneRenderForwardClustered::_setup_giprobes(const PagedArray<RID> &p_giprobes) { scene_state.giprobes_used = MIN(p_giprobes.size(), uint32_t(MAX_GI_PROBES)); for (uint32_t i = 0; i < scene_state.giprobes_used; i++) { scene_state.giprobe_ids[i] = p_giprobes[i]; } } -void RendererSceneRenderForward::_setup_lightmaps(const PagedArray<RID> &p_lightmaps, const Transform &p_cam_transform) { +void RendererSceneRenderForwardClustered::_setup_lightmaps(const PagedArray<RID> &p_lightmaps, const Transform &p_cam_transform) { scene_state.lightmaps_used = 0; for (int i = 0; i < (int)p_lightmaps.size(); i++) { if (i >= (int)scene_state.max_lightmaps) { @@ -1479,29 +1574,22 @@ void RendererSceneRenderForward::_setup_lightmaps(const PagedArray<RID> &p_light 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, true); + RD::get_singleton()->buffer_update(scene_state.lightmap_buffer, 0, sizeof(LightmapData) * scene_state.lightmaps_used, scene_state.lightmaps, RD::BARRIER_MASK_RASTER); } } -void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, int p_directional_light_count, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_bg_color, float p_screen_lod_threshold) { +void RendererSceneRenderForwardClustered::_render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_cluster_buffer, uint32_t p_cluster_size, uint32_t p_max_cluster_elements, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_bg_color, float p_screen_lod_threshold) { RenderBufferDataForward *render_buffer = nullptr; if (p_render_buffer.is_valid()) { render_buffer = (RenderBufferDataForward *)render_buffers_get_data(p_render_buffer); } + RendererSceneEnvironmentRD *env = get_environment(p_environment); //first of all, make a new render pass - render_pass++; - //fill up ubo RENDER_TIMESTAMP("Setup 3D Scene"); - if (p_reflection_probe.is_valid()) { - scene_state.ubo.reflection_multiplier = 0.0; - } else { - scene_state.ubo.reflection_multiplier = 1.0; - } - float lod_distance_multiplier = p_cam_projection.get_lod_multiplier(); Plane lod_camera_plane(p_cam_transform.get_origin(), -p_cam_transform.basis.get_axis(Vector3::AXIS_Z)); @@ -1514,9 +1602,8 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf Vector2 vp_he = p_cam_projection.get_viewport_half_extents(); scene_state.ubo.viewport_size[0] = vp_he.x; scene_state.ubo.viewport_size[1] = vp_he.y; - scene_state.ubo.directional_light_count = p_directional_light_count; + scene_state.ubo.directional_light_count = 0; - Size2 screen_pixel_size; Size2i screen_size; RID opaque_framebuffer; RID opaque_specular_framebuffer; @@ -1531,8 +1618,6 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf bool using_giprobe = false; if (render_buffer) { - screen_pixel_size.width = 1.0 / render_buffer->width; - screen_pixel_size.height = 1.0 / render_buffer->height; screen_size.x = render_buffer->width; screen_size.y = render_buffer->height; @@ -1540,7 +1625,6 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf if (!low_end && p_gi_probes.size() > 0) { using_giprobe = true; - render_buffer->ensure_gi(); } if (!p_environment.is_valid() && using_giprobe) { @@ -1550,7 +1634,6 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf if (environment_is_sdfgi_enabled(p_environment)) { depth_pass_mode = using_giprobe ? PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE : PASS_MODE_DEPTH_NORMAL_ROUGHNESS; // also giprobe using_sdfgi = true; - render_buffer->ensure_gi(); } else { depth_pass_mode = using_giprobe ? PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE : PASS_MODE_DEPTH_NORMAL_ROUGHNESS; } @@ -1589,8 +1672,6 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf alpha_framebuffer = opaque_framebuffer; } else if (p_reflection_probe.is_valid()) { uint32_t resolution = reflection_probe_instance_get_resolution(p_reflection_probe); - screen_pixel_size.width = 1.0 / resolution; - screen_pixel_size.height = 1.0 / resolution; screen_size.x = resolution; screen_size.y = resolution; @@ -1605,13 +1686,21 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf ERR_FAIL(); //bug? } + RD::get_singleton()->draw_command_begin_label("Render Setup"); + _setup_lightmaps(p_lightmaps, p_cam_transform); _setup_giprobes(p_gi_probes); - _setup_environment(p_environment, p_render_buffer, p_cam_projection, p_cam_transform, p_reflection_probe, p_reflection_probe.is_valid(), screen_pixel_size, p_shadow_atlas, !p_reflection_probe.is_valid(), p_default_bg_color, p_cam_projection.get_z_near(), p_cam_projection.get_z_far(), false); + _setup_environment(p_environment, p_render_buffer, p_cam_projection, p_cam_transform, p_reflection_probe, p_reflection_probe.is_valid(), screen_size, p_cluster_size, p_max_cluster_elements, p_shadow_atlas, !p_reflection_probe.is_valid(), p_default_bg_color, p_cam_projection.get_z_near(), p_cam_projection.get_z_far(), false); _update_render_base_uniform_set(); //may have changed due to the above (light buffer enlarged, as an example) - _fill_render_list(p_instances, PASS_MODE_COLOR, p_cam_projection, p_cam_transform, using_sdfgi, using_sdfgi || using_giprobe); + _fill_render_list(RENDER_LIST_OPAQUE, p_instances, PASS_MODE_COLOR, p_cam_projection, p_cam_transform, using_sdfgi, using_sdfgi || using_giprobe, lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); + render_list[RENDER_LIST_OPAQUE].sort_by_key(); + render_list[RENDER_LIST_ALPHA].sort_by_depth(); + _fill_instance_data(RENDER_LIST_OPAQUE); + _fill_instance_data(RENDER_LIST_ALPHA); + + RD::get_singleton()->draw_command_end_label(); bool using_sss = !low_end && render_buffer && scene_state.used_sss && sub_surface_scattering_get_quality() != RS::SUB_SURFACE_SCATTERING_QUALITY_DISABLED; @@ -1641,7 +1730,7 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf clear_color.b *= bg_energy; if (render_buffers_has_volumetric_fog(p_render_buffer) || environment_is_fog_enabled(p_environment)) { draw_sky_fog_only = true; - storage->material_set_param(sky_scene_state.fog_material, "clear_color", Variant(clear_color.to_linear())); + storage->material_set_param(sky.sky_scene_state.fog_material, "clear_color", Variant(clear_color.to_linear())); } } break; case RS::ENV_BG_COLOR: { @@ -1651,7 +1740,7 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf clear_color.b *= bg_energy; if (render_buffers_has_volumetric_fog(p_render_buffer) || environment_is_fog_enabled(p_environment)) { draw_sky_fog_only = true; - storage->material_set_param(sky_scene_state.fog_material, "clear_color", Variant(clear_color.to_linear())); + storage->material_set_param(sky.sky_scene_state.fog_material, "clear_color", Variant(clear_color.to_linear())); } } break; case RS::ENV_BG_SKY: { @@ -1671,6 +1760,7 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf // setup sky if used for ambient, reflections, or background if (draw_sky || draw_sky_fog_only || environment_get_reflection_source(p_environment) == RS::ENV_REFLECTION_SOURCE_SKY || environment_get_ambient_source(p_environment) == RS::ENV_AMBIENT_SOURCE_SKY) { RENDER_TIMESTAMP("Setup Sky"); + RD::get_singleton()->draw_command_begin_label("Setup Sky"); CameraMatrix projection = p_cam_projection; if (p_reflection_probe.is_valid()) { CameraMatrix correction; @@ -1678,65 +1768,87 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf projection = correction * p_cam_projection; } - _setup_sky(p_environment, p_render_buffer, projection, p_cam_transform, screen_size); + sky.setup(env, p_render_buffer, projection, p_cam_transform, screen_size, this); - RID sky = environment_get_sky(p_environment); - if (sky.is_valid()) { - _update_sky(p_environment, projection, p_cam_transform); - radiance_texture = sky_get_radiance_texture_rd(sky); + RID sky_rid = env->sky; + if (sky_rid.is_valid()) { + sky.update(env, projection, p_cam_transform, time); + 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; } - render_list.sort_by_key(false); - bool debug_giprobes = get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_ALBEDO || get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_LIGHTING || get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_EMISSION; bool debug_sdfgi_probes = get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_SDFGI_PROBES; - bool depth_pre_pass = !low_end && depth_framebuffer.is_valid(); bool using_ssao = depth_pre_pass && p_render_buffer.is_valid() && p_environment.is_valid() && environment_is_ssao_enabled(p_environment); bool continue_depth = false; if (depth_pre_pass) { //depth pre pass - RENDER_TIMESTAMP("Render Depth Pre-Pass"); - RID rp_uniform_set = _setup_render_pass_uniform_set(RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); + bool needs_pre_resolve = _needs_post_prepass_render(using_sdfgi || using_giprobe); + 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(using_sdfgi || using_giprobe); + } + + RD::get_singleton()->draw_command_begin_label("Render Depth Pre-Pass"); + + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_OPAQUE, RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); bool finish_depth = using_ssao || using_sdfgi || using_giprobe; - RenderListParameters render_list_params(render_list.elements, render_list.element_count, false, depth_pass_mode, render_buffer == nullptr, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); - _render_list_with_threads(&render_list_params, depth_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, finish_depth ? RD::FINAL_ACTION_READ : RD::FINAL_ACTION_CONTINUE, depth_pass_clear); + 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(), false, depth_pass_mode, render_buffer == nullptr, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); + _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(using_sdfgi || using_giprobe); + } if (render_buffer && render_buffer->msaa != RS::VIEWPORT_MSAA_DISABLED) { RENDER_TIMESTAMP("Resolve Depth Pre-Pass"); + RD::get_singleton()->draw_command_begin_label("Resolve Depth Pre-Pass"); if (depth_pass_mode == PASS_MODE_DEPTH_NORMAL_ROUGHNESS || depth_pass_mode == PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE) { + if (needs_pre_resolve) { + RD::get_singleton()->barrier(RD::BARRIER_MASK_RASTER, RD::BARRIER_MASK_COMPUTE); + } static int texture_samples[RS::VIEWPORT_MSAA_MAX] = { 1, 2, 4, 8, 16 }; storage->get_effects()->resolve_gi(render_buffer->depth_msaa, render_buffer->normal_roughness_buffer_msaa, using_giprobe ? render_buffer->giprobe_buffer_msaa : RID(), render_buffer->depth, render_buffer->normal_roughness_buffer, using_giprobe ? render_buffer->giprobe_buffer : RID(), Vector2i(render_buffer->width, render_buffer->height), texture_samples[render_buffer->msaa]); } else if (finish_depth) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->depth_msaa, render_buffer->depth, true); + RD::get_singleton()->texture_resolve_multisample(render_buffer->depth_msaa, render_buffer->depth); } + RD::get_singleton()->draw_command_end_label(); } continue_depth = !finish_depth; } - if (using_ssao) { - _process_ssao(p_render_buffer, p_environment, render_buffer->normal_roughness_buffer, p_cam_projection); - } + _pre_opaque_render(using_ssao, using_sdfgi || using_giprobe, render_buffer ? render_buffer->normal_roughness_buffer : RID(), render_buffer ? render_buffer->giprobe_buffer : RID()); - if (using_sdfgi || using_giprobe) { - _process_gi(p_render_buffer, render_buffer->normal_roughness_buffer, render_buffer->ambient_buffer, render_buffer->reflection_buffer, render_buffer->giprobe_buffer, p_environment, p_cam_projection, p_cam_transform, p_gi_probes); - } + RD::get_singleton()->draw_command_begin_label("Render Opaque Pass"); - _setup_environment(p_environment, p_render_buffer, p_cam_projection, p_cam_transform, p_reflection_probe, p_reflection_probe.is_valid(), screen_pixel_size, p_shadow_atlas, !p_reflection_probe.is_valid(), p_default_bg_color, p_cam_projection.get_z_near(), p_cam_projection.get_z_far(), p_render_buffer.is_valid()); + scene_state.ubo.directional_light_count = _get_render_state_directional_light_count(); + + _setup_environment(p_environment, p_render_buffer, p_cam_projection, p_cam_transform, p_reflection_probe, p_reflection_probe.is_valid(), screen_size, p_cluster_size, p_max_cluster_elements, p_shadow_atlas, !p_reflection_probe.is_valid(), p_default_bg_color, p_cam_projection.get_z_near(), p_cam_projection.get_z_far(), p_render_buffer.is_valid()); RENDER_TIMESTAMP("Render Opaque Pass"); - RID rp_uniform_set = _setup_render_pass_uniform_set(p_render_buffer, radiance_texture, p_shadow_atlas, p_reflection_atlas, p_gi_probes, p_lightmaps); + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_OPAQUE, p_render_buffer, radiance_texture, p_shadow_atlas, p_reflection_atlas, p_cluster_buffer, p_gi_probes, p_lightmaps, 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; @@ -1757,10 +1869,8 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf } RID framebuffer = using_separate_specular ? opaque_specular_framebuffer : opaque_framebuffer; - RenderListParameters render_list_params(render_list.elements, render_list.element_count, false, using_separate_specular ? PASS_MODE_COLOR_SPECULAR : PASS_MODE_COLOR, render_buffer == nullptr, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); - - _render_list_with_threads(&render_list_params, 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_KEEP : RD::INITIAL_ACTION_CONTINUE) : RD::INITIAL_ACTION_CLEAR, will_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, c, 1.0, 0); - + 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(), false, using_separate_specular ? PASS_MODE_COLOR_SPECULAR : PASS_MODE_COLOR, render_buffer == nullptr, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); + _render_list_with_threads(&render_list_params, 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(render_buffer->specular_only_fb, RD::INITIAL_ACTION_CONTINUE, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CONTINUE, RD::FINAL_ACTION_CONTINUE); @@ -1768,6 +1878,8 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf } } + RD::get_singleton()->draw_command_end_label(); + if (debug_giprobes) { //debug giprobes bool will_continue_color = (can_continue_color || draw_sky || draw_sky_fog_only); @@ -1777,9 +1889,11 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf dc.set_depth_correction(true); CameraMatrix cm = (dc * p_cam_projection) * CameraMatrix(p_cam_transform.affine_inverse()); RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(opaque_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 GIProbes"); for (int i = 0; i < (int)p_gi_probes.size(); i++) { - _debug_giprobe(p_gi_probes[i], draw_list, opaque_framebuffer, cm, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_LIGHTING, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_EMISSION, 1.0); + gi.debug_giprobe(p_gi_probes[i], draw_list, opaque_framebuffer, cm, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_LIGHTING, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_EMISSION, 1.0); } + RD::get_singleton()->draw_command_end_label(); RD::get_singleton()->draw_list_end(); } @@ -1792,7 +1906,9 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf dc.set_depth_correction(true); CameraMatrix cm = (dc * p_cam_projection) * CameraMatrix(p_cam_transform.affine_inverse()); RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(opaque_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 SDFGI"); _debug_sdfgi_probes(p_render_buffer, draw_list, opaque_framebuffer, cm); + RD::get_singleton()->draw_command_end_label(); RD::get_singleton()->draw_list_end(); } @@ -1805,30 +1921,35 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf correction.set_depth_correction(true); projection = correction * p_cam_projection; } - - _draw_sky(can_continue_color, can_continue_depth, opaque_framebuffer, p_environment, projection, p_cam_transform); + RD::get_singleton()->draw_command_begin_label("Draw Sky"); + sky.draw(env, can_continue_color, can_continue_depth, opaque_framebuffer, projection, p_cam_transform, time); + RD::get_singleton()->draw_command_end_label(); } if (render_buffer && !can_continue_color && render_buffer->msaa != RS::VIEWPORT_MSAA_DISABLED) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->color_msaa, render_buffer->color, true); + RD::get_singleton()->texture_resolve_multisample(render_buffer->color_msaa, render_buffer->color); if (using_separate_specular) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->specular_msaa, render_buffer->specular, true); + RD::get_singleton()->texture_resolve_multisample(render_buffer->specular_msaa, render_buffer->specular); } } if (render_buffer && !can_continue_depth && render_buffer->msaa != RS::VIEWPORT_MSAA_DISABLED) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->depth_msaa, render_buffer->depth, true); + RD::get_singleton()->texture_resolve_multisample(render_buffer->depth_msaa, render_buffer->depth); } 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(p_render_buffer, p_cam_projection); + RD::get_singleton()->draw_command_end_label(); } if (using_ssr) { RENDER_TIMESTAMP("Screen Space Reflection"); + RD::get_singleton()->draw_command_begin_label("Process Screen Space Reflections"); _process_ssr(p_render_buffer, render_buffer->color_fb, render_buffer->normal_roughness_buffer, render_buffer->specular, render_buffer->specular, Color(0, 0, 0, 1), p_environment, p_cam_projection, render_buffer->msaa == RS::VIEWPORT_MSAA_DISABLED); + RD::get_singleton()->draw_command_end_label(); } else { //just mix specular back RENDER_TIMESTAMP("Merge Specular"); @@ -1838,30 +1959,44 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf RENDER_TIMESTAMP("Render Transparent Pass"); - _setup_environment(p_environment, p_render_buffer, p_cam_projection, p_cam_transform, p_reflection_probe, p_reflection_probe.is_valid(), screen_pixel_size, p_shadow_atlas, !p_reflection_probe.is_valid(), p_default_bg_color, p_cam_projection.get_z_near(), p_cam_projection.get_z_far(), false); + RD::get_singleton()->draw_command_begin_label("Render Transparent Pass"); - render_list.sort_by_reverse_depth_and_priority(true); + rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_ALPHA, p_render_buffer, radiance_texture, p_shadow_atlas, p_reflection_atlas, p_cluster_buffer, p_gi_probes, p_lightmaps, true); + + _setup_environment(p_environment, p_render_buffer, p_cam_projection, p_cam_transform, p_reflection_probe, p_reflection_probe.is_valid(), screen_size, p_cluster_size, p_max_cluster_elements, p_shadow_atlas, !p_reflection_probe.is_valid(), p_default_bg_color, p_cam_projection.get_z_near(), p_cam_projection.get_z_far(), false); { - RenderListParameters render_list_params(&render_list.elements[render_list.max_elements - render_list.alpha_element_count], render_list.alpha_element_count, false, PASS_MODE_COLOR, render_buffer == nullptr, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); + 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, render_buffer == nullptr, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); _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(); + + RD::get_singleton()->draw_command_begin_label("Resolve"); + if (render_buffer && render_buffer->msaa != RS::VIEWPORT_MSAA_DISABLED) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->color_msaa, render_buffer->color, true); + RD::get_singleton()->texture_resolve_multisample(render_buffer->color_msaa, render_buffer->color); } -} -void RendererSceneRenderForward::_render_shadow(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform &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_lod_threshold) { - RENDER_TIMESTAMP("Setup Rendering Shadow"); + RD::get_singleton()->draw_command_end_label(); +} +void RendererSceneRenderForwardClustered::_render_shadow_begin() { + scene_state.shadow_passes.clear(); + RD::get_singleton()->draw_command_begin_label("Shadow Setup"); _update_render_base_uniform_set(); - render_pass++; + render_list[RENDER_LIST_SECONDARY].clear(); + scene_state.instance_data[RENDER_LIST_SECONDARY].clear(); +} +void RendererSceneRenderForwardClustered::_render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform &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_lod_threshold, const Rect2i &p_rect, bool p_flip_y, bool p_clear_region, bool p_begin, bool p_end) { + uint32_t shadow_pass_index = scene_state.shadow_passes.size(); + + SceneState::ShadowPass shadow_pass; scene_state.ubo.dual_paraboloid_side = p_use_dp_flip ? -1 : 1; - _setup_environment(RID(), RID(), p_projection, p_transform, RID(), true, Vector2(1, 1), RID(), true, Color(), 0, p_zfar, false, p_use_pancake); + _setup_environment(RID(), RID(), p_projection, p_transform, RID(), true, Vector2(1, 1), 1, 32, RID(), !p_flip_y, Color(), 0, p_zfar, false, p_use_pancake, shadow_pass_index); if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_DISABLE_LOD) { p_screen_lod_threshold = 0.0; @@ -1869,72 +2004,116 @@ void RendererSceneRenderForward::_render_shadow(RID p_framebuffer, const PagedAr PassMode pass_mode = p_use_dp ? PASS_MODE_SHADOW_DP : PASS_MODE_SHADOW; - _fill_render_list(p_instances, pass_mode, p_projection, p_transform); + uint32_t render_list_from = render_list[RENDER_LIST_SECONDARY].elements.size(); + _fill_render_list(RENDER_LIST_SECONDARY, p_instances, pass_mode, p_projection, p_transform, false, false, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold, 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, render_list_from, render_list_size, false); - RID rp_uniform_set = _setup_render_pass_uniform_set(RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); + { + //regular forward for now + bool flip_cull = p_use_dp_flip; + if (p_flip_y) { + flip_cull = !flip_cull; + } - RENDER_TIMESTAMP("Render Shadow"); + 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; - render_list.sort_by_key(false); + 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_lod_threshold = p_screen_lod_threshold; + shadow_pass.lod_distance_multiplier = p_lod_distance_multiplier; - { - //regular forward for now - RenderListParameters render_list_params(render_list.elements, render_list.element_count, p_use_dp_flip, pass_mode, true, rp_uniform_set, false, Vector2(), p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold); - _render_list_with_threads(&render_list_params, p_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ); + 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 RendererSceneRenderForward::_render_particle_collider_heightfield(RID p_fb, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<GeometryInstance *> &p_instances) { - RENDER_TIMESTAMP("Setup Render Collider Heightfield"); +void RendererSceneRenderForwardClustered::_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) - _update_render_base_uniform_set(); + 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, RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>(), false, i); + } + + RD::get_singleton()->draw_command_end_label(); +} +void RendererSceneRenderForwardClustered::_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, true, shadow_pass.rp_uniform_set, false, Vector2(), shadow_pass.camera_plane, shadow_pass.lod_distance_multiplier, shadow_pass.screen_lod_threshold, 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 RendererSceneRenderForwardClustered::_render_particle_collider_heightfield(RID p_fb, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<GeometryInstance *> &p_instances) { + RENDER_TIMESTAMP("Setup Render Collider Heightfield"); - render_pass++; + RD::get_singleton()->draw_command_begin_label("Render Collider Heightfield"); + _update_render_base_uniform_set(); scene_state.ubo.dual_paraboloid_side = 0; - _setup_environment(RID(), RID(), p_cam_projection, p_cam_transform, RID(), true, Vector2(1, 1), RID(), true, Color(), 0, p_cam_projection.get_z_far(), false, false); + _setup_environment(RID(), RID(), p_cam_projection, p_cam_transform, RID(), true, Vector2(1, 1), 1, 32, RID(), true, Color(), 0, p_cam_projection.get_z_far(), false, false); PassMode pass_mode = PASS_MODE_SHADOW; - _fill_render_list(p_instances, pass_mode, p_cam_projection, p_cam_transform); - - RID rp_uniform_set = _setup_render_pass_uniform_set(RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); + _fill_render_list(RENDER_LIST_SECONDARY, p_instances, pass_mode, p_cam_projection, p_cam_transform); + render_list[RENDER_LIST_SECONDARY].sort_by_key(); + _fill_instance_data(RENDER_LIST_SECONDARY); - RENDER_TIMESTAMP("Render Collider Heightield"); + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); - render_list.sort_by_key(false); + RENDER_TIMESTAMP("Render Collider Heightfield"); { //regular forward for now - RenderListParameters render_list_params(render_list.elements, render_list.element_count, false, pass_mode, true, rp_uniform_set); + 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, true, 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 RendererSceneRenderForward::_render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) { +void RendererSceneRenderForwardClustered::_render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) { RENDER_TIMESTAMP("Setup Rendering Material"); - _update_render_base_uniform_set(); + RD::get_singleton()->draw_command_begin_label("Render Material"); - render_pass++; + _update_render_base_uniform_set(); scene_state.ubo.dual_paraboloid_side = 0; - scene_state.ubo.material_uv2_mode = true; + scene_state.ubo.material_uv2_mode = false; - _setup_environment(RID(), RID(), p_cam_projection, p_cam_transform, RID(), true, Vector2(1, 1), RID(), false, Color(), 0, 0); + _setup_environment(RID(), RID(), p_cam_projection, p_cam_transform, RID(), true, Vector2(1, 1), 1, 32, RID(), false, Color(), 0, 0); PassMode pass_mode = PASS_MODE_DEPTH_MATERIAL; - _fill_render_list(p_instances, pass_mode, p_cam_projection, p_cam_transform); + _fill_render_list(RENDER_LIST_SECONDARY, p_instances, pass_mode, p_cam_projection, p_cam_transform); + render_list[RENDER_LIST_SECONDARY].sort_by_key(); + _fill_instance_data(RENDER_LIST_SECONDARY); - RID rp_uniform_set = _setup_render_pass_uniform_set(RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); RENDER_TIMESTAMP("Render Material"); - render_list.sort_by_key(false); - { - RenderListParameters render_list_params(render_list.elements, render_list.element_count, true, pass_mode, true, rp_uniform_set); + 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, true, rp_uniform_set); //regular forward for now Vector<Color> clear; clear.push_back(Color(0, 0, 0, 0)); @@ -1946,31 +2125,33 @@ void RendererSceneRenderForward::_render_material(const Transform &p_cam_transfo _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 RendererSceneRenderForward::_render_uv2(const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) { +void RendererSceneRenderForwardClustered::_render_uv2(const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) { RENDER_TIMESTAMP("Setup Rendering UV2"); - _update_render_base_uniform_set(); + RD::get_singleton()->draw_command_begin_label("Render UV2"); - render_pass++; + _update_render_base_uniform_set(); scene_state.ubo.dual_paraboloid_side = 0; scene_state.ubo.material_uv2_mode = true; - _setup_environment(RID(), RID(), CameraMatrix(), Transform(), RID(), true, Vector2(1, 1), RID(), false, Color(), 0, 0); + _setup_environment(RID(), RID(), CameraMatrix(), Transform(), RID(), true, Vector2(1, 1), 1, 32, RID(), false, Color(), 0, 0); PassMode pass_mode = PASS_MODE_DEPTH_MATERIAL; - _fill_render_list(p_instances, pass_mode, CameraMatrix(), Transform()); + _fill_render_list(RENDER_LIST_SECONDARY, p_instances, pass_mode, CameraMatrix(), Transform()); + render_list[RENDER_LIST_SECONDARY].sort_by_key(); + _fill_instance_data(RENDER_LIST_SECONDARY); - RID rp_uniform_set = _setup_render_pass_uniform_set(RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); RENDER_TIMESTAMP("Render Material"); - render_list.sort_by_key(false); - { - RenderListParameters render_list_params(render_list.elements, render_list.element_count, true, pass_mode, true, rp_uniform_set, true); + 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, true, rp_uniform_set, true); //regular forward for now Vector<Color> clear; clear.push_back(Color(0, 0, 0, 0)); @@ -2006,23 +2187,24 @@ void RendererSceneRenderForward::_render_uv2(const PagedArray<GeometryInstance * RD::get_singleton()->draw_list_end(); } + + RD::get_singleton()->draw_command_end_label(); } -void RendererSceneRenderForward::_render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<GeometryInstance *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture) { +void RendererSceneRenderForwardClustered::_render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<GeometryInstance *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture) { RENDER_TIMESTAMP("Render SDFGI"); + RD::get_singleton()->draw_command_begin_label("Render SDFGI Voxel"); + _update_render_base_uniform_set(); RenderBufferDataForward *render_buffer = (RenderBufferDataForward *)render_buffers_get_data(p_render_buffers); ERR_FAIL_COND(!render_buffer); - render_pass++; - PassMode pass_mode = PASS_MODE_SDF; - _fill_render_list(p_instances, pass_mode, CameraMatrix(), Transform()); - render_list.sort_by_key(false); - - RID rp_uniform_set = _setup_sdfgi_render_pass_uniform_set(p_albedo_texture, p_emission_texture, p_emission_aniso_texture, p_geom_facing_texture); + _fill_render_list(RENDER_LIST_SECONDARY, p_instances, pass_mode, CameraMatrix(), Transform()); + 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; @@ -2073,7 +2255,9 @@ void RendererSceneRenderForward::_render_sdfgi(RID p_render_buffers, const Vecto RendererStorageRD::store_transform(to_bounds.affine_inverse() * cam_xform, scene_state.ubo.sdf_to_bounds); - _setup_environment(RID(), RID(), camera_proj, cam_xform, RID(), true, Vector2(1, 1), RID(), false, Color(), 0, 0); + _setup_environment(RID(), RID(), camera_proj, cam_xform, RID(), true, Vector2(1, 1), 1, 32, RID(), false, Color(), 0, 0); + + RID rp_uniform_set = _setup_sdfgi_render_pass_uniform_set(p_albedo_texture, p_emission_texture, p_emission_aniso_texture, p_geom_facing_texture); Map<Size2i, RID>::Element *E = sdfgi_framebuffer_size_cache.find(fb_size); if (!E) { @@ -2081,19 +2265,21 @@ void RendererSceneRenderForward::_render_sdfgi(RID p_render_buffers, const Vecto E = sdfgi_framebuffer_size_cache.insert(fb_size, fb); } - RenderListParameters render_list_params(render_list.elements, render_list.element_count, true, pass_mode, true, rp_uniform_set, false); + 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, true, rp_uniform_set, false); _render_list_with_threads(&render_list_params, E->get(), 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 RendererSceneRenderForward::_base_uniforms_changed() { +void RendererSceneRenderForwardClustered::_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 RendererSceneRenderForward::_update_render_base_uniform_set() { +void RendererSceneRenderForwardClustered::_update_render_base_uniform_set() { if (render_base_uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(render_base_uniform_set) || (lightmap_texture_array_version != 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); @@ -2135,50 +2321,49 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() { { RD::Uniform u; u.binding = 3; - u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.ids.push_back(scene_state.uniform_buffer); + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.ids.push_back(get_omni_light_buffer()); uniforms.push_back(u); } - { RD::Uniform u; - u.binding = 5; + u.binding = 4; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.ids.push_back(get_positional_light_buffer()); + u.ids.push_back(get_spot_light_buffer()); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 6; + u.binding = 5; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(get_reflection_probe_buffer()); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 7; + u.binding = 6; u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; u.ids.push_back(get_directional_light_buffer()); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 10; + u.binding = 7; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(scene_state.lightmap_buffer); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 11; + u.binding = 8; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(scene_state.lightmap_capture_buffer); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 12; + u.binding = 9; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID decal_atlas = storage->decal_atlas_get_texture(); u.ids.push_back(decal_atlas); @@ -2186,7 +2371,7 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() { } { RD::Uniform u; - u.binding = 13; + u.binding = 10; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID decal_atlas = storage->decal_atlas_get_texture_srgb(); u.ids.push_back(decal_atlas); @@ -2194,7 +2379,7 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() { } { RD::Uniform u; - u.binding = 14; + u.binding = 11; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(get_decal_buffer()); uniforms.push_back(u); @@ -2202,35 +2387,8 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() { { RD::Uniform u; - u.binding = 15; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.ids.push_back(get_cluster_builder_texture()); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.binding = 16; - u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.ids.push_back(get_cluster_builder_indices_buffer()); - uniforms.push_back(u); - } - - { - RD::Uniform u; - u.binding = 17; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - if (directional_shadow_get_texture().is_valid()) { - u.ids.push_back(directional_shadow_get_texture()); - } else { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE)); - } - uniforms.push_back(u); - } - - { - RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.binding = 18; + u.binding = 12; u.ids.push_back(storage->global_variables_get_storage_buffer()); uniforms.push_back(u); } @@ -2238,7 +2396,7 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() { if (!low_end) { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.binding = 19; + u.binding = 13; u.ids.push_back(sdfgi_get_ubo()); uniforms.push_back(u); } @@ -2247,10 +2405,9 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() { } } -RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buffers, RID p_radiance_texture, RID p_shadow_atlas, RID p_reflection_atlas, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps) { - if (render_pass_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(render_pass_uniform_set)) { - RD::get_singleton()->free(render_pass_uniform_set); - } +RID RendererSceneRenderForwardClustered::_setup_render_pass_uniform_set(RenderListType p_render_list, RID p_render_buffers, RID p_radiance_texture, RID p_shadow_atlas, RID p_reflection_atlas, RID p_cluster_buffer, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, bool p_use_directional_shadow_atlas, int p_index) { + //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()); RenderBufferDataForward *rb = nullptr; if (p_render_buffers.is_valid()) { @@ -2262,6 +2419,24 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff Vector<RD::Uniform> uniforms; { + RD::Uniform u; + u.binding = 0; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.ids.push_back(scene_state.uniform_buffers[p_index]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 1; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + RID instance_buffer = scene_state.instance_buffer[p_render_list]; + if (instance_buffer == RID()) { + instance_buffer = default_vec4_xform_buffer; // any buffer will do since its not used + } + u.ids.push_back(instance_buffer); + uniforms.push_back(u); + } + { RID radiance_texture; if (p_radiance_texture.is_valid()) { radiance_texture = p_radiance_texture; @@ -2269,7 +2444,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff radiance_texture = storage->texture_rd_get_default(is_using_radiance_cubemap_array() ? RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_BLACK : RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK); } RD::Uniform u; - u.binding = 0; + u.binding = 2; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.ids.push_back(radiance_texture); uniforms.push_back(u); @@ -2278,7 +2453,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff { RID ref_texture = p_reflection_atlas.is_valid() ? reflection_atlas_get_texture(p_reflection_atlas) : RID(); RD::Uniform u; - u.binding = 1; + u.binding = 3; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; if (ref_texture.is_valid()) { u.ids.push_back(ref_texture); @@ -2290,7 +2465,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff { RD::Uniform u; - u.binding = 2; + u.binding = 4; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID texture; if (p_shadow_atlas.is_valid()) { @@ -2304,7 +2479,18 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff } { RD::Uniform u; - u.binding = 3; + u.binding = 5; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + if (p_use_directional_shadow_atlas && directional_shadow_get_texture().is_valid()) { + u.ids.push_back(directional_shadow_get_texture()); + } else { + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE)); + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 6; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.ids.resize(scene_state.max_lightmaps); RID default_tex = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE); @@ -2323,13 +2509,13 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff } { RD::Uniform u; - u.binding = 4; + u.binding = 7; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.ids.resize(MAX_GI_PROBES); RID default_tex = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE); for (int i = 0; i < MAX_GI_PROBES; i++) { if (i < (int)p_gi_probes.size()) { - RID tex = gi_probe_instance_get_texture(p_gi_probes[i]); + RID tex = gi.gi_probe_instance_get_texture(p_gi_probes[i]); if (!tex.is_valid()) { tex = default_tex; } @@ -2344,7 +2530,16 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff { RD::Uniform u; - u.binding = 5; + 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.ids.push_back(cb); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 9; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID texture = (false && rb && rb->depth.is_valid()) ? rb->depth : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE); u.ids.push_back(texture); @@ -2352,17 +2547,18 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff } { RD::Uniform u; - u.binding = 6; + u.binding = 10; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID bbt = rb ? render_buffers_get_back_buffer_texture(p_render_buffers) : RID(); RID texture = bbt.is_valid() ? bbt : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); u.ids.push_back(texture); uniforms.push_back(u); } + if (!low_end) { { RD::Uniform u; - u.binding = 7; + u.binding = 11; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID texture = rb && rb->normal_roughness_buffer.is_valid() ? rb->normal_roughness_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_NORMAL); u.ids.push_back(texture); @@ -2371,7 +2567,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff { RD::Uniform u; - u.binding = 8; + u.binding = 12; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID aot = rb ? render_buffers_get_ao_texture(p_render_buffers) : RID(); RID texture = aot.is_valid() ? aot : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); @@ -2381,24 +2577,26 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff { RD::Uniform u; - u.binding = 9; + u.binding = 13; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - RID texture = rb && rb->ambient_buffer.is_valid() ? rb->ambient_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); + RID ambient_buffer = p_render_buffers.is_valid() ? render_buffers_get_gi_ambient_texture(p_render_buffers) : RID(); + RID texture = ambient_buffer.is_valid() ? ambient_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); u.ids.push_back(texture); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 10; + u.binding = 14; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - RID texture = rb && rb->reflection_buffer.is_valid() ? rb->reflection_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); + RID reflection_buffer = p_render_buffers.is_valid() ? render_buffers_get_gi_reflection_texture(p_render_buffers) : RID(); + RID texture = reflection_buffer.is_valid() ? reflection_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); u.ids.push_back(texture); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 11; + u.binding = 15; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID t; if (rb && render_buffers_is_sdfgi_enabled(p_render_buffers)) { @@ -2411,7 +2609,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff } { RD::Uniform u; - u.binding = 12; + u.binding = 16; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; if (rb && render_buffers_is_sdfgi_enabled(p_render_buffers)) { u.ids.push_back(render_buffers_get_sdfgi_occlusion_texture(p_render_buffers)); @@ -2422,14 +2620,14 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff } { RD::Uniform u; - u.binding = 13; + u.binding = 17; u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; u.ids.push_back(rb ? render_buffers_get_gi_probe_buffer(p_render_buffers) : render_buffers_get_default_gi_probe_buffer()); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 14; + u.binding = 18; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID vfog = RID(); if (rb && render_buffers_has_volumetric_fog(p_render_buffers)) { @@ -2445,11 +2643,19 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff } } - render_pass_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, default_shader_rd, RENDER_PASS_UNIFORM_SET); - return render_pass_uniform_set; + 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, default_shader_rd, RENDER_PASS_UNIFORM_SET); + return render_pass_uniform_sets[p_index]; } -RID RendererSceneRenderForward::_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 RendererSceneRenderForwardClustered::_setup_sdfgi_render_pass_uniform_set(RID p_albedo_texture, RID p_emission_texture, RID p_emission_aniso_texture, RID p_geom_facing_texture) { if (sdfgi_pass_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(sdfgi_pass_uniform_set)) { RD::get_singleton()->free(sdfgi_pass_uniform_set); } @@ -2457,10 +2663,24 @@ RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albed Vector<RD::Uniform> uniforms; { + RD::Uniform u; + u.binding = 0; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.ids.push_back(scene_state.uniform_buffers[0]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 1; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.ids.push_back(scene_state.instance_buffer[RENDER_LIST_SECONDARY]); + uniforms.push_back(u); + } + { // No radiance texture. RID radiance_texture = storage->texture_rd_get_default(is_using_radiance_cubemap_array() ? RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_BLACK : RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK); RD::Uniform u; - u.binding = 0; + u.binding = 2; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.ids.push_back(radiance_texture); uniforms.push_back(u); @@ -2470,7 +2690,7 @@ RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albed // No reflection atlas. RID ref_texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_BLACK); RD::Uniform u; - u.binding = 1; + u.binding = 3; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.ids.push_back(ref_texture); uniforms.push_back(u); @@ -2479,7 +2699,17 @@ RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albed { // No shadow atlas. RD::Uniform u; - u.binding = 2; + u.binding = 4; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE); + u.ids.push_back(texture); + uniforms.push_back(u); + } + + { + // No directional shadow atlas. + RD::Uniform u; + u.binding = 5; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE); u.ids.push_back(texture); @@ -2489,7 +2719,7 @@ RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albed { // No Lightmaps RD::Uniform u; - u.binding = 3; + u.binding = 6; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.ids.resize(scene_state.max_lightmaps); RID default_tex = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE); @@ -2503,7 +2733,7 @@ RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albed { // No GIProbes RD::Uniform u; - u.binding = 4; + u.binding = 7; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.ids.resize(MAX_GI_PROBES); RID default_tex = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE); @@ -2513,33 +2743,43 @@ RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albed uniforms.push_back(u); } + + { + RD::Uniform u; + u.binding = 8; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + RID cb = default_vec4_xform_buffer; + u.ids.push_back(cb); + uniforms.push_back(u); + } + // actual sdfgi stuff { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 5; + u.binding = 9; u.ids.push_back(p_albedo_texture); uniforms.push_back(u); } { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 6; + u.binding = 10; u.ids.push_back(p_emission_texture); uniforms.push_back(u); } { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 7; + u.binding = 11; u.ids.push_back(p_emission_aniso_texture); uniforms.push_back(u); } { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 8; + u.binding = 12; u.ids.push_back(p_geom_facing_texture); uniforms.push_back(u); } @@ -2548,42 +2788,25 @@ RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albed return sdfgi_pass_uniform_set; } -void RendererSceneRenderForward::_render_buffers_clear_uniform_set(RenderBufferDataForward *rb) { +void RendererSceneRenderForwardClustered::_render_buffers_clear_uniform_set(RenderBufferDataForward *rb) { } -void RendererSceneRenderForward::_render_buffers_uniform_set_changed(RID p_render_buffers) { +void RendererSceneRenderForwardClustered::_render_buffers_uniform_set_changed(RID p_render_buffers) { RenderBufferDataForward *rb = (RenderBufferDataForward *)render_buffers_get_data(p_render_buffers); _render_buffers_clear_uniform_set(rb); } -RID RendererSceneRenderForward::_render_buffers_get_normal_texture(RID p_render_buffers) { +RID RendererSceneRenderForwardClustered::_render_buffers_get_normal_texture(RID p_render_buffers) { RenderBufferDataForward *rb = (RenderBufferDataForward *)render_buffers_get_data(p_render_buffers); return rb->normal_roughness_buffer; } -RID RendererSceneRenderForward::_render_buffers_get_ambient_texture(RID p_render_buffers) { - RenderBufferDataForward *rb = (RenderBufferDataForward *)render_buffers_get_data(p_render_buffers); - - return rb->ambient_buffer; -} - -RID RendererSceneRenderForward::_render_buffers_get_reflection_texture(RID p_render_buffers) { - RenderBufferDataForward *rb = (RenderBufferDataForward *)render_buffers_get_data(p_render_buffers); - - return rb->reflection_buffer; -} - -RendererSceneRenderForward *RendererSceneRenderForward::singleton = nullptr; - -void RendererSceneRenderForward::set_time(double p_time, double p_step) { - time = p_time; - RendererSceneRenderRD::set_time(p_time, p_step); -} +RendererSceneRenderForwardClustered *RendererSceneRenderForwardClustered::singleton = nullptr; -void RendererSceneRenderForward::_geometry_instance_mark_dirty(GeometryInstance *p_geometry_instance) { - GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); +void RendererSceneRenderForwardClustered::_geometry_instance_mark_dirty(GeometryInstance *p_geometry_instance) { + GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_geometry_instance); if (ginstance->dirty_list_element.in_list()) { return; } @@ -2602,7 +2825,7 @@ void RendererSceneRenderForward::_geometry_instance_mark_dirty(GeometryInstance geometry_instance_dirty_list.add(&ginstance->dirty_list_element); } -void RendererSceneRenderForward::_geometry_instance_add_surface_with_material(GeometryInstanceForward *ginstance, uint32_t p_surface, MaterialData *p_material, uint32_t p_material_id, uint32_t p_shader_id, RID p_mesh) { +void RendererSceneRenderForwardClustered::_geometry_instance_add_surface_with_material(GeometryInstanceForwardClustered *ginstance, uint32_t p_surface, MaterialData *p_material, uint32_t p_material_id, uint32_t p_shader_id, RID p_mesh) { 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 || has_read_screen_alpha); bool has_blend_alpha = p_material->shader_data->uses_blend_alpha; @@ -2626,7 +2849,7 @@ void RendererSceneRenderForward::_geometry_instance_add_surface_with_material(Ge flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_NORMAL_TEXTURE; } - if (ginstance->data->cast_double_sided_shaodows) { + if (ginstance->data->cast_double_sided_shadows) { flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_DOUBLE_SIDED_SHADOWS; } @@ -2644,10 +2867,17 @@ void RendererSceneRenderForward::_geometry_instance_add_surface_with_material(Ge } MaterialData *material_shadow = nullptr; - //void *surface_shadow = nullptr; + void *surface_shadow = nullptr; if (!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) { flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_SHARED_SHADOW_MATERIAL; material_shadow = (MaterialData *)storage->material_get_data(default_material, RendererStorageRD::SHADER_TYPE_3D); + + RID shadow_mesh = storage->mesh_get_shadow_mesh(p_mesh); + + if (shadow_mesh.is_valid()) { + surface_shadow = storage->mesh_get_surface(shadow_mesh, p_surface); + } + } else { material_shadow = p_material; } @@ -2669,7 +2899,8 @@ void RendererSceneRenderForward::_geometry_instance_add_surface_with_material(Ge //shadow sdcache->shader_shadow = material_shadow->shader_data; sdcache->material_uniform_set_shadow = material_shadow->uniform_set; - sdcache->surface_shadow = sdcache->surface; //when adding special shadow meshes, will use this + + sdcache->surface_shadow = surface_shadow ? surface_shadow : sdcache->surface; sdcache->owner = ginstance; @@ -2681,15 +2912,16 @@ void RendererSceneRenderForward::_geometry_instance_add_surface_with_material(Ge sdcache->sort.sort_key1 = 0; sdcache->sort.sort_key2 = 0; - sdcache->sort.surface_type = ginstance->data->base_type; - sdcache->sort.material_id = p_material_id; + sdcache->sort.surface_index = p_surface; + sdcache->sort.material_id_low = p_material_id & 0x3FFF; + sdcache->sort.material_id_hi = p_material_id >> 14; sdcache->sort.shader_id = p_shader_id; - sdcache->sort.geometry_id = p_mesh.get_local_index(); + 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; } -void RendererSceneRenderForward::_geometry_instance_add_surface(GeometryInstanceForward *ginstance, uint32_t p_surface, RID p_material, RID p_mesh) { +void RendererSceneRenderForwardClustered::_geometry_instance_add_surface(GeometryInstanceForwardClustered *ginstance, uint32_t p_surface, RID p_material, RID p_mesh) { RID m_src; m_src = ginstance->data->material_override.is_valid() ? ginstance->data->material_override : p_material; @@ -2729,8 +2961,8 @@ void RendererSceneRenderForward::_geometry_instance_add_surface(GeometryInstance } } -void RendererSceneRenderForward::_geometry_instance_update(GeometryInstance *p_geometry_instance) { - GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); +void RendererSceneRenderForwardClustered::_geometry_instance_update(GeometryInstance *p_geometry_instance) { + GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_geometry_instance); if (ginstance->data->dirty_dependencies) { ginstance->data->dependency_tracker.update_begin(); @@ -2814,11 +3046,6 @@ void RendererSceneRenderForward::_geometry_instance_update(GeometryInstance *p_g //Fill push constant - ginstance->push_constant.instance_uniforms_ofs = ginstance->data->shader_parameters_offset >= 0 ? ginstance->data->shader_parameters_offset : 0; - ginstance->push_constant.layer_mask = ginstance->data->layer_mask; - ginstance->push_constant.flags = 0; - ginstance->push_constant.gi_offset = 0xFFFFFFFF; //disabled - bool store_transform = true; if (ginstance->data->base_type == RS::INSTANCE_MULTIMESH) { @@ -2875,21 +3102,10 @@ void RendererSceneRenderForward::_geometry_instance_update(GeometryInstance *p_g } } - if (store_transform) { - RendererStorageRD::store_transform(ginstance->data->transform, ginstance->push_constant.transform); - } else { - RendererStorageRD::store_transform(Transform(), ginstance->push_constant.transform); - } - + ginstance->store_transform_cache = store_transform; ginstance->can_sdfgi = false; - if (lightmap_instance_is_valid(ginstance->lightmap_instance)) { - ginstance->push_constant.gi_offset = ginstance->data->lightmap_slice_index << 16; - ginstance->push_constant.lightmap_uv_scale[0] = ginstance->data->lightmap_uv_scale.position.x; - ginstance->push_constant.lightmap_uv_scale[1] = ginstance->data->lightmap_uv_scale.position.y; - ginstance->push_constant.lightmap_uv_scale[2] = ginstance->data->lightmap_uv_scale.size.width; - ginstance->push_constant.lightmap_uv_scale[3] = ginstance->data->lightmap_uv_scale.size.height; - } else if (!low_end) { + if (!lightmap_instance_is_valid(ginstance->lightmap_instance) && !low_end) { if (ginstance->gi_probes[0].is_null() && (ginstance->data->use_baked_light || ginstance->data->use_dynamic_gi)) { ginstance->can_sdfgi = true; } @@ -2903,24 +3119,24 @@ void RendererSceneRenderForward::_geometry_instance_update(GeometryInstance *p_g ginstance->dirty_list_element.remove_from_list(); } -void RendererSceneRenderForward::_update_dirty_geometry_instances() { +void RendererSceneRenderForwardClustered::_update_dirty_geometry_instances() { while (geometry_instance_dirty_list.first()) { _geometry_instance_update(geometry_instance_dirty_list.first()->self()); } } -void RendererSceneRenderForward::_geometry_instance_dependency_changed(RendererStorage::DependencyChangedNotification p_notification, RendererStorage::DependencyTracker *p_tracker) { +void RendererSceneRenderForwardClustered::_geometry_instance_dependency_changed(RendererStorage::DependencyChangedNotification p_notification, RendererStorage::DependencyTracker *p_tracker) { switch (p_notification) { case RendererStorage::DEPENDENCY_CHANGED_MATERIAL: case RendererStorage::DEPENDENCY_CHANGED_MESH: case RendererStorage::DEPENDENCY_CHANGED_MULTIMESH: case RendererStorage::DEPENDENCY_CHANGED_SKELETON_DATA: { - static_cast<RendererSceneRenderForward *>(singleton)->_geometry_instance_mark_dirty(static_cast<GeometryInstance *>(p_tracker->userdata)); + static_cast<RendererSceneRenderForwardClustered *>(singleton)->_geometry_instance_mark_dirty(static_cast<GeometryInstance *>(p_tracker->userdata)); } break; case RendererStorage::DEPENDENCY_CHANGED_MULTIMESH_VISIBLE_INSTANCES: { - GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_tracker->userdata); + GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_tracker->userdata); if (ginstance->data->base_type == RS::INSTANCE_MULTIMESH) { - ginstance->instance_count = static_cast<RendererSceneRenderForward *>(singleton)->storage->multimesh_get_instances_to_draw(ginstance->data->base); + ginstance->instance_count = static_cast<RendererSceneRenderForwardClustered *>(singleton)->storage->multimesh_get_instances_to_draw(ginstance->data->base); } } break; default: { @@ -2928,16 +3144,16 @@ void RendererSceneRenderForward::_geometry_instance_dependency_changed(RendererS } break; } } -void RendererSceneRenderForward::_geometry_instance_dependency_deleted(const RID &p_dependency, RendererStorage::DependencyTracker *p_tracker) { - static_cast<RendererSceneRenderForward *>(singleton)->_geometry_instance_mark_dirty(static_cast<GeometryInstance *>(p_tracker->userdata)); +void RendererSceneRenderForwardClustered::_geometry_instance_dependency_deleted(const RID &p_dependency, RendererStorage::DependencyTracker *p_tracker) { + static_cast<RendererSceneRenderForwardClustered *>(singleton)->_geometry_instance_mark_dirty(static_cast<GeometryInstance *>(p_tracker->userdata)); } -RendererSceneRender::GeometryInstance *RendererSceneRenderForward::geometry_instance_create(RID p_base) { +RendererSceneRender::GeometryInstance *RendererSceneRenderForwardClustered::geometry_instance_create(RID p_base) { RS::InstanceType type = storage->get_base_type(p_base); ERR_FAIL_COND_V(!((1 << type) & RS::INSTANCE_GEOMETRY_MASK), nullptr); - GeometryInstanceForward *ginstance = geometry_instance_alloc.alloc(); - ginstance->data = memnew(GeometryInstanceForward::Data); + GeometryInstanceForwardClustered *ginstance = geometry_instance_alloc.alloc(); + ginstance->data = memnew(GeometryInstanceForwardClustered::Data); ginstance->data->base = p_base; ginstance->data->base_type = type; @@ -2949,38 +3165,37 @@ RendererSceneRender::GeometryInstance *RendererSceneRenderForward::geometry_inst return ginstance; } -void RendererSceneRenderForward::geometry_instance_set_skeleton(GeometryInstance *p_geometry_instance, RID p_skeleton) { - GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); +void RendererSceneRenderForwardClustered::geometry_instance_set_skeleton(GeometryInstance *p_geometry_instance, RID p_skeleton) { + GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_geometry_instance); ERR_FAIL_COND(!ginstance); ginstance->data->skeleton = p_skeleton; _geometry_instance_mark_dirty(ginstance); ginstance->data->dirty_dependencies = true; } -void RendererSceneRenderForward::geometry_instance_set_material_override(GeometryInstance *p_geometry_instance, RID p_override) { - GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); +void RendererSceneRenderForwardClustered::geometry_instance_set_material_override(GeometryInstance *p_geometry_instance, RID p_override) { + GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_geometry_instance); ERR_FAIL_COND(!ginstance); ginstance->data->material_override = p_override; _geometry_instance_mark_dirty(ginstance); ginstance->data->dirty_dependencies = true; } -void RendererSceneRenderForward::geometry_instance_set_surface_materials(GeometryInstance *p_geometry_instance, const Vector<RID> &p_materials) { - GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); +void RendererSceneRenderForwardClustered::geometry_instance_set_surface_materials(GeometryInstance *p_geometry_instance, const Vector<RID> &p_materials) { + GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_geometry_instance); ERR_FAIL_COND(!ginstance); ginstance->data->surface_materials = p_materials; _geometry_instance_mark_dirty(ginstance); ginstance->data->dirty_dependencies = true; } -void RendererSceneRenderForward::geometry_instance_set_mesh_instance(GeometryInstance *p_geometry_instance, RID p_mesh_instance) { - GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); +void RendererSceneRenderForwardClustered::geometry_instance_set_mesh_instance(GeometryInstance *p_geometry_instance, RID p_mesh_instance) { + GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_geometry_instance); ERR_FAIL_COND(!ginstance); ginstance->mesh_instance = p_mesh_instance; _geometry_instance_mark_dirty(ginstance); } -void RendererSceneRenderForward::geometry_instance_set_transform(GeometryInstance *p_geometry_instance, const Transform &p_transform, const AABB &p_aabb, const AABB &p_transformed_aabb) { - GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); +void RendererSceneRenderForwardClustered::geometry_instance_set_transform(GeometryInstance *p_geometry_instance, const Transform &p_transform, const AABB &p_aabb, const AABB &p_transformed_aabb) { + GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_geometry_instance); ERR_FAIL_COND(!ginstance); - RendererStorageRD::store_transform(p_transform, ginstance->push_constant.transform); - ginstance->data->transform = p_transform; + ginstance->transform = p_transform; ginstance->mirror = p_transform.basis.determinant() < 0; ginstance->data->aabb = p_aabb; ginstance->transformed_aabb = p_transformed_aabb; @@ -2994,33 +3209,33 @@ void RendererSceneRenderForward::geometry_instance_set_transform(GeometryInstanc ginstance->lod_model_scale = max_scale; } -void RendererSceneRenderForward::geometry_instance_set_lod_bias(GeometryInstance *p_geometry_instance, float p_lod_bias) { - GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); +void RendererSceneRenderForwardClustered::geometry_instance_set_lod_bias(GeometryInstance *p_geometry_instance, float p_lod_bias) { + GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_geometry_instance); ERR_FAIL_COND(!ginstance); ginstance->lod_bias = p_lod_bias; } -void RendererSceneRenderForward::geometry_instance_set_use_baked_light(GeometryInstance *p_geometry_instance, bool p_enable) { - GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); +void RendererSceneRenderForwardClustered::geometry_instance_set_use_baked_light(GeometryInstance *p_geometry_instance, bool p_enable) { + GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_geometry_instance); ERR_FAIL_COND(!ginstance); ginstance->data->use_baked_light = p_enable; _geometry_instance_mark_dirty(ginstance); } -void RendererSceneRenderForward::geometry_instance_set_use_dynamic_gi(GeometryInstance *p_geometry_instance, bool p_enable) { - GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); +void RendererSceneRenderForwardClustered::geometry_instance_set_use_dynamic_gi(GeometryInstance *p_geometry_instance, bool p_enable) { + GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_geometry_instance); ERR_FAIL_COND(!ginstance); ginstance->data->use_dynamic_gi = p_enable; _geometry_instance_mark_dirty(ginstance); } -void RendererSceneRenderForward::geometry_instance_set_use_lightmap(GeometryInstance *p_geometry_instance, RID p_lightmap_instance, const Rect2 &p_lightmap_uv_scale, int p_lightmap_slice_index) { - GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); +void RendererSceneRenderForwardClustered::geometry_instance_set_use_lightmap(GeometryInstance *p_geometry_instance, RID p_lightmap_instance, const Rect2 &p_lightmap_uv_scale, int p_lightmap_slice_index) { + GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_geometry_instance); ERR_FAIL_COND(!ginstance); ginstance->lightmap_instance = p_lightmap_instance; - ginstance->data->lightmap_uv_scale = p_lightmap_uv_scale; - ginstance->data->lightmap_slice_index = p_lightmap_slice_index; + ginstance->lightmap_uv_scale = p_lightmap_uv_scale; + ginstance->lightmap_slice_index = p_lightmap_slice_index; _geometry_instance_mark_dirty(ginstance); } -void RendererSceneRenderForward::geometry_instance_set_lightmap_capture(GeometryInstance *p_geometry_instance, const Color *p_sh9) { - GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); +void RendererSceneRenderForwardClustered::geometry_instance_set_lightmap_capture(GeometryInstance *p_geometry_instance, const Color *p_sh9) { + GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_geometry_instance); ERR_FAIL_COND(!ginstance); if (p_sh9) { if (ginstance->lightmap_sh == nullptr) { @@ -3036,29 +3251,28 @@ void RendererSceneRenderForward::geometry_instance_set_lightmap_capture(Geometry } _geometry_instance_mark_dirty(ginstance); } -void RendererSceneRenderForward::geometry_instance_set_instance_shader_parameters_offset(GeometryInstance *p_geometry_instance, int32_t p_offset) { - GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); +void RendererSceneRenderForwardClustered::geometry_instance_set_instance_shader_parameters_offset(GeometryInstance *p_geometry_instance, int32_t p_offset) { + GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_geometry_instance); ERR_FAIL_COND(!ginstance); - ginstance->data->shader_parameters_offset = p_offset; + ginstance->shader_parameters_offset = p_offset; _geometry_instance_mark_dirty(ginstance); } -void RendererSceneRenderForward::geometry_instance_set_cast_double_sided_shadows(GeometryInstance *p_geometry_instance, bool p_enable) { - GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); +void RendererSceneRenderForwardClustered::geometry_instance_set_cast_double_sided_shadows(GeometryInstance *p_geometry_instance, bool p_enable) { + GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_geometry_instance); ERR_FAIL_COND(!ginstance); - ginstance->data->cast_double_sided_shaodows = p_enable; + ginstance->data->cast_double_sided_shadows = p_enable; _geometry_instance_mark_dirty(ginstance); } -void RendererSceneRenderForward::geometry_instance_set_layer_mask(GeometryInstance *p_geometry_instance, uint32_t p_layer_mask) { - GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); +void RendererSceneRenderForwardClustered::geometry_instance_set_layer_mask(GeometryInstance *p_geometry_instance, uint32_t p_layer_mask) { + GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_geometry_instance); ERR_FAIL_COND(!ginstance); - ginstance->data->layer_mask = p_layer_mask; - ginstance->push_constant.layer_mask = p_layer_mask; + ginstance->layer_mask = p_layer_mask; } -void RendererSceneRenderForward::geometry_instance_free(GeometryInstance *p_geometry_instance) { - GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); +void RendererSceneRenderForwardClustered::geometry_instance_free(GeometryInstance *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); @@ -3073,29 +3287,29 @@ void RendererSceneRenderForward::geometry_instance_free(GeometryInstance *p_geom geometry_instance_alloc.free(ginstance); } -uint32_t RendererSceneRenderForward::geometry_instance_get_pair_mask() { +uint32_t RendererSceneRenderForwardClustered::geometry_instance_get_pair_mask() { return (1 << RS::INSTANCE_GI_PROBE); } -void RendererSceneRenderForward::geometry_instance_pair_light_instances(GeometryInstance *p_geometry_instance, const RID *p_light_instances, uint32_t p_light_instance_count) { +void RendererSceneRenderForwardClustered::geometry_instance_pair_light_instances(GeometryInstance *p_geometry_instance, const RID *p_light_instances, uint32_t p_light_instance_count) { } -void RendererSceneRenderForward::geometry_instance_pair_reflection_probe_instances(GeometryInstance *p_geometry_instance, const RID *p_reflection_probe_instances, uint32_t p_reflection_probe_instance_count) { +void RendererSceneRenderForwardClustered::geometry_instance_pair_reflection_probe_instances(GeometryInstance *p_geometry_instance, const RID *p_reflection_probe_instances, uint32_t p_reflection_probe_instance_count) { } -void RendererSceneRenderForward::geometry_instance_pair_decal_instances(GeometryInstance *p_geometry_instance, const RID *p_decal_instances, uint32_t p_decal_instance_count) { +void RendererSceneRenderForwardClustered::geometry_instance_pair_decal_instances(GeometryInstance *p_geometry_instance, const RID *p_decal_instances, uint32_t p_decal_instance_count) { } -Transform RendererSceneRenderForward::geometry_instance_get_transform(GeometryInstance *p_instance) { - GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_instance); +Transform RendererSceneRenderForwardClustered::geometry_instance_get_transform(GeometryInstance *p_instance) { + GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_instance); ERR_FAIL_COND_V(!ginstance, Transform()); - return ginstance->data->transform; + return ginstance->transform; } -AABB RendererSceneRenderForward::geometry_instance_get_aabb(GeometryInstance *p_instance) { - GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_instance); +AABB RendererSceneRenderForwardClustered::geometry_instance_get_aabb(GeometryInstance *p_instance) { + GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_instance); ERR_FAIL_COND_V(!ginstance, AABB()); return ginstance->data->aabb; } -void RendererSceneRenderForward::geometry_instance_pair_gi_probe_instances(GeometryInstance *p_geometry_instance, const RID *p_gi_probe_instances, uint32_t p_gi_probe_instance_count) { - GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); +void RendererSceneRenderForwardClustered::geometry_instance_pair_gi_probe_instances(GeometryInstance *p_geometry_instance, const RID *p_gi_probe_instances, uint32_t p_gi_probe_instance_count) { + GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_geometry_instance); ERR_FAIL_COND(!ginstance); if (p_gi_probe_instance_count > 0) { ginstance->gi_probes[0] = p_gi_probe_instances[0]; @@ -3110,11 +3324,10 @@ void RendererSceneRenderForward::geometry_instance_pair_gi_probe_instances(Geome } } -RendererSceneRenderForward::RendererSceneRenderForward(RendererStorageRD *p_storage) : +RendererSceneRenderForwardClustered::RendererSceneRenderForwardClustered(RendererStorageRD *p_storage) : RendererSceneRenderRD(p_storage) { singleton = this; low_end = is_low_end(); - storage = p_storage; /* SCENE SHADER */ @@ -3128,7 +3341,7 @@ RendererSceneRenderForward::RendererSceneRenderForward(RendererStorageRD *p_stor if (is_using_radiance_cubemap_array()) { defines += "\n#define USE_RADIANCE_CUBEMAP_ARRAY \n"; } - defines += "\n#define SDFGI_OCT_SIZE " + itos(sdfgi_get_lightprobe_octahedron_size()) + "\n"; + defines += "\n#define SDFGI_OCT_SIZE " + itos(gi.sdfgi_get_lightprobe_octahedron_size()) + "\n"; defines += "\n#define MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS " + itos(get_max_directional_lights()) + "\n"; { @@ -3311,7 +3524,7 @@ RendererSceneRenderForward::RendererSceneRenderForward(RendererStorageRD *p_stor actions.render_mode_defines["cull_front"] = "#define DO_SIDE_CHECK\n"; actions.render_mode_defines["cull_disabled"] = "#define DO_SIDE_CHECK\n"; - bool force_lambert = GLOBAL_GET("rendering/quality/shading/force_lambert_over_burley"); + 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"; @@ -3323,7 +3536,7 @@ RendererSceneRenderForward::RendererSceneRenderForward(RendererStorageRD *p_stor actions.render_mode_defines["sss_mode_skin"] = "#define SSS_MODE_SKIN\n"; - bool force_blinn = GLOBAL_GET("rendering/quality/shading/force_blinn_over_ggx"); + bool force_blinn = GLOBAL_GET("rendering/shading/overrides/force_blinn_over_ggx"); if (!force_blinn) { actions.render_mode_defines["specular_schlick_ggx"] = "#define SPECULAR_SCHLICK_GGX\n"; @@ -3349,23 +3562,18 @@ RendererSceneRenderForward::RendererSceneRenderForward(RendererStorageRD *p_stor actions.default_filter = ShaderLanguage::FILTER_LINEAR_MIPMAP; actions.default_repeat = ShaderLanguage::REPEAT_ENABLE; actions.global_buffer_array_variable = "global_variables.data"; - actions.instance_uniform_index_variable = "instances.data[instance_index].instance_uniforms_ofs"; + actions.instance_uniform_index_variable = "draw_call.instance_uniforms_ofs"; shader.compiler.initialize(actions); } - //render list - render_list.max_elements = GLOBAL_DEF_RST("rendering/limits/rendering/max_renderable_elements", (int)128000); - render_list.init(); - render_pass = 0; - - scene_state.uniform_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(SceneState::UBO)); - { //default material and shader - default_shader = storage->shader_create(); + default_shader = storage->shader_allocate(); + storage->shader_initialize(default_shader); storage->shader_set_code(default_shader, "shader_type spatial; void vertex() { ROUGHNESS = 0.8; } void fragment() { ALBEDO=vec3(0.6); ROUGHNESS=0.8; METALLIC=0.2; } \n"); - default_material = storage->material_create(); + default_material = storage->material_allocate(); + storage->material_initialize(default_material); storage->material_set_shader(default_material, default_shader); MaterialData *md = (MaterialData *)storage->material_get_data(default_material, RendererStorageRD::SHADER_TYPE_3D); @@ -3376,14 +3584,18 @@ RendererSceneRenderForward::RendererSceneRenderForward(RendererStorageRD *p_stor } { - overdraw_material_shader = storage->shader_create(); + overdraw_material_shader = storage->shader_allocate(); + storage->shader_initialize(overdraw_material_shader); storage->shader_set_code(overdraw_material_shader, "shader_type spatial;\nrender_mode blend_add,unshaded;\n void fragment() { ALBEDO=vec3(0.4,0.8,0.8); ALPHA=0.2; }"); - overdraw_material = storage->material_create(); + overdraw_material = storage->material_allocate(); + storage->material_initialize(overdraw_material); storage->material_set_shader(overdraw_material, overdraw_material_shader); - wireframe_material_shader = storage->shader_create(); + wireframe_material_shader = storage->shader_allocate(); + storage->shader_initialize(wireframe_material_shader); storage->shader_set_code(wireframe_material_shader, "shader_type spatial;\nrender_mode wireframe,unshaded;\n void fragment() { ALBEDO=vec3(0.0,0.0,0.0); }"); - wireframe_material = storage->material_create(); + wireframe_material = storage->material_allocate(); + storage->material_initialize(wireframe_material); storage->material_set_shader(wireframe_material, wireframe_material_shader); } @@ -3407,15 +3619,17 @@ RendererSceneRenderForward::RendererSceneRenderForward(RendererStorageRD *p_stor shadow_sampler = RD::get_singleton()->sampler_create(sampler); } - render_list_thread_threshold = GLOBAL_GET("rendering/forward_renderer/threaded_render_minimum_instances"); + render_list_thread_threshold = GLOBAL_GET("rendering/limits/forward_renderer/threaded_render_minimum_instances"); } -RendererSceneRenderForward::~RendererSceneRenderForward() { +RendererSceneRenderForwardClustered::~RendererSceneRenderForwardClustered() { directional_shadow_atlas_set_size(0); //clear base uniform set if still valid - if (render_pass_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(render_pass_uniform_set)) { - RD::get_singleton()->free(render_pass_uniform_set); + 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]); + } } if (sdfgi_pass_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(sdfgi_pass_uniform_set)) { @@ -3434,9 +3648,16 @@ RendererSceneRenderForward::~RendererSceneRenderForward() { storage->free(default_material); { - RD::get_singleton()->free(scene_state.uniform_buffer); + 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); + 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); } diff --git a/servers/rendering/renderer_rd/renderer_scene_render_forward.h b/servers/rendering/renderer_rd/renderer_scene_render_forward_clustered.h index 8a6f268c46..98e2a7efcc 100644 --- a/servers/rendering/renderer_rd/renderer_scene_render_forward.h +++ b/servers/rendering/renderer_rd/renderer_scene_render_forward_clustered.h @@ -1,5 +1,5 @@ /*************************************************************************/ -/* renderer_scene_render_forward.h */ +/* renderer_scene_render_forward_clustered.h */ /*************************************************************************/ /* This file is part of: */ /* GODOT ENGINE */ @@ -28,16 +28,16 @@ /* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ /*************************************************************************/ -#ifndef RENDERING_SERVER_SCENE_RENDER_FORWARD_H -#define RENDERING_SERVER_SCENE_RENDER_FORWARD_H +#ifndef RENDERING_SERVER_SCENE_RENDER_FORWARD_CLUSTERED_H +#define RENDERING_SERVER_SCENE_RENDER_FORWARD_CLUSTERED_H #include "core/templates/paged_allocator.