summaryrefslogtreecommitdiff
path: root/servers/rendering/rasterizer_rd
diff options
context:
space:
mode:
Diffstat (limited to 'servers/rendering/rasterizer_rd')
-rw-r--r--servers/rendering/rasterizer_rd/rasterizer_effects_rd.cpp53
-rw-r--r--servers/rendering/rasterizer_rd/rasterizer_effects_rd.h30
-rw-r--r--servers/rendering/rasterizer_rd/rasterizer_rd.h1
-rw-r--r--servers/rendering/rasterizer_rd/rasterizer_scene_high_end_rd.cpp903
-rw-r--r--servers/rendering/rasterizer_rd/rasterizer_scene_high_end_rd.h91
-rw-r--r--servers/rendering/rasterizer_rd/rasterizer_scene_rd.cpp2753
-rw-r--r--servers/rendering/rasterizer_rd/rasterizer_scene_rd.h502
-rw-r--r--servers/rendering/rasterizer_rd/rasterizer_storage_rd.cpp73
-rw-r--r--servers/rendering/rasterizer_rd/rasterizer_storage_rd.h28
-rw-r--r--servers/rendering/rasterizer_rd/shaders/SCsub7
-rw-r--r--servers/rendering/rasterizer_rd/shaders/copy_to_fb.glsl20
-rw-r--r--servers/rendering/rasterizer_rd/shaders/gi.glsl663
-rw-r--r--servers/rendering/rasterizer_rd/shaders/resolve.glsl110
-rw-r--r--servers/rendering/rasterizer_rd/shaders/scene_high_end.glsl591
-rw-r--r--servers/rendering/rasterizer_rd/shaders/scene_high_end_inc.glsl133
-rw-r--r--servers/rendering/rasterizer_rd/shaders/screen_space_reflection.glsl10
-rw-r--r--servers/rendering/rasterizer_rd/shaders/sdfgi_debug.glsl275
-rw-r--r--servers/rendering/rasterizer_rd/shaders/sdfgi_debug_probes.glsl231
-rw-r--r--servers/rendering/rasterizer_rd/shaders/sdfgi_direct_light.glsl472
-rw-r--r--servers/rendering/rasterizer_rd/shaders/sdfgi_fields.glsl182
-rw-r--r--servers/rendering/rasterizer_rd/shaders/sdfgi_integrate.glsl605
-rw-r--r--servers/rendering/rasterizer_rd/shaders/sdfgi_preprocess.glsl1155
22 files changed, 8013 insertions, 875 deletions
diff --git a/servers/rendering/rasterizer_rd/rasterizer_effects_rd.cpp b/servers/rendering/rasterizer_rd/rasterizer_effects_rd.cpp
index 303cb7ad42..0c036b886c 100644
--- a/servers/rendering/rasterizer_rd/rasterizer_effects_rd.cpp
+++ b/servers/rendering/rasterizer_rd/rasterizer_effects_rd.cpp
@@ -218,7 +218,7 @@ void RasterizerEffectsRD::copy_to_atlas_fb(RID p_source_rd_texture, RID p_dest_f
RD::get_singleton()->draw_list_draw(draw_list, true);
}
-void RasterizerEffectsRD::copy_to_fb_rect(RID p_source_rd_texture, RID p_dest_framebuffer, const Rect2i &p_rect, bool p_flip_y, bool p_force_luminance, bool p_alpha_to_zero) {
+void RasterizerEffectsRD::copy_to_fb_rect(RID p_source_rd_texture, RID p_dest_framebuffer, const Rect2i &p_rect, bool p_flip_y, bool p_force_luminance, bool p_alpha_to_zero, bool p_srgb, RID p_secondary) {
zeromem(&copy_to_fb.push_constant, sizeof(CopyToFbPushConstant));
if (p_flip_y) {
@@ -230,10 +230,16 @@ void RasterizerEffectsRD::copy_to_fb_rect(RID p_source_rd_texture, RID p_dest_fr
if (p_alpha_to_zero) {
copy_to_fb.push_constant.alpha_to_zero = true;
}
+ if (p_srgb) {
+ copy_to_fb.push_constant.srgb = true;
+ }
RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD, Vector<Color>(), 1.0, 0, p_rect);
- RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, copy_to_fb.pipelines[COPY_TO_FB_COPY].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dest_framebuffer)));
+ RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, copy_to_fb.pipelines[p_secondary.is_valid() ? COPY_TO_FB_COPY2 : COPY_TO_FB_COPY].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dest_framebuffer)));
RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_source_rd_texture), 0);
+ if (p_secondary.is_valid()) {
+ RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_secondary), 1);
+ }
RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array);
RD::get_singleton()->draw_list_set_push_constant(draw_list, &copy_to_fb.push_constant, sizeof(CopyToFbPushConstant));
RD::get_singleton()->draw_list_draw(draw_list, true);
@@ -434,7 +440,7 @@ void RasterizerEffectsRD::gaussian_glow(RID p_source_rd_texture, RID p_texture,
RD::get_singleton()->compute_list_end();
}
-void RasterizerEffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal, RenderingServer::EnvironmentSSRRoughnessQuality p_roughness_quality, RID p_roughness, 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) {
+void RasterizerEffectsRD::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;
@@ -451,7 +457,7 @@ void RasterizerEffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal, R
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssr_scale.pipeline);
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_diffuse), 0);
- RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture_pair(p_depth, p_normal), 1);
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture_pair(p_depth, p_normal_roughness), 1);
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_output_blur), 2);
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_image_pair(p_scale_depth, p_scale_normal), 3);
@@ -491,7 +497,7 @@ void RasterizerEffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal, R
if (p_roughness_quality != RS::ENV_SSR_ROUGNESS_QUALITY_DISABLED) {
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_image_pair(p_output, p_blur_radius), 1);
- RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture_pair(p_metallic, p_roughness), 3);
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture_pair(p_metallic, p_normal_roughness), 3);
} else {
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_output), 1);
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_metallic), 3);
@@ -1201,6 +1207,28 @@ void RasterizerEffectsRD::render_sky(RD::DrawListID p_list, float p_time, RID p_
RD::get_singleton()->draw_list_draw(draw_list, true);
}
+void RasterizerEffectsRD::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) {
+ ResolvePushConstant push_constant;
+ push_constant.screen_size[0] = p_screen_size.x;
+ push_constant.screen_size[1] = p_screen_size.y;
+ push_constant.samples = p_samples;
+
+ RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin();
+ RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, resolve.pipelines[p_source_giprobe.is_valid() ? RESOLVE_MODE_GI_GIPROBE : RESOLVE_MODE_GI]);
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture_pair(p_source_depth, p_source_normal_roughness), 0);
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_image_pair(p_dest_depth, p_dest_normal_roughness), 1);
+ if (p_source_giprobe.is_valid()) {
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_giprobe), 2);
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_giprobe), 3);
+ }
+
+ RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ResolvePushConstant));
+
+ RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.x, p_screen_size.y, 1, 8, 8, 1);
+
+ RD::get_singleton()->compute_list_end();
+}
+
RasterizerEffectsRD::RasterizerEffectsRD() {
{ // Initialize copy
Vector<String> copy_modes;
@@ -1228,6 +1256,7 @@ RasterizerEffectsRD::RasterizerEffectsRD() {
Vector<String> copy_modes;
copy_modes.push_back("\n");
copy_modes.push_back("\n#define MODE_PANORAMA_TO_DP\n");
+ copy_modes.push_back("\n#define MODE_TWO_SOURCES\n");
copy_to_fb.shader.initialize(copy_modes);
@@ -1517,6 +1546,20 @@ RasterizerEffectsRD::RasterizerEffectsRD() {
}
}
+ {
+ Vector<String> resolve_modes;
+ resolve_modes.push_back("\n#define MODE_RESOLVE_GI\n");
+ resolve_modes.push_back("\n#define MODE_RESOLVE_GI\n#define GIPROBE_RESOLVE\n");
+
+ resolve.shader.initialize(resolve_modes);
+
+ resolve.shader_version = resolve.shader.version_create();
+
+ for (int i = 0; i < RESOLVE_MODE_MAX; i++) {
+ resolve.pipelines[i] = RD::get_singleton()->compute_pipeline_create(resolve.shader.version_get_shader(resolve.shader_version, i));
+ }
+ }
+
RD::SamplerState sampler;
sampler.mag_filter = RD::SAMPLER_FILTER_LINEAR;
sampler.min_filter = RD::SAMPLER_FILTER_LINEAR;
diff --git a/servers/rendering/rasterizer_rd/rasterizer_effects_rd.h b/servers/rendering/rasterizer_rd/rasterizer_effects_rd.h
index 8a55d2d13c..80849654de 100644
--- a/servers/rendering/rasterizer_rd/rasterizer_effects_rd.h
+++ b/servers/rendering/rasterizer_rd/rasterizer_effects_rd.h
@@ -41,6 +41,7 @@
#include "servers/rendering/rasterizer_rd/shaders/cubemap_filter.glsl.gen.h"
#include "servers/rendering/rasterizer_rd/shaders/cubemap_roughness.glsl.gen.h"
#include "servers/rendering/rasterizer_rd/shaders/luminance_reduce.glsl.gen.h"
+#include "servers/rendering/rasterizer_rd/shaders/resolve.glsl.gen.h"
#include "servers/rendering/rasterizer_rd/shaders/roughness_limiter.glsl.gen.h"
#include "servers/rendering/rasterizer_rd/shaders/screen_space_reflection.glsl.gen.h"
#include "servers/rendering/rasterizer_rd/shaders/screen_space_reflection_filter.glsl.gen.h"
@@ -114,6 +115,7 @@ class RasterizerEffectsRD {
enum CopyToFBMode {
COPY_TO_FB_COPY,
COPY_TO_FB_COPY_PANORAMA_TO_DP,
+ COPY_TO_FB_COPY2,
COPY_TO_FB_MAX,
};
@@ -126,7 +128,8 @@ class RasterizerEffectsRD {
uint32_t force_luminance;
uint32_t alpha_to_zero;
- uint32_t pad[2];
+ uint32_t srgb;
+ uint32_t pad;
};
struct CopyToFb {
@@ -512,6 +515,25 @@ class RasterizerEffectsRD {
RID pipelines[3]; //3 quality levels
} sss;
+ struct ResolvePushConstant {
+ int32_t screen_size[2];
+ int32_t samples;
+ uint32_t pad;
+ };
+
+ enum ResolveMode {
+ RESOLVE_MODE_GI,
+ RESOLVE_MODE_GI_GIPROBE,
+ RESOLVE_MODE_MAX
+ };
+
+ struct Resolve {
+ ResolvePushConstant push_constant;
+ ResolveShaderRD shader;
+ RID shader_version;
+ RID pipelines[RESOLVE_MODE_MAX]; //3 quality levels
+ } resolve;
+
RID default_sampler;
RID default_mipmap_sampler;
RID index_buffer;
@@ -544,7 +566,7 @@ class RasterizerEffectsRD {
RID _get_compute_uniform_set_from_image_pair(RID p_texture, RID p_texture2);
public:
- void copy_to_fb_rect(RID p_source_rd_texture, RID p_dest_framebuffer, const Rect2i &p_rect, bool p_flip_y = false, bool p_force_luminance = false, bool p_alpha_to_zero = false);
+ void copy_to_fb_rect(RID p_source_rd_texture, RID p_dest_framebuffer, const Rect2i &p_rect, bool p_flip_y = false, bool p_force_luminance = false, bool p_alpha_to_zero = false, bool p_srgb = false, RID p_secondary = RID());
void copy_to_rect(RID p_source_rd_texture, RID p_dest_texture, const Rect2i &p_rect, bool p_flip_y = false, bool p_force_luminance = false, bool p_all_source = false, bool p_8_bit_dst = false);
void copy_cubemap_to_panorama(RID p_source_cube, RID p_dest_panorama, const Size2i &p_panorama_size, float p_lod, bool p_is_array);
void copy_depth_to_rect(RID p_source_rd_texture, RID p_dest_framebuffer, const Rect2i &p_rect, bool p_flip_y = false);
@@ -605,10 +627,12 @@ public:
void cubemap_filter(RID p_source_cubemap, Vector<RID> p_dest_cubemap, bool p_use_array);
void render_sky(RD::DrawListID p_list, float p_time, RID p_fb, RID p_samplers, RID p_lights, RenderPipelineVertexFormatCacheRD *p_pipeline, RID p_uniform_set, RID p_texture_set, const CameraMatrix &p_camera, const Basis &p_orientation, float p_multiplier, const Vector3 &p_position);
- void screen_space_reflection(RID p_diffuse, RID p_normal, RS::EnvironmentSSRRoughnessQuality p_roughness_quality, RID p_roughness, 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);
+ void screen_space_reflection(RID p_diffuse, RID p_normal_roughness, RS::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);
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);
+
RasterizerEffectsRD();
~RasterizerEffectsRD();
};
diff --git a/servers/rendering/rasterizer_rd/rasterizer_rd.h b/servers/rendering/rasterizer_rd/rasterizer_rd.h
index cb53a531ac..cdcc6bfd73 100644
--- a/servers/rendering/rasterizer_rd/rasterizer_rd.h
+++ b/servers/rendering/rasterizer_rd/rasterizer_rd.h
@@ -74,6 +74,7 @@ public:
_ALWAYS_INLINE_ uint64_t get_frame_number() const { return frame; }
_ALWAYS_INLINE_ float get_frame_delta_time() const { return delta; }
+ _ALWAYS_INLINE_ double get_total_time() const { return time; }