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/renderer_storage_rd.h" -#include "servers/rendering/renderer_rd/shaders/scene_forward.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl.gen.h" -class RendererSceneRenderForward : public RendererSceneRenderRD { +class RendererSceneRenderForwardClustered : public RendererSceneRenderRD { enum { SCENE_UNIFORM_SET = 0, RENDER_PASS_UNIFORM_SET = 1, @@ -50,6 +50,15 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { MAX_GI_PROBES = 8, MAX_LIGHTMAPS = 8, MAX_GI_PROBES_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 */ @@ -70,12 +79,10 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { }; struct { - SceneForwardShaderRD scene_shader; + SceneForwardClusteredShaderRD scene_shader; ShaderCompilerRD compiler; } shader; - RendererStorageRD *storage; - /* Material */ struct ShaderData : public RendererStorageRD::ShaderData { @@ -169,13 +176,15 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { 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; + ShaderData(); virtual ~ShaderData(); }; RendererStorageRD::ShaderData *_create_shader_func(); static RendererStorageRD::ShaderData *_create_shader_funcs() { - return static_cast<RendererSceneRenderForward *>(singleton)->_create_shader_func(); + return static_cast<RendererSceneRenderForwardClustered *>(singleton)->_create_shader_func(); } struct MaterialData : public RendererStorageRD::MaterialData { @@ -197,7 +206,7 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { RendererStorageRD::MaterialData *_create_material_func(ShaderData *p_shader); static RendererStorageRD::MaterialData *_create_material_funcs(RendererStorageRD::ShaderData *p_shader) { - return static_cast<RendererSceneRenderForward *>(singleton)->_create_material_func(static_cast<ShaderData *>(p_shader)); + return static_cast<RendererSceneRenderForwardClustered *>(singleton)->_create_material_func(static_cast<ShaderData *>(p_shader)); } /* Framebuffer */ @@ -211,9 +220,6 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { RID normal_roughness_buffer; RID giprobe_buffer; - RID ambient_buffer; - RID reflection_buffer; - RS::ViewportMSAA msaa; RD::TextureSamples texture_samples; @@ -234,7 +240,6 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { RID render_sdfgi_uniform_set; void ensure_specular(); - void ensure_gi(); void ensure_giprobe(); void clear(); virtual void configure(RID p_color_buffer, RID p_depth_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa); @@ -247,7 +252,7 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { RID shadow_sampler; RID render_base_uniform_set; - RID render_pass_uniform_set; + LocalVector<RID> render_pass_uniform_sets; RID sdfgi_pass_uniform_set; uint64_t lightmap_texture_array_version = 0xFFFFFFFF; @@ -256,12 +261,61 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { void _render_buffers_clear_uniform_set(RenderBufferDataForward *rb); virtual void _render_buffers_uniform_set_changed(RID p_render_buffers); virtual RID _render_buffers_get_normal_texture(RID p_render_buffers); - virtual RID _render_buffers_get_ambient_texture(RID p_render_buffers); - virtual RID _render_buffers_get_reflection_texture(RID p_render_buffers); 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(RID p_render_buffers, RID p_radiance_texture, RID p_shadow_atlas, RID p_reflection_atlas, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps); + RID _setup_render_pass_uniform_set(RenderListType p_render_list, RID p_render_buffers, RID p_radiance_texture, RID p_shadow_atlas, RID p_reflection_atlas, RID p_cluster_buffer, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, bool p_use_directional_shadow_atlas = false, int p_index = 0); + + 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_GIPROBE, + PASS_MODE_DEPTH_MATERIAL, + PASS_MODE_SDF, + }; + + 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; + bool no_gi = false; + RID render_pass_uniform_set; + bool force_wireframe = false; + Vector2 uv_offset; + Plane lod_plane; + float lod_distance_multiplier = 0.0; + float screen_lod_threshold = 0.0; + RD::FramebufferFormatID framebuffer_format = 0; + uint32_t element_offset = 0; + uint32_t barrier = RD::BARRIER_MASK_ALL; + + RenderListParameters(GeometryInstanceSurfaceDataCache **p_elements, RenderElementInfo *p_element_info, int p_element_count, bool p_reverse_cull, PassMode p_pass_mode, bool p_no_gi, RID p_render_pass_uniform_set, bool p_force_wireframe = false, const Vector2 &p_uv_offset = Vector2(), const Plane &p_lod_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, 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; + render_pass_uniform_set = p_render_pass_uniform_set; + force_wireframe = p_force_wireframe; + uv_offset = p_uv_offset; + lod_plane = p_lod_plane; + lod_distance_multiplier = p_lod_distance_multiplier; + screen_lod_threshold = p_screen_lod_threshold; + element_offset = p_element_offset; + barrier = p_barrier; + } + }; struct LightmapData { float normal_xform[12]; @@ -298,6 +352,11 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { float viewport_size[2]; float screen_pixel_size[2]; + uint32_t cluster_shift; + uint32_t cluster_width; + uint32_t cluster_type_size; + uint32_t max_cluster_element_count_div_32; + float directional_penumbra_shadow_kernel[128]; //32 vec4s float directional_soft_shadow_kernel[128]; float penumbra_shadow_kernel[128]; @@ -366,9 +425,24 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { uint32_t pancake_shadows; }; + struct PushConstant { + uint32_t base_index; // + uint32_t uv_offset; //packed + uint32_t pad[2]; + }; + + struct InstanceData { + 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; + float lightmap_uv_scale[4]; + }; + UBO ubo; - RID uniform_buffer; + LocalVector<RID> uniform_buffers; LightmapData lightmaps[MAX_LIGHTMAPS]; RID lightmap_ids[MAX_LIGHTMAPS]; @@ -377,6 +451,10 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { 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; uint32_t max_lightmap_captures; RID lightmap_capture_buffer; @@ -389,11 +467,29 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { 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_lod_threshold; + + RID framebuffer; + RD::InitialAction initial_depth_action; + RD::FinalAction final_depth_action; + Rect2i rect; + }; + + LocalVector<ShadowPass> shadow_passes; + } scene_state; - static RendererSceneRenderForward *singleton; - uint64_t render_pass; - double time; + static RendererSceneRenderForwardClustered *singleton; + RID default_shader; RID default_material; RID overdraw_material_shader; @@ -406,51 +502,15 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { RID default_vec4_xform_buffer; RID default_vec4_xform_uniform_set; - 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_GIPROBE, - PASS_MODE_DEPTH_MATERIAL, - PASS_MODE_SDF, - }; - - void _setup_environment(RID p_environment, RID p_render_buffers, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_reflection_probe, bool p_no_fog, const Size2 &p_screen_pixel_size, RID p_shadow_atlas, bool p_flip_y, const Color &p_default_bg_color, float p_znear, float p_zfar, bool p_opaque_render_buffers = false, bool p_pancake_shadows = false); + void _setup_environment(RID p_environment, RID p_render_buffers, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_reflection_probe, bool p_no_fog, const Size2i &p_screen_size, uint32_t p_cluster_size, uint32_t p_max_cluster_elements, RID p_shadow_atlas, bool p_flip_y, const Color &p_default_bg_color, float p_znear, float p_zfar, bool p_opaque_render_buffers = false, bool p_pancake_shadows = false, int p_index = 0); void _setup_giprobes(const PagedArray<RID> &p_giprobes); void _setup_lightmaps(const PagedArray<RID> &p_lightmaps, const Transform &p_cam_transform); - struct GeometryInstanceSurfaceDataCache; - - struct RenderListParameters { - GeometryInstanceSurfaceDataCache **elements = nullptr; - int element_count = 0; - bool reverse_cull = false; - PassMode pass_mode = PASS_MODE_COLOR; - bool no_gi = false; - RID render_pass_uniform_set; - bool force_wireframe = false; - Vector2 uv_offset; - Plane lod_plane; - float lod_distance_multiplier = 0.0; - float screen_lod_threshold = 0.0; - RD::FramebufferFormatID framebuffer_format = 0; - RenderListParameters(GeometryInstanceSurfaceDataCache **p_elements, int p_element_count, bool p_reverse_cull, PassMode p_pass_mode, bool p_no_gi, RID p_render_pass_uniform_set, bool p_force_wireframe = false, const Vector2 &p_uv_offset = Vector2(), const Plane &p_lod_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0) { - elements = p_elements; - element_count = p_element_count; - reverse_cull = p_reverse_cull; - pass_mode = p_pass_mode; - no_gi = p_no_gi; - render_pass_uniform_set = p_render_pass_uniform_set; - force_wireframe = p_force_wireframe; - uv_offset = p_uv_offset; - lod_plane = p_lod_plane; - lod_distance_multiplier = p_lod_distance_multiplier; - screen_lod_threshold = p_screen_lod_threshold; - } + struct RenderElementInfo { + uint32_t repeat : 22; + uint32_t uses_lightmap : 1; + uint32_t uses_forward_gi : 1; + uint32_t lod_index : 8; }; template <PassMode p_pass_mode> @@ -464,12 +524,14 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { uint32_t render_list_thread_threshold = 500; - void _fill_render_list(const PagedArray<GeometryInstance *> &p_instances, PassMode p_pass_mode, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, bool p_using_sdfgi = false, bool p_using_opaque_gi = false); + void _update_instance_data_buffer(RenderListType p_render_list); + void _fill_instance_data(RenderListType p_render_list, 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 PagedArray<GeometryInstance *> &p_instances, PassMode p_pass_mode, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, bool p_using_sdfgi = false, bool p_using_opaque_gi = false, const Plane &p_lod_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, bool p_append = false); Map<Size2i, RID> sdfgi_framebuffer_size_cache; struct GeometryInstanceData; - struct GeometryInstanceForward; + struct GeometryInstanceForwardClustered; struct GeometryInstanceLightmapSH { Color sh[9]; @@ -492,14 +554,17 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { union { struct { - uint32_t geometry_id; - uint32_t material_id; - uint32_t shader_id; - uint32_t surface_type : 4; - uint32_t uses_forward_gi : 1; //set during addition - uint32_t uses_lightmap : 1; //set during addition - uint32_t depth_layer : 4; //set during addition - uint32_t priority : 8; + uint64_t lod_index : 8; + uint64_t surface_index : 10; + uint64_t geometry_id : 32; + uint64_t material_id_low : 14; + + uint64_t material_id_hi : 18; + uint64_t shader_id : 32; + 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; @@ -520,10 +585,10 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { ShaderData *shader_shadow = nullptr; GeometryInstanceSurfaceDataCache *next = nullptr; - GeometryInstanceForward *owner = nullptr; + GeometryInstanceForwardClustered *owner = nullptr; }; - struct GeometryInstanceForward : public GeometryInstance { + struct GeometryInstanceForwardClustered : public GeometryInstance { //used during rendering bool mirror = false; bool non_uniform_scale = false; @@ -531,25 +596,25 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { float lod_model_scale = 1.0; AABB transformed_aabb; //needed for LOD float depth = 0; - 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; - float lightmap_uv_scale[4]; - } push_constant; + uint32_t gi_offset_cache = 0; + uint32_t flags_cache = 0; + bool store_transform_cache = true; + int32_t shader_parameters_offset = -1; + uint32_t lightmap_slice_index; + Rect2 lightmap_uv_scale; + uint32_t layer_mask = 1; RID transforms_uniform_set; uint32_t instance_count = 0; RID mesh_instance; bool can_sdfgi = false; //used during setup uint32_t base_flags = 0; + Transform transform; RID gi_probes[MAX_GI_PROBES_PER_INSTANCE]; RID lightmap_instance; GeometryInstanceLightmapSH *lightmap_sh = nullptr; GeometryInstanceSurfaceDataCache *surface_caches = nullptr; - SelfList<GeometryInstanceForward> dirty_list_element; + SelfList<GeometryInstanceForwardClustered> dirty_list_element; struct Data { //data used less often goes into regular heap @@ -557,21 +622,14 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { RS::InstanceType base_type; RID skeleton; - - uint32_t layer_mask = 1; - Vector<RID> surface_materials; RID material_override; - Transform transform; AABB aabb; - int32_t shader_parameters_offset = -1; bool use_dynamic_gi = false; bool use_baked_light = false; - bool cast_double_sided_shaodows = false; + bool cast_double_sided_shadows = false; bool mirror = false; - Rect2 lightmap_uv_scale; - uint32_t lightmap_slice_index = 0; bool dirty_dependencies = false; RendererStorage::DependencyTracker dependency_tracker; @@ -579,21 +637,21 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { Data *data = nullptr; - GeometryInstanceForward() : + GeometryInstanceForwardClustered() : dirty_list_element(this) {} }; static void _geometry_instance_dependency_changed(RendererStorage::DependencyChangedNotification p_notification, RendererStorage::DependencyTracker *p_tracker); static void _geometry_instance_dependency_deleted(const RID &p_dependency, RendererStorage::DependencyTracker *p_tracker); - SelfList<GeometryInstanceForward>::List geometry_instance_dirty_list; + SelfList<GeometryInstanceForwardClustered>::List geometry_instance_dirty_list; - PagedAllocator<GeometryInstanceForward> geometry_instance_alloc; + PagedAllocator<GeometryInstanceForwardClustered> geometry_instance_alloc; PagedAllocator<GeometryInstanceSurfaceDataCache> geometry_instance_surface_alloc; PagedAllocator<GeometryInstanceLightmapSH> geometry_instance_lightmap_sh; - void _geometry_instance_add_surface_with_material(GeometryInstanceForward *ginstance, uint32_t p_surface, MaterialData *p_material, uint32_t p_material_id, uint32_t p_shader_id, RID p_mesh); - void _geometry_instance_add_surface(GeometryInstanceForward *ginstance, uint32_t p_surface, RID p_material, RID p_mesh); + void _geometry_instance_add_surface_with_material(GeometryInstanceForwardClustered *ginstance, uint32_t p_surface, MaterialData *p_material, uint32_t p_material_id, uint32_t p_shader_id, RID p_mesh); + void _geometry_instance_add_surface(GeometryInstanceForwardClustered *ginstance, uint32_t p_surface, RID p_material, RID p_mesh); void _geometry_instance_mark_dirty(GeometryInstance *p_geometry_instance); void _geometry_instance_update(GeometryInstance *p_geometry_instance); void _update_dirty_geometry_instances(); @@ -603,16 +661,12 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { /* Render List */ struct RenderList { - int max_elements; - - GeometryInstanceSurfaceDataCache **elements = nullptr; - - int element_count; - int alpha_element_count; + LocalVector<GeometryInstanceSurfaceDataCache *> elements; + LocalVector<RenderElementInfo> element_info; void clear() { - element_count = 0; - alpha_element_count = 0; + elements.clear(); + element_info.clear(); } //should eventually be replaced by radix @@ -623,13 +677,14 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { } }; - void sort_by_key(bool p_alpha) { + void sort_by_key() { SortArray<GeometryInstanceSurfaceDataCache *, SortByKey> sorter; - if (p_alpha) { - sorter.sort(&elements[max_elements - alpha_element_count], alpha_element_count); - } else { - sorter.sort(elements, element_count); - } + 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 { @@ -638,14 +693,10 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { } }; - void sort_by_depth(bool p_alpha) { //used for shadows + void sort_by_depth() { //used for shadows SortArray<GeometryInstanceSurfaceDataCache *, SortByDepth> sorter; - if (p_alpha) { - sorter.sort(&elements[max_elements - alpha_element_count], alpha_element_count); - } else { - sorter.sort(elements, element_count); - } + sorter.sort(elements.ptr(), elements.size()); } struct SortByReverseDepthAndPriority { @@ -657,50 +708,24 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { void sort_by_reverse_depth_and_priority(bool p_alpha) { //used for alpha SortArray<GeometryInstanceSurfaceDataCache *, SortByReverseDepthAndPriority> sorter; - if (p_alpha) { - sorter.sort(&elements[max_elements - alpha_element_count], alpha_element_count); - } else { - sorter.sort(elements, element_count); - } + sorter.sort(elements.ptr(), elements.size()); } _FORCE_INLINE_ void add_element(GeometryInstanceSurfaceDataCache *p_element) { - if (element_count + alpha_element_count >= max_elements) { - return; - } - elements[element_count] = p_element; - element_count++; - } - - _FORCE_INLINE_ void add_alpha_element(GeometryInstanceSurfaceDataCache *p_element) { - if (element_count + alpha_element_count >= max_elements) { - return; - } - int idx = max_elements - alpha_element_count - 1; - elements[idx] = p_element; - alpha_element_count++; - } - - void init() { - element_count = 0; - alpha_element_count = 0; - elements = memnew_arr(GeometryInstanceSurfaceDataCache *, max_elements); - } - - RenderList() { - max_elements = 0; - } - - ~RenderList() { - memdelete_arr(elements); + elements.push_back(p_element); } }; - RenderList render_list; + RenderList render_list[RENDER_LIST_MAX]; protected: - virtual void _render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, int p_directional_light_count, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_bg_color, float p_lod_threshold); - virtual void _render_shadow(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform &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_lod_threshold = 0.0); + virtual void _render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_cluster_buffer, uint32_t p_cluster_size, uint32_t p_max_cluster_elements, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_bg_color, float p_lod_threshold); + + virtual void _render_shadow_begin(); + virtual void _render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform &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_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); + virtual void _render_shadow_process(); + virtual void _render_shadow_end(uint32_t p_barrier = RD::BARRIER_MASK_ALL); + virtual void _render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region); virtual void _render_uv2(const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region); virtual void _render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<GeometryInstance *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture); @@ -733,11 +758,9 @@ public: virtual void geometry_instance_pair_decal_instances(GeometryInstance *p_geometry_instance, const RID *p_decal_instances, uint32_t p_decal_instance_count); virtual void geometry_instance_pair_gi_probe_instances(GeometryInstance *p_geometry_instance, const RID *p_gi_probe_instances, uint32_t p_gi_probe_instance_count); - virtual void set_time(double p_time, double p_step); - virtual bool free(RID p_rid); - RendererSceneRenderForward(RendererStorageRD *p_storage); - ~RendererSceneRenderForward(); + RendererSceneRenderForwardClustered(RendererStorageRD *p_storage); + ~RendererSceneRenderForwardClustered(); }; -#endif // RASTERIZER_SCENE_HIGHEND_RD_H +#endif // !RENDERING_SERVER_SCENE_RENDER_FORWARD_CLUSTERED_H diff --git a/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp b/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp index dad08179e7..4cf296f0db 100644 --- a/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp @@ -35,8 +35,6 @@ #include "renderer_compositor_rd.h" #include "servers/rendering/rendering_server_default.h" -uint64_t RendererSceneRenderRD::auto_exposure_counter = 2; - void get_vogel_disk(float *r_kernel, int p_sample_count) { const float golden_angle = 2.4; @@ -49,975 +47,42 @@ void get_vogel_disk(float *r_kernel, int p_sample_count) { } } -void RendererSceneRenderRD::_clear_reflection_data(ReflectionData &rd) { - rd.layers.clear(); - rd.radiance_base_cubemap = RID(); - if (rd.downsampled_radiance_cubemap.is_valid()) { - RD::get_singleton()->free(rd.downsampled_radiance_cubemap); - } - rd.downsampled_radiance_cubemap = RID(); - rd.downsampled_layer.mipmaps.clear(); - rd.coefficient_buffer = RID(); -} - -void RendererSceneRenderRD::_update_reflection_data(ReflectionData &rd, int p_size, int p_mipmaps, bool p_use_array, RID p_base_cube, int p_base_layer, bool p_low_quality) { - //recreate radiance and all data - - int mipmaps = p_mipmaps; - uint32_t w = p_size, h = p_size; - - if (p_use_array) { - int layers = p_low_quality ? 8 : roughness_layers; - - for (int i = 0; i < 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, RD::TEXTURE_SLICE_CUBEMAP); - - mmw = MAX(1, mmw >> 1); - mmh = MAX(1, mmh >> 1); - } - - rd.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, RD::TEXTURE_SLICE_CUBEMAP); - - mmw = MAX(1, mmw >> 1); - mmh = MAX(1, mmh >> 1); - } - - rd.layers.push_back(layer); - } - - rd.radiance_base_cubemap = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_base_cube, p_base_layer, 0, RD::TEXTURE_SLICE_CUBEMAP); - - RD::TextureFormat tf; - tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; - tf.width = 64; // Always 64x64 - tf.height = 64; - tf.texture_type = RD::TEXTURE_TYPE_CUBE; - tf.array_layers = 6; - tf.mipmaps = 7; - tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; - - rd.downsampled_radiance_cubemap = RD::get_singleton()->texture_create(tf, RD::TextureView()); - { - uint32_t mmw = 64; - uint32_t mmh = 64; - rd.downsampled_layer.mipmaps.resize(7); - for (int j = 0; j < rd.downsampled_layer.mipmaps.size(); j++) { - ReflectionData::DownsampleLayer::Mipmap &mm = rd.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(), rd.downsampled_radiance_cubemap, 0, j, RD::TEXTURE_SLICE_CUBEMAP); - - mmw = MAX(1, mmw >> 1); - mmh = MAX(1, mmh >> 1); - } - } -} - -void RendererSceneRenderRD::_create_reflection_fast_filter(ReflectionData &rd, bool p_use_arrays) { - storage->get_effects()->cubemap_downsample(rd.radiance_base_cubemap, rd.downsampled_layer.mipmaps[0].view, rd.downsampled_layer.mipmaps[0].size); - - for (int i = 1; i < rd.downsampled_layer.mipmaps.size(); i++) { - storage->get_effects()->cubemap_downsample(rd.downsampled_layer.mipmaps[i - 1].view, rd.downsampled_layer.mipmaps[i].view, rd.downsampled_layer.mipmaps[i].size); - } - - Vector<RID> views; - if (p_use_arrays) { - for (int i = 1; i < rd.layers.size(); i++) { - views.push_back(rd.layers[i].views[0]); - } - } else { - for (int i = 1; i < rd.layers[0].views.size(); i++) { - views.push_back(rd.layers[0].views[i]); - } - } - - storage->get_effects()->cubemap_filter(rd.downsampled_radiance_cubemap, views, p_use_arrays); -} - -void RendererSceneRenderRD::_create_reflection_importance_sample(ReflectionData &rd, bool p_use_arrays, int p_cube_side, int p_base_layer) { - if (p_use_arrays) { - //render directly to the layers - storage->get_effects()->cubemap_roughness(rd.radiance_base_cubemap, rd.layers[p_base_layer].views[0], p_cube_side, sky_ggx_samples_quality, float(p_base_layer) / (rd.layers.size() - 1.0), rd.layers[p_base_layer].mipmaps[0].size.x); - } else { - storage->get_effects()->cubemap_roughness(rd.layers[0].views[p_base_layer - 1], rd.layers[0].views[p_base_layer], p_cube_side, sky_ggx_samples_quality, float(p_base_layer) / (rd.layers[0].mipmaps.size() - 1.0), rd.layers[0].mipmaps[p_base_layer].size.x); - } -} - -void RendererSceneRenderRD::_update_reflection_mipmaps(ReflectionData &rd, int p_start, int p_end) { - for (int i = p_start; i < p_end; i++) { - for (int j = 0; j < rd.layers[i].mipmaps.size() - 1; j++) { - for (int k = 0; k < 6; k++) { - RID view = rd.layers[i].mipmaps[j].views[k]; - RID texture = rd.layers[i].mipmaps[j + 1].views[k]; - Size2i size = rd.layers[i].mipmaps[j + 1].size; - storage->get_effects()->make_mipmap(view, texture, size); - } - } - } -} - -void RendererSceneRenderRD::_sdfgi_erase(RenderBuffers *rb) { - for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) { - const SDFGI::Cascade &c = rb->sdfgi->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(rb->sdfgi->render_albedo); - RD::get_singleton()->free(rb->sdfgi->render_emission); - RD::get_singleton()->free(rb->sdfgi->render_emission_aniso); - - RD::get_singleton()->free(rb->sdfgi->render_sdf[0]); - RD::get_singleton()->free(rb->sdfgi->render_sdf[1]); - - RD::get_singleton()->free(rb->sdfgi->render_sdf_half[0]); - RD::get_singleton()->free(rb->sdfgi->render_sdf_half[1]); - - for (int i = 0; i < 8; i++) { - RD::get_singleton()->free(rb->sdfgi->render_occlusion[i]); - } - - RD::get_singleton()->free(rb->sdfgi->render_geom_facing); - - RD::get_singleton()->free(rb->sdfgi->lightprobe_data); - RD::get_singleton()->free(rb->sdfgi->lightprobe_history_scroll); - RD::get_singleton()->free(rb->sdfgi->occlusion_data); - RD::get_singleton()->free(rb->sdfgi->ambient_texture); - - RD::get_singleton()->free(rb->sdfgi->cascades_ubo); - - memdelete(rb->sdfgi); - - rb->sdfgi = nullptr; -} - -const Vector3i RendererSceneRenderRD::SDFGI::Cascade::DIRTY_ALL = Vector3i(0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF); - void RendererSceneRenderRD::sdfgi_update(RID p_render_buffers, RID p_environment, const Vector3 &p_world_position) { - Environment *env = environment_owner.getornull(p_environment); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_environment); RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); bool needs_sdfgi = env && env->sdfgi_enabled; if (!needs_sdfgi) { if (rb->sdfgi != nullptr) { //erase it - _sdfgi_erase(rb); + rb->sdfgi->erase(); + memdelete(rb->sdfgi); + rb->sdfgi = nullptr; + _render_buffers_uniform_set_changed(p_render_buffers); } 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[sdfgi_frames_to_converge]; + uint32_t requested_history_size = history_frames_to_converge[gi.sdfgi_frames_to_converge]; if (rb->sdfgi && (rb->sdfgi->cascade_mode != env->sdfgi_cascades || rb->sdfgi->min_cell_size != env->sdfgi_min_cell_size || requested_history_size != rb->sdfgi->history_size || rb->sdfgi->uses_occlusion != env->sdfgi_use_occlusion || rb->sdfgi->y_scale_mode != env->sdfgi_y_scale)) { //configuration changed, erase - _sdfgi_erase(rb); + rb->sdfgi->erase(); + memdelete(rb->sdfgi); + rb->sdfgi = nullptr; } - SDFGI *sdfgi = rb->sdfgi; + RendererSceneGIRD::SDFGI *sdfgi = rb->sdfgi; if (sdfgi == nullptr) { - //re-create - rb->sdfgi = memnew(SDFGI); - sdfgi = rb->sdfgi; - sdfgi->cascade_mode = env->sdfgi_cascades; - sdfgi->min_cell_size = env->sdfgi_min_cell_size; - sdfgi->uses_occlusion = env->sdfgi_use_occlusion; - sdfgi->y_scale_mode = env->sdfgi_y_scale; - static const float y_scale[3] = { 1.0, 1.5, 2.0 }; - sdfgi->y_mult = y_scale[sdfgi->y_scale_mode]; - static const int cascasde_size[3] = { 4, 6, 8 }; - sdfgi->cascades.resize(cascasde_size[sdfgi->cascade_mode]); - sdfgi->probe_axis_count = SDFGI::PROBE_DIVISOR + 1; - sdfgi->solid_cell_ratio = sdfgi_solid_cell_ratio; - sdfgi->solid_cell_count = uint32_t(float(sdfgi->cascade_size * sdfgi->cascade_size * sdfgi->cascade_size) * sdfgi->solid_cell_ratio); - - float base_cell_size = sdfgi->min_cell_size; - - RD::TextureFormat tf_sdf; - tf_sdf.format = RD::DATA_FORMAT_R8_UNORM; - tf_sdf.width = sdfgi->cascade_size; // Always 64x64 - tf_sdf.height = sdfgi->cascade_size; - tf_sdf.depth = sdfgi->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; - sdfgi->render_albedo = RD::get_singleton()->texture_create(tf_render, RD::TextureView()); - tf_render.format = RD::DATA_FORMAT_R32_UINT; - sdfgi->render_emission = RD::get_singleton()->texture_create(tf_render, RD::TextureView()); - sdfgi->render_emission_aniso = RD::get_singleton()->texture_create(tf_render, RD::TextureView()); - - tf_render.format = RD::DATA_FORMAT_R8_UNORM; //at least its easy to visualize - - for (int i = 0; i < 8; i++) { - sdfgi->render_occlusion[i] = RD::get_singleton()->texture_create(tf_render, RD::TextureView()); - } - - tf_render.format = RD::DATA_FORMAT_R32_UINT; - sdfgi->render_geom_facing = RD::get_singleton()->texture_create(tf_render, RD::TextureView()); - - tf_render.format = RD::DATA_FORMAT_R8G8B8A8_UINT; - sdfgi->render_sdf[0] = RD::get_singleton()->texture_create(tf_render, RD::TextureView()); - sdfgi->render_sdf[1] = RD::get_singleton()->texture_create(tf_render, RD::TextureView()); - - tf_render.width /= 2; - tf_render.height /= 2; - tf_render.depth /= 2; - - sdfgi->render_sdf_half[0] = RD::get_singleton()->texture_create(tf_render, RD::TextureView()); - sdfgi->render_sdf_half[1] = RD::get_singleton()->texture_create(tf_render, RD::TextureView()); - } - - 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 *= sdfgi->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(sdfgi->cascade_size) - 1; - - //store lightprobe SH - RD::TextureFormat tf_probes; - tf_probes.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; - tf_probes.width = sdfgi->probe_axis_count * sdfgi->probe_axis_count; - tf_probes.height = sdfgi->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; - - sdfgi->history_size = 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 = sdfgi->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; - - sdfgi->lightprobe_history_scroll = RD::get_singleton()->texture_create(tf_probe_history, RD::TextureView()); - sdfgi->lightprobe_average_scroll = RD::get_singleton()->texture_create(tf_probe_average, RD::TextureView()); - - { - //octahedral lightprobes - RD::TextureFormat tf_octprobes = tf_probes; - tf_octprobes.array_layers = sdfgi->cascades.size() * 2; - tf_octprobes.format = RD::DATA_FORMAT_R32_UINT; //pack well with RGBE - tf_octprobes.width = sdfgi->probe_axis_count * sdfgi->probe_axis_count * (SDFGI::LIGHTPROBE_OCT_SIZE + 2); - tf_octprobes.height = sdfgi->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 - - sdfgi->lightprobe_data = RD::get_singleton()->texture_create(tf_octprobes, RD::TextureView()); - RD::TextureView tv; - tv.format_override = RD::DATA_FORMAT_E5B9G9R9_UFLOAT_PACK32; - sdfgi->lightprobe_texture = RD::get_singleton()->texture_create_shared(tv, sdfgi->lightprobe_data); - - //texture handling ambient data, to integrate with volumetric foc - RD::TextureFormat tf_ambient = tf_probes; - tf_ambient.array_layers = sdfgi->cascades.size(); - tf_ambient.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; //pack well with RGBE - tf_ambient.width = sdfgi->probe_axis_count * sdfgi->probe_axis_count; - tf_ambient.height = sdfgi->probe_axis_count; - tf_ambient.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; - //lightprobe texture is an octahedral texture - sdfgi->ambient_texture = RD::get_singleton()->texture_create(tf_ambient, RD::TextureView()); - } - - sdfgi->cascades_ubo = RD::get_singleton()->uniform_buffer_create(sizeof(SDFGI::Cascade::UBO) * SDFGI::MAX_CASCADES); - - sdfgi->occlusion_data = RD::get_singleton()->texture_create(tf_occlusion, RD::TextureView()); - { - RD::TextureView tv; - tv.format_override = RD::DATA_FORMAT_R4G4B4A4_UNORM_PACK16; - sdfgi->occlusion_texture = RD::get_singleton()->texture_create_shared(tv, sdfgi->occlusion_data); - } - - for (uint32_t i = 0; i < sdfgi->cascades.size(); i++) { - SDFGI::Cascade &cascade = sdfgi->cascades[i]; - - /* 3D Textures */ - - cascade.sdf_tex = RD::get_singleton()->texture_create(tf_sdf, RD::TextureView()); - - cascade.light_data = RD::get_singleton()->texture_create(tf_light, RD::TextureView()); - - cascade.light_aniso_0_tex = RD::get_singleton()->texture_create(tf_aniso0, RD::TextureView()); - cascade.light_aniso_1_tex = RD::get_singleton()->texture_create(tf_aniso1, RD::TextureView()); - - { - 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 *= sdfgi->y_mult; - int32_t probe_cells = sdfgi->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()->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()->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) * sdfgi->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(SDGIShader::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.ids.push_back(sdfgi->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.ids.push_back(sdfgi->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.ids.push_back(sdfgi->render_occlusion[j]); - } - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 4; - u.ids.push_back(sdfgi->render_emission); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 5; - u.ids.push_back(sdfgi->render_emission_aniso); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 6; - u.ids.push_back(sdfgi->render_geom_facing); - uniforms.push_back(u); - } - - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 7; - u.ids.push_back(cascade.sdf_tex); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 8; - u.ids.push_back(sdfgi->occlusion_data); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.binding = 10; - u.ids.push_back(cascade.solid_cell_dispatch_buffer); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.binding = 11; - u.ids.push_back(cascade.solid_cell_buffer); - uniforms.push_back(u); - } - - cascade.sdf_store_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.preprocess.version_get_shader(sdfgi_shader.preprocess_shader, SDGIShader::PRE_PROCESS_STORE), 0); - } - - { - Vector<RD::Uniform> uniforms; - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 1; - u.ids.push_back(sdfgi->render_albedo); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 2; - u.ids.push_back(sdfgi->render_geom_facing); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 3; - u.ids.push_back(sdfgi->render_emission); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 4; - u.ids.push_back(sdfgi->render_emission_aniso); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.binding = 5; - u.ids.push_back(cascade.solid_cell_dispatch_buffer); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.binding = 6; - u.ids.push_back(cascade.solid_cell_buffer); - uniforms.push_back(u); - } - - cascade.scroll_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.preprocess.version_get_shader(sdfgi_shader.preprocess_shader, SDGIShader::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.ids.push_back(sdfgi->render_occlusion[j]); - } - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 2; - u.ids.push_back(sdfgi->occlusion_data); - uniforms.push_back(u); - } - - cascade.scroll_occlusion_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.preprocess.version_get_shader(sdfgi_shader.preprocess_shader, SDGIShader::PRE_PROCESS_SCROLL_OCCLUSION), 0); - } - } - - //direct light - for (uint32_t i = 0; i < sdfgi->cascades.size(); i++) { - SDFGI::Cascade &cascade = sdfgi->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 < rb->sdfgi->cascades.size()) { - u.ids.push_back(rb->sdfgi->cascades[j].sdf_tex); - } else { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE)); - } - } - uniforms.push_back(u); - } - { - RD::Uniform u; - u.binding = 2; - u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; - u.ids.push_back(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.ids.push_back(cascade.solid_cell_dispatch_buffer); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.binding = 4; - u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.ids.push_back(cascade.solid_cell_buffer); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.binding = 5; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.ids.push_back(cascade.light_data); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.binding = 6; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.ids.push_back(cascade.light_aniso_0_tex); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.binding = 7; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.ids.push_back(cascade.light_aniso_1_tex); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.binding = 8; - u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.ids.push_back(rb->sdfgi->cascades_ubo); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.binding = 9; - u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.ids.push_back(cascade.lights_buffer); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.binding = 10; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.ids.push_back(rb->sdfgi->lightprobe_texture); - uniforms.push_back(u); - } - - cascade.sdf_direct_light_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.direct_light.version_get_shader(sdfgi_shader.direct_light_shader, 0), 0); - } - - //preprocess initialize uniform set - { - Vector<RD::Uniform> uniforms; - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 1; - u.ids.push_back(sdfgi->render_albedo); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 2; - u.ids.push_back(sdfgi->render_sdf[0]); - uniforms.push_back(u); - } - - sdfgi->sdf_initialize_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.preprocess.version_get_shader(sdfgi_shader.preprocess_shader, SDGIShader::PRE_PROCESS_JUMP_FLOOD_INITIALIZE), 0); - } - - { - Vector<RD::Uniform> uniforms; - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 1; - u.ids.push_back(sdfgi->render_albedo); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 2; - u.ids.push_back(sdfgi->render_sdf_half[0]); - uniforms.push_back(u); - } - - sdfgi->sdf_initialize_half_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.preprocess.version_get_shader(sdfgi_shader.preprocess_shader, SDGIShader::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.ids.push_back(sdfgi->render_sdf[0]); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 2; - u.ids.push_back(sdfgi->render_sdf[1]); - uniforms.push_back(u); - } - - sdfgi->jump_flood_uniform_set[0] = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.preprocess.version_get_shader(sdfgi_shader.preprocess_shader, SDGIShader::PRE_PROCESS_JUMP_FLOOD), 0); - SWAP(uniforms.write[0].ids.write[0], uniforms.write[1].ids.write[0]); - sdfgi->jump_flood_uniform_set[1] = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.preprocess.version_get_shader(sdfgi_shader.preprocess_shader, SDGIShader::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.ids.push_back(sdfgi->render_sdf_half[0]); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 2; - u.ids.push_back(sdfgi->render_sdf_half[1]); - uniforms.push_back(u); - } - - sdfgi->jump_flood_half_uniform_set[0] = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.preprocess.version_get_shader(sdfgi_shader.preprocess_shader, SDGIShader::PRE_PROCESS_JUMP_FLOOD), 0); - SWAP(uniforms.write[0].ids.write[0], uniforms.write[1].ids.write[0]); - sdfgi->jump_flood_half_uniform_set[1] = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.preprocess.version_get_shader(sdfgi_shader.preprocess_shader, SDGIShader::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.ids.push_back(sdfgi->render_albedo); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 2; - u.ids.push_back(sdfgi->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.ids.push_back(sdfgi->render_sdf[(passes & 1) ? 0 : 1]); //reverse pass order because it needs an extra JFA pass - uniforms.push_back(u); - } - - sdfgi->upscale_jfa_uniform_set_index = (passes & 1) ? 0 : 1; - sdfgi->sdf_upscale_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.preprocess.version_get_shader(sdfgi_shader.preprocess_shader, SDGIShader::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.ids.push_back(sdfgi->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.ids.push_back(sdfgi->render_occlusion[i]); - } - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 3; - u.ids.push_back(sdfgi->render_geom_facing); - uniforms.push_back(u); - } - - sdfgi->occlusion_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.preprocess.version_get_shader(sdfgi_shader.preprocess_shader, SDGIShader::PRE_PROCESS_OCCLUSION), 0); - } - - for (uint32_t i = 0; i < sdfgi->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 < sdfgi->cascades.size()) { - u.ids.push_back(sdfgi->cascades[j].sdf_tex); - } else { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::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 < sdfgi->cascades.size()) { - u.ids.push_back(sdfgi->cascades[j].light_tex); - } else { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::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 < sdfgi->cascades.size()) { - u.ids.push_back(sdfgi->cascades[j].light_aniso_0_tex); - } else { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::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 < sdfgi->cascades.size()) { - u.ids.push_back(sdfgi->cascades[j].light_aniso_1_tex); - } else { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE)); - } - } - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; - u.binding = 6; - u.ids.push_back(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.ids.push_back(sdfgi->cascades_ubo); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 8; - u.ids.push_back(sdfgi->lightprobe_data); - uniforms.push_back(u); - } - - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 9; - u.ids.push_back(sdfgi->cascades[i].lightprobe_history_tex); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 10; - u.ids.push_back(sdfgi->cascades[i].lightprobe_average_tex); - uniforms.push_back(u); - } - - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 11; - u.ids.push_back(sdfgi->lightprobe_history_scroll); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 12; - u.ids.push_back(sdfgi->lightprobe_average_scroll); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 13; - RID parent_average; - if (i < sdfgi->cascades.size() - 1) { - parent_average = sdfgi->cascades[i + 1].lightprobe_average_tex; - } else { - parent_average = sdfgi->cascades[i - 1].lightprobe_average_tex; //to use something, but it won't be used - } - u.ids.push_back(parent_average); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 14; - u.ids.push_back(sdfgi->ambient_texture); - uniforms.push_back(u); - } - - sdfgi->cascades[i].integrate_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.integrate.version_get_shader(sdfgi_shader.integrate_shader, 0), 0); - } - - sdfgi->uses_multibounce = env->sdfgi_use_multibounce; - sdfgi->energy = env->sdfgi_energy; - sdfgi->normal_bias = env->sdfgi_normal_bias; - sdfgi->probe_bias = env->sdfgi_probe_bias; - sdfgi->reads_sky = env->sdfgi_read_sky_light; + // re-create + rb->sdfgi = gi.create_sdfgi(env, p_world_position, requested_history_size); _render_buffers_uniform_set_changed(p_render_buffers); - - return; //done. all levels will need to be rendered which its going to take a bit - } - - //check for updates - - sdfgi->uses_multibounce = env->sdfgi_use_multibounce; - sdfgi->energy = env->sdfgi_energy; - sdfgi->normal_bias = env->sdfgi_normal_bias; - sdfgi->probe_bias = env->sdfgi_probe_bias; - sdfgi->reads_sky = env->sdfgi_read_sky_light; - - int32_t drag_margin = (sdfgi->cascade_size / SDFGI::PROBE_DIVISOR) / 2; - - for (uint32_t i = 0; i < sdfgi->cascades.size(); i++) { - SDFGI::Cascade &cascade = sdfgi->cascades[i]; - cascade.dirty_regions = Vector3i(); - - Vector3 probe_half_size = Vector3(1, 1, 1) * cascade.cell_size * float(sdfgi->cascade_size / SDFGI::PROBE_DIVISOR) * 0.5; - probe_half_size = Vector3(0, 0, 0); - - Vector3 world_position = p_world_position; - world_position.y *= sdfgi->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])) >= sdfgi->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 = sdfgi->cascade_size * sdfgi->cascade_size * sdfgi->cascade_size; - uint32_t safe_volume = 1; - for (int j = 0; j < 3; j++) { - safe_volume *= sdfgi->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; - } - } + } else { + //check for updates + rb->sdfgi->update(env, p_world_position); } } @@ -1032,9 +97,9 @@ int RendererSceneRenderRD::sdfgi_get_pending_region_count(RID p_render_buffers) int dirty_count = 0; for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) { - const SDFGI::Cascade &c = rb->sdfgi->cascades[i]; + const RendererSceneGIRD::SDFGI::Cascade &c = rb->sdfgi->cascades[i]; - if (c.dirty_regions == SDFGI::Cascade::DIRTY_ALL) { + if (c.dirty_regions == RendererSceneGIRD::SDFGI::Cascade::DIRTY_ALL) { dirty_count++; } else { for (int j = 0; j < 3; j++) { @@ -1048,72 +113,15 @@ int RendererSceneRenderRD::sdfgi_get_pending_region_count(RID p_render_buffers) return dirty_count; } -int RendererSceneRenderRD::_sdfgi_get_pending_region_data(RID p_render_buffers, int p_region, Vector3i &r_local_offset, Vector3i &r_local_size, AABB &r_bounds) const { - RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); - ERR_FAIL_COND_V(rb == nullptr, -1); - ERR_FAIL_COND_V(rb->sdfgi == nullptr, -1); - - int dirty_count = 0; - for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) { - const SDFGI::Cascade &c = rb->sdfgi->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) * rb->sdfgi->cascade_size; - - r_bounds.position = Vector3((Vector3i(1, 1, 1) * -int32_t(rb->sdfgi->cascade_size >> 1) + c.position)) * c.cell_size * Vector3(1, 1.0 / rb->sdfgi->y_mult, 1); - r_bounds.size = Vector3(r_local_size) * c.cell_size * Vector3(1, 1.0 / rb->sdfgi->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) * rb->sdfgi->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(rb->sdfgi->cascade_size >> 1) + c.position) * c.cell_size * Vector3(1, 1.0 / rb->sdfgi->y_mult, 1); - r_bounds.size = Vector3(r_local_size) * c.cell_size * Vector3(1, 1.0 / rb->sdfgi->y_mult, 1); - - return i; - } - - dirty_count++; - } - } - } - } - return -1; -} - AABB RendererSceneRenderRD::sdfgi_get_pending_region_bounds(RID p_render_buffers, int p_region) const { AABB bounds; Vector3i from; Vector3i size; + RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); + ERR_FAIL_COND_V(rb == nullptr, AABB()); + ERR_FAIL_COND_V(rb->sdfgi == nullptr, AABB()); - int c = _sdfgi_get_pending_region_data(p_render_buffers, p_region, from, size, bounds); + int c = rb->sdfgi->get_pending_region_data(p_region, from, size, bounds); ERR_FAIL_COND_V(c == -1, AABB()); return bounds; } @@ -1122,1841 +130,179 @@ uint32_t RendererSceneRenderRD::sdfgi_get_pending_region_cascade(RID p_render_bu AABB bounds; Vector3i from; Vector3i size; - - return _sdfgi_get_pending_region_data(p_render_buffers, p_region, from, size, bounds); -} - -void RendererSceneRenderRD::_sdfgi_update_cascades(RID p_render_buffers) { - RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); - ERR_FAIL_COND(rb == nullptr); - if (rb->sdfgi == nullptr) { - return; - } - - //update cascades - SDFGI::Cascade::UBO cascade_data[SDFGI::MAX_CASCADES]; - int32_t probe_divisor = rb->sdfgi->cascade_size / SDFGI::PROBE_DIVISOR; - - for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) { - Vector3 pos = Vector3((Vector3i(1, 1, 1) * -int32_t(rb->sdfgi->cascade_size >> 1) + rb->sdfgi->cascades[i].position)) * rb->sdfgi->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 / rb->sdfgi->cascades[i].cell_size; - cascade_data[i].probe_offset[0] = rb->sdfgi->cascades[i].position.x / probe_divisor; - cascade_data[i].probe_offset[1] = rb->sdfgi->cascades[i].position.y / probe_divisor; - cascade_data[i].probe_offset[2] = rb->sdfgi->cascades[i].position.z / probe_divisor; - cascade_data[i].pad = 0; - } - - RD::get_singleton()->buffer_update(rb->sdfgi->cascades_ubo, 0, sizeof(SDFGI::Cascade::UBO) * SDFGI::MAX_CASCADES, cascade_data, true); -} - -void RendererSceneRenderRD::sdfgi_update_probes(RID p_render_buffers, RID p_environment, const Vector<RID> &p_directional_lights, const RID *p_positional_light_instances, uint32_t p_positional_light_count) { - RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); - ERR_FAIL_COND(rb == nullptr); - if (rb->sdfgi == nullptr) { - return; - } - Environment *env = environment_owner.getornull(p_environment); - - RENDER_TIMESTAMP(">SDFGI Update Probes"); - - /* Update Cascades UBO */ - _sdfgi_update_cascades(p_render_buffers); - /* Update Dynamic Lights Buffer */ - - RENDER_TIMESTAMP("Update Lights"); - - /* Update dynamic lights */ - - { - int32_t cascade_light_count[SDFGI::MAX_CASCADES]; - - for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) { - SDFGI::Cascade &cascade = rb->sdfgi->cascades[i]; - - SDGIShader::Light lights[SDFGI::MAX_DYNAMIC_LIGHTS]; - uint32_t idx = 0; - for (uint32_t j = 0; j < (uint32_t)p_directional_lights.size(); j++) { - if (idx == SDFGI::MAX_DYNAMIC_LIGHTS) { - break; - } - - LightInstance *li = light_instance_owner.getornull(p_directional_lights[j]); - ERR_CONTINUE(!li); - - if (storage->light_directional_is_sky_only(li->light)) { - continue; - } - - Vector3 dir = -li->transform.basis.get_axis(Vector3::AXIS_Z); - dir.y *= rb->sdfgi->y_mult; - dir.normalize(); - lights[idx].direction[0] = dir.x; - lights[idx].direction[1] = dir.y; - lights[idx].direction[2] = dir.z; - Color color = storage->light_get_color(li->light); - color = color.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 = storage->light_get_param(li->light, RS::LIGHT_PARAM_ENERGY); - lights[idx].has_shadow = storage->light_has_shadow(li->light); - - idx++; - } - - AABB cascade_aabb; - cascade_aabb.position = Vector3((Vector3i(1, 1, 1) * -int32_t(rb->sdfgi->cascade_size >> 1) + cascade.position)) * cascade.cell_size; - cascade_aabb.size = Vector3(1, 1, 1) * rb->sdfgi->cascade_size * cascade.cell_size; - - for (uint32_t j = 0; j < p_positional_light_count; j++) { - if (idx == SDFGI::MAX_DYNAMIC_LIGHTS) { - break; - } - - LightInstance *li = light_instance_owner.getornull(p_positional_light_instances[j]); - ERR_CONTINUE(!li); - - uint32_t max_sdfgi_cascade = storage->light_get_max_sdfgi_cascade(li->light); - if (i > max_sdfgi_cascade) { - continue; - } - - if (!cascade_aabb.intersects(li->aabb)) { - continue; - } - - Vector3 dir = -li->transform.basis.get_axis(Vector3::AXIS_Z); - //faster to not do this here - //dir.y *= rb->sdfgi->y_mult; - //dir.normalize(); - lights[idx].direction[0] = dir.x; - lights[idx].direction[1] = dir.y; - lights[idx].direction[2] = dir.z; - Vector3 pos = li->transform.origin; - pos.y *= rb->sdfgi->y_mult; - lights[idx].position[0] = pos.x; - lights[idx].position[1] = pos.y; - lights[idx].position[2] = pos.z; - Color color = storage->light_get_color(li->light); - color = color.to_linear(); - lights[idx].color[0] = color.r; - lights[idx].color[1] = color.g; - lights[idx].color[2] = color.b; - lights[idx].type = storage->light_get_type(li->light); - lights[idx].energy = storage->light_get_param(li->light, RS::LIGHT_PARAM_ENERGY); - lights[idx].has_shadow = storage->light_has_shadow(li->light); - lights[idx].attenuation = storage->light_get_param(li->light, RS::LIGHT_PARAM_ATTENUATION); - lights[idx].radius = storage->light_get_param(li->light, RS::LIGHT_PARAM_RANGE); - lights[idx].spot_angle = Math::deg2rad(storage->light_get_param(li->light, RS::LIGHT_PARAM_SPOT_ANGLE)); - lights[idx].spot_attenuation = storage->light_get_param(li->light, RS::LIGHT_PARAM_SPOT_ATTENUATION); - - idx++; - } - - if (idx > 0) { - RD::get_singleton()->buffer_update(cascade.lights_buffer, 0, idx * sizeof(SDGIShader::Light), lights, true); - } - - cascade_light_count[i] = idx; - } - - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.direct_light_pipeline[SDGIShader::DIRECT_LIGHT_MODE_DYNAMIC]); - - SDGIShader::DirectLightPushConstant push_constant; - - push_constant.grid_size[0] = rb->sdfgi->cascade_size; - push_constant.grid_size[1] = rb->sdfgi->cascade_size; - push_constant.grid_size[2] = rb->sdfgi->cascade_size; - push_constant.max_cascades = rb->sdfgi->cascades.size(); - push_constant.probe_axis_size = rb->sdfgi->probe_axis_count; - push_constant.multibounce = rb->sdfgi->uses_multibounce; - push_constant.y_mult = rb->sdfgi->y_mult; - - push_constant.process_offset = 0; - push_constant.process_increment = 1; - - for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) { - SDFGI::Cascade &cascade = rb->sdfgi->cascades[i]; - push_constant.light_count = cascade_light_count[i]; - push_constant.cascade = i; - - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, cascade.sdf_direct_light_uniform_set, 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::DirectLightPushConstant)); - RD::get_singleton()->compute_list_dispatch_indirect(compute_list, cascade.solid_cell_dispatch_buffer, 0); - } - RD::get_singleton()->compute_list_end(); - } - - RENDER_TIMESTAMP("Raytrace"); - - SDGIShader::IntegratePushConstant push_constant; - push_constant.grid_size[1] = rb->sdfgi->cascade_size; - push_constant.grid_size[2] = rb->sdfgi->cascade_size; - push_constant.grid_size[0] = rb->sdfgi->cascade_size; - push_constant.max_cascades = rb->sdfgi->cascades.size(); - push_constant.probe_axis_size = rb->sdfgi->probe_axis_count; - push_constant.history_index = rb->sdfgi->render_pass % rb->sdfgi->history_size; - push_constant.history_size = rb->sdfgi->history_size; - static const uint32_t ray_count[RS::ENV_SDFGI_RAY_COUNT_MAX] = { 8, 16, 32, 64, 96, 128 }; - push_constant.ray_count = ray_count[sdfgi_ray_count]; - push_constant.ray_bias = rb->sdfgi->probe_bias; - push_constant.image_size[0] = rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count; - push_constant.image_size[1] = rb->sdfgi->probe_axis_count; - push_constant.store_ambient_texture = env->volumetric_fog_enabled; - - RID sky_uniform_set = sdfgi_shader.integrate_default_sky_uniform_set; - push_constant.sky_mode = SDGIShader::IntegratePushConstant::SKY_MODE_DISABLED; - push_constant.y_mult = rb->sdfgi->y_mult; - - if (rb->sdfgi->reads_sky && env) { - push_constant.sky_energy = env->bg_energy; - - if (env->background == RS::ENV_BG_CLEAR_COLOR) { - push_constant.sky_mode = SDGIShader::IntegratePushConstant::SKY_MODE_COLOR; - Color c = storage->get_default_clear_color().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 (env->background == RS::ENV_BG_COLOR) { - push_constant.sky_mode = SDGIShader::IntegratePushConstant::SKY_MODE_COLOR; - Color c = env->bg_color; - push_constant.sky_color[0] = c.r; - push_constant.sky_color[1] = c.g; - push_constant.sky_color[2] = c.b; - - } else if (env->background == RS::ENV_BG_SKY) { - Sky *sky = sky_owner.getornull(env->sky); - if (sky && sky->radiance.is_valid()) { - if (sky->sdfgi_integrate_sky_uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(sky->sdfgi_integrate_sky_uniform_set)) { - Vector<RD::Uniform> uniforms; - - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 0; - u.ids.push_back(sky->radiance); - uniforms.push_back(u); - } - - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; - u.binding = 1; - u.ids.push_back(storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED)); - uniforms.push_back(u); - } - - sky->sdfgi_integrate_sky_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.integrate.version_get_shader(sdfgi_shader.integrate_shader, 0), 1); - } - sky_uniform_set = sky->sdfgi_integrate_sky_uniform_set; - push_constant.sky_mode = SDGIShader::IntegratePushConstant::SKY_MODE_SKY; - } - } - } - - rb->sdfgi->render_pass++; - - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.integrate_pipeline[SDGIShader::INTEGRATE_MODE_PROCESS]); - - int32_t probe_divisor = rb->sdfgi->cascade_size / SDFGI::PROBE_DIVISOR; - for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) { - push_constant.cascade = i; - push_constant.world_offset[0] = rb->sdfgi->cascades[i].position.x / probe_divisor; - push_constant.world_offset[1] = rb->sdfgi->cascades[i].position.y / probe_divisor; - push_constant.world_offset[2] = rb->sdfgi->cascades[i].position.z / probe_divisor; - - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->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(SDGIShader::IntegratePushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count, rb->sdfgi->probe_axis_count, 1, 8, 8, 1); - } - - RD::get_singleton()->compute_list_add_barrier(compute_list); //wait until done - - // Then store values into the lightprobe texture. Separating these steps has a small performance hit, but it allows for multiple bounces - RENDER_TIMESTAMP("Average Probes"); - - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.integrate_pipeline[SDGIShader::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 < rb->sdfgi->cascades.size(); i++) { - push_constant.cascade = i; - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->cascades[i].integrate_uniform_set, 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::IntegratePushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count * SDFGI::LIGHTPROBE_OCT_SIZE, rb->sdfgi->probe_axis_count * SDFGI::LIGHTPROBE_OCT_SIZE, 1, 8, 8, 1); - } - - RD::get_singleton()->compute_list_end(); - - RENDER_TIMESTAMP("<SDFGI Update Probes"); -} - -void RendererSceneRenderRD::_setup_giprobes(RID p_render_buffers, const Transform &p_transform, const PagedArray<RID> &p_gi_probes, uint32_t &r_gi_probes_used) { - r_gi_probes_used = 0; RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); - ERR_FAIL_COND(rb == nullptr); - - RID gi_probe_buffer = render_buffers_get_gi_probe_buffer(p_render_buffers); - GI::GIProbeData gi_probe_data[RenderBuffers::MAX_GIPROBES]; - - bool giprobes_changed = false; - - Transform to_camera; - to_camera.origin = p_transform.origin; //only translation, make local - - for (int i = 0; i < RenderBuffers::MAX_GIPROBES; i++) { - RID texture; - if (i < (int)p_gi_probes.size()) { - GIProbeInstance *gipi = gi_probe_instance_owner.getornull(p_gi_probes[i]); - - if (gipi) { - texture = gipi->texture; - GI::GIProbeData &gipd = gi_probe_data[i]; - - RID base_probe = gipi->probe; - - Transform to_cell = storage->gi_probe_get_to_cell_xform(gipi->probe) * gipi->transform.affine_inverse() * to_camera; - - gipd.xform[0] = to_cell.basis.elements[0][0]; - gipd.xform[1] = to_cell.basis.elements[1][0]; - gipd.xform[2] = to_cell.basis.elements[2][0]; - gipd.xform[3] = 0; - gipd.xform[4] = to_cell.basis.elements[0][1]; - gipd.xform[5] = to_cell.basis.elements[1][1]; - gipd.xform[6] = to_cell.basis.elements[2][1]; - gipd.xform[7] = 0; - gipd.xform[8] = to_cell.basis.elements[0][2]; - gipd.xform[9] = to_cell.basis.elements[1][2]; - gipd.xform[10] = to_cell.basis.elements[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 = storage->gi_probe_get_octree_size(base_probe); - - gipd.bounds[0] = bounds.x; - gipd.bounds[1] = bounds.y; - gipd.bounds[2] = bounds.z; - - gipd.dynamic_range = storage->gi_probe_get_dynamic_range(base_probe) * storage->gi_probe_get_energy(base_probe); - gipd.bias = storage->gi_probe_get_bias(base_probe); - gipd.normal_bias = storage->gi_probe_get_normal_bias(base_probe); - gipd.blend_ambient = !storage->gi_probe_is_interior(base_probe); - gipd.anisotropy_strength = 0; - gipd.ao = storage->gi_probe_get_ao(base_probe); - gipd.ao_size = Math::pow(storage->gi_probe_get_ao_size(base_probe), 4.0f); - gipd.mipmaps = gipi->mipmaps.size(); - } - - r_gi_probes_used++; - } - - if (texture == RID()) { - texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE); - } - - if (texture != rb->giprobe_textures[i]) { - giprobes_changed = true; - rb->giprobe_textures[i] = texture; - } - } - - if (giprobes_changed) { - if (RD::get_singleton()->uniform_set_is_valid(rb->gi_uniform_set)) { - RD::get_singleton()->free(rb->gi_uniform_set); - } - rb->gi_uniform_set = RID(); - if (rb->volumetric_fog) { - if (RD::get_singleton()->uniform_set_is_valid(rb->volumetric_fog->uniform_set)) { - RD::get_singleton()->free(rb->volumetric_fog->uniform_set); - RD::get_singleton()->free(rb->volumetric_fog->uniform_set2); - } - rb->volumetric_fog->uniform_set = RID(); - rb->volumetric_fog->uniform_set2 = RID(); - } - } - - if (p_gi_probes.size() > 0) { - RD::get_singleton()->buffer_update(gi_probe_buffer, 0, sizeof(GI::GIProbeData) * MIN((uint64_t)RenderBuffers::MAX_GIPROBES, p_gi_probes.size()), gi_probe_data, true); - } -} - -void RendererSceneRenderRD::_process_gi(RID p_render_buffers, RID p_normal_roughness_buffer, RID p_ambient_buffer, RID p_reflection_buffer, RID p_gi_probe_buffer, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform, const PagedArray<RID> &p_gi_probes) { - RENDER_TIMESTAMP("Render GI"); - - RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); - ERR_FAIL_COND(rb == nullptr); - Environment *env = environment_owner.getornull(p_environment); - - GI::PushConstant push_constant; - - push_constant.screen_size[0] = rb->width; - push_constant.screen_size[1] = rb->height; - push_constant.z_near = p_projection.get_z_near(); - push_constant.z_far = p_projection.get_z_far(); - push_constant.orthogonal = p_projection.is_orthogonal(); - push_constant.proj_info[0] = -2.0f / (rb->width * p_projection.matrix[0][0]); - push_constant.proj_info[1] = -2.0f / (rb->height * p_projection.matrix[1][1]); - push_constant.proj_info[2] = (1.0f - p_projection.matrix[0][2]) / p_projection.matrix[0][0]; - push_constant.proj_info[3] = (1.0f + p_projection.matrix[1][2]) / p_projection.matrix[1][1]; - push_constant.max_giprobes = MIN((uint64_t)RenderBuffers::MAX_GIPROBES, p_gi_probes.size()); - push_constant.high_quality_vct = gi_probe_quality == RS::GI_PROBE_QUALITY_HIGH; - push_constant.use_sdfgi = rb->sdfgi != nullptr; - - if (env) { - push_constant.ao_color[0] = env->ao_color.r; - push_constant.ao_color[1] = env->ao_color.g; - push_constant.ao_color[2] = env->ao_color.b; - } else { - push_constant.ao_color[0] = 0; - push_constant.ao_color[1] = 0; - push_constant.ao_color[2] = 0; - } - - push_constant.cam_rotation[0] = p_transform.basis[0][0]; - push_constant.cam_rotation[1] = p_transform.basis[1][0]; - push_constant.cam_rotation[2] = p_transform.basis[2][0]; - push_constant.cam_rotation[3] = 0; - push_constant.cam_rotation[4] = p_transform.basis[0][1]; - push_constant.cam_rotation[5] = p_transform.basis[1][1]; - push_constant.cam_rotation[6] = p_transform.basis[2][1]; - push_constant.cam_rotation[7] = 0; - push_constant.cam_rotation[8] = p_transform.basis[0][2]; - push_constant.cam_rotation[9] = p_transform.basis[1][2]; - push_constant.cam_rotation[10] = p_transform.basis[2][2]; - push_constant.cam_rotation[11] = 0; - - if (rb->sdfgi) { - GI::SDFGIData sdfgi_data; - - sdfgi_data.grid_size[0] = rb->sdfgi->cascade_size; - sdfgi_data.grid_size[1] = rb->sdfgi->cascade_size; - sdfgi_data.grid_size[2] = rb->sdfgi->cascade_size; - - sdfgi_data.max_cascades = rb->sdfgi->cascades.size(); - sdfgi_data.probe_axis_size = rb->sdfgi->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 = rb->sdfgi->cascade_size; - sdfgi_data.probe_to_uvw = 1.0 / float(sdfgi_data.cascade_probe_size[0]); - sdfgi_data.use_occlusion = rb->sdfgi->uses_occlusion; - //sdfgi_data.energy = rb->sdfgi->energy; - - sdfgi_data.y_mult = rb->sdfgi->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 = (rb->sdfgi->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 = rb->sdfgi->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 = rb->sdfgi->cascade_size / SDFGI::PROBE_DIVISOR; - - for (uint32_t i = 0; i < sdfgi_data.max_cascades; i++) { - GI::SDFGIData::ProbeCascadeData &c = sdfgi_data.cascades[i]; - Vector3 pos = Vector3((Vector3i(1, 1, 1) * -int32_t(rb->sdfgi->cascade_size >> 1) + rb->sdfgi->cascades[i].position)) * rb->sdfgi->cascades[i].cell_size; - Vector3 cam_origin = p_transform.origin; - cam_origin.y *= rb->sdfgi->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(rb->sdfgi->cascade_size) * rb->sdfgi->cascades[i].cell_size / float(rb->sdfgi->probe_axis_count - 1)); - - Vector3i probe_ofs = rb->sdfgi->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 / rb->sdfgi->cascades[i].cell_size; - } - - RD::get_singleton()->buffer_update(gi.sdfgi_ubo, 0, sizeof(GI::SDFGIData), &sdfgi_data, true); - } - - if (rb->gi_uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(rb->gi_uniform_set)) { - 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 (rb->sdfgi && j < rb->sdfgi->cascades.size()) { - u.ids.push_back(rb->sdfgi->cascades[j].sdf_tex); - } else { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::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 (rb->sdfgi && j < rb->sdfgi->cascades.size()) { - u.ids.push_back(rb->sdfgi->cascades[j].light_tex); - } else { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::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 (rb->sdfgi && j < rb->sdfgi->cascades.size()) { - u.ids.push_back(rb->sdfgi->cascades[j].light_aniso_0_tex); - } else { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::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 (rb->sdfgi && j < rb->sdfgi->cascades.size()) { - u.ids.push_back(rb->sdfgi->cascades[j].light_aniso_1_tex); - } else { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE)); - } - } - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 5; - if (rb->sdfgi) { - u.ids.push_back(rb->sdfgi->occlusion_texture); - } else { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE)); - } - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; - u.binding = 6; - u.ids.push_back(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.ids.push_back(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.ids.push_back(p_ambient_buffer); - uniforms.push_back(u); - } - - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 10; - u.ids.push_back(p_reflection_buffer); - uniforms.push_back(u); - } - - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 11; - if (rb->sdfgi) { - u.ids.push_back(rb->sdfgi->lightprobe_texture); - } else { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE)); - } - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 12; - u.ids.push_back(rb->depth_texture); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 13; - u.ids.push_back(p_normal_roughness_buffer); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 14; - RID buffer = p_gi_probe_buffer.is_valid() ? p_gi_probe_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); - u.ids.push_back(buffer); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.binding = 15; - u.ids.push_back(gi.sdfgi_ubo); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.binding = 16; - u.ids.push_back(rb->giprobe_buffer); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 17; - for (int i = 0; i < RenderBuffers::MAX_GIPROBES; i++) { - u.ids.push_back(rb->giprobe_textures[i]); - } - uniforms.push_back(u); - } - - rb->gi_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi.shader.version_get_shader(gi.shader_version, 0), 0); - } + ERR_FAIL_COND_V(rb == nullptr, -1); + ERR_FAIL_COND_V(rb->sdfgi == nullptr, -1); - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi.pipelines[0]); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->gi_uniform_set, 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(GI::PushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->width, rb->height, 1, 8, 8, 1); - RD::get_singleton()->compute_list_end(); + return rb->sdfgi->get_pending_region_data(p_region, from, size, bounds); } -RID RendererSceneRenderRD::sky_create() { - return sky_owner.make_rid(Sky()); +RID RendererSceneRenderRD::sky_allocate() { + return sky.allocate_sky_rid(); } - -void RendererSceneRenderRD::_sky_invalidate(Sky *p_sky) { - if (!p_sky->dirty) { - p_sky->dirty = true; - p_sky->dirty_list = dirty_sky_list; - dirty_sky_list = p_sky; - } +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 = sky_owner.getornull(p_sky); - ERR_FAIL_COND(!sky); - ERR_FAIL_COND(p_radiance_size < 32 || p_radiance_size > 2048); - if (sky->radiance_size == p_radiance_size) { - return; - } - sky->radiance_size = p_radiance_size; - - if (sky->mode == RS::SKY_MODE_REALTIME && sky->radiance_size != 256) { - WARN_PRINT("Realtime Skies can only use a radiance size of 256. Radiance size will be set to 256 internally."); - sky->radiance_size = 256; - } - - _sky_invalidate(sky); - if (sky->radiance.is_valid()) { - RD::get_singleton()->free(sky->radiance); - sky->radiance = RID(); - } - _clear_reflection_data(sky->reflection); + sky.sky_set_radiance_size(p_sky, p_radiance_size); } void RendererSceneRenderRD::sky_set_mode(RID p_sky, RS::SkyMode p_mode) { - Sky *sky = sky_owner.getornull(p_sky); - ERR_FAIL_COND(!sky); - - if (sky->mode == p_mode) { - return; - } - - sky->mode = p_mode; - - if (sky->mode == RS::SKY_MODE_REALTIME && sky->radiance_size != 256) { - WARN_PRINT("Realtime Skies can only use a radiance size of 256. Radiance size will be set to 256 internally."); - sky_set_radiance_size(p_sky, 256); - } - - _sky_invalidate(sky); - if (sky->radiance.is_valid()) { - RD::get_singleton()->free(sky->radiance); - sky->radiance = RID(); - } - _clear_reflection_data(sky->reflection); + sky.sky_set_mode(p_sky, p_mode); } void RendererSceneRenderRD::sky_set_material(RID p_sky, RID p_material) { - Sky *sky = sky_owner.getornull(p_sky); - ERR_FAIL_COND(!sky); - sky->material = p_material; - _sky_invalidate(sky); + 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) { - Sky *sky = sky_owner.getornull(p_sky); - ERR_FAIL_COND_V(!sky, Ref<Image>()); - - _update_dirty_skys(); - - if (sky->radiance.is_valid()) { - RD::TextureFormat tf; - tf.format = RD::DATA_FORMAT_R32G32B32A32_SFLOAT; - 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()); - storage->get_effects()->copy_cubemap_to_panorama(sky->radiance, rad_tex, p_size, p_bake_irradiance ? roughness_layers : 0, sky->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; - img.instance(); - img->create(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>(); -} - -void RendererSceneRenderRD::_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 = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; - 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()); - - _update_reflection_data(sky->reflection, sky->radiance_size, mipmaps, true, sky->radiance, 0, sky->mode == RS::SKY_MODE_REALTIME); - - } else { - //regular cubemap, lower quality (aliasing, less memory) - RD::TextureFormat tf; - tf.array_layers = 6; - tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; - 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()); - - _update_reflection_data(sky->reflection, sky->radiance_size, MIN(mipmaps, layers), false, sky->radiance, 0, sky->mode == RS::SKY_MODE_REALTIME); - } - 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 = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; - 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 = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; - 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; + return sky.sky_bake_panorama(p_sky, p_energy, p_bake_irradiance, p_size); } -RID RendererSceneRenderRD::sky_get_radiance_texture_rd(RID p_sky) const { - Sky *sky = sky_owner.getornull(p_sky); - ERR_FAIL_COND_V(!sky, RID()); - - return sky->radiance; +RID RendererSceneRenderRD::environment_allocate() { + return environment_owner.allocate_rid(); } - -RID RendererSceneRenderRD::sky_get_radiance_uniform_set_rd(RID p_sky, RID p_shader, int p_set) const { - Sky *sky = sky_owner.getornull(p_sky); - ERR_FAIL_COND_V(!sky, RID()); - - if (sky->uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(sky->uniform_set)) { - sky->uniform_set = RID(); - if (sky->radiance.is_valid()) { - Vector<RD::Uniform> uniforms; - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 0; - u.ids.push_back(sky->radiance); - uniforms.push_back(u); - } - - sky->uniform_set = RD::get_singleton()->uniform_set_create(uniforms, p_shader, p_set); - } - } - - return sky->uniform_set; -} - -RID RendererSceneRenderRD::_get_sky_textures(Sky *p_sky, SkyTextureSetVersion p_version) { - if (p_sky->texture_uniform_sets[p_version].is_valid() && RD::get_singleton()->uniform_set_is_valid(p_sky->texture_uniform_sets[p_version])) { - return p_sky->texture_uniform_sets[p_version]; - } - Vector<RD::Uniform> uniforms; - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 0; - if (p_sky->radiance.is_valid() && p_version <= SKY_TEXTURE_SET_QUARTER_RES) { - u.ids.push_back(p_sky->radiance); - } else { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::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 (p_sky->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.ids.push_back(p_sky->reflection.layers[0].views[1]); - } else { - u.ids.push_back(p_sky->half_res_pass); - } - } else { - if (p_version < SKY_TEXTURE_SET_CUBEMAP) { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE)); - } else { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::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 (p_sky->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.ids.push_back(p_sky->reflection.layers[0].views[2]); - } else { - u.ids.push_back(p_sky->quarter_res_pass); - } - } else { - if (p_version < SKY_TEXTURE_SET_CUBEMAP) { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE)); - } else { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK)); - } - } - uniforms.push_back(u); - } - - p_sky->texture_uniform_sets[p_version] = RD::get_singleton()->uniform_set_create(uniforms, sky_shader.default_shader_rd, SKY_SET_TEXTURES); - return p_sky->texture_uniform_sets[p_version]; -} - -RID RendererSceneRenderRD::sky_get_material(RID p_sky) const { - Sky *sky = sky_owner.getornull(p_sky); - ERR_FAIL_COND_V(!sky, RID()); - - return sky->material; -} - -void RendererSceneRenderRD::_draw_sky(bool p_can_continue_color, bool p_can_continue_depth, RID p_fb, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform) { - ERR_FAIL_COND(!is_environment(p_environment)); - - SkyMaterialData *material = nullptr; - - Sky *sky = sky_owner.getornull(environment_get_sky(p_environment)); - - RID sky_material; - - RS::EnvironmentBG background = environment_get_background(p_environment); - - if (!