static Error is_viable() {
return OK;
diff --git a/servers/rendering/rasterizer_rd/rasterizer_scene_high_end_rd.cpp b/servers/rendering/rasterizer_rd/rasterizer_scene_high_end_rd.cpp
index 7d351f249a..890ada019f 100644
--- a/servers/rendering/rasterizer_rd/rasterizer_scene_high_end_rd.cpp
+++ b/servers/rendering/rasterizer_rd/rasterizer_scene_high_end_rd.cpp
@@ -268,8 +268,8 @@ void RasterizerSceneHighEndRD::ShaderData::set_code(const String &p_code) {
blend_state_blend.attachments.push_back(blend_attachment);
RD::PipelineColorBlendState blend_state_opaque = RD::PipelineColorBlendState::create_disabled(1);
RD::PipelineColorBlendState blend_state_opaque_specular = RD::PipelineColorBlendState::create_disabled(2);
- RD::PipelineColorBlendState blend_state_depth_normal = RD::PipelineColorBlendState::create_disabled(1);
- RD::PipelineColorBlendState blend_state_depth_normal_roughness = RD::PipelineColorBlendState::create_disabled(2);
+ RD::PipelineColorBlendState blend_state_depth_normal_roughness = RD::PipelineColorBlendState::create_disabled(1);
+ RD::PipelineColorBlendState blend_state_depth_normal_roughness_giprobe = RD::PipelineColorBlendState::create_disabled(2);
//update pipelines
@@ -310,12 +310,12 @@ void RasterizerSceneHighEndRD::ShaderData::set_code(const String &p_code) {
RD::PipelineDepthStencilState depth_stencil = depth_stencil_state;
if (uses_alpha || uses_blend_alpha) {
- if (k == SHADER_VERSION_COLOR_PASS || k == SHADER_VERSION_VCT_COLOR_PASS || k == SHADER_VERSION_LIGHTMAP_COLOR_PASS) {
+ if (k == SHADER_VERSION_COLOR_PASS || k == SHADER_VERSION_COLOR_PASS_WITH_FORWARD_GI || k == SHADER_VERSION_LIGHTMAP_COLOR_PASS) {
blend_state = blend_state_blend;
if (depth_draw == DEPTH_DRAW_OPAQUE) {
depth_stencil.enable_depth_write = false; //alpha does not draw depth
}
- } else if (uses_depth_pre_pass && (k == SHADER_VERSION_DEPTH_PASS || k == SHADER_VERSION_DEPTH_PASS_DP || k == SHADER_VERSION_DEPTH_PASS_WITH_NORMAL || k == SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS || k == SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL)) {
+ } else if (uses_depth_pre_pass && (k == SHADER_VERSION_DEPTH_PASS || k == SHADER_VERSION_DEPTH_PASS_DP || k == SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS || k == SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL)) {
if (k == SHADER_VERSION_DEPTH_PASS || k == SHADER_VERSION_DEPTH_PASS_DP) {
//none, blend state contains nothing
} else if (k == SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL) {
@@ -328,20 +328,23 @@ void RasterizerSceneHighEndRD::ShaderData::set_code(const String &p_code) {
continue; // do not use this version (will error if using it is attempted)
}
} else {
- if (k == SHADER_VERSION_COLOR_PASS || k == SHADER_VERSION_VCT_COLOR_PASS || k == SHADER_VERSION_LIGHTMAP_COLOR_PASS) {
+ if (k == SHADER_VERSION_COLOR_PASS || k == SHADER_VERSION_COLOR_PASS_WITH_FORWARD_GI || k == SHADER_VERSION_LIGHTMAP_COLOR_PASS) {
blend_state = blend_state_opaque;
} else if (k == SHADER_VERSION_DEPTH_PASS || k == SHADER_VERSION_DEPTH_PASS_DP) {
//none, leave empty
- } else if (k == SHADER_VERSION_DEPTH_PASS_WITH_NORMAL) {
- blend_state = blend_state_depth_normal;
} else if (k == SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS) {
blend_state = blend_state_depth_normal_roughness;
+ } else if (k == SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_GIPROBE) {
+ blend_state = blend_state_depth_normal_roughness_giprobe;
} else if (k == SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL) {
blend_state = RD::PipelineColorBlendState::create_disabled(5); //writes to normal and roughness in opaque way
-
+ } else if (k == SHADER_VERSION_DEPTH_PASS_WITH_SDF) {
+ blend_state = RD::PipelineColorBlendState(); //no color targets for SDF
} else {
//specular write
blend_state = blend_state_opaque_specular;
+ depth_stencil.enable_depth_test = false;
+ depth_stencil.enable_depth_write = false;
}
}
@@ -607,7 +610,78 @@ void RasterizerSceneHighEndRD::RenderBufferDataHighEnd::ensure_specular() {
}
}
+void RasterizerSceneHighEndRD::RenderBufferDataHighEnd::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 RasterizerSceneHighEndRD::RenderBufferDataHighEnd::ensure_giprobe() {
+ if (!giprobe_buffer.is_valid()) {
+ RD::TextureFormat tf;
+ tf.format = RD::DATA_FORMAT_R8G8_UINT;
+ tf.width = width;
+ tf.height = height;
+ tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT;
+
+ if (msaa != RS::VIEWPORT_MSAA_DISABLED) {
+ RD::TextureFormat tf_aa = tf;
+ tf_aa.usage_bits |= RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT;
+ tf_aa.samples = texture_samples;
+ giprobe_buffer_msaa = RD::get_singleton()->texture_create(tf_aa, RD::TextureView());
+ } else {
+ tf.usage_bits |= RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT;
+ }
+
+ tf.usage_bits |= RD::TEXTURE_USAGE_STORAGE_BIT;
+
+ giprobe_buffer = RD::get_singleton()->texture_create(tf, RD::TextureView());
+
+ Vector<RID> fb;
+ if (msaa != RS::VIEWPORT_MSAA_DISABLED) {
+ fb.push_back(depth_msaa);
+ fb.push_back(normal_roughness_buffer_msaa);
+ fb.push_back(giprobe_buffer_msaa);
+ } else {
+ fb.push_back(depth);
+ fb.push_back(normal_roughness_buffer);
+ fb.push_back(giprobe_buffer);
+ }
+
+ depth_normal_roughness_giprobe_fb = RD::get_singleton()->framebuffer_create(fb);
+ }
+}
+
void RasterizerSceneHighEndRD::RenderBufferDataHighEnd::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();
+ }
+
+ if (giprobe_buffer != RID()) {
+ RD::get_singleton()->free(giprobe_buffer);
+ giprobe_buffer = RID();
+
+ if (giprobe_buffer_msaa.is_valid()) {
+ RD::get_singleton()->free(giprobe_buffer_msaa);
+ giprobe_buffer_msaa = RID();
+ }
+
+ depth_normal_roughness_giprobe_fb = RID();
+ }
+
if (color_msaa.is_valid()) {
RD::get_singleton()->free(color_msaa);
color_msaa = RID();
@@ -634,24 +708,18 @@ void RasterizerSceneHighEndRD::RenderBufferDataHighEnd::clear() {
color_fb = RID();
depth_fb = RID();
- if (normal_buffer.is_valid()) {
- RD::get_singleton()->free(normal_buffer);
- if (normal_buffer_msaa.is_valid()) {
- RD::get_singleton()->free(normal_buffer_msaa);
- normal_buffer_msaa = RID();
+ if (normal_roughness_buffer.is_valid()) {
+ RD::get_singleton()->free(normal_roughness_buffer);
+ if (normal_roughness_buffer_msaa.is_valid()) {
+ RD::get_singleton()->free(normal_roughness_buffer_msaa);
+ normal_roughness_buffer_msaa = RID();
}
- normal_buffer = RID();
- depth_normal_fb = RID();
+ normal_roughness_buffer = RID();
+ depth_normal_roughness_fb = RID();
}
- if (roughness_buffer.is_valid()) {
- RD::get_singleton()->free(roughness_buffer);
- if (roughness_buffer_msaa.is_valid()) {
- RD::get_singleton()->free(roughness_buffer_msaa);
- roughness_buffer_msaa = RID();
- }
- roughness_buffer = RID();
- depth_normal_roughness_fb = RID();
+ if (!render_sdfgi_uniform_set.is_null() && RD::get_singleton()->uniform_set_is_valid(render_sdfgi_uniform_set)) {
+ RD::get_singleton()->free(render_sdfgi_uniform_set);
}
}
@@ -686,7 +754,7 @@ void RasterizerSceneHighEndRD::RenderBufferDataHighEnd::configure(RID p_color_bu
tf.width = p_width;
tf.height = p_height;
tf.type = RD::TEXTURE_TYPE_2D;
- tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT;
+ tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT;
RD::TextureSamples ts[RS::VIEWPORT_MSAA_MAX] = {
RD::TEXTURE_SAMPLES_1,
@@ -702,7 +770,7 @@ void RasterizerSceneHighEndRD::RenderBufferDataHighEnd::configure(RID p_color_bu
color_msaa = RD::get_singleton()->texture_create(tf, RD::TextureView());
tf.format = RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_D24_UNORM_S8_UINT, RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) ? RD::DATA_FORMAT_D24_UNORM_S8_UINT : RD::DATA_FORMAT_D32_SFLOAT_S8_UINT;
- tf.usage_bits = RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT;
+ tf.usage_bits = RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT;
depth_msaa = RD::get_singleton()->texture_create(tf, RD::TextureView());
@@ -722,80 +790,38 @@ void RasterizerSceneHighEndRD::RenderBufferDataHighEnd::configure(RID p_color_bu
}
}
-void RasterizerSceneHighEndRD::_allocate_normal_texture(RenderBufferDataHighEnd *rb) {
- if (rb->normal_buffer.is_valid()) {
+void RasterizerSceneHighEndRD::_allocate_normal_roughness_texture(RenderBufferDataHighEnd *rb) {
+ if (rb->normal_roughness_buffer.is_valid()) {
return;
}
RD::TextureFormat tf;
- tf.format = RD::DATA_FORMAT_A2B10G10R10_UNORM_PACK32;
+ tf.format = RD::DATA_FORMAT_R8G8B8A8_UNORM;
tf.width = rb->width;
tf.height = rb->height;
tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT;
if (rb->msaa != RS::VIEWPORT_MSAA_DISABLED) {
- tf.usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT;
- } else {
- tf.usage_bits |= RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT;
- }
-
- rb->normal_buffer = RD::get_singleton()->texture_create(tf, RD::TextureView());
-
- if (rb->msaa == RS::VIEWPORT_MSAA_DISABLED) {
- Vector<RID> fb;
- fb.push_back(rb->depth);
- fb.push_back(rb->normal_buffer);
- rb->depth_normal_fb = RD::get_singleton()->framebuffer_create(fb);
- } else {
- tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT;
- tf.samples = rb->texture_samples;
- rb->normal_buffer_msaa = RD::get_singleton()->texture_create(tf, RD::TextureView());
-
- Vector<RID> fb;
- fb.push_back(rb->depth_msaa);
- fb.push_back(rb->normal_buffer_msaa);
- rb->depth_normal_fb = RD::get_singleton()->framebuffer_create(fb);
- }
-
- _render_buffers_clear_uniform_set(rb);
-}
-
-void RasterizerSceneHighEndRD::_allocate_roughness_texture(RenderBufferDataHighEnd *rb) {
- if (rb->roughness_buffer.is_valid()) {
- return;
- }
-
- ERR_FAIL_COND(rb->normal_buffer.is_null());
-
- RD::TextureFormat tf;
- tf.format = RD::DATA_FORMAT_R8_UNORM;
- tf.width = rb->width;
- tf.height = rb->height;
- tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT;
-
- if (rb->msaa != RS::VIEWPORT_MSAA_DISABLED) {
- tf.usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT;
+ tf.usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_STORAGE_BIT;
} else {
tf.usage_bits |= RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT;
}
- rb->roughness_buffer = RD::get_singleton()->texture_create(tf, RD::TextureView());
+ rb->normal_roughness_buffer = RD::get_singleton()->texture_create(tf, RD::TextureView());
if (rb->msaa == RS::VIEWPORT_MSAA_DISABLED) {
Vector<RID> fb;
fb.push_back(rb->depth);
- fb.push_back(rb->normal_buffer);
- fb.push_back(rb->roughness_buffer);
+ fb.push_back(rb->normal_roughness_buffer);
rb->depth_normal_roughness_fb = RD::get_singleton()->framebuffer_create(fb);
} else {
- tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT;
+ tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT;
tf.samples = rb->texture_samples;
- rb->roughness_buffer_msaa = RD::get_singleton()->texture_create(tf, RD::TextureView());
+ rb->normal_roughness_buffer_msaa = RD::get_singleton()->texture_create(tf, RD::TextureView());
Vector<RID> fb;
fb.push_back(rb->depth_msaa);
- fb.push_back(rb->normal_buffer_msaa);
- fb.push_back(rb->roughness_buffer_msaa);
+ fb.push_back(rb->normal_roughness_buffer_msaa);
rb->depth_normal_roughness_fb = RD::get_singleton()->framebuffer_create(fb);
}
@@ -813,7 +839,7 @@ bool RasterizerSceneHighEndRD::free(RID p_rid) {
return false;
}
-void RasterizerSceneHighEndRD::_fill_instances(RenderList::Element **p_elements, int p_element_count, bool p_for_depth) {
+void RasterizerSceneHighEndRD::_fill_instances(RenderList::Element **p_elements, int p_element_count, bool p_for_depth, bool p_has_sdfgi, bool p_has_opaque_gi) {
uint32_t lightmap_captures_used = 0;
for (int i = 0; i < p_element_count; i++) {
@@ -855,37 +881,7 @@ void RasterizerSceneHighEndRD::_fill_instances(RenderList::Element **p_elements,
continue;
}
- if (!e->instance->gi_probe_instances.empty()) {