(background == RS::ENV_BG_CLEAR_COLOR || background == RS::ENV_BG_COLOR) || sky) { - ERR_FAIL_COND(!sky); - sky_material = sky_get_material(environment_get_sky(p_environment)); - - if (sky_material.is_valid()) { - material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); - if (!material || !material->shader_data->valid) { - material = nullptr; - } - } - - if (!material) { - sky_material = sky_shader.default_material; - material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); - } - } - - if (background == RS::ENV_BG_CLEAR_COLOR || background == RS::ENV_BG_COLOR) { - sky_material = sky_scene_state.fog_material; - material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); - } - - ERR_FAIL_COND(!material); - - SkyShaderData *shader_data = material->shader_data; - - ERR_FAIL_COND(!shader_data); - - Basis sky_transform = environment_get_sky_orientation(p_environment); - sky_transform.invert(); - - float multiplier = environment_get_bg_energy(p_environment); - float custom_fov = environment_get_sky_custom_fov(p_environment); - // Camera - CameraMatrix camera; - - if (custom_fov) { - float near_plane = p_projection.get_z_near(); - float far_plane = p_projection.get_z_far(); - float aspect = p_projection.get_aspect(); - - camera.set_perspective(custom_fov, aspect, near_plane, far_plane); - - } else { - camera = p_projection; - } - - sky_transform = p_transform.basis * sky_transform; - - if (shader_data->uses_quarter_res) { - PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_QUARTER_RES]; - - RID texture_uniform_set = _get_sky_textures(sky, SKY_TEXTURE_SET_QUARTER_RES); - - 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); - storage->get_effects()->render_sky(draw_list, time, sky->quarter_res_framebuffer, sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, camera, sky_transform, multiplier, p_transform.origin); - RD::get_singleton()->draw_list_end(); - } - - if (shader_data->uses_half_res) { - PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_HALF_RES]; - - RID texture_uniform_set = _get_sky_textures(sky, SKY_TEXTURE_SET_HALF_RES); - - 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); - storage->get_effects()->render_sky(draw_list, time, sky->half_res_framebuffer, sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, camera, sky_transform, multiplier, p_transform.origin); - RD::get_singleton()->draw_list_end(); - } - - PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_BACKGROUND]; - - RID texture_uniform_set; - if (sky) { - texture_uniform_set = _get_sky_textures(sky, SKY_TEXTURE_SET_BACKGROUND); - } 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); - storage->get_effects()->render_sky(draw_list, time, p_fb, sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, camera, sky_transform, multiplier, p_transform.origin); - RD::get_singleton()->draw_list_end(); -} - -void RendererSceneRenderRD::_setup_sky(RID p_environment, RID p_render_buffers, const CameraMatrix &p_projection, const Transform &p_transform, const Size2i p_screen_size) { - ERR_FAIL_COND(!is_environment(p_environment)); - - SkyMaterialData *material = nullptr; - - Sky *sky = sky_owner.getornull(environment_get_sky(p_environment)); - - RID sky_material; - - SkyShaderData *shader_data = nullptr; - - RS::EnvironmentBG background = environment_get_background(p_environment); - - if (!(background == RS::ENV_BG_CLEAR_COLOR || background == RS::ENV_BG_COLOR) || sky) { - ERR_FAIL_COND(!sky); - sky_material = sky_get_material(environment_get_sky(p_environment)); - - if (sky_material.is_valid()) { - material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); - if (!material || !material->shader_data->valid) { - material = nullptr; - } - } - - if (!material) { - sky_material = sky_shader.default_material; - material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); - } - - ERR_FAIL_COND(!material); - - shader_data = material->shader_data; - - ERR_FAIL_COND(!shader_data); - } - - if (sky) { - // 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(); - } - _sky_invalidate(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(); - } - _sky_invalidate(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()) { - _sky_invalidate(sky); - _update_dirty_skys(); - } - - if (shader_data->uses_time && time - sky->prev_time > 0.00001) { - sky->prev_time = 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) { - // Check whether the directional_light_buffer changes - bool light_data_dirty = false; - - 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; - } - } - 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, true); - - RendererSceneRenderRD::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 (render_buffers_has_volumetric_fog(p_render_buffers)) { - sky_scene_state.ubo.volumetric_fog_enabled = true; - - float fog_end = render_buffers_get_volumetric_fog_end(p_render_buffers); - 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 = render_buffers_get_volumetric_fog_detail_spread(p_render_buffers); //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; - } - } - - RID fog_uniform_set = render_buffers_get_volumetric_fog_sky_uniform_set(p_render_buffers); - - if (fog_uniform_set != RID()) { - sky_scene_state.fog_uniform_set = fog_uniform_set; - } else { - sky_scene_state.fog_uniform_set = sky_scene_state.default_fog_uniform_set; - } - } - - sky_scene_state.ubo.z_far = p_projection.get_z_far(); - sky_scene_state.ubo.fog_enabled = environment_is_fog_enabled(p_environment); - sky_scene_state.ubo.fog_density = environment_get_fog_density(p_environment); - sky_scene_state.ubo.fog_aerial_perspective = environment_get_fog_aerial_perspective(p_environment); - Color fog_color = environment_get_fog_light_color(p_environment).to_linear(); - float fog_energy = environment_get_fog_light_energy(p_environment); - 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 = environment_get_fog_sun_scatter(p_environment); - - RD::get_singleton()->buffer_update(sky_scene_state.uniform_buffer, 0, sizeof(SkySceneState::UBO), &sky_scene_state.ubo, true); -} - -void RendererSceneRenderRD::_update_sky(RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform) { - ERR_FAIL_COND(!is_environment(p_environment)); - - Sky *sky = sky_owner.getornull(environment_get_sky(p_environment)); - ERR_FAIL_COND(!sky); - - RID sky_material = sky_get_material(environment_get_sky(p_environment)); - - SkyMaterialData *material = nullptr; - - if (sky_material.is_valid()) { - material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); - if (!material || !material->shader_data->valid) { - material = nullptr; - } - } - - if (!material) { - sky_material = sky_shader.default_material; - material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); - } - - ERR_FAIL_COND(!material); - - SkyShaderData *shader_data = material->shader_data; - - ERR_FAIL_COND(!shader_data); - - float multiplier = environment_get_bg_energy(p_environment); - - 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) - }; - - CameraMatrix cm; - cm.set_perspective(90, 1, 0.01, 10.0); - CameraMatrix correction; - correction.set_depth_correction(true); - cm = correction * cm; - - if (shader_data->uses_quarter_res) { - 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++) { - Transform local_view; - local_view.set_look_at(Vector3(0, 0, 0), view_normals[i], view_up[i]); - RID texture_uniform_set = _get_sky_textures(sky, SKY_TEXTURE_SET_CUBEMAP_QUARTER_RES); - - 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); - storage->get_effects()->render_sky(cubemap_draw_list, time, sky->reflection.layers[0].mipmaps[2].framebuffers[i], sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, cm, local_view.basis, multiplier, p_transform.origin); - RD::get_singleton()->draw_list_end(); - } - } - - if (shader_data->uses_half_res) { - 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++) { - Transform local_view; - local_view.set_look_at(Vector3(0, 0, 0), view_normals[i], view_up[i]); - RID texture_uniform_set = _get_sky_textures(sky, SKY_TEXTURE_SET_CUBEMAP_HALF_RES); - - 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); - storage->get_effects()->render_sky(cubemap_draw_list, time, sky->reflection.layers[0].mipmaps[1].framebuffers[i], sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, cm, local_view.basis, multiplier, p_transform.origin); - RD::get_singleton()->draw_list_end(); - } - } - - RD::DrawListID cubemap_draw_list; - PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_CUBEMAP]; - - for (int i = 0; i < 6; i++) { - Transform local_view; - local_view.set_look_at(Vector3(0, 0, 0), view_normals[i], view_up[i]); - RID texture_uniform_set = _get_sky_textures(sky, SKY_TEXTURE_SET_CUBEMAP); - - 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); - storage->get_effects()->render_sky(cubemap_draw_list, time, sky->reflection.layers[0].mipmaps[0].framebuffers[i], sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, cm, local_view.basis, multiplier, p_transform.origin); - RD::get_singleton()->draw_list_end(); - } - - if (sky_mode == RS::SKY_MODE_REALTIME) { - _create_reflection_fast_filter(sky->reflection, sky_use_cubemap_array); - if (sky_use_cubemap_array) { - _update_reflection_mipmaps(sky->reflection, 0, sky->reflection.layers.size()); - } - } else { - if (update_single_frame) { - for (int i = 1; i < max_processing_layer; i++) { - _create_reflection_importance_sample(sky->reflection, sky_use_cubemap_array, 10, i); - } - if (sky_use_cubemap_array) { - _update_reflection_mipmaps(sky->reflection, 0, sky->reflection.layers.size()); - } - } else { - if (sky_use_cubemap_array) { - // Multi-Frame so just update the first array level - _update_reflection_mipmaps(sky->reflection, 0, 1); - } - } - sky->processing_layer = 1; - } - - sky->reflection.dirty = false; - - } else { - if (sky_mode == RS::SKY_MODE_INCREMENTAL && sky->processing_layer < max_processing_layer) { - _create_reflection_importance_sample(sky->reflection, sky_use_cubemap_array, 10, sky->processing_layer); - - if (sky_use_cubemap_array) { - _update_reflection_mipmaps(sky->reflection, sky->processing_layer, sky->processing_layer + 1); - } - - sky->processing_layer++; - } - } -} - -/* SKY SHADER */ - -void RendererSceneRenderRD::SkyShaderData::set_code(const String &p_code) { - //compile - - code = p_code; - valid = false; - ubo_size = 0; - uniforms.clear(); - - if (code == String()) { - return; //just invalid, but no error - } - - ShaderCompilerRD::GeneratedCode gen_code; - ShaderCompilerRD::IdentifierActions actions; - - 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; - - RendererSceneRenderRD *scene_singleton = (RendererSceneRenderRD *)RendererSceneRenderRD::singleton; - - Error err = scene_singleton->sky_shader.compiler.compile(RS::SHADER_SKY, code, &actions, path, gen_code); - - ERR_FAIL_COND(err != OK); - - if (version.is_null()) { - version = scene_singleton->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]); - } - print_line("\n**uniforms:\n" + gen_code.uniforms); - // print_line("\n**vertex_globals:\n" + gen_code.vertex_global); - // print_line("\n**vertex_code:\n" + gen_code.vertex); - print_line("\n**fragment_globals:\n" + gen_code.fragment_global); - print_line("\n**fragment_code:\n" + gen_code.fragment); - print_line("\n**light_code:\n" + gen_code.light); -#endif - - scene_singleton->sky_shader.shader.version_set_code(version, gen_code.uniforms, gen_code.vertex_global, gen_code.vertex, gen_code.fragment_global, gen_code.light, gen_code.fragment, gen_code.defines); - ERR_FAIL_COND(!scene_singleton->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; - - RID shader_variant = scene_singleton->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); - } - - valid = true; -} - -void RendererSceneRenderRD::SkyShaderData::set_default_texture_param(const StringName &p_name, RID p_texture) { - if (!p_texture.is_valid()) { - default_texture_params.erase(p_name); - } else { - default_texture_params[p_name] = p_texture; - } -} - -void RendererSceneRenderRD::SkyShaderData::get_param_list(List<PropertyInfo> *p_param_list) const { - Map<int, StringName> order; - - for (Map<StringName, ShaderLanguage::ShaderNode::Uniform>::Element *E = uniforms.front(); E; E = E->next()) { - if (E->get().scope == ShaderLanguage::ShaderNode::Uniform::SCOPE_GLOBAL || E->get().scope == ShaderLanguage::ShaderNode::Uniform::SCOPE_INSTANCE) { - continue; - } - - if (E->get().texture_order >= 0) { - order[E->get().texture_order + 100000] = E->key(); - } else { - order[E->get().order] = E->key(); - } - } - - for (Map<int, StringName>::Element *E = order.front(); E; E = E->next()) { - PropertyInfo pi = ShaderLanguage::uniform_to_property_info(uniforms[E->get()]); - pi.name = E->get(); - p_param_list->push_back(pi); - } -} - -void RendererSceneRenderRD::SkyShaderData::get_instance_param_list(List<RendererStorage::InstanceShaderParam> *p_param_list) const { - for (Map<StringName, ShaderLanguage::ShaderNode::Uniform>::Element *E = uniforms.front(); E; E = E->next()) { - if (E->get().scope != ShaderLanguage::ShaderNode::Uniform::SCOPE_INSTANCE) { - continue; - } - - RendererStorage::InstanceShaderParam p; - p.info = ShaderLanguage::uniform_to_property_info(E->get()); - p.info.name = E->key(); //supply name - p.index = E->get().instance_index; - p.default_value = ShaderLanguage::constant_value_to_variant(E->get().default_value, E->get().type, E->get().hint); - p_param_list->push_back(p); - } -} - -bool RendererSceneRenderRD::SkyShaderData::is_param_texture(const StringName &p_param) const { - if (!uniforms.has(p_param)) { - return false; - } - - return uniforms[p_param].texture_order >= 0; -} - -bool RendererSceneRenderRD::SkyShaderData::is_animated() const { - return false; -} - -bool RendererSceneRenderRD::SkyShaderData::casts_shadows() const { - return false; -} - -Variant RendererSceneRenderRD::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.hint); - } - return Variant(); -} - -RendererSceneRenderRD::SkyShaderData::SkyShaderData() { - valid = false; -} - -RendererSceneRenderRD::SkyShaderData::~SkyShaderData() { - RendererSceneRenderRD *scene_singleton = (RendererSceneRenderRD *)RendererSceneRenderRD::singleton; - ERR_FAIL_COND(!scene_singleton); - //pipeline variants will clear themselves if shader is gone - if (version.is_valid()) { - scene_singleton->sky_shader.shader.version_free(version); - } -} - -RendererStorageRD::ShaderData *RendererSceneRenderRD::_create_sky_shader_func() { - SkyShaderData *shader_data = memnew(SkyShaderData); - return shader_data; -} - -void RendererSceneRenderRD::SkyMaterialData::update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { - RendererSceneRenderRD *scene_singleton = (RendererSceneRenderRD *)RendererSceneRenderRD::singleton; - - uniform_set_updated = true; - - if ((uint32_t)ubo_data.size() != shader_data->ubo_size) { - p_uniform_dirty = true; - if (uniform_buffer.is_valid()) { - RD::get_singleton()->free(uniform_buffer); - uniform_buffer = RID(); - } - - ubo_data.resize(shader_data->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()->free(uniform_set); - uniform_set = RID(); - } - } - - //check whether buffer changed - if (p_uniform_dirty && ubo_data.size()) { - update_uniform_buffer(shader_data->uniforms, shader_data->ubo_offsets.ptr(), p_parameters, ubo_data.ptrw(), ubo_data.size(), false); - RD::get_singleton()->buffer_update(uniform_buffer, 0, ubo_data.size(), ubo_data.ptrw()); - } - - uint32_t tex_uniform_count = shader_data->texture_uniforms.size(); - - if ((uint32_t)texture_cache.size() != tex_uniform_count) { - 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()->free(uniform_set); - uniform_set = RID(); - } - } - - if (p_textures_dirty && tex_uniform_count) { - update_textures(p_parameters, shader_data->default_texture_params, shader_data->texture_uniforms, texture_cache.ptrw(), true); - } - - if (shader_data->ubo_size == 0 && shader_data->texture_uniforms.size() == 0) { - // This material does not require an uniform set, so don't create it. - return; - } - - 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; - } - - Vector<RD::Uniform> uniforms; - - { - if (shader_data->ubo_size) { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.binding = 0; - u.ids.push_back(uniform_buffer); - uniforms.push_back(u); - } - - const RID *textures = texture_cache.ptrw(); - for (uint32_t i = 0; i < tex_uniform_count; i++) { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 1 + i; - u.ids.push_back(textures[i]); - uniforms.push_back(u); - } - } - - uniform_set = RD::get_singleton()->uniform_set_create(uniforms, scene_singleton->sky_shader.shader.version_get_shader(shader_data->version, 0), SKY_SET_MATERIAL); -} - -RendererSceneRenderRD::SkyMaterialData::~SkyMaterialData() { - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - } - - if (uniform_buffer.is_valid()) { - RD::get_singleton()->free(uniform_buffer); - } -} - -RendererStorageRD::MaterialData *RendererSceneRenderRD::_create_sky_material_func(SkyShaderData *p_shader) { - SkyMaterialData *material_data = memnew(SkyMaterialData); - material_data->shader_data = p_shader; - material_data->last_frame = false; - //update will happen later anyway so do nothing. - return material_data; -} - -RID RendererSceneRenderRD::environment_create() { - return environment_owner.make_rid(Environment()); +void RendererSceneRenderRD::environment_initialize(RID p_rid) { + environment_owner.initialize_rid(p_rid, RendererSceneEnvironmentRD()); } void RendererSceneRenderRD::environment_set_background(RID p_env, RS::EnvironmentBG p_bg) { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND(!env); env->background = p_bg; } void RendererSceneRenderRD::environment_set_sky(RID p_env, RID p_sky) { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND(!env); env->sky = p_sky; } void RendererSceneRenderRD::environment_set_sky_custom_fov(RID p_env, float p_scale) { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND(!env); env->sky_custom_fov = p_scale; } void RendererSceneRenderRD::environment_set_sky_orientation(RID p_env, const Basis &p_orientation) { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND(!env); env->sky_orientation = p_orientation; } void RendererSceneRenderRD::environment_set_bg_color(RID p_env, const Color &p_color) { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND(!env); env->bg_color = p_color; } void RendererSceneRenderRD::environment_set_bg_energy(RID p_env, float p_energy) { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND(!env); env->bg_energy = p_energy; } void RendererSceneRenderRD::environment_set_canvas_max_layer(RID p_env, int p_max_layer) { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND(!env); env->canvas_max_layer = p_max_layer; } void RendererSceneRenderRD::environment_set_ambient_light(RID p_env, const Color &p_color, RS::EnvironmentAmbientSource p_ambient, float p_energy, float p_sky_contribution, RS::EnvironmentReflectionSource p_reflection_source, const Color &p_ao_color) { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND(!env); - env->ambient_light = p_color; - env->ambient_source = p_ambient; - env->ambient_light_energy = p_energy; - env->ambient_sky_contribution = p_sky_contribution; - env->reflection_source = p_reflection_source; - env->ao_color = p_ao_color; + env->set_ambient_light(p_color, p_ambient, p_energy, p_sky_contribution, p_reflection_source, p_ao_color); } RS::EnvironmentBG RendererSceneRenderRD::environment_get_background(RID p_env) const { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, RS::ENV_BG_MAX); return env->background; } RID RendererSceneRenderRD::environment_get_sky(RID p_env) const { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, RID()); return env->sky; } float RendererSceneRenderRD::environment_get_sky_custom_fov(RID p_env) const { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, 0); return env->sky_custom_fov; } Basis RendererSceneRenderRD::environment_get_sky_orientation(RID p_env) const { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, Basis()); return env->sky_orientation; } Color RendererSceneRenderRD::environment_get_bg_color(RID p_env) const { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, Color()); return env->bg_color; } float RendererSceneRenderRD::environment_get_bg_energy(RID p_env) const { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, 0); return env->bg_energy; } int RendererSceneRenderRD::environment_get_canvas_max_layer(RID p_env) const { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, 0); return env->canvas_max_layer; } Color RendererSceneRenderRD::environment_get_ambient_light_color(RID p_env) const { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, Color()); return env->ambient_light; } RS::EnvironmentAmbientSource RendererSceneRenderRD::environment_get_ambient_source(RID p_env) const { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, RS::ENV_AMBIENT_SOURCE_BG); return env->ambient_source; } float RendererSceneRenderRD::environment_get_ambient_light_energy(RID p_env) const { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, 0); return env->ambient_light_energy; } float RendererSceneRenderRD::environment_get_ambient_sky_contribution(RID p_env) const { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, 0); return env->ambient_sky_contribution; } RS::EnvironmentReflectionSource RendererSceneRenderRD::environment_get_reflection_source(RID p_env) const { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, RS::ENV_REFLECTION_SOURCE_DISABLED); return env->reflection_source; } Color RendererSceneRenderRD::environment_get_ao_color(RID p_env) const { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, Color()); return env->ao_color; } void RendererSceneRenderRD::environment_set_tonemap(RID p_env, RS::EnvironmentToneMapper p_tone_mapper, float p_exposure, float p_white, bool p_auto_exposure, float p_min_luminance, float p_max_luminance, float p_auto_exp_speed, float p_auto_exp_scale) { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND(!env); - env->exposure = p_exposure; - env->tone_mapper = p_tone_mapper; - if (!env->auto_exposure && p_auto_exposure) { - env->auto_exposure_version = ++auto_exposure_counter; - } - env->auto_exposure = p_auto_exposure; - env->white = p_white; - env->min_luminance = p_min_luminance; - env->max_luminance = p_max_luminance; - env->auto_exp_speed = p_auto_exp_speed; - env->auto_exp_scale = p_auto_exp_scale; + env->set_tonemap(p_tone_mapper, p_exposure, p_white, p_auto_exposure, p_min_luminance, p_max_luminance, p_auto_exp_speed, p_auto_exp_scale); } void RendererSceneRenderRD::environment_set_glow(RID p_env, bool p_enable, Vector<float> p_levels, float p_intensity, float p_strength, float p_mix, float p_bloom_threshold, RS::EnvironmentGlowBlendMode p_blend_mode, float p_hdr_bleed_threshold, float p_hdr_bleed_scale, float p_hdr_luminance_cap) { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND(!env); - ERR_FAIL_COND_MSG(p_levels.size() != 7, "Size of array of glow levels must be 7"); - env->glow_enabled = p_enable; - env->glow_levels = p_levels; - env->glow_intensity = p_intensity; - env->glow_strength = p_strength; - env->glow_mix = p_mix; - env->glow_bloom = p_bloom_threshold; - env->glow_blend_mode = p_blend_mode; - env->glow_hdr_bleed_threshold = p_hdr_bleed_threshold; - env->glow_hdr_bleed_scale = p_hdr_bleed_scale; - env->glow_hdr_luminance_cap = p_hdr_luminance_cap; + env->set_glow(p_enable, p_levels, p_intensity, p_strength, p_mix, p_bloom_threshold, p_blend_mode, p_hdr_bleed_threshold, p_hdr_bleed_scale, p_hdr_luminance_cap); } void RendererSceneRenderRD::environment_glow_set_use_bicubic_upscale(bool p_enable) { @@ -2967,100 +313,77 @@ void RendererSceneRenderRD::environment_glow_set_use_high_quality(bool p_enable) glow_high_quality = p_enable; } -void RendererSceneRenderRD::environment_set_sdfgi(RID p_env, bool p_enable, RS::EnvironmentSDFGICascades p_cascades, float p_min_cell_size, RS::EnvironmentSDFGIYScale p_y_scale, bool p_use_occlusion, bool p_use_multibounce, bool p_read_sky, float p_energy, float p_normal_bias, float p_probe_bias) { - Environment *env = environment_owner.getornull(p_env); +void RendererSceneRenderRD::environment_set_sdfgi(RID p_env, bool p_enable, RS::EnvironmentSDFGICascades p_cascades, float p_min_cell_size, RS::EnvironmentSDFGIYScale p_y_scale, bool p_use_occlusion, float p_bounce_feedback, bool p_read_sky, float p_energy, float p_normal_bias, float p_probe_bias) { + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND(!env); if (low_end) { return; } - env->sdfgi_enabled = p_enable; - env->sdfgi_cascades = p_cascades; - env->sdfgi_min_cell_size = p_min_cell_size; - env->sdfgi_use_occlusion = p_use_occlusion; - env->sdfgi_use_multibounce = p_use_multibounce; - env->sdfgi_read_sky_light = p_read_sky; - env->sdfgi_energy = p_energy; - env->sdfgi_normal_bias = p_normal_bias; - env->sdfgi_probe_bias = p_probe_bias; - env->sdfgi_y_scale = p_y_scale; + env->set_sdfgi(p_enable, p_cascades, p_min_cell_size, p_y_scale, p_use_occlusion, p_bounce_feedback, p_read_sky, p_energy, p_normal_bias, p_probe_bias); } void RendererSceneRenderRD::environment_set_fog(RID p_env, bool p_enable, const Color &p_light_color, float p_light_energy, float p_sun_scatter, float p_density, float p_height, float p_height_density, float p_fog_aerial_perspective) { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND(!env); - env->fog_enabled = p_enable; - env->fog_light_color = p_light_color; - env->fog_light_energy = p_light_energy; - env->fog_sun_scatter = p_sun_scatter; - env->fog_density = p_density; - env->fog_height = p_height; - env->fog_height_density = p_height_density; - env->fog_aerial_perspective = p_fog_aerial_perspective; + env->set_fog(p_enable, p_light_color, p_light_energy, p_sun_scatter, p_density, p_height, p_height_density, p_fog_aerial_perspective); } bool RendererSceneRenderRD::environment_is_fog_enabled(RID p_env) const { - const Environment *env = environment_owner.getornull(p_env); + const RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, false); return env->fog_enabled; } Color RendererSceneRenderRD::environment_get_fog_light_color(RID p_env) const { - const Environment *env = environment_owner.getornull(p_env); + const RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, Color()); return env->fog_light_color; } float RendererSceneRenderRD::environment_get_fog_light_energy(RID p_env) const { - const Environment *env = environment_owner.getornull(p_env); + const RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, 0); return env->fog_light_energy; } float RendererSceneRenderRD::environment_get_fog_sun_scatter(RID p_env) const { - const Environment *env = environment_owner.getornull(p_env); + const RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, 0); return env->fog_sun_scatter; } float RendererSceneRenderRD::environment_get_fog_density(RID p_env) const { - const Environment *env = environment_owner.getornull(p_env); + const RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, 0); return env->fog_density; } float RendererSceneRenderRD::environment_get_fog_height(RID p_env) const { - const Environment *env = environment_owner.getornull(p_env); + const RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, 0); return env->fog_height; } float RendererSceneRenderRD::environment_get_fog_height_density(RID p_env) const { - const Environment *env = environment_owner.getornull(p_env); + const RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, 0); return env->fog_height_density; } float RendererSceneRenderRD::environment_get_fog_aerial_perspective(RID p_env) const { - const Environment *env = environment_owner.getornull(p_env); + const RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, 0); return env->fog_aerial_perspective; } -void RendererSceneRenderRD::environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, RenderingServer::EnvVolumetricFogShadowFilter p_shadow_filter) { - Environment *env = environment_owner.getornull(p_env); +void RendererSceneRenderRD::environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, bool p_temporal_reprojection, float p_temporal_reprojection_amount) { + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND(!env); if (low_end) { return; } - env->volumetric_fog_enabled = p_enable; - env->volumetric_fog_density = p_density; - env->volumetric_fog_light = p_light; - env->volumetric_fog_light_energy = p_light_energy; - env->volumetric_fog_length = p_length; - env->volumetric_fog_detail_spread = p_detail_spread; - env->volumetric_fog_shadow_filter = p_shadow_filter; - env->volumetric_fog_gi_inject = p_gi_inject; + env->set_volumetric_fog(p_enable, p_density, p_light, p_light_energy, p_length, p_detail_spread, p_gi_inject, p_temporal_reprojection, p_temporal_reprojection_amount); } void RendererSceneRenderRD::environment_set_volumetric_fog_volume_size(int p_size, int p_depth) { @@ -3071,47 +394,27 @@ void RendererSceneRenderRD::environment_set_volumetric_fog_volume_size(int p_siz void RendererSceneRenderRD::environment_set_volumetric_fog_filter_active(bool p_enable) { volumetric_fog_filter_active = p_enable; } -void RendererSceneRenderRD::environment_set_volumetric_fog_directional_shadow_shrink_size(int p_shrink_size) { - p_shrink_size = nearest_power_of_2_templated(p_shrink_size); - if (volumetric_fog_directional_shadow_shrink == (uint32_t)p_shrink_size) { - return; - } - - _clear_shadow_shrink_stages(directional_shadow.shrink_stages); -} -void RendererSceneRenderRD::environment_set_volumetric_fog_positional_shadow_shrink_size(int p_shrink_size) { - p_shrink_size = nearest_power_of_2_templated(p_shrink_size); - if (volumetric_fog_positional_shadow_shrink == (uint32_t)p_shrink_size) { - return; - } - - for (uint32_t i = 0; i < shadow_atlas_owner.get_rid_count(); i++) { - ShadowAtlas *sa = shadow_atlas_owner.get_ptr_by_index(i); - _clear_shadow_shrink_stages(sa->shrink_stages); - } -} void RendererSceneRenderRD::environment_set_sdfgi_ray_count(RS::EnvironmentSDFGIRayCount p_ray_count) { - sdfgi_ray_count = p_ray_count; + gi.sdfgi_ray_count = p_ray_count; } void RendererSceneRenderRD::environment_set_sdfgi_frames_to_converge(RS::EnvironmentSDFGIFramesToConverge p_frames) { - sdfgi_frames_to_converge = 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; } void RendererSceneRenderRD::environment_set_ssr(RID p_env, bool p_enable, int p_max_steps, float p_fade_int, float p_fade_out, float p_depth_tolerance) { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND(!env); if (low_end) { return; } - env->ssr_enabled = p_enable; - env->ssr_max_steps = p_max_steps; - env->ssr_fade_in = p_fade_int; - env->ssr_fade_out = p_fade_out; - env->ssr_depth_tolerance = p_depth_tolerance; + env->set_ssr(p_enable, p_max_steps, p_fade_int, p_fade_out, p_depth_tolerance); } void RendererSceneRenderRD::environment_set_ssr_roughness_quality(RS::EnvironmentSSRRoughnessQuality p_quality) { @@ -3123,22 +426,14 @@ RS::EnvironmentSSRRoughnessQuality RendererSceneRenderRD::environment_get_ssr_ro } void RendererSceneRenderRD::environment_set_ssao(RID p_env, bool p_enable, float p_radius, float p_intensity, float p_power, float p_detail, float p_horizon, float p_sharpness, float p_light_affect, float p_ao_channel_affect) { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND(!env); if (low_end) { return; } - env->ssao_enabled = p_enable; - env->ssao_radius = p_radius; - env->ssao_intensity = p_intensity; - env->ssao_power = p_power; - env->ssao_detail = p_detail; - env->ssao_horizon = p_horizon; - env->ssao_sharpness = p_sharpness; - env->ssao_direct_light_affect = p_light_affect; - env->ssao_ao_channel_affect = p_ao_channel_affect; + env->set_ssao(p_enable, p_radius, p_intensity, p_power, p_detail, p_horizon, p_sharpness, p_light_affect, p_ao_channel_affect); } void RendererSceneRenderRD::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) { @@ -3151,30 +446,30 @@ void RendererSceneRenderRD::environment_set_ssao_quality(RS::EnvironmentSSAOQual } bool RendererSceneRenderRD::environment_is_ssao_enabled(RID p_env) const { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, false); return env->ssao_enabled; } float RendererSceneRenderRD::environment_get_ssao_ao_affect(RID p_env) const { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, 0.0); return env->ssao_ao_channel_affect; } float RendererSceneRenderRD::environment_get_ssao_light_affect(RID p_env) const { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, 0.0); return env->ssao_direct_light_affect; } bool RendererSceneRenderRD::environment_is_ssr_enabled(RID p_env) const { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, false); return env->ssr_enabled; } bool RendererSceneRenderRD::environment_is_sdfgi_enabled(RID p_env) const { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, false); return env->sdfgi_enabled; } @@ -3184,7 +479,7 @@ bool RendererSceneRenderRD::is_environment(RID p_env) const { } Ref<Image> RendererSceneRenderRD::environment_bake_panorama(RID p_env, bool p_bake_irradiance, const Size2i &p_size) { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND_V(!env, Ref<Image>()); if (env->background == RS::ENV_BG_CAMERA_FEED || env->background == RS::ENV_BG_CANVAS || env->background == RS::ENV_BG_KEEP) { @@ -3224,8 +519,12 @@ Ref<Image> RendererSceneRenderRD::environment_bake_panorama(RID p_env, bool p_ba RID RendererSceneRenderRD::reflection_atlas_create() { ReflectionAtlas ra; - ra.count = GLOBAL_GET("rendering/quality/reflection_atlas/reflection_count"); - ra.size = GLOBAL_GET("rendering/quality/reflection_atlas/reflection_size"); + ra.count = GLOBAL_GET("rendering/reflections/reflection_atlas/reflection_count"); + ra.size = GLOBAL_GET("rendering/reflections/reflection_atlas/reflection_size"); + + ra.cluster_builder = memnew(ClusterBuilderRD); + ra.cluster_builder->set_shared(&cluster_builder_shared); + ra.cluster_builder->setup(Size2i(ra.size, ra.size), max_cluster_elements, RID(), RID(), RID()); return reflection_atlas_owner.make_rid(ra); } @@ -3238,6 +537,8 @@ void RendererSceneRenderRD::reflection_atlas_set_size(RID p_ref_atlas, int p_ref return; //no changes } + 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; @@ -3247,9 +548,8 @@ void RendererSceneRenderRD::reflection_atlas_set_size(RID p_ref_atlas, int p_ref ra->reflection = RID(); RD::get_singleton()->free(ra->depth_buffer); ra->depth_buffer = RID(); - for (int i = 0; i < ra->reflections.size(); i++) { - _clear_reflection_data(ra->reflections.write[i].data); + ra->reflections.write[i].data.clear_reflection_data(); if (ra->reflections[i].owner.is_null()) { continue; } @@ -3353,7 +653,7 @@ bool RendererSceneRenderRD::reflection_probe_instance_begin_render(RID p_instanc } if (atlas->reflection.is_null()) { - int mipmaps = MIN(roughness_layers, Image::get_image_required_mipmaps(atlas->size, atlas->size, Image::FORMAT_RGBAH) + 1); + int mipmaps = MIN(sky.roughness_layers, Image::get_image_required_mipmaps(atlas->size, atlas->size, Image::FORMAT_RGBAH) + 1); mipmaps = storage->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: @@ -3378,7 +678,7 @@ bool RendererSceneRenderRD::reflection_probe_instance_begin_render(RID p_instanc } atlas->reflections.resize(atlas->count); for (int i = 0; i < atlas->count; i++) { - _update_reflection_data(atlas->reflections.write[i].data, atlas->size, mipmaps, false, atlas->reflection, i * 6, storage->reflection_probe_get_update_mode(rpi->probe) == RS::REFLECTION_PROBE_UPDATE_ALWAYS); + atlas->reflections.write[i].data.update_reflection_data(atlas->size, mipmaps, false, atlas->reflection, i * 6, storage->reflection_probe_get_update_mode(rpi->probe) == RS::REFLECTION_PROBE_UPDATE_ALWAYS, sky.roughness_layers); for (int j = 0; j < 6; j++) { Vector<RID> fb; fb.push_back(atlas->reflections.write[i].data.layers[0].mipmaps[0].views[j]); @@ -3438,7 +738,7 @@ bool RendererSceneRenderRD::reflection_probe_instance_postprocess_step(RID p_ins if (storage->reflection_probe_get_update_mode(rpi->probe) == RS::REFLECTION_PROBE_UPDATE_ALWAYS) { // Using real time reflections, all roughness is done in one step - _create_reflection_fast_filter(atlas->reflections.write[rpi->atlas_index].data, false); + atlas->reflections.write[rpi->atlas_index].data.create_reflection_fast_filter(storage, false); rpi->rendering = false; rpi->processing_side = 0; rpi->processing_layer = 1; @@ -3446,7 +746,7 @@ bool RendererSceneRenderRD::reflection_probe_instance_postprocess_step(RID p_ins } if (rpi->processing_layer > 1) { - _create_reflection_importance_sample(atlas->reflections.write[rpi->atlas_index].data, false, 10, rpi->processing_layer); + atlas->reflections.write[rpi->atlas_index].data.create_reflection_importance_sample(storage, false, 10, rpi->processing_layer, 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; @@ -3457,7 +757,7 @@ bool RendererSceneRenderRD::reflection_probe_instance_postprocess_step(RID p_ins return false; } else { - _create_reflection_importance_sample(atlas->reflections.write[rpi->atlas_index].data, false, rpi->processing_side, rpi->processing_layer); + atlas->reflections.write[rpi->atlas_index].data.create_reflection_importance_sample(storage, false, rpi->processing_side, rpi->processing_layer, sky.sky_ggx_samples_quality); } rpi->processing_side++; @@ -3504,13 +804,28 @@ RID RendererSceneRenderRD::shadow_atlas_create() { return shadow_atlas_owner.make_rid(ShadowAtlas()); } -void RendererSceneRenderRD::shadow_atlas_set_size(RID p_atlas, int p_size) { +void RendererSceneRenderRD::_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 RendererSceneRenderRD::shadow_atlas_set_size(RID p_atlas, int p_size, bool p_16_bits) { ShadowAtlas *shadow_atlas = shadow_atlas_owner.getornull(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) { + if (p_size == shadow_atlas->size && p_16_bits == shadow_atlas->use_16_bits) { return; } @@ -3518,7 +833,6 @@ void RendererSceneRenderRD::shadow_atlas_set_size(RID p_atlas, int p_size) { if (shadow_atlas->depth.is_valid()) { RD::get_singleton()->free(shadow_atlas->depth); shadow_atlas->depth = RID(); - _clear_shadow_shrink_stages(shadow_atlas->shrink_stages); } for (int i = 0; i < 4; i++) { //clear subdivisions @@ -3537,16 +851,7 @@ void RendererSceneRenderRD::shadow_atlas_set_size(RID p_atlas, int p_size) { shadow_atlas->shadow_owners.clear(); shadow_atlas->size = p_size; - - if (shadow_atlas->size) { - RD::TextureFormat tf; - tf.format = RD::DATA_FORMAT_R32_SFLOAT; - tf.width = shadow_atlas->size; - tf.height = shadow_atlas->size; - tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; - - shadow_atlas->depth = RD::get_singleton()->texture_create(tf, RD::TextureView()); - } + shadow_atlas->use_16_bits = p_size; } void RendererSceneRenderRD::shadow_atlas_set_quadrant_subdivision(RID p_atlas, int p_quadrant, int p_subdivision) { @@ -3801,10 +1106,24 @@ bool RendererSceneRenderRD::shadow_atlas_update_light(RID p_atlas, RID p_light_i return false; } -void RendererSceneRenderRD::directional_shadow_atlas_set_size(int p_size) { +void RendererSceneRenderRD::_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 RendererSceneRenderRD::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) { + if (directional_shadow.size == p_size && directional_shadow.use_16_bits == p_16_bits) { return; } @@ -3812,21 +1131,9 @@ void RendererSceneRenderRD::directional_shadow_atlas_set_size(int p_size) { if (directional_shadow.depth.is_valid()) { RD::get_singleton()->free(directional_shadow.depth); - _clear_shadow_shrink_stages(directional_shadow.shrink_stages); directional_shadow.depth = RID(); + _base_uniforms_changed(); } - - if (p_size > 0) { - RD::TextureFormat tf; - tf.format = RD::DATA_FORMAT_R32_SFLOAT; - tf.width = p_size; - tf.height = p_size; - tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; - - directional_shadow.depth = RD::get_singleton()->texture_create(tf, RD::TextureView()); - } - - _base_uniforms_changed(); } void RendererSceneRenderRD::set_directional_shadow_count(int p_count) { @@ -3880,8 +1187,11 @@ int RendererSceneRenderRD::get_directional_light_shadow_size(RID p_light_intance ////////////////////////////////////////////////// -RID RendererSceneRenderRD::camera_effects_create() { - return camera_effects_owner.make_rid(CameraEffects()); +RID RendererSceneRenderRD::camera_effects_allocate() { + return camera_effects_owner.allocate_rid(); +} +void RendererSceneRenderRD::camera_effects_initialize(RID p_rid) { + camera_effects_owner.initialize_rid(p_rid, CameraEffects()); } void RendererSceneRenderRD::camera_effects_set_dof_blur_quality(RS::DOFBlurQuality p_quality, bool p_use_jitter) { @@ -3946,11 +1256,7 @@ void RendererSceneRenderRD::light_instance_set_shadow_transform(RID p_light_inst LightInstance *light_instance = light_instance_owner.getornull(p_light_instance); ERR_FAIL_COND(!light_instance); - if (storage->light_get_type(light_instance->light) != RS::LIGHT_DIRECTIONAL) { - p_pass = 0; - } - - ERR_FAIL_INDEX(p_pass, 4); + ERR_FAIL_INDEX(p_pass, 6); light_instance->shadow_transform[p_pass].camera = p_projection; light_instance->shadow_transform[p_pass].transform = p_transform; @@ -3996,29 +1302,6 @@ RendererSceneRenderRD::ShadowCubemap *RendererSceneRenderRD::_get_shadow_cubemap return &shadow_cubemaps[p_size]; } -RendererSceneRenderRD::ShadowMap *RendererSceneRenderRD::_get_shadow_map(const Size2i &p_size) { - if (!shadow_maps.has(p_size)) { - ShadowMap sm; - { - 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.width; - tf.height = p_size.height; - tf.usage_bits = RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; - - sm.depth = RD::get_singleton()->texture_create(tf, RD::TextureView()); - } - - Vector<RID> fbtex; - fbtex.push_back(sm.depth); - sm.fb = RD::get_singleton()->framebuffer_create(fbtex); - - shadow_maps[p_size] = sm; - } - - return &shadow_maps[p_size]; -} - ////////////////////////// RID RendererSceneRenderRD::decal_instance_create(RID p_decal) { @@ -4049,805 +1332,27 @@ void RendererSceneRenderRD::lightmap_instance_set_transform(RID p_lightmap, cons ///////////////////////////////// RID RendererSceneRenderRD::gi_probe_instance_create(RID p_base) { - GIProbeInstance gi_probe; - gi_probe.probe = p_base; - RID rid = gi_probe_instance_owner.make_rid(gi_probe); - return rid; + return gi.gi_probe_instance_create(p_base); } void RendererSceneRenderRD::gi_probe_instance_set_transform_to_data(RID p_probe, const Transform &p_xform) { - GIProbeInstance *gi_probe = gi_probe_instance_owner.getornull(p_probe); - ERR_FAIL_COND(!gi_probe); - - gi_probe->transform = p_xform; + gi.gi_probe_instance_set_transform_to_data(p_probe, p_xform); } bool RendererSceneRenderRD::gi_probe_needs_update(RID p_probe) const { - GIProbeInstance *gi_probe = gi_probe_instance_owner.getornull(p_probe); - ERR_FAIL_COND_V(!gi_probe, false); - if (low_end) { return false; } - //return true; - return gi_probe->last_probe_version != storage->gi_probe_get_version(gi_probe->probe); + return gi.gi_probe_needs_update(p_probe); } void RendererSceneRenderRD::gi_probe_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<GeometryInstance *> &p_dynamic_objects) { - GIProbeInstance *gi_probe = gi_probe_instance_owner.getornull(p_probe); - ERR_FAIL_COND(!gi_probe); - if (low_end) { return; } - uint32_t data_version = storage->gi_probe_get_data_version(gi_probe->probe); - - // (RE)CREATE IF NEEDED - - if (gi_probe->last_probe_data_version != data_version) { - //need to re-create everything - if (gi_probe->texture.is_valid()) { - RD::get_singleton()->free(gi_probe->texture); - RD::get_singleton()->free(gi_probe->write_buffer); - gi_probe->mipmaps.clear(); - } - - for (int i = 0; i < gi_probe->dynamic_maps.size(); i++) { - RD::get_singleton()->free(gi_probe->dynamic_maps[i].texture); - RD::get_singleton()->free(gi_probe->dynamic_maps[i].depth); - } - - gi_probe->dynamic_maps.clear(); - - Vector3i octree_size = storage->gi_probe_get_octree_size(gi_probe->probe); - - if (octree_size != Vector3i()) { - //can create a 3D texture - Vector<int> levels = storage->gi_probe_get_level_counts(gi_probe->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; - - gi_probe->texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); - - RD::get_singleton()->texture_clear(gi_probe->texture, Color(0, 0, 0, 0), 0, levels.size(), 0, 1, false); - - { - int total_elements = 0; - for (int i = 0; i < levels.size(); i++) { - total_elements += levels[i]; - } - - gi_probe->write_buffer = RD::get_singleton()->storage_buffer_create(total_elements * 16); - } - - for (int i = 0; i < levels.size(); i++) { - GIProbeInstance::Mipmap mipmap; - mipmap.texture = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), gi_probe->texture, 0, i, 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.ids.push_back(storage->gi_probe_get_octree_buffer(gi_probe->probe)); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.binding = 2; - u.ids.push_back(storage->gi_probe_get_data_buffer(gi_probe->probe)); - uniforms.push_back(u); - } - - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.binding = 4; - u.ids.push_back(gi_probe->write_buffer); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 9; - u.ids.push_back(storage->gi_probe_get_sdf_texture(gi_probe->probe)); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; - u.binding = 10; - u.ids.push_back(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.ids.push_back(gi_probe_lights_uniform); - copy_uniforms.push_back(u); - } - - mipmap.uniform_set = RD::get_singleton()->uniform_set_create(copy_uniforms, giprobe_lighting_shader_version_shaders[GI_PROBE_SHADER_VERSION_COMPUTE_LIGHT], 0); - - copy_uniforms = uniforms; //restore - - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 5; - u.ids.push_back(gi_probe->texture); - copy_uniforms.push_back(u); - } - mipmap.second_bounce_uniform_set = RD::get_singleton()->uniform_set_create(copy_uniforms, giprobe_lighting_shader_version_shaders[GI_PROBE_SHADER_VERSION_COMPUTE_SECOND_BOUNCE], 0); - } else { - mipmap.uniform_set = RD::get_singleton()->uniform_set_create(copy_uniforms, giprobe_lighting_shader_version_shaders[GI_PROBE_SHADER_VERSION_COMPUTE_MIPMAP], 0); - } - } - - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 5; - u.ids.push_back(mipmap.texture); - uniforms.push_back(u); - } - - mipmap.write_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, giprobe_lighting_shader_version_shaders[GI_PROBE_SHADER_VERSION_WRITE_TEXTURE], 0); - - gi_probe->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 < gi_probe->mipmaps.size()) { - GIProbeInstance::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 (gi_probe->dynamic_maps.size() == 0) { - dtf.usage_bits |= RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; - } - dmap.texture = RD::get_singleton()->texture_create(dtf, RD::TextureView()); - - if (gi_probe->dynamic_maps.size() == 0) { - //render depth for first one - dtf.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; - dtf.usage_bits = RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT; - dmap.fb_depth = RD::get_singleton()->texture_create(dtf, RD::TextureView()); - } - - //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()); - - if (gi_probe->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()); - dmap.normal = RD::get_singleton()->texture_create(dtf, RD::TextureView()); - dmap.orm = RD::get_singleton()->texture_create(dtf, RD::TextureView()); - - 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.ids.push_back(gi_probe_lights_uniform); - uniforms.push_back(u); - } - - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 5; - u.ids.push_back(dmap.albedo); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 6; - u.ids.push_back(dmap.normal); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 7; - u.ids.push_back(dmap.orm); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 8; - u.ids.push_back(dmap.fb_depth); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 9; - u.ids.push_back(storage->gi_probe_get_sdf_texture(gi_probe->probe)); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; - u.binding = 10; - u.ids.push_back(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.ids.push_back(dmap.texture); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 12; - u.ids.push_back(dmap.depth); - uniforms.push_back(u); - } - - dmap.uniform_set = RD::get_singleton()->uniform_set_create(uniforms, giprobe_lighting_shader_version_shaders[GI_PROBE_SHADER_VERSION_DYNAMIC_OBJECT_LIGHTING], 0); - } - } else { - bool plot = dmap.mipmap >= 0; - bool write = dmap.mipmap < (gi_probe->mipmaps.size() - 1); - - Vector<RD::Uniform> uniforms; - - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 5; - u.ids.push_back(gi_probe->dynamic_maps[gi_probe->dynamic_maps.size() - 1].texture); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 6; - u.ids.push_back(gi_probe->dynamic_maps[gi_probe->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.ids.push_back(dmap.texture); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 8; - u.ids.push_back(dmap.depth); - uniforms.push_back(u); - } - } - - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 9; - u.ids.push_back(storage->gi_probe_get_sdf_texture(gi_probe->probe)); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; - u.binding = 10; - u.ids.push_back(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.ids.push_back(gi_probe->mipmaps[dmap.mipmap].texture); - uniforms.push_back(u); - } - } - - dmap.uniform_set = RD::get_singleton()->uniform_set_create(uniforms, giprobe_lighting_shader_version_shaders[(write && plot) ? GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE_PLOT : write ? GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE : GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_PLOT], 0); - } - - gi_probe->dynamic_maps.push_back(dmap); - } - } - } - - gi_probe->last_probe_data_version = data_version; - p_update_light_instances = true; //just in case - - _base_uniforms_changed(); - } - - // UDPDATE TIME - - if (gi_probe->has_dynamic_object_data) { - //if it has dynamic object data, it needs to be cleared - RD::get_singleton()->texture_clear(gi_probe->texture, Color(0, 0, 0, 0), 0, gi_probe->mipmaps.size(), 0, 1, true); - } - - uint32_t light_count = 0; - - if (p_update_light_instances || p_dynamic_objects.size() > 0) { - light_count = MIN(gi_probe_max_lights, (uint32_t)p_light_instances.size()); - - { - Transform to_cell = storage->gi_probe_get_to_cell_xform(gi_probe->probe); - Transform to_probe_xform = (gi_probe->transform * to_cell.affine_inverse()).affine_inverse(); - //update lights - - for (uint32_t i = 0; i < light_count; i++) { - GIProbeLight &l = gi_probe_lights[i]; - RID light_instance = p_light_instances[i]; - RID light = light_instance_get_base_light(light_instance); - - l.type = storage->light_get_type(light); - if (l.type == RS::LIGHT_DIRECTIONAL && storage->light_directional_is_sky_only(light)) { - light_count--; - continue; - } - - l.attenuation = storage->light_get_param(light, RS::LIGHT_PARAM_ATTENUATION); - l.energy = storage->light_get_param(light, RS::LIGHT_PARAM_ENERGY) * storage->light_get_param(light, RS::LIGHT_PARAM_INDIRECT_ENERGY); - l.radius = to_cell.basis.xform(Vector3(storage->light_get_param(light, RS::LIGHT_PARAM_RANGE), 0, 0)).length(); - Color color = storage->light_get_color(light).to_linear(); - l.color[0] = color.r; - l.color[1] = color.g; - l.color[2] = color.b; - - l.spot_angle_radians = Math::deg2rad(storage->light_get_param(light, RS::LIGHT_PARAM_SPOT_ANGLE)); - l.spot_attenuation = storage->light_get_param(light, RS::LIGHT_PARAM_SPOT_ATTENUATION); - - Transform xform = 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_axis(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 = storage->light_has_shadow(light); - } - - RD::get_singleton()->buffer_update(gi_probe_lights_uniform, 0, sizeof(GIProbeLight) * light_count, gi_probe_lights, true); - } - } - - if (gi_probe->has_dynamic_object_data || p_update_light_instances || p_dynamic_objects.size()) { - // PROCESS MIPMAPS - if (gi_probe->mipmaps.size()) { - //can update mipmaps - - Vector3i probe_size = storage->gi_probe_get_octree_size(gi_probe->probe); - - GIProbePushConstant 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 = gi_probe->mipmaps.size(); - push_constant.emission_scale = 1.0; - push_constant.propagation = storage->gi_probe_get_propagation(gi_probe->probe); - push_constant.dynamic_range = storage->gi_probe_get_dynamic_range(gi_probe->probe); - push_constant.light_count = light_count; - push_constant.aniso_strength = 0; - - /* print_line("probe update to version " + itos(gi_probe->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 = storage->gi_probe_is_using_two_bounces(gi_probe->probe) ? 2 : 1; - } else { - passes = 1; //only re-blitting is necessary - } - int wg_size = 64; - int wg_limit_x = 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 < gi_probe->mipmaps.size(); i++) { - if (i == 0) { - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, giprobe_lighting_shader_version_pipelines[pass == 0 ? GI_PROBE_SHADER_VERSION_COMPUTE_LIGHT : GI_PROBE_SHADER_VERSION_COMPUTE_SECOND_BOUNCE]); - } else if (i == 1) { - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, giprobe_lighting_shader_version_pipelines[GI_PROBE_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, gi_probe->mipmaps[i].uniform_set, 0); - } else { - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, gi_probe->mipmaps[i].second_bounce_uniform_set, 0); - } - - push_constant.cell_offset = gi_probe->mipmaps[i].cell_offset; - push_constant.cell_count = gi_probe->mipmaps[i].cell_count; - - int wg_todo = (gi_probe->mipmaps[i].cell_count - 1) / wg_size + 1; - while (wg_todo) { - int wg_count = MIN(wg_todo, wg_limit_x); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(GIProbePushConstant)); - 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, giprobe_lighting_shader_version_pipelines[GI_PROBE_SHADER_VERSION_WRITE_TEXTURE]); - - for (int i = 0; i < gi_probe->mipmaps.size(); i++) { - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, gi_probe->mipmaps[i].write_uniform_set, 0); - - push_constant.cell_offset = gi_probe->mipmaps[i].cell_offset; - push_constant.cell_count = gi_probe->mipmaps[i].cell_count; - - int wg_todo = (gi_probe->mipmaps[i].cell_count - 1) / wg_size + 1; - while (wg_todo) { - int wg_count = MIN(wg_todo, wg_limit_x); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(GIProbePushConstant)); - 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(); - } - } - - gi_probe->has_dynamic_object_data = false; //clear until dynamic object data is used again - - if (p_dynamic_objects.size() && gi_probe->dynamic_maps.size()) { - Vector3i octree_size = storage->gi_probe_get_octree_size(gi_probe->probe); - int multiplier = gi_probe->dynamic_maps[0].size / MAX(MAX(octree_size.x, octree_size.y), octree_size.z); - - Transform oversample_scale; - oversample_scale.basis.scale(Vector3(multiplier, multiplier, multiplier)); - - Transform to_cell = oversample_scale * storage->gi_probe_get_to_cell_xform(gi_probe->probe); - Transform to_world_xform = gi_probe->transform * to_cell.affine_inverse(); - Transform 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++) { - GeometryInstance *instance = p_dynamic_objects[i]; - - //transform aabb to giprobe - AABB aabb = (to_probe_xform * geometry_instance_get_transform(instance)).xform(geometry_instance_get_aabb(instance)); - - //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.position + aabb.size * 0.5; - Transform xform; - xform.set_look_at(center - aabb.size * 0.5 * render_dir, center, up_dir); - - Vector3 x_dir = xform.basis.get_axis(0).abs(); - int x_axis = int(Vector3(0, 1, 2).dot(x_dir)); - Vector3 y_dir = xform.basis.get_axis(1).abs(); - int y_axis = int(Vector3(0, 1, 2).dot(y_dir)); - Vector3 z_dir = -xform.basis.get_axis(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_axis(0)) < 0); - bool y_flip = bool(Vector3(1, 1, 1).dot(xform.basis.get_axis(1)) < 0); - bool z_flip = bool(Vector3(1, 1, 1).dot(xform.basis.get_axis(2)) > 0); - - CameraMatrix 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 (cull_argument.size() == 0) { - cull_argument.push_back(nullptr); - } - cull_argument[0] = instance; - - _render_material(to_world_xform * xform, cm, true, cull_argument, gi_probe->dynamic_maps[0].fb, Rect2i(Vector2i(), rect.size)); - - GIProbeDynamicPushConstant push_constant; - zeromem(&push_constant, sizeof(GIProbeDynamicPushConstant)); - 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 = storage->gi_probe_get_dynamic_range(gi_probe->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 = storage->gi_probe_get_propagation(gi_probe->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, giprobe_lighting_shader_version_pipelines[GI_PROBE_SHADER_VERSION_DYNAMIC_OBJECT_LIGHTING]); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, gi_probe->dynamic_maps[0].uniform_set, 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(GIProbeDynamicPushConstant)); - 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 < gi_probe->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 (gi_probe->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 = gi_probe->dynamic_maps[k].mipmap > 0; - - RD::get_singleton()->compute_list_add_barrier(compute_list); - - if (gi_probe->dynamic_maps[k].mipmap < 0) { - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, giprobe_lighting_shader_version_pipelines[GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE]); - } else if (k < gi_probe->dynamic_maps.size() - 1) { - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, giprobe_lighting_shader_version_pipelines[GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE_PLOT]); - } else { - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, giprobe_lighting_shader_version_pipelines[GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_PLOT]); - } - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, gi_probe->dynamic_maps[k].uniform_set, 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(GIProbeDynamicPushConstant)); - 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(); - } - } - - gi_probe->has_dynamic_object_data = true; //clear until dynamic object data is used again - } - - gi_probe->last_probe_version = storage->gi_probe_get_version(gi_probe->probe); -} - -void RendererSceneRenderRD::_debug_giprobe(RID p_gi_probe, RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform, bool p_lighting, bool p_emission, float p_alpha) { - GIProbeInstance *gi_probe = gi_probe_instance_owner.getornull(p_gi_probe); - ERR_FAIL_COND(!gi_probe); - - if (gi_probe->mipmaps.size() == 0) { - return; - } - - CameraMatrix transform = (p_camera_with_transform * CameraMatrix(gi_probe->transform)) * CameraMatrix(storage->gi_probe_get_to_cell_xform(gi_probe->probe).affine_inverse()); - - int level = 0; - Vector3i octree_size = storage->gi_probe_get_octree_size(gi_probe->probe); - - GIProbeDebugPushConstant push_constant; - push_constant.alpha = p_alpha; - push_constant.dynamic_range = storage->gi_probe_get_dynamic_range(gi_probe->probe); - push_constant.cell_offset = gi_probe->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] = transform.matrix[i][j]; - } - } - - if (giprobe_debug_uniform_set.is_valid()) { - RD::get_singleton()->free(giprobe_debug_uniform_set); - } - Vector<RD::Uniform> uniforms; - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.binding = 1; - u.ids.push_back(storage->gi_probe_get_data_buffer(gi_probe->probe)); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 2; - u.ids.push_back(gi_probe->texture); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; - u.binding = 3; - u.ids.push_back(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 && gi_probe->has_dynamic_object_data) { - cell_count = push_constant.bounds[0] * push_constant.bounds[1] * push_constant.bounds[2]; - } else { - cell_count = gi_probe->mipmaps[level].cell_count; - } - - giprobe_debug_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, giprobe_debug_shader_version_shaders[0], 0); - RD::get_singleton()->draw_list_bind_render_pipeline(p_draw_list, giprobe_debug_shader_version_pipelines[p_emission ? GI_PROBE_DEBUG_EMISSION : p_lighting ? (gi_probe->has_dynamic_object_data ? GI_PROBE_DEBUG_LIGHT_FULL : GI_PROBE_DEBUG_LIGHT) : GI_PROBE_DEBUG_COLOR].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, giprobe_debug_uniform_set, 0); - RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(GIProbeDebugPushConstant)); - RD::get_singleton()->draw_list_draw(p_draw_list, false, cell_count, 36); + gi.gi_probe_update(p_probe, p_update_light_instances, p_light_instances, p_dynamic_objects, this); } void RendererSceneRenderRD::_debug_sdfgi_probes(RID p_render_buffers, RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform) { @@ -4858,132 +1363,7 @@ void RendererSceneRenderRD::_debug_sdfgi_probes(RID p_render_buffers, RD::DrawLi return; //nothing to debug } - SDGIShader::DebugProbesPushConstant push_constant; - - for (int i = 0; i < 4; i++) { - for (int j = 0; j < 4; j++) { - push_constant.projection[i * 4 + j] = p_camera_with_transform.matrix[i][j]; - } - } - - //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_PI * 2.0) / float(push_constant.sections_in_band); - push_constant.y_mult = rb->sdfgi->y_mult; - - uint32_t total_points = push_constant.sections_in_band * band_points; - uint32_t total_probes = rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count; - - push_constant.grid_size[0] = rb->sdfgi->cascade_size; - push_constant.grid_size[1] = rb->sdfgi->cascade_size; - push_constant.grid_size[2] = rb->sdfgi->cascade_size; - push_constant.cascade = 0; - - push_constant.probe_axis_size = rb->sdfgi->probe_axis_count; - - if (!rb->sdfgi->debug_probes_uniform_set.is_valid() || !RD::get_singleton()->uniform_set_is_valid(rb->sdfgi->debug_probes_uniform_set)) { - Vector<RD::Uniform> uniforms; - { - RD::Uniform u; - u.binding = 1; - u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.ids.push_back(rb->sdfgi->cascades_ubo); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.binding = 2; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.ids.push_back(rb->sdfgi->lightprobe_texture); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.binding = 3; - u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; - u.ids.push_back(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.ids.push_back(rb->sdfgi->occlusion_texture); - uniforms.push_back(u); - } - - rb->sdfgi->debug_probes_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.debug_probes.version_get_shader(sdfgi_shader.debug_probes_shader, 0), 0); - } - - RD::get_singleton()->draw_list_bind_render_pipeline(p_draw_list, sdfgi_shader.debug_probes_pipeline[SDGIShader::PROBE_DEBUG_PROBES].get_render_pipeline(RD::INVALID_FORMAT_ID, RD::get_singleton()->framebuffer_get_format(p_framebuffer))); - RD::get_singleton()->draw_list_bind_uniform_set(p_draw_list, rb->sdfgi->debug_probes_uniform_set, 0); - RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(SDGIShader::DebugProbesPushConstant)); - RD::get_singleton()->draw_list_draw(p_draw_list, false, total_probes, total_points); - - if (sdfgi_debug_probe_dir != Vector3()) { - print_line("CLICK DEBUG ME?"); - uint32_t cascade = 0; - Vector3 offset = Vector3((Vector3i(1, 1, 1) * -int32_t(rb->sdfgi->cascade_size >> 1) + rb->sdfgi->cascades[cascade].position)) * rb->sdfgi->cascades[cascade].cell_size * Vector3(1.0, 1.0 / rb->sdfgi->y_mult, 1.0); - Vector3 probe_size = rb->sdfgi->cascades[cascade].cell_size * (rb->sdfgi->cascade_size / SDFGI::PROBE_DIVISOR) * Vector3(1.0, 1.0 / rb->sdfgi->y_mult, 1.0); - Vector3 ray_from = sdfgi_debug_probe_pos; - Vector3 ray_to = sdfgi_debug_probe_pos + sdfgi_debug_probe_dir * rb->sdfgi->cascades[cascade].cell_size * Math::sqrt(3.0) * rb->sdfgi->cascade_size; - float sphere_radius = 0.2; - float closest_dist = 1e20; - sdfgi_debug_probe_enabled = false; - - Vector3i probe_from = rb->sdfgi->cascades[cascade].position / (rb->sdfgi->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; - sdfgi_debug_probe_enabled = true; - sdfgi_debug_probe_index = probe_from + Vector3i(i, j, k); - } - } - } - } - } - - if (sdfgi_debug_probe_enabled) { - print_line("found: " + sdfgi_debug_probe_index); - } else { - print_line("no found"); - } - sdfgi_debug_probe_dir = Vector3(); - } - - if (sdfgi_debug_probe_enabled) { - uint32_t cascade = 0; - uint32_t probe_cells = (rb->sdfgi->cascade_size / SDFGI::PROBE_DIVISOR); - Vector3i probe_from = rb->sdfgi->cascades[cascade].position / probe_cells; - Vector3i ofs = 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(p_draw_list, sdfgi_shader.debug_probes_pipeline[SDGIShader::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(p_draw_list, rb->sdfgi->debug_probes_uniform_set, 0); - RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(SDGIShader::DebugProbesPushConstant)); - RD::get_singleton()->draw_list_draw(p_draw_list, false, cell_count, total_points); - } + rb->sdfgi->debug_probes(p_draw_list, p_framebuffer, p_camera_with_transform); } //////////////////////////////// @@ -5092,9 +1472,6 @@ void RendererSceneRenderRD::_free_render_buffer_data(RenderBuffers *rb) { RD::get_singleton()->free(rb->luminance.reduce[i]); } - for (int i = 0; i < rb->luminance.reduce.size(); i++) { - RD::get_singleton()->free(rb->luminance.reduce[i]); - } rb->luminance.reduce.clear(); if (rb->luminance.current.is_valid()) { @@ -5135,6 +1512,13 @@ void RendererSceneRenderRD::_free_render_buffer_data(RenderBuffers *rb) { RD::get_singleton()->free(rb->ssr.normal_scaled); rb->ssr.normal_scaled = RID(); } + + if (rb->ambient_buffer.is_valid()) { + RD::get_singleton()->free(rb->ambient_buffer); + RD::get_singleton()->free(rb->reflection_buffer); + rb->ambient_buffer = RID(); + rb->reflection_buffer = RID(); + } } void RendererSceneRenderRD::_process_sss(RID p_render_buffers, const CameraMatrix &p_camera) { @@ -5168,7 +1552,7 @@ void RendererSceneRenderRD::_process_ssr(RID p_render_buffers, RID p_dest_frameb return; } - Environment *env = environment_owner.getornull(p_environment); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_environment); ERR_FAIL_COND(!env); ERR_FAIL_COND(!env->ssr_enabled); @@ -5213,7 +1597,7 @@ void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environmen RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); ERR_FAIL_COND(!rb); - Environment *env = environment_owner.getornull(p_environment); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_environment); ERR_FAIL_COND(!env); RENDER_TIMESTAMP("Process SSAO"); @@ -5267,9 +1651,11 @@ void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environmen tf.array_layers = 4; tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; rb->ssao.depth = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(rb->ssao.depth, "SSAO Depth"); for (uint32_t i = 0; i < tf.mipmaps; i++) { RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), rb->ssao.depth, 0, i, RD::TEXTURE_SLICE_2D_ARRAY); rb->ssao.depth_slices.push_back(slice); + RD::get_singleton()->set_resource_name(rb->ssao.depth_slices[i], "SSAO Depth Mip " + itos(i) + " "); } } @@ -5282,9 +1668,11 @@ void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environmen tf.array_layers = 4; tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; rb->ssao.ao_deinterleaved = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(rb->ssao.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(), rb->ssao.ao_deinterleaved, i, 0); rb->ssao.ao_deinterleaved_slices.push_back(slice); + RD::get_singleton()->set_resource_name(rb->ssao.ao_deinterleaved_slices[i], "SSAO De-interleaved Array Layer " + itos(i) + " "); } } @@ -5297,9 +1685,11 @@ void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environmen tf.array_layers = 4; tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; rb->ssao.ao_pong = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(rb->ssao.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(), rb->ssao.ao_pong, i, 0); rb->ssao.ao_pong_slices.push_back(slice); + RD::get_singleton()->set_resource_name(rb->ssao.ao_deinterleaved_slices[i], "SSAO De-interleaved Array Layer " + itos(i) + " Pong"); } } @@ -5310,7 +1700,9 @@ void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environmen tf.height = half_height; tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; rb->ssao.importance_map[0] = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(rb->ssao.importance_map[0], "SSAO Importance Map"); rb->ssao.importance_map[1] = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(rb->ssao.importance_map[1], "SSAO Importance Map Pong"); } { RD::TextureFormat tf; @@ -5319,6 +1711,7 @@ void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environmen tf.height = rb->height; tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; rb->ssao.ao_final = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(rb->ssao.ao_final, "SSAO Final"); _render_buffers_uniform_set_changed(p_render_buffers); } ssao_using_half_size = ssao_half_size; @@ -5350,7 +1743,7 @@ void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(RID p_rende RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); ERR_FAIL_COND(!rb); - Environment *env = environment_owner.getornull(p_environment); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_environment); //glow (if enabled) CameraEffects *camfx = camera_effects_owner.getornull(p_camera_effects); @@ -5534,16 +1927,16 @@ void RendererSceneRenderRD::_render_buffers_debug_draw(RID p_render_buffers, RID effects->copy_to_fb_rect(_render_buffers_get_normal_texture(p_render_buffers), storage->render_target_get_rd_framebuffer(rb->render_target), Rect2(Vector2(), rtsize), false, false); } - if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_GI_BUFFER && _render_buffers_get_ambient_texture(p_render_buffers).is_valid()) { + if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_GI_BUFFER && rb->ambient_buffer.is_valid()) { Size2 rtsize = storage->render_target_get_size(rb->render_target); - RID ambient_texture = _render_buffers_get_ambient_texture(p_render_buffers); - RID reflection_texture = _render_buffers_get_reflection_texture(p_render_buffers); + RID ambient_texture = rb->ambient_buffer; + RID reflection_texture = rb->reflection_buffer; effects->copy_to_fb_rect(ambient_texture, storage->render_target_get_rd_framebuffer(rb->render_target), Rect2(Vector2(), rtsize), false, false, false, true, reflection_texture); } } void RendererSceneRenderRD::environment_set_adjustment(RID p_env, bool p_enable, float p_brightness, float p_contrast, float p_saturation, bool p_use_1d_color_correction, RID p_color_correction) { - Environment *env = environment_owner.getornull(p_env); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env); ERR_FAIL_COND(!env); env->adjustments_enabled = p_enable; @@ -5554,152 +1947,6 @@ void RendererSceneRenderRD::environment_set_adjustment(RID p_env, bool p_enable, env->color_correction = p_color_correction; } -void RendererSceneRenderRD::_sdfgi_debug_draw(RID p_render_buffers, const CameraMatrix &p_projection, const Transform &p_transform) { - RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); - ERR_FAIL_COND(!rb); - - if (!rb->sdfgi) { - return; //eh - } - - if (!rb->sdfgi->debug_uniform_set.is_valid() || !RD::get_singleton()->uniform_set_is_valid(rb->sdfgi->debug_uniform_set)) { - 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 < rb->sdfgi->cascades.size()) { - u.ids.push_back(rb->sdfgi->cascades[i].sdf_tex); - } else { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::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 < rb->sdfgi->cascades.size()) { - u.ids.push_back(rb->sdfgi->cascades[i].light_tex); - } else { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::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 < rb->sdfgi->cascades.size()) { - u.ids.push_back(rb->sdfgi->cascades[i].light_aniso_0_tex); - } else { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::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 < rb->sdfgi->cascades.size()) { - u.ids.push_back(rb->sdfgi->cascades[i].light_aniso_1_tex); - } else { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE)); - } - } - uniforms.push_back(u); - } - { - RD::Uniform u; - u.binding = 5; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.ids.push_back(rb->sdfgi->occlusion_texture); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.binding = 8; - u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; - u.ids.push_back(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.ids.push_back(rb->sdfgi->cascades_ubo); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.binding = 10; - u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.ids.push_back(rb->texture); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.binding = 11; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.ids.push_back(rb->sdfgi->lightprobe_texture); - uniforms.push_back(u); - } - rb->sdfgi->debug_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, 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, sdfgi_shader.debug_pipeline); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->debug_uniform_set, 0); - - SDGIShader::DebugPushConstant push_constant; - push_constant.grid_size[0] = rb->sdfgi->cascade_size; - push_constant.grid_size[1] = rb->sdfgi->cascade_size; - push_constant.grid_size[2] = rb->sdfgi->cascade_size; - push_constant.max_cascades = rb->sdfgi->cascades.size(); - push_constant.screen_size[0] = rb->width; - push_constant.screen_size[1] = rb->height; - push_constant.probe_axis_size = rb->sdfgi->probe_axis_count; - push_constant.use_occlusion = rb->sdfgi->uses_occlusion; - push_constant.y_mult = rb->sdfgi->y_mult; - - Vector2 vp_half = p_projection.get_viewport_half_extents(); - push_constant.cam_extent[0] = vp_half.x; - push_constant.cam_extent[1] = vp_half.y; - push_constant.cam_extent[2] = -p_projection.get_z_near(); - - push_constant.cam_transform[0] = p_transform.basis.elements[0][0]; - push_constant.cam_transform[1] = p_transform.basis.elements[1][0]; - push_constant.cam_transform[2] = p_transform.basis.elements[2][0]; - push_constant.cam_transform[3] = 0; - push_constant.cam_transform[4] = p_transform.basis.elements[0][1]; - push_constant.cam_transform[5] = p_transform.basis.elements[1][1]; - push_constant.cam_transform[6] = p_transform.basis.elements[2][1]; - push_constant.cam_transform[7] = 0; - push_constant.cam_transform[8] = p_transform.basis.elements[0][2]; - push_constant.cam_transform[9] = p_transform.basis.elements[1][2]; - push_constant.cam_transform[10] = p_transform.basis.elements[2][2]; - push_constant.cam_transform[11] = 0; - push_constant.cam_transform[12] = p_transform.origin.x; - push_constant.cam_transform[13] = p_transform.origin.y; - push_constant.cam_transform[14] = p_transform.origin.z; - push_constant.cam_transform[15] = 1; - - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::DebugPushConstant)); - - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->width, rb->height, 1, 8, 8, 1); - RD::get_singleton()->compute_list_end(); - - Size2 rtsize = storage->render_target_get_size(rb->render_target); - storage->get_effects()->copy_to_fb_rect(rb->texture, storage->render_target_get_rd_framebuffer(rb->render_target), Rect2(Vector2(), rtsize), true); -} - RID RendererSceneRenderRD::render_buffers_get_back_buffer_texture(RID p_render_buffers) { RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); ERR_FAIL_COND_V(!rb, RID()); @@ -5719,14 +1966,25 @@ RID RendererSceneRenderRD::render_buffers_get_ao_texture(RID p_render_buffers) { RID RendererSceneRenderRD::render_buffers_get_gi_probe_buffer(RID p_render_buffers) { RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); ERR_FAIL_COND_V(!rb, RID()); - if (rb->giprobe_buffer.is_null()) { - rb->giprobe_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(GI::GIProbeData) * RenderBuffers::MAX_GIPROBES); + if (rb->gi.giprobe_buffer.is_null()) { + rb->gi.giprobe_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(RendererSceneGIRD::GIProbeData) * RendererSceneGIRD::MAX_GIPROBES); } - return rb->giprobe_buffer; + return rb->gi.giprobe_buffer; } RID RendererSceneRenderRD::render_buffers_get_default_gi_probe_buffer() { - return default_giprobe_buffer; + return gi.default_giprobe_buffer; +} + +RID RendererSceneRenderRD::render_buffers_get_gi_ambient_texture(RID p_render_buffers) { + RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); + ERR_FAIL_COND_V(!rb, RID()); + return rb->ambient_buffer; +} +RID RendererSceneRenderRD::render_buffers_get_gi_reflection_texture(RID p_render_buffers) { + RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); + ERR_FAIL_COND_V(!rb, RID()); + return rb->reflection_buffer; } uint32_t RendererSceneRenderRD::render_buffers_get_sdfgi_cascade_count(RID p_render_buffers) const { @@ -5764,7 +2022,7 @@ Vector3i RendererSceneRenderRD::render_buffers_get_sdfgi_cascade_probe_offset(RI ERR_FAIL_COND_V(!rb, Vector3i()); ERR_FAIL_COND_V(!rb->sdfgi, Vector3i()); ERR_FAIL_UNSIGNED_INDEX_V(p_cascade, rb->sdfgi->cascades.size(), Vector3i()); - int32_t probe_divisor = rb->sdfgi->cascade_size / SDFGI::PROBE_DIVISOR; + int32_t probe_divisor = rb->sdfgi->cascade_size / RendererSceneGIRD::SDFGI::PROBE_DIVISOR; return rb->sdfgi->cascades[p_cascade].position / probe_divisor; } @@ -5866,6 +2124,11 @@ void RendererSceneRenderRD::render_buffers_configure(RID p_render_buffers, RID p rb->msaa = p_msaa; rb->screen_space_aa = p_screen_space_aa; rb->use_debanding = p_use_debanding; + if (rb->cluster_builder == nullptr) { + rb->cluster_builder = memnew(ClusterBuilderRD); + } + rb->cluster_builder->set_shared(&cluster_builder_shared); + _free_render_buffer_data(rb); { @@ -5906,6 +2169,12 @@ void RendererSceneRenderRD::render_buffers_configure(RID p_render_buffers, RID p rb->data->configure(rb->texture, rb->depth_texture, p_width, p_height, p_msaa); _render_buffers_uniform_set_changed(p_render_buffers); + + rb->cluster_builder->setup(Size2i(p_width, p_height), max_cluster_elements, rb->depth_texture, storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED), rb->texture); +} + +void RendererSceneRenderRD::gi_set_use_half_resolution(bool p_enable) { + gi.half_resolution = p_enable; } void RendererSceneRenderRD::sub_surface_scattering_set_quality(RS::SubSurfaceScatteringQuality p_quality) { @@ -6002,11 +2271,11 @@ void RendererSceneRenderRD::directional_shadow_quality_set(RS::ShadowQuality p_q } int RendererSceneRenderRD::get_roughness_layers() const { - return roughness_layers; + return sky.roughness_layers; } bool RendererSceneRenderRD::is_using_radiance_cubemap_array() const { - return sky_use_cubemap_array; + return sky.sky_use_cubemap_array; } RendererSceneRenderRD::RenderBufferData *RendererSceneRenderRD::render_buffers_get_data(RID p_render_buffers) { @@ -6016,17 +2285,34 @@ RendererSceneRenderRD::RenderBufferData *RendererSceneRenderRD::render_buffers_g } void RendererSceneRenderRD::_setup_reflections(const PagedArray<RID> &p_reflections, const Transform &p_camera_inverse_transform, RID p_environment) { + cluster.reflection_count = 0; + for (uint32_t i = 0; i < (uint32_t)p_reflections.size(); i++) { - RID rpi = p_reflections[i]; + if (cluster.reflection_count == cluster.max_reflections) { + break; + } - if (i >= cluster.max_reflections) { - reflection_probe_instance_set_render_index(rpi, 0); //invalid, but something needs to be set + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_reflections[i]); + if (!rpi) { continue; } - reflection_probe_instance_set_render_index(rpi, i); + cluster.reflection_sort[cluster.reflection_count].instance = rpi; + cluster.reflection_sort[cluster.reflection_count].depth = -p_camera_inverse_transform.xform(rpi->transform.origin).z; + cluster.reflection_count++; + } + + if (cluster.reflection_count > 0) { + SortArray<Cluster::InstanceSort<ReflectionProbeInstance>> sort_array; + sort_array.sort(cluster.reflection_sort, cluster.reflection_count); + } + + for (uint32_t i = 0; i < cluster.reflection_count; i++) { + ReflectionProbeInstance *rpi = cluster.reflection_sort[i].instance; + + rpi->render_index = i; - RID base_probe = reflection_probe_instance_get_probe(rpi); + RID base_probe = rpi->probe; Cluster::ReflectionData &reflection_ubo = cluster.reflections[i]; @@ -6035,7 +2321,7 @@ void RendererSceneRenderRD::_setup_reflections(const PagedArray<RID> &p_reflecti reflection_ubo.box_extents[0] = extents.x; reflection_ubo.box_extents[1] = extents.y; reflection_ubo.box_extents[2] = extents.z; - reflection_ubo.index = reflection_probe_instance_get_atlas_index(rpi); + reflection_ubo.index = rpi->atlas_index; Vector3 origin_offset = storage->reflection_probe_get_origin_offset(base_probe); @@ -6044,46 +2330,50 @@ void RendererSceneRenderRD::_setup_reflections(const PagedArray<RID> &p_reflecti reflection_ubo.box_offset[2] = origin_offset.z; reflection_ubo.mask = storage->reflection_probe_get_cull_mask(base_probe); - float intensity = storage->reflection_probe_get_intensity(base_probe); - bool interior = storage->reflection_probe_is_interior(base_probe); - bool box_projection = storage->reflection_probe_is_box_projection(base_probe); + reflection_ubo.intensity = storage->reflection_probe_get_intensity(base_probe); + reflection_ubo.ambient_mode = storage->reflection_probe_get_ambient_mode(base_probe); - reflection_ubo.params[0] = intensity; - reflection_ubo.params[1] = 0; - reflection_ubo.params[2] = interior ? 1.0 : 0.0; - reflection_ubo.params[3] = box_projection ? 1.0 : 0.0; + reflection_ubo.exterior = !storage->reflection_probe_is_interior(base_probe); + reflection_ubo.box_project = storage->reflection_probe_is_box_projection(base_probe); Color ambient_linear = storage->reflection_probe_get_ambient_color(base_probe).to_linear(); float interior_ambient_energy = storage->reflection_probe_get_ambient_color_energy(base_probe); - uint32_t ambient_mode = storage->reflection_probe_get_ambient_mode(base_probe); 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; - reflection_ubo.ambient_mode = ambient_mode; - Transform transform = reflection_probe_instance_get_transform(rpi); + Transform transform = rpi->transform; Transform proj = (p_camera_inverse_transform * transform).inverse(); RendererStorageRD::store_transform(proj, reflection_ubo.local_matrix); - cluster.builder.add_reflection_probe(transform, extents); + current_cluster_builder->add_box(ClusterBuilderRD::BOX_TYPE_REFLECTION_PROBE, transform, extents); - reflection_probe_instance_set_render_pass(rpi, RSG::rasterizer->get_frame_number()); + rpi->last_pass = RSG::rasterizer->get_frame_number(); } - if (p_reflections.size()) { - RD::get_singleton()->buffer_update(cluster.reflection_buffer, 0, MIN(cluster.max_reflections, (unsigned int)p_reflections.size()) * sizeof(ReflectionData), cluster.reflections, true); + if (cluster.reflection_count) { + RD::get_singleton()->buffer_update(cluster.reflection_buffer, 0, cluster.reflection_count * sizeof(RendererSceneSkyRD::ReflectionData), cluster.reflections, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); } } -void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const Transform &p_camera_inverse_transform, RID p_shadow_atlas, bool p_using_shadows, uint32_t &r_directional_light_count, uint32_t &r_positional_light_count) { - uint32_t light_count = 0; +void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const Transform &p_camera_transform, RID p_shadow_atlas, bool p_using_shadows, uint32_t &r_directional_light_count, uint32_t &r_positional_light_count) { + Transform inverse_transform = p_camera_transform.affine_inverse(); + r_directional_light_count = 0; r_positional_light_count = 0; - sky_scene_state.ubo.directional_light_count = 0; + sky.sky_scene_state.ubo.directional_light_count = 0; + + Plane camera_plane(p_camera_transform.origin, -p_camera_transform.basis.get_axis(Vector3::AXIS_Z).normalized()); + + cluster.omni_light_count = 0; + cluster.spot_light_count = 0; for (int i = 0; i < (int)p_lights.size(); i++) { - RID li = p_lights[i]; - RID base = light_instance_get_base_light(li); + LightInstance *li = light_instance_owner.getornull(p_lights[i]); + if (!li) { + continue; + } + RID base = li->light; ERR_CONTINUE(base.is_null()); @@ -6091,9 +2381,9 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const switch (type) { case RS::LIGHT_DIRECTIONAL: { // Copy to SkyDirectionalLightData - if (r_directional_light_count < sky_scene_state.max_directional_lights) { - SkyDirectionalLightData &sky_light_data = sky_scene_state.directional_lights[r_directional_light_count]; - Transform light_transform = light_instance_get_base_transform(li); + if (r_directional_light_count < sky.sky_scene_state.max_directional_lights) { + RendererSceneSkyRD::SkyDirectionalLightData &sky_light_data = sky.sky_scene_state.directional_lights[r_directional_light_count]; + Transform light_transform = li->transform; Vector3 world_direction = light_transform.basis.xform(Vector3(0, 0, 1)).normalized(); sky_light_data.direction[0] = world_direction.x; @@ -6120,7 +2410,7 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const angular_diameter = 0.0; } sky_light_data.size = angular_diameter; - sky_scene_state.ubo.directional_light_count++; + sky.sky_scene_state.ubo.directional_light_count++; } if (r_directional_light_count >= cluster.max_directional_lights || storage->light_directional_is_sky_only(base)) { @@ -6129,9 +2419,9 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const Cluster::DirectionalLightData &light_data = cluster.directional_lights[r_directional_light_count]; - Transform light_transform = light_instance_get_base_transform(li); + Transform light_transform = li->transform; - Vector3 direction = p_camera_inverse_transform.basis.xform(light_transform.basis.xform(Vector3(0, 0, 1))).normalized(); + 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; @@ -6210,28 +2500,28 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const int limit = smode == RS::LIGHT_DIRECTIONAL_SHADOW_ORTHOGONAL ? 0 : (smode == RS::LIGHT_DIRECTIONAL_SHADOW_PARALLEL_2_SPLITS ? 1 : 3); light_data.blend_splits = storage->light_directional_get_blend_splits(base); for (int j = 0; j < 4; j++) { - Rect2 atlas_rect = light_instance_get_directional_shadow_atlas_rect(li, j); - CameraMatrix matrix = light_instance_get_shadow_camera(li, j); - float split = light_instance_get_directional_shadow_split(li, MIN(limit, j)); + Rect2 atlas_rect = li->shadow_transform[j].atlas_rect; + CameraMatrix matrix = li->shadow_transform[j].camera; + float split = li->shadow_transform[MIN(limit, j)].split; CameraMatrix bias; bias.set_light_bias(); CameraMatrix rectm; rectm.set_light_atlas_rect(atlas_rect); - Transform modelview = (p_camera_inverse_transform * light_instance_get_shadow_transform(li, j)).inverse(); + Transform modelview = (inverse_transform * li->shadow_transform[j].transform).inverse(); CameraMatrix shadow_mtx = rectm * bias * matrix * modelview; light_data.shadow_split_offsets[j] = split; - float bias_scale = light_instance_get_shadow_bias_scale(li, j); + float bias_scale = li->shadow_transform[j].bias_scale; light_data.shadow_bias[j] = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) * bias_scale; - light_data.shadow_normal_bias[j] = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * light_instance_get_directional_shadow_texel_size(li, j); + light_data.shadow_normal_bias[j] = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * li->shadow_transform[j].shadow_texel_size; light_data.shadow_transmittance_bias[j] = storage->light_get_transmittance_bias(base) * bias_scale; - light_data.shadow_z_range[j] = light_instance_get_shadow_range(li, j); - light_data.shadow_range_begin[j] = light_instance_get_shadow_range_begin(li, j); + light_data.shadow_z_range[j] = li->shadow_transform[j].farplane; + light_data.shadow_range_begin[j] = li->shadow_transform[j].range_begin; RendererStorageRD::store_camera(shadow_mtx, light_data.shadow_matrices[j]); - Vector2 uv_scale = light_instance_get_shadow_uv_scale(li, j); + Vector2 uv_scale = li->shadow_transform[j].uv_scale; uv_scale *= atlas_rect.size; //adapt to atlas size switch (j) { case 0: { @@ -6268,170 +2558,201 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const r_directional_light_count++; } break; - case RS::LIGHT_SPOT: case RS::LIGHT_OMNI: { - if (light_count >= cluster.max_lights) { + if (cluster.omni_light_count >= cluster.max_lights) { continue; } - Transform light_transform = light_instance_get_base_transform(li); + cluster.omni_light_sort[cluster.omni_light_count].instance = li; + cluster.omni_light_sort[cluster.omni_light_count].depth = camera_plane.distance_to(li->transform.origin); + cluster.omni_light_count++; + } break; + case RS::LIGHT_SPOT: { + if (cluster.spot_light_count >= cluster.max_lights) { + continue; + } - Cluster::LightData &light_data = cluster.lights[light_count]; - cluster.lights_instances[light_count] = li; + cluster.spot_light_sort[cluster.spot_light_count].instance = li; + cluster.spot_light_sort[cluster.spot_light_count].depth = camera_plane.distance_to(li->transform.origin); + cluster.spot_light_count++; + } break; + } - float sign = storage->light_is_negative(base) ? -1 : 1; - Color linear_col = storage->light_get_color(base).to_linear(); + li->last_pass = RSG::rasterizer->get_frame_number(); + } - light_data.attenuation_energy[0] = Math::make_half_float(storage->light_get_param(base, RS::LIGHT_PARAM_ATTENUATION)); - light_data.attenuation_energy[1] = Math::make_half_float(sign * storage->light_get_param(base, RS::LIGHT_PARAM_ENERGY) * Math_PI); + if (cluster.omni_light_count) { + SortArray<Cluster::InstanceSort<LightInstance>> sorter; + sorter.sort(cluster.omni_light_sort, cluster.omni_light_count); + } - light_data.color_specular[0] = MIN(uint32_t(linear_col.r * 255), 255); - light_data.color_specular[1] = MIN(uint32_t(linear_col.g * 255), 255); - light_data.color_specular[2] = MIN(uint32_t(linear_col.b * 255), 255); - light_data.color_specular[3] = MIN(uint32_t(storage->light_get_param(base, RS::LIGHT_PARAM_SPECULAR) * 255), 255); + if (cluster.spot_light_count) { + SortArray<Cluster::InstanceSort<LightInstance>> sorter; + sorter.sort(cluster.spot_light_sort, cluster.spot_light_count); + } - float radius = MAX(0.001, storage->light_get_param(base, RS::LIGHT_PARAM_RANGE)); - light_data.inv_radius = 1.0 / radius; + ShadowAtlas *shadow_atlas = nullptr; - Vector3 pos = p_camera_inverse_transform.xform(light_transform.origin); + if (p_shadow_atlas.is_valid() && p_using_shadows) { + shadow_atlas = shadow_atlas_owner.getornull(p_shadow_atlas); + } - light_data.position[0] = pos.x; - light_data.position[1] = pos.y; - light_data.position[2] = pos.z; + for (uint32_t i = 0; i < (cluster.omni_light_count + cluster.spot_light_count); i++) { + uint32_t index = (i < cluster.omni_light_count) ? i : i - (cluster.omni_light_count); + Cluster::LightData &light_data = (i < cluster.omni_light_count) ? cluster.omni_lights[index] : cluster.spot_lights[index]; + RS::LightType type = (i < cluster.omni_light_count) ? RS::LIGHT_OMNI : RS::LIGHT_SPOT; + LightInstance *li = (i < cluster.omni_light_count) ? cluster.omni_light_sort[index].instance : cluster.spot_light_sort[index].instance; + RID base = li->light; - Vector3 direction = p_camera_inverse_transform.basis.xform(light_transform.basis.xform(Vector3(0, 0, -1))).normalized(); + Transform light_transform = li->transform; - light_data.direction[0] = direction.x; - light_data.direction[1] = direction.y; - light_data.direction[2] = direction.z; + float sign = storage->light_is_negative(base) ? -1 : 1; + Color linear_col = storage->light_get_color(base).to_linear(); - float size = storage->light_get_param(base, RS::LIGHT_PARAM_SIZE); + light_data.attenuation = storage->light_get_param(base, RS::LIGHT_PARAM_ATTENUATION); - light_data.size = size; + float energy = sign * storage->light_get_param(base, RS::LIGHT_PARAM_ENERGY) * Math_PI; - light_data.cone_attenuation_angle[0] = Math::make_half_float(storage->light_get_param(base, RS::LIGHT_PARAM_SPOT_ATTENUATION)); - float spot_angle = storage->light_get_param(base, RS::LIGHT_PARAM_SPOT_ANGLE); - light_data.cone_attenuation_angle[1] = Math::make_half_float(Math::cos(Math::deg2rad(spot_angle))); + 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 = storage->light_get_param(base, RS::LIGHT_PARAM_SPECULAR) * 2.0; - light_data.mask = storage->light_get_cull_mask(base); + float radius = MAX(0.001, storage->light_get_param(base, RS::LIGHT_PARAM_RANGE)); + light_data.inv_radius = 1.0 / radius; - light_data.atlas_rect[0] = 0; - light_data.atlas_rect[1] = 0; - light_data.atlas_rect[2] = 0; - light_data.atlas_rect[3] = 0; + Vector3 pos = inverse_transform.xform(light_transform.origin); - RID projector = storage->light_get_projector(base); + light_data.position[0] = pos.x; + light_data.position[1] = pos.y; + light_data.position[2] = pos.z; - if (projector.is_valid()) { - Rect2 rect = storage->decal_atlas_get_texture_rect(projector); + Vector3 direction = inverse_transform.basis.xform(light_transform.basis.xform(Vector3(0, 0, -1))).normalized(); - 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; - } + light_data.direction[0] = direction.x; + light_data.direction[1] = direction.y; + light_data.direction[2] = direction.z; - if (p_using_shadows && p_shadow_atlas.is_valid() && shadow_atlas_owns_light_instance(p_shadow_atlas, li)) { - // fill in the shadow information + float size = storage->light_get_param(base, RS::LIGHT_PARAM_SIZE); - Color shadow_color = storage->light_get_shadow_color(base); + light_data.size = size; - light_data.shadow_color_enabled[0] = MIN(uint32_t(shadow_color.r * 255), 255); - light_data.shadow_color_enabled[1] = MIN(uint32_t(shadow_color.g * 255), 255); - light_data.shadow_color_enabled[2] = MIN(uint32_t(shadow_color.b * 255), 255); - light_data.shadow_color_enabled[3] = 255; + light_data.inv_spot_attenuation = 1.0f / storage->light_get_param(base, RS::LIGHT_PARAM_SPOT_ATTENUATION); + float spot_angle = storage->light_get_param(base, RS::LIGHT_PARAM_SPOT_ANGLE); + light_data.cos_spot_angle = Math::cos(Math::deg2rad(spot_angle)); - if (type == RS::LIGHT_SPOT) { - light_data.shadow_bias = (storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) * radius / 10.0); - float shadow_texel_size = Math::tan(Math::deg2rad(spot_angle)) * radius * 2.0; - shadow_texel_size *= light_instance_get_shadow_texel_size(li, p_shadow_atlas); + light_data.mask = storage->light_get_cull_mask(base); - light_data.shadow_normal_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * shadow_texel_size; + light_data.atlas_rect[0] = 0; + light_data.atlas_rect[1] = 0; + light_data.atlas_rect[2] = 0; + light_data.atlas_rect[3] = 0; - } else { //omni - light_data.shadow_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) * radius / 10.0; - float shadow_texel_size = light_instance_get_shadow_texel_size(li, p_shadow_atlas); - light_data.shadow_normal_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * shadow_texel_size * 2.0; // applied in -1 .. 1 space - } + RID projector = storage->light_get_projector(base); - light_data.transmittance_bias = storage->light_get_transmittance_bias(base); + if (projector.is_valid()) { + Rect2 rect = storage->decal_atlas_get_texture_rect(projector); - Rect2 rect = light_instance_get_shadow_atlas_rect(li, p_shadow_atlas); + 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; + } - 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; + if (shadow_atlas && shadow_atlas->shadow_owners.has(li->self)) { + // fill in the shadow information - light_data.soft_shadow_scale = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BLUR); - light_data.shadow_volumetric_fog_fade = 1.0 / storage->light_get_shadow_volumetric_fog_fade(base); + light_data.shadow_enabled = true; - if (type == RS::LIGHT_OMNI) { - light_data.atlas_rect[3] *= 0.5; //one paraboloid on top of another - Transform proj = (p_camera_inverse_transform * light_transform).inverse(); + if (type == RS::LIGHT_SPOT) { + light_data.shadow_bias = (storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) * radius / 10.0); + float shadow_texel_size = Math::tan(Math::deg2rad(spot_angle)) * radius * 2.0; + shadow_texel_size *= light_instance_get_shadow_texel_size(li->self, p_shadow_atlas); - RendererStorageRD::store_transform(proj, light_data.shadow_matrix); + light_data.shadow_normal_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * shadow_texel_size; - if (size > 0.0) { - light_data.soft_shadow_size = size; - } else { - light_data.soft_shadow_size = 0.0; - light_data.soft_shadow_scale *= shadows_quality_radius_get(); // Only use quality radius for PCF - } + } else { //omni + light_data.shadow_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) * radius / 10.0; + float shadow_texel_size = light_instance_get_shadow_texel_size(li->self, p_shadow_atlas); + light_data.shadow_normal_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * shadow_texel_size * 2.0; // applied in -1 .. 1 space + } - } else if (type == RS::LIGHT_SPOT) { - Transform modelview = (p_camera_inverse_transform * light_transform).inverse(); - CameraMatrix bias; - bias.set_light_bias(); + light_data.transmittance_bias = storage->light_get_transmittance_bias(base); - CameraMatrix shadow_mtx = bias * light_instance_get_shadow_camera(li, 0) * modelview; - RendererStorageRD::store_camera(shadow_mtx, light_data.shadow_matrix); + Rect2 rect = light_instance_get_shadow_atlas_rect(li->self, p_shadow_atlas); - if (size > 0.0) { - CameraMatrix cm = light_instance_get_shadow_camera(li, 0); - float half_np = cm.get_z_near() * Math::tan(Math::deg2rad(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 *= shadows_quality_radius_get(); // Only use quality radius for PCF - } - } + 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 = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BLUR); + light_data.shadow_volumetric_fog_fade = 1.0 / storage->light_get_shadow_volumetric_fog_fade(base); + + if (type == RS::LIGHT_OMNI) { + light_data.atlas_rect[3] *= 0.5; //one paraboloid on top of another + Transform proj = (inverse_transform * light_transform).inverse(); + + RendererStorageRD::store_transform(proj, light_data.shadow_matrix); + + if (size > 0.0) { + light_data.soft_shadow_size = size; } else { - light_data.shadow_color_enabled[3] = 0; + light_data.soft_shadow_size = 0.0; + light_data.soft_shadow_scale *= shadows_quality_radius_get(); // Only use quality radius for PCF } - light_instance_set_index(li, light_count); + } else if (type == RS::LIGHT_SPOT) { + Transform modelview = (inverse_transform * light_transform).inverse(); + CameraMatrix bias; + bias.set_light_bias(); - cluster.builder.add_light(type == RS::LIGHT_SPOT ? LightClusterBuilder::LIGHT_TYPE_SPOT : LightClusterBuilder::LIGHT_TYPE_OMNI, light_transform, radius, spot_angle); + CameraMatrix shadow_mtx = bias * li->shadow_transform[0].camera * modelview; + RendererStorageRD::store_camera(shadow_mtx, light_data.shadow_matrix); - light_count++; - r_positional_light_count++; - } break; + if (size > 0.0) { + CameraMatrix cm = li->shadow_transform[0].camera; + float half_np = cm.get_z_near() * Math::tan(Math::deg2rad(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 *= shadows_quality_radius_get(); // Only use quality radius for PCF + } + } + } else { + light_data.shadow_enabled = false; } - light_instance_set_render_pass(li, RSG::rasterizer->get_frame_number()); + li->light_index = index; - //update UBO for forward rendering, blit to texture for clustered + current_cluster_builder->add_light(type == RS::LIGHT_SPOT ? ClusterBuilderRD::LIGHT_TYPE_SPOT : ClusterBuilderRD::LIGHT_TYPE_OMNI, light_transform, radius, spot_angle); + + r_positional_light_count++; } - if (light_count) { - RD::get_singleton()->buffer_update(cluster.light_buffer, 0, sizeof(Cluster::LightData) * light_count, cluster.lights, true); + //update without barriers + if (cluster.omni_light_count) { + RD::get_singleton()->buffer_update(cluster.omni_light_buffer, 0, sizeof(Cluster::LightData) * cluster.omni_light_count, cluster.omni_lights, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); + } + + if (cluster.spot_light_count) { + RD::get_singleton()->buffer_update(cluster.spot_light_buffer, 0, sizeof(Cluster::LightData) * cluster.spot_light_count, cluster.spot_lights, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); } if (r_directional_light_count) { - RD::get_singleton()->buffer_update(cluster.directional_light_buffer, 0, sizeof(Cluster::DirectionalLightData) * r_directional_light_count, cluster.directional_lights, true); + RD::get_singleton()->buffer_update(cluster.directional_light_buffer, 0, sizeof(Cluster::DirectionalLightData) * r_directional_light_count, cluster.directional_lights, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); } } @@ -6440,18 +2761,26 @@ void RendererSceneRenderRD::_setup_decals(const PagedArray<RID> &p_decals, const uv_xform.basis.scale(Vector3(2.0, 1.0, 2.0)); uv_xform.origin = Vector3(-1.0, 0.0, -1.0); - uint32_t decal_count = MIN((uint32_t)p_decals.size(), cluster.max_decals); - int idx = 0; + uint32_t decal_count = p_decals.size(); + + cluster.decal_count = 0; + for (uint32_t i = 0; i < decal_count; i++) { - RID di = p_decals[i]; - RID decal = decal_instance_get_base(di); + if (cluster.decal_count == cluster.max_decals) { + break; + } - Transform xform = decal_instance_get_transform(di); + DecalInstance *di = decal_instance_owner.getornull(p_decals[i]); + if (!di) { + continue; + } + RID decal = di->decal; - float fade = 1.0; + Transform xform = di->transform; + + real_t distance = -p_camera_inverse_xform.xform(xform.origin).z; if (storage->decal_is_distance_fade_enabled(decal)) { - real_t distance = -p_camera_inverse_xform.xform(xform.origin).z; float fade_begin = storage->decal_get_distance_fade_begin(decal); float fade_length = storage->decal_get_distance_fade_length(decal); @@ -6459,18 +2788,43 @@ void RendererSceneRenderRD::_setup_decals(const PagedArray<RID> &p_decals, const if (distance > fade_begin + fade_length) { continue; // do not use this decal, its invisible } + } + } + + cluster.decal_sort[cluster.decal_count].instance = di; + cluster.decal_sort[cluster.decal_count].depth = distance; + cluster.decal_count++; + } + + if (cluster.decal_count > 0) { + SortArray<Cluster::InstanceSort<DecalInstance>> sort_array; + sort_array.sort(cluster.decal_sort, cluster.decal_count); + } + + for (uint32_t i = 0; i < cluster.decal_count; i++) { + DecalInstance *di = cluster.decal_sort[i].instance; + RID decal = di->decal; + + Transform xform = di->transform; + float fade = 1.0; + + if (storage->decal_is_distance_fade_enabled(decal)) { + real_t distance = -p_camera_inverse_xform.xform(xform.origin).z; + float fade_begin = storage->decal_get_distance_fade_begin(decal); + float fade_length = storage->decal_get_distance_fade_length(decal); + if (distance > fade_begin) { fade = 1.0 - (distance - fade_begin) / fade_length; } } - Cluster::DecalData &dd = cluster.decals[idx]; + Cluster::DecalData &dd = cluster.decals[i]; Vector3 decal_extents = storage->decal_get_extents(decal); Transform scale_xform; scale_xform.basis.scale(Vector3(decal_extents.x, decal_extents.y, decal_extents.z)); - Transform to_decal_xform = (p_camera_inverse_xform * decal_instance_get_transform(di) * scale_xform * uv_xform).affine_inverse(); + Transform to_decal_xform = (p_camera_inverse_xform * di->transform * scale_xform * uv_xform).affine_inverse(); RendererStorageRD::store_transform(to_decal_xform, dd.xform); Vector3 normal = xform.basis.get_axis(Vector3::AXIS_Y).normalized(); @@ -6555,19 +2909,18 @@ void RendererSceneRenderRD::_setup_decals(const PagedArray<RID> &p_decals, const dd.upper_fade = storage->decal_get_upper_fade(decal); dd.lower_fade = storage->decal_get_lower_fade(decal); - cluster.builder.add_decal(xform, decal_extents); - - idx++; + current_cluster_builder->add_box(ClusterBuilderRD::BOX_TYPE_DECAL, xform, decal_extents); } - if (idx > 0) { - RD::get_singleton()->buffer_update(cluster.decal_buffer, 0, sizeof(Cluster::DecalData) * idx, cluster.decals, true); + if (cluster.decal_count > 0) { + RD::get_singleton()->buffer_update(cluster.decal_buffer, 0, sizeof(Cluster::DecalData) * cluster.decal_count, cluster.decals, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); } } void RendererSceneRenderRD::_volumetric_fog_erase(RenderBuffers *rb) { ERR_FAIL_COND(!rb->volumetric_fog); + RD::get_singleton()->free(rb->volumetric_fog->prev_light_density_map); RD::get_singleton()->free(rb->volumetric_fog->light_density_map); RD::get_singleton()->free(rb->volumetric_fog->fog_map); @@ -6589,53 +2942,10 @@ void RendererSceneRenderRD::_volumetric_fog_erase(RenderBuffers *rb) { rb->volumetric_fog = nullptr; } -void RendererSceneRenderRD::_allocate_shadow_shrink_stages(RID p_base, int p_base_size, Vector<ShadowShrinkStage> &shrink_stages, uint32_t p_target_size) { - //create fog mipmaps - uint32_t fog_texture_size = p_target_size; - uint32_t base_texture_size = p_base_size; - - ShadowShrinkStage first; - first.size = base_texture_size; - first.texture = p_base; - shrink_stages.push_back(first); //put depth first in case we dont find smaller ones - - while (fog_texture_size < base_texture_size) { - base_texture_size = MAX(base_texture_size / 8, fog_texture_size); - - ShadowShrinkStage s; - s.size = base_texture_size; - - RD::TextureFormat tf; - tf.format = RD::DATA_FORMAT_R32_SFLOAT; - tf.width = base_texture_size; - tf.height = base_texture_size; - tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT; - - if (base_texture_size == fog_texture_size) { - s.filter_texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); - tf.usage_bits |= RD::TEXTURE_USAGE_SAMPLING_BIT; - } - - s.texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); - - shrink_stages.push_back(s); - } -} - -void RendererSceneRenderRD::_clear_shadow_shrink_stages(Vector<ShadowShrinkStage> &shrink_stages) { - for (int i = 1; i < shrink_stages.size(); i++) { - RD::get_singleton()->free(shrink_stages[i].texture); - if (shrink_stages[i].filter_texture.is_valid()) { - RD::get_singleton()->free(shrink_stages[i].filter_texture); - } - } - shrink_stages.clear(); -} - void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_environment, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_shadow_atlas, int p_directional_light_count, bool p_use_directional_shadows, int p_positional_light_count, int p_gi_probe_count) { RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); ERR_FAIL_COND(!rb); - Environment *env = environment_owner.getornull(p_environment); + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_environment); float ratio = float(rb->width) / float((rb->width + rb->height) / 2); uint32_t target_width = uint32_t(float(volumetric_fog_size) * ratio); @@ -6654,6 +2964,8 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e return; } + RENDER_TIMESTAMP(">Volumetric Fog"); + if (env && env->volumetric_fog_enabled && !rb->volumetric_fog) { //required volumetric fog but not existing, create rb->volumetric_fog = memnew(VolumetricFog); @@ -6667,11 +2979,16 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e tf.height = target_height; tf.depth = volumetric_fog_depth; tf.texture_type = RD::TEXTURE_TYPE_3D; - tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT; + tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT; rb->volumetric_fog->light_density_map = RD::get_singleton()->texture_create(tf, RD::TextureView()); - tf.usage_bits |= RD::TEXTURE_USAGE_SAMPLING_BIT; + tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; + + rb->volumetric_fog->prev_light_density_map = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->texture_clear(rb->volumetric_fog->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; rb->volumetric_fog->fog_map = RD::get_singleton()->texture_create(tf, RD::TextureView()); _render_buffers_uniform_set_changed(p_render_buffers); @@ -6685,163 +3002,7 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e uniforms.push_back(u); } - rb->volumetric_fog->sky_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sky_shader.default_shader_rd, SKY_SET_FOG); - } - - //update directional shadow - - if (p_use_directional_shadows) { - if (directional_shadow.shrink_stages.is_empty()) { - if (rb->volumetric_fog->uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(rb->volumetric_fog->uniform_set)) { - //invalidate uniform set, we will need a new one - RD::get_singleton()->free(rb->volumetric_fog->uniform_set); - rb->volumetric_fog->uniform_set = RID(); - } - _allocate_shadow_shrink_stages(directional_shadow.depth, directional_shadow.size, directional_shadow.shrink_stages, volumetric_fog_directional_shadow_shrink); - } - - if (directional_shadow.shrink_stages.size() > 1) { - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - for (int i = 1; i < directional_shadow.shrink_stages.size(); i++) { - int32_t src_size = directional_shadow.shrink_stages[i - 1].size; - int32_t dst_size = directional_shadow.shrink_stages[i].size; - Rect2i r(0, 0, src_size, src_size); - int32_t shrink_limit = 8 / (src_size / dst_size); - - storage->get_effects()->reduce_shadow(directional_shadow.shrink_stages[i - 1].texture, directional_shadow.shrink_stages[i].texture, Size2i(src_size, src_size), r, shrink_limit, compute_list); - RD::get_singleton()->compute_list_add_barrier(compute_list); - if (env->volumetric_fog_shadow_filter != RS::ENV_VOLUMETRIC_FOG_SHADOW_FILTER_DISABLED && directional_shadow.shrink_stages[i].filter_texture.is_valid()) { - Rect2i rf(0, 0, dst_size, dst_size); - storage->get_effects()->filter_shadow(directional_shadow.shrink_stages[i].texture, directional_shadow.shrink_stages[i].filter_texture, Size2i(dst_size, dst_size), rf, env->volumetric_fog_shadow_filter, compute_list); - } - } - RD::get_singleton()->compute_list_end(); - } - } - - ShadowAtlas *shadow_atlas = shadow_atlas_owner.getornull(p_shadow_atlas); - - if (shadow_atlas) { - //shrink shadows that need to be shrunk - - bool force_shrink_shadows = false; - - if (shadow_atlas->shrink_stages.is_empty()) { - if (rb->volumetric_fog->uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(rb->volumetric_fog->uniform_set)) { - //invalidate uniform set, we will need a new one - RD::get_singleton()->free(rb->volumetric_fog->uniform_set); - rb->volumetric_fog->uniform_set = RID(); - } - _allocate_shadow_shrink_stages(shadow_atlas->depth, shadow_atlas->size, shadow_atlas->shrink_stages, volumetric_fog_positional_shadow_shrink); - force_shrink_shadows = true; - } - - if (rb->volumetric_fog->last_shadow_filter != env->volumetric_fog_shadow_filter) { - //if shadow filter changed, invalidate caches - rb->volumetric_fog->last_shadow_filter = env->volumetric_fog_shadow_filter; - force_shrink_shadows = true; - } - - cluster.lights_shadow_rect_cache_count = 0; - - for (int i = 0; i < p_positional_light_count; i++) { - if (cluster.lights[i].shadow_color_enabled[3] > 127) { - RID li = cluster.lights_instances[i]; - - ERR_CONTINUE(!shadow_atlas->shadow_owners.has(li)); - - uint32_t key = shadow_atlas->shadow_owners[li]; - - uint32_t quadrant = (key >> ShadowAtlas::QUADRANT_SHIFT) & 0x3; - uint32_t shadow = key & ShadowAtlas::SHADOW_INDEX_MASK; - - ERR_CONTINUE((int)shadow >= shadow_atlas->quadrants[quadrant].shadows.size()); - - ShadowAtlas::Quadrant::Shadow &s = shadow_atlas->quadrants[quadrant].shadows.write[shadow]; - - if (!force_shrink_shadows && s.fog_version == s.version) { - continue; //do not update, no need - } - - s.fog_version = s.version; - - uint32_t quadrant_size = shadow_atlas->size >> 1; - - Rect2i atlas_rect; - - atlas_rect.position.x = (quadrant & 1) * quadrant_size; - atlas_rect.position.y = (quadrant >> 1) * quadrant_size; - - uint32_t shadow_size = (quadrant_size / shadow_atlas->quadrants[quadrant].subdivision); - atlas_rect.position.x += (shadow % shadow_atlas->quadrants[quadrant].subdivision) * shadow_size; - atlas_rect.position.y += (shadow / shadow_atlas->quadrants[quadrant].subdivision) * shadow_size; - - atlas_rect.size.x = shadow_size; - atlas_rect.size.y = shadow_size; - - cluster.lights_shadow_rect_cache[cluster.lights_shadow_rect_cache_count] = atlas_rect; - - cluster.lights_shadow_rect_cache_count++; - - if (cluster.lights_shadow_rect_cache_count == cluster.max_lights) { - break; //light limit reached - } - } - } - - if (cluster.lights_shadow_rect_cache_count > 0) { - //there are shadows to be shrunk, try to do them in parallel - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - - for (int i = 1; i < shadow_atlas->shrink_stages.size(); i++) { - int32_t base_size = shadow_atlas->shrink_stages[0].size; - int32_t src_size = shadow_atlas->shrink_stages[i - 1].size; - int32_t dst_size = shadow_atlas->shrink_stages[i].size; - - uint32_t rect_divisor = base_size / src_size; - - int32_t shrink_limit = 8 / (src_size / dst_size); - - //shrink in parallel for more performance - for (uint32_t j = 0; j < cluster.lights_shadow_rect_cache_count; j++) { - Rect2i src_rect = cluster.lights_shadow_rect_cache[j]; - - src_rect.position /= rect_divisor; - src_rect.size /= rect_divisor; - - storage->get_effects()->reduce_shadow(shadow_atlas->shrink_stages[i - 1].texture, shadow_atlas->shrink_stages[i].texture, Size2i(src_size, src_size), src_rect, shrink_limit, compute_list); - } - - RD::get_singleton()->compute_list_add_barrier(compute_list); - - if (env->volumetric_fog_shadow_filter != RS::ENV_VOLUMETRIC_FOG_SHADOW_FILTER_DISABLED && shadow_atlas->shrink_stages[i].filter_texture.is_valid()) { - uint32_t filter_divisor = base_size / dst_size; - - //filter in parallel for more performance - for (uint32_t j = 0; j < cluster.lights_shadow_rect_cache_count; j++) { - Rect2i dst_rect = cluster.lights_shadow_rect_cache[j]; - - dst_rect.position /= filter_divisor; - dst_rect.size /= filter_divisor; - - storage->get_effects()->filter_shadow(shadow_atlas->shrink_stages[i].texture, shadow_atlas->shrink_stages[i].filter_texture, Size2i(dst_size, dst_size), dst_rect, env->volumetric_fog_shadow_filter, compute_list, true, false); - } - - RD::get_singleton()->compute_list_add_barrier(compute_list); - - for (uint32_t j = 0; j < cluster.lights_shadow_rect_cache_count; j++) { - Rect2i dst_rect = cluster.lights_shadow_rect_cache[j]; - - dst_rect.position /= filter_divisor; - dst_rect.size /= filter_divisor; - - storage->get_effects()->filter_shadow(shadow_atlas->shrink_stages[i].texture, shadow_atlas->shrink_stages[i].filter_texture, Size2i(dst_size, dst_size), dst_rect, env->volumetric_fog_shadow_filter, compute_list, false, true); - } - } - } - - RD::get_singleton()->compute_list_end(); - } + rb->volumetric_fog->sky_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sky.sky_shader.default_shader_rd, RendererSceneSkyRD::SKY_SET_FOG); } //update volumetric fog @@ -6855,10 +3016,11 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.binding = 1; - if (shadow_atlas == nullptr || shadow_atlas->shrink_stages.size() == 0) { + ShadowAtlas *shadow_atlas = shadow_atlas_owner.getornull(p_shadow_atlas); + if (shadow_atlas == nullptr || shadow_atlas->depth.is_null()) { u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK)); } else { - u.ids.push_back(shadow_atlas->shrink_stages[shadow_atlas->shrink_stages.size() - 1].texture); + u.ids.push_back(shadow_atlas->depth); } uniforms.push_back(u); @@ -6868,10 +3030,10 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.binding = 2; - if (directional_shadow.shrink_stages.size() == 0) { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK)); + if (directional_shadow.depth.is_valid()) { + u.ids.push_back(directional_shadow.depth); } else { - u.ids.push_back(directional_shadow.shrink_stages[directional_shadow.shrink_stages.size() - 1].texture); + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK)); } uniforms.push_back(u); } @@ -6880,23 +3042,22 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.binding = 3; - u.ids.push_back(get_positional_light_buffer()); + u.ids.push_back(get_omni_light_buffer()); uniforms.push_back(u); } - { RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.binding = 4; - u.ids.push_back(get_directional_light_buffer()); + u.ids.push_back(get_spot_light_buffer()); uniforms.push_back(u); } { RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; u.binding = 5; - u.ids.push_back(get_cluster_builder_texture()); + u.ids.push_back(get_directional_light_buffer()); uniforms.push_back(u); } @@ -6904,7 +3065,7 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.binding = 6; - u.ids.push_back(get_cluster_builder_indices_buffer()); + u.ids.push_back(rb->cluster_builder->get_cluster_buffer()); uniforms.push_back(u); } @@ -6952,8 +3113,8 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.binding = 12; - for (int i = 0; i < RenderBuffers::MAX_GIPROBES; i++) { - u.ids.push_back(rb->giprobe_textures[i]); + for (int i = 0; i < RendererSceneGIRD::MAX_GIPROBES; i++) { + u.ids.push_back(rb->gi.giprobe_textures[i]); } uniforms.push_back(u); } @@ -6964,6 +3125,20 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e u.ids.push_back(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_UNIFORM_BUFFER; + u.binding = 14; + u.ids.push_back(volumetric_fog.params_ubo); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 15; + u.ids.push_back(rb->volumetric_fog->prev_light_density_map); + uniforms.push_back(u); + } rb->volumetric_fog->uniform_set = RD::get_singleton()->uniform_set_create(uniforms, volumetric_fog.shader.version_get_shader(volumetric_fog.shader_version, 0), 0); @@ -7009,7 +3184,7 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e rb->volumetric_fog->length = env->volumetric_fog_length; rb->volumetric_fog->spread = env->volumetric_fog_detail_spread; - VolumetricFogShader::PushConstant push_constant; + 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(); @@ -7025,51 +3200,78 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e fog_near_size = Vector2(); } - push_constant.fog_frustum_size_begin[0] = fog_near_size.x; - push_constant.fog_frustum_size_begin[1] = fog_near_size.y; + params.fog_frustum_size_begin[0] = fog_near_size.x; + params.fog_frustum_size_begin[1] = fog_near_size.y; - push_constant.fog_frustum_size_end[0] = fog_far_size.x; - push_constant.fog_frustum_size_end[1] = fog_far_size.y; + params.fog_frustum_size_end[0] = fog_far_size.x; + params.fog_frustum_size_end[1] = fog_far_size.y; - push_constant.z_near = z_near; - push_constant.z_far = z_far; + params.z_near = z_near; + params.z_far = z_far; - push_constant.fog_frustum_end = fog_end; + params.fog_frustum_end = fog_end; - push_constant.fog_volume_size[0] = rb->volumetric_fog->width; - push_constant.fog_volume_size[1] = rb->volumetric_fog->height; - push_constant.fog_volume_size[2] = rb->volumetric_fog->depth; + params.fog_volume_size[0] = rb->volumetric_fog->width; + params.fog_volume_size[1] = rb->volumetric_fog->height; + params.fog_volume_size[2] = rb->volumetric_fog->depth; - push_constant.directional_light_count = p_directional_light_count; + params.directional_light_count = p_directional_light_count; Color light = env->volumetric_fog_light.to_linear(); - push_constant.light_energy[0] = light.r * env->volumetric_fog_light_energy; - push_constant.light_energy[1] = light.g * env->volumetric_fog_light_energy; - push_constant.light_energy[2] = light.b * env->volumetric_fog_light_energy; - push_constant.base_density = env->volumetric_fog_density; - - push_constant.detail_spread = env->volumetric_fog_detail_spread; - push_constant.gi_inject = env->volumetric_fog_gi_inject; - - push_constant.cam_rotation[0] = p_cam_transform.basis[0][0]; - push_constant.cam_rotation[1] = p_cam_transform.basis[1][0]; - push_constant.cam_rotation[2] = p_cam_transform.basis[2][0]; - push_constant.cam_rotation[3] = 0; - push_constant.cam_rotation[4] = p_cam_transform.basis[0][1]; - push_constant.cam_rotation[5] = p_cam_transform.basis[1][1]; - push_constant.cam_rotation[6] = p_cam_transform.basis[2][1]; - push_constant.cam_rotation[7] = 0; - push_constant.cam_rotation[8] = p_cam_transform.basis[0][2]; - push_constant.cam_rotation[9] = p_cam_transform.basis[1][2]; - push_constant.cam_rotation[10] = p_cam_transform.basis[2][2]; - push_constant.cam_rotation[11] = 0; - push_constant.filter_axis = 0; - push_constant.max_gi_probes = env->volumetric_fog_gi_inject > 0.001 ? p_gi_probe_count : 0; + params.light_energy[0] = light.r * env->volumetric_fog_light_energy; + params.light_energy[1] = light.g * env->volumetric_fog_light_energy; + params.light_energy[2] = light.b * env->volumetric_fog_light_energy; + params.base_density = env->volumetric_fog_density; + + params.detail_spread = env->volumetric_fog_detail_spread; + params.gi_inject = env->volumetric_fog_gi_inject; + + 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_gi_probes = env->volumetric_fog_gi_inject > 0.001 ? p_gi_probe_count : 0; + params.temporal_frame = RSG::rasterizer->get_frame_number() % VolumetricFog::MAX_TEMPORAL_FRAMES; + + Transform to_prev_cam_view = rb->volumetric_fog->prev_cam_transform.affine_inverse() * p_cam_transform; + storage->store_transform(to_prev_cam_view, params.to_prev_view); + + params.use_temporal_reprojection = env->volumetric_fog_temporal_reprojection; + params.temporal_blend = env->volumetric_fog_temporal_reprojection_amount; + + { + uint32_t cluster_size = rb->cluster_builder->get_cluster_size(); + params.cluster_shift = get_shift_from_power_of_2(cluster_size); + + uint32_t cluster_screen_width = (rb->width - 1) / cluster_size + 1; + uint32_t cluster_screen_height = (rb->height - 1) / cluster_size + 1; + params.cluster_type_size = cluster_screen_width * cluster_screen_height * (32 + 32); + params.cluster_width = cluster_screen_width; + params.max_cluster_element_count_div_32 = max_cluster_elements / 32; + + params.screen_size[0] = rb->width; + params.screen_size[1] = rb->height; + } /* Vector2 dssize = directional_shadow_get_size(); push_constant.directional_shadow_pixel_size[0] = 1.0 / dssize.x; push_constant.directional_shadow_pixel_size[1] = 1.0 / dssize.y; */ + + 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(); bool use_filter = volumetric_fog_filter_active; @@ -7077,93 +3279,225 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, volumetric_fog.pipelines[using_sdfgi ? VOLUMETRIC_FOG_SHADER_DENSITY_WITH_SDFGI : VOLUMETRIC_FOG_SHADER_DENSITY]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->volumetric_fog->uniform_set, 0); + if (using_sdfgi) { RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->volumetric_fog->sdfgi_uniform_set, 1); } - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(VolumetricFogShader::PushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth, 4, 4, 4); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth); + + RD::get_singleton()->draw_command_end_label(); - RD::get_singleton()->compute_list_add_barrier(compute_list); + RD::get_singleton()->compute_list_end(); + + RD::get_singleton()->texture_copy(rb->volumetric_fog->light_density_map, rb->volumetric_fog->prev_light_density_map, Vector3(0, 0, 0), Vector3(0, 0, 0), Vector3(rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth), 0, 0, 0, 0); + + compute_list = RD::get_singleton()->compute_list_begin(); if (use_filter) { + 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.pipelines[VOLUMETRIC_FOG_SHADER_FILTER]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->volumetric_fog->uniform_set, 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(VolumetricFogShader::PushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth); - RD::get_singleton()->compute_list_add_barrier(compute_list); + RD::get_singleton()->compute_list_end(); + //need restart for buffer update - push_constant.filter_axis = 1; + 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.pipelines[VOLUMETRIC_FOG_SHADER_FILTER]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->volumetric_fog->uniform_set2, 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(VolumetricFogShader::PushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth, 8, 8, 1); + if (using_sdfgi) { + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->volumetric_fog->sdfgi_uniform_set, 1); + } + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_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.pipelines[VOLUMETRIC_FOG_SHADER_FOG]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->volumetric_fog->uniform_set, 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(VolumetricFogShader::PushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, 1, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, 1); - RD::get_singleton()->compute_list_end(); + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_RASTER); + + RENDER_TIMESTAMP("<Volumetric Fog"); + RD::get_singleton()->draw_command_end_label(); + + rb->volumetric_fog->prev_cam_transform = p_cam_transform; } -void RendererSceneRenderRD::render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold) { - Color clear_color; - if (p_render_buffers.is_valid()) { - RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); - ERR_FAIL_COND(!rb); - clear_color = storage->render_target_get_clear_request_color(rb->render_target); - } else { - clear_color = storage->get_default_clear_color(); +uint32_t RendererSceneRenderRD::_get_render_state_directional_light_count() const { + return render_state.directional_light_count; +} + +bool RendererSceneRenderRD::_needs_post_prepass_render(bool p_use_gi) { + if (render_state.render_buffers.is_valid()) { + RenderBuffers *rb = render_buffers_owner.getornull(render_state.render_buffers); + if (rb->sdfgi != nullptr) { + return true; + } } + return false; +} - //assign render indices to giprobes - for (uint32_t i = 0; i < (uint32_t)p_gi_probes.size(); i++) { - GIProbeInstance *giprobe_inst = gi_probe_instance_owner.getornull(p_gi_probes[i]); - if (giprobe_inst) { - giprobe_inst->render_index = i; +void RendererSceneRenderRD::_post_prepass_render(bool p_use_gi) { + if (render_state.render_buffers.is_valid()) { + if (p_use_gi) { + RenderBuffers *rb = render_buffers_owner.getornull(render_state.render_buffers); + ERR_FAIL_COND(rb == nullptr); + if (rb->sdfgi == nullptr) { + return; + } + + RendererSceneEnvironmentRD *env = environment_owner.getornull(render_state.environment); + rb->sdfgi->update_probes(env, sky.sky_owner.getornull(env->sky)); } } +} - const PagedArray<RID> *lights = &p_lights; - const PagedArray<RID> *reflections = &p_reflection_probes; - const PagedArray<RID> *gi_probes = &p_gi_probes; +void RendererSceneRenderRD::_pre_resolve_render(bool p_use_gi) { + if (render_state.render_buffers.is_valid()) { + if (p_use_gi) { + RD::get_singleton()->compute_list_end(); + } + } +} - PagedArray<RID> empty; +void RendererSceneRenderRD::_pre_opaque_render(bool p_use_ssao, bool p_use_gi, RID p_normal_roughness_buffer, RID p_gi_probe_buffer) { + // Render shadows while GI is rendering, due to how barriers are handled, this should happen at the same time - if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_UNSHADED) { - lights = ∅ - reflections = ∅ - gi_probes = ∅ + if (render_state.render_buffers.is_valid() && p_use_gi) { + RenderBuffers *rb = render_buffers_owner.getornull(render_state.render_buffers); + ERR_FAIL_COND(rb == nullptr); + if (rb->sdfgi == nullptr) { + return; + } + + rb->sdfgi->store_probes(); } - cluster.builder.begin(p_cam_transform.affine_inverse(), p_cam_projection); //prepare cluster + render_state.cube_shadows.clear(); + render_state.shadows.clear(); + render_state.directional_shadows.clear(); + + Plane camera_plane(render_state.cam_transform.origin, -render_state.cam_transform.basis.get_axis(Vector3::AXIS_Z)); + float lod_distance_multiplier = render_state.cam_projection.get_lod_multiplier(); + + { + for (int i = 0; i < render_state.render_shadow_count; i++) { + LightInstance *li = light_instance_owner.getornull(render_state.render_shadows[i].light); + + if (storage->light_get_type(li->light) == RS::LIGHT_DIRECTIONAL) { + render_state.directional_shadows.push_back(i); + } else if (storage->light_get_type(li->light) == RS::LIGHT_OMNI && storage->light_omni_get_shadow_mode(li->light) == RS::LIGHT_OMNI_SHADOW_CUBE) { + render_state.cube_shadows.push_back(i); + } else { + render_state.shadows.push_back(i); + } + } + + //cube shadows are rendered in their own way + for (uint32_t i = 0; i < render_state.cube_shadows.size(); i++) { + _render_shadow_pass(render_state.render_shadows[render_state.cube_shadows[i]].light, render_state.shadow_atlas, render_state.render_shadows[render_state.cube_shadows[i]].pass, render_state.render_shadows[render_state.cube_shadows[i]].instances, camera_plane, lod_distance_multiplier, render_state.screen_lod_threshold, true, true, true); + } + + if (render_state.directional_shadows.size()) { + //open the pass for directional shadows + _update_directional_shadow_atlas(); + RD::get_singleton()->draw_list_begin(directional_shadow.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 = render_state.directional_shadows.size() || render_state.shadows.size(); + bool render_gi = render_state.render_buffers.