- uint32_t written = 0;
- for (int j = 0; j < e->instance->gi_probe_instances.size(); j++) {
- RID probe = e->instance->gi_probe_instances[j];
- int slot = gi_probe_instance_get_slot(probe);
- if (slot < 0) {
- continue; //unallocated, dont render
- }
-
- if (render_pass != gi_probe_instance_get_render_pass(probe)) {
- continue; //not rendered in this frame
- }
-
- uint32_t index = gi_probe_instance_get_render_index(probe);
-
- if (written == 0) {
- id.gi_offset = index;
- id.flags |= INSTANCE_DATA_FLAG_USE_GIPROBE;
- written = 1;
- } else {
- id.gi_offset = index << 16;
- written = 2;
- break;
- }
- }
- if (written == 0) {
- id.gi_offset = 0xFFFFFFFF;
- } else if (written == 1) {
- id.gi_offset |= 0xFFFF0000;
- }
- } else if (e->instance->lightmap) {
+ if (e->instance->lightmap) {
int32_t lightmap_index = storage->lightmap_get_array_index(e->instance->lightmap->base);
if (lightmap_index >= 0) {
id.gi_offset = lightmap_index;
@@ -916,8 +912,40 @@ void RasterizerSceneHighEndRD::_fill_instances(RenderList::Element **p_elements,
id.gi_offset = lightmap_captures_used;
lightmap_captures_used++;
}
+
} else {
- id.gi_offset = 0xFFFFFFFF;
+ if (p_has_opaque_gi) {
+ id.flags |= INSTANCE_DATA_FLAG_USE_GI_BUFFERS;
+ }
+
+ if (!e->instance->gi_probe_instances.empty()) {
+ uint32_t written = 0;
+ for (int j = 0; j < e->instance->gi_probe_instances.size(); j++) {
+ RID probe = e->instance->gi_probe_instances[j];
+
+ uint32_t index = gi_probe_instance_get_render_index(probe);
+
+ if (written == 0) {
+ id.gi_offset = index;
+ id.flags |= INSTANCE_DATA_FLAG_USE_GIPROBE;
+ written = 1;
+ } else {
+ id.gi_offset = index << 16;
+ written = 2;
+ break;
+ }
+ }
+ if (written == 0) {
+ id.gi_offset = 0xFFFFFFFF;
+ } else if (written == 1) {
+ id.gi_offset |= 0xFFFF0000;
+ }
+ } else {
+ if (p_has_sdfgi && (e->instance->baked_light || e->instance->dynamic_gi)) {
+ id.flags |= INSTANCE_DATA_FLAG_USE_SDFGI;
+ }
+ id.gi_offset = 0xFFFFFFFF;
+ }
}
}
@@ -970,7 +998,7 @@ void RasterizerSceneHighEndRD::_render_list(RenderingDevice::DrawListID p_draw_l
//find cull variant
ShaderData::CullVariant cull_variant;
- if (p_pass_mode == PASS_MODE_DEPTH_MATERIAL || ((p_pass_mode == PASS_MODE_SHADOW || p_pass_mode == PASS_MODE_SHADOW_DP) && e->instance->cast_shadows == RS::SHADOW_CASTING_SETTING_DOUBLE_SIDED)) {
+ if (p_pass_mode == PASS_MODE_DEPTH_MATERIAL || p_pass_mode == PASS_MODE_SDF || ((p_pass_mode == PASS_MODE_SHADOW || p_pass_mode == PASS_MODE_SHADOW_DP) && e->instance->cast_shadows == RS::SHADOW_CASTING_SETTING_DOUBLE_SIDED)) {
cull_variant = ShaderData::CULL_VARIANT_DOUBLE_SIDED;
} else {
bool mirror = e->instance->mirror;
@@ -1016,8 +1044,8 @@ void RasterizerSceneHighEndRD::_render_list(RenderingDevice::DrawListID p_draw_l
case PASS_MODE_COLOR_TRANSPARENT: {
if (e->uses_lightmap) {
shader_version = SHADER_VERSION_LIGHTMAP_COLOR_PASS;
- } else if (e->uses_vct) {
- shader_version = SHADER_VERSION_VCT_COLOR_PASS;
+ } else if (e->uses_forward_gi) {
+ shader_version = SHADER_VERSION_COLOR_PASS_WITH_FORWARD_GI;
} else {
shader_version = SHADER_VERSION_COLOR_PASS;
}
@@ -1025,8 +1053,6 @@ void RasterizerSceneHighEndRD::_render_list(RenderingDevice::DrawListID p_draw_l
case PASS_MODE_COLOR_SPECULAR: {
if (e->uses_lightmap) {
shader_version = SHADER_VERSION_LIGHTMAP_COLOR_PASS_WITH_SEPARATE_SPECULAR;
- } else if (e->uses_vct) {
- shader_version = SHADER_VERSION_VCT_COLOR_PASS_WITH_SEPARATE_SPECULAR;
} else {
shader_version = SHADER_VERSION_COLOR_PASS_WITH_SEPARATE_SPECULAR;
}
@@ -1038,15 +1064,18 @@ void RasterizerSceneHighEndRD::_render_list(RenderingDevice::DrawListID p_draw_l
case PASS_MODE_SHADOW_DP: {
shader_version = SHADER_VERSION_DEPTH_PASS_DP;
} break;
- case PASS_MODE_DEPTH_NORMAL: {
- shader_version = SHADER_VERSION_DEPTH_PASS_WITH_NORMAL;
- } break;
case PASS_MODE_DEPTH_NORMAL_ROUGHNESS: {
shader_version = SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS;
} break;
+ case PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE: {
+ shader_version = SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_GIPROBE;
+ } break;
case PASS_MODE_DEPTH_MATERIAL: {
shader_version = SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL;
} break;
+ case PASS_MODE_SDF: {
+ shader_version = SHADER_VERSION_DEPTH_PASS_WITH_SDF;
+ } break;
}
RenderPipelineVertexFormatCacheRD *pipeline = nullptr;
@@ -1134,7 +1163,7 @@ void RasterizerSceneHighEndRD::_render_list(RenderingDevice::DrawListID p_draw_l
}
}
-void RasterizerSceneHighEndRD::_setup_environment(RID p_environment, 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 RasterizerSceneHighEndRD::_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) {
//CameraMatrix projection = p_cam_projection;
//projection.flip_y(); // Vulkan and modern APIs use Y-Down
CameraMatrix correction;
@@ -1178,6 +1207,70 @@ void RasterizerSceneHighEndRD::_setup_environment(RID p_environment, const Camer
//time global variables
scene_state.ubo.time = time;
+ scene_state.ubo.gi_upscale_for_msaa = false;
+
+ if (p_render_buffers.is_valid()) {
+ RenderBufferDataHighEnd *render_buffers = (RenderBufferDataHighEnd *)render_buffers_get_data(p_render_buffers);
+ if (render_buffers->msaa != RS::VIEWPORT_MSAA_DISABLED) {
+ scene_state.ubo.gi_upscale_for_msaa = true;
+ }
+ }
+#if 0
+ if (p_render_buffers.is_valid() && render_buffers_is_sdfgi_enabled(p_render_buffers)) {
+
+ scene_state.ubo.sdfgi_cascade_count = render_buffers_get_sdfgi_cascade_count(p_render_buffers);
+ scene_state.ubo.sdfgi_probe_axis_size = render_buffers_get_sdfgi_cascade_probe_count(p_render_buffers);
+ scene_state.ubo.sdfgi_cascade_probe_size[0] = scene_state.ubo.sdfgi_probe_axis_size - 1; //float version for performance
+ scene_state.ubo.sdfgi_cascade_probe_size[1] = scene_state.ubo.sdfgi_probe_axis_size - 1;
+ scene_state.ubo.sdfgi_cascade_probe_size[2] = scene_state.ubo.sdfgi_probe_axis_size - 1;
+