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 < render_state.directional_shadows.size(); i++) { + _render_shadow_pass(render_state.render_shadows[render_state.directional_shadows[i]].light, render_state.shadow_atlas, render_state.render_shadows[render_state.directional_shadows[i]].pass, render_state.render_shadows[render_state.directional_shadows[i]].instances, camera_plane, lod_distance_multiplier, render_state.screen_lod_threshold, false, i == render_state.directional_shadows.size() - 1, false); + } + //render positional shadows + for (uint32_t i = 0; i < render_state.shadows.size(); i++) { + _render_shadow_pass(render_state.render_shadows[render_state.shadows[i]].light, render_state.shadow_atlas, render_state.render_shadows[render_state.shadows[i]].pass, render_state.render_shadows[render_state.shadows[i]].instances, camera_plane, lod_distance_multiplier, render_state.screen_lod_threshold, i == 0, i == render_state.shadows.size() - 1, true); + } + + _render_shadow_process(); + } + + //start GI + if (render_gi) { + gi.process_gi(render_state.render_buffers, p_normal_roughness_buffer, p_gi_probe_buffer, render_state.environment, render_state.cam_projection, render_state.cam_transform, *render_state.gi_probes, this); + } + + //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 (render_state.render_buffers.is_valid()) { + if (p_use_ssao) { + _process_ssao(render_state.render_buffers, render_state.environment, p_normal_roughness_buffer, render_state.cam_projection); + } + } + + //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(render_state.cam_transform, render_state.cam_projection, !render_state.reflection_probe.is_valid()); + } bool using_shadows = true; - if (p_reflection_probe.is_valid()) { - if (!storage->reflection_probe_renders_shadows(reflection_probe_instance_get_probe(p_reflection_probe))) { + if (render_state.reflection_probe.is_valid()) { + if (!storage->reflection_probe_renders_shadows(reflection_probe_instance_get_probe(render_state.reflection_probe))) { using_shadows = false; } } else { //do not render reflections when rendering a reflection probe - _setup_reflections(*reflections, p_cam_transform.affine_inverse(), p_environment); + _setup_reflections(*render_state.reflection_probes, render_state.cam_transform.affine_inverse(), render_state.environment); } uint32_t directional_light_count = 0; uint32_t positional_light_count = 0; - _setup_lights(*lights, p_cam_transform.affine_inverse(), p_shadow_atlas, using_shadows, directional_light_count, positional_light_count); - _setup_decals(p_decals, p_cam_transform.affine_inverse()); - cluster.builder.bake_cluster(); //bake to cluster + _setup_lights(*render_state.lights, render_state.cam_transform, render_state.shadow_atlas, using_shadows, directional_light_count, positional_light_count); + _setup_decals(*render_state.decals, render_state.cam_transform.affine_inverse()); - uint32_t gi_probe_count = 0; - _setup_giprobes(p_render_buffers, p_cam_transform, *gi_probes, gi_probe_count); + render_state.directional_light_count = directional_light_count; - if (p_render_buffers.is_valid()) { + if (current_cluster_builder) { + current_cluster_builder->bake_cluster(); + } + + if (render_state.render_buffers.is_valid()) { bool directional_shadows = false; for (uint32_t i = 0; i < directional_light_count; i++) { if (cluster.directional_lights[i].shadow_enabled) { @@ -7171,43 +3505,164 @@ void RendererSceneRenderRD::render_scene(RID p_render_buffers, const Transform & break; } } - _update_volumetric_fog(p_render_buffers, p_environment, p_cam_projection, p_cam_transform, p_shadow_atlas, directional_light_count, directional_shadows, positional_light_count, gi_probe_count); + _update_volumetric_fog(render_state.render_buffers, render_state.environment, render_state.cam_projection, render_state.cam_transform, render_state.shadow_atlas, directional_light_count, directional_shadows, positional_light_count, render_state.gi_probe_count); + } +} + +void RendererSceneRenderRD::render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_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) { + // getting this here now so we can direct call a bunch of things more easily + RenderBuffers *rb = nullptr; + if (p_render_buffers.is_valid()) { + rb = render_buffers_owner.getornull(p_render_buffers); + ERR_FAIL_COND(!rb); // !BAS! Do we fail here or skip the parts that won't work. can't really see a case why we would be rendering without buffers.... + } + + //assign render data + { + render_state.render_buffers = p_render_buffers; + render_state.cam_transform = p_cam_transform; + render_state.cam_projection = p_cam_projection; + render_state.cam_ortogonal = p_cam_projection.is_orthogonal(); + render_state.instances = &p_instances; + render_state.lights = &p_lights; + render_state.reflection_probes = &p_reflection_probes; + render_state.gi_probes = &p_gi_probes; + render_state.decals = &p_decals; + render_state.lightmaps = &p_lightmaps; + render_state.environment = p_environment; + render_state.camera_effects = p_camera_effects; + render_state.shadow_atlas = p_shadow_atlas; + render_state.reflection_atlas = p_reflection_atlas; + render_state.reflection_probe = p_reflection_probe; + render_state.reflection_probe_pass = p_reflection_probe_pass; + render_state.screen_lod_threshold = p_screen_lod_threshold; + + render_state.render_shadows = p_render_shadows; + render_state.render_shadow_count = p_render_shadow_count; + render_state.render_sdfgi_regions = p_render_sdfgi_regions; + render_state.render_sdfgi_region_count = p_render_sdfgi_region_count; + render_state.sdfgi_update_data = p_sdfgi_update_data; + } + + PagedArray<RID> empty; + + if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_UNSHADED) { + render_state.lights = ∅ + render_state.reflection_probes = ∅ + render_state.gi_probes = ∅ + } + + //sdfgi first + if (rb != nullptr && rb->sdfgi != nullptr) { + for (int i = 0; i < render_state.render_sdfgi_region_count; i++) { + rb->sdfgi->render_region(p_render_buffers, render_state.render_sdfgi_regions[i].region, render_state.render_sdfgi_regions[i].instances, this); + } + if (render_state.sdfgi_update_data->update_static) { + rb->sdfgi->render_static_lights(p_render_buffers, render_state.sdfgi_update_data->static_cascade_count, p_sdfgi_update_data->static_cascade_indices, render_state.sdfgi_update_data->static_positional_lights, this); + } + } + + Color clear_color; + if (p_render_buffers.is_valid()) { + clear_color = storage->render_target_get_clear_request_color(rb->render_target); + } else { + clear_color = storage->get_default_clear_color(); + } + + //assign render indices to giprobes + for (uint32_t i = 0; i < (uint32_t)p_gi_probes.size(); i++) { + RendererSceneGIRD::GIProbeInstance *giprobe_inst = gi.gi_probe_instance_owner.getornull(p_gi_probes[i]); + if (giprobe_inst) { + giprobe_inst->render_index = i; + } + } + + if (render_buffers_owner.owns(render_state.render_buffers)) { + RenderBuffers *rs_rb = render_buffers_owner.getornull(render_state.render_buffers); + current_cluster_builder = rs_rb->cluster_builder; + } else if (reflection_probe_instance_owner.owns(render_state.reflection_probe)) { + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(render_state.reflection_probe); + ReflectionAtlas *ra = reflection_atlas_owner.getornull(rpi->atlas); + if (!ra) { + ERR_PRINT("reflection probe has no reflection atlas! Bug?"); + current_cluster_builder = nullptr; + } else { + current_cluster_builder = ra->cluster_builder; + } + } else { + ERR_PRINT("No cluster builder, bug"); //should never happen, will crash + current_cluster_builder = nullptr; } - _render_scene(p_render_buffers, p_cam_transform, p_cam_projection, p_cam_ortogonal, p_instances, directional_light_count, *gi_probes, p_lightmaps, p_environment, p_camera_effects, p_shadow_atlas, p_reflection_atlas, p_reflection_probe, p_reflection_probe_pass, clear_color, p_screen_lod_threshold); + if (rb != nullptr && rb->sdfgi != nullptr) { + rb->sdfgi->update_cascades(); + + rb->sdfgi->pre_process_gi(p_cam_transform, this); + } + + render_state.gi_probe_count = 0; + if (rb != nullptr && rb->sdfgi != nullptr) { + gi.setup_giprobes(render_state.render_buffers, render_state.cam_transform, *render_state.gi_probes, render_state.gi_probe_count, this); + + rb->sdfgi->update_light(); + } + + render_state.depth_prepass_used = false; + //calls _pre_opaque_render between depth pre-pass and opaque pass + _render_scene(p_render_buffers, p_cam_transform, p_cam_projection, p_cam_ortogonal, p_instances, *render_state.gi_probes, p_lightmaps, p_environment, current_cluster_builder->get_cluster_buffer(), current_cluster_builder->get_cluster_size(), current_cluster_builder->get_max_cluster_elements(), p_camera_effects, p_shadow_atlas, p_reflection_atlas, p_reflection_probe, p_reflection_probe_pass, clear_color, p_screen_lod_threshold); if (p_render_buffers.is_valid()) { + if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_OMNI_LIGHTS || debug_draw == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_SPOT_LIGHTS || debug_draw == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_DECALS || debug_draw == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_REFLECTION_PROBES) { + ClusterBuilderRD::ElementType elem_type = ClusterBuilderRD::ELEMENT_TYPE_MAX; + switch (debug_draw) { + 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); + } + RENDER_TIMESTAMP("Tonemap"); _render_buffers_post_process_and_tonemap(p_render_buffers, p_environment, p_camera_effects, p_cam_projection); _render_buffers_debug_draw(p_render_buffers, p_shadow_atlas); - if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_SDFGI) { - _sdfgi_debug_draw(p_render_buffers, p_cam_projection, p_cam_transform); + if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_SDFGI && rb != nullptr && rb->sdfgi != nullptr) { + rb->sdfgi->debug_draw(p_cam_projection, p_cam_transform, rb->width, rb->height, rb->render_target, rb->texture); } } } -void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<GeometryInstance *> &p_instances, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold) { +void RendererSceneRenderRD::_render_shadow_pass(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<GeometryInstance *> &p_instances, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold, bool p_open_pass, bool p_close_pass, bool p_clear_region) { LightInstance *light_instance = light_instance_owner.getornull(p_light); ERR_FAIL_COND(!light_instance); Rect2i atlas_rect; - RID atlas_texture; + uint32_t atlas_size; + RID atlas_fb; bool using_dual_paraboloid = false; bool using_dual_paraboloid_flip = false; - float znear = 0; - float zfar = 0; RID render_fb; RID render_texture; - float bias = 0; - float normal_bias = 0; + float zfar; bool use_pancake = false; - bool use_linear_depth = false; bool render_cubemap = false; bool finalize_cubemap = false; + bool flip_y = false; + CameraMatrix light_projection; Transform light_transform; @@ -7240,7 +3695,6 @@ void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p atlas_rect.position.x += atlas_rect.size.width; atlas_rect.position.y += atlas_rect.size.height; } - } else if (storage->light_directional_get_shadow_mode(light_instance->light) == RS::LIGHT_DIRECTIONAL_SHADOW_PARALLEL_2_SPLITS) { atlas_rect.size.height /= 2; @@ -7255,15 +3709,11 @@ void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p light_instance->shadow_transform[p_pass].atlas_rect.position /= directional_shadow.size; light_instance->shadow_transform[p_pass].atlas_rect.size /= directional_shadow.size; - float bias_mult = light_instance->shadow_transform[p_pass].bias_scale; zfar = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_RANGE); - bias = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_SHADOW_BIAS) * bias_mult; - normal_bias = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * bias_mult; - ShadowMap *shadow_map = _get_shadow_map(atlas_rect.size); - render_fb = shadow_map->fb; - render_texture = shadow_map->depth; - atlas_texture = directional_shadow.depth; + render_fb = directional_shadow.fb; + render_texture = RID(); + flip_y = true; } else { //set from shadow atlas @@ -7272,6 +3722,8 @@ void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p ERR_FAIL_COND(!shadow_atlas); ERR_FAIL_COND(!shadow_atlas->shadow_owners.has(p_light)); + _update_shadow_atlas(shadow_atlas); + uint32_t key = shadow_atlas->shadow_owners[p_light]; uint32_t quadrant = (key >> ShadowAtlas::QUADRANT_SHIFT) & 0x3; @@ -7290,11 +3742,8 @@ void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p atlas_rect.size.width = shadow_size; atlas_rect.size.height = shadow_size; - atlas_texture = shadow_atlas->depth; zfar = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_RANGE); - bias = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_SHADOW_BIAS); - normal_bias = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS); if (storage->light_get_type(light_instance->light) == RS::LIGHT_OMNI) { if (storage->light_omni_get_shadow_mode(light_instance->light) == RS::LIGHT_OMNI_SHADOW_CUBE) { @@ -7303,10 +3752,17 @@ void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p render_fb = cubemap->side_fb[p_pass]; render_texture = cubemap->cubemap; - light_projection = light_instance->shadow_transform[0].camera; - light_transform = light_instance->shadow_transform[0].transform; + light_projection = light_instance->shadow_transform[p_pass].camera; + light_transform = light_instance->shadow_transform[p_pass].transform; render_cubemap = true; finalize_cubemap = p_pass == 5; + atlas_fb = shadow_atlas->fb; + + atlas_size = shadow_atlas->size; + + if (p_pass == 0) { + _render_shadow_begin(); + } } else { light_projection = light_instance->shadow_transform[0].camera; @@ -7317,49 +3773,44 @@ void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p using_dual_paraboloid = true; using_dual_paraboloid_flip = p_pass == 1; - - ShadowMap *shadow_map = _get_shadow_map(atlas_rect.size); - render_fb = shadow_map->fb; - render_texture = shadow_map->depth; + render_fb = shadow_atlas->fb; + flip_y = true; } } else if (storage->light_get_type(light_instance->light) == RS::LIGHT_SPOT) { light_projection = light_instance->shadow_transform[0].camera; light_transform = light_instance->shadow_transform[0].transform; - ShadowMap *shadow_map = _get_shadow_map(atlas_rect.size); - render_fb = shadow_map->fb; - render_texture = shadow_map->depth; + render_fb = shadow_atlas->fb; - znear = light_instance->shadow_transform[0].camera.get_z_near(); - use_linear_depth = true; + flip_y = true; } } if (render_cubemap) { //rendering to cubemap - _render_shadow(render_fb, p_instances, light_projection, light_transform, zfar, 0, 0, false, false, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold); + _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_lod_threshold, Rect2(), false, true, true, true); if (finalize_cubemap) { + _render_shadow_process(); + _render_shadow_end(); //reblit - atlas_rect.size.height /= 2; - storage->get_effects()->copy_cubemap_to_dp(render_texture, atlas_texture, atlas_rect, light_projection.get_z_near(), light_projection.get_z_far(), 0.0, false); - atlas_rect.position.y += atlas_rect.size.height; - storage->get_effects()->copy_cubemap_to_dp(render_texture, atlas_texture, atlas_rect, light_projection.get_z_near(), light_projection.get_z_far(), 0.0, true); - } - } else { - //render shadow - - _render_shadow(render_fb, p_instances, light_projection, light_transform, zfar, bias, normal_bias, using_dual_paraboloid, using_dual_paraboloid_flip, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold); + Rect2 atlas_rect_norm = atlas_rect; + atlas_rect_norm.position.x /= float(atlas_size); + atlas_rect_norm.position.y /= float(atlas_size); + atlas_rect_norm.size.x /= float(atlas_size); + atlas_rect_norm.size.y /= float(atlas_size); + atlas_rect_norm.size.height /= 2; + storage->get_effects()->copy_cubemap_to_dp(render_texture, atlas_fb, atlas_rect_norm, light_projection.get_z_near(), light_projection.get_z_far(), false); + atlas_rect_norm.position.y += atlas_rect_norm.size.height; + storage->get_effects()->copy_cubemap_to_dp(render_texture, atlas_fb, atlas_rect_norm, light_projection.get_z_near(), light_projection.get_z_far(), true); - //copy to atlas - if (use_linear_depth) { - storage->get_effects()->copy_depth_to_rect_and_linearize(render_texture, atlas_texture, atlas_rect, true, znear, zfar); - } else { - storage->get_effects()->copy_depth_to_rect(render_texture, atlas_texture, atlas_rect, true); + //restore transform so it can be properly used + light_instance_set_shadow_transform(p_light, CameraMatrix(), light_instance->transform, zfar, 0, 0, 0); } - //does not work from depth to color - //RD::get_singleton()->texture_copy(render_texture, atlas_texture, Vector3(0, 0, 0), Vector3(atlas_rect.position.x, atlas_rect.position.y, 0), Vector3(atlas_rect.size.x, atlas_rect.size.y, 1), 0, 0, 0, 0, true); + } 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_lod_threshold, atlas_rect, flip_y, p_clear_region, p_open_pass, p_close_pass); } } @@ -7367,343 +3818,6 @@ void RendererSceneRenderRD::render_material(const Transform &p_cam_transform, co _render_material(p_cam_transform, p_cam_projection, p_cam_ortogonal, p_instances, p_framebuffer, p_region); } -void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, const PagedArray<GeometryInstance *> &p_instances) { - //print_line("rendering region " + itos(p_region)); - RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); - ERR_FAIL_COND(!rb); - ERR_FAIL_COND(!rb->sdfgi); - AABB bounds; - Vector3i from; - Vector3i size; - - int cascade_prev = _sdfgi_get_pending_region_data(p_render_buffers, p_region - 1, from, size, bounds); - int cascade_next = _sdfgi_get_pending_region_data(p_render_buffers, p_region + 1, from, size, bounds); - int cascade = _sdfgi_get_pending_region_data(p_render_buffers, p_region, from, size, bounds); - ERR_FAIL_COND(cascade < 0); - - if (cascade_prev != cascade) { - //initialize render - RD::get_singleton()->texture_clear(rb->sdfgi->render_albedo, Color(0, 0, 0, 0), 0, 1, 0, 1, true); - RD::get_singleton()->texture_clear(rb->sdfgi->render_emission, Color(0, 0, 0, 0), 0, 1, 0, 1, true); - RD::get_singleton()->texture_clear(rb->sdfgi->render_emission_aniso, Color(0, 0, 0, 0), 0, 1, 0, 1, true); - RD::get_singleton()->texture_clear(rb->sdfgi->render_geom_facing, Color(0, 0, 0, 0), 0, 1, 0, 1, true); - } - - //print_line("rendering cascade " + itos(p_region) + " objects: " + itos(p_cull_count) + " bounds: " + bounds + " from: " + from + " size: " + size + " cell size: " + rtos(rb->sdfgi->cascades[cascade].cell_size)); - _render_sdfgi(p_render_buffers, from, size, bounds, p_instances, rb->sdfgi->render_albedo, rb->sdfgi->render_emission, rb->sdfgi->render_emission_aniso, rb->sdfgi->render_geom_facing); - - if (cascade_next != cascade) { - RENDER_TIMESTAMP(">SDFGI Update SDF"); - //done rendering! must update SDF - //clear dispatch indirect data - - SDGIShader::PreprocessPushConstant push_constant; - zeromem(&push_constant, sizeof(SDGIShader::PreprocessPushConstant)); - - RENDER_TIMESTAMP("Scroll SDF"); - - //scroll - if (rb->sdfgi->cascades[cascade].dirty_regions != SDFGI::Cascade::DIRTY_ALL) { - //for scroll - Vector3i dirty = rb->sdfgi->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; - } - push_constant.grid_size = rb->sdfgi->cascade_size; - push_constant.cascade = cascade; - - if (rb->sdfgi->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, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_SCROLL]); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->cascades[cascade].scroll_uniform_set, 0); - - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant)); - RD::get_singleton()->compute_list_dispatch_indirect(compute_list, rb->sdfgi->cascades[cascade].solid_cell_dispatch_buffer, 0); - // no barrier do all together - - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_SCROLL_OCCLUSION]); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->cascades[cascade].scroll_occlusion_uniform_set, 0); - - Vector3i dirty = rb->sdfgi->cascades[cascade].dirty_regions; - Vector3i groups; - groups.x = rb->sdfgi->cascade_size - ABS(dirty.x); - groups.y = rb->sdfgi->cascade_size - ABS(dirty.y); - groups.z = rb->sdfgi->cascade_size - ABS(dirty.z); - - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, groups.x, groups.y, groups.z, 4, 4, 4); - - //no barrier, continue together - - { - //scroll probes and their history also - - SDGIShader::IntegratePushConstant ipush_constant; - ipush_constant.grid_size[1] = rb->sdfgi->cascade_size; - ipush_constant.grid_size[2] = rb->sdfgi->cascade_size; - ipush_constant.grid_size[0] = rb->sdfgi->cascade_size; - ipush_constant.max_cascades = rb->sdfgi->cascades.size(); - ipush_constant.probe_axis_size = rb->sdfgi->probe_axis_count; - ipush_constant.history_index = 0; - ipush_constant.history_size = rb->sdfgi->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 = rb->sdfgi->y_mult; - ipush_constant.store_ambient_texture = false; - - ipush_constant.image_size[0] = rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count; - ipush_constant.image_size[1] = rb->sdfgi->probe_axis_count; - - int32_t probe_divisor = rb->sdfgi->cascade_size / SDFGI::PROBE_DIVISOR; - ipush_constant.cascade = cascade; - ipush_constant.world_offset[0] = rb->sdfgi->cascades[cascade].position.x / probe_divisor; - ipush_constant.world_offset[1] = rb->sdfgi->cascades[cascade].position.y / probe_divisor; - ipush_constant.world_offset[2] = rb->sdfgi->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, sdfgi_shader.integrate_pipeline[SDGIShader::INTEGRATE_MODE_SCROLL]); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->cascades[cascade].integrate_uniform_set, 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, sdfgi_shader.integrate_default_sky_uniform_set, 1); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &ipush_constant, sizeof(SDGIShader::IntegratePushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count, rb->sdfgi->probe_axis_count, 1, 8, 8, 1); - - RD::get_singleton()->compute_list_add_barrier(compute_list); - - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.integrate_pipeline[SDGIShader::INTEGRATE_MODE_SCROLL_STORE]); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->cascades[cascade].integrate_uniform_set, 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, sdfgi_shader.integrate_default_sky_uniform_set, 1); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &ipush_constant, sizeof(SDGIShader::IntegratePushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count, rb->sdfgi->probe_axis_count, 1, 8, 8, 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(rb->sdfgi->cascades[cascade].solid_cell_dispatch_buffer, 0, sizeof(uint32_t) * 4, dispatch_indirct_data, true); - - 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 = rb->sdfgi->cascade_size >> 1; - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_JUMP_FLOOD_INITIALIZE_HALF]); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->sdf_initialize_half_uniform_set, 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, cascade_half_size, cascade_half_size, cascade_half_size, 4, 4, 4); - 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, sdfgi_shader.preprocess_pipeline[SDGIShader::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, rb->sdfgi->jump_flood_half_uniform_set[jf_us], 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, cascade_half_size, cascade_half_size, cascade_half_size, 4, 4, 4); - 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, sdfgi_shader.preprocess_pipeline[SDGIShader::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, rb->sdfgi->jump_flood_half_uniform_set[jf_us], 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, cascade_half_size, cascade_half_size, cascade_half_size, optimized_jf_group_size, optimized_jf_group_size, optimized_jf_group_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 = rb->sdfgi->cascade_size; - - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_JUMP_FLOOD_UPSCALE]); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->sdf_upscale_uniform_set, 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, 4, 4, 4); - 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, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_JUMP_FLOOD_OPTIMIZED]); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->jump_flood_uniform_set[rb->sdfgi->upscale_jfa_uniform_set_index], 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, optimized_jf_group_size, optimized_jf_group_size, optimized_jf_group_size); - RD::get_singleton()->compute_list_add_barrier(compute_list); - - } else { - //full size jumpflood - RENDER_TIMESTAMP("SDFGI Jump Flood"); - - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_JUMP_FLOOD_INITIALIZE]); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->sdf_initialize_uniform_set, 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, 4, 4, 4); - - RD::get_singleton()->compute_list_add_barrier(compute_list); - - push_constant.half_size = false; - { - uint32_t s = rb->sdfgi->cascade_size; - - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.preprocess_pipeline[SDGIShader::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, rb->sdfgi->jump_flood_uniform_set[jf_us], 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, 4, 4, 4); - RD::get_singleton()->compute_list_add_barrier(compute_list); - jf_us = jf_us == 0 ? 1 : 0; - - if (rb->sdfgi->cascade_size / (s / 2) >= optimized_jf_group_size) { - break; - } - } - - RENDER_TIMESTAMP("SDFGI Jump Flood Optimized"); - - //continue with optimized jump flood for smaller reads - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.preprocess_pipeline[SDGIShader::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, rb->sdfgi->jump_flood_uniform_set[jf_us], 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, optimized_jf_group_size, optimized_jf_group_size, optimized_jf_group_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 = rb->sdfgi->cascade_size / SDFGI::PROBE_DIVISOR; - Vector3i probe_global_pos = rb->sdfgi->cascades[cascade].position / probe_size; - - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_OCCLUSION]); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->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(SDGIShader::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, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_STORE]); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->cascades[cascade].sdf_store_uniform_set, 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, 4, 4, 4); - - RD::get_singleton()->compute_list_end(); - - //clear these textures, as they will have previous garbage on next draw - RD::get_singleton()->texture_clear(rb->sdfgi->cascades[cascade].light_tex, Color(0, 0, 0, 0), 0, 1, 0, 1, true); - RD::get_singleton()->texture_clear(rb->sdfgi->cascades[cascade].light_aniso_0_tex, Color(0, 0, 0, 0), 0, 1, 0, 1, true); - RD::get_singleton()->texture_clear(rb->sdfgi->cascades[cascade].light_aniso_1_tex, Color(0, 0, 0, 0), 0, 1, 0, 1, true); - -#if 0 - Vector<uint8_t> data = RD::get_singleton()->texture_get_data(rb->sdfgi->cascades[cascade].sdf, 0); - Ref<Image> img; - img.instance(); - for (uint32_t i = 0; i < rb->sdfgi->cascade_size; i++) { - Vector<uint8_t> subarr = data.subarray(128 * 128 * i, 128 * 128 * (i + 1) - 1); - img->create(rb->sdfgi->cascade_size, rb->sdfgi->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(rb->sdfgi->render_albedo, 0); - Ref<Image> img; - img.instance(); - for (uint32_t i = 0; i < rb->sdfgi->cascade_size; i++) { - Vector<uint8_t> subarr = data.subarray(128 * 128 * i * 2, 128 * 128 * (i + 1) * 2 - 1); - img->create(rb->sdfgi->cascade_size, rb->sdfgi->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"); - } -} - void RendererSceneRenderRD::render_particle_collider_heightfield(RID p_collider, const Transform &p_transform, const PagedArray<GeometryInstance *> &p_instances) { ERR_FAIL_COND(!storage->particles_collision_is_heightfield(p_collider)); Vector3 extents = storage->particles_collision_get_extents(p_collider) * p_transform.basis.get_scale(); @@ -7721,122 +3835,22 @@ void RendererSceneRenderRD::render_particle_collider_heightfield(RID p_collider, _render_particle_collider_heightfield(fb, cam_xform, cm, p_instances); } -void RendererSceneRenderRD::render_sdfgi_static_lights(RID p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const PagedArray<RID> *p_positional_light_cull_result) { - RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); - ERR_FAIL_COND(!rb); - ERR_FAIL_COND(!rb->sdfgi); - - _sdfgi_update_cascades(p_render_buffers); //need cascades updated for this - - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.direct_light_pipeline[SDGIShader::DIRECT_LIGHT_MODE_STATIC]); - - SDGIShader::DirectLightPushConstant dl_push_constant; - - dl_push_constant.grid_size[0] = rb->sdfgi->cascade_size; - dl_push_constant.grid_size[1] = rb->sdfgi->cascade_size; - dl_push_constant.grid_size[2] = rb->sdfgi->cascade_size; - dl_push_constant.max_cascades = rb->sdfgi->cascades.size(); - dl_push_constant.probe_axis_size = rb->sdfgi->probe_axis_count; - dl_push_constant.multibounce = false; // this is static light, do not multibounce yet - dl_push_constant.y_mult = rb->sdfgi->y_mult; - - //all must be processed - dl_push_constant.process_offset = 0; - dl_push_constant.process_increment = 1; - - SDGIShader::Light lights[SDFGI::MAX_STATIC_LIGHTS]; - - for (uint32_t i = 0; i < p_cascade_count; i++) { - ERR_CONTINUE(p_cascade_indices[i] >= rb->sdfgi->cascades.size()); - - SDFGI::Cascade &cc = rb->sdfgi->cascades[p_cascade_indices[i]]; - - { //fill light buffer - - AABB cascade_aabb; - cascade_aabb.position = Vector3((Vector3i(1, 1, 1) * -int32_t(rb->sdfgi->cascade_size >> 1) + cc.position)) * cc.cell_size; - cascade_aabb.size = Vector3(1, 1, 1) * rb->sdfgi->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; - } - - LightInstance *li = light_instance_owner.getornull(p_positional_light_cull_result[i][j]); - ERR_CONTINUE(!li); - - uint32_t max_sdfgi_cascade = storage->light_get_max_sdfgi_cascade(li->light); - if (p_cascade_indices[i] > max_sdfgi_cascade) { - continue; - } - - if (!cascade_aabb.intersects(li->aabb)) { - continue; - } - - lights[idx].type = storage->light_get_type(li->light); - - Vector3 dir = -li->transform.basis.get_axis(Vector3::AXIS_Z); - if (lights[idx].type == RS::LIGHT_DIRECTIONAL) { - dir.y *= rb->sdfgi->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 = li->transform.origin; - pos.y *= rb->sdfgi->y_mult; - lights[idx].position[0] = pos.x; - lights[idx].position[1] = pos.y; - lights[idx].position[2] = pos.z; - Color color = storage->light_get_color(li->light); - color = color.to_linear(); - lights[idx].color[0] = color.r; - lights[idx].color[1] = color.g; - lights[idx].color[2] = color.b; - lights[idx].energy = storage->light_get_param(li->light, RS::LIGHT_PARAM_ENERGY); - lights[idx].has_shadow = storage->light_has_shadow(li->light); - lights[idx].attenuation = storage->light_get_param(li->light, RS::LIGHT_PARAM_ATTENUATION); - lights[idx].radius = storage->light_get_param(li->light, RS::LIGHT_PARAM_RANGE); - lights[idx].spot_angle = Math::deg2rad(storage->light_get_param(li->light, RS::LIGHT_PARAM_SPOT_ANGLE)); - lights[idx].spot_attenuation = storage->light_get_param(li->light, RS::LIGHT_PARAM_SPOT_ATTENUATION); - - idx++; - } - - if (idx > 0) { - RD::get_singleton()->buffer_update(cc.lights_buffer, 0, idx * sizeof(SDGIShader::Light), lights, true); - } - dl_push_constant.light_count = idx; - } - - 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_uniform_set, 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &dl_push_constant, sizeof(SDGIShader::DirectLightPushConstant)); - RD::get_singleton()->compute_list_dispatch_indirect(compute_list, cc.solid_cell_dispatch_buffer, 0); - } - } - - RD::get_singleton()->compute_list_end(); -} - bool RendererSceneRenderRD::free(RID p_rid) { if (render_buffers_owner.owns(p_rid)) { RenderBuffers *rb = render_buffers_owner.getornull(p_rid); _free_render_buffer_data(rb); memdelete(rb->data); if (rb->sdfgi) { - _sdfgi_erase(rb); + rb->sdfgi->erase(); + memdelete(rb->sdfgi); + rb->sdfgi = nullptr; } if (rb->volumetric_fog) { _volumetric_fog_erase(rb); } + if (rb->cluster_builder) { + memdelete(rb->cluster_builder); + } render_buffers_owner.free(p_rid); } else if (environment_owner.owns(p_rid)) { //not much to delete, just free it @@ -7846,6 +3860,10 @@ bool RendererSceneRenderRD::free(RID p_rid) { camera_effects_owner.free(p_rid); } else if (reflection_atlas_owner.owns(p_rid)) { reflection_atlas_set_size(p_rid, 0, 0); + ReflectionAtlas *ra = reflection_atlas_owner.getornull(p_rid); + if (ra->cluster_builder) { + memdelete(ra->cluster_builder); + } reflection_atlas_owner.free(p_rid); } else if (reflection_probe_instance_owner.owns(p_rid)) { //not much to delete, just free it @@ -7856,8 +3874,8 @@ bool RendererSceneRenderRD::free(RID p_rid) { decal_instance_owner.free(p_rid); } else if (lightmap_instance_owner.owns(p_rid)) { lightmap_instance_owner.free(p_rid); - } else if (gi_probe_instance_owner.owns(p_rid)) { - GIProbeInstance *gi_probe = gi_probe_instance_owner.getornull(p_rid); + } else if (gi.gi_probe_instance_owner.owns(p_rid)) { + RendererSceneGIRD::GIProbeInstance *gi_probe = gi.gi_probe_instance_owner.getornull(p_rid); if (gi_probe->texture.is_valid()) { RD::get_singleton()->free(gi_probe->texture); RD::get_singleton()->free(gi_probe->write_buffer); @@ -7868,37 +3886,10 @@ bool RendererSceneRenderRD::free(RID p_rid) { RD::get_singleton()->free(gi_probe->dynamic_maps[i].depth); } - gi_probe_instance_owner.free(p_rid); - } else if (sky_owner.owns(p_rid)) { - _update_dirty_skys(); - Sky *sky = sky_owner.getornull(p_rid); - - if (sky->radiance.is_valid()) { - RD::get_singleton()->free(sky->radiance); - sky->radiance = RID(); - } - _clear_reflection_data(sky->reflection); - - if (sky->uniform_buffer.is_valid()) { - RD::get_singleton()->free(sky->uniform_buffer); - sky->uniform_buffer = RID(); - } - - if (sky->half_res_pass.is_valid()) { - RD::get_singleton()->free(sky->half_res_pass); - sky->half_res_pass = RID(); - } - - if (sky->quarter_res_pass.is_valid()) { - RD::get_singleton()->free(sky->quarter_res_pass); - sky->quarter_res_pass = RID(); - } - - if (sky->material.is_valid()) { - storage->free(sky->material); - } - - sky_owner.free(p_rid); + gi.gi_probe_instance_owner.free(p_rid); + } else if (sky.sky_owner.owns(p_rid)) { + sky.update_dirty_skys(); + sky.free_sky(p_rid); } else if (light_instance_owner.owns(p_rid)) { LightInstance *light_instance = light_instance_owner.getornull(p_rid); @@ -7932,7 +3923,7 @@ void RendererSceneRenderRD::set_debug_draw_mode(RS::ViewportDebugDraw p_debug_dr } void RendererSceneRenderRD::update() { - _update_dirty_skys(); + sky.update_dirty_skys(); } void RendererSceneRenderRD::set_time(double p_time, double p_step) { @@ -8058,26 +4049,23 @@ TypedArray<Image> RendererSceneRenderRD::bake_render_uv2(RID p_base, const Vecto } void RendererSceneRenderRD::sdfgi_set_debug_probe_select(const Vector3 &p_position, const Vector3 &p_dir) { - sdfgi_debug_probe_pos = p_position; - sdfgi_debug_probe_dir = p_dir; + gi.sdfgi_debug_probe_pos = p_position; + gi.sdfgi_debug_probe_dir = p_dir; } RendererSceneRenderRD *RendererSceneRenderRD::singleton = nullptr; -RID RendererSceneRenderRD::get_cluster_builder_texture() { - return cluster.builder.get_cluster_texture(); -} - -RID RendererSceneRenderRD::get_cluster_builder_indices_buffer() { - return cluster.builder.get_cluster_indices_buffer(); -} - RID RendererSceneRenderRD::get_reflection_probe_buffer() { return cluster.reflection_buffer; } -RID RendererSceneRenderRD::get_positional_light_buffer() { - return cluster.light_buffer; +RID RendererSceneRenderRD::get_omni_light_buffer() { + return cluster.omni_light_buffer; } + +RID RendererSceneRenderRD::get_spot_light_buffer() { + return cluster.spot_light_buffer; +} + RID RendererSceneRenderRD::get_directional_light_buffer() { return cluster.directional_light_buffer; } @@ -8093,413 +4081,59 @@ bool RendererSceneRenderRD::is_low_end() const { } RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { + max_cluster_elements = GLOBAL_GET("rendering/limits/cluster_builder/max_clustered_elements"); + storage = p_storage; singleton = this; - roughness_layers = GLOBAL_GET("rendering/quality/reflections/roughness_layers"); - sky_ggx_samples_quality = GLOBAL_GET("rendering/quality/reflections/ggx_samples"); - sky_use_cubemap_array = GLOBAL_GET("rendering/quality/reflections/texture_array_reflections"); - // sky_use_cubemap_array = false; + directional_shadow.size = GLOBAL_GET("rendering/shadows/directional_shadow/size"); + directional_shadow.use_16_bits = GLOBAL_GET("rendering/shadows/directional_shadow/16_bits"); uint32_t textures_per_stage = RD::get_singleton()->limit_get(RD::LIMIT_MAX_TEXTURES_PER_SHADER_STAGE); - low_end = GLOBAL_GET("rendering/quality/rd_renderer/use_low_end_renderer"); + low_end = GLOBAL_GET("rendering/driver/rd_renderer/use_low_end_renderer"); if (textures_per_stage < 48) { low_end = true; } - if (!low_end) { - //kinda complicated to compute the amount of slots, we try to use as many as we can - - gi_probe_max_lights = 32; - - gi_probe_lights = memnew_arr(GIProbeLight, gi_probe_max_lights); - gi_probe_lights_uniform = RD::get_singleton()->uniform_buffer_create(gi_probe_max_lights * sizeof(GIProbeLight)); - gi_probe_quality = RS::GIProbeQuality(CLAMP(int(GLOBAL_GET("rendering/quality/gi_probes/quality")), 0, 1)); - - String defines = "\n#define MAX_LIGHTS " + itos(gi_probe_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"); - - giprobe_shader.initialize(versions, defines); - giprobe_lighting_shader_version = giprobe_shader.version_create(); - for (int i = 0; i < GI_PROBE_SHADER_VERSION_MAX; i++) { - giprobe_lighting_shader_version_shaders[i] = giprobe_shader.version_get_shader(giprobe_lighting_shader_version, i); - giprobe_lighting_shader_version_pipelines[i] = RD::get_singleton()->compute_pipeline_create(giprobe_lighting_shader_version_shaders[i]); - } - } - - if (!low_end) { - 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"); - - giprobe_debug_shader.initialize(versions, defines); - giprobe_debug_shader_version = giprobe_debug_shader.version_create(); - for (int i = 0; i < GI_PROBE_DEBUG_MAX; i++) { - giprobe_debug_shader_version_shaders[i] = giprobe_debug_shader.version_get_shader(giprobe_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; - - giprobe_debug_shader_version_pipelines[i].setup(giprobe_debug_shader_version_shaders[i], RD::RENDER_PRIMITIVE_TRIANGLES, rs, RD::PipelineMultisampleState(), ds, RD::PipelineColorBlendState::create_disabled(), 0); - } - } - /* SKY SHADER */ - { - // 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_shader.shader.initialize(sky_modes, defines); - } - - // register our shader funds - storage->shader_set_data_request_function(RendererStorageRD::SHADER_TYPE_SKY, _create_sky_shader_funcs); - storage->material_set_data_request_function(RendererStorageRD::SHADER_TYPE_SKY, _create_sky_material_funcs); + sky.init(storage); - { - ShaderCompilerRD::DefaultIdentifierActions actions; - - actions.renames["COLOR"] = "color"; - actions.renames["ALPHA"] = "alpha"; - actions.renames["EYEDIR"] = "cube_normal"; - actions.renames["POSITION"] = "params.position_multiplier.xyz"; - actions.renames["SKY_COORDS"] = "panorama_coords"; - actions.renames["SCREEN_UV"] = "uv"; - actions.renames["TIME"] = "params.time"; - 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.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_variables.data"; - - sky_shader.compiler.initialize(actions); - } - - { - // default material and shader for sky shader - sky_shader.default_shader = storage->shader_create(); - storage->shader_set_code(sky_shader.default_shader, "shader_type sky; void fragment() { COLOR = vec3(0.0); } \n"); - sky_shader.default_material = storage->material_create(); - storage->material_set_shader(sky_shader.default_material, sky_shader.default_shader); - - SkyMaterialData *md = (SkyMaterialData *)storage->material_get_data(sky_shader.default_material, RendererStorageRD::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; - - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; - u.binding = 0; - u.ids.resize(12); - RID *ids_ptr = u.ids.ptrw(); - ids_ptr[0] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); - ids_ptr[1] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); - ids_ptr[2] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); - ids_ptr[3] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); - ids_ptr[4] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); - ids_ptr[5] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); - ids_ptr[6] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); - ids_ptr[7] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); - ids_ptr[8] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); - ids_ptr[9] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); - ids_ptr[10] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); - ids_ptr[11] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); - uniforms.push_back(u); - } - - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.binding = 1; - u.ids.push_back(storage->global_variables_get_storage_buffer()); - uniforms.push_back(u); - } - - { - RD::Uniform u; - u.binding = 2; - u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.ids.push_back(sky_scene_state.uniform_buffer); - uniforms.push_back(u); - } - - { - RD::Uniform u; - u.binding = 3; - u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.ids.push_back(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 = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE); - u.ids.push_back(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 = storage->shader_create(); - storage->shader_set_code(sky_scene_state.fog_shader, "shader_type sky; uniform vec4 clear_color; void fragment() { COLOR = clear_color.rgb; } \n"); - sky_scene_state.fog_material = storage->material_create(); - 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.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK)); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 1; - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE)); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 2; - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::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); - } + /* GI */ if (!low_end) { - //SDFGI - { - 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 < SDGIShader::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 < SDGIShader::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"; - - 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 < SDGIShader::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; - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_WHITE)); - uniforms.push_back(u); - } - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; - u.binding = 1; - u.ids.push_back(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); - } - } - { - //calculate tables - String defines = "\n#define SDFGI_OCT_SIZE " + itos(SDFGI::LIGHTPROBE_OCT_SIZE) + "\n"; - Vector<String> gi_modes; - gi_modes.push_back(""); - gi.shader.initialize(gi_modes, defines); - gi.shader_version = gi.shader.version_create(); - for (int i = 0; i < GI::MODE_MAX; i++) { - gi.pipelines[i] = RD::get_singleton()->compute_pipeline_create(gi.shader.version_get_shader(gi.shader_version, i)); - } - - gi.sdfgi_ubo = RD::get_singleton()->uniform_buffer_create(sizeof(GI::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_VISIBILITY\n"); - - sdfgi_shader.debug_probes.initialize(versions, defines); - 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 < SDGIShader::PROBE_DEBUG_MAX; i++) { - 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_giprobe_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(GI::GIProbeData) * RenderBuffers::MAX_GIPROBES); + gi.init(storage, &sky); } - //cluster setup - uint32_t uniform_max_size = RD::get_singleton()->limit_get(RD::LIMIT_MAX_UNIFORM_BUFFER_SIZE); + { //decals + cluster.max_decals = max_cluster_elements; + uint32_t decal_buffer_size = cluster.max_decals * sizeof(Cluster::DecalData); + cluster.decals = memnew_arr(Cluster::DecalData, cluster.max_decals); + cluster.decal_sort = memnew_arr(Cluster::InstanceSort<DecalInstance>, cluster.max_decals); + cluster.decal_buffer = RD::get_singleton()->storage_buffer_create(decal_buffer_size); + } { //reflections - uint32_t reflection_buffer_size; - if (uniform_max_size < 65536) { - //Yes, you guessed right, ARM again - reflection_buffer_size = uniform_max_size; - } else { - reflection_buffer_size = 65536; - } - cluster.max_reflections = reflection_buffer_size / sizeof(Cluster::ReflectionData); + cluster.max_reflections = max_cluster_elements; cluster.reflections = memnew_arr(Cluster::ReflectionData, cluster.max_reflections); - cluster.reflection_buffer = RD::get_singleton()->storage_buffer_create(reflection_buffer_size); + cluster.reflection_sort = memnew_arr(Cluster::InstanceSort<ReflectionProbeInstance>, cluster.max_reflections); + cluster.reflection_buffer = RD::get_singleton()->storage_buffer_create(sizeof(Cluster::ReflectionData) * cluster.max_reflections); } { //lights - cluster.max_lights = MIN(1024 * 1024, uniform_max_size) / sizeof(Cluster::LightData); //1mb of lights + cluster.max_lights = max_cluster_elements; + uint32_t light_buffer_size = cluster.max_lights * sizeof(Cluster::LightData); - cluster.lights = memnew_arr(Cluster::LightData, cluster.max_lights); - cluster.light_buffer = RD::get_singleton()->storage_buffer_create(light_buffer_size); + cluster.omni_lights = memnew_arr(Cluster::LightData, cluster.max_lights); + cluster.omni_light_buffer = RD::get_singleton()->storage_buffer_create(light_buffer_size); + cluster.omni_light_sort = memnew_arr(Cluster::InstanceSort<LightInstance>, cluster.max_lights); + cluster.spot_lights = memnew_arr(Cluster::LightData, cluster.max_lights); + cluster.spot_light_buffer = RD::get_singleton()->storage_buffer_create(light_buffer_size); + cluster.spot_light_sort = memnew_arr(Cluster::InstanceSort<LightInstance>, cluster.max_lights); //defines += "\n#define MAX_LIGHT_DATA_STRUCTS " + itos(cluster.max_lights) + "\n"; - cluster.lights_instances = memnew_arr(RID, cluster.max_lights); - cluster.lights_shadow_rect_cache = memnew_arr(Rect2i, cluster.max_lights); cluster.max_directional_lights = MAX_DIRECTIONAL_LIGHTS; uint32_t directional_light_buffer_size = cluster.max_directional_lights * sizeof(Cluster::DirectionalLightData); @@ -8507,15 +4141,6 @@ RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { cluster.directional_light_buffer = RD::get_singleton()->uniform_buffer_create(directional_light_buffer_size); } - { //decals - cluster.max_decals = MIN(1024 * 1024, uniform_max_size) / sizeof(Cluster::DecalData); //1mb of decals - uint32_t decal_buffer_size = cluster.max_decals * sizeof(Cluster::DecalData); - cluster.decals = memnew_arr(Cluster::DecalData, cluster.max_decals); - cluster.decal_buffer = RD::get_singleton()->storage_buffer_create(decal_buffer_size); - } - - cluster.builder.setup(16, 8, 24); - if (!low_end) { String defines = "\n#define MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS " + itos(cluster.max_directional_lights) + "\n"; Vector<String> volumetric_fog_modes; @@ -8528,6 +4153,7 @@ RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { for (int i = 0; i < VOLUMETRIC_FOG_SHADER_MAX; i++) { volumetric_fog.pipelines[i] = RD::get_singleton()->compute_pipeline_create(volumetric_fog.shader.version_get_shader(volumetric_fog.shader_version, i)); } + volumetric_fog.params_ubo = RD::get_singleton()->uniform_buffer_create(sizeof(VolumetricFogShader::ParamsUBO)); } { @@ -8539,74 +4165,59 @@ RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { shadow_sampler = RD::get_singleton()->sampler_create(sampler); } - camera_effects_set_dof_blur_bokeh_shape(RS::DOFBokehShape(int(GLOBAL_GET("rendering/quality/depth_of_field/depth_of_field_bokeh_shape")))); - camera_effects_set_dof_blur_quality(RS::DOFBlurQuality(int(GLOBAL_GET("rendering/quality/depth_of_field/depth_of_field_bokeh_quality"))), GLOBAL_GET("rendering/quality/depth_of_field/depth_of_field_use_jitter")); - environment_set_ssao_quality(RS::EnvironmentSSAOQuality(int(GLOBAL_GET("rendering/quality/ssao/quality"))), GLOBAL_GET("rendering/quality/ssao/half_size"), GLOBAL_GET("rendering/quality/ssao/adaptive_target"), GLOBAL_GET("rendering/quality/ssao/blur_passes"), GLOBAL_GET("rendering/quality/ssao/fadeout_from"), GLOBAL_GET("rendering/quality/ssao/fadeout_to")); - screen_space_roughness_limiter = GLOBAL_GET("rendering/quality/screen_filters/screen_space_roughness_limiter_enabled"); - screen_space_roughness_limiter_amount = GLOBAL_GET("rendering/quality/screen_filters/screen_space_roughness_limiter_amount"); - screen_space_roughness_limiter_limit = GLOBAL_GET("rendering/quality/screen_filters/screen_space_roughness_limiter_limit"); - glow_bicubic_upscale = int(GLOBAL_GET("rendering/quality/glow/upscale_mode")) > 0; - glow_high_quality = GLOBAL_GET("rendering/quality/glow/use_high_quality"); - ssr_roughness_quality = RS::EnvironmentSSRRoughnessQuality(int(GLOBAL_GET("rendering/quality/screen_space_reflection/roughness_quality"))); - sss_quality = RS::SubSurfaceScatteringQuality(int(GLOBAL_GET("rendering/quality/subsurface_scattering/subsurface_scattering_quality"))); - sss_scale = GLOBAL_GET("rendering/quality/subsurface_scattering/subsurface_scattering_scale"); - sss_depth_scale = GLOBAL_GET("rendering/quality/subsurface_scattering/subsurface_scattering_depth_scale"); + camera_effects_set_dof_blur_bokeh_shape(RS::DOFBokehShape(int(GLOBAL_GET("rendering/camera/depth_of_field/depth_of_field_bokeh_shape")))); + camera_effects_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")); + environment_set_ssao_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")); + 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"); + ssr_roughness_quality = RS::EnvironmentSSRRoughnessQuality(int(GLOBAL_GET("rendering/environment/screen_space_reflection/roughness_quality"))); + 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"); 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); - shadows_quality_set(RS::ShadowQuality(int(GLOBAL_GET("rendering/quality/shadows/soft_shadow_quality")))); - directional_shadow_quality_set(RS::ShadowQuality(int(GLOBAL_GET("rendering/quality/directional_shadow/soft_shadow_quality")))); + shadows_quality_set(RS::ShadowQuality(int(GLOBAL_GET("rendering/shadows/shadows/soft_shadow_quality")))); + directional_shadow_quality_set(RS::ShadowQuality(int(GLOBAL_GET("rendering/shadows/directional_shadow/soft_shadow_quality")))); - environment_set_volumetric_fog_volume_size(GLOBAL_GET("rendering/volumetric_fog/volume_size"), GLOBAL_GET("rendering/volumetric_fog/volume_depth")); - environment_set_volumetric_fog_filter_active(GLOBAL_GET("rendering/volumetric_fog/use_filter")); - environment_set_volumetric_fog_directional_shadow_shrink_size(GLOBAL_GET("rendering/volumetric_fog/directional_shadow_shrink")); - environment_set_volumetric_fog_positional_shadow_shrink_size(GLOBAL_GET("rendering/volumetric_fog/positional_shadow_shrink")); + 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")); cull_argument.set_page_pool(&cull_argument_pool); + + gi.half_resolution = GLOBAL_GET("rendering/global_illumination/gi/use_half_resolution"); } RendererSceneRenderRD::~RendererSceneRenderRD() { - for (Map<Vector2i, ShadowMap>::Element *E = shadow_maps.front(); E; E = E->next()) { - RD::get_singleton()->free(E->get().depth); - } for (Map<int, ShadowCubemap>::Element *E = shadow_cubemaps.front(); E; E = E->next()) { RD::get_singleton()->free(E->get().cubemap); } - if (sky_scene_state.uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(sky_scene_state.uniform_set)) { - RD::get_singleton()->free(sky_scene_state.uniform_set); + 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 (!low_end) { - RD::get_singleton()->free(default_giprobe_buffer); - RD::get_singleton()->free(gi_probe_lights_uniform); - RD::get_singleton()->free(gi.sdfgi_ubo); - - giprobe_debug_shader.version_free(giprobe_debug_shader_version); - giprobe_shader.version_free(giprobe_lighting_shader_version); - gi.shader.version_free(gi.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); + gi.free(); volumetric_fog.shader.version_free(volumetric_fog.shader_version); - - memdelete_arr(gi_probe_lights); - } - - SkyMaterialData *md = (SkyMaterialData *)storage->material_get_data(sky_shader.default_material, RendererStorageRD::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); - storage->free(sky_shader.default_shader); - storage->free(sky_shader.default_material); - storage->free(sky_scene_state.fog_shader); - storage->free(sky_scene_state.fog_material); + RD::get_singleton()->free(volumetric_fog.params_ubo); + } + + RendererSceneSkyRD::SkyMaterialData *md = (RendererSceneSkyRD::SkyMaterialData *)storage->material_get_data(sky.sky_shader.default_material, RendererStorageRD::SHADER_TYPE_SKY); + sky.sky_shader.shader.version_free(md->shader_data->version); + RD::get_singleton()->free(sky.sky_scene_state.directional_light_buffer); + RD::get_singleton()->free(sky.sky_scene_state.uniform_buffer); + memdelete_arr(sky.sky_scene_state.directional_lights); + memdelete_arr(sky.sky_scene_state.last_frame_directional_lights); + storage->free(sky.sky_shader.default_shader); + storage->free(sky.sky_shader.default_material); + storage->free(sky.sky_scene_state.fog_shader); + storage->free(sky.sky_scene_state.fog_material); memdelete_arr(directional_penumbra_shadow_kernel); memdelete_arr(directional_soft_shadow_kernel); memdelete_arr(penumbra_shadow_kernel); @@ -8614,15 +4225,19 @@ RendererSceneRenderRD::~RendererSceneRenderRD() { { RD::get_singleton()->free(cluster.directional_light_buffer); - RD::get_singleton()->free(cluster.light_buffer); + RD::get_singleton()->free(cluster.omni_light_buffer); + RD::get_singleton()->free(cluster.spot_light_buffer); RD::get_singleton()->free(cluster.reflection_buffer); RD::get_singleton()->free(cluster.decal_buffer); memdelete_arr(cluster.directional_lights); - memdelete_arr(cluster.lights); - memdelete_arr(cluster.lights_shadow_rect_cache); - memdelete_arr(cluster.lights_instances); + memdelete_arr(cluster.omni_lights); + memdelete_arr(cluster.spot_lights); + memdelete_arr(cluster.omni_light_sort); + memdelete_arr(cluster.spot_light_sort); memdelete_arr(cluster.reflections); + memdelete_arr(cluster.reflection_sort); memdelete_arr(cluster.decals); + memdelete_arr(cluster.decal_sort); } RD::get_singleton()->free(shadow_sampler); diff --git a/servers/rendering/renderer_rd/renderer_scene_render_rd.h b/servers/rendering/renderer_rd/renderer_scene_render_rd.h index f81a35f025..001cfeb74d 100644 --- a/servers/rendering/renderer_rd/renderer_scene_render_rd.h +++ b/servers/rendering/renderer_rd/renderer_scene_render_rd.h @@ -34,69 +34,23 @@ #include "core/templates/local_vector.h" #include "core/templates/rid_owner.h" #include "servers/rendering/renderer_compositor.h" -#include "servers/rendering/renderer_rd/light_cluster_builder.h" +#include "servers/rendering/renderer_rd/cluster_builder_rd.h" +#include "servers/rendering/renderer_rd/renderer_scene_environment_rd.h" +#include "servers/rendering/renderer_rd/renderer_scene_gi_rd.h" +#include "servers/rendering/renderer_rd/renderer_scene_sky_rd.h" #include "servers/rendering/renderer_rd/renderer_storage_rd.h" -#include "servers/rendering/renderer_rd/shaders/gi.glsl.gen.h" -#include "servers/rendering/renderer_rd/shaders/giprobe.glsl.gen.h" -#include "servers/rendering/renderer_rd/shaders/giprobe_debug.glsl.gen.h" -#include "servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl.gen.h" -#include "servers/rendering/renderer_rd/shaders/sdfgi_debug_probes.glsl.gen.h" -#include "servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl.gen.h" -#include "servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl.gen.h" -#include "servers/rendering/renderer_rd/shaders/sdfgi_preprocess.glsl.gen.h" -#include "servers/rendering/renderer_rd/shaders/sky.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/volumetric_fog.glsl.gen.h" #include "servers/rendering/renderer_scene_render.h" #include "servers/rendering/rendering_device.h" class RendererSceneRenderRD : public RendererSceneRender { + friend RendererSceneSkyRD; + friend RendererSceneGIRD; + protected: + RendererStorageRD *storage; double time; - - // 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]; - }; - - struct SkySceneState { - struct UBO { - uint32_t volumetric_fog_enabled; - float volumetric_fog_inv_length; - float volumetric_fog_detail_spread; - - float fog_aerial_perspective; - - float fog_light_color[3]; - float fog_sun_scatter; - - uint32_t fog_enabled; - float fog_density; - - float z_far; - uint32_t directional_light_count; - }; - - UBO ubo; - - SkyDirectionalLightData *directional_lights; - SkyDirectionalLightData *last_frame_directional_lights; - 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; + double time_step = 0; struct RenderBufferData { virtual void configure(RID p_color_buffer, RID p_depth_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa) = 0; @@ -104,19 +58,22 @@ protected: }; virtual RenderBufferData *_create_render_buffer_data() = 0; - void _setup_lights(const PagedArray<RID> &p_lights, const Transform &p_camera_inverse_transform, RID p_shadow_atlas, bool p_using_shadows, uint32_t &r_directional_light_count, uint32_t &r_positional_light_count); + void _setup_lights(const PagedArray<RID> &p_lights, const Transform &p_camera_transform, RID p_shadow_atlas, bool p_using_shadows, uint32_t &r_directional_light_count, uint32_t &r_positional_light_count); void _setup_decals(const PagedArray<RID> &p_decals, const Transform &p_camera_inverse_xform); void _setup_reflections(const PagedArray<RID> &p_reflections, const Transform &p_camera_inverse_transform, RID p_environment); - void _setup_giprobes(RID p_render_buffers, const Transform &p_transform, const PagedArray<RID> &p_gi_probes, uint32_t &r_gi_probes_used); - virtual void _render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, int p_directional_light_count, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_color, float p_screen_lod_threshold) = 0; - virtual void _render_shadow(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0) = 0; + virtual void _render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_cluster_buffer, uint32_t p_cluster_size, uint32_t p_cluster_max_elements, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_color, float p_screen_lod_threshold) = 0; + + virtual void _render_shadow_begin() = 0; + virtual void _render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform &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_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) = 0; + virtual void _render_shadow_process() = 0; + virtual void _render_shadow_end(uint32_t p_barrier = RD::BARRIER_MASK_ALL) = 0; + virtual void _render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) = 0; virtual void _render_uv2(const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) = 0; virtual void _render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<GeometryInstance *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture) = 0; virtual void _render_particle_collider_heightfield(RID p_fb, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<GeometryInstance *> &p_instances) = 0; - virtual void _debug_giprobe(RID p_gi_probe, RenderingDevice::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform, bool p_lighting, bool p_emission, float p_alpha); void _debug_sdfgi_probes(RID p_render_buffers, RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform); RenderBufferData *render_buffers_get_data(RID p_render_buffers); @@ -124,204 +81,36 @@ protected: virtual void _base_uniforms_changed() = 0; virtual void _render_buffers_uniform_set_changed(RID p_render_buffers) = 0; virtual RID _render_buffers_get_normal_texture(RID p_render_buffers) = 0; - virtual RID _render_buffers_get_ambient_texture(RID p_render_buffers) = 0; - virtual RID _render_buffers_get_reflection_texture(RID p_render_buffers) = 0; void _process_ssao(RID p_render_buffers, RID p_environment, RID p_normal_buffer, const CameraMatrix &p_projection); void _process_ssr(RID p_render_buffers, RID p_dest_framebuffer, RID p_normal_buffer, RID p_specular_buffer, RID p_metallic, const Color &p_metallic_mask, RID p_environment, const CameraMatrix &p_projection, bool p_use_additive); void _process_sss(RID p_render_buffers, const CameraMatrix &p_camera); - void _setup_sky(RID p_environment, RID p_render_buffers, const CameraMatrix &p_projection, const Transform &p_transform, const Size2i p_screen_size); - void _update_sky(RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform); - void _draw_sky(bool p_can_continue_color, bool p_can_continue_depth, RID p_fb, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform); - void _process_gi(RID p_render_buffers, RID p_normal_roughness_buffer, RID p_ambient_buffer, RID p_reflection_buffer, RID p_gi_probe_buffer, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform, const PagedArray<RID> &p_gi_probes); + bool _needs_post_prepass_render(bool p_use_gi); + void _post_prepass_render(bool p_use_gi); + void _pre_resolve_render(bool p_use_gi); + + void _pre_opaque_render(bool p_use_ssao, bool p_use_gi, RID p_normal_roughness_buffer, RID p_gi_probe_buffer); + uint32_t _get_render_state_directional_light_count() const; // needed for a single argument calls (material and uv2) PagedArrayPool<GeometryInstance *> cull_argument_pool; PagedArray<GeometryInstance *> cull_argument; //need this to exist -private: - RS::ViewportDebugDraw debug_draw = RS::VIEWPORT_DEBUG_DRAW_DISABLED; - double time_step = 0; - static RendererSceneRenderRD *singleton; - - int roughness_layers; - RendererStorageRD *storage; - - 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 - }; + RendererSceneGIRD gi; + RendererSceneSkyRD sky; - struct DownsampleLayer { - struct Mipmap { - RID view; - Size2i size; - }; - 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(ReflectionData &rd); - void _update_reflection_data(ReflectionData &rd, int p_size, int p_mipmaps, bool p_use_array, RID p_base_cube, int p_base_layer, bool p_low_quality); - void _create_reflection_fast_filter(ReflectionData &rd, bool p_use_arrays); - void _create_reflection_importance_sample(ReflectionData &rd, bool p_use_arrays, int p_cube_side, int p_base_layer); - void _update_reflection_mipmaps(ReflectionData &rd, int p_start, int p_end); - - /* Sky shader */ - - 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_MAX - }; - - struct SkyShader { - SkyShaderRD shader; - ShaderCompilerRD compiler; - - RID default_shader; - RID default_material; - RID default_shader_rd; - } sky_shader; - - struct SkyShaderData : public RendererStorageRD::ShaderData { - bool valid; - RID version; - - PipelineCacheRD pipelines[SKY_VERSION_MAX]; - Map<StringName, ShaderLanguage::ShaderNode::Uniform> uniforms; - Vector<ShaderCompilerRD::GeneratedCode::Texture> texture_uniforms; - - Vector<uint32_t> ubo_offsets; - uint32_t ubo_size; - - String path; - String code; - Map<StringName, RID> default_texture_params; - - bool uses_time; - bool uses_position; - bool uses_half_res; - bool uses_quarter_res; - bool uses_light; - - virtual void set_code(const String &p_Code); - virtual void set_default_texture_param(const StringName &p_name, RID p_texture); - virtual void get_param_list(List<PropertyInfo> *p_param_list) const; - virtual void get_instance_param_list(List<RendererStorage::InstanceShaderParam> *p_param_list) const; - virtual bool is_param_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; - SkyShaderData(); - virtual ~SkyShaderData(); - }; - - RendererStorageRD::ShaderData *_create_sky_shader_func(); - static RendererStorageRD::ShaderData *_create_sky_shader_funcs() { - return static_cast<RendererSceneRenderRD *>(singleton)->_create_sky_shader_func(); - }; - - struct SkyMaterialData : public RendererStorageRD::MaterialData { - uint64_t last_frame; - SkyShaderData *shader_data; - RID uniform_buffer; - RID uniform_set; - Vector<RID> texture_cache; - Vector<uint8_t> ubo_data; - bool uniform_set_updated; - - virtual void set_render_priority(int p_priority) {} - virtual void set_next_pass(RID p_pass) {} - virtual void update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); - virtual ~SkyMaterialData(); - }; - - RendererStorageRD::MaterialData *_create_sky_material_func(SkyShaderData *p_shader); - static RendererStorageRD::MaterialData *_create_sky_material_funcs(RendererStorageRD::ShaderData *p_shader) { - return static_cast<RendererSceneRenderRD *>(singleton)->_create_sky_material_func(static_cast<SkyShaderData *>(p_shader)); - }; - - 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 SkySet { - SKY_SET_UNIFORMS, - SKY_SET_MATERIAL, - SKY_SET_TEXTURES, - SKY_SET_FOG, - SKY_SET_MAX - }; - - /* SKY */ - 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; - - //State to track when radiance cubemap needs updating - SkyMaterialData *prev_material; - Vector3 prev_position; - float prev_time; - - RID sdfgi_integrate_sky_uniform_set; - }; - - Sky *dirty_sky_list = nullptr; - - void _sky_invalidate(Sky *p_sky); - void _update_dirty_skys(); - RID _get_sky_textures(Sky *p_sky, SkyTextureSetVersion p_version); - - uint32_t sky_ggx_samples_quality; - bool sky_use_cubemap_array; + RendererSceneEnvironmentRD *get_environment(RID p_environment) { + if (p_environment.is_valid()) { + return environment_owner.getornull(p_environment); + } else { + return nullptr; + } + } - mutable RID_Owner<Sky> sky_owner; +private: + RS::ViewportDebugDraw debug_draw = RS::VIEWPORT_DEBUG_DRAW_DISABLED; + static RendererSceneRenderRD *singleton; /* REFLECTION ATLAS */ @@ -335,11 +124,13 @@ private: struct Reflection { RID owner; - ReflectionData data; + RendererSceneSkyRD::ReflectionData data; RID fbs[6]; }; Vector<Reflection> reflections; + + ClusterBuilderRD *cluster_builder = nullptr; }; mutable RID_Owner<ReflectionAtlas> reflection_atlas_owner; @@ -383,151 +174,6 @@ private: mutable RID_Owner<LightmapInstance> lightmap_instance_owner; - /* GIPROBE INSTANCE */ - - struct GIProbeLight { - uint32_t type; - float energy; - float radius; - float attenuation; - - float color[3]; - float spot_angle_radians; - - float position[3]; - float spot_attenuation; - - float direction[3]; - uint32_t has_shadow; - }; - - struct GIProbePushConstant { - 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 GIProbeDynamicPushConstant { - 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]; - }; - - struct GIProbeInstance { - 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; - - Transform transform; - }; - - GIProbeLight *gi_probe_lights; - uint32_t gi_probe_max_lights; - RID gi_probe_lights_uniform; - - enum { - GI_PROBE_SHADER_VERSION_COMPUTE_LIGHT, - GI_PROBE_SHADER_VERSION_COMPUTE_SECOND_BOUNCE, - GI_PROBE_SHADER_VERSION_COMPUTE_MIPMAP, - GI_PROBE_SHADER_VERSION_WRITE_TEXTURE, - GI_PROBE_SHADER_VERSION_DYNAMIC_OBJECT_LIGHTING, - GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE, - GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_PLOT, - GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE_PLOT, - GI_PROBE_SHADER_VERSION_MAX - }; - GiprobeShaderRD giprobe_shader; - RID giprobe_lighting_shader_version; - RID giprobe_lighting_shader_version_shaders[GI_PROBE_SHADER_VERSION_MAX]; - RID giprobe_lighting_shader_version_pipelines[GI_PROBE_SHADER_VERSION_MAX]; - - mutable RID_Owner<GIProbeInstance> gi_probe_instance_owner; - - RS::GIProbeQuality gi_probe_quality = RS::GI_PROBE_QUALITY_HIGH; - - enum { - GI_PROBE_DEBUG_COLOR, - GI_PROBE_DEBUG_LIGHT, - GI_PROBE_DEBUG_EMISSION, - GI_PROBE_DEBUG_LIGHT_FULL, - GI_PROBE_DEBUG_MAX - }; - - struct GIProbeDebugPushConstant { - float projection[16]; - uint32_t cell_offset; - float dynamic_range; - float alpha; - uint32_t level; - int32_t bounds[3]; - uint32_t pad; - }; - - GiprobeDebugShaderRD giprobe_debug_shader; - RID giprobe_debug_shader_version; - RID giprobe_debug_shader_version_shaders[GI_PROBE_DEBUG_MAX]; - PipelineCacheRD giprobe_debug_shader_version_pipelines[GI_PROBE_DEBUG_MAX]; - RID giprobe_debug_uniform_set; - /* SHADOW ATLAS */ struct ShadowShrinkStage { @@ -571,17 +217,18 @@ private: uint32_t smallest_subdiv = 0; int size = 0; + bool use_16_bits = false; RID depth; RID fb; //for copying Map<RID, uint32_t> shadow_owners; - - Vector<ShadowShrinkStage> shrink_stages; }; RID_Owner<ShadowAtlas> shadow_atlas_owner; + void _update_shadow_atlas(ShadowAtlas *shadow_atlas); + 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); RS::ShadowQuality shadows_quality = RS::SHADOW_QUALITY_MAX; //So it always updates when first set @@ -602,17 +249,16 @@ private: struct DirectionalShadow { RID depth; + RID fb; //when renderign direct int light_count = 0; int size = 0; + bool use_16_bits = false; int current_light = 0; - Vector<ShadowShrinkStage> shrink_stages; - } directional_shadow; - void _allocate_shadow_shrink_stages(RID p_base, int p_base_size, Vector<ShadowShrinkStage> &shrink_stages, uint32_t p_target_size); - void _clear_shadow_shrink_stages(Vector<ShadowShrinkStage> &shrink_stages); + void _update_directional_shadow_atlas(); /* SHADOW CUBEMAPS */ @@ -624,14 +270,6 @@ private: Map<int, ShadowCubemap> shadow_cubemaps; ShadowCubemap *_get_shadow_cubemap(int p_size); - struct ShadowMap { - RID depth; - RID fb; - }; - - Map<Vector2i, ShadowMap> shadow_maps; - ShadowMap *_get_shadow_map(const Size2i &p_size); - void _create_shadow_cubemaps(); /* LIGHT INSTANCE */ @@ -651,7 +289,7 @@ private: RS::LightType light_type = RS::LIGHT_DIRECTIONAL; - ShadowTransform shadow_transform[4]; + ShadowTransform shadow_transform[6]; AABB aabb; RID self; @@ -684,110 +322,6 @@ private: /* ENVIRONMENT */ - struct Environment { - // BG - RS::EnvironmentBG background = RS::ENV_BG_CLEAR_COLOR; - RID sky; - float sky_custom_fov = 0.0; - Basis sky_orientation; - Color bg_color; - float bg_energy = 1.0; - int canvas_max_layer = 0; - RS::EnvironmentAmbientSource ambient_source = RS::ENV_AMBIENT_SOURCE_BG; - Color ambient_light; - float ambient_light_energy = 1.0; - float ambient_sky_contribution = 1.0; - RS::EnvironmentReflectionSource reflection_source = RS::ENV_REFLECTION_SOURCE_BG; - Color ao_color; - - /// Tonemap - - RS::EnvironmentToneMapper tone_mapper; - float exposure = 1.0; - float white = 1.0; - bool auto_exposure = false; - float min_luminance = 0.2; - float max_luminance = 8.0; - float auto_exp_speed = 0.2; - float auto_exp_scale = 0.5; - uint64_t auto_exposure_version = 0; - - // Fog - bool fog_enabled = false; - Color fog_light_color = Color(0.5, 0.6, 0.7); - float fog_light_energy = 1.0; - float fog_sun_scatter = 0.0; - float fog_density = 0.001; - float fog_height = 0.0; - float fog_height_density = 0.0; //can be negative to invert effect - float fog_aerial_perspective = 0.0; - - /// Volumetric Fog - /// - bool volumetric_fog_enabled = false; - float volumetric_fog_density = 0.01; - Color volumetric_fog_light = Color(0, 0, 0); - float volumetric_fog_light_energy = 0.0; - float volumetric_fog_length = 64.0; - float volumetric_fog_detail_spread = 2.0; - RS::EnvVolumetricFogShadowFilter volumetric_fog_shadow_filter = RS::ENV_VOLUMETRIC_FOG_SHADOW_FILTER_LOW; - float volumetric_fog_gi_inject = 0.0; - - /// Glow - - bool glow_enabled = false; - Vector<float> glow_levels; - float glow_intensity = 0.8; - float glow_strength = 1.0; - float glow_bloom = 0.0; - float glow_mix = 0.01; - RS::EnvironmentGlowBlendMode glow_blend_mode = RS::ENV_GLOW_BLEND_MODE_SOFTLIGHT; - float glow_hdr_bleed_threshold = 1.0; - float glow_hdr_luminance_cap = 12.0; - float glow_hdr_bleed_scale = 2.0; - - /// SSAO - - bool ssao_enabled = false; - float ssao_radius = 1.0; - float ssao_intensity = 2.0; - float ssao_power = 1.5; - float ssao_detail = 0.5; - float ssao_horizon = 0.06; - float ssao_sharpness = 0.98; - float ssao_direct_light_affect = 0.0; - float ssao_ao_channel_affect = 0.0; - - /// SSR - /// - bool ssr_enabled = false; - int ssr_max_steps = 64; - float ssr_fade_in = 0.15; - float ssr_fade_out = 2.0; - float ssr_depth_tolerance = 0.2; - - /// SDFGI - bool sdfgi_enabled = false; - RS::EnvironmentSDFGICascades sdfgi_cascades; - float sdfgi_min_cell_size = 0.2; - bool sdfgi_use_occlusion = false; - bool sdfgi_use_multibounce = false; - bool sdfgi_read_sky_light = false; - float sdfgi_energy = 1.0; - float sdfgi_normal_bias = 1.1; - float sdfgi_probe_bias = 1.1; - RS::EnvironmentSDFGIYScale sdfgi_y_scale = RS::ENV_SDFGI_Y_SCALE_DISABLED; - - /// Adjustments - - bool adjustments_enabled = false; - float adjustments_brightness = 1.0f; - float adjustments_contrast = 1.0f; - float adjustments_saturation = 1.0f; - bool use_1d_color_correction = false; - RID color_correction = RID(); - }; - RS::EnvironmentSSAOQuality ssao_quality = RS::ENV_SSAO_QUALITY_MEDIUM; bool ssao_half_size = false; bool ssao_using_half_size = false; @@ -800,9 +334,7 @@ private: bool glow_high_quality = false; RS::EnvironmentSSRRoughnessQuality ssr_roughness_quality = RS::ENV_SSR_ROUGNESS_QUALITY_LOW; - static uint64_t auto_exposure_counter; - - mutable RID_Owner<Environment> environment_owner; + mutable RID_Owner<RendererSceneEnvironmentRD, true> environment_owner; /* CAMERA EFFECTS */ @@ -828,18 +360,16 @@ private: float sss_scale = 0.05; float sss_depth_scale = 0.01; - mutable RID_Owner<CameraEffects> camera_effects_owner; + mutable RID_Owner<CameraEffects, true> camera_effects_owner; /* RENDER BUFFERS */ - struct SDFGI; + ClusterBuilderSharedDataRD cluster_builder_shared; + ClusterBuilderRD *current_cluster_builder = nullptr; + struct VolumetricFog; struct RenderBuffers { - enum { - MAX_GIPROBES = 8 - }; - RenderBufferData *data = nullptr; int width = 0, height = 0; RS::ViewportMSAA msaa = RS::VIEWPORT_MSAA_DISABLED; @@ -854,9 +384,11 @@ private: RID depth_texture; //main depth texture RID gi_uniform_set; - SDFGI *sdfgi = nullptr; + RendererSceneGIRD::SDFGI *sdfgi = nullptr; VolumetricFog *volumetric_fog = nullptr; + ClusterBuilderRD *cluster_builder = nullptr; + //built-in textures used for ping pong image processing and blurring struct Blur { RID texture; @@ -894,389 +426,14 @@ private: RID blur_radius[2]; } ssr; - RID giprobe_textures[MAX_GIPROBES]; - RID giprobe_buffer; - }; - - RID default_giprobe_buffer; - - /* SDFGI */ - - struct SDFGI { - 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; - }; - - //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_uniform_set; - RID scroll_uniform_set; - RID scroll_occlusion_uniform_set; - RID integrate_uniform_set; - RID lights_buffer; - }; + RID ambient_buffer; + RID reflection_buffer; + bool using_half_size_gi = false; - //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; - - RS::EnvironmentSDFGICascades cascade_mode; - 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; - RID debug_probes_uniform_set; - RID cascades_ubo; - - bool uses_occlusion = false; - bool uses_multibounce = false; - bool reads_sky = false; - float energy = 1.0; - float normal_bias = 1.1; - float probe_bias = 1.1; - RS::EnvironmentSDFGIYScale y_scale_mode = RS::ENV_SDFGI_Y_SCALE_DISABLED; - - float y_mult = 1.0; - - uint32_t render_pass = 0; + RendererSceneGIRD::RenderBuffersGI gi; }; - RS::EnvironmentSDFGIRayCount sdfgi_ray_count = RS::ENV_SDFGI_RAY_COUNT_16; - RS::EnvironmentSDFGIFramesToConverge sdfgi_frames_to_converge = RS::ENV_SDFGI_CONVERGE_IN_10_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; - - struct SDGIShader { - 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]; - uint32_t use_occlusion; - float y_mult; - - float cam_extent[3]; - uint32_t probe_axis_size; - - float cam_transform[16]; - }; - - SdfgiDebugShaderRD debug; - RID debug_shader; - RID debug_shader_version; - RID debug_pipeline; - - enum ProbeDebugMode { - PROBE_DEBUG_PROBES, - PROBE_DEBUG_VISIBILITY, - PROBE_DEBUG_MAX - }; - - struct DebugProbesPushConstant { - float projection[16]; - - 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 spot_angle; - float spot_attenuation; - float radius; - - float shadow_color[4]; - }; - - 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; - uint32_t multibounce; - float y_mult; - uint32_t pad; - }; - - 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; - - void _sdfgi_erase(RenderBuffers *rb); - int _sdfgi_get_pending_region_data(RID p_render_buffers, int p_region, Vector3i &r_local_offset, Vector3i &r_local_size, AABB &r_bounds) const; - void _sdfgi_update_cascades(RID p_render_buffers); - /* GI */ - - struct GI { - 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 - }; - - ProbeCascadeData cascades[SDFGI::MAX_CASCADES]; - }; - - struct GIProbeData { - float xform[16]; - float bounds[3]; - float dynamic_range; - - float bias; - float normal_bias; - uint32_t blend_ambient; - uint32_t texture_slot; - - float anisotropy_strength; - float ao; - float ao_size; - uint32_t mipmaps; - }; - - struct PushConstant { - int32_t screen_size[2]; - float z_near; - float z_far; - - float proj_info[4]; - - uint32_t max_giprobes; - uint32_t high_quality_vct; - uint32_t use_sdfgi; - uint32_t orthogonal; - - float ao_color[3]; - uint32_t pad; - - float cam_rotation[12]; - }; - - RID sdfgi_ubo; - enum { - MODE_MAX = 1 - }; - - GiShaderRD shader; - RID shader_version; - RID pipelines[MODE_MAX]; - } gi; - bool screen_space_roughness_limiter = false; float screen_space_roughness_limiter_amount = 0.25; float screen_space_roughness_limiter_limit = 0.18; @@ -1289,21 +446,29 @@ private: void _render_buffers_debug_draw(RID p_render_buffers, RID p_shadow_atlas); void _render_buffers_post_process_and_tonemap(RID p_render_buffers, RID p_environment, RID p_camera_effects, const CameraMatrix &p_projection); - void _sdfgi_debug_draw(RID p_render_buffers, const CameraMatrix &p_projection, const Transform &p_transform); /* Cluster */ struct Cluster { /* Scene State UBO */ - struct ReflectionData { //should always be 128 bytes + 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 params[4]; // intensity, 0, interior , boxproject float ambient[3]; // ambient color, + float intensity; + bool exterior; + bool box_project; uint32_t ambient_mode; + uint32_t pad; float local_matrix[16]; // up to here for spot and omni, rest is for directional }; @@ -1312,10 +477,15 @@ private: float inv_radius; float direction[3]; float size; - uint16_t attenuation_energy[2]; //16 bits attenuation, then energy - uint8_t color_specular[4]; //rgb color, a specular (8 bit unorm) - uint16_t cone_attenuation_angle[2]; // attenuation and angle, (16bit float) - uint8_t shadow_color_enabled[4]; //shadow rgb color, a>0.5 enabled (8bit unorm) + + float color[3]; + float attenuation; + + float inv_spot_attenuation; + float cos_spot_angle; + float specular_amount; + uint32_t shadow_enabled; + float atlas_rect[4]; // in omni, used for atlas uv, in spot, used for projector uv float shadow_matrix[16]; float shadow_bias; @@ -1379,31 +549,85 @@ private: float normal_fade; }; + template <class T> + struct InstanceSort { + float depth; + T *instance; + bool operator<(const InstanceSort &p_sort) const { + return depth < p_sort.depth; + } + }; + ReflectionData *reflections; + InstanceSort<ReflectionProbeInstance> *reflection_sort; uint32_t max_reflections; RID reflection_buffer; uint32_t max_reflection_probes_per_instance; + uint32_t reflection_count = 0; DecalData *decals; + InstanceSort<DecalInstance> *decal_sort; uint32_t max_decals; RID decal_buffer; + uint32_t decal_count; - LightData *lights; + LightData *omni_lights; + LightData *spot_lights; + + InstanceSort<LightInstance> *omni_light_sort; + InstanceSort<LightInstance> *spot_light_sort; uint32_t max_lights; - RID light_buffer; - RID *lights_instances; - Rect2i *lights_shadow_rect_cache; - uint32_t lights_shadow_rect_cache_count = 0; + RID omni_light_buffer; + RID spot_light_buffer; + uint32_t omni_light_count = 0; + uint32_t spot_light_count = 0; DirectionalLightData *directional_lights; uint32_t max_directional_lights; RID directional_light_buffer; - LightClusterBuilder builder; - } cluster; + struct RenderState { + RID render_buffers; + Transform cam_transform; + CameraMatrix cam_projection; + bool cam_ortogonal = false; + const PagedArray<GeometryInstance *> *instances = nullptr; + const PagedArray<RID> *lights = nullptr; + const PagedArray<RID> *reflection_probes = nullptr; + const PagedArray<RID> *gi_probes = nullptr; + const PagedArray<RID> *decals = nullptr; + const PagedArray<RID> *lightmaps = nullptr; + RID environment; + RID camera_effects; + RID shadow_atlas; + RID reflection_atlas; + RID reflection_probe; + int reflection_probe_pass = 0; + float screen_lod_threshold = 0.0; + + const RenderShadowData *render_shadows = nullptr; + int render_shadow_count = 0; + const RenderSDFGIData *render_sdfgi_regions = nullptr; + int render_sdfgi_region_count = 0; + const RenderSDFGIUpdateData *sdfgi_update_data = nullptr; + + uint32_t directional_light_count = 0; + uint32_t gi_probe_count = 0; + + LocalVector<int> cube_shadows; + LocalVector<int> shadows; + LocalVector<int> directional_shadows; + + bool depth_prepass_used; + } render_state; + struct VolumetricFog { + enum { + MAX_TEMPORAL_FRAMES = 16 + }; + uint32_t width = 0; uint32_t height = 0; uint32_t depth = 0; @@ -1412,6 +636,8 @@ private: float spread; RID light_density_map; + RID prev_light_density_map; + RID fog_map; RID uniform_set; RID uniform_set2; @@ -1419,6 +645,8 @@ private: RID sky_uniform_set; int last_shadow_filter = -1; + + Transform prev_cam_transform; }; enum { @@ -1430,7 +658,7 @@ private: }; struct VolumetricFogShader { - struct PushConstant { + struct ParamsUBO { float fog_frustum_size_begin[2]; float fog_frustum_size_end[2]; @@ -1448,13 +676,24 @@ private: float detail_spread; float gi_inject; uint32_t max_gi_probes; - uint32_t pad; + 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]; }; VolumetricFogShaderRD shader; + RID params_ubo; RID shader_version; RID pipelines[VOLUMETRIC_FOG_SHADER_MAX]; @@ -1462,9 +701,7 @@ private: uint32_t volumetric_fog_depth = 128; uint32_t volumetric_fog_size = 128; - bool volumetric_fog_filter_active = false; - uint32_t volumetric_fog_directional_shadow_shrink = 512; - uint32_t volumetric_fog_positional_shadow_shrink = 512; + bool volumetric_fog_filter_active = true; void _volumetric_fog_erase(RenderBuffers *rb); void _update_volumetric_fog(RID p_render_buffers, RID p_environment, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_shadow_atlas, int p_directional_light_count, bool p_use_directional_shadows, int p_positional_light_count, int p_gi_probe_count); @@ -1474,13 +711,18 @@ private: uint64_t scene_pass = 0; uint64_t shadow_atlas_realloc_tolerance_msec = 500; + /* !BAS! is this used anywhere? struct SDFGICosineNeighbour { uint32_t neighbour; float weight; }; + */ + uint32_t max_cluster_elements = 512; bool low_end = false; + void _render_shadow_pass(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<GeometryInstance *> &p_instances, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0, float p_screen_lod_threshold = 0.0, bool p_open_pass = true, bool p_close_pass = true, bool p_clear_region = true); + public: virtual Transform geometry_instance_get_transform(GeometryInstance *p_instance) = 0; virtual AABB geometry_instance_get_aabb(GeometryInstance *p_instance) = 0; @@ -1488,7 +730,7 @@ public: /* SHADOW ATLAS API */ RID shadow_atlas_create(); - void shadow_atlas_set_size(RID p_atlas, int p_size); + void shadow_atlas_set_size(RID p_atlas, int p_size, bool p_16_bits = false); void shadow_atlas_set_quadrant_subdivision(RID p_atlas, int p_quadrant, int p_subdivision); bool shadow_atlas_update_light(RID p_atlas, RID p_light_intance, float p_coverage, uint64_t p_light_version); _FORCE_INLINE_ bool shadow_atlas_owns_light_instance(RID p_atlas, RID p_light_intance) { @@ -1509,7 +751,7 @@ public: return Size2(atlas->size, atlas->size); } - void directional_shadow_atlas_set_size(int p_size); + void directional_shadow_atlas_set_size(int p_size, bool p_16_bits = false); int get_directional_light_shadow_size(RID p_light_intance); void set_directional_shadow_count(int p_count); @@ -1523,28 +765,26 @@ public: /* SDFGI UPDATE */ - int sdfgi_get_lightprobe_octahedron_size() const { return SDFGI::LIGHTPROBE_OCT_SIZE; } virtual void sdfgi_update(RID p_render_buffers, RID p_environment, const Vector3 &p_world_position); virtual int sdfgi_get_pending_region_count(RID p_render_buffers) const; virtual AABB sdfgi_get_pending_region_bounds(RID p_render_buffers, int p_region) const; virtual uint32_t sdfgi_get_pending_region_cascade(RID p_render_buffers, int p_region) const; - virtual void sdfgi_update_probes(RID p_render_buffers, RID p_environment, const Vector<RID> &p_directional_lights, const RID *p_positional_light_instances, uint32_t p_positional_light_count); RID sdfgi_get_ubo() const { return gi.sdfgi_ubo; } + /* SKY API */ - RID sky_create(); + virtual RID sky_allocate(); + virtual void sky_initialize(RID p_rid); + 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); - RID sky_get_radiance_texture_rd(RID p_sky) const; - RID sky_get_radiance_uniform_set_rd(RID p_sky, RID p_shader, int p_set) const; - RID sky_get_material(RID p_sky) const; - /* ENVIRONMENT API */ - RID environment_create(); + virtual RID environment_allocate(); + virtual void environment_initialize(RID p_rid); void environment_set_background(RID p_env, RS::EnvironmentBG p_bg); void environment_set_sky(RID p_env, RID p_sky); @@ -1585,12 +825,10 @@ public: float environment_get_fog_height_density(RID p_env) const; float environment_get_fog_aerial_perspective(RID p_env) const; - void environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, RS::EnvVolumetricFogShadowFilter p_shadow_filter); + void environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, bool p_temporal_reprojection, float p_temporal_reprojection_amount); virtual void environment_set_volumetric_fog_volume_size(int p_size, int p_depth); virtual void environment_set_volumetric_fog_filter_active(bool p_enable); - virtual void environment_set_volumetric_fog_directional_shadow_shrink_size(int p_shrink_size); - virtual void environment_set_volumetric_fog_positional_shadow_shrink_size(int p_shrink_size); void environment_set_ssr(RID p_env, bool p_enable, int p_max_steps, float p_fade_int, float p_fade_out, float p_depth_tolerance); void environment_set_ssao(RID p_env, bool p_enable, float p_radius, float p_intensity, float p_power, float p_detail, float p_horizon, float p_sharpness, float p_light_affect, float p_ao_channel_affect); @@ -1601,9 +839,10 @@ public: bool environment_is_ssr_enabled(RID p_env) const; bool environment_is_sdfgi_enabled(RID p_env) const; - virtual void environment_set_sdfgi(RID p_env, bool p_enable, RS::EnvironmentSDFGICascades p_cascades, float p_min_cell_size, RS::EnvironmentSDFGIYScale p_y_scale, bool p_use_occlusion, bool p_use_multibounce, bool p_read_sky, float p_energy, float p_normal_bias, float p_probe_bias); + virtual void environment_set_sdfgi(RID p_env, bool p_enable, RS::EnvironmentSDFGICascades p_cascades, float p_min_cell_size, RS::EnvironmentSDFGIYScale p_y_scale, bool p_use_occlusion, float p_bounce_feedback, bool p_read_sky, float p_energy, float p_normal_bias, float p_probe_bias); virtual void environment_set_sdfgi_ray_count(RS::EnvironmentSDFGIRayCount p_ray_count); virtual void environment_set_sdfgi_frames_to_converge(RS::EnvironmentSDFGIFramesToConverge p_frames); + virtual void environment_set_sdfgi_frames_to_update_light(RS::EnvironmentSDFGIFramesToUpdateLight p_update); void environment_set_ssr_roughness_quality(RS::EnvironmentSSRRoughnessQuality p_quality); RS::EnvironmentSSRRoughnessQuality environment_get_ssr_roughness_quality() const; @@ -1613,7 +852,8 @@ public: virtual Ref<Image> environment_bake_panorama(RID p_env, bool p_bake_irradiance, const Size2i &p_size); - virtual RID camera_effects_create(); + virtual RID camera_effects_allocate(); + virtual void camera_effects_initialize(RID p_rid); virtual void camera_effects_set_dof_blur_quality(RS::DOFBlurQuality p_quality, bool p_use_jitter); virtual void camera_effects_set_dof_blur_bokeh_shape(RS::DOFBokehShape p_shape); @@ -1849,64 +1089,26 @@ public: return li->transform; } + /* gi light probes */ + RID gi_probe_instance_create(RID p_base); void gi_probe_instance_set_transform_to_data(RID p_probe, const Transform &p_xform); bool gi_probe_needs_update(RID p_probe) const; void gi_probe_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RendererSceneRender::GeometryInstance *> &p_dynamic_objects); + void gi_probe_set_quality(RS::GIProbeQuality p_quality) { gi.gi_probe_quality = p_quality; } - void gi_probe_set_quality(RS::GIProbeQuality p_quality) { gi_probe_quality = p_quality; } - - _FORCE_INLINE_ uint32_t gi_probe_instance_get_slot(RID p_probe) { - GIProbeInstance *gi_probe = gi_probe_instance_owner.getornull(p_probe); - return gi_probe->slot; - } - _FORCE_INLINE_ RID gi_probe_instance_get_base_probe(RID p_probe) { - GIProbeInstance *gi_probe = gi_probe_instance_owner.getornull(p_probe); - return gi_probe->probe; - } - _FORCE_INLINE_ Transform gi_probe_instance_get_transform_to_cell(RID p_probe) { - GIProbeInstance *gi_probe = gi_probe_instance_owner.getornull(p_probe); - return storage->gi_probe_get_to_cell_xform(gi_probe->probe) * gi_probe->transform.affine_inverse(); - } - - _FORCE_INLINE_ RID gi_probe_instance_get_texture(RID p_probe) { - GIProbeInstance *gi_probe = gi_probe_instance_owner.getornull(p_probe); - return gi_probe->texture; - } - - _FORCE_INLINE_ void gi_probe_instance_set_render_index(RID p_instance, uint32_t p_render_index) { - GIProbeInstance *gi_probe = gi_probe_instance_owner.getornull(p_instance); - ERR_FAIL_COND(!gi_probe); - gi_probe->render_index = p_render_index; - } + /* render buffers */ - _FORCE_INLINE_ uint32_t gi_probe_instance_get_render_index(RID p_instance) { - GIProbeInstance *gi_probe = gi_probe_instance_owner.getornull(p_instance); - ERR_FAIL_COND_V(!gi_probe, 0); - - return gi_probe->render_index; - } - /* - _FORCE_INLINE_ void gi_probe_instance_set_render_pass(RID p_instance, uint32_t p_render_pass) { - GIProbeInstance *g_probe = gi_probe_instance_owner.getornull(p_instance); - ERR_FAIL_COND(!g_probe); - g_probe->last_pass = p_render_pass; - } - - _FORCE_INLINE_ uint32_t gi_probe_instance_get_render_pass(RID p_instance) { - GIProbeInstance *g_probe = gi_probe_instance_owner.getornull(p_instance); - ERR_FAIL_COND_V(!g_probe, 0); - - return g_probe->last_pass; - } -*/ RID render_buffers_create(); void render_buffers_configure(RID p_render_buffers, RID p_render_target, int p_width, int p_height, RS::ViewportMSAA p_msaa, RS::ViewportScreenSpaceAA p_screen_space_aa, bool p_use_debanding); + void gi_set_use_half_resolution(bool p_enable); RID render_buffers_get_ao_texture(RID p_render_buffers); RID render_buffers_get_back_buffer_texture(RID p_render_buffers); RID render_buffers_get_gi_probe_buffer(RID p_render_buffers); RID render_buffers_get_default_gi_probe_buffer(); + RID render_buffers_get_gi_ambient_texture(RID p_render_buffers); + RID render_buffers_get_gi_reflection_texture(RID p_render_buffers); uint32_t render_buffers_get_sdfgi_cascade_count(RID p_render_buffers) const; bool render_buffers_is_sdfgi_enabled(RID p_render_buffers) const; @@ -1927,15 +1129,10 @@ public: float render_buffers_get_volumetric_fog_end(RID p_render_buffers); float render_buffers_get_volumetric_fog_detail_spread(RID p_render_buffers); - void render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold); - - void render_shadow(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<GeometryInstance *> &p_instances, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0, float p_screen_lod_threshold = 0.0); + void render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_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); void render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region); - void render_sdfgi(RID p_render_buffers, int p_region, const PagedArray<GeometryInstance *> &p_instances); - void render_sdfgi_static_lights(RID p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const PagedArray<RID> *p_positional_light_cull_result); - void render_particle_collider_heightfield(RID p_collider, const Transform &p_transform, const PagedArray<GeometryInstance *> &p_instances); virtual void set_scene_pass(uint64_t p_pass) { @@ -1985,12 +1182,11 @@ public: return debug_draw; } - virtual void set_time(double p_time, double p_step); + void set_time(double p_time, double p_step); - RID get_cluster_builder_texture(); - RID get_cluster_builder_indices_buffer(); RID get_reflection_probe_buffer(); - RID get_positional_light_buffer(); + RID get_omni_light_buffer(); + RID get_spot_light_buffer(); RID get_directional_light_buffer(); RID get_decal_buffer(); int get_max_directional_lights() const; diff --git a/servers/rendering/renderer_rd/renderer_scene_sky_rd.cpp b/servers/rendering/renderer_rd/renderer_scene_sky_rd.cpp new file mode 100644 index 0000000000..769335ac16 --- /dev/null +++ b/servers/rendering/renderer_rd/renderer_scene_sky_rd.cpp @@ -0,0 +1,1491 @@ +/*************************************************************************/ +/* renderer_scene_sky_rd.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2021 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2021 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_sky_rd.h" +#include "core/config/project_settings.h" +#include "renderer_scene_render_rd.h" +#include "servers/rendering/rendering_server_default.h" + +//////////////////////////////////////////////////////////////////////////////// +// SKY SHADER + +void RendererSceneSkyRD::SkyShaderData::set_code(const String &p_code) { + //compile + + code = p_code; + valid = false; + ubo_size = 0; + uniforms.clear(); + + if (code == String()) { + return; //just invalid, but no error + } + + ShaderCompilerRD::GeneratedCode gen_code; + ShaderCompilerRD::IdentifierActions actions; + + 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 = (RendererSceneRenderRD *)RendererSceneRenderRD::singleton; + + Error err = scene_singleton->sky.sky_shader.compiler.compile(RS::SHADER_SKY, code, &actions, path, gen_code); + + ERR_FAIL_COND(err != OK); + + 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]); + } + print_line("\n**uniforms:\n" + gen_code.uniforms); + // print_line("\n**vertex_globals:\n" + gen_code.vertex_global); + // print_line("\n**vertex_code:\n" + gen_code.vertex); + print_line("\n**fragment_globals:\n" + gen_code.fragment_global); + print_line("\n**fragment_code:\n" + gen_code.fragment); + print_line("\n**light_code:\n" + gen_code.light); +#endif + + scene_singleton->sky.sky_shader.shader.version_set_code(version, gen_code.uniforms, gen_code.vertex_global, gen_code.vertex, gen_code.fragment_global, gen_code.light, gen_code.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; + + 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); + } + + valid = true; +} + +void RendererSceneSkyRD::SkyShaderData::set_default_texture_param(const StringName &p_name, RID p_texture) { + if (!p_texture.is_valid()) { + default_texture_params.erase(p_name); + } else { + default_texture_params[p_name] = p_texture; + } +} + +void RendererSceneSkyRD::SkyShaderData::get_param_list(List<PropertyInfo> *p_param_list) const { + Map<int, StringName> order; + + for (Map<StringName, ShaderLanguage::ShaderNode::Uniform>::Element *E = uniforms.front(); E; E = E->next()) { + if (E->get().scope == ShaderLanguage::ShaderNode::Uniform::SCOPE_GLOBAL || E->get().scope == ShaderLanguage::ShaderNode::Uniform::SCOPE_INSTANCE) { + continue; + } + + if (E->get().texture_order >= 0) { + order[E->get().texture_order + 100000] = E->key(); + } else { + order[E->get().order] = E->key(); + } + } + + for (Map<int, StringName>::Element *E = order.front(); E; E = E->next()) { + PropertyInfo pi = ShaderLanguage::uniform_to_property_info(uniforms[E->get()]); + pi.name = E->get(); + p_param_list->push_back(pi); + } +} + +void RendererSceneSkyRD::SkyShaderData::get_instance_param_list(List<RendererStorage::InstanceShaderParam> *p_param_list) const { + for (Map<StringName, ShaderLanguage::ShaderNode::Uniform>::Element *E = uniforms.front(); E; E = E->next()) { + if (E->get().scope != ShaderLanguage::ShaderNode::Uniform::SCOPE_INSTANCE) { + continue; + } + + RendererStorage::InstanceShaderParam p; + p.info = ShaderLanguage::uniform_to_property_info(E->get()); + p.info.name = E->key(); //supply name + p.index = E->get().instance_index; + p.default_value = ShaderLanguage::constant_value_to_variant(E->get().default_value, E->get().type, E->get().hint); + p_param_list->push_back(p); + } +} + +bool RendererSceneSkyRD::SkyShaderData::is_param_texture(const StringName &p_param) const { + if (!uniforms.has(p_param)) { + return false; + } + + return uniforms[p_param].texture_order >= 0; +} + +bool RendererSceneSkyRD::SkyShaderData::is_animated() const { + return false; +} + +bool RendererSceneSkyRD::SkyShaderData::casts_shadows() const { + return false; +} + +Variant RendererSceneSkyRD::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.hint); + } + return Variant(); +} + +RS::ShaderNativeSourceCode RendererSceneSkyRD::SkyShaderData::get_native_source_code() const { + RendererSceneRenderRD *scene_singleton = (RendererSceneRenderRD *)RendererSceneRenderRD::singleton; + + return scene_singleton->sky.sky_shader.shader.version_get_native_source_code(version); +} + +RendererSceneSkyRD::SkyShaderData::SkyShaderData() { + valid = false; +} + +RendererSceneSkyRD::SkyShaderData::~SkyShaderData() { + RendererSceneRenderRD *scene_singleton = (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 + +void RendererSceneSkyRD::SkyMaterialData::update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { + RendererSceneRenderRD *scene_singleton = (RendererSceneRenderRD *)RendererSceneRenderRD::singleton; + + uniform_set_updated = true; + + if ((uint32_t)ubo_data.size() != shader_data->ubo_size) { + p_uniform_dirty = true; + if (uniform_buffer.is_valid()) { + RD::get_singleton()->free(uniform_buffer); + uniform_buffer = RID(); + } + + ubo_data.resize(shader_data->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()->free(uniform_set); + uniform_set = RID(); + } + } + + //check whether buffer changed + if (p_uniform_dirty && ubo_data.size()) { + update_uniform_buffer(shader_data->uniforms, shader_data->ubo_offsets.ptr(), p_parameters, ubo_data.ptrw(), ubo_data.size(), false); + RD::get_singleton()->buffer_update(uniform_buffer, 0, ubo_data.size(), ubo_data.ptrw()); + } + + uint32_t tex_uniform_count = shader_data->texture_uniforms.size(); + + if ((uint32_t)texture_cache.size() != tex_uniform_count) { + 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()->free(uniform_set); + uniform_set = RID(); + } + } + + if (p_textures_dirty && tex_uniform_count) { + update_textures(p_parameters, shader_data->default_texture_params, shader_data->texture_uniforms, texture_cache.ptrw(), true); + } + + if (shader_data->ubo_size == 0 && shader_data->texture_uniforms.size() == 0) { + // This material does not require an uniform set, so don't create it. + return; + } + + 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; + } + + Vector<RD::Uniform> uniforms; + + { + if (shader_data->ubo_size) { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 0; + u.ids.push_back(uniform_buffer); + uniforms.push_back(u); + } + + const RID *textures = texture_cache.ptrw(); + for (uint32_t i = 0; i < tex_uniform_count; i++) { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 1 + i; + u.ids.push_back(textures[i]); + uniforms.push_back(u); + } + } + + uniform_set = RD::get_singleton()->uniform_set_create(uniforms, scene_singleton->sky.sky_shader.shader.version_get_shader(shader_data->version, 0), SKY_SET_MATERIAL); +} + +RendererSceneSkyRD::SkyMaterialData::~SkyMaterialData() { + if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { + RD::get_singleton()->free(uniform_set); + } + + if (uniform_buffer.is_valid()) { + RD::get_singleton()->free(uniform_buffer); + } +} + +//////////////////////////////////////////////////////////////////////////////// +// ReflectionData + +void RendererSceneSkyRD::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 RendererSceneSkyRD::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) { + //recreate radiance and all data + + int mipmaps = p_mipmaps; + uint32_t w = p_size, h = p_size; + + 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, RD::TEXTURE_SLICE_CUBEMAP); + + mmw = MAX(1, mmw >> 1); + mmh = MAX(1, 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, RD::TEXTURE_SLICE_CUBEMAP); + + mmw = MAX(1, mmw >> 1); + mmh = MAX(1, 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, RD::TEXTURE_SLICE_CUBEMAP); + + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.width = 64; // Always 64x64 + tf.height = 64; + tf.texture_type = RD::TEXTURE_TYPE_CUBE; + tf.array_layers = 6; + tf.mipmaps = 7; + 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()); + { + uint32_t mmw = 64; + uint32_t mmh = 64; + downsampled_layer.mipmaps.resize(7); + 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, RD::TEXTURE_SLICE_CUBEMAP); + + mmw = MAX(1, mmw >> 1); + mmh = MAX(1, mmh >> 1); + } + } +} + +void RendererSceneSkyRD::ReflectionData::create_reflection_fast_filter(RendererStorageRD *p_storage, bool p_use_arrays) { + p_storage->get_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++) { + p_storage->get_effects()->cubemap_downsample(downsampled_layer.mipmaps[i - 1].view, downsampled_layer.mipmaps[i].view, downsampled_layer.mipmaps[i].size); + } + + 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]); + } + } + + p_storage->get_effects()->cubemap_filter(downsampled_radiance_cubemap, views, p_use_arrays); +} + +void RendererSceneSkyRD::ReflectionData::create_reflection_importance_sample(RendererStorageRD *p_storage, bool p_use_arrays, int p_cube_side, int p_base_layer, uint32_t p_sky_ggx_samples_quality) { + if (p_use_arrays) { + //render directly to the layers + p_storage->get_effects()->cubemap_roughness(radiance_base_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 { + p_storage->get_effects()->cubemap_roughness( + layers[0].views[p_base_layer - 1], + 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); + } +} + +void RendererSceneSkyRD::ReflectionData::update_reflection_mipmaps(RendererStorageRD *p_storage, int p_start, int p_end) { + 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]; + RID texture = layers[i].views[j + 1]; + Size2i size = layers[i].mipmaps[j + 1].size; + p_storage->get_effects()->cubemap_downsample(view, texture, size); + } + } +} + +//////////////////////////////////////////////////////////////////////////////// +// RendererSceneSkyRD::Sky + +void RendererSceneSkyRD::Sky::free(RendererStorageRD *p_storage) { + 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()) { + p_storage->free(material); + } +} + +RID RendererSceneSkyRD::Sky::get_textures(RendererStorageRD *p_storage, SkyTextureSetVersion p_version, RID p_default_shader_rd) { + 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.ids.push_back(radiance); + } else { + u.ids.push_back(p_storage->texture_rd_get_default(RendererStorageRD::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.ids.push_back(reflection.layers[0].views[1]); + } else { + u.ids.push_back(half_res_pass); + } + } else { + if (p_version < SKY_TEXTURE_SET_CUBEMAP) { + u.ids.push_back(p_storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE)); + } else { + u.ids.push_back(p_storage->texture_rd_get_default(RendererStorageRD::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.ids.push_back(reflection.layers[0].views[2]); + } else { + u.ids.push_back(quarter_res_pass); + } + } else { + if (p_version < SKY_TEXTURE_SET_CUBEMAP) { + u.ids.push_back(p_storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE)); + } else { + u.ids.push_back(p_storage->texture_rd_get_default(RendererStorageRD::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 RendererSceneSkyRD::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 RendererSceneSkyRD::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 RendererSceneSkyRD::Sky::set_material(RID p_material) { + if (material == p_material) { + return false; + } + + material = p_material; + return true; +} + +Ref<Image> RendererSceneSkyRD::Sky::bake_panorama(RendererStorageRD *p_storage, float p_energy, int p_roughness_layers, const Size2i &p_size) { + if (radiance.is_valid()) { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R32G32B32A32_SFLOAT; + 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()); + p_storage->get_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; + img.instance(); + img->create(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>(); +} + +//////////////////////////////////////////////////////////////////////////////// +// RendererSceneSkyRD + +RendererStorageRD::ShaderData *RendererSceneSkyRD::_create_sky_shader_func() { + SkyShaderData *shader_data = memnew(SkyShaderData); + return shader_data; +} + +RendererStorageRD::ShaderData *RendererSceneSkyRD::_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(); +}; + +RendererStorageRD::MaterialData *RendererSceneSkyRD::_create_sky_material_func(SkyShaderData *p_shader) { + SkyMaterialData *material_data = memnew(SkyMaterialData); + material_data->shader_data = p_shader; + material_data->last_frame = false; + //update will happen later anyway so do nothing. + return material_data; +} + +RendererStorageRD::MaterialData *RendererSceneSkyRD::_create_sky_material_funcs(RendererStorageRD::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)); +}; + +RendererSceneSkyRD::RendererSceneSkyRD() { + 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 RendererSceneSkyRD::init(RendererStorageRD *p_storage) { + storage = p_storage; + + { + // 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_shader.shader.initialize(sky_modes, defines); + } + + // register our shader funds + storage->shader_set_data_request_function(RendererStorageRD::SHADER_TYPE_SKY, _create_sky_shader_funcs); + storage->material_set_data_request_function(RendererStorageRD::SHADER_TYPE_SKY, _create_sky_material_funcs); + + { + ShaderCompilerRD::DefaultIdentifierActions actions; + + actions.renames["COLOR"] = "color"; + actions.renames["ALPHA"] = "alpha"; + actions.renames["EYEDIR"] = "cube_normal"; + actions.renames["POSITION"] = "params.position_multiplier.xyz"; + actions.renames["SKY_COORDS"] = "panorama_coords"; + actions.renames["SCREEN_UV"] = "uv"; + actions.renames["TIME"] = "params.time"; + 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.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_variables.data"; + + sky_shader.compiler.initialize(actions); + } + + { + // default material and shader for sky shader + sky_shader.default_shader = storage->shader_allocate(); + storage->shader_initialize(sky_shader.default_shader); + + storage->shader_set_code(sky_shader.default_shader, "shader_type sky; void fragment() { COLOR = vec3(0.0); } \n"); + + sky_shader.default_material = storage->material_allocate(); + storage->material_initialize(sky_shader.default_material); + + storage->material_set_shader(sky_shader.default_material, sky_shader.default_shader); + + SkyMaterialData *md = (SkyMaterialData *)storage->material_get_data(sky_shader.default_material, RendererStorageRD::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; + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + u.binding = 0; + u.ids.resize(12); + RID *ids_ptr = u.ids.ptrw(); + ids_ptr[0] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[1] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[2] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[3] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[4] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[5] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[6] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[7] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[8] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[9] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[10] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[11] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 1; + u.ids.push_back(storage->global_variables_get_storage_buffer()); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 2; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.ids.push_back(sky_scene_state.uniform_buffer); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 3; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.ids.push_back(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 = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE); + u.ids.push_back(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 = storage->shader_allocate(); + storage->shader_initialize(sky_scene_state.fog_shader); + + storage->shader_set_code(sky_scene_state.fog_shader, "shader_type sky; uniform vec4 clear_color; void fragment() { COLOR = clear_color.rgb; } \n"); + sky_scene_state.fog_material = storage->material_allocate(); + storage->material_initialize(sky_scene_state.fog_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.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK)); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 1; + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE)); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 2; + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::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); + } +} + +void RendererSceneSkyRD::setup(RendererSceneEnvironmentRD *p_env, RID p_render_buffers, const CameraMatrix &p_projection, const Transform &p_transform, const Size2i p_screen_size, RendererSceneRenderRD *p_scene_render) { + ERR_FAIL_COND(!p_env); // I guess without an environment we also can't have a sky... + + SkyMaterialData *material = nullptr; + Sky *sky = get_sky(p_env->sky); + + RID sky_material; + + SkyShaderData *shader_data = nullptr; + + RS::EnvironmentBG background = p_env->background; + + if (!(background == RS::ENV_BG_CLEAR_COLOR || background == RS::ENV_BG_COLOR) || sky) { + // !BAS! Possibly silently fail here, we now get error spam when you select sky as the background but haven't setup the sky yet. + ERR_FAIL_COND(!sky); + sky_material = sky_get_material(p_env->sky); + + if (sky_material.is_valid()) { + material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); + if (!material || !material->shader_data->valid) { + material = nullptr; + } + } + + if (!material) { + sky_material = sky_shader.default_material; + material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); + } + + ERR_FAIL_COND(!material); + + shader_data = material->shader_data; + + ERR_FAIL_COND(!shader_data); + } + + if (sky) { + // 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) { + // Check whether the directional_light_buffer changes + bool light_data_dirty = false; + + 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; + } + } + 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_scene_render->render_buffers_has_volumetric_fog(p_render_buffers)) { + sky_scene_state.ubo.volumetric_fog_enabled = true; + + float fog_end = p_scene_render->render_buffers_get_volumetric_fog_end(p_render_buffers); + 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 = p_scene_render->render_buffers_get_volumetric_fog_detail_spread(p_render_buffers); //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; + } + } + + RID fog_uniform_set = p_scene_render->render_buffers_get_volumetric_fog_sky_uniform_set(p_render_buffers); + + if (fog_uniform_set != RID()) { + sky_scene_state.fog_uniform_set = fog_uniform_set; + } else { + sky_scene_state.fog_uniform_set = sky_scene_state.default_fog_uniform_set; + } + } + + sky_scene_state.ubo.z_far = p_projection.get_z_far(); + sky_scene_state.ubo.fog_enabled = p_env->fog_enabled; + sky_scene_state.ubo.fog_density = p_env->fog_density; + sky_scene_state.ubo.fog_aerial_perspective = p_env->fog_aerial_perspective; + Color fog_color = p_env->fog_light_color.to_linear(); + float fog_energy = p_env->fog_light_energy; + 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 = p_env->fog_sun_scatter; + + RD::get_singleton()->buffer_update(sky_scene_state.uniform_buffer, 0, sizeof(SkySceneState::UBO), &sky_scene_state.ubo); +} + +void RendererSceneSkyRD::update(RendererSceneEnvironmentRD *p_env, const CameraMatrix &p_projection, const Transform &p_transform, double p_time) { + ERR_FAIL_COND(!p_env); + + Sky *sky = get_sky(p_env->sky); + ERR_FAIL_COND(!sky); + + RID sky_material = sky_get_material(p_env->sky); + + SkyMaterialData *material = nullptr; + + if (sky_material.is_valid()) { + material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); + if (!material || !material->shader_data->valid) { + material = nullptr; + } + } + + if (!material) { + sky_material = sky_shader.default_material; + material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); + } + + ERR_FAIL_COND(!material); + + SkyShaderData *shader_data = material->shader_data; + + ERR_FAIL_COND(!shader_data); + + float multiplier = p_env->bg_energy; + + 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) + }; + + CameraMatrix cm; + cm.set_perspective(90, 1, 0.01, 10.0); + CameraMatrix correction; + correction.set_depth_correction(true); + cm = correction * cm; + + if (shader_data->uses_quarter_res) { + 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++) { + Transform local_view; + local_view.set_look_at(Vector3(0, 0, 0), view_normals[i], view_up[i]); + RID texture_uniform_set = sky->get_textures(storage, 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); + storage->get_effects()->render_sky(cubemap_draw_list, p_time, sky->reflection.layers[0].mipmaps[2].framebuffers[i], sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, cm, local_view.basis, multiplier, p_transform.origin); + RD::get_singleton()->draw_list_end(); + } + } + + if (shader_data->uses_half_res) { + 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++) { + Transform local_view; + local_view.set_look_at(Vector3(0, 0, 0), view_normals[i], view_up[i]); + RID texture_uniform_set = sky->get_textures(storage, 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); + storage->get_effects()->render_sky(cubemap_draw_list, p_time, sky->reflection.layers[0].mipmaps[1].framebuffers[i], sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, cm, local_view.basis, multiplier, p_transform.origin); + RD::get_singleton()->draw_list_end(); + } + } + + RD::DrawListID cubemap_draw_list; + PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_CUBEMAP]; + + for (int i = 0; i < 6; i++) { + Transform local_view; + local_view.set_look_at(Vector3(0, 0, 0), view_normals[i], view_up[i]); + RID texture_uniform_set = sky->get_textures(storage, 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); + storage->get_effects()->render_sky(cubemap_draw_list, p_time, sky->reflection.layers[0].mipmaps[0].framebuffers[i], sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, cm, local_view.basis, multiplier, p_transform.origin); + RD::get_singleton()->draw_list_end(); + } + + if (sky_mode == RS::SKY_MODE_REALTIME) { + sky->reflection.create_reflection_fast_filter(storage, sky_use_cubemap_array); + if (sky_use_cubemap_array) { + sky->reflection.update_reflection_mipmaps(storage, 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(storage, sky_use_cubemap_array, 10, i, sky_ggx_samples_quality); + } + if (sky_use_cubemap_array) { + sky->reflection.update_reflection_mipmaps(storage, 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(storage, 0, 1); + } + } + sky->processing_layer = 1; + } + + sky->reflection.dirty = false; + + } else { + if (sky_mode == RS::SKY_MODE_INCREMENTAL && sky->processing_layer < max_processing_layer) { + sky->reflection.create_reflection_importance_sample(storage, sky_use_cubemap_array, 10, sky->processing_layer, sky_ggx_samples_quality); + + if (sky_use_cubemap_array) { + sky->reflection.update_reflection_mipmaps(storage, sky->processing_layer, sky->processing_layer + 1); + } + + sky->processing_layer++; + } + } +} + +void RendererSceneSkyRD::draw(RendererSceneEnvironmentRD *p_env, bool p_can_continue_color, bool p_can_continue_depth, RID p_fb, const CameraMatrix &p_projection, const Transform &p_transform, double p_time) { + ERR_FAIL_COND(!p_env); + + Sky *sky = get_sky(p_env->sky); + ERR_FAIL_COND(!sky); + + SkyMaterialData *material = nullptr; + RID sky_material; + + RS::EnvironmentBG background = p_env->background; + + if (!(background == RS::ENV_BG_CLEAR_COLOR || background == RS::ENV_BG_COLOR) || sky) { + ERR_FAIL_COND(!sky); + sky_material = sky_get_material(p_env->sky); + + if (sky_material.is_valid()) { + material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); + if (!material || !material->shader_data->valid) { + material = nullptr; + } + } + + if (!material) { + sky_material = sky_shader.default_material; + material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); + } + } + + if (background == RS::ENV_BG_CLEAR_COLOR || background == RS::ENV_BG_COLOR) { + sky_material = sky_scene_state.fog_material; + material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); + } + + ERR_FAIL_COND(!material); + + SkyShaderData *shader_data = material->shader_data; + + ERR_FAIL_COND(!shader_data); + + Basis sky_transform = p_env->sky_orientation; + sky_transform.invert(); + + float multiplier = p_env->bg_energy; + float custom_fov = p_env->sky_custom_fov; + // Camera + CameraMatrix camera; + + if (custom_fov) { + float near_plane = p_projection.get_z_near(); + float far_plane = p_projection.get_z_far(); + float aspect = p_projection.get_aspect(); + + camera.set_perspective(custom_fov, aspect, near_plane, far_plane); + + } else { + camera = p_projection; + } + + sky_transform = p_transform.basis * sky_transform; + + if (shader_data->uses_quarter_res) { + PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_QUARTER_RES]; + + RID texture_uniform_set = sky->get_textures(storage, 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); + storage->get_effects()->render_sky(draw_list, p_time, sky->quarter_res_framebuffer, sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, camera, sky_transform, multiplier, p_transform.origin); + RD::get_singleton()->draw_list_end(); + } + + if (shader_data->uses_half_res) { + PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_HALF_RES]; + + RID texture_uniform_set = sky->get_textures(storage, 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); + storage->get_effects()->render_sky(draw_list, p_time, sky->half_res_framebuffer, sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, camera, sky_transform, multiplier, p_transform.origin); + RD::get_singleton()->draw_list_end(); + } + + PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_BACKGROUND]; + + RID texture_uniform_set; + if (sky) { + texture_uniform_set = sky->get_textures(storage, 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); + storage->get_effects()->render_sky(draw_list, p_time, p_fb, sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, camera, sky_transform, multiplier, p_transform.origin); + RD::get_singleton()->draw_list_end(); +} + +void RendererSceneSkyRD::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 RendererSceneSkyRD::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 = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + 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); + + } else { + //regular cubemap, lower quality (aliasing, less memory) + RD::TextureFormat tf; + tf.array_layers = 6; + tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + 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_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 = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + 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 = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + 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 RendererSceneSkyRD::sky_get_material(RID p_sky) const { + Sky *sky = get_sky(p_sky); + ERR_FAIL_COND_V(!sky, RID()); + + return sky->material; +} + +RID RendererSceneSkyRD::allocate_sky_rid() { + return sky_owner.allocate_rid(); +} + +void RendererSceneSkyRD::initialize_sky_rid(RID p_rid) { + sky_owner.initialize_rid(p_rid, Sky()); +} + +RendererSceneSkyRD::Sky *RendererSceneSkyRD::get_sky(RID p_sky) const { + return sky_owner.getornull(p_sky); +} + +void RendererSceneSkyRD::free_sky(RID p_sky) { + Sky *sky = get_sky(p_sky); + ERR_FAIL_COND(!sky); + + sky->free(storage); + sky_owner.free(p_sky); +} + +void RendererSceneSkyRD::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 RendererSceneSkyRD::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 RendererSceneSkyRD::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> RendererSceneSkyRD::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(storage, p_energy, p_bake_irradiance ? roughness_layers : 0, p_size); +} + +RID RendererSceneSkyRD::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/renderer_scene_sky_rd.h b/servers/rendering/renderer_rd/renderer_scene_sky_rd.h new file mode 100644 index 0000000000..73390a586b --- /dev/null +++ b/servers/rendering/renderer_rd/renderer_scene_sky_rd.h @@ -0,0 +1,292 @@ +/*************************************************************************/ +/* renderer_scene_sky_rd.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2021 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2021 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 RENDERING_SERVER_SCENE_SKY_RD_H +#define RENDERING_SERVER_SCENE_SKY_RD_H + +#include "core/templates/rid_owner.h" +#include "servers/rendering/renderer_compositor.h" +#include "servers/rendering/renderer_rd/renderer_scene_environment_rd.h" +#include "servers/rendering/renderer_rd/renderer_storage_rd.h" +#include "servers/rendering/renderer_rd/shaders/sky.glsl.gen.h" +#include "servers/rendering/renderer_scene_render.h" +#include "servers/rendering/rendering_device.h" + +// Forward declare RendererSceneRenderRD so we can pass it into some of our methods, these classes are pretty tightly bound +class RendererSceneRenderRD; + +class RendererSceneSkyRD { +private: + RendererStorageRD *storage; + +public: + enum SkySet { + SKY_SET_UNIFORMS, + SKY_SET_MATERIAL, + SKY_SET_TEXTURES, + SKY_SET_FOG, + SKY_SET_MAX + }; + + 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_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]; + }; + + struct SkySceneState { + struct UBO { + uint32_t volumetric_fog_enabled; + float volumetric_fog_inv_length; + float volumetric_fog_detail_spread; + + float fog_aerial_perspective; + + float fog_light_color[3]; + float fog_sun_scatter; + + uint32_t fog_enabled; + float fog_density; + + float z_far; + uint32_t directional_light_count; + }; + + UBO ubo; + + SkyDirectionalLightData *directional_lights; + SkyDirectionalLightData *last_frame_directional_lights; + 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; + }; + 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); + void create_reflection_fast_filter(RendererStorageRD *p_storage, bool p_use_arrays); + void create_reflection_importance_sample(RendererStorageRD *p_storage, bool p_use_arrays, int p_cube_side, int p_base_layer, uint32_t p_sky_ggx_samples_quality); + void update_reflection_mipmaps(RendererStorageRD *p_storage, int p_start, int p_end); + }; + + struct SkyShaderData : public RendererStorageRD::ShaderData { + bool valid; + RID version; + + PipelineCacheRD pipelines[SKY_VERSION_MAX]; + Map<StringName, ShaderLanguage::ShaderNode::Uniform> uniforms; + Vector<ShaderCompilerRD::GeneratedCode::Texture> texture_uniforms; + + Vector<uint32_t> ubo_offsets; + uint32_t ubo_size; + + String path; + String code; + Map<StringName, RID> default_texture_params; + + bool uses_time; + bool uses_position; + bool uses_half_res; + bool uses_quarter_res; + bool uses_light; + + virtual void set_code(const String &p_Code); + virtual void set_default_texture_param(const StringName &p_name, RID p_texture); + virtual void get_param_list(List<PropertyInfo> *p_param_list) const; + virtual void get_instance_param_list(List<RendererStorage::InstanceShaderParam> *p_param_list) const; + virtual bool is_param_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(); + }; + + /* Sky shader */ + + struct SkyShader { + SkyShaderRD shader; + ShaderCompilerRD compiler; + + RID default_shader; + RID default_material; + RID default_shader_rd; + } sky_shader; + + struct SkyMaterialData : public RendererStorageRD::MaterialData { + uint64_t last_frame; + SkyShaderData *shader_data; + RID uniform_buffer; + RID uniform_set; + Vector<RID> texture_cache; + Vector<uint8_t> ubo_data; + bool uniform_set_updated; + + virtual void set_render_priority(int p_priority) {} + virtual void set_next_pass(RID p_pass) {} + virtual void update_parameters(const Map<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; + + //State to track when radiance cubemap needs updating + SkyMaterialData *prev_material; + Vector3 prev_position; + float prev_time; + + void free(RendererStorageRD *p_storage); + + RID get_textures(RendererStorageRD *p_storage, 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(RendererStorageRD *p_storage, 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; + + RendererStorageRD::ShaderData *_create_sky_shader_func(); + static RendererStorageRD::ShaderData *_create_sky_shader_funcs(); + + RendererStorageRD::MaterialData *_create_sky_material_func(SkyShaderData *p_shader); + static RendererStorageRD::MaterialData *_create_sky_material_funcs(RendererStorageRD::ShaderData *p_shader); + + RendererSceneSkyRD(); + + void init(RendererStorageRD *p_storage); + + void setup(RendererSceneEnvironmentRD *p_env, RID p_render_buffers, const CameraMatrix &p_projection, const Transform &p_transform, const Size2i p_screen_size, RendererSceneRenderRD *p_scene_render); + void update(RendererSceneEnvironmentRD *p_env, const CameraMatrix &p_projection, const Transform &p_transform, double p_time); + void draw(RendererSceneEnvironmentRD *p_env, bool p_can_continue_color, bool p_can_continue_depth, RID p_fb, const CameraMatrix &p_projection, const Transform &p_transform, double p_time); + + void invalidate_sky(Sky *p_sky); + void update_dirty_skys(); + + RID sky_get_material(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); + + RID sky_get_radiance_texture_rd(RID p_sky) const; +}; + +#endif /* RENDERING_SERVER_SCENE_SKY_RD_H */ diff --git a/servers/rendering/renderer_rd/renderer_storage_rd.cpp b/servers/rendering/renderer_rd/renderer_storage_rd.cpp index bf7237cad0..ba5ace8f31 100644 --- a/servers/rendering/renderer_rd/renderer_storage_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_storage_rd.cpp @@ -36,6 +36,10 @@ #include "renderer_compositor_rd.h" #include "servers/rendering/shader_language.h" +bool RendererStorageRD::can_create_resources_async() const { + return true; +} + Ref<Image> RendererStorageRD::_validate_texture_format(const Ref<Image> &p_image, TextureToRDFormat &r_format) { Ref<Image> image = p_image->duplicate(); @@ -535,9 +539,13 @@ Ref<Image> RendererStorageRD::_validate_texture_format(const Ref<Image> &p_image return image; } -RID RendererStorageRD::texture_2d_create(const Ref<Image> &p_image) { - ERR_FAIL_COND_V(p_image.is_null(), RID()); - ERR_FAIL_COND_V(p_image->is_empty(), RID()); +RID RendererStorageRD::texture_allocate() { + return texture_owner.allocate_rid(); +} + +void RendererStorageRD::texture_2d_initialize(RID p_texture, const Ref<Image> &p_image) { + ERR_FAIL_COND(p_image.is_null()); + ERR_FAIL_COND(p_image->is_empty()); TextureToRDFormat ret_format; Ref<Image> image = _validate_texture_format(p_image, ret_format); @@ -585,13 +593,13 @@ RID RendererStorageRD::texture_2d_create(const Ref<Image> &p_image) { 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_V(texture.rd_texture.is_null(), RID()); + 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_V(texture.rd_texture_srgb.is_null(), RID()); + ERR_FAIL_COND(texture.rd_texture_srgb.is_null()); } } @@ -602,14 +610,14 @@ RID RendererStorageRD::texture_2d_create(const Ref<Image> &p_image) { texture.rd_view = rd_view; texture.is_proxy = false; - return texture_owner.make_rid(texture); + texture_owner.initialize_rid(p_texture, texture); } -RID RendererStorageRD::texture_2d_layered_create(const Vector<Ref<Image>> &p_layers, RS::TextureLayeredType p_layered_type) { - ERR_FAIL_COND_V(p_layers.size() == 0, RID()); +void RendererStorageRD::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_V(p_layered_type == RS::TEXTURE_LAYERED_CUBEMAP && p_layers.size() != 6, RID()); - ERR_FAIL_COND_V(p_layered_type == RS::TEXTURE_LAYERED_CUBEMAP_ARRAY && (p_layers.size() < 6 || (p_layers.size() % 6) != 0), RID()); + 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; @@ -620,7 +628,7 @@ RID RendererStorageRD::texture_2d_layered_create(const Vector<Ref<Image>> &p_lay Image::Format valid_format = Image::FORMAT_MAX; for (int i = 0; i < p_layers.size(); i++) { - ERR_FAIL_COND_V(p_layers[i]->is_empty(), RID()); + ERR_FAIL_COND(p_layers[i]->is_empty()); if (i == 0) { valid_width = p_layers[i]->get_width(); @@ -628,10 +636,10 @@ RID RendererStorageRD::texture_2d_layered_create(const Vector<Ref<Image>> &p_lay valid_format = p_layers[i]->get_format(); valid_mipmaps = p_layers[i]->has_mipmaps(); } else { - ERR_FAIL_COND_V(p_layers[i]->get_width() != valid_width, RID()); - ERR_FAIL_COND_V(p_layers[i]->get_height() != valid_height, RID()); - ERR_FAIL_COND_V(p_layers[i]->get_format() != valid_format, RID()); - ERR_FAIL_COND_V(p_layers[i]->has_mipmaps() != valid_mipmaps, RID()); + 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)); @@ -695,13 +703,13 @@ RID RendererStorageRD::texture_2d_layered_create(const Vector<Ref<Image>> &p_lay data_slices.push_back(data); } texture.rd_texture = RD::get_singleton()->texture_create(rd_format, rd_view, data_slices); - ERR_FAIL_COND_V(texture.rd_texture.is_null(), RID()); + 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_V(texture.rd_texture_srgb.is_null(), RID()); + ERR_FAIL_COND(texture.rd_texture_srgb.is_null()); } } @@ -712,14 +720,14 @@ RID RendererStorageRD::texture_2d_layered_create(const Vector<Ref<Image>> &p_lay texture.rd_view = rd_view; texture.is_proxy = false; - return texture_owner.make_rid(texture); + texture_owner.initialize_rid(p_texture, texture); } -RID RendererStorageRD::texture_3d_create(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_V(p_data.size() == 0, RID()); +void RendererStorageRD::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_V_MSG(RID(), Image::get_3d_image_validation_error_text(verr)); + ERR_FAIL_MSG(Image::get_3d_image_validation_error_text(verr)); } TextureToRDFormat ret_format; @@ -811,13 +819,13 @@ RID RendererStorageRD::texture_3d_create(Image::Format p_format, int p_width, in 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_V(texture.rd_texture.is_null(), RID()); + 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_V(texture.rd_texture_srgb.is_null(), RID()); + ERR_FAIL_COND(texture.rd_texture_srgb.is_null()); } } @@ -828,12 +836,12 @@ RID RendererStorageRD::texture_3d_create(Image::Format p_format, int p_width, in texture.rd_view = rd_view; texture.is_proxy = false; - return texture_owner.make_rid(texture); + texture_owner.initialize_rid(p_texture, texture); } -RID RendererStorageRD::texture_proxy_create(RID p_base) { +void RendererStorageRD::texture_proxy_initialize(RID p_texture, RID p_base) { Texture *tex = texture_owner.getornull(p_base); - ERR_FAIL_COND_V(!tex, RID()); + ERR_FAIL_COND(!tex); Texture proxy_tex = *tex; proxy_tex.rd_view.format_override = tex->rd_format; @@ -847,11 +855,9 @@ RID RendererStorageRD::texture_proxy_create(RID p_base) { proxy_tex.is_proxy = true; proxy_tex.proxies.clear(); - RID rid = texture_owner.make_rid(proxy_tex); + texture_owner.initialize_rid(p_texture, proxy_tex); - tex->proxies.push_back(rid); - - return rid; + tex->proxies.push_back(p_texture); } void RendererStorageRD::_texture_2d_update(RID p_texture, const Ref<Image> &p_image, int p_layer, bool p_immediate) { @@ -873,7 +879,7 @@ void RendererStorageRD::_texture_2d_update(RID p_texture, const Ref<Image> &p_im TextureToRDFormat f; Ref<Image> validated = _validate_texture_format(p_image, f); - RD::get_singleton()->texture_update(tex->rd_texture, p_layer, validated->get_data(), !p_immediate); + RD::get_singleton()->texture_update(tex->rd_texture, p_layer, validated->get_data()); } void RendererStorageRD::texture_2d_update_immediate(RID p_texture, const Ref<Image> &p_image, int p_layer) { @@ -918,7 +924,7 @@ void RendererStorageRD::texture_3d_update(RID p_texture, const Vector<Ref<Image> } } - RD::get_singleton()->texture_update(tex->rd_texture, 0, all_data, true); + RD::get_singleton()->texture_update(tex->rd_texture, 0, all_data); } void RendererStorageRD::texture_proxy_update(RID p_texture, RID p_proxy_to) { @@ -961,7 +967,7 @@ void RendererStorageRD::texture_proxy_update(RID p_texture, RID p_proxy_to) { } //these two APIs can be used together or in combination with the others. -RID RendererStorageRD::texture_2d_placeholder_create() { +void RendererStorageRD::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; @@ -974,10 +980,10 @@ RID RendererStorageRD::texture_2d_placeholder_create() { } } - return texture_2d_create(image); + texture_2d_initialize(p_texture, image); } -RID RendererStorageRD::texture_2d_layered_placeholder_create(RS::TextureLayeredType p_layered_type) { +void RendererStorageRD::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; @@ -1000,10 +1006,10 @@ RID RendererStorageRD::texture_2d_layered_placeholder_create(RS::TextureLayeredT } } - return texture_2d_layered_create(images, p_layered_type); + texture_2d_layered_initialize(p_texture, images, p_layered_type); } -RID RendererStorageRD::texture_3d_placeholder_create() { +void RendererStorageRD::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; @@ -1022,7 +1028,7 @@ RID RendererStorageRD::texture_3d_placeholder_create() { images.push_back(image); } - return texture_3d_create(Image::FORMAT_RGBA8, 4, 4, 4, false, images); + texture_3d_initialize(p_texture, Image::FORMAT_RGBA8, 4, 4, 4, false, images); } Ref<Image> RendererStorageRD::texture_2d_get(RID p_texture) const { @@ -1223,8 +1229,11 @@ RendererStorageRD::CanvasTexture::~CanvasTexture() { clear_sets(); } -RID RendererStorageRD::canvas_texture_create() { - return canvas_texture_owner.make_rid(memnew(CanvasTexture)); +RID RendererStorageRD::canvas_texture_allocate() { + return canvas_texture_owner.allocate_rid(); +} +void RendererStorageRD::canvas_texture_initialize(RID p_rid) { + canvas_texture_owner.initialize_rid(p_rid, memnew(CanvasTexture)); } void RendererStorageRD::canvas_texture_set_channel(RID p_canvas_texture, RS::CanvasTextureChannel p_channel, RID p_texture) { @@ -1365,12 +1374,15 @@ bool RendererStorageRD::canvas_texture_get_uniform_set(RID p_texture, RS::Canvas /* SHADER API */ -RID RendererStorageRD::shader_create() { +RID RendererStorageRD::shader_allocate() { + return shader_owner.allocate_rid(); +} +void RendererStorageRD::shader_initialize(RID p_rid) { Shader shader; shader.data = nullptr; shader.type = SHADER_TYPE_MAX; - return shader_owner.make_rid(shader); + shader_owner.initialize_rid(p_rid, shader); } void RendererStorageRD::shader_set_code(RID p_shader, const String &p_code) { @@ -1499,9 +1511,21 @@ void RendererStorageRD::shader_set_data_request_function(ShaderType p_shader_typ shader_data_request_func[p_shader_type] = p_function; } +RS::ShaderNativeSourceCode RendererStorageRD::shader_get_native_source_code(RID p_shader) const { + Shader *shader = shader_owner.getornull(p_shader); + ERR_FAIL_COND_V(!shader, RS::ShaderNativeSourceCode()); + if (shader->data) { + return shader->data->get_native_source_code(); + } + return RS::ShaderNativeSourceCode(); +} + /* COMMON MATERIAL API */ -RID RendererStorageRD::material_create() { +RID RendererStorageRD::material_allocate() { + return material_owner.allocate_rid(); +} +void RendererStorageRD::material_initialize(RID p_rid) { Material material; material.data = nullptr; material.shader = nullptr; @@ -1511,12 +1535,8 @@ RID RendererStorageRD::material_create() { material.uniform_dirty = false; material.texture_dirty = false; material.priority = 0; - RID id = material_owner.make_rid(material); - { - Material *material_ptr = material_owner.getornull(id); - material_ptr->self = id; - } - return id; + material.self = p_rid; + material_owner.initialize_rid(p_rid, material); } void RendererStorageRD::_material_queue_update(Material *material, bool p_uniform, bool p_texture) { @@ -2390,8 +2410,11 @@ void RendererStorageRD::_update_queued_materials() { /* MESH API */ -RID RendererStorageRD::mesh_create() { - return mesh_owner.make_rid(Mesh()); +RID RendererStorageRD::mesh_allocate() { + return mesh_owner.allocate_rid(); +} +void RendererStorageRD::mesh_initialize(RID p_rid) { + mesh_owner.initialize_rid(p_rid, Mesh()); } void RendererStorageRD::mesh_set_blend_shape_count(RID p_mesh, int p_blend_shape_count) { @@ -2600,6 +2623,12 @@ void RendererStorageRD::mesh_add_surface(RID p_mesh, const RS::SurfaceData &p_su mesh->dependency.changed_notify(DEPENDENCY_CHANGED_MESH); + for (Set<Mesh *>::Element *E = mesh->shadow_owners.front(); E; E = E->next()) { + Mesh *shadow_owner = E->get(); + shadow_owner->shadow_mesh = RID(); + shadow_owner->dependency.changed_notify(DEPENDENCY_CHANGED_MESH); + } + mesh->material_cache.clear(); } @@ -2815,6 +2844,25 @@ AABB RendererStorageRD::mesh_get_aabb(RID p_mesh, RID p_skeleton) { return aabb; } +void RendererStorageRD::mesh_set_shadow_mesh(RID p_mesh, RID p_shadow_mesh) { + Mesh *mesh = mesh_owner.getornull(p_mesh); + ERR_FAIL_COND(!mesh); + + Mesh *shadow_mesh = mesh_owner.getornull(mesh->shadow_mesh); + if (shadow_mesh) { + shadow_mesh->shadow_owners.erase(mesh); + } + mesh->shadow_mesh = p_shadow_mesh; + + shadow_mesh = mesh_owner.getornull(mesh->shadow_mesh); + + if (shadow_mesh) { + shadow_mesh->shadow_owners.insert(mesh); + } + + mesh->dependency.changed_notify(DEPENDENCY_CHANGED_MESH); +} + void RendererStorageRD::mesh_clear(RID p_mesh) { Mesh *mesh = mesh_owner.getornull(p_mesh); ERR_FAIL_COND(!mesh); @@ -2862,6 +2910,12 @@ void RendererStorageRD::mesh_clear(RID p_mesh) { } mesh->has_bone_weights = false; mesh->dependency.changed_notify(DEPENDENCY_CHANGED_MESH); + + for (Set<Mesh *>::Element *E = mesh->shadow_owners.front(); E; E = E->next()) { + Mesh *shadow_owner = E->get(); + shadow_owner->shadow_mesh = RID(); + shadow_owner->dependency.changed_notify(DEPENDENCY_CHANGED_MESH); + } } bool RendererStorageRD::mesh_needs_instance(RID p_mesh, bool p_has_skeleton) { @@ -3004,7 +3058,7 @@ void RendererStorageRD::update_mesh_instances() { 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(), true); + 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; @@ -3058,7 +3112,7 @@ void RendererStorageRD::update_mesh_instances() { 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, 64, 1, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, push_constant.vertex_count, 1, 1); } mi->dirty = false; @@ -3258,11 +3312,14 @@ void RendererStorageRD::_mesh_surface_generate_version_for_input_mask(Mesh::Surf ////////////////// MULTIMESH -RID RendererStorageRD::multimesh_create() { - return multimesh_owner.make_rid(MultiMesh()); +RID RendererStorageRD::multimesh_allocate() { + return multimesh_owner.allocate_rid(); +} +void RendererStorageRD::multimesh_initialize(RID p_rid) { + multimesh_owner.initialize_rid(p_rid, MultiMesh()); } -void RendererStorageRD::multimesh_allocate(RID p_multimesh, int p_instances, RS::MultimeshTransformFormat p_transform_format, bool p_use_colors, bool p_use_custom_data) { +void RendererStorageRD::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.getornull(p_multimesh); ERR_FAIL_COND(!multimesh); @@ -3672,7 +3729,7 @@ void RendererStorageRD::multimesh_set_buffer(RID p_multimesh, const Vector<float { const float *r = p_buffer.ptr(); - RD::get_singleton()->buffer_update(multimesh->buffer, 0, p_buffer.size() * sizeof(float), r, false); + RD::get_singleton()->buffer_update(multimesh->buffer, 0, p_buffer.size() * sizeof(float), r); multimesh->buffer_set = true; } @@ -3771,14 +3828,14 @@ void RendererStorageRD::_update_dirty_multimeshes() { if (multimesh->data_cache_used_dirty_regions > 32 || multimesh->data_cache_used_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, 0, MIN(visible_region_count * region_size, multimesh->instances * multimesh->stride_cache * sizeof(float)), data, false); + RD::get_singleton()->buffer_update(multimesh->buffer, 0, MIN(visible_region_count * region_size, multimesh->instances * multimesh->stride_cache * 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]) { uint64_t offset = i * region_size; uint64_t size = multimesh->stride_cache * multimesh->instances * sizeof(float); - RD::get_singleton()->buffer_update(multimesh->buffer, offset, MIN(region_size, size - offset), &data[i * region_size], false); + RD::get_singleton()->buffer_update(multimesh->buffer, offset, MIN(region_size, size - offset), &data[i * region_size]); } } } @@ -3809,8 +3866,11 @@ void RendererStorageRD::_update_dirty_multimeshes() { /* PARTICLES */ -RID RendererStorageRD::particles_create() { - return particles_owner.make_rid(Particles()); +RID RendererStorageRD::particles_allocate() { + return particles_owner.allocate_rid(); +} +void RendererStorageRD::particles_initialize(RID p_rid) { + particles_owner.initialize_rid(p_rid, Particles()); } void RendererStorageRD::particles_set_emitting(RID p_particles, bool p_emitting) { @@ -4469,7 +4529,7 @@ void RendererStorageRD::_particles_process(Particles *p_particles, float p_delta 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, true); + 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) { @@ -4487,13 +4547,13 @@ void RendererStorageRD::_particles_process(Particles *p_particles, float p_delta } 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, true); + 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; - RD::get_singleton()->buffer_update(p_particles->frame_params_buffer, 0, sizeof(ParticlesFrameParams), &frame_params, true); + RD::get_singleton()->buffer_update(p_particles->frame_params_buffer, 0, sizeof(ParticlesFrameParams), &frame_params); ParticlesMaterialData *m = (ParticlesMaterialData *)material_get_data(p_particles->process_material, SHADER_TYPE_PARTICLES); if (!m) { @@ -4515,7 +4575,7 @@ void RendererStorageRD::_particles_process(Particles *p_particles, float p_delta RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ParticlesShader::PushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_particles->amount, 1, 1, 64, 1, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_particles->amount, 1, 1); RD::get_singleton()->compute_list_end(); } @@ -4569,7 +4629,7 @@ void RendererStorageRD::particles_set_view_axis(RID p_particles, const Vector3 & RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->particles_sort_uniform_set, 1); 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, 64, 1, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, particles->amount, 1, 1); RD::get_singleton()->compute_list_end(); @@ -4581,7 +4641,7 @@ void RendererStorageRD::particles_set_view_axis(RID p_particles, const Vector3 & RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->particles_sort_uniform_set, 1); 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, 64, 1, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, particles->amount, 1, 1); RD::get_singleton()->compute_list_end(); } @@ -4688,7 +4748,7 @@ void RendererStorageRD::update_particles() { RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->particles_copy_uniform_set, 0); 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, 64, 1, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, particles->amount, 1, 1); RD::get_singleton()->compute_list_end(); } @@ -4823,6 +4883,10 @@ Variant RendererStorageRD::ParticlesShaderData::get_default_parameter(const Stri return Variant(); } +RS::ShaderNativeSourceCode RendererStorageRD::ParticlesShaderData::get_native_source_code() const { + return base_singleton->particles_shader.shader.version_get_native_source_code(version); +} + RendererStorageRD::ParticlesShaderData::ParticlesShaderData() { valid = false; } @@ -4940,8 +5004,11 @@ RendererStorageRD::MaterialData *RendererStorageRD::_create_particles_material_f /* PARTICLES COLLISION API */ -RID RendererStorageRD::particles_collision_create() { - return particles_collision_owner.make_rid(ParticlesCollision()); +RID RendererStorageRD::particles_collision_allocate() { + return particles_collision_owner.allocate_rid(); +} +void RendererStorageRD::particles_collision_initialize(RID p_rid) { + particles_collision_owner.initialize_rid(p_rid, ParticlesCollision()); } RID RendererStorageRD::particles_collision_get_heightfield_framebuffer(RID p_particles_collision) const { @@ -5054,6 +5121,7 @@ void RendererStorageRD::particles_collision_height_field_update(RID p_particles_ void RendererStorageRD::particles_collision_set_height_field_resolution(RID p_particles_collision, RS::ParticlesCollisionHeightfieldResolution p_resolution) { ParticlesCollision *particles_collision = particles_collision_owner.getornull(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; @@ -5120,8 +5188,11 @@ void RendererStorageRD::particles_collision_instance_set_active(RID p_collision_ /* SKELETON API */ -RID RendererStorageRD::skeleton_create() { - return skeleton_owner.make_rid(Skeleton()); +RID RendererStorageRD::skeleton_allocate() { + return skeleton_owner.allocate_rid(); +} +void RendererStorageRD::skeleton_initialize(RID p_rid) { + skeleton_owner.initialize_rid(p_rid, Skeleton()); } void RendererStorageRD::_skeleton_make_dirty(Skeleton *skeleton) { @@ -5132,7 +5203,7 @@ void RendererStorageRD::_skeleton_make_dirty(Skeleton *skeleton) { } } -void RendererStorageRD::skeleton_allocate(RID p_skeleton, int p_bones, bool p_2d_skeleton) { +void RendererStorageRD::skeleton_allocate_data(RID p_skeleton, int p_bones, bool p_2d_skeleton) { Skeleton *skeleton = skeleton_owner.getornull(p_skeleton); ERR_FAIL_COND(!skeleton); ERR_FAIL_COND(p_bones < 0); @@ -5288,7 +5359,7 @@ void RendererStorageRD::_update_dirty_skeletons() { Skeleton *skeleton = skeleton_dirty_list; if (skeleton->size) { - RD::get_singleton()->buffer_update(skeleton->buffer, 0, skeleton->data.size() * sizeof(float), skeleton->data.ptr(), false); + RD::get_singleton()->buffer_update(skeleton->buffer, 0, skeleton->data.size() * sizeof(float), skeleton->data.ptr()); } skeleton_dirty_list = skeleton->dirty_list; @@ -5306,7 +5377,7 @@ void RendererStorageRD::_update_dirty_skeletons() { /* LIGHT */ -RID RendererStorageRD::light_create(RS::LightType p_type) { +void RendererStorageRD::_light_initialize(RID p_light, RS::LightType p_type) { Light light; light.type = p_type; @@ -5327,10 +5398,31 @@ RID RendererStorageRD::light_create(RS::LightType p_type) { light.param[RS::LIGHT_PARAM_SHADOW_BIAS] = 0.02; light.param[RS::LIGHT_PARAM_SHADOW_BLUR] = 0; light.param[RS::LIGHT_PARAM_SHADOW_PANCAKE_SIZE] = 20.0; - light.param[RS::LIGHT_PARAM_SHADOW_VOLUMETRIC_FOG_FADE] = 1.0; + light.param[RS::LIGHT_PARAM_SHADOW_VOLUMETRIC_FOG_FADE] = 0.1; light.param[RS::LIGHT_PARAM_TRANSMITTANCE_BIAS] = 0.05; - return light_owner.make_rid(light); + light_owner.initialize_rid(p_light, light); +} + +RID RendererStorageRD::directional_light_allocate() { + return light_owner.allocate_rid(); +} +void RendererStorageRD::directional_light_initialize(RID p_light) { + _light_initialize(p_light, RS::LIGHT_DIRECTIONAL); +} + +RID RendererStorageRD::omni_light_allocate() { + return light_owner.allocate_rid(); +} +void RendererStorageRD::omni_light_initialize(RID p_light) { + _light_initialize(p_light, RS::LIGHT_OMNI); +} + +RID RendererStorageRD::spot_light_allocate() { + return light_owner.allocate_rid(); +} +void RendererStorageRD::spot_light_initialize(RID p_light) { + _light_initialize(p_light, RS::LIGHT_SPOT); } void RendererStorageRD::light_set_color(RID p_light, const Color &p_color) { @@ -5568,8 +5660,11 @@ AABB RendererStorageRD::light_get_aabb(RID p_light) const { /* REFLECTION PROBE */ -RID RendererStorageRD::reflection_probe_create() { - return reflection_probe_owner.make_rid(ReflectionProbe()); +RID RendererStorageRD::reflection_probe_allocate() { + return reflection_probe_owner.allocate_rid(); +} +void RendererStorageRD::reflection_probe_initialize(RID p_reflection_probe) { + reflection_probe_owner.initialize_rid(p_reflection_probe, ReflectionProbe()); } void RendererStorageRD::reflection_probe_set_update_mode(RID p_probe, RS::ReflectionProbeUpdateMode p_mode) { @@ -5791,8 +5886,11 @@ float RendererStorageRD::reflection_probe_get_ambient_color_energy(RID p_probe) return reflection_probe->ambient_color_energy; } -RID RendererStorageRD::decal_create() { - return decal_owner.make_rid(Decal()); +RID RendererStorageRD::decal_allocate() { + return decal_owner.allocate_rid(); +} +void RendererStorageRD::decal_initialize(RID p_decal) { + decal_owner.initialize_rid(p_decal, Decal()); } void RendererStorageRD::decal_set_extents(RID p_decal, const Vector3 &p_extents) { @@ -5879,11 +5977,14 @@ AABB RendererStorageRD::decal_get_aabb(RID p_decal) const { return AABB(-decal->extents, decal->extents * 2.0); } -RID RendererStorageRD::gi_probe_create() { - return gi_probe_owner.make_rid(GIProbe()); +RID RendererStorageRD::gi_probe_allocate() { + return gi_probe_owner.allocate_rid(); +} +void RendererStorageRD::gi_probe_initialize(RID p_gi_probe) { + gi_probe_owner.initialize_rid(p_gi_probe, GIProbe()); } -void RendererStorageRD::gi_probe_allocate(RID p_gi_probe, const Transform &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) { +void RendererStorageRD::gi_probe_allocate_data(RID p_gi_probe, const Transform &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) { GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); ERR_FAIL_COND(!gi_probe); @@ -6232,8 +6333,12 @@ RID RendererStorageRD::gi_probe_get_sdf_texture(RID p_gi_probe) { /* LIGHTMAP API */ -RID RendererStorageRD::lightmap_create() { - return lightmap_owner.make_rid(Lightmap()); +RID RendererStorageRD::lightmap_allocate() { + return lightmap_owner.allocate_rid(); +} + +void RendererStorageRD::lightmap_initialize(RID p_lightmap) { + lightmap_owner.initialize_rid(p_lightmap, Lightmap()); } void RendererStorageRD::lightmap_set_textures(RID p_lightmap, RID p_light, bool p_uses_spherical_haromics) { @@ -6436,7 +6541,8 @@ void RendererStorageRD::_clear_render_target(RenderTarget *rt) { void RendererStorageRD::_update_render_target(RenderTarget *rt) { if (rt->texture.is_null()) { //create a placeholder until updated - rt->texture = texture_2d_placeholder_create(); + rt->texture = texture_allocate(); + texture_2d_placeholder_initialize(rt->texture); Texture *tex = texture_owner.getornull(rt->texture); tex->is_render_target = true; } @@ -6936,7 +7042,7 @@ void RendererStorageRD::render_target_sdf_process(RID p_render_target) { 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, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, push_constant.size[0], push_constant.size[1], 1); /* Process */ @@ -6952,7 +7058,7 @@ void RendererStorageRD::render_target_sdf_process(RID p_render_target) { 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, 8, 8, 1); + 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); @@ -6963,7 +7069,7 @@ void RendererStorageRD::render_target_sdf_process(RID p_render_target) { 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, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, push_constant.size[0], push_constant.size[1], 1); RD::get_singleton()->compute_list_end(); } @@ -7327,6 +7433,7 @@ void RendererStorageRD::_update_decal_atlas() { 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 @@ -7381,7 +7488,7 @@ void RendererStorageRD::_update_decal_atlas() { prev_texture = mm.texture; } } else { - RD::get_singleton()->texture_clear(mm.texture, clear_color, 0, 1, 0, 1, false); + RD::get_singleton()->texture_clear(mm.texture, clear_color, 0, 1, 0, 1); } } } @@ -8147,29 +8254,38 @@ bool RendererStorageRD::free(RID p_rid) { material_owner.free(p_rid); } else if (mesh_owner.owns(p_rid)) { mesh_clear(p_rid); + mesh_set_shadow_mesh(p_rid, RID()); Mesh *mesh = mesh_owner.getornull(p_rid); mesh->dependency.deleted_notify(p_rid); if (mesh->instances.size()) { ERR_PRINT("deleting mesh with active instances"); } + if (mesh->shadow_owners.size()) { + for (Set<Mesh *>::Element *E = mesh->shadow_owners.front(); E; E = E->next()) { + Mesh *shadow_owner = E->get(); + shadow_owner->shadow_mesh = RID(); + shadow_owner->dependency.changed_notify(DEPENDENCY_CHANGED_MESH); + } + } mesh_owner.free(p_rid); } else if (mesh_instance_owner.owns(p_rid)) { MeshInstance *mi = mesh_instance_owner.getornull(p_rid); _mesh_instance_clear(mi); mi->mesh->instances.erase(mi->I); mi->I = nullptr; + mesh_instance_owner.free(p_rid); memdelete(mi); } else if (multimesh_owner.owns(p_rid)) { _update_dirty_multimeshes(); - multimesh_allocate(p_rid, 0, RS::MULTIMESH_TRANSFORM_2D); + multimesh_allocate_data(p_rid, 0, RS::MULTIMESH_TRANSFORM_2D); MultiMesh *multimesh = multimesh_owner.getornull(p_rid); multimesh->dependency.deleted_notify(p_rid); multimesh_owner.free(p_rid); } else if (skeleton_owner.owns(p_rid)) { _update_dirty_skeletons(); - skeleton_allocate(p_rid, 0); + skeleton_allocate_data(p_rid, 0); Skeleton *skeleton = skeleton_owner.getornull(p_rid); skeleton->dependency.deleted_notify(p_rid); skeleton_owner.free(p_rid); @@ -8187,7 +8303,7 @@ bool RendererStorageRD::free(RID p_rid) { decal->dependency.deleted_notify(p_rid); decal_owner.free(p_rid); } else if (gi_probe_owner.owns(p_rid)) { - gi_probe_allocate(p_rid, Transform(), AABB(), Vector3i(), Vector<uint8_t>(), Vector<uint8_t>(), Vector<uint8_t>(), Vector<int>()); //deallocate + gi_probe_allocate_data(p_rid, Transform(), AABB(), Vector3i(), Vector<uint8_t>(), Vector<uint8_t>(), Vector<uint8_t>(), Vector<int>()); //deallocate GIProbe *gi_probe = gi_probe_owner.getornull(p_rid); gi_probe->dependency.deleted_notify(p_rid); gi_probe_owner.free(p_rid); @@ -8243,11 +8359,11 @@ EffectsRD *RendererStorageRD::get_effects() { } void RendererStorageRD::capture_timestamps_begin() { - RD::get_singleton()->capture_timestamp("Frame Begin", false); + RD::get_singleton()->capture_timestamp("Frame Begin"); } void RendererStorageRD::capture_timestamp(const String &p_name) { - RD::get_singleton()->capture_timestamp(p_name, true); + RD::get_singleton()->capture_timestamp(p_name); } uint32_t RendererStorageRD::get_captured_timestamps_count() const { @@ -8281,7 +8397,7 @@ RendererStorageRD::RendererStorageRD() { static_assert(sizeof(GlobalVariables::Value) == 16); - global_variables.buffer_size = GLOBAL_GET("rendering/high_end/global_shader_variables_buffer_size"); + global_variables.buffer_size = GLOBAL_GET("rendering/limits/global_shader_variables/buffer_size"); global_variables.buffer_size = MAX(4096, global_variables.buffer_size); global_variables.buffer_values = memnew_arr(GlobalVariables::Value, global_variables.buffer_size); zeromem(global_variables.buffer_values, sizeof(GlobalVariables::Value) * global_variables.buffer_size); @@ -8549,14 +8665,14 @@ RendererStorageRD::RendererStorageRD() { sampler_state.min_filter = RD::SAMPLER_FILTER_LINEAR; sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; sampler_state.use_anisotropy = true; - sampler_state.anisotropy_max = 1 << int(GLOBAL_GET("rendering/quality/texture_filters/anisotropic_filtering_level")); + 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; sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; sampler_state.use_anisotropy = true; - sampler_state.anisotropy_max = 1 << int(GLOBAL_GET("rendering/quality/texture_filters/anisotropic_filtering_level")); + sampler_state.anisotropy_max = 1 << int(GLOBAL_GET("rendering/textures/default_filters/anisotropic_filtering_level")); } break; default: { @@ -8726,7 +8842,7 @@ RendererStorageRD::RendererStorageRD() { } } - lightmap_probe_capture_update_speed = GLOBAL_GET("rendering/lightmapper/probe_capture_update_speed"); + lightmap_probe_capture_update_speed = GLOBAL_GET("rendering/lightmapping/probe_capture/update_speed"); /* Particles */ @@ -8767,7 +8883,7 @@ RendererStorageRD::RendererStorageRD() { actions.renames["RESTART_VELOCITY"] = "restart_velocity"; actions.renames["RESTART_COLOR"] = "restart_color"; actions.renames["RESTART_CUSTOM"] = "restart_custom"; - actions.renames["emit_particle"] = "emit_particle"; + actions.renames["emit_subparticle"] = "emit_subparticle"; actions.renames["COLLIDED"] = "collided"; actions.renames["COLLISION_NORMAL"] = "collision_normal"; actions.renames["COLLISION_DEPTH"] = "collision_depth"; @@ -8793,9 +8909,11 @@ RendererStorageRD::RendererStorageRD() { { // default material and shader for particles shader - particles_shader.default_shader = shader_create(); + particles_shader.default_shader = shader_allocate(); + shader_initialize(particles_shader.default_shader); shader_set_code(particles_shader.default_shader, "shader_type particles; void compute() { COLOR = vec4(1.0); } \n"); - particles_shader.default_material = material_create(); + particles_shader.default_material = material_allocate(); + material_initialize(particles_shader.default_material); material_set_shader(particles_shader.default_material, particles_shader.default_shader); ParticlesMaterialData *md = (ParticlesMaterialData *)material_get_data(particles_shader.default_material, RendererStorageRD::SHADER_TYPE_PARTICLES); diff --git a/servers/rendering/renderer_rd/renderer_storage_rd.h b/servers/rendering/renderer_rd/renderer_storage_rd.h index bd27936e38..cd3d4604eb 100644 --- a/servers/rendering/renderer_rd/renderer_storage_rd.h +++ b/servers/rendering/renderer_rd/renderer_storage_rd.h @@ -95,6 +95,21 @@ public: p_array[11] = 0; } + static _FORCE_INLINE_ void store_transform_transposed_3x4(const Transform &p_mtx, float *p_array) { + p_array[0] = p_mtx.basis.elements[0][0]; + p_array[1] = p_mtx.basis.elements[0][1]; + p_array[2] = p_mtx.basis.elements[0][2]; + p_array[3] = p_mtx.origin.x; + p_array[4] = p_mtx.basis.elements[1][0]; + p_array[5] = p_mtx.basis.elements[1][1]; + p_array[6] = p_mtx.basis.elements[1][2]; + p_array[7] = p_mtx.origin.y; + p_array[8] = p_mtx.basis.elements[2][0]; + p_array[9] = p_mtx.basis.elements[2][1]; + p_array[10] = p_mtx.basis.elements[2][2]; + p_array[11] = p_mtx.origin.z; + } + static _FORCE_INLINE_ void store_camera(const CameraMatrix &p_mtx, float *p_array) { for (int i = 0; i < 4; i++) { for (int j = 0; j < 4; j++) { @@ -127,6 +142,8 @@ public: 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() {} }; @@ -204,7 +221,7 @@ private: ~CanvasTexture(); }; - RID_PtrOwner<CanvasTexture> canvas_texture_owner; + RID_PtrOwner<CanvasTexture, true> canvas_texture_owner; /* TEXTURE API */ struct Texture { @@ -350,7 +367,7 @@ private: }; ShaderDataRequestFunction shader_data_request_func[SHADER_TYPE_MAX]; - mutable RID_Owner<Shader> shader_owner; + mutable RID_Owner<Shader, true> shader_owner; /* Material */ @@ -372,7 +389,7 @@ private: }; MaterialDataRequestFunction material_data_request_func[SHADER_TYPE_MAX]; - mutable RID_Owner<Material> material_owner; + mutable RID_Owner<Material, true> material_owner; Material *material_update_list; void _material_queue_update(Material *material, bool p_uniform, bool p_texture); @@ -461,10 +478,13 @@ private: List<MeshInstance *> instances; + RID shadow_mesh; + Set<Mesh *> shadow_owners; + Dependency dependency; }; - mutable RID_Owner<Mesh> mesh_owner; + mutable RID_Owner<Mesh, true> mesh_owner; struct MeshInstance { Mesh *mesh; @@ -567,7 +587,7 @@ private: Dependency dependency; }; - mutable RID_Owner<MultiMesh> multimesh_owner; + mutable RID_Owner<MultiMesh, true> multimesh_owner; MultiMesh *multimesh_dirty_list = nullptr; @@ -840,6 +860,8 @@ private: 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(); }; @@ -871,7 +893,7 @@ private: void update_particles(); - mutable RID_Owner<Particles> particles_owner; + mutable RID_Owner<Particles, true> particles_owner; /* Particles Collision */ @@ -893,7 +915,7 @@ private: Dependency dependency; }; - mutable RID_Owner<ParticlesCollision> particles_collision_owner; + mutable RID_Owner<ParticlesCollision, true> particles_collision_owner; struct ParticlesCollisionInstance { RID collision; @@ -923,7 +945,7 @@ private: Dependency dependency; }; - mutable RID_Owner<Skeleton> skeleton_owner; + mutable RID_Owner<Skeleton, true> skeleton_owner; _FORCE_INLINE_ void _skeleton_make_dirty(Skeleton *skeleton); @@ -955,7 +977,7 @@ private: Dependency dependency; }; - mutable RID_Owner<Light> light_owner; + mutable RID_Owner<Light, true> light_owner; /* REFLECTION PROBE */ @@ -978,7 +1000,7 @@ private: Dependency dependency; }; - mutable RID_Owner<ReflectionProbe> reflection_probe_owner; + mutable RID_Owner<ReflectionProbe, true> reflection_probe_owner; /* DECAL */ @@ -999,7 +1021,7 @@ private: Dependency dependency; }; - mutable RID_Owner<Decal> decal_owner; + mutable RID_Owner<Decal, true> decal_owner; /* GI PROBE */ @@ -1042,7 +1064,7 @@ private: RID giprobe_sdf_shader_version_shader; RID giprobe_sdf_shader_pipeline; - mutable RID_Owner<GIProbe> gi_probe_owner; + mutable RID_Owner<GIProbe, true> gi_probe_owner; /* REFLECTION PROBE */ @@ -1073,7 +1095,7 @@ private: uint64_t lightmap_array_version = 0; - mutable RID_Owner<Lightmap> lightmap_owner; + mutable RID_Owner<Lightmap, true> lightmap_owner; float lightmap_probe_capture_update_speed = 4; @@ -1227,12 +1249,16 @@ private: EffectsRD effects; public: + virtual bool can_create_resources_async() const; + /* TEXTURE API */ - virtual RID texture_2d_create(const Ref<Image> &p_image); - virtual RID texture_2d_layered_create(const Vector<Ref<Image>> &p_layers, RS::TextureLayeredType p_layered_type); - virtual RID texture_3d_create(Image::Format p_format, int p_width, int p_height, int p_depth, bool p_mipmaps, const Vector<Ref<Image>> &p_data); //all slices, then all the mipmaps, must be coherent - virtual RID texture_proxy_create(RID p_base); + virtual RID texture_allocate(); + + virtual void texture_2d_initialize(RID p_texture, const Ref<Image> &p_image); + virtual void texture_2d_layered_initialize(RID p_texture, const Vector<Ref<Image>> &p_layers, RS::TextureLayeredType p_layered_type); + virtual void 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); //all slices, then all the mipmaps, must be coherent + virtual void texture_proxy_initialize(RID p_texture, RID p_base); virtual void _texture_2d_update(RID p_texture, const Ref<Image> &p_image, int p_layer, bool p_immediate); @@ -1242,9 +1268,9 @@ public: virtual void texture_proxy_update(RID p_texture, RID p_proxy_to); //these two APIs can be used together or in combination with the others. - virtual RID texture_2d_placeholder_create(); - virtual RID texture_2d_layered_placeholder_create(RenderingServer::TextureLayeredType p_layered_type); - virtual RID texture_3d_placeholder_create(); + virtual void texture_2d_placeholder_initialize(RID p_texture); + virtual void texture_2d_layered_placeholder_initialize(RID p_texture, RenderingServer::TextureLayeredType p_layered_type); + virtual void texture_3d_placeholder_initialize(RID p_texture); virtual Ref<Image> texture_2d_get(RID p_texture) const; virtual Ref<Image> texture_2d_layer_get(RID p_texture, int p_layer) const; @@ -1316,7 +1342,8 @@ public: /* CANVAS TEXTURE API */ - virtual RID canvas_texture_create(); + RID canvas_texture_allocate(); + void canvas_texture_initialize(RID p_canvas_texture); virtual void canvas_texture_set_channel(RID p_canvas_texture, RS::CanvasTextureChannel p_channel, RID p_texture); virtual void canvas_texture_set_shading_parameters(RID p_canvas_texture, const Color &p_specular_color, float p_shininess); @@ -1328,7 +1355,8 @@ public: /* SHADER API */ - RID shader_create(); + RID shader_allocate(); + void shader_initialize(RID p_shader); void shader_set_code(RID p_shader, const String &p_code); String shader_get_code(RID p_shader) const; @@ -1339,9 +1367,12 @@ public: Variant shader_get_param_default(RID p_shader, const StringName &p_param) const; 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; + /* COMMON MATERIAL API */ - RID material_create(); + RID material_allocate(); + void material_initialize(RID p_material); void material_set_shader(RID p_material, RID p_shader); @@ -1377,7 +1408,8 @@ public: /* MESH API */ - virtual RID mesh_create(); + RID mesh_allocate(); + void mesh_initialize(RID p_mesh); virtual void mesh_set_blend_shape_count(RID p_mesh, int p_blend_shape_count); @@ -1402,6 +1434,7 @@ public: virtual AABB mesh_get_custom_aabb(RID p_mesh) const; virtual AABB mesh_get_aabb(RID p_mesh, RID p_skeleton = RID()); + virtual void mesh_set_shadow_mesh(RID p_mesh, RID p_shadow_mesh); virtual void mesh_clear(RID p_mesh); @@ -1440,6 +1473,13 @@ public: return mesh->surfaces[p_surface_index]; } + _FORCE_INLINE_ RID mesh_get_shadow_mesh(RID p_mesh) { + Mesh *mesh = mesh_owner.getornull(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; @@ -1450,13 +1490,7 @@ public: return s->lod_count > 0; } - _FORCE_INLINE_ RID mesh_surface_get_index_array(void *p_surface) const { - Mesh::Surface *s = reinterpret_cast<Mesh::Surface *>(p_surface); - - return s->index_array; - } - - _FORCE_INLINE_ RID mesh_surface_get_index_array_with_lod(void *p_surface, float p_model_scale, float p_distance_threshold, float p_lod_threshold) const { + _FORCE_INLINE_ uint32_t mesh_surface_get_lod(void *p_surface, float p_model_scale, float p_distance_threshold, float p_lod_threshold) const { Mesh::Surface *s = reinterpret_cast<Mesh::Surface *>(p_surface); int32_t current_lod = -1; @@ -1468,9 +1502,19 @@ public: current_lod = i; } if (current_lod == -1) { + return 0; + } else { + 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[current_lod].index_array; + return s->lods[p_lod - 1].index_array; } } @@ -1586,9 +1630,10 @@ public: /* MULTIMESH API */ - RID multimesh_create(); + RID multimesh_allocate(); + void multimesh_initialize(RID p_multimesh); - void multimesh_allocate(RID p_multimesh, int p_instances, RS::MultimeshTransformFormat p_transform_format, bool p_use_colors = false, bool p_use_custom_data = false); + 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); int multimesh_get_instance_count(RID p_multimesh) const; void multimesh_set_mesh(RID p_multimesh, RID p_mesh); @@ -1652,24 +1697,28 @@ public: /* IMMEDIATE API */ - RID immediate_create() { return RID(); } - void immediate_begin(RID p_immediate, RS::PrimitiveType p_rimitive, RID p_texture = RID()) {} - void immediate_vertex(RID p_immediate, const Vector3 &p_vertex) {} - void immediate_normal(RID p_immediate, const Vector3 &p_normal) {} - void immediate_tangent(RID p_immediate, const Plane &p_tangent) {} - void immediate_color(RID p_immediate, const Color &p_color) {} - void immediate_uv(RID p_immediate, const Vector2 &tex_uv) {} - void immediate_uv2(RID p_immediate, const Vector2 &tex_uv) {} - void immediate_end(RID p_immediate) {} - void immediate_clear(RID p_immediate) {} - void immediate_set_material(RID p_immediate, RID p_material) {} - RID immediate_get_material(RID p_immediate) const { return RID(); } - AABB immediate_get_aabb(RID p_immediate) const { return AABB(); } + RID immediate_allocate() { return RID(); } + void immediate_initialize(RID p_immediate) {} + + virtual void immediate_begin(RID p_immediate, RS::PrimitiveType p_rimitive, RID p_texture = RID()) {} + virtual void immediate_vertex(RID p_immediate, const Vector3 &p_vertex) {} + virtual void immediate_normal(RID p_immediate, const Vector3 &p_normal) {} + virtual void immediate_tangent(RID p_immediate, const Plane &p_tangent) {} + virtual void immediate_color(RID p_immediate, const Color &p_color) {} + virtual void immediate_uv(RID p_immediate, const Vector2 &tex_uv) {} + virtual void immediate_uv2(RID p_immediate, const Vector2 &tex_uv) {} + virtual void immediate_end(RID p_immediate) {} + virtual void immediate_clear(RID p_immediate) {} + virtual void immediate_set_material(RID p_immediate, RID p_material) {} + virtual RID immediate_get_material(RID p_immediate) const { return RID(); } + virtual AABB immediate_get_aabb(RID p_immediate) const { return AABB(); } /* SKELETON API */ - RID skeleton_create(); - void skeleton_allocate(RID p_skeleton, int p_bones, bool p_2d_skeleton = false); + RID skeleton_allocate(); + void skeleton_initialize(RID p_skeleton); + + void skeleton_allocate_data(RID p_skeleton, int p_bones, bool p_2d_skeleton = false); void skeleton_set_base_transform_2d(RID p_skeleton, const Transform2D &p_base_transform); void skeleton_set_world_transform(RID p_skeleton, bool p_enable, const Transform &p_world_transform); int skeleton_get_bone_count(RID p_skeleton) const; @@ -1703,11 +1752,16 @@ public: } /* Light API */ - RID light_create(RS::LightType p_type); + void _light_initialize(RID p_rid, RS::LightType p_type); + + RID directional_light_allocate(); + void directional_light_initialize(RID p_light); - RID directional_light_create() { return light_create(RS::LIGHT_DIRECTIONAL); } - RID omni_light_create() { return light_create(RS::LIGHT_OMNI); } - RID spot_light_create() { return light_create(RS::LIGHT_SPOT); } + RID omni_light_allocate(); + void omni_light_initialize(RID p_light); + + RID spot_light_allocate(); + void spot_light_initialize(RID p_light); void light_set_color(RID p_light, const Color &p_color); void light_set_param(RID p_light, RS::LightParam p_param, float p_value); @@ -1810,7 +1864,8 @@ public: /* PROBE API */ - RID reflection_probe_create(); + RID reflection_probe_allocate(); + void reflection_probe_initialize(RID p_reflection_probe); void reflection_probe_set_update_mode(RID p_probe, RS::ReflectionProbeUpdateMode p_mode); void reflection_probe_set_intensity(RID p_probe, float p_intensity); @@ -1850,7 +1905,9 @@ public: /* DECAL API */ - virtual RID decal_create(); + RID decal_allocate(); + void decal_initialize(RID p_decal); + virtual void decal_set_extents(RID p_decal, const Vector3 &p_extents); virtual void decal_set_texture(RID p_decal, RS::DecalTexture p_type, RID p_texture); virtual void decal_set_emission_energy(RID p_decal, float p_energy); @@ -1925,9 +1982,10 @@ public: /* GI PROBE API */ - RID gi_probe_create(); + RID gi_probe_allocate(); + void gi_probe_initialize(RID p_gi_probe); - void gi_probe_allocate(RID p_gi_probe, const Transform &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); + void gi_probe_allocate_data(RID p_gi_probe, const Transform &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); AABB gi_probe_get_bounds(RID p_gi_probe) const; Vector3i gi_probe_get_octree_size(RID p_gi_probe) const; @@ -1978,7 +2036,8 @@ public: /* LIGHTMAP CAPTURE */ - virtual RID lightmap_create(); + RID lightmap_allocate(); + void lightmap_initialize(RID p_lightmap); virtual void lightmap_set_textures(RID p_lightmap, RID p_light, bool p_uses_spherical_haromics); virtual void lightmap_set_probe_bounds(RID p_lightmap, const AABB &p_bounds); @@ -2027,7 +2086,8 @@ public: /* PARTICLES */ - RID particles_create(); + RID particles_allocate(); + void particles_initialize(RID p_particles_collision); void particles_set_emitting(RID p_particles, bool p_emitting); void particles_set_amount(RID p_particles, int p_amount); @@ -2105,7 +2165,9 @@ public: /* PARTICLES COLLISION */ - virtual RID particles_collision_create(); + RID particles_collision_allocate(); + void particles_collision_initialize(RID p_particles_collision); + virtual void particles_collision_set_collision_type(RID p_particles_collision, RS::ParticlesCollisionType p_type); virtual void particles_collision_set_cull_mask(RID p_particles_collision, uint32_t p_cull_mask); virtual void particles_collision_set_sphere_radius(RID p_particles_collision, float p_radius); //for spheres @@ -2201,7 +2263,7 @@ public: void render_info_end_capture() {} int get_captured_render_info(RS::RenderInfo p_info) { return 0; } - int get_render_info(RS::RenderInfo p_info) { return 0; } + uint64_t get_render_info(RS::RenderInfo p_info) { return 0; } String get_video_adapter_name() const { return String(); } String get_video_adapter_vendor() const { return String(); } diff --git a/servers/rendering/renderer_rd/shader_compiler_rd.cpp b/servers/rendering/renderer_rd/shader_compiler_rd.cpp index e77141b26c..8135d388e1 100644 --- a/servers/rendering/renderer_rd/shader_compiler_rd.cpp +++ b/servers/rendering/renderer_rd/shader_compiler_rd.cpp @@ -687,7 +687,15 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge uint32_t index = p_default_actions.base_varying_index; + List<Pair<StringName, SL::ShaderNode::Varying>> var_frag_to_light; + for (Map<StringName, SL::ShaderNode::Varying>::Element *E = pnode->varyings.front(); E; E = E->next()) { + if (E->get().stage == SL::ShaderNode::Varying::STAGE_FRAGMENT_TO_LIGHT || E->get().stage == SL::ShaderNode::Varying::STAGE_FRAGMENT) { + var_frag_to_light.push_back(Pair<StringName, SL::ShaderNode::Varying>(E->key(), E->get())); + fragment_varyings.insert(E->key()); + continue; + } + String vcode; String interp_mode = _interpstr(E->get().interpolation); vcode += _prestr(E->get().precision); @@ -705,6 +713,21 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge index++; } + if (var_frag_to_light.size() > 0) { + String gcode = "\n\nstruct {\n"; + for (List<Pair<StringName, SL::ShaderNode::Varying>>::Element *E = var_frag_to_light.front(); E; E = E->next()) { + gcode += "\t" + _prestr(E->get().second.precision) + _typestr(E->get().second.type) + " " + _mkid(E->get().first); + if (E->get().second.array_size > 0) { + gcode += "["; + gcode += itos(E->get().second.array_size); + gcode += "]"; + } + gcode += ";\n"; + } + gcode += "} frag_to_light;\n"; + r_gen_code.fragment_global += gcode; + } + for (int i = 0; i < pnode->vconstants.size(); i++) { const SL::ShaderNode::Constant &cnode = pnode->vconstants[i]; String gcode; @@ -833,6 +856,19 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge } break; case SL::Node::TYPE_VARIABLE: { SL::VariableNode *vnode = (SL::VariableNode *)p_node; + bool use_fragment_varying = false; + + if (current_func_name != vertex_name) { + if (p_assigning) { + if (shader->varyings.has(vnode->name)) { + use_fragment_varying = true; + } + } else { + if (fragment_varyings.has(vnode->name)) { + use_fragment_varying = true; + } + } + } if (p_assigning && p_actions.write_flag_pointers.has(vnode->name)) { *p_actions.write_flag_pointers[vnode->name] = true; @@ -877,7 +913,10 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge } } else { - code = _mkid(vnode->name); //its something else (local var most likely) use as is + if (use_fragment_varying) { + code = "frag_to_light."; + } + code += _mkid(vnode->name); //its something else (local var most likely) use as is } } @@ -962,6 +1001,23 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge } break; case SL::Node::TYPE_ARRAY: { SL::ArrayNode *anode = (SL::ArrayNode *)p_node; + bool use_fragment_varying = false; + + if (current_func_name != vertex_name) { + if (anode->assign_expression != nullptr) { + use_fragment_varying = true; + } else { + if (p_assigning) { + if (shader->varyings.has(anode->name)) { + use_fragment_varying = true; + } + } else { + if (fragment_varyings.has(anode->name)) { + use_fragment_varying = true; + } + } + } + } if (p_assigning && p_actions.write_flag_pointers.has(anode->name)) { *p_actions.write_flag_pointers[anode->name] = true; @@ -984,7 +1040,10 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge if (p_default_actions.renames.has(anode->name)) { code = p_default_actions.renames[anode->name]; } else { - code = _mkid(anode->name); + if (use_fragment_varying) { + code = "frag_to_light."; + } + code += _mkid(anode->name); } if (anode->call_expression != nullptr) { @@ -1277,6 +1336,7 @@ Error ShaderCompilerRD::compile(RS::ShaderMode p_mode, const String &p_code, Ide used_name_defines.clear(); used_rmode_defines.clear(); used_flag_pointers.clear(); + fragment_varyings.clear(); shader = parser.get_shader(); function = nullptr; @@ -1408,7 +1468,7 @@ ShaderCompilerRD::ShaderCompilerRD() { actions[RS::SHADER_SPATIAL].render_mode_defines["cull_front"] = "#define DO_SIDE_CHECK\n"; actions[RS::SHADER_SPATIAL].render_mode_defines["cull_disabled"] = "#define DO_SIDE_CHECK\n"; - bool force_lambert = GLOBAL_GET("rendering/quality/shading/force_lambert_over_burley"); + bool force_lambert = GLOBAL_GET("rendering/shading/overrides/force_lambert_over_burley"); if (!force_lambert) { actions[RS::SHADER_SPATIAL].render_mode_defines["diffuse_burley"] = "#define DIFFUSE_BURLEY\n"; @@ -1418,7 +1478,7 @@ ShaderCompilerRD::ShaderCompilerRD() { actions[RS::SHADER_SPATIAL].render_mode_defines["diffuse_lambert_wrap"] = "#define DIFFUSE_LAMBERT_WRAP\n"; actions[RS::SHADER_SPATIAL].render_mode_defines["diffuse_toon"] = "#define DIFFUSE_TOON\n"; - bool force_blinn = GLOBAL_GET("rendering/quality/shading/force_blinn_over_ggx"); + bool force_blinn = GLOBAL_GET("rendering/shading/overrides/force_blinn_over_ggx"); if (!force_blinn) { actions[RS::SHADER_SPATIAL].render_mode_defines["specular_schlick_ggx"] = "#define SPECULAR_SCHLICK_GGX\n"; diff --git a/servers/rendering/renderer_rd/shader_compiler_rd.h b/servers/rendering/renderer_rd/shader_compiler_rd.h index d127d8e01c..6575829e73 100644 --- a/servers/rendering/renderer_rd/shader_compiler_rd.h +++ b/servers/rendering/renderer_rd/shader_compiler_rd.h @@ -114,6 +114,7 @@ private: Set<StringName> used_flag_pointers; Set<StringName> used_rmode_defines; Set<StringName> internal_functions; + Set<StringName> fragment_varyings; DefaultIdentifierActions actions; diff --git a/servers/rendering/renderer_rd/shader_rd.cpp b/servers/rendering/renderer_rd/shader_rd.cpp index e955cead05..e4a39ff813 100644 --- a/servers/rendering/renderer_rd/shader_rd.cpp +++ b/servers/rendering/renderer_rd/shader_rd.cpp @@ -301,6 +301,7 @@ void ShaderRD::_compile_variant(uint32_t p_variant, Version *p_version) { builder.append(compute_codev.get_data()); // version info (if exists) builder.append("\n"); //make sure defines begin at newline + builder.append(base_compute_defines.get_data()); builder.append(general_defines.get_data()); builder.append(variant_defines[p_variant].get_data()); @@ -351,6 +352,127 @@ void ShaderRD::_compile_variant(uint32_t p_variant, Version *p_version) { } } +RS::ShaderNativeSourceCode ShaderRD::version_get_native_source_code(RID p_version) { + Version *version = version_owner.getornull(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; + + builder.append(vertex_codev.get_data()); // version info (if exists) + builder.append("\n"); //make sure defines begin at newline + builder.append(general_defines.get_data()); + builder.append(variant_defines[i].get_data()); + + for (int j = 0; j < version->custom_defines.size(); j++) { + builder.append(version->custom_defines[j].get_data()); + } + + builder.append(vertex_code0.get_data()); //first part of vertex + + builder.append(version->uniforms.get_data()); //uniforms (same for vertex and fragment) + + builder.append(vertex_code1.get_data()); //second part of vertex + + builder.append(version->vertex_globals.get_data()); // vertex globals + + builder.append(vertex_code2.get_data()); //third part of vertex + + builder.append(version->vertex_code.get_data()); // code + + builder.append(vertex_code3.get_data()); //fourth of 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; + + builder.append(fragment_codev.get_data()); // version info (if exists) + builder.append("\n"); //make sure defines begin at newline + builder.append(general_defines.get_data()); + builder.append(variant_defines[i].get_data()); + for (int j = 0; j < version->custom_defines.size(); j++) { + builder.append(version->custom_defines[j].get_data()); + } + + builder.append(fragment_code0.get_data()); //first part of fragment + + builder.append(version->uniforms.get_data()); //uniforms (same for fragment and fragment) + + builder.append(fragment_code1.get_data()); //first part of fragment + + builder.append(version->fragment_globals.get_data()); // fragment globals + + builder.append(fragment_code2.get_data()); //third part of fragment + + builder.append(version->fragment_light.get_data()); // fragment light + + builder.append(fragment_code3.get_data()); //fourth part of fragment + + builder.append(version->fragment_code.get_data()); // fragment code + + builder.append(fragment_code4.get_data()); //fourth part of 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; + + builder.append(compute_codev.get_data()); // version info (if exists) + builder.append("\n"); //make sure defines begin at newline + builder.append(base_compute_defines.get_data()); + builder.append(general_defines.get_data()); + builder.append(variant_defines[i].get_data()); + + for (int j = 0; j < version->custom_defines.size(); j++) { + builder.append(version->custom_defines[j].get_data()); + } + + builder.append(compute_code0.get_data()); //first part of compute + + builder.append(version->uniforms.get_data()); //uniforms (same for compute and fragment) + + builder.append(compute_code1.get_data()); //second part of compute + + builder.append(version->compute_globals.get_data()); // compute globals + + builder.append(compute_code2.get_data()); //third part of compute + + builder.append(version->compute_code.get_data()); // code + + builder.append(compute_code3.get_data()); //fourth of 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; +} + void ShaderRD::_compile_version(Version *p_version) { _clear_version(p_version); @@ -475,6 +597,22 @@ bool ShaderRD::is_variant_enabled(int p_variant) const { return variants_enabled[p_variant]; } +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); diff --git a/servers/rendering/renderer_rd/shader_rd.h b/servers/rendering/renderer_rd/shader_rd.h index a80d08050a..e0f4dcf2d0 100644 --- a/servers/rendering/renderer_rd/shader_rd.h +++ b/servers/rendering/renderer_rd/shader_rd.h @@ -36,6 +36,7 @@ #include "core/templates/map.h" #include "core/templates/rid_owner.h" #include "core/variant/variant.h" +#include "servers/rendering_server.h" #include <stdio.h> /** @@ -98,8 +99,10 @@ class ShaderRD { const char *name; + CharString base_compute_defines; + protected: - ShaderRD() {} + ShaderRD(); void setup(const char *p_vertex_code, const char *p_fragment_code, const char *p_compute_code, const char *p_name); public: @@ -133,6 +136,8 @@ public: void set_variant_enabled(int p_variant, bool p_enabled); bool is_variant_enabled(int p_variant) const; + 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(); }; diff --git a/servers/rendering/renderer_rd/shaders/SCsub b/servers/rendering/renderer_rd/shaders/SCsub index deaa9668df..fc513d3fb9 100644 --- a/servers/rendering/renderer_rd/shaders/SCsub +++ b/servers/rendering/renderer_rd/shaders/SCsub @@ -3,44 +3,15 @@ Import("env") if "RD_GLSL" in env["BUILDERS"]: - env.RD_GLSL("canvas.glsl") - env.RD_GLSL("canvas_occlusion.glsl") - env.RD_GLSL("canvas_sdf.glsl") - env.RD_GLSL("copy.glsl") - env.RD_GLSL("copy_to_fb.glsl") - env.RD_GLSL("cubemap_roughness.glsl") - env.RD_GLSL("cubemap_downsampler.glsl") - env.RD_GLSL("cubemap_filter.glsl") - env.RD_GLSL("scene_forward.glsl") - env.RD_GLSL("sky.glsl") - env.RD_GLSL("tonemap.glsl") - env.RD_GLSL("cube_to_dp.glsl") - env.RD_GLSL("giprobe.glsl") - env.RD_GLSL("giprobe_debug.glsl") - env.RD_GLSL("giprobe_sdf.glsl") - env.RD_GLSL("luminance_reduce.glsl") - env.RD_GLSL("bokeh_dof.glsl") - env.RD_GLSL("ssao.glsl") - env.RD_GLSL("ssao_downsample.glsl") - env.RD_GLSL("ssao_importance_map.glsl") - env.RD_GLSL("ssao_blur.glsl") - env.RD_GLSL("ssao_interleave.glsl") - env.RD_GLSL("roughness_limiter.glsl") - env.RD_GLSL("screen_space_reflection.glsl") - env.RD_GLSL("screen_space_reflection_filter.glsl") - env.RD_GLSL("screen_space_reflection_scale.glsl") - env.RD_GLSL("subsurface_scattering.glsl") - env.RD_GLSL("specular_merge.glsl") - env.RD_GLSL("gi.glsl") - env.RD_GLSL("resolve.glsl") - env.RD_GLSL("sdfgi_preprocess.glsl") - env.RD_GLSL("sdfgi_integrate.glsl") - env.RD_GLSL("sdfgi_direct_light.glsl") - env.RD_GLSL("sdfgi_debug.glsl") - env.RD_GLSL("sdfgi_debug_probes.glsl") - env.RD_GLSL("volumetric_fog.glsl") - env.RD_GLSL("shadow_reduce.glsl") - env.RD_GLSL("particles.glsl") - env.RD_GLSL("particles_copy.glsl") - env.RD_GLSL("sort.glsl") - env.RD_GLSL("skeleton.glsl") + # 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) + + # compile shaders + for glsl_file in glsl_files: + env.RD_GLSL(glsl_file) diff --git a/servers/rendering/renderer_rd/shaders/canvas.glsl b/servers/rendering/renderer_rd/shaders/canvas.glsl index 9c4e95a7c2..3b39edc70e 100644 --- a/servers/rendering/renderer_rd/shaders/canvas.glsl +++ b/servers/rendering/renderer_rd/shaders/canvas.glsl @@ -396,7 +396,7 @@ vec4 light_shadow_compute(uint light_base, vec4 light_color, vec4 shadow_uv vec4 shadow_color = unpackUnorm4x8(light_array.data[light_base].shadow_color); #ifdef LIGHT_SHADER_CODE_USED - shadow_color *= shadow_modulate; + shadow_color.rgb *= shadow_modulate; #endif shadow_color.a *= light_color.a; //respect light alpha @@ -546,7 +546,7 @@ FRAGMENT_SHADER_CODE #ifdef LIGHT_SHADER_CODE_USED vec4 shadow_modulate = vec4(1.0); - light_color = light_compute(light_vertex, direction, normal, light_color, light_color.a, specular_shininess, shadow_modulate, screen_uv, color, uv, true); + 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, color, true); #else if (normal_used) { @@ -563,7 +563,7 @@ FRAGMENT_SHADER_CODE light_color = light_shadow_compute(light_base, light_color, shadow_uv #ifdef LIGHT_SHADER_CODE_USED , - shadow_modulate + shadow_modulate.rgb #endif ); } @@ -605,7 +605,7 @@ FRAGMENT_SHADER_CODE 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, color, uv, false); + light_color = light_compute(light_vertex, light_position, normal, light_color, light_base_color.a, specular_shininess, shadow_modulate, screen_uv, uv, color, false); #else light_color.rgb *= light_base_color.rgb * light_base_color.a; @@ -659,7 +659,7 @@ FRAGMENT_SHADER_CODE light_color = light_shadow_compute(light_base, light_color, shadow_uv #ifdef LIGHT_SHADER_CODE_USED , - shadow_modulate + shadow_modulate.rgb #endif ); } diff --git a/servers/rendering/renderer_rd/shaders/cluster_data_inc.glsl b/servers/rendering/renderer_rd/shaders/cluster_data_inc.glsl index e723468dd8..3a4bf4da07 100644 --- a/servers/rendering/renderer_rd/shaders/cluster_data_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/cluster_data_inc.glsl @@ -6,12 +6,18 @@ struct LightData { //this structure needs to be as packed as possible vec3 position; float inv_radius; + vec3 direction; float size; - uint attenuation_energy; //attenuation - uint color_specular; //rgb color, a specular (8 bit unorm) - uint cone_attenuation_angle; // attenuation and angle, (16bit float) - uint shadow_color_enabled; //shadow rgb color, a>0.5 enabled (8bit unorm) + + vec3 color; + float attenuation; + + float cone_attenuation; + float cone_angle; + float specular_amount; + bool shadow_enabled; + vec4 atlas_rect; // rect in the shadow atlas mat4 shadow_matrix; float shadow_bias; @@ -34,9 +40,13 @@ struct ReflectionData { float index; vec3 box_offset; uint mask; - vec4 params; // intensity, 0, interior , boxproject vec3 ambient; // ambient color + float intensity; + bool exterior; + bool box_project; uint ambient_mode; + uint pad; + //0-8 is intensity,8-9 is ambient, mode 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 }; 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..70a875192c --- /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, binding = 0, 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..8723ea78e4 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cluster_render.glsl @@ -0,0 +1,168 @@ +#[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, binding = 0, 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 + +#if defined(GL_KHR_shader_subgroup_ballot) && defined(GL_KHR_shader_subgroup_arithmetic) && defined(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 + +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) { + //http://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..5be0893c4f --- /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, binding = 0, 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/cube_to_dp.glsl b/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl index 54d67db6c6..c3ac0bee57 100644 --- a/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl +++ b/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl @@ -1,33 +1,48 @@ -#[compute] +#[vertex] #version 450 VERSION_DEFINES -layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; +layout(push_constant, binding = 1, std430) uniform Params { + float z_far; + float z_near; + bool z_flip; + uint pad; + 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, binding = 1, std430) uniform Params { - ivec2 screen_size; - ivec2 offset; - float bias; float z_far; float z_near; bool z_flip; + uint pad; + vec4 screen_rect; } params; -layout(r32f, set = 1, binding = 0) uniform restrict writeonly image2D depth_buffer; - void main() { - ivec2 pos = ivec2(gl_GlobalInvocationID.xy); - if (any(greaterThan(pos, params.screen_size))) { //too large, do nothing - return; - } - - vec2 pixel_size = 1.0 / vec2(params.screen_size); - vec2 uv = (vec2(pos) + 0.5) * pixel_size; + vec2 uv = uv_interp; vec3 normal = vec3(uv * 2.0 - 1.0, 0.0); @@ -65,5 +80,5 @@ void main() { 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; - imageStore(depth_buffer, pos + params.offset, vec4(depth)); + gl_FragDepth = depth; } diff --git a/servers/rendering/renderer_rd/shaders/gi.glsl b/servers/rendering/renderer_rd/shaders/gi.glsl index 8011dadc72..92a5682572 100644 --- a/servers/rendering/renderer_rd/shaders/gi.glsl +++ b/servers/rendering/renderer_rd/shaders/gi.glsl @@ -97,13 +97,12 @@ layout(push_constant, binding = 0, std430) uniform Params { vec4 proj_info; + vec3 ao_color; uint max_giprobes; + bool high_quality_vct; - bool use_sdfgi; bool orthogonal; - - vec3 ao_color; - uint pad; + uint pad[2]; mat3x4 cam_rotation; } @@ -331,7 +330,7 @@ void sdfgi_process(vec3 vertex, vec3 normal, vec3 reflection, float roughness, o } ambient_light.rgb = diffuse; -#if 1 + if (roughness < 0.2) { vec3 pos_to_uvw = 1.0 / sdfgi.grid_size; vec4 light_accum = vec4(0.0); @@ -363,59 +362,63 @@ void sdfgi_process(vec3 vertex, vec3 normal, vec3 reflection, float roughness, o //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 - while (length(ray_pos) < max_distance) { - for (uint i = 0; i < sdfgi.max_cascades; i++) { - if (i >= cascade && length(ray_pos) < radius_sizes[i]) { - 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 distance = texture(sampler3D(sdf_cascades[i], linear_sampler), pos).r * 255.0 - 1.1; - - vec4 hit_light = vec4(0.0); - if (distance < softness) { - hit_light.rgb = texture(sampler3D(light_cascades[i], linear_sampler), pos).rgb; - hit_light.rgb *= 0.5; //approximation given value read is actually meant for anisotropy - hit_light.a = clamp(1.0 - (distance / softness), 0.0, 1.0); - hit_light.rgb *= hit_light.a; - } + 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 - distance /= sdfgi.cascades[i].to_cell; + vec3 pos = ray_pos - sdfgi.cascades[i].position; + pos *= sdfgi.cascades[i].to_cell * pos_to_uvw; - if (i < (sdfgi.max_cascades - 1)) { - pos = ray_pos - sdfgi.cascades[i + 1].position; - pos *= sdfgi.cascades[i + 1].to_cell * pos_to_uvw; + float fdistance = textureLod(sampler3D(sdf_cascades[i], linear_sampler), pos, 0.0).r * 255.0 - 1.1; - float distance2 = texture(sampler3D(sdf_cascades[i + 1], linear_sampler), pos).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; + } - vec4 hit_light2 = vec4(0.0); - if (distance2 < softness) { - hit_light2.rgb = texture(sampler3D(light_cascades[i + 1], linear_sampler), pos).rgb; - hit_light2.rgb *= 0.5; //approximation given value read is actually meant for anisotropy - hit_light2.a = clamp(1.0 - (distance2 / softness), 0.0, 1.0); - hit_light2.rgb *= hit_light2.a; - } + fdistance /= sdfgi.cascades[i].to_cell; - float prev_radius = i == 0 ? 0.0 : radius_sizes[i - 1]; - float blend = clamp((length(ray_pos) - prev_radius) / (radius_sizes[i] - prev_radius), 0.0, 1.0); + if (i < (sdfgi.max_cascades - 1)) { + pos = ray_pos - sdfgi.cascades[next_i].position; + pos *= sdfgi.cascades[next_i].to_cell * pos_to_uvw; - distance2 /= sdfgi.cascades[i + 1].to_cell; + float fdistance2 = textureLod(sampler3D(sdf_cascades[next_i], linear_sampler), pos, 0.0).r * 255.0 - 1.1; - hit_light = mix(hit_light, hit_light2, blend); - distance = mix(distance, distance2, blend); + 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; } - light_accum += hit_light; - ray_pos += ray_dir * distance; - break; + 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); } - } - if (light_accum.a > 0.99) { - break; + light_accum += hit_light; + ray_pos += ray_dir * fdistance; + found = true; + } + i++; + if (i == sdfgi.max_cascades) { + i = 0; + found = false; } } @@ -434,8 +437,6 @@ void sdfgi_process(vec3 vertex, vec3 normal, vec3 reflection, float roughness, o } } -#endif - reflection_light.rgb = specular; ambient_light.rgb *= sdfgi.energy; @@ -597,35 +598,24 @@ vec4 fetch_normal_and_roughness(ivec2 pos) { return normal_roughness; } -void main() { - // Pixel being shaded - ivec2 pos = ivec2(gl_GlobalInvocationID.xy); - if (any(greaterThanEqual(pos, params.screen_size))) { //too large, do nothing - return; - } - - vec3 vertex = reconstruct_position(pos); - vertex.y = -vertex.y; - +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; - vec4 ambient_light = vec4(0.0), reflection_light = vec4(0.0); + vec3 normal = normal_roughness.xyz; if (normal.length() > 0.5) { //valid normal, can do GI float roughness = normal_roughness.w; - vertex = mat3(params.cam_rotation) * vertex; normal = normalize(mat3(params.cam_rotation) * normal); - vec3 reflection = normalize(reflect(normalize(vertex), normal)); - if (params.use_sdfgi) { - sdfgi_process(vertex, normal, reflection, roughness, ambient_light, reflection_light); - } +#ifdef USE_SDFGI + sdfgi_process(vertex, normal, reflection, roughness, ambient_light, reflection_light); +#endif - if (params.max_giprobes > 0) { +#ifdef USE_GIPROBES + { uvec2 giprobe_tex = texelFetch(usampler2D(giprobe_buffer, linear_sampler), pos, 0).rg; roughness *= roughness; //find arbitrary tangent and bitangent, then build a matrix @@ -648,16 +638,40 @@ void main() { spec_accum /= blend_accum; } - if (params.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; - } +#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); + +#ifdef MODE_HALF_RES + pos <<= 1; +#endif + if (any(greaterThanEqual(pos, params.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); + +#ifdef MODE_HALF_RES + pos >>= 1; +#endif + imageStore(ambient_buffer, pos, ambient_light); imageStore(reflection_buffer, pos, reflection_light); } diff --git a/servers/rendering/renderer_rd/shaders/giprobe.glsl b/servers/rendering/renderer_rd/shaders/giprobe.glsl index 4f4753d147..b931461b31 100644 --- a/servers/rendering/renderer_rd/shaders/giprobe.glsl +++ b/servers/rendering/renderer_rd/shaders/giprobe.glsl @@ -51,10 +51,10 @@ struct Light { float attenuation; vec3 color; - float spot_angle_radians; + float cos_spot_angle; vec3 position; - float spot_attenuation; + float inv_spot_attenuation; vec3 direction; bool has_shadow; @@ -233,13 +233,15 @@ bool compute_light_vector(uint light, vec3 pos, out float attenuation, out vec3 if (lights.data[light].type == LIGHT_TYPE_SPOT) { vec3 rel = normalize(pos - light_pos); - float angle = acos(dot(rel, lights.data[light].direction)); - if (angle > lights.data[light].spot_angle_radians) { + 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 d = clamp(angle / lights.data[light].spot_angle_radians, 0, 1); - attenuation *= pow(1.0 - d, lights.data[light].spot_attenuation); + 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); } } diff --git a/servers/rendering/renderer_rd/shaders/giprobe_write.glsl b/servers/rendering/renderer_rd/shaders/giprobe_write.glsl index 9c794f1bcc..56b3b7ccb4 100644 --- a/servers/rendering/renderer_rd/shaders/giprobe_write.glsl +++ b/servers/rendering/renderer_rd/shaders/giprobe_write.glsl @@ -43,10 +43,10 @@ struct Light { float attenuation; vec3 color; - float spot_angle_radians; + float cos_spot_angle; vec3 position; - float spot_attenuation; + float inv_spot_attenuation; vec3 direction; bool has_shadow; @@ -146,13 +146,15 @@ bool compute_light_vector(uint light, uint cell, vec3 pos, out float attenuation if (lights.data[light].type == LIGHT_TYPE_SPOT) { vec3 rel = normalize(pos - light_pos); - float angle = acos(dot(rel, lights.data[light].direction)); - if (angle > lights.data[light].spot_angle_radians) { + 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 d = clamp(angle / lights.data[light].spot_angle_radians, 0, 1); - attenuation *= pow(1.0 - d, lights.data[light].spot_attenuation); + 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); } } diff --git a/servers/rendering/renderer_rd/shaders/particles.glsl b/servers/rendering/renderer_rd/shaders/particles.glsl index 926c7ef9fc..cb6d8dc7f6 100644 --- a/servers/rendering/renderer_rd/shaders/particles.glsl +++ b/servers/rendering/renderer_rd/shaders/particles.glsl @@ -173,7 +173,7 @@ uint hash(uint x) { return x; } -bool emit_particle(mat4 p_xform, vec3 p_velocity, vec4 p_color, vec4 p_custom, uint p_flags) { +bool emit_subparticle(mat4 p_xform, vec3 p_velocity, vec4 p_color, vec4 p_custom, uint p_flags) { if (!params.can_emit) { return false; } diff --git a/servers/rendering/renderer_rd/shaders/resolve.glsl b/servers/rendering/renderer_rd/shaders/resolve.glsl index 9429a66dc9..e83c4ca93b 100644 --- a/servers/rendering/renderer_rd/shaders/resolve.glsl +++ b/servers/rendering/renderer_rd/shaders/resolve.glsl @@ -58,6 +58,116 @@ void main() { #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]; @@ -91,7 +201,7 @@ void main() { 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 GIPROBE_RESOLVE diff --git a/servers/rendering/renderer_rd/shaders/scene_forward.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl index 0518976322..7b86dac143 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl @@ -4,7 +4,7 @@ VERSION_DEFINES -#include "scene_forward_inc.glsl" +#include "scene_forward_clustered_inc.glsl" /* INPUT ATTRIBS */ @@ -89,12 +89,6 @@ MATERIAL_UNIFORMS } material; #endif -/* clang-format off */ - -VERTEX_SHADER_GLOBALS - -/* clang-format on */ - invariant gl_Position; #ifdef MODE_DUAL_PARABOLOID @@ -103,28 +97,43 @@ layout(location = 8) out float dp_clip; #endif +layout(location = 9) out flat uint instance_index; + +/* clang-format off */ + +VERTEX_SHADER_GLOBALS + +/* clang-format on */ + void main() { vec4 instance_custom = vec4(0.0); #if defined(COLOR_USED) color_interp = color_attrib; #endif - mat4 world_matrix = draw_call.transform; + 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; + } + + mat4 world_matrix = instances.data[instance_index].transform; mat3 world_normal_matrix; - if (bool(draw_call.flags & INSTANCE_FLAGS_NON_UNIFORM_SCALE)) { + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_NON_UNIFORM_SCALE)) { world_normal_matrix = inverse(mat3(world_matrix)); } else { world_normal_matrix = mat3(world_matrix); } - if (bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH)) { + if (is_multimesh) { //multimesh, instances are for it - uint offset = (draw_call.flags >> INSTANCE_FLAGS_MULTIMESH_STRIDE_SHIFT) & INSTANCE_FLAGS_MULTIMESH_STRIDE_MASK; + uint offset = (instances.data[instance_index].flags >> INSTANCE_FLAGS_MULTIMESH_STRIDE_SHIFT) & INSTANCE_FLAGS_MULTIMESH_STRIDE_MASK; offset *= gl_InstanceIndex; mat4 matrix; - if (bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH_FORMAT_2D)) { + 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 { @@ -132,14 +141,14 @@ void main() { offset += 3; } - if (bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH_HAS_COLOR)) { + 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(draw_call.flags & INSTANCE_FLAGS_MULTIMESH_HAS_CUSTOM_DATA)) { + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH_HAS_CUSTOM_DATA)) { instance_custom = transforms.data[offset]; } @@ -161,7 +170,7 @@ void main() { #endif #if 0 - if (bool(draw_call.flags & INSTANCE_FLAGS_SKELETON)) { + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_SKELETON)) { //multimesh, instances are for it uvec2 bones_01 = uvec2(bone_attrib.x & 0xFFFF, bone_attrib.x >> 16) * 3; @@ -194,7 +203,7 @@ void main() { uv2_interp = uv2_attrib; #endif -#ifdef USE_OVERRIDE_POSITION +#ifdef OVERRIDE_POSITION vec4 position; #endif @@ -289,7 +298,7 @@ VERTEX_SHADER_CODE #endif //MODE_RENDER_DEPTH -#ifdef USE_OVERRIDE_POSITION +#ifdef OVERRIDE_POSITION gl_Position = position; #else gl_Position = projection_matrix * vec4(vertex_interp, 1.0); @@ -304,7 +313,8 @@ VERTEX_SHADER_CODE #endif #ifdef MODE_RENDER_MATERIAL if (scene_data.material_uv2_mode) { - gl_Position.xy = (uv2_attrib.xy + draw_call.lightmap_uv_scale.xy) * 2.0 - 1.0; + 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; } @@ -317,7 +327,7 @@ VERTEX_SHADER_CODE VERSION_DEFINES -#include "scene_forward_inc.glsl" +#include "scene_forward_clustered_inc.glsl" /* Varyings */ @@ -350,9 +360,11 @@ layout(location = 8) in float dp_clip; #endif +layout(location = 9) in flat uint instance_index; + //defines to keep compatibility with vertex -#define world_matrix draw_call.transform +#define world_matrix instances.data[instance_index].transform #define projection_matrix scene_data.projection_matrix #if defined(ENABLE_SSS) && defined(ENABLE_TRANSMITTANCE) @@ -541,7 +553,7 @@ vec3 F0(float metallic, float specular, vec3 albedo) { return mix(vec3(dielectric), albedo, vec3(metallic)); } -void light_compute(vec3 N, vec3 L, vec3 V, float A, vec3 light_color, float attenuation, vec3 shadow_attenuation, vec3 diffuse_color, float roughness, float metallic, float specular, float specular_blob_intensity, +void light_compute(vec3 N, vec3 L, vec3 V, vec3 light_color, float attenuation, vec3 f0, uint orms, float specular_amount, #ifdef LIGHT_BACKLIGHT_USED vec3 backlight, #endif @@ -553,7 +565,7 @@ void light_compute(vec3 N, vec3 L, vec3 V, float A, vec3 light_color, float atte float transmittance_z, #endif #ifdef LIGHT_RIM_USED - float rim, float rim_tint, + float rim, float rim_tint, vec3 rim_color, #endif #ifdef LIGHT_CLEARCOAT_USED float clearcoat, float clearcoat_gloss, @@ -561,6 +573,9 @@ void light_compute(vec3 N, vec3 L, vec3 V, float A, vec3 light_color, float atte #ifdef LIGHT_ANISOTROPY_USED vec3 B, vec3 T, float anisotropy, #endif +#ifdef USE_SOFT_SHADOWS + float A, +#endif #ifdef USE_SHADOW_TO_OPACITY inout float alpha, #endif @@ -570,7 +585,6 @@ void light_compute(vec3 N, vec3 L, vec3 V, float A, vec3 light_color, float atte // light is written by the light shader vec3 normal = N; - vec3 albedo = diffuse_color; vec3 light = L; vec3 view = V; @@ -581,7 +595,12 @@ LIGHT_SHADER_CODE /* clang-format on */ #else + +#ifdef USE_SOFT_SHADOWS float NdotL = min(A + dot(N, L), 1.0); +#else + float NdotL = dot(N, L); +#endif float cNdotL = max(NdotL, 0.0); // clamped NdotL float NdotV = dot(N, V); float cNdotV = max(NdotV, 0.0); @@ -591,14 +610,25 @@ LIGHT_SHADER_CODE #endif #if defined(SPECULAR_BLINN) || defined(SPECULAR_SCHLICK_GGX) || defined(LIGHT_CLEARCOAT_USED) +#ifdef USE_SOFT_SHADOWS float cNdotH = clamp(A + dot(N, H), 0.0, 1.0); +#else + float cNdotH = clamp(dot(N, H), 0.0, 1.0); +#endif #endif #if defined(DIFFUSE_BURLEY) || defined(SPECULAR_SCHLICK_GGX) || defined(LIGHT_CLEARCOAT_USED) +#ifdef USE_SOFT_SHADOWS float cLdotH = clamp(A + dot(L, H), 0.0, 1.0); +#else + float cLdotH = clamp(dot(L, H), 0.0, 1.0); +#endif #endif + float metallic = unpackUnorm4x8(orms).z; if (metallic < 1.0) { + float roughness = unpackUnorm4x8(orms).y; + #if defined(DIFFUSE_OREN_NAYAR) vec3 diffuse_brdf_NL; #else @@ -608,23 +638,6 @@ LIGHT_SHADER_CODE #if defined(DIFFUSE_LAMBERT_WRAP) // energy conserving lambert wrap shader diffuse_brdf_NL = max(0.0, (NdotL + roughness) / ((1.0 + roughness) * (1.0 + roughness))); - -#elif defined(DIFFUSE_OREN_NAYAR) - - { - // see http://mimosa-pudica.net/improved-oren-nayar.html - float LdotV = dot(L, V); - - float s = LdotV - NdotL * NdotV; - float t = mix(1.0, max(NdotL, NdotV), step(0.0, s)); - - float sigma2 = roughness * roughness; // TODO: this needs checking - vec3 A = 1.0 + sigma2 * (-0.5 / (sigma2 + 0.33) + 0.17 * diffuse_color / (sigma2 + 0.13)); - float B = 0.45 * sigma2 / (sigma2 + 0.09); - - diffuse_brdf_NL = cNdotL * (A + vec3(B) * s / t) * (1.0 / M_PI); - } - #elif defined(DIFFUSE_TOON) diffuse_brdf_NL = smoothstep(-roughness, max(roughness, 0.01), NdotL); @@ -652,15 +665,15 @@ LIGHT_SHADER_CODE diffuse_brdf_NL = cNdotL * (1.0 / M_PI); #endif - diffuse_light += light_color * diffuse_color * shadow_attenuation * diffuse_brdf_NL * attenuation; + diffuse_light += light_color * diffuse_brdf_NL * attenuation; #if defined(LIGHT_BACKLIGHT_USED) - diffuse_light += light_color * diffuse_color * (vec3(1.0 / M_PI) - diffuse_brdf_NL) * backlight * attenuation; + 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), diffuse_color, rim_tint) * light_color; + diffuse_light += rim_light * rim * mix(vec3(1.0), rim_color, rim_tint) * light_color; #endif #ifdef LIGHT_TRANSMITTANCE_USED @@ -678,7 +691,7 @@ LIGHT_SHADER_CODE 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 * diffuse_color * light_color * clamp(transmittance_boost - NdotL, 0.0, 1.0) * (1.0 / M_PI) * attenuation; + diffuse_light += profile * transmittance_color.a * light_color * clamp(transmittance_boost - NdotL, 0.0, 1.0) * (1.0 / M_PI); } #else @@ -688,7 +701,7 @@ LIGHT_SHADER_CODE fade = pow(max(0.0, 1.0 - fade), transmittance_curve); fade *= clamp(transmittance_boost - NdotL, 0.0, 1.0); - diffuse_light += diffuse_color * transmittance_color.rgb * light_color * (1.0 / M_PI) * transmittance_color.a * fade * attenuation; + diffuse_light += transmittance_color.rgb * light_color * (1.0 / M_PI) * transmittance_color.a * fade; } #endif //SSS_MODE_SKIN @@ -696,6 +709,7 @@ LIGHT_SHADER_CODE #endif //LIGHT_TRANSMITTANCE_USED } + float roughness = unpackUnorm4x8(orms).y; if (roughness > 0.0) { // FIXME: roughness == 0 should not disable specular light entirely // D @@ -708,7 +722,7 @@ LIGHT_SHADER_CODE blinn *= (shininess + 8.0) * (1.0 / (8.0 * M_PI)); float intensity = blinn; - specular_light += light_color * shadow_attenuation * intensity * specular_blob_intensity * attenuation; + specular_light += light_color * intensity * attenuation * specular_amount; #elif defined(SPECULAR_PHONG) @@ -719,7 +733,7 @@ LIGHT_SHADER_CODE phong *= (shininess + 8.0) * (1.0 / (8.0 * M_PI)); float intensity = (phong) / max(4.0 * cNdotV * cNdotL, 0.75); - specular_light += light_color * shadow_attenuation * intensity * specular_blob_intensity * attenuation; + specular_light += light_color * intensity * attenuation * specular_amount; #elif defined(SPECULAR_TOON) @@ -728,7 +742,7 @@ LIGHT_SHADER_CODE float mid = 1.0 - roughness; mid *= mid; float intensity = smoothstep(mid - roughness * 0.5, mid + roughness * 0.5, RdotV) * mid; - diffuse_light += light_color * shadow_attenuation * intensity * specular_blob_intensity * attenuation; // write to diffuse_light, as in toon shading you generally want no reflection + 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.. @@ -753,13 +767,12 @@ LIGHT_SHADER_CODE float G = G_GGX_2cos(cNdotL, alpha_ggx) * G_GGX_2cos(cNdotV, alpha_ggx); #endif // F - vec3 f0 = F0(metallic, specular, diffuse_color); float cLdotH5 = SchlickFresnel(cLdotH); vec3 F = mix(vec3(cLdotH5), vec3(1.0), f0); vec3 specular_brdf_NL = cNdotL * D * F * G; - specular_light += specular_brdf_NL * light_color * shadow_attenuation * specular_blob_intensity * attenuation; + specular_light += specular_brdf_NL * light_color * attenuation * specular_amount; #endif #if defined(LIGHT_CLEARCOAT_USED) @@ -773,12 +786,12 @@ LIGHT_SHADER_CODE float clearcoat_specular_brdf_NL = 0.25 * clearcoat * Gr * Fr * Dr * cNdotL; - specular_light += clearcoat_specular_brdf_NL * light_color * shadow_attenuation * specular_blob_intensity * attenuation; + specular_light += clearcoat_specular_brdf_NL * light_color * attenuation * specular_amount; #endif } #ifdef USE_SHADOW_TO_OPACITY - alpha = min(alpha, clamp(1.0 - length(shadow_attenuation * attenuation), 0.0, 1.0)); + alpha = min(alpha, clamp(1.0 - attenuation), 0.0, 1.0)); #endif #endif //defined(USE_LIGHT_SHADER_CODE) @@ -786,13 +799,11 @@ LIGHT_SHADER_CODE #ifndef USE_NO_SHADOWS -// Produces cheap white noise, optimized for window-space -// Comes from: https://www.shadertoy.com/view/4djSRW -// Copyright: Dave Hoskins, MIT License +// Interleaved Gradient Noise +// http://www.iryoku.com/next-generation-post-processing-in-call-of-duty-advanced-warfare float quick_hash(vec2 pos) { - vec3 p3 = fract(vec3(pos.xyx) * .1031); - p3 += dot(p3, p3.yzx + 33.33); - return fract((p3.x + p3.y) * p3.z); + 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) { @@ -900,68 +911,30 @@ float get_omni_attenuation(float distance, float inv_range, float decay) { return nd * pow(max(distance, 0.0001), -decay); } -void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 vertex_ddx, vec3 vertex_ddy, vec3 albedo, float roughness, float metallic, float specular, float p_blob_intensity, -#ifdef LIGHT_BACKLIGHT_USED - vec3 backlight, -#endif -#ifdef LIGHT_TRANSMITTANCE_USED - vec4 transmittance_color, - float transmittance_depth, - float transmittance_curve, - float transmittance_boost, -#endif -#ifdef LIGHT_RIM_USED - float rim, float rim_tint, -#endif -#ifdef LIGHT_CLEARCOAT_USED - float clearcoat, float clearcoat_gloss, -#endif -#ifdef LIGHT_ANISOTROPY_USED - vec3 binormal, vec3 tangent, float anisotropy, -#endif -#ifdef USE_SHADOW_TO_OPACITY - inout float alpha, -#endif - inout vec3 diffuse_light, inout vec3 specular_light) { - vec3 light_rel_vec = lights.data[idx].position - vertex; - float light_length = length(light_rel_vec); - vec2 attenuation_energy = unpackHalf2x16(lights.data[idx].attenuation_energy); - float omni_attenuation = get_omni_attenuation(light_length, lights.data[idx].inv_radius, attenuation_energy.x); - float light_attenuation = omni_attenuation; - vec3 shadow_attenuation = vec3(1.0); - vec4 color_specular = unpackUnorm4x8(lights.data[idx].color_specular); - color_specular.rgb *= attenuation_energy.y; - float size_A = 0.0; - - if (lights.data[idx].size > 0.0) { - float t = 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 -#endif - +float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { #ifndef USE_NO_SHADOWS - vec4 shadow_color_enabled = unpackUnorm4x8(lights.data[idx].shadow_color_enabled); - if (shadow_color_enabled.w > 0.5) { + if (omni_lights.data[idx].shadow_enabled) { // there is a shadowmap + vec3 light_rel_vec = omni_lights.data[idx].position - vertex; + float light_length = length(light_rel_vec); + vec4 v = vec4(vertex, 1.0); - vec4 splane = (lights.data[idx].shadow_matrix * v); + vec4 splane = (omni_lights.data[idx].shadow_matrix * v); float shadow_len = length(splane.xyz); //need to remember shadow len from here { - vec3 nofs = normal_interp * lights.data[idx].shadow_normal_bias / lights.data[idx].inv_radius; + vec3 nofs = normal_interp * omni_lights.data[idx].shadow_normal_bias / omni_lights.data[idx].inv_radius; nofs *= (1.0 - max(0.0, dot(normalize(light_rel_vec), normalize(normal_interp)))); v.xyz += nofs; - splane = (lights.data[idx].shadow_matrix * v); + splane = (omni_lights.data[idx].shadow_matrix * v); } float shadow; - if (lights.data[idx].soft_shadow_size > 0.0) { +#ifdef USE_SOFT_SHADOWS + if (omni_lights.data[idx].soft_shadow_size > 0.0) { //soft shadow //find blocker @@ -981,10 +954,10 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v 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)); - float z_norm = shadow_len * lights.data[idx].inv_radius; + float z_norm = shadow_len * omni_lights.data[idx].inv_radius; - tangent *= lights.data[idx].soft_shadow_size * lights.data[idx].soft_shadow_scale; - bitangent *= lights.data[idx].soft_shadow_size * lights.data[idx].soft_shadow_scale; + 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 < scene_data.penumbra_shadow_samples; i++) { vec2 disk = disk_rotation * scene_data.penumbra_shadow_kernel[i].xy; @@ -992,7 +965,7 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v vec3 pos = splane.xyz + tangent * disk.x + bitangent * disk.y; pos = normalize(pos); - vec4 uv_rect = lights.data[idx].atlas_rect; + vec4 uv_rect = omni_lights.data[idx].atlas_rect; if (pos.z >= 0.0) { pos.z += 1.0; @@ -1020,7 +993,7 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v tangent *= penumbra; bitangent *= penumbra; - z_norm -= lights.data[idx].inv_radius * lights.data[idx].shadow_bias; + z_norm -= omni_lights.data[idx].inv_radius * omni_lights.data[idx].shadow_bias; shadow = 0.0; for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { @@ -1028,7 +1001,7 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v vec3 pos = splane.xyz + tangent * disk.x + bitangent * disk.y; pos = normalize(pos); - vec4 uv_rect = lights.data[idx].atlas_rect; + vec4 uv_rect = omni_lights.data[idx].atlas_rect; if (pos.z >= 0.0) { pos.z += 1.0; @@ -1051,8 +1024,9 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v shadow = 1.0; } } else { +#endif splane.xyz = normalize(splane.xyz); - vec4 clamp_rect = lights.data[idx].atlas_rect; + vec4 clamp_rect = omni_lights.data[idx].atlas_rect; if (splane.z >= 0.0) { splane.z += 1.0; @@ -1066,101 +1040,149 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v splane.xy /= splane.z; splane.xy = splane.xy * 0.5 + 0.5; - splane.z = (shadow_len - lights.data[idx].shadow_bias) * lights.data[idx].inv_radius; + splane.z = (shadow_len - omni_lights.data[idx].shadow_bias) * omni_lights.data[idx].inv_radius; splane.xy = clamp_rect.xy + splane.xy * clamp_rect.zw; splane.w = 1.0; //needed? i think it should be 1 already - shadow = sample_pcf_shadow(shadow_atlas, lights.data[idx].soft_shadow_scale * scene_data.shadow_atlas_pixel_size, splane); + shadow = sample_pcf_shadow(shadow_atlas, omni_lights.data[idx].soft_shadow_scale * scene_data.shadow_atlas_pixel_size, splane); +#ifdef USE_SOFT_SHADOWS } +#endif + + 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, +#ifdef LIGHT_BACKLIGHT_USED + vec3 backlight, +#endif #ifdef LIGHT_TRANSMITTANCE_USED - { - vec4 clamp_rect = lights.data[idx].atlas_rect; + vec4 transmittance_color, + float transmittance_depth, + float transmittance_curve, + float transmittance_boost, +#endif +#ifdef LIGHT_RIM_USED + float rim, float rim_tint, vec3 rim_color, +#endif +#ifdef LIGHT_CLEARCOAT_USED + float clearcoat, float clearcoat_gloss, +#endif +#ifdef LIGHT_ANISOTROPY_USED + vec3 binormal, vec3 tangent, float anisotropy, +#endif +#ifdef USE_SHADOW_TO_OPACITY + inout float alpha, +#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; - //redo shadowmapping, but shrink the model a bit to avoid arctifacts - splane = (lights.data[idx].shadow_matrix * vec4(vertex - normalize(normal_interp) * lights.data[idx].transmittance_bias, 1.0)); +#ifdef USE_SOFT_SHADOWS + float size_A = 0.0; - shadow_len = length(splane.xyz); - splane = normalize(splane.xyz); + if (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)); + } +#endif - if (splane.z >= 0.0) { - splane.z += 1.0; +#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; - } else { - splane.z = 1.0 - splane.z; - } + //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)); - splane.xy /= splane.z; - splane.xy = splane.xy * 0.5 + 0.5; - splane.z = shadow_len * lights.data[idx].inv_radius; - splane.xy = clamp_rect.xy + splane.xy * clamp_rect.zw; - splane.w = 1.0; //needed? i think it should be 1 already + shadow_len = length(splane.xyz); + splane = normalize(splane.xyz); + + if (splane.z >= 0.0) { + splane.z += 1.0; - float shadow_z = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), splane.xy, 0.0).r; - transmittance_z = (splane.z - shadow_z) / lights.data[idx].inv_radius; + } else { + splane.z = 1.0 - splane.z; } -#endif - vec3 no_shadow = vec3(1.0); + 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.w = 1.0; //needed? i think it should be 1 already - if (lights.data[idx].projector_rect != vec4(0.0)) { - vec3 local_v = (lights.data[idx].shadow_matrix * vec4(vertex, 1.0)).xyz; - local_v = normalize(local_v); + 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 - vec4 atlas_rect = lights.data[idx].projector_rect; +#if 0 - if (local_v.z >= 0.0) { - local_v.z += 1.0; - atlas_rect.y += atlas_rect.w; + if (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); - } else { - local_v.z = 1.0 - local_v.z; - } + vec4 atlas_rect = omni_lights.data[idx].projector_rect; - 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 (local_v.z >= 0.0) { + local_v.z += 1.0; + atlas_rect.y += atlas_rect.w; - vec2 proj_uv_ddx; - vec2 proj_uv_ddy; - { - vec3 local_v_ddx = (lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddx, 1.0)).xyz; - local_v_ddx = normalize(local_v_ddx); + } else { + local_v.z = 1.0 - local_v.z; + } - 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.xy /= local_v.z; + local_v.xy = local_v.xy * 0.5 + 0.5; + vec2 proj_uv = local_v.xy * atlas_rect.zw; - local_v_ddx.xy /= local_v_ddx.z; - local_v_ddx.xy = local_v_ddx.xy * 0.5 + 0.5; + 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); - proj_uv_ddx = local_v_ddx.xy * atlas_rect.zw - proj_uv; + if (local_v_ddx.z >= 0.0) { + local_v_ddx.z += 1.0; + } else { + local_v_ddx.z = 1.0 - local_v_ddx.z; + } - vec3 local_v_ddy = (lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddy, 1.0)).xyz; - local_v_ddy = normalize(local_v_ddy); + local_v_ddx.xy /= local_v_ddx.z; + local_v_ddx.xy = local_v_ddx.xy * 0.5 + 0.5; - if (local_v_ddy.z >= 0.0) { - local_v_ddy.z += 1.0; - } else { - local_v_ddy.z = 1.0 - local_v_ddy.z; - } + proj_uv_ddx = local_v_ddx.xy * atlas_rect.zw - proj_uv; - local_v_ddy.xy /= local_v_ddy.z; - local_v_ddy.xy = local_v_ddy.xy * 0.5 + 0.5; + vec3 local_v_ddy = (omni_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddy, 1.0)).xyz; + local_v_ddy = normalize(local_v_ddy); - proj_uv_ddy = local_v_ddy.xy * atlas_rect.zw - proj_uv; + if (local_v_ddy.z >= 0.0) { + local_v_ddy.z += 1.0; + } else { + local_v_ddy.z = 1.0 - local_v_ddy.z; } - vec4 proj = textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), proj_uv + atlas_rect.xy, proj_uv_ddx, proj_uv_ddy); - no_shadow = mix(no_shadow, proj.rgb, proj.a); + 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; } - shadow_attenuation = mix(shadow_color_enabled.rgb, no_shadow, shadow); + vec4 proj = textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), proj_uv + atlas_rect.xy, proj_uv_ddx, proj_uv_ddy); + no_shadow = mix(no_shadow, proj.rgb, proj.a); } -#endif //USE_NO_SHADOWS +#endif + + light_attenuation *= shadow; - light_compute(normal, normalize(light_rel_vec), eye_vec, size_A, color_specular.rgb, light_attenuation, shadow_attenuation, albedo, roughness, metallic, specular, color_specular.a * p_blob_intensity, + light_compute(normal, normalize(light_rel_vec), eye_vec, color, light_attenuation, f0, orms, omni_lights.data[idx].specular_amount, #ifdef LIGHT_BACKLIGHT_USED backlight, #endif @@ -1172,7 +1194,7 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v transmittance_z, #endif #ifdef LIGHT_RIM_USED - rim * omni_attenuation, rim_tint, + rim * omni_attenuation, rim_tint, rim_color, #endif #ifdef LIGHT_CLEARCOAT_USED clearcoat, clearcoat_gloss, @@ -1180,6 +1202,9 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v #ifdef LIGHT_ANISOTROPY_USED binormal, tangent, anisotropy, #endif +#ifdef USE_SOFT_SHADOWS + size_A, +#endif #ifdef USE_SHADOW_TO_OPACITY alpha, #endif @@ -1187,88 +1212,39 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v specular_light); } -void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 vertex_ddx, vec3 vertex_ddy, vec3 albedo, float roughness, float metallic, float specular, float p_blob_intensity, -#ifdef LIGHT_BACKLIGHT_USED - vec3 backlight, -#endif -#ifdef LIGHT_TRANSMITTANCE_USED - vec4 transmittance_color, - float transmittance_depth, - float transmittance_curve, - float transmittance_boost, -#endif -#ifdef LIGHT_RIM_USED - float rim, float rim_tint, -#endif -#ifdef LIGHT_CLEARCOAT_USED - float clearcoat, float clearcoat_gloss, -#endif -#ifdef LIGHT_ANISOTROPY_USED - vec3 binormal, vec3 tangent, float anisotropy, -#endif -#ifdef USE_SHADOW_TO_OPACITY - inout float alpha, -#endif - inout vec3 diffuse_light, - inout vec3 specular_light) { - vec3 light_rel_vec = lights.data[idx].position - vertex; - float light_length = length(light_rel_vec); - vec2 attenuation_energy = unpackHalf2x16(lights.data[idx].attenuation_energy); - float spot_attenuation = get_omni_attenuation(light_length, lights.data[idx].inv_radius, attenuation_energy.x); - vec3 spot_dir = lights.data[idx].direction; - vec2 spot_att_angle = unpackHalf2x16(lights.data[idx].cone_attenuation_angle); - float scos = max(dot(-normalize(light_rel_vec), spot_dir), spot_att_angle.y); - float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - spot_att_angle.y)); - spot_attenuation *= 1.0 - pow(spot_rim, spot_att_angle.x); - float light_attenuation = spot_attenuation; - vec3 shadow_attenuation = vec3(1.0); - vec4 color_specular = unpackUnorm4x8(lights.data[idx].color_specular); - color_specular.rgb *= attenuation_energy.y; - - float size_A = 0.0; - - if (lights.data[idx].size > 0.0) { - float t = lights.data[idx].size / max(0.001, light_length); - size_A = max(0.0, 1.0 - 1 / sqrt(1 + t * t)); - } -/* - if (lights.data[idx].atlas_rect!=vec4(0.0)) { - //use projector texture - } - */ -#ifdef LIGHT_TRANSMITTANCE_USED - float transmittance_z = transmittance_depth; -#endif - +float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { #ifndef USE_NO_SHADOWS - vec4 shadow_color_enabled = unpackUnorm4x8(lights.data[idx].shadow_color_enabled); - if (shadow_color_enabled.w > 0.5) { + if (spot_lights.data[idx].shadow_enabled) { + 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; //there is a shadowmap vec4 v = vec4(vertex, 1.0); - v.xyz -= spot_dir * lights.data[idx].shadow_bias; + v.xyz -= spot_dir * spot_lights.data[idx].shadow_bias; - float z_norm = dot(spot_dir, -light_rel_vec) * lights.data[idx].inv_radius; + float z_norm = dot(spot_dir, -light_rel_vec) * spot_lights.data[idx].inv_radius; float depth_bias_scale = 1.0 / (max(0.0001, z_norm)); //the closer to the light origin, the more you have to offset to reach 1px in the map - vec3 normal_bias = normalize(normal_interp) * (1.0 - max(0.0, dot(spot_dir, -normalize(normal_interp)))) * lights.data[idx].shadow_normal_bias * depth_bias_scale; + vec3 normal_bias = normalize(normal_interp) * (1.0 - max(0.0, dot(spot_dir, -normalize(normal_interp)))) * spot_lights.data[idx].shadow_normal_bias * depth_bias_scale; normal_bias -= spot_dir * dot(spot_dir, normal_bias); //only XY, no Z v.xyz += normal_bias; //adjust with bias - z_norm = dot(spot_dir, v.xyz - lights.data[idx].position) * lights.data[idx].inv_radius; + z_norm = dot(spot_dir, v.xyz - spot_lights.data[idx].position) * spot_lights.data[idx].inv_radius; float shadow; - vec4 splane = (lights.data[idx].shadow_matrix * v); + vec4 splane = (spot_lights.data[idx].shadow_matrix * v); splane /= splane.w; - if (lights.data[idx].soft_shadow_size > 0.0) { +#ifdef USE_SOFT_SHADOWS + if (spot_lights.data[idx].soft_shadow_size > 0.0) { //soft shadow //find blocker - vec2 shadow_uv = splane.xy * lights.data[idx].atlas_rect.zw + lights.data[idx].atlas_rect.xy; + 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; @@ -1281,11 +1257,11 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); } - float uv_size = lights.data[idx].soft_shadow_size * z_norm * lights.data[idx].soft_shadow_scale; - vec2 clamp_max = lights.data[idx].atlas_rect.xy + lights.data[idx].atlas_rect.zw; + 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 < scene_data.penumbra_shadow_samples; i++) { vec2 suv = shadow_uv + (disk_rotation * scene_data.penumbra_shadow_kernel[i].xy) * uv_size; - suv = clamp(suv, lights.data[idx].atlas_rect.xy, clamp_max); + 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 < z_norm) { blocker_average += d; @@ -1302,7 +1278,7 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v shadow = 0.0; for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { vec2 suv = shadow_uv + (disk_rotation * scene_data.penumbra_shadow_kernel[i].xy) * uv_size; - suv = clamp(suv, lights.data[idx].atlas_rect.xy, clamp_max); + suv = clamp(suv, spot_lights.data[idx].atlas_rect.xy, clamp_max); shadow += textureProj(sampler2DShadow(shadow_atlas, shadow_sampler), vec4(suv, z_norm, 1.0)); } @@ -1314,54 +1290,93 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v } } else { +#endif //hard shadow - vec4 shadow_uv = vec4(splane.xy * lights.data[idx].atlas_rect.zw + lights.data[idx].atlas_rect.xy, z_norm, 1.0); + vec4 shadow_uv = vec4(splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy, splane.z, 1.0); - shadow = sample_pcf_shadow(shadow_atlas, lights.data[idx].soft_shadow_scale * scene_data.shadow_atlas_pixel_size, shadow_uv); + shadow = sample_pcf_shadow(shadow_atlas, spot_lights.data[idx].soft_shadow_scale * scene_data.shadow_atlas_pixel_size, shadow_uv); +#ifdef USE_SOFT_SHADOWS } +#endif - vec3 no_shadow = vec3(1.0); + return shadow; + } - if (lights.data[idx].projector_rect != vec4(0.0)) { - splane = (lights.data[idx].shadow_matrix * vec4(vertex, 1.0)); - splane /= splane.w; +#endif //USE_NO_SHADOWS - vec2 proj_uv = splane.xy * lights.data[idx].projector_rect.zw; + return 1.0; +} - //ensure we have proper mipmaps - vec4 splane_ddx = (lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddx, 1.0)); - splane_ddx /= splane_ddx.w; - vec2 proj_uv_ddx = splane_ddx.xy * lights.data[idx].projector_rect.zw - proj_uv; +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, +#ifdef LIGHT_BACKLIGHT_USED + vec3 backlight, +#endif +#ifdef LIGHT_TRANSMITTANCE_USED + vec4 transmittance_color, + float transmittance_depth, + float transmittance_curve, + float transmittance_boost, +#endif +#ifdef LIGHT_RIM_USED + float rim, float rim_tint, vec3 rim_color, +#endif +#ifdef LIGHT_CLEARCOAT_USED + float clearcoat, float clearcoat_gloss, +#endif +#ifdef LIGHT_ANISOTROPY_USED + vec3 binormal, vec3 tangent, float anisotropy, +#endif +#ifdef USE_SHADOW_TO_OPACITY + inout float alpha, +#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; - vec4 splane_ddy = (lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddy, 1.0)); - splane_ddy /= splane_ddy.w; - vec2 proj_uv_ddy = splane_ddy.xy * lights.data[idx].projector_rect.zw - proj_uv; +#ifdef USE_SOFT_SHADOWS + float size_A = 0.0; - vec4 proj = textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), proj_uv + lights.data[idx].projector_rect.xy, proj_uv_ddx, proj_uv_ddy); - no_shadow = mix(no_shadow, proj.rgb, proj.a); - } + if (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)); + } +#endif - shadow_attenuation = mix(shadow_color_enabled.rgb, no_shadow, shadow); + /* + if (spot_lights.data[idx].atlas_rect!=vec4(0.0)) { + //use projector texture + } + */ #ifdef LIGHT_TRANSMITTANCE_USED - { - splane = (lights.data[idx].shadow_matrix * vec4(vertex - normalize(normal_interp) * lights.data[idx].transmittance_bias, 1.0)); - splane /= splane.w; - splane.xy = splane.xy * lights.data[idx].atlas_rect.zw + lights.data[idx].atlas_rect.xy; - - float shadow_z = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), splane.xy, 0.0).r; - //reconstruct depth - shadow_z /= lights.data[idx].inv_radius; - //distance to light plane - float z = dot(spot_dir, -light_rel_vec); - transmittance_z = z - shadow_z; - } -#endif //LIGHT_TRANSMITTANCE_USED + float transmittance_z = transmittance_depth; + transmittance_color.a *= light_attenuation; + { + 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; + //reconstruct depth + shadow_z /= spot_lights.data[idx].inv_radius; + //distance to light plane + float z = dot(spot_dir, -light_rel_vec); + transmittance_z = z - shadow_z; } +#endif //LIGHT_TRANSMITTANCE_USED -#endif //USE_NO_SHADOWS + light_attenuation *= shadow; - light_compute(normal, normalize(light_rel_vec), eye_vec, size_A, color_specular.rgb, light_attenuation, shadow_attenuation, albedo, roughness, metallic, specular, color_specular.a * p_blob_intensity, + light_compute(normal, normalize(light_rel_vec), eye_vec, color, light_attenuation, f0, orms, spot_lights.data[idx].specular_amount, #ifdef LIGHT_BACKLIGHT_USED backlight, #endif @@ -1373,7 +1388,7 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v transmittance_z, #endif #ifdef LIGHT_RIM_USED - rim * spot_attenuation, rim_tint, + rim * spot_attenuation, rim_tint, rim_color, #endif #ifdef LIGHT_CLEARCOAT_USED clearcoat, clearcoat_gloss, @@ -1381,6 +1396,9 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v #ifdef LIGHT_ANISOTROPY_USED binormal, tangent, anisotropy, #endif +#ifdef USE_SOFT_SHADOW + size_A, +#endif #ifdef USE_SHADOW_TO_OPACITY alpha, #endif @@ -1404,11 +1422,11 @@ void reflection_process(uint ref_index, vec3 vertex, vec3 normal, float roughnes blend *= blend; blend = max(0.0, 1.0 - blend); - if (reflections.data[ref_index].params.x > 0.0) { // compute reflection + 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].params.w > 0.5) { //box project + if (reflections.data[ref_index].box_project) { //box project vec3 nrdir = normalize(local_ref_vec); vec3 rbmax = (box_extents - local_pos) / nrdir; @@ -1425,11 +1443,11 @@ void reflection_process(uint ref_index, vec3 vertex, vec3 normal, float roughnes 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; - if (reflections.data[ref_index].params.z < 0.5) { + if (reflections.data[ref_index].exterior) { reflection.rgb = mix(specular_light, reflection.rgb, blend); } - reflection.rgb *= reflections.data[ref_index].params.x; + reflection.rgb *= reflections.data[ref_index].intensity; //intensity reflection.a = blend; reflection.rgb *= reflection.a; @@ -1448,7 +1466,7 @@ void reflection_process(uint ref_index, vec3 vertex, vec3 normal, float roughnes 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.a = blend; - if (reflections.data[ref_index].params.z < 0.5) { //interior + if (reflections.data[ref_index].exterior) { ambient_out.rgb = mix(ambient_light, ambient_out.rgb, blend); } @@ -1459,7 +1477,7 @@ void reflection_process(uint ref_index, vec3 vertex, vec3 normal, float roughnes vec4 ambient_out; ambient_out.a = blend; ambient_out.rgb = reflections.data[ref_index].ambient; - if (reflections.data[ref_index].params.z < 0.5) { + if (reflections.data[ref_index].exterior) { ambient_out.rgb = mix(ambient_light, ambient_out.rgb, blend); } ambient_out.rgb *= ambient_out.a; @@ -1762,7 +1780,7 @@ vec4 fog_process(vec3 vertex) { } } - float fog_amount = 1.0 - exp(vertex.z * scene_data.fog_density); + float fog_amount = 1.0 - exp(min(0.0, vertex.z * scene_data.fog_density)); if (abs(scene_data.fog_height_density) > 0.001) { float y = (scene_data.camera_matrix * vec4(vertex, 1.0)).y; @@ -1777,7 +1795,43 @@ vec4 fog_process(vec3 vertex) { 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 & 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 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 +} + +#endif //!MODE_RENDER DEPTH void main() { #ifdef MODE_DUAL_PARABOLOID @@ -1805,9 +1859,7 @@ void main() { float clearcoat_gloss = 0.0; float anisotropy = 0.0; vec2 anisotropy_flow = vec2(1.0, 0.0); -#if defined(CUSTOM_FOG_USED) - vec4 custom_fog = vec4(0.0); -#endif + vec4 fog = vec4(0.0); #if defined(CUSTOM_RADIANCE_USED) vec4 custom_radiance = vec4(0.0); #endif @@ -1815,10 +1867,8 @@ void main() { vec4 custom_irradiance = vec4(0.0); #endif -#if defined(AO_USED) float ao = 1.0; float ao_light_affect = 0.0; -#endif float alpha = 1.0; @@ -1956,77 +2006,147 @@ FRAGMENT_SHADER_CODE 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); + } + +#ifndef LOW_END_MODE + if (scene_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 //!LOW_END_MODE +#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 - uvec4 cluster_cell = texture(usampler3D(cluster_texture, material_samplers[SAMPLER_NEAREST_CLAMP]), vec3(screen_uv, (abs(vertex.z) - scene_data.z_near) / (scene_data.z_far - scene_data.z_near))); + uvec2 cluster_pos = uvec2(gl_FragCoord.xy) >> scene_data.cluster_shift; + uint cluster_offset = (scene_data.cluster_width * cluster_pos.y + cluster_pos.x) * (scene_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 decal_count = cluster_cell.w >> CLUSTER_COUNTER_SHIFT; - uint decal_pointer = cluster_cell.w & CLUSTER_POINTER_MASK; + uint cluster_decal_offset = cluster_offset + scene_data.cluster_type_size * 2; - //do outside for performance and avoiding arctifacts + uint item_min; + uint item_max; + uint item_from; + uint item_to; - for (uint i = 0; i < decal_count; i++) { - uint decal_index = cluster_data.indices[decal_pointer + i]; - if (!bool(decals.data[decal_index].mask & draw_call.layer_mask)) { - continue; //not masked - } + cluster_get_item_range(cluster_decal_offset + scene_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); - 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 - } +#ifdef USE_SUBGROUPS + item_from = subgroupBroadcastFirst(subgroupMin(item_from)); + item_to = subgroupBroadcastFirst(subgroupMax(item_to)); +#endif - //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; + 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 - 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); + 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 decal_index = 32 * i + bit; - 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); - } + if (!bool(decals.data[decal_index].mask & instances.data[instance_index].layer_mask)) { + continue; //not masked + } - if (decals.data[decal_index].albedo_rect != vec4(0.0)) { - //has albedo - vec4 decal_albedo = textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), 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); - 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 = textureGrad(sampler2D(decal_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), 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; - 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)); + 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 } - if (decals.data[decal_index].orm_rect != vec4(0.0)) { - vec3 decal_orm = textureGrad(sampler2D(decal_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), 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; -#if defined(AO_USED) - ao = mix(ao, decal_orm.r, decal_albedo.a); -#endif - roughness = mix(roughness, decal_orm.g, decal_albedo.a); - metallic = mix(metallic, decal_orm.b, decal_albedo.a); + //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; + + 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); + } + + if (decals.data[decal_index].albedo_rect != vec4(0.0)) { + //has albedo + vec4 decal_albedo = textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), 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); + 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 = textureGrad(sampler2D(decal_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), 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; + 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 = textureGrad(sampler2D(decal_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), 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; + 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 - emission += textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), 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; + if (decals.data[decal_index].emission_rect != vec4(0.0)) { + //emission is additive, so its independent from albedo + emission += textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), 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; + } } } } + //pack albedo until needed again, saves 2 VGPRs in the meantime + #endif //not render depth /////////////////////// LIGHTING ////////////////////////////// @@ -2094,19 +2214,14 @@ FRAGMENT_SHADER_CODE //radiance - float specular_blob_intensity = 1.0; - -#if defined(SPECULAR_TOON) - specular_blob_intensity *= specular * 2.0; -#endif - +/// GI /// #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; + 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.camera_matrix) * normal; const float c1 = 0.429043; @@ -2125,12 +2240,12 @@ FRAGMENT_SHADER_CODE 2.0 * c2 * lightmap_captures.data[index].sh[1].rgb * wnormal.y + 2.0 * c2 * lightmap_captures.data[index].sh[2].rgb * wnormal.z); - } 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; + } 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 * draw_call.lightmap_uv_scale.zw + draw_call.lightmap_uv_scale.xy; - uvw.z = float((draw_call.gi_offset >> 16) & 0xFFFF); + 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 @@ -2139,7 +2254,7 @@ FRAGMENT_SHADER_CODE 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 = draw_call.gi_offset >> 20; + uint idx = instances.data[instance_index].gi_offset >> 20; vec3 n = normalize(lightmaps.data[idx].normal_xform * normal); ambient_light += lm_light_l0 * 0.282095f; @@ -2159,7 +2274,7 @@ FRAGMENT_SHADER_CODE } #elif defined(USE_FORWARD_GI) - if (bool(draw_call.flags & INSTANCE_FLAGS_USE_SDFGI)) { //has lightmap capture + if (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.camera_matrix) * vertex; @@ -2231,9 +2346,9 @@ FRAGMENT_SHADER_CODE } } - if (bool(draw_call.flags & INSTANCE_FLAGS_USE_GIPROBE)) { // process giprobes + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_GIPROBE)) { // process giprobes - uint index1 = draw_call.gi_offset & 0xFFFF; + uint index1 = instances.data[instance_index].gi_offset & 0xFFFF; vec3 ref_vec = normalize(reflect(normalize(vertex), normal)); //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); @@ -2245,7 +2360,7 @@ FRAGMENT_SHADER_CODE vec4 spec_accum = vec4(0.0); gi_probe_compute(index1, vertex, normal, ref_vec, normal_mat, roughness * roughness, ambient_light, specular_light, spec_accum, amb_accum); - uint index2 = draw_call.gi_offset >> 16; + uint index2 = instances.data[instance_index].gi_offset >> 16; if (index2 != 0xFFFF) { gi_probe_compute(index2, vertex, normal, ref_vec, normal_mat, roughness * roughness, ambient_light, specular_light, spec_accum, amb_accum); @@ -2264,19 +2379,19 @@ FRAGMENT_SHADER_CODE } #elif !defined(LOW_END_MODE) - if (bool(draw_call.flags & INSTANCE_FLAGS_USE_GI_BUFFERS)) { //use GI buffers + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_GI_BUFFERS)) { //use GI buffers - ivec2 coord; + vec2 coord; if (scene_data.gi_upscale_for_msaa) { - ivec2 base_coord = ivec2(gl_FragCoord.xy); - ivec2 closest_coord = base_coord; - float closest_ang = dot(normal, texelFetch(sampler2D(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), base_coord, 0).xyz * 2.0 - 1.0); + vec2 base_coord = screen_uv; + vec2 closest_coord = base_coord; + float closest_ang = dot(normal, textureLod(sampler2D(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), base_coord, 0.0).xyz * 2.0 - 1.0); for (int i = 0; i < 4; i++) { - const ivec2 neighbours[4] = ivec2[](ivec2(-1, 0), ivec2(1, 0), ivec2(0, -1), ivec2(0, 1)); - ivec2 neighbour_coord = base_coord + neighbours[i]; - float neighbour_ang = dot(normal, texelFetch(sampler2D(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), neighbour_coord, 0).xyz * 2.0 - 1.0); + 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; + float neighbour_ang = dot(normal, textureLod(sampler2D(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), neighbour_coord, 0.0).xyz * 2.0 - 1.0); if (neighbour_ang > closest_ang) { closest_ang = neighbour_ang; closest_coord = neighbour_coord; @@ -2286,28 +2401,69 @@ FRAGMENT_SHADER_CODE coord = closest_coord; } else { - coord = ivec2(gl_FragCoord.xy); + coord = screen_uv; } - vec4 buffer_ambient = texelFetch(sampler2D(ambient_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), coord, 0); - vec4 buffer_reflection = texelFetch(sampler2D(reflection_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), coord, 0); + 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); ambient_light = mix(ambient_light, buffer_ambient.rgb, buffer_ambient.a); specular_light = mix(specular_light, buffer_reflection.rgb, buffer_reflection.a); } #endif +#ifndef LOW_END_MODE + if (scene_data.ssao_enabled) { + 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, scene_data.ssao_light_affect), scene_data.ssao_ao_affect); + } +#endif //LOW_END_MODE + { // 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 reflection_probe_count = cluster_cell.z >> CLUSTER_COUNTER_SHIFT; - uint reflection_probe_pointer = cluster_cell.z & CLUSTER_POINTER_MASK; + uint cluster_reflection_offset = cluster_offset + scene_data.cluster_type_size * 3; - for (uint i = 0; i < reflection_probe_count; i++) { - uint ref_index = cluster_data.indices[reflection_probe_pointer + i]; - reflection_process(ref_index, vertex, normal, roughness, ambient_light, specular_light, ambient_accum, reflection_accum); + uint item_min; + uint item_max; + uint item_from; + uint item_to; + + cluster_get_item_range(cluster_reflection_offset + scene_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_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 &= ~(1 << bit); +#ifdef USE_SUBGROUPS + if (((1 << 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, normal, roughness, ambient_light, specular_light, ambient_accum, reflection_accum); + } } if (reflection_accum.a > 0.0) { @@ -2321,6 +2477,16 @@ FRAGMENT_SHADER_CODE #endif } + //finalize ambient light here + ambient_light *= albedo.rgb; + ambient_light *= ao; + + // 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 @@ -2338,24 +2504,39 @@ FRAGMENT_SHADER_CODE 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; - vec3 f0 = F0(metallic, specular, albedo); specular_light *= env.x * f0 + env.y; #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 - for (uint i = 0; i < scene_data.directional_light_count; i++) { - if (!bool(directional_lights.data[i].mask & draw_call.layer_mask)) { - continue; //not masked + // 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; } - vec3 shadow_attenuation = vec3(1.0); + 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; -#endif + float shadow = 1.0; +#ifdef USE_SOFT_SHADOWS + //version with soft shadows, more expensive if (directional_lights.data[i].shadow_enabled) { float depth_z = -vertex.z; @@ -2369,8 +2550,6 @@ FRAGMENT_SHADER_CODE normal_bias -= light_dir * dot(light_dir, normal_bias); \ m_var.xyz += normal_bias; - float shadow = 0.0; - if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { vec4 v = vec4(vertex, 1.0); @@ -2391,19 +2570,6 @@ FRAGMENT_SHADER_CODE shadow_color = directional_lights.data[i].shadow_color1.rgb; -#ifdef LIGHT_TRANSMITTANCE_USED - { - 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; - } -#endif } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { vec4 v = vec4(vertex, 1.0); @@ -2423,19 +2589,6 @@ FRAGMENT_SHADER_CODE } shadow_color = directional_lights.data[i].shadow_color2.rgb; -#ifdef LIGHT_TRANSMITTANCE_USED - { - 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; - } -#endif } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { vec4 v = vec4(vertex, 1.0); @@ -2455,19 +2608,6 @@ FRAGMENT_SHADER_CODE } shadow_color = directional_lights.data[i].shadow_color3.rgb; -#ifdef LIGHT_TRANSMITTANCE_USED - { - 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; - } -#endif } else { vec4 v = vec4(vertex, 1.0); @@ -2488,20 +2628,6 @@ FRAGMENT_SHADER_CODE } shadow_color = directional_lights.data[i].shadow_color4.rgb; - -#ifdef LIGHT_TRANSMITTANCE_USED - { - 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 } if (directional_lights.data[i].blend_splits) { @@ -2575,130 +2701,407 @@ FRAGMENT_SHADER_CODE 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 - shadow_attenuation = mix(shadow_color, vec3(1.0), shadow); +#undef BIAS_FUNC + } +#else + // Soft shadow disabled version + + if (directional_lights.data[i].shadow_enabled) { + float depth_z = -vertex.z; + + vec4 pssm_coord; + 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); +#ifdef LIGHT_TRANSMITTANCE_USED + { + 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_transmittance_z_scale.x; + float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.x; + + transmittance_z = z - shadow_z; + } +#endif + } 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); +#ifdef LIGHT_TRANSMITTANCE_USED + { + 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_transmittance_z_scale.y; + float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.y; + + transmittance_z = z - shadow_z; + } +#endif + } 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); +#ifdef LIGHT_TRANSMITTANCE_USED + { + 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_transmittance_z_scale.z; + float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.z; + + transmittance_z = z - shadow_z; + } +#endif + + } else { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 3) + + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); +#ifdef LIGHT_TRANSMITTANCE_USED + { + 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_transmittance_z_scale.w; + float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.w; + + transmittance_z = z - shadow_z; + } +#endif + } + + 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, pssm_coord); + + if (directional_lights.data[i].blend_splits) { + float pssm_blend; + + 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); + } 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); + } 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); + } 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_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, 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); + } + } + + 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_enabled) { + 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_transmittance_z_scale.x; + float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.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; - light_compute(normal, directional_lights.data[i].direction, normalize(view), directional_lights.data[i].size, directional_lights.data[i].color * directional_lights.data[i].energy, 1.0, shadow_attenuation, albedo, roughness, metallic, specular, directional_lights.data[i].specular * specular_blob_intensity, + 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_transmittance_z_scale.y; + float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.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_transmittance_z_scale.z; + float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.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_transmittance_z_scale.w; + float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.w; + + transmittance_z = z - shadow_z; + } +#endif + + float shadow = 1.0; + + if (i < 4) { + shadow = float(shadow0 >> (i * 8) & 0xFF) / 255.0; + } else { + shadow = float(shadow1 >> ((i - 4) * 8) & 0xFF) / 255.0; + } + + blur_shadow(shadow); + + light_compute(normal, directional_lights.data[i].direction, normalize(view), directional_lights.data[i].color * directional_lights.data[i].energy, shadow, f0, orms, 1.0, #ifdef LIGHT_BACKLIGHT_USED - backlight, + backlight, #endif #ifdef LIGHT_TRANSMITTANCE_USED - transmittance_color, - transmittance_depth, - transmittance_curve, - transmittance_boost, - transmittance_z, + transmittance_color, + transmittance_depth, + transmittance_curve, + transmittance_boost, + transmittance_z, #endif #ifdef LIGHT_RIM_USED - rim, rim_tint, + rim, rim_tint, albedo, #endif #ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, + clearcoat, clearcoat_gloss, #endif #ifdef LIGHT_ANISOTROPY_USED - binormal, tangent, anisotropy, + binormal, tangent, anisotropy, +#endif +#ifdef USE_SOFT_SHADOW + directional_lights.data[i].size, #endif #ifdef USE_SHADOW_TO_OPACITY - alpha, + alpha, #endif - diffuse_light, - specular_light); + diffuse_light, + specular_light); + } } - } - { //omni lights + { //omni lights - uint omni_light_count = cluster_cell.x >> CLUSTER_COUNTER_SHIFT; - uint omni_light_pointer = cluster_cell.x & CLUSTER_POINTER_MASK; + uint cluster_omni_offset = cluster_offset; - for (uint i = 0; i < omni_light_count; i++) { - uint light_index = cluster_data.indices[omni_light_pointer + i]; + uint item_min; + uint item_max; + uint item_from; + uint item_to; - if (!bool(lights.data[light_index].mask & draw_call.layer_mask)) { - continue; //not masked - } + cluster_get_item_range(cluster_omni_offset + scene_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 &= ~(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; - light_process_omni(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, albedo, roughness, metallic, specular, specular_blob_intensity, + if (!bool(omni_lights.data[light_index].mask & instances.data[instance_index].layer_mask)) { + continue; //not masked + } + + float shadow = light_process_omni_shadow(light_index, vertex, view); + + shadow = blur_shadow(shadow); + + light_process_omni(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, #ifdef LIGHT_BACKLIGHT_USED - backlight, + backlight, #endif #ifdef LIGHT_TRANSMITTANCE_USED - transmittance_color, - transmittance_depth, - transmittance_curve, - transmittance_boost, + transmittance_color, + transmittance_depth, + transmittance_curve, + transmittance_boost, #endif #ifdef LIGHT_RIM_USED - rim, - rim_tint, + rim, + rim_tint, + albedo, #endif #ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, + clearcoat, clearcoat_gloss, #endif #ifdef LIGHT_ANISOTROPY_USED - tangent, binormal, anisotropy, + tangent, binormal, anisotropy, #endif #ifdef USE_SHADOW_TO_OPACITY - alpha, + alpha, #endif - diffuse_light, specular_light); + diffuse_light, specular_light); + } + } } - } - { //spot lights - uint spot_light_count = cluster_cell.y >> CLUSTER_COUNTER_SHIFT; - uint spot_light_pointer = cluster_cell.y & CLUSTER_POINTER_MASK; + { //spot lights - for (uint i = 0; i < spot_light_count; i++) { - uint light_index = cluster_data.indices[spot_light_pointer + i]; + uint cluster_spot_offset = cluster_offset + scene_data.cluster_type_size; - if (!bool(lights.data[light_index].mask & draw_call.layer_mask)) { - continue; //not masked - } + uint item_min; + uint item_max; + uint item_from; + uint item_to; - light_process_spot(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, albedo, roughness, metallic, specular, specular_blob_intensity, + cluster_get_item_range(cluster_spot_offset + scene_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 &= ~(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(spot_lights.data[light_index].mask & instances.data[instance_index].layer_mask)) { + continue; //not masked + } + + float shadow = light_process_spot_shadow(light_index, vertex, view); + + shadow = blur_shadow(shadow); + + light_process_spot(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, #ifdef LIGHT_BACKLIGHT_USED - backlight, + backlight, #endif #ifdef LIGHT_TRANSMITTANCE_USED - transmittance_color, - transmittance_depth, - transmittance_curve, - transmittance_boost, + transmittance_color, + transmittance_depth, + transmittance_curve, + transmittance_boost, #endif #ifdef LIGHT_RIM_USED - rim, - rim_tint, + rim, + rim_tint, + albedo, #endif #ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, + clearcoat, clearcoat_gloss, #endif #ifdef LIGHT_ANISOTROPY_USED - tangent, binormal, anisotropy, + tangent, binormal, anisotropy, #endif #ifdef USE_SHADOW_TO_OPACITY - alpha, + alpha, #endif - diffuse_light, specular_light); + diffuse_light, specular_light); + } + } } - } #ifdef USE_SHADOW_TO_OPACITY - alpha = min(alpha, clamp(length(ambient_light), 0.0, 1.0)); + alpha = min(alpha, clamp(length(ambient_light), 0.0, 1.0)); #if defined(ALPHA_SCISSOR_USED) - if (alpha < alpha_scissor) { - discard; - } + if (alpha < alpha_scissor) { + discard; + } #endif // ALPHA_SCISSOR_USED #ifdef USE_OPAQUE_PREPASS - if (alpha < opaque_prepass_threshold) { - discard; - } + if (alpha < opaque_prepass_threshold) { + discard; + } #endif // USE_OPAQUE_PREPASS @@ -2710,173 +3113,149 @@ FRAGMENT_SHADER_CODE #ifdef MODE_RENDER_SDF - { - vec3 local_pos = (scene_data.sdf_to_bounds * vec4(vertex, 1.0)).xyz; - ivec3 grid_pos = scene_data.sdf_offset + ivec3(local_pos * vec3(scene_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.camera_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); + { + vec3 local_pos = (scene_data.sdf_to_bounds * vec4(vertex, 1.0)).xyz; + ivec3 grid_pos = scene_data.sdf_offset + ivec3(local_pos * vec3(scene_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.camera_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); + } } - } - imageAtomicOr(geom_facing_grid, grid_pos, facing_bits); //store facing bits + imageAtomicOr(geom_facing_grid, grid_pos, facing_bits); //store facing bits - if (length(emission) > 0.001) { - float lumas[6]; - vec3 light_total = vec3(0); + 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)); - } + 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)); + float luma_total = max(light_total.r, max(light_total.g, light_total.b)); - uint light_aniso = 0; + uint light_aniso = 0; - for (int i = 0; i < 6; i++) { - light_aniso |= min(31, uint((lumas[i] / luma_total) * 31.0)) << (i * 5); - } + 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 + //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; + 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 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 cMax = max(cRed, max(cGreen, cBlue)); - float expp = max(-B - 1.0f, floor(log(cMax) / LN2)) + 1.0f + B; + 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 sMax = floor((cMax / pow(2.0f, expp - B - N)) + 0.5f); - float exps = expp + 1.0f; + float exps = expp + 1.0f; - if (0.0 <= sMax && sMax < pow2to9) { - exps = expp; - } + 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) & 0x1FF) >> 1) | ((uint(sGreen) & 0x1FF) << 8) | (((uint(sBlue) & 0x1FF) >> 1) << 17) | ((uint(exps) & 0x1F) << 25); + 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) & 0x1FF) >> 1) | ((uint(sGreen) & 0x1FF) << 8) | (((uint(sBlue) & 0x1FF) >> 1) << 17) | ((uint(exps) & 0x1F) << 25); - imageStore(emission_grid, grid_pos, uvec4(light_rgbe)); - imageStore(emission_aniso_grid, grid_pos, uvec4(light_aniso)); + 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; + 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; + normal_output_buffer.rgb = normal * 0.5 + 0.5; + normal_output_buffer.a = 0.0; + depth_output_buffer.r = -vertex.z; -#if defined(AO_USED) - orm_output_buffer.r = ao; -#else - orm_output_buffer.r = 0.0; -#endif - orm_output_buffer.g = roughness; - orm_output_buffer.b = metallic; - orm_output_buffer.a = sss_strength; + 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; + 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); + normal_roughness_output_buffer = vec4(normal * 0.5 + 0.5, roughness); #ifdef MODE_RENDER_GIPROBE - if (bool(draw_call.flags & INSTANCE_FLAGS_USE_GIPROBE)) { // process giprobes - uint index1 = draw_call.gi_offset & 0xFFFF; - uint index2 = draw_call.gi_offset >> 16; - giprobe_buffer.x = index1 & 0xFF; - giprobe_buffer.y = index2 & 0xFF; - } else { - giprobe_buffer.x = 0xFF; - giprobe_buffer.y = 0xFF; - } + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_GIPROBE)) { // process giprobes + uint index1 = instances.data[instance_index].gi_offset & 0xFFFF; + uint index2 = instances.data[instance_index].gi_offset >> 16; + giprobe_buffer.x = index1 & 0xFF; + giprobe_buffer.y = index2 & 0xFF; + } else { + giprobe_buffer.x = 0xFF; + giprobe_buffer.y = 0xFF; + } #endif -#endif //MODE_RENDER_NORMAL +#endif //MODE_RENDER_NORMAL_ROUGHNESS //nothing happens, so a tree-ssa optimizer will result in no fragment shader :) #else - specular_light *= scene_data.reflection_multiplier; - ambient_light *= albedo; //ambient must be multiplied by albedo at the end - -//ambient occlusion -#if defined(AO_USED) - -#ifndef LOW_END_MODE - if (scene_data.ssao_enabled && scene_data.ssao_ao_affect > 0.0) { - float ssao = texture(sampler2D(ao_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), screen_uv).r; - ao = mix(ao, min(ao, ssao), scene_data.ssao_ao_affect); - ao_light_affect = mix(ao_light_affect, max(ao_light_affect, scene_data.ssao_light_affect), scene_data.ssao_ao_affect); - } -#endif //LOW_END_MODE - - ambient_light = mix(scene_data.ao_color.rgb, ambient_light, ao); - ao_light_affect = mix(1.0, ao, ao_light_affect); - specular_light = mix(scene_data.ao_color.rgb, specular_light, ao_light_affect); - diffuse_light = mix(scene_data.ao_color.rgb, diffuse_light, ao_light_affect); -#else - -#ifndef LOW_END_MODE - if (scene_data.ssao_enabled) { - float ao = texture(sampler2D(ao_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), screen_uv).r; - ambient_light = mix(scene_data.ao_color.rgb, ambient_light, ao); - float ao_light_affect = mix(1.0, ao, scene_data.ssao_light_affect); - specular_light = mix(scene_data.ao_color.rgb, specular_light, ao_light_affect); - diffuse_light = mix(scene_data.ao_color.rgb, diffuse_light, ao_light_affect); - } -#endif //LOW_END_MODE + // multiply by albedo + diffuse_light *= albedo; // ambient must be multiplied by albedo at the end -#endif // AO_USED + // apply direct light AO + ao = unpackUnorm4x8(orms).x; + specular_light *= ao; + diffuse_light *= ao; - // base color remapping - diffuse_light *= 1.0 - metallic; // TODO: avoid all diffuse and ambient light calculations when metallic == 1 up to this point + // 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 @@ -2892,25 +3271,8 @@ FRAGMENT_SHADER_CODE specular_buffer = vec4(specular_light, metallic); #endif - // 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(vertex); - diffuse_buffer.rgb = mix(diffuse_buffer.rgb, fog.rgb, fog.a); - specular_buffer.rgb = mix(specular_buffer.rgb, vec3(0.0), fog.a); - } - -#ifndef LOW_END_MODE - if (scene_data.volumetric_fog_enabled) { - vec4 fog = volumetric_fog_process(screen_uv, -vertex.z); - diffuse_buffer.rgb = mix(diffuse_buffer.rgb, fog.rgb, fog.a); - specular_buffer.rgb = mix(specular_buffer.rgb, vec3(0.0), fog.a); - } -#endif // LOW_END_MODE - -#if defined(CUSTOM_FOG_USED) - diffuse_buffer.rgb = mix(diffuse_buffer.rgb, custom_fog.rgb, custom_fog.a); - specular_buffer.rgb = mix(specular_buffer.rgb, vec3(0.0), custom_fog.a); -#endif //CUSTOM_FOG_USED + 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 @@ -2922,22 +3284,10 @@ FRAGMENT_SHADER_CODE #endif //USE_NO_SHADING // 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(vertex); - frag_color.rgb = mix(frag_color.rgb, fog.rgb, fog.a); - } -#ifndef LOW_END_MODE - if (scene_data.volumetric_fog_enabled) { - vec4 fog = volumetric_fog_process(screen_uv, -vertex.z); - frag_color.rgb = mix(frag_color.rgb, fog.rgb, fog.a); - } -#endif - -#if defined(CUSTOM_FOG_USED) - frag_color.rgb = mix(frag_color.rgb, custom_fog.rgb, custom_fog.a); -#endif //CUSTOM_FOG_USED + frag_color.rgb = mix(frag_color.rgb, fog.rgb, fog.a); + ; #endif //MODE_MULTIPLE_RENDER_TARGETS #endif //MODE_RENDER_DEPTH -} + } diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl index 87ce74ba88..d78890fa9e 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl @@ -3,6 +3,15 @@ #define MAX_GI_PROBES 8 +#if defined(GL_KHR_shader_subgroup_ballot) && defined(GL_KHR_shader_subgroup_arithmetic) + +#extension GL_KHR_shader_subgroup_ballot : enable +#extension GL_KHR_shader_subgroup_arithmetic : enable + +#define USE_SUBGROUPS + +#endif + #include "cluster_data_inc.glsl" #if !defined(MODE_RENDER_DEPTH) || defined(MODE_RENDER_MATERIAL) || defined(MODE_RENDER_SDF) || defined(MODE_RENDER_NORMAL_ROUGHNESS) || defined(MODE_RENDER_GIPROBE) || defined(TANGENT_USED) || defined(NORMAL_MAP_USED) @@ -12,12 +21,10 @@ #endif layout(push_constant, binding = 0, std430) uniform DrawCall { - mat4 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; + uint instance_index; + uint uv_offset; + uint pad0; + uint pad1; } draw_call; @@ -36,91 +43,13 @@ draw_call; #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; - #define SDFGI_MAX_CASCADES 8 -layout(set = 0, binding = 3, std140) uniform SceneData { - mat4 projection_matrix; - mat4 inv_projection_matrix; - - mat4 camera_matrix; - mat4 inv_camera_matrix; - - vec2 viewport_size; - vec2 screen_pixel_size; - - //use vec4s because std140 doesnt play nice with vec2s, z and w are wasted - vec4 directional_penumbra_shadow_kernel[32]; - vec4 directional_soft_shadow_kernel[32]; - vec4 penumbra_shadow_kernel[32]; - vec4 soft_shadow_kernel[32]; - - uint directional_penumbra_shadow_samples; - uint directional_soft_shadow_samples; - uint penumbra_shadow_samples; - uint soft_shadow_samples; - - vec4 ambient_light_color_energy; - - float ambient_color_sky_mix; - bool use_ambient_light; - bool use_ambient_cubemap; - bool use_reflection_cubemap; - - mat3 radiance_inverse_xform; - - vec2 shadow_atlas_pixel_size; - vec2 directional_shadow_pixel_size; - - uint directional_light_count; - float dual_paraboloid_side; - float z_far; - float z_near; - - bool ssao_enabled; - float ssao_light_affect; - float ssao_ao_affect; - bool roughness_limiter_enabled; - - float roughness_limiter_amount; - float roughness_limiter_limit; - uvec2 roughness_limiter_pad; - - vec4 ao_color; - - mat4 sdf_to_bounds; - - ivec3 sdf_offset; - bool material_uv2_mode; - - 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; - - bool fog_enabled; - float fog_density; - float fog_height; - float fog_height_density; - - vec3 fog_light_color; - float fog_sun_scatter; - - float fog_aerial_perspective; - - float time; - float reflection_multiplier; // one normally, zero when rendering reflections +/* Set 1: Base Pass (never changes) */ - bool pancake_shadows; -} +layout(set = 0, binding = 1) uniform sampler material_samplers[12]; -scene_data; +layout(set = 0, binding = 2) uniform sampler shadow_sampler; #define INSTANCE_FLAGS_USE_GI_BUFFERS (1 << 6) #define INSTANCE_FLAGS_USE_SDFGI (1 << 7) @@ -139,17 +68,22 @@ scene_data; #define INSTANCE_FLAGS_SKELETON (1 << 19) #define INSTANCE_FLAGS_NON_UNIFORM_SCALE (1 << 20) -layout(set = 0, binding = 5, std430) restrict readonly buffer Lights { +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[]; } -lights; +spot_lights; -layout(set = 0, binding = 6) buffer restrict readonly ReflectionProbeData { +layout(set = 0, binding = 5) buffer restrict readonly ReflectionProbeData { ReflectionData data[]; } reflections; -layout(set = 0, binding = 7, std140) uniform DirectionalLights { +layout(set = 0, binding = 6, std140) uniform DirectionalLights { DirectionalLightData data[MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS]; } directional_lights; @@ -161,7 +95,7 @@ struct Lightmap { mat3 normal_xform; }; -layout(set = 0, binding = 10, std140) restrict readonly buffer Lightmaps { +layout(set = 0, binding = 7, std140) restrict readonly buffer Lightmaps { Lightmap data[]; } lightmaps; @@ -170,29 +104,20 @@ struct LightmapCapture { vec4 sh[9]; }; -layout(set = 0, binding = 11, std140) restrict readonly buffer LightmapCaptures { +layout(set = 0, binding = 8, std140) restrict readonly buffer LightmapCaptures { LightmapCapture data[]; } lightmap_captures; -layout(set = 0, binding = 12) uniform texture2D decal_atlas; -layout(set = 0, binding = 13) uniform texture2D decal_atlas_srgb; +layout(set = 0, binding = 9) uniform texture2D decal_atlas; +layout(set = 0, binding = 10) uniform texture2D decal_atlas_srgb; -layout(set = 0, binding = 14, std430) restrict readonly buffer Decals { +layout(set = 0, binding = 11, std430) restrict readonly buffer Decals { DecalData data[]; } decals; -layout(set = 0, binding = 15) uniform utexture3D cluster_texture; - -layout(set = 0, binding = 16, std430) restrict readonly buffer ClusterData { - uint indices[]; -} -cluster_data; - -layout(set = 0, binding = 17) uniform texture2D directional_shadow_atlas; - -layout(set = 0, binding = 18, std430) restrict readonly buffer GlobalVariableData { +layout(set = 0, binding = 12, std430) restrict readonly buffer GlobalVariableData { vec4 data[]; } global_variables; @@ -206,7 +131,7 @@ struct SDFGIProbeCascadeData { float to_cell; // 1/bounds * grid_size }; -layout(set = 0, binding = 19, std140) uniform SDFGI { +layout(set = 0, binding = 13, std140) uniform SDFGI { vec3 grid_size; uint max_cascades; @@ -236,40 +161,140 @@ sdfgi; #endif //LOW_END_MODE -// decal atlas +/* Set 2: Render Pass (changes per render pass) */ + +layout(set = 1, binding = 0, std140) uniform SceneData { + mat4 projection_matrix; + mat4 inv_projection_matrix; + + mat4 camera_matrix; + mat4 inv_camera_matrix; + + vec2 viewport_size; + vec2 screen_pixel_size; + + uint cluster_shift; + uint cluster_width; + uint cluster_type_size; + uint max_cluster_element_count_div_32; + + //use vec4s because std140 doesnt play nice with vec2s, z and w are wasted + vec4 directional_penumbra_shadow_kernel[32]; + vec4 directional_soft_shadow_kernel[32]; + vec4 penumbra_shadow_kernel[32]; + vec4 soft_shadow_kernel[32]; + + uint directional_penumbra_shadow_samples; + uint directional_soft_shadow_samples; + uint penumbra_shadow_samples; + uint soft_shadow_samples; + + vec4 ambient_light_color_energy; + + float ambient_color_sky_mix; + bool use_ambient_light; + bool use_ambient_cubemap; + bool use_reflection_cubemap; + + mat3 radiance_inverse_xform; + + vec2 shadow_atlas_pixel_size; + vec2 directional_shadow_pixel_size; + + uint directional_light_count; + float dual_paraboloid_side; + float z_far; + float z_near; -/* Set 1, Radiance */ + bool ssao_enabled; + float ssao_light_affect; + float ssao_ao_affect; + bool roughness_limiter_enabled; + + float roughness_limiter_amount; + float roughness_limiter_limit; + uvec2 roughness_limiter_pad; + + vec4 ao_color; + + mat4 sdf_to_bounds; + + ivec3 sdf_offset; + bool material_uv2_mode; + + 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; + + bool fog_enabled; + float fog_density; + float fog_height; + float fog_height_density; + + vec3 fog_light_color; + float fog_sun_scatter; + + float fog_aerial_perspective; + + float time; + float reflection_multiplier; // one normally, zero when rendering reflections + + bool pancake_shadows; +} + +scene_data; + +struct InstanceData { + mat4 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 = 1, std430) buffer restrict readonly InstanceDataBuffer { + InstanceData data[]; +} +instances; #ifdef USE_RADIANCE_CUBEMAP_ARRAY -layout(set = 1, binding = 0) uniform textureCubeArray radiance_cubemap; +layout(set = 1, binding = 2) uniform textureCubeArray radiance_cubemap; #else -layout(set = 1, binding = 0) uniform textureCube radiance_cubemap; +layout(set = 1, binding = 2) uniform textureCube radiance_cubemap; #endif -/* Set 2, Reflection and Shadow Atlases (view dependent) */ +layout(set = 1, binding = 3) uniform textureCubeArray reflection_atlas; -layout(set = 1, binding = 1) uniform textureCubeArray reflection_atlas; +layout(set = 1, binding = 4) uniform texture2D shadow_atlas; -layout(set = 1, binding = 2) uniform texture2D shadow_atlas; +layout(set = 1, binding = 5) uniform texture2D directional_shadow_atlas; -layout(set = 1, binding = 3) uniform texture2DArray lightmap_textures[MAX_LIGHTMAP_TEXTURES]; +layout(set = 1, binding = 6) uniform texture2DArray lightmap_textures[MAX_LIGHTMAP_TEXTURES]; -#ifndef LOW_END_MODE -layout(set = 1, binding = 4) uniform texture3D gi_probe_textures[MAX_GI_PROBES]; +#ifndef LOW_END_MOD +layout(set = 1, binding = 7) uniform texture3D gi_probe_textures[MAX_GI_PROBES]; #endif -/* Set 3, Render Buffers */ +layout(set = 1, binding = 8, std430) buffer restrict readonly ClusterBuffer { + uint data[]; +} +cluster_buffer; #ifdef MODE_RENDER_SDF -layout(r16ui, set = 1, binding = 5) uniform restrict writeonly uimage3D albedo_volume_grid; -layout(r32ui, set = 1, binding = 6) uniform restrict writeonly uimage3D emission_grid; -layout(r32ui, set = 1, binding = 7) uniform restrict writeonly uimage3D emission_aniso_grid; -layout(r32ui, set = 1, binding = 8) uniform restrict uimage3D geom_facing_grid; +layout(r16ui, set = 1, binding = 9) uniform restrict writeonly uimage3D albedo_volume_grid; +layout(r32ui, set = 1, binding = 10) uniform restrict writeonly uimage3D emission_grid; +layout(r32ui, set = 1, binding = 11) uniform restrict writeonly uimage3D emission_aniso_grid; +layout(r32ui, set = 1, binding = 12) 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 @@ -278,17 +303,17 @@ layout(r32ui, set = 1, binding = 8) uniform restrict uimage3D geom_facing_grid; #else -layout(set = 1, binding = 5) uniform texture2D depth_buffer; -layout(set = 1, binding = 6) uniform texture2D color_buffer; +layout(set = 1, binding = 9) uniform texture2D depth_buffer; +layout(set = 1, binding = 10) uniform texture2D color_buffer; #ifndef LOW_END_MODE -layout(set = 1, binding = 7) uniform texture2D normal_roughness_buffer; -layout(set = 1, binding = 8) uniform texture2D ao_buffer; -layout(set = 1, binding = 9) uniform texture2D ambient_buffer; -layout(set = 1, binding = 10) uniform texture2D reflection_buffer; -layout(set = 1, binding = 11) uniform texture2DArray sdfgi_lightprobe_texture; -layout(set = 1, binding = 12) uniform texture3D sdfgi_occlusion_cascades; +layout(set = 1, binding = 11) uniform texture2D normal_roughness_buffer; +layout(set = 1, binding = 12) uniform texture2D ao_buffer; +layout(set = 1, binding = 13) uniform texture2D ambient_buffer; +layout(set = 1, binding = 14) uniform texture2D reflection_buffer; +layout(set = 1, binding = 15) uniform texture2DArray sdfgi_lightprobe_texture; +layout(set = 1, binding = 16) uniform texture3D sdfgi_occlusion_cascades; struct GIProbeData { mat4 xform; @@ -306,22 +331,22 @@ struct GIProbeData { uint mipmaps; }; -layout(set = 1, binding = 13, std140) uniform GIProbes { +layout(set = 1, binding = 17, std140) uniform GIProbes { GIProbeData data[MAX_GI_PROBES]; } gi_probes; -layout(set = 1, binding = 14) uniform texture3D volumetric_fog_texture; +layout(set = 1, binding = 18) uniform texture3D volumetric_fog_texture; #endif // LOW_END_MODE #endif -/* Set 4 Skeleton & Instancing (Multimesh) */ +/* Set 2 Skeleton & Instancing (can change per item) */ layout(set = 2, binding = 0, std430) restrict readonly buffer Transforms { vec4 data[]; } transforms; -/* Set 5 User Material */ +/* Set 3 User Material */ diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl index 813ea29fa1..e4c3f3a84b 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl @@ -97,6 +97,8 @@ void main() { float blend = 0.0; #if 1 + // No interpolation + vec3 inv_dir = 1.0 / ray_dir; float rough = 0.5; @@ -161,114 +163,11 @@ void main() { 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))); - if (blend > 0.0) { - light = mix(light, hit_light, blend); - blend = 0.0; - } else { - light = hit_light; - - //process blend - float blend_from = (float(params.probe_axis_size - 1) / 2.0) - 2.5; - float blend_to = blend_from + 2.0; - - vec3 cam_pos = params.cam_transform[3].xyz - cascades.data[i].offset; - cam_pos *= cascades.data[i].to_cell; - - pos += ray_dir * min(advance, max_advance); - vec3 inner_pos = pos - cam_pos; - - inner_pos = inner_pos * float(params.probe_axis_size - 1) / params.grid_size.x; - - 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); - - pos /= cascades.data[i].to_cell; - pos += cascades.data[i].offset; - ray_pos = pos; - hit = false; //continue trace for blend - - continue; - } - } + light = hit_light; break; } - light = mix(light, vec3(0.0), blend); - -#else - - vec3 inv_dir = 1.0 / ray_dir; - - bool hit = false; - vec4 light_accum = vec4(0.0); - - float blend_size = (params.grid_size.x / float(params.probe_axis_size - 1)) * 0.5; - - float radius_sizes[MAX_CASCADES]; - for (uint i = 0; i < params.max_cascades; i++) { - radius_sizes[i] = (1.0 / cascades.data[i].to_cell) * (params.grid_size.x * 0.5 - blend_size); - } - - float max_distance = radius_sizes[params.max_cascades - 1]; - float advance = 0; - while (advance < max_distance) { - for (uint i = 0; i < params.max_cascades; i++) { - if (advance < radius_sizes[i]) { - vec3 pos = (ray_pos + ray_dir * advance) - cascades.data[i].offset; - pos *= cascades.data[i].to_cell * pos_to_uvw; - - float distance = texture(sampler3D(sdf_cascades[i], linear_sampler), pos).r * 255.0 - 1.0; - - vec4 hit_light = vec4(0.0); - if (distance < 1.0) { - hit_light.a = max(0.0, 1.0 - distance); - hit_light.rgb = texture(sampler3D(light_cascades[i], linear_sampler), pos).rgb; - hit_light.rgb *= hit_light.a; - } - - distance /= cascades.data[i].to_cell; - - if (i < (params.max_cascades - 1)) { - pos = (ray_pos + ray_dir * advance) - cascades.data[i + 1].offset; - pos *= cascades.data[i + 1].to_cell * pos_to_uvw; - - float distance2 = texture(sampler3D(sdf_cascades[i + 1], linear_sampler), pos).r * 255.0 - 1.0; - - vec4 hit_light2 = vec4(0.0); - if (distance2 < 1.0) { - hit_light2.a = max(0.0, 1.0 - distance2); - hit_light2.rgb = texture(sampler3D(light_cascades[i + 1], linear_sampler), pos).rgb; - hit_light2.rgb *= hit_light2.a; - } - - float prev_radius = i == 0 ? 0.0 : radius_sizes[i - 1]; - float blend = (advance - prev_radius) / (radius_sizes[i] - prev_radius); - - distance2 /= cascades.data[i + 1].to_cell; - - hit_light = mix(hit_light, hit_light2, blend); - distance = mix(distance, distance2, blend); - } - - light_accum += hit_light; - advance += distance; - break; - } - } - - if (light_accum.a > 0.98) { - break; - } - } - - light = light_accum.rgb / light_accum.a; - #endif imageStore(screen_buffer, screen_pos, vec4(linear_to_srgb(light), 1.0)); diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl index 30dbf5871f..dc7238abed 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl @@ -67,8 +67,8 @@ struct Light { float attenuation; uint type; - float spot_angle; - float spot_attenuation; + float cos_spot_angle; + float inv_spot_attenuation; float radius; vec4 shadow_color; @@ -80,6 +80,7 @@ layout(set = 0, binding = 9, std140) buffer restrict readonly Lights { lights; layout(set = 0, binding = 10) uniform texture2DArray lightprobe_texture; +layout(set = 0, binding = 11) uniform texture3D occlusion_texture; layout(push_constant, binding = 0, std430) uniform Params { vec3 grid_size; @@ -91,9 +92,9 @@ layout(push_constant, binding = 0, std430) uniform Params { uint process_increment; int probe_axis_size; - bool multibounce; + float bounce_feedback; float y_mult; - uint pad; + bool use_occlusion; } params; @@ -125,7 +126,10 @@ void main() { uint voxel_index = uint(gl_GlobalInvocationID.x); //used for skipping voxels every N frames - voxel_index = params.process_offset + voxel_index * params.process_increment; + if (params.process_increment > 1) { + voxel_index *= params.process_increment; + voxel_index += params.process_offset; + } if (voxel_index >= dispatch_data.total_count) { return; @@ -143,10 +147,96 @@ void main() { 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 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; @@ -162,18 +252,10 @@ void main() { 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; + light_accum[i] += l * strength; } } - 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)); - // Raytrace light vec3 pos_to_uvw = 1.0 / params.grid_size; @@ -203,13 +285,16 @@ void main() { rel_vec.y /= params.y_mult; attenuation = get_omni_attenuation(light_distance, 1.0 / lights.data[i].radius, lights.data[i].attenuation); - float angle = acos(dot(normalize(rel_vec), -lights.data[i].direction)); - if (angle > lights.data[i].spot_angle) { - attenuation = 0.0; - } else { - float d = clamp(angle / lights.data[i].spot_angle, 0, 1); - attenuation *= pow(1.0 - d, lights.data[i].spot_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; } @@ -292,65 +377,6 @@ void main() { } } - // Add indirect light - - if (params.multibounce) { - vec3 pos = (vec3(positioni) + vec3(0.5)) * float(params.probe_axis_size - 1) / params.grid_size; - ivec3 probe_base_pos = ivec3(pos); - - vec4 probe_accum[6] = vec4[](vec4(0.0), vec4(0.0), vec4(0.0), vec4(0.0), vec4(0.0), vec4(0.0)); - 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 = (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(0.005, dot(n, probe_dir)); - - 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; - vec4 indirect_light = textureLod(sampler2DArray(lightprobe_texture, linear_sampler), pos_uvw, 0.0); - - probe_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] += probe_accum[k].rgb * albedo / weight_accum[k]; - } - } - } - // Store the light in the light texture float lumas[6]; diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl index d516ab22c3..007e4c113a 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl @@ -39,8 +39,11 @@ layout(rgba32i, set = 0, binding = 13) uniform restrict iimage2D lightprobe_aver 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 @@ -136,12 +139,24 @@ uint rgbe_encode(vec3 color) { 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; @@ -154,27 +169,9 @@ void main() { vec3 probe_pos = cascades.data[params.cascade].offset + vec3(probe_cell) * probe_cell_size; vec3 pos_to_uvw = 1.0 / params.grid_size; - vec4 probe_sh_accum[SH_SIZE] = vec4[]( - vec4(0.0), - vec4(0.0), - vec4(0.0), - vec4(0.0), - vec4(0.0), - vec4(0.0), - vec4(0.0), - vec4(0.0), - vec4(0.0) -#if (SH_SIZE == 16) - , - vec4(0.0), - vec4(0.0), - vec4(0.0), - vec4(0.0), - vec4(0.0), - vec4(0.0), - vec4(0.0) -#endif - ); + 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)); @@ -195,14 +192,12 @@ void main() { vec3 inv_dir = 1.0 / ray_dir; bool hit = false; - vec3 hit_normal; - vec3 hit_light; - vec3 hit_aniso0; - vec3 hit_aniso1; + 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 @@ -221,14 +216,12 @@ void main() { float advance = 0.0; - vec3 uvw; - 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 < 0.001) { + if (distance < 0.05) { //consider hit hit = true; break; @@ -238,17 +231,7 @@ void main() { } if (hit) { - const float EPSILON = 0.001; - hit_normal = normalize(vec3( - texture(sampler3D(sdf_cascades[j], linear_sampler), uvw + vec3(EPSILON, 0.0, 0.0)).r - texture(sampler3D(sdf_cascades[j], linear_sampler), uvw - vec3(EPSILON, 0.0, 0.0)).r, - texture(sampler3D(sdf_cascades[j], linear_sampler), uvw + vec3(0.0, EPSILON, 0.0)).r - texture(sampler3D(sdf_cascades[j], linear_sampler), uvw - vec3(0.0, EPSILON, 0.0)).r, - texture(sampler3D(sdf_cascades[j], linear_sampler), uvw + vec3(0.0, 0.0, EPSILON)).r - texture(sampler3D(sdf_cascades[j], linear_sampler), uvw - vec3(0.0, 0.0, EPSILON)).r)); - - hit_light = texture(sampler3D(light_cascades[j], linear_sampler), uvw).rgb; - vec4 aniso0 = texture(sampler3D(aniso0_cascades[j], linear_sampler), uvw); - hit_aniso0 = aniso0.rgb; - hit_aniso1 = vec3(aniso0.a, texture(sampler3D(aniso1_cascades[j], linear_sampler), uvw).rg); - + hit_cascade = j; break; } @@ -261,11 +244,32 @@ void main() { vec4 light; if (hit) { - //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; + //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 dont 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 dont usually throw a lot of rays, so this compensates +#endif light.rgb *= params.sky_energy; light.a = 0.0; @@ -278,33 +282,33 @@ void main() { } vec3 ray_dir2 = ray_dir * ray_dir; - float c[SH_SIZE] = float[]( - - 0.282095, //l0 - 0.488603 * ray_dir.y, //l1n1 - 0.488603 * ray_dir.z, //l1n0 - 0.488603 * ray_dir.x, //l1p1 - 1.092548 * ray_dir.x * ray_dir.y, //l2n2 - 1.092548 * ray_dir.y * ray_dir.z, //l2n1 - 0.315392 * (3.0 * ray_dir2.z - 1.0), //l20 - 1.092548 * ray_dir.x * ray_dir.z, //l2p1 - 0.546274 * (ray_dir2.x - ray_dir2.y) //l2p2 + +#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) - , - 0.590043 * ray_dir.y * (3.0f * ray_dir2.x - ray_dir2.y), - 2.890611 * ray_dir.y * ray_dir.x * ray_dir.z, - 0.646360 * ray_dir.y * (-1.0f + 5.0f * ray_dir2.z), - 0.373176 * (5.0f * ray_dir2.z * ray_dir.z - 3.0f * ray_dir.z), - 0.457045 * ray_dir.x * (-1.0f + 5.0f * ray_dir2.z), - 1.445305 * (ray_dir2.x - ray_dir2.y) * ray_dir.z, - 0.590043 * ray_dir.x * (ray_dir2.x - 3.0f * ray_dir2.y) + 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 j = 0; j < SH_SIZE; j++) { - probe_sh_accum[j] += light * c[j]; - } } for (uint i = 0; i < SH_SIZE; i++) { @@ -312,7 +316,7 @@ void main() { ivec3 prev_pos = ivec3(pos.x, pos.y * SH_SIZE + i, int(params.history_index)); ivec2 average_pos = prev_pos.xy; - vec4 value = probe_sh_accum[i] * 4.0 / float(params.ray_count); + 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 @@ -344,37 +348,11 @@ void main() { ivec2 oct_pos = (pos / OCT_SIZE) * (OCT_SIZE + 2) + ivec2(1); ivec2 local_pos = pos % OCT_SIZE; - //fill the spherical harmonic - vec4 sh[SH_SIZE]; - - 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); - - sh[i] = (vec4(average) / float(params.history_size)) / float(1 << HISTORY_BITS); - } - //compute the octahedral normal for this texel vec3 normal = octahedron_encode(vec2(local_pos) / float(OCT_SIZE)); - /* + // read the spherical harmonic - 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; - vec4 light = (c1 * sh[8] * (normal.x * normal.x - normal.y * normal.y) + - c3 * sh[6] * normal.z * normal.z + - c4 * sh[0] - - c5 * sh[6] + - 2.0 * c1 * sh[4] * normal.x * normal.y + - 2.0 * c1 * sh[7] * normal.x * normal.z + - 2.0 * c1 * sh[5] * normal.y * normal.z + - 2.0 * c2 * sh[3] * normal.x + - 2.0 * c2 * sh[1] * normal.y + - 2.0 * c2 * sh[2] * normal.z); -*/ + vec3 normal2 = normal * normal; float c[SH_SIZE] = float[]( @@ -426,7 +404,14 @@ void main() { vec3 radiance = vec3(0.0); for (uint i = 0; i < SH_SIZE; i++) { - vec3 m = sh[i].rgb * c[i] * 4.0; + // 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; } @@ -515,13 +500,15 @@ void main() { //can't scroll, must look for position in parent cascade //to global coords - float probe_cell_size = float(params.grid_size.x / float(params.probe_axis_size - 1)) / cascades.data[params.cascade].to_cell; + 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 *= cascades.data[params.cascade + 1].to_cell; - probe_pos = probe_pos * float(params.probe_axis_size - 1) / float(params.grid_size.x); + 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 @@ -574,20 +561,28 @@ void main() { } } else { - // clear and let it re-raytrace, only for the last cascade, which happens very un-often - //scroll + //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)); - imageStore(lightprobe_history_scroll_texture, dst_pos, ivec4(0)); + 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 dst_pos = ivec2(pos.x, pos.y * SH_SIZE + i); - imageStore(lightprobe_average_scroll_texture, dst_pos, ivec4(0)); + ivec2 spos = ivec2(pos.x, pos.y * SH_SIZE + i); + ivec4 average = imageLoad(lightprobe_average_texture, spos); + imageStore(lightprobe_average_scroll_texture, spos, average); } } diff --git a/servers/rendering/renderer_rd/shaders/shadow_reduce.glsl b/servers/rendering/renderer_rd/shaders/shadow_reduce.glsl deleted file mode 100644 index 29443ae7db..0000000000 --- a/servers/rendering/renderer_rd/shaders/shadow_reduce.glsl +++ /dev/null @@ -1,105 +0,0 @@ -#[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_REDUCE - -shared float tmp_data[BLOCK_SIZE * BLOCK_SIZE]; -const uint swizzle_table[BLOCK_SIZE] = uint[](0, 4, 2, 6, 1, 5, 3, 7); -const uint unswizzle_table[BLOCK_SIZE] = uint[](0, 0, 0, 1, 0, 2, 1, 3); - -#endif - -layout(r32f, set = 0, binding = 0) uniform restrict readonly image2D source_depth; -layout(r32f, set = 0, binding = 1) uniform restrict writeonly image2D dst_depth; - -layout(push_constant, binding = 1, std430) uniform Params { - ivec2 source_size; - ivec2 source_offset; - uint min_size; - uint gaussian_kernel_version; - ivec2 filter_dir; -} -params; - -void main() { -#ifdef MODE_REDUCE - - uvec2 pos = gl_LocalInvocationID.xy; - - ivec2 image_offset = params.source_offset; - ivec2 image_pos = image_offset + ivec2(gl_GlobalInvocationID.xy); - uint dst_t = swizzle_table[pos.y] * BLOCK_SIZE + swizzle_table[pos.x]; - tmp_data[dst_t] = imageLoad(source_depth, min(image_pos, params.source_size - ivec2(1))).r; - ivec2 image_size = params.source_size; - - uint t = pos.y * BLOCK_SIZE + pos.x; - - //neighbours - uint size = BLOCK_SIZE; - - do { - groupMemoryBarrier(); - barrier(); - - size >>= 1; - image_size >>= 1; - image_offset >>= 1; - - if (all(lessThan(pos, uvec2(size)))) { - uint nx = t + size; - uint ny = t + (BLOCK_SIZE * size); - uint nxy = ny + size; - - tmp_data[t] += tmp_data[nx]; - tmp_data[t] += tmp_data[ny]; - tmp_data[t] += tmp_data[nxy]; - tmp_data[t] /= 4.0; - } - - } while (size > params.min_size); - - if (all(lessThan(pos, uvec2(size)))) { - image_pos = ivec2(unswizzle_table[size + pos.x], unswizzle_table[size + pos.y]); - image_pos += image_offset + ivec2(gl_WorkGroupID.xy) * int(size); - - image_size = max(ivec2(1), image_size); //in case image size became 0 - - if (all(lessThan(image_pos, uvec2(image_size)))) { - imageStore(dst_depth, image_pos, vec4(tmp_data[t])); - } - } -#endif - -#ifdef MODE_FILTER - - ivec2 image_pos = params.source_offset + ivec2(gl_GlobalInvocationID.xy); - if (any(greaterThanEqual(image_pos, params.source_size))) { - return; - } - - ivec2 clamp_min = ivec2(params.source_offset); - ivec2 clamp_max = ivec2(params.source_size) - 1; - - //gaussian kernel, size 9, sigma 4 - const int kernel_size = 9; - const float gaussian_kernel[kernel_size * 3] = float[]( - 0.000229, 0.005977, 0.060598, 0.241732, 0.382928, 0.241732, 0.060598, 0.005977, 0.000229, - 0.028532, 0.067234, 0.124009, 0.179044, 0.20236, 0.179044, 0.124009, 0.067234, 0.028532, - 0.081812, 0.101701, 0.118804, 0.130417, 0.134535, 0.130417, 0.118804, 0.101701, 0.081812); - float accum = 0.0; - for (int i = 0; i < kernel_size; i++) { - ivec2 ofs = clamp(image_pos + params.filter_dir * (i - kernel_size / 2), clamp_min, clamp_max); - accum += imageLoad(source_depth, ofs).r * gaussian_kernel[params.gaussian_kernel_version + i]; - } - - imageStore(dst_depth, image_pos, vec4(accum)); - -#endif -} diff --git a/servers/rendering/renderer_rd/shaders/skeleton.glsl b/servers/rendering/renderer_rd/shaders/skeleton.glsl index b19f5a9ad3..680d1045cd 100644 --- a/servers/rendering/renderer_rd/shaders/skeleton.glsl +++ b/servers/rendering/renderer_rd/shaders/skeleton.glsl @@ -100,7 +100,7 @@ void main() { for (uint i = 0; i < params.blend_shape_count; i++) { float w = blend_shape_weights.data[i]; - if (w > 0.0001) { + 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; diff --git a/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl b/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl index 498a6ddb5b..e7ba8feb80 100644 --- a/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl +++ b/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl @@ -4,6 +4,15 @@ VERSION_DEFINES +/* Do not use subgroups here, seems there is not much advantage and causes glitches +#extension GL_KHR_shader_subgroup_ballot: enable +#extension GL_KHR_shader_subgroup_arithmetic: enable + +#if defined(GL_KHR_shader_subgroup_ballot) && defined(GL_KHR_shader_subgroup_arithmetic) +#define USE_SUBGROUPS +#endif +*/ + #if defined(MODE_FOG) || defined(MODE_FILTER) layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; @@ -23,22 +32,25 @@ layout(local_size_x = 4, local_size_y = 4, local_size_z = 4) in; 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 Lights { +layout(set = 0, binding = 3, std430) restrict readonly buffer OmniLights { LightData data[]; } -lights; +omni_lights; -layout(set = 0, binding = 4, std140) uniform DirectionalLights { +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 = 5) uniform utexture3D cluster_texture; - -layout(set = 0, binding = 6, std430) restrict readonly buffer ClusterData { - uint indices[]; +layout(set = 0, binding = 6, std430) buffer restrict readonly ClusterBuffer { + uint data[]; } -cluster_data; +cluster_buffer; layout(set = 0, binding = 7) uniform sampler linear_sampler; @@ -132,7 +144,7 @@ layout(set = 1, binding = 2) uniform texture3D sdfgi_occlusion_texture; #endif //SDFGI -layout(push_constant, binding = 0, std430) uniform Params { +layout(set = 0, binding = 14, std140) uniform Params { vec2 fog_frustum_size_begin; vec2 fog_frustum_size_end; @@ -150,12 +162,24 @@ layout(push_constant, binding = 0, std430) uniform Params { float detail_spread; float gi_inject; uint max_gi_probes; - uint pad; + 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; } params; +layout(set = 0, binding = 15) uniform texture3D prev_density_texture; + 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); @@ -178,6 +202,42 @@ float get_omni_attenuation(float distance, float inv_range, float decay) { return nd * pow(max(distance, 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); +} + +#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(params.fog_volume_size); @@ -193,6 +253,12 @@ void main() { //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; @@ -200,6 +266,47 @@ void main() { 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 = params.light_color; float total_density = params.base_density; @@ -266,108 +373,160 @@ void main() { //compute lights from cluster - vec3 cluster_pos; - cluster_pos.xy = fog_unit_pos.xy; - cluster_pos.z = clamp((abs(view_pos.z) - params.z_near) / (params.z_far - params.z_near), 0.0, 1.0); + { //omni lights - uvec4 cluster_cell = texture(usampler3D(cluster_texture, linear_sampler), cluster_pos); + uint cluster_omni_offset = cluster_offset; - uint omni_light_count = cluster_cell.x >> CLUSTER_COUNTER_SHIFT; - uint omni_light_pointer = cluster_cell.x & CLUSTER_POINTER_MASK; + uint item_min; + uint item_max; + uint item_from; + uint item_to; - for (uint i = 0; i < omni_light_count; i++) { - uint light_index = cluster_data.indices[omni_light_pointer + i]; + cluster_get_item_range(cluster_omni_offset + params.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); - vec3 light_pos = lights.data[i].position; - float d = distance(lights.data[i].position, view_pos); - vec3 shadow_attenuation = vec3(1.0); +#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 (d * lights.data[i].inv_radius < 1.0) { - vec2 attenuation_energy = unpackHalf2x16(lights.data[i].attenuation_energy); - vec4 color_specular = unpackUnorm4x8(lights.data[i].color_specular); + //if (!bool(omni_omni_lights.data[light_index].mask & draw_call.layer_mask)) { + // continue; //not masked + //} - float attenuation = get_omni_attenuation(d, lights.data[i].inv_radius, attenuation_energy.x); + 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; - vec3 light = attenuation_energy.y * color_specular.rgb / M_PI; + if (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); - vec4 shadow_color_enabled = unpackUnorm4x8(lights.data[i].shadow_color_enabled); + vec3 light = omni_lights.data[light_index].color / M_PI; - if (shadow_color_enabled.a > 0.5) { - //has shadow - vec4 v = vec4(view_pos, 1.0); + if (omni_lights.data[light_index].shadow_enabled) { + //has shadow + vec4 v = vec4(view_pos, 1.0); - vec4 splane = (lights.data[i].shadow_matrix * v); - float shadow_len = length(splane.xyz); //need to remember shadow len from here + vec4 splane = (omni_lights.data[light_index].shadow_matrix * v); + float shadow_len = length(splane.xyz); //need to remember shadow len from here - splane.xyz = normalize(splane.xyz); - vec4 clamp_rect = lights.data[i].atlas_rect; + splane.xyz = normalize(splane.xyz); + vec4 clamp_rect = omni_lights.data[light_index].atlas_rect; - if (splane.z >= 0.0) { - splane.z += 1.0; + if (splane.z >= 0.0) { + splane.z += 1.0; - clamp_rect.y += clamp_rect.w; + clamp_rect.y += clamp_rect.w; - } else { - splane.z = 1.0 - splane.z; - } + } else { + splane.z = 1.0 - splane.z; + } - splane.xy /= splane.z; + splane.xy /= splane.z; - splane.xy = splane.xy * 0.5 + 0.5; - splane.z = shadow_len * lights.data[i].inv_radius; - splane.xy = clamp_rect.xy + splane.xy * clamp_rect.zw; - splane.w = 1.0; //needed? i think it should be 1 already + splane.xy = splane.xy * 0.5 + 0.5; + splane.z = shadow_len * omni_lights.data[light_index].inv_radius; + splane.xy = clamp_rect.xy + splane.xy * clamp_rect.zw; + splane.w = 1.0; //needed? i think it should be 1 already - float depth = texture(sampler2D(shadow_atlas, linear_sampler), splane.xy).r; - float shadow = exp(min(0.0, (depth - splane.z)) / lights.data[i].inv_radius * lights.data[i].shadow_volumetric_fog_fade); + float depth = texture(sampler2D(shadow_atlas, linear_sampler), splane.xy).r; - shadow_attenuation = mix(shadow_color_enabled.rgb, vec3(1.0), shadow); + shadow_attenuation = exp(min(0.0, (depth - splane.z)) / omni_lights.data[light_index].inv_radius * omni_lights.data[light_index].shadow_volumetric_fog_fade); + } + total_light += light * attenuation * shadow_attenuation; + } } - total_light += light * attenuation * shadow_attenuation; } } - uint spot_light_count = cluster_cell.y >> CLUSTER_COUNTER_SHIFT; - uint spot_light_pointer = cluster_cell.y & CLUSTER_POINTER_MASK; + { //spot lights - for (uint i = 0; i < spot_light_count; i++) { - uint light_index = cluster_data.indices[spot_light_pointer + i]; + uint cluster_spot_offset = cluster_offset + params.cluster_type_size; - vec3 light_pos = lights.data[i].position; - vec3 light_rel_vec = lights.data[i].position - view_pos; - float d = length(light_rel_vec); - vec3 shadow_attenuation = vec3(1.0); + 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); - if (d * lights.data[i].inv_radius < 1.0) { - vec2 attenuation_energy = unpackHalf2x16(lights.data[i].attenuation_energy); - vec4 color_specular = unpackUnorm4x8(lights.data[i].color_specular); +#ifdef USE_SUBGROUPS + item_from = subgroupBroadcastFirst(subgroupMin(item_from)); + item_to = subgroupBroadcastFirst(subgroupMax(item_to)); +#endif - float attenuation = get_omni_attenuation(d, lights.data[i].inv_radius, attenuation_energy.x); + 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 - vec3 spot_dir = lights.data[i].direction; - vec2 spot_att_angle = unpackHalf2x16(lights.data[i].cone_attenuation_angle); - float scos = max(dot(-normalize(light_rel_vec), spot_dir), spot_att_angle.y); - float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - spot_att_angle.y)); - attenuation *= 1.0 - pow(spot_rim, spot_att_angle.x); + 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 - vec3 light = attenuation_energy.y * color_specular.rgb / M_PI; + //if (!bool(omni_lights.data[light_index].mask & draw_call.layer_mask)) { + // continue; //not masked + //} - vec4 shadow_color_enabled = unpackUnorm4x8(lights.data[i].shadow_color_enabled); + uint light_index = 32 * i + bit; - if (shadow_color_enabled.a > 0.5) { - //has shadow - vec4 v = vec4(view_pos, 1.0); + 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; - vec4 splane = (lights.data[i].shadow_matrix * v); - splane /= splane.w; + if (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); - float depth = texture(sampler2D(shadow_atlas, linear_sampler), splane.xy).r; - float shadow = exp(min(0.0, (depth - splane.z)) / lights.data[i].inv_radius * lights.data[i].shadow_volumetric_fog_fade); + 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); - shadow_attenuation = mix(shadow_color_enabled.rgb, vec3(1.0), shadow); - } + vec3 light = spot_lights.data[light_index].color / M_PI; + + if (spot_lights.data[light_index].shadow_enabled) { + //has shadow + vec4 v = vec4(view_pos, 1.0); + + vec4 splane = (spot_lights.data[light_index].shadow_matrix * v); + splane /= splane.w; - total_light += light * attenuation * shadow_attenuation; + float depth = texture(sampler2D(shadow_atlas, linear_sampler), splane.xy).r; + + shadow_attenuation = exp(min(0.0, (depth - splane.z)) / spot_lights.data[light_index].inv_radius * spot_lights.data[light_index].shadow_volumetric_fog_fade); + } + + total_light += light * attenuation * shadow_attenuation; + } + } } } @@ -470,7 +629,11 @@ void main() { #endif - imageStore(density_map, pos, vec4(total_light, total_density)); + vec4 final_density = vec4(total_light, total_density); + + final_density = mix(final_density, reprojected_density, reproject_amount); + + imageStore(density_map, pos, final_density); #endif #ifdef MODE_FOG |