+ float csize = render_buffers_get_sdfgi_cascade_size(p_render_buffers);
+ scene_state.ubo.sdfgi_probe_to_uvw = 1.0 / float(scene_state.ubo.sdfgi_cascade_probe_size[0]);
+ float occ_bias = 0.0;
+ scene_state.ubo.sdfgi_occlusion_bias = occ_bias / csize;
+ scene_state.ubo.sdfgi_use_occlusion = render_buffers_is_sdfgi_using_occlusion(p_render_buffers);
+ scene_state.ubo.sdfgi_energy = render_buffers_get_sdfgi_energy(p_render_buffers);
+
+ float cascade_voxel_size = (csize / scene_state.ubo.sdfgi_cascade_probe_size[0]);
+ float occlusion_clamp = (cascade_voxel_size - 0.5) / cascade_voxel_size;
+ scene_state.ubo.sdfgi_occlusion_clamp[0] = occlusion_clamp;
+ scene_state.ubo.sdfgi_occlusion_clamp[1] = occlusion_clamp;
+ scene_state.ubo.sdfgi_occlusion_clamp[2] = occlusion_clamp;
+ scene_state.ubo.sdfgi_normal_bias = (render_buffers_get_sdfgi_normal_bias(p_render_buffers) / csize) * scene_state.ubo.sdfgi_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_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);
+ scene_state.ubo.sdfgi_lightprobe_tex_pixel_size[2] = 1.0;
+
+ scene_state.ubo.sdfgi_probe_uv_offset[0] = float(oct_size + 2) * scene_state.ubo.sdfgi_lightprobe_tex_pixel_size[0];
+ scene_state.ubo.sdfgi_probe_uv_offset[1] = float(oct_size + 2) * scene_state.ubo.sdfgi_lightprobe_tex_pixel_size[1];
+ scene_state.ubo.sdfgi_probe_uv_offset[2] = float((oct_size + 2) * scene_state.ubo.sdfgi_probe_axis_size) * scene_state.ubo.sdfgi_lightprobe_tex_pixel_size[0];
+
+ scene_state.ubo.sdfgi_occlusion_renormalize[0] = 0.5;
+ scene_state.ubo.sdfgi_occlusion_renormalize[1] = 1.0;
+ scene_state.ubo.sdfgi_occlusion_renormalize[2] = 1.0 / float(scene_state.ubo.sdfgi_cascade_count);
+
+ for (uint32_t i = 0; i < scene_state.ubo.sdfgi_cascade_count; i++) {
+ SceneState::UBO::SDFGICascade &c = scene_state.ubo.sdfgi_cascades[i];
+ Vector3 pos = render_buffers_get_sdfgi_cascade_offset(p_render_buffers, i);
+ pos -= p_cam_transform.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 / render_buffers_get_sdfgi_cascade_probe_size(p_render_buffers, i);
+
+ Vector3i probe_ofs = render_buffers_get_sdfgi_cascade_probe_offset(p_render_buffers, i);
+ c.probe_world_offset[0] = probe_ofs.x;
+ c.probe_world_offset[1] = probe_ofs.y;
+ c.probe_world_offset[2] = probe_ofs.z;
+ }
+ }
+#endif
if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_UNSHADED) {
scene_state.ubo.use_ambient_light = true;
scene_state.ubo.ambient_light_color_energy[0] = 1;
@@ -1260,11 +1353,13 @@ void RasterizerSceneHighEndRD::_setup_environment(RID p_environment, const Camer
}
scene_state.ubo.roughness_limiter_enabled = p_opaque_render_buffers && screen_space_roughness_limiter_is_active();
+ 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);
}
-void RasterizerSceneHighEndRD::_add_geometry(InstanceBase *p_instance, uint32_t p_surface, RID p_material, PassMode p_pass_mode, uint32_t p_geometry_index) {
+void RasterizerSceneHighEndRD::_add_geometry(InstanceBase *p_instance, uint32_t p_surface, RID p_material, PassMode p_pass_mode, uint32_t p_geometry_index, bool p_using_sdfgi) {
RID m_src;
m_src = p_instance->material_override.is_valid() ? p_instance->material_override : p_material;
@@ -1293,18 +1388,18 @@ void RasterizerSceneHighEndRD::_add_geometry(InstanceBase *p_instance, uint32_t
ERR_FAIL_COND(!material);
- _add_geometry_with_material(p_instance, p_surface, material, m_src, p_pass_mode, p_geometry_index);
+ _add_geometry_with_material(p_instance, p_surface, material, m_src, p_pass_mode, p_geometry_index, p_using_sdfgi);
while (material->next_pass.is_valid()) {
material = (MaterialData *)storage->material_get_data(material->next_pass, RasterizerStorageRD::SHADER_TYPE_3D);
if (!material || !material->shader_data->valid) {
break;
+ _add_geometry_with_material(p_instance, p_surface, material, material->next_pass, p_pass_mode, p_geometry_index, p_using_sdfgi);
}
- _add_geometry_with_material(p_instance, p_surface, material, material->next_pass, p_pass_mode, p_geometry_index);
}
}
-void RasterizerSceneHighEndRD::_add_geometry_with_material(InstanceBase *p_instance, uint32_t p_surface, MaterialData *p_material, RID p_material_rid, PassMode p_pass_mode, uint32_t p_geometry_index) {
+void RasterizerSceneHighEndRD::_add_geometry_with_material(InstanceBase *p_instance, uint32_t p_surface, MaterialData *p_material, RID p_material_rid, PassMode p_pass_mode, uint32_t p_geometry_index, bool p_using_sdfgi) {
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;
@@ -1332,13 +1427,11 @@ void RasterizerSceneHighEndRD::_add_geometry_with_material(InstanceBase *p_insta
return;
}
- if (p_pass_mode != PASS_MODE_DEPTH_MATERIAL && !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) {
+ if ((p_pass_mode != PASS_MODE_DEPTH_MATERIAL && p_pass_mode != PASS_MODE_SDF) && !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) {
//shader does not use discard and does not write a vertex position, use generic material
if (p_pass_mode == PASS_MODE_SHADOW || p_pass_mode == PASS_MODE_DEPTH) {
p_material = (MaterialData *)storage->material_get_data(default_material, RasterizerStorageRD::SHADER_TYPE_3D);
- } else if (p_pass_mode == PASS_MODE_DEPTH_NORMAL && !p_material->shader_data->uses_normal) {
- p_material = (MaterialData *)storage->material_get_data(default_material, RasterizerStorageRD::SHADER_TYPE_3D);
- } else if (p_pass_mode == PASS_MODE_DEPTH_NORMAL_ROUGHNESS && !p_material->shader_data->uses_normal && !p_material->shader_data->uses_roughness) {
+ } else if ((p_pass_mode == PASS_MODE_DEPTH_NORMAL_ROUGHNESS || p_pass_mode == PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE) && !p_material->shader_data->uses_normal && !p_material->shader_data->uses_roughness) {
p_material = (MaterialData *)storage->material_get_data(default_material, RasterizerStorageRD::SHADER_TYPE_3D);
}
}
@@ -1346,7 +1439,9 @@ void RasterizerSceneHighEndRD::_add_geometry_with_material(InstanceBase *p_insta
has_alpha = false;
}
- RenderList::Element *e = (has_alpha || p_material->shader_data->depth_test == ShaderData::DEPTH_TEST_DISABLED) ? render_list.add_alpha_element() : render_list.add_element();
+ has_alpha = has_alpha || p_material->shader_data->depth_test == ShaderData::DEPTH_TEST_DISABLED;
+
+ RenderList::Element *e = has_alpha ? render_list.add_alpha_element() : render_list.add_element();
if (!e) {
return;
@@ -1373,7 +1468,7 @@ void RasterizerSceneHighEndRD::_add_geometry_with_material(InstanceBase *p_insta
e->material_index = e->material->index;
e->uses_instancing = e->instance->base_type == RS::INSTANCE_MULTIMESH;
e->uses_lightmap = e->instance->lightmap != nullptr || !e->instance->lightmap_sh.empty();
- e->uses_vct = e->instance->gi_probe_instances.size();
+ e->uses_forward_gi = has_alpha && (e->instance->gi_probe_instances.size() || p_using_sdfgi);
e->shader_index = e->shader_index;
e->depth_layer = e->instance->depth_layer;
e->priority = p_material->priority;
@@ -1383,7 +1478,7 @@ void RasterizerSceneHighEndRD::_add_geometry_with_material(InstanceBase *p_insta
}
}
-void RasterizerSceneHighEndRD::_fill_render_list(InstanceBase **p_cull_result, int p_cull_count, PassMode p_pass_mode, bool p_no_gi) {
+void RasterizerSceneHighEndRD::_fill_render_list(InstanceBase **p_cull_result, int p_cull_count, PassMode p_pass_mode, bool p_using_sdfgi) {
scene_state.current_shader_index = 0;
scene_state.current_material_index = 0;
scene_state.used_sss = false;
@@ -1415,7 +1510,7 @@ void RasterizerSceneHighEndRD::_fill_render_list(InstanceBase **p_cull_result, i
RID material = inst_materials[j].is_valid() ? inst_materials[j] : materials[j];
uint32_t surface_index = storage->mesh_surface_get_render_pass_index(inst->base, j, render_pass, &geometry_index);
- _add_geometry(inst, j, material, p_pass_mode, surface_index);
+ _add_geometry(inst, j, material, p_pass_mode, surface_index, p_using_sdfgi);
}
//mesh->last_pass=frame;
@@ -1443,7 +1538,7 @@ void RasterizerSceneHighEndRD::_fill_render_list(InstanceBase **p_cull_result, i
for (uint32_t j = 0; j < surface_count; j++) {
uint32_t surface_index = storage->mesh_surface_get_multimesh_render_pass_index(mesh, j, render_pass, &geometry_index);
- _add_geometry(inst, j, materials[j], p_pass_mode, surface_index);
+ _add_geometry(inst, j, materials[j], p_pass_mode, surface_index, p_using_sdfgi);
}
} break;
@@ -1525,30 +1620,13 @@ void RasterizerSceneHighEndRD::_setup_reflections(RID *p_reflection_probe_cull_r
reflection_ubo.params[2] = interior ? 1.0 : 0.0;
reflection_ubo.params[3] = box_projection ? 1.0 : 0.0;
- if (interior) {
- Color ambient_linear = storage->reflection_probe_get_interior_ambient(base_probe).to_linear();
- float interior_ambient_energy = storage->reflection_probe_get_interior_ambient_energy(base_probe);
- float interior_ambient_probe_contrib = storage->reflection_probe_get_interior_ambient_probe_contribution(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[3] = interior_ambient_probe_contrib;
- } else {
- Color ambient_linear = storage->reflection_probe_get_interior_ambient(base_probe).to_linear();
- if (is_environment(p_environment)) {
- Color env_ambient_color = environment_get_ambient_light_color(p_environment).to_linear();
- float env_ambient_energy = environment_get_ambient_light_energy(p_environment);
- ambient_linear = env_ambient_color;
- ambient_linear.r *= env_ambient_energy;
- ambient_linear.g *= env_ambient_energy;
- ambient_linear.b *= env_ambient_energy;
- }
-
- reflection_ubo.ambient[0] = ambient_linear.r;
- reflection_ubo.ambient[1] = ambient_linear.g;
- reflection_ubo.ambient[2] = ambient_linear.b;
- reflection_ubo.ambient[3] = 0; //not used in exterior mode, since it just blends with regular ambient light
- }
+ 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 proj = (p_camera_inverse_transform * transform).inverse();
@@ -1583,59 +1661,6 @@ void RasterizerSceneHighEndRD::_setup_lightmaps(InstanceBase **p_lightmap_cull_r
}
}
-void RasterizerSceneHighEndRD::_setup_gi_probes(RID *p_gi_probe_probe_cull_result, int p_gi_probe_probe_cull_count, const Transform &p_camera_transform) {
- int index = 0;
-
- for (int i = 0; i < p_gi_probe_probe_cull_count; i++) {
- RID rpi = p_gi_probe_probe_cull_result[i];
-
- if (index >= (int)scene_state.max_gi_probes) {
- continue;
- }
-
- int slot = gi_probe_instance_get_slot(rpi);
- if (slot < 0) {
- continue; //not usable
- }
-
- RID base_probe = gi_probe_instance_get_base_probe(rpi);
-
- GIProbeData &gi_probe_ubo = scene_state.gi_probes[index];
-
- Transform to_cell = gi_probe_instance_get_transform_to_cell(rpi) * p_camera_transform;
-
- store_transform(to_cell, gi_probe_ubo.xform);
-
- Vector3 bounds = storage->gi_probe_get_octree_size(base_probe);
-
- gi_probe_ubo.bounds[0] = bounds.x;
- gi_probe_ubo.bounds[1] = bounds.y;
- gi_probe_ubo.bounds[2] = bounds.z;
-
- gi_probe_ubo.dynamic_range = storage->gi_probe_get_dynamic_range(base_probe) * storage->gi_probe_get_energy(base_probe);
- gi_probe_ubo.bias = storage->gi_probe_get_bias(base_probe);
- gi_probe_ubo.normal_bias = storage->gi_probe_get_normal_bias(base_probe);
- gi_probe_ubo.blend_ambient = !storage->gi_probe_is_interior(base_probe);
- gi_probe_ubo.texture_slot = gi_probe_instance_get_slot(rpi);
- gi_probe_ubo.anisotropy_strength = storage->gi_probe_get_anisotropy_strength(base_probe);
- gi_probe_ubo.ao = storage->gi_probe_get_ao(base_probe);
- gi_probe_ubo.ao_size = Math::pow(storage->gi_probe_get_ao_size(base_probe), 4.0f);
-
- if (gi_probe_is_anisotropic()) {
- gi_probe_ubo.texture_slot *= 3;
- }
-
- gi_probe_instance_set_render_index(rpi, index);
- gi_probe_instance_set_render_pass(rpi, render_pass);
-
- index++;
- }
-
- if (index) {
- RD::get_singleton()->buffer_update(scene_state.gi_probe_buffer, 0, index * sizeof(GIProbeData), scene_state.gi_probes, true);
- }
-}
-
void RasterizerSceneHighEndRD::_setup_lights(RID *p_light_cull_result, int p_light_cull_count, const Transform &p_camera_inverse_transform, RID p_shadow_atlas, bool p_using_shadows) {
uint32_t light_count = 0;
scene_state.ubo.directional_light_count = 0;
@@ -2156,6 +2181,8 @@ void RasterizerSceneHighEndRD::_render_scene(RID p_render_buffer, const Transfor
Vector<Color> depth_pass_clear;
bool using_separate_specular = false;
bool using_ssr = false;
+ bool using_sdfgi = false;
+ bool using_giprobe = false;
if (render_buffer) {
screen_pixel_size.width = 1.0 / render_buffer->width;
@@ -2165,43 +2192,55 @@ void RasterizerSceneHighEndRD::_render_scene(RID p_render_buffer, const Transfor
opaque_framebuffer = render_buffer->color_fb;
- if (p_environment.is_valid() && environment_is_ssr_enabled(p_environment)) {
- depth_pass_mode = PASS_MODE_DEPTH_NORMAL_ROUGHNESS;
- render_buffer->ensure_specular();
- using_separate_specular = true;
- using_ssr = true;
- opaque_specular_framebuffer = render_buffer->color_specular_fb;
- } else if (screen_space_roughness_limiter_is_active()) {
- depth_pass_mode = PASS_MODE_DEPTH_NORMAL;
- //we need to allocate both these, if not allocated
- _allocate_normal_texture(render_buffer);
- _allocate_roughness_texture(render_buffer);
+ if (p_gi_probe_cull_count > 0) {
+ using_giprobe = true;
+ render_buffer->ensure_gi();
+ }
+
+ if (!p_environment.is_valid() && using_giprobe) {
+ depth_pass_mode = PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE;
+
+ } else if (p_environment.is_valid() && (environment_is_ssr_enabled(p_environment) || environment_is_sdfgi_enabled(p_environment) || using_giprobe)) {
+ 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;
+ }
+
+ if (environment_is_ssr_enabled(p_environment)) {
+ render_buffer->ensure_specular();
+ using_separate_specular = true;
+ using_ssr = true;
+ opaque_specular_framebuffer = render_buffer->color_specular_fb;
+ }
+
} else if (p_environment.is_valid() && (environment_is_ssao_enabled(p_environment) || get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_NORMAL_BUFFER)) {
- depth_pass_mode = PASS_MODE_DEPTH_NORMAL;
+ depth_pass_mode = PASS_MODE_DEPTH_NORMAL_ROUGHNESS;
}
switch (depth_pass_mode) {
case PASS_MODE_DEPTH: {
depth_framebuffer = render_buffer->depth_fb;
} break;
- case PASS_MODE_DEPTH_NORMAL: {
- _allocate_normal_texture(render_buffer);
- depth_framebuffer = render_buffer->depth_normal_fb;
- depth_pass_clear.push_back(Color(0.5, 0.5, 0.5, 0));
- } break;
case PASS_MODE_DEPTH_NORMAL_ROUGHNESS: {
- _allocate_normal_texture(render_buffer);
- _allocate_roughness_texture(render_buffer);
+ _allocate_normal_roughness_texture(render_buffer);
depth_framebuffer = render_buffer->depth_normal_roughness_fb;
depth_pass_clear.push_back(Color(0.5, 0.5, 0.5, 0));
- depth_pass_clear.push_back(Color());
+ } break;
+ case PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE: {
+ _allocate_normal_roughness_texture(render_buffer);
+ render_buffer->ensure_giprobe();
+ depth_framebuffer = render_buffer->depth_normal_roughness_giprobe_fb;
+ depth_pass_clear.push_back(Color(0.5, 0.5, 0.5, 0));
+ depth_pass_clear.push_back(Color(0, 0, 0, 0));
} break;
default: {
};
}
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;
@@ -2216,7 +2255,6 @@ void RasterizerSceneHighEndRD::_render_scene(RID p_render_buffer, const Transfor
if (storage->reflection_probe_is_interior(reflection_probe_instance_get_probe(p_reflection_probe))) {
p_environment = RID(); //no environment on interiors
}
-
} else {
ERR_FAIL(); //bug?
}
@@ -2226,16 +2264,15 @@ void RasterizerSceneHighEndRD::_render_scene(RID p_render_buffer, const Transfor
_setup_lights(p_light_cull_result, p_light_cull_count, p_cam_transform.affine_inverse(), p_shadow_atlas, using_shadows);
_setup_decals(p_decal_cull_result, p_decal_cull_count, p_cam_transform.affine_inverse());
_setup_reflections(p_reflection_probe_cull_result, p_reflection_probe_cull_count, p_cam_transform.affine_inverse(), p_environment);
- _setup_gi_probes(p_gi_probe_cull_result, p_gi_probe_cull_count, p_cam_transform);
_setup_lightmaps(p_lightmap_cull_result, p_lightmap_cull_count, p_cam_transform);
- _setup_environment(p_environment, 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_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);
cluster_builder.bake_cluster(); //bake to cluster
_update_render_base_uniform_set(); //may have changed due to the above (light buffer enlarged, as an example)
render_list.clear();
- _fill_render_list(p_cull_result, p_cull_count, PASS_MODE_COLOR, render_buffer == nullptr);
+ _fill_render_list(p_cull_result, p_cull_count, PASS_MODE_COLOR, using_sdfgi);
bool using_sss = render_buffer && scene_state.used_sss && sub_surface_scattering_get_quality() != RS::SUB_SURFACE_SCATTERING_QUALITY_DISABLED;
@@ -2307,47 +2344,46 @@ void RasterizerSceneHighEndRD::_render_scene(RID p_render_buffer, const Transfor
clear_color = p_default_bg_color;
}
- _setup_view_dependant_uniform_set(p_shadow_atlas, p_reflection_atlas);
+ _setup_view_dependant_uniform_set(p_shadow_atlas, p_reflection_atlas, p_gi_probe_cull_result, p_gi_probe_cull_count);
render_list.sort_by_key(false);
- _fill_instances(render_list.elements, render_list.element_count, false);
+ _fill_instances(render_list.elements, render_list.element_count, false, false, using_sdfgi || using_giprobe);
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 = depth_framebuffer.is_valid();
RID render_buffers_uniform_set;
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");
- bool finish_depth = using_ssao;
+ bool finish_depth = using_ssao || using_sdfgi || using_giprobe;
RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(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);
_render_list(draw_list, RD::get_singleton()->framebuffer_get_format(depth_framebuffer), render_list.elements, render_list.element_count, false, depth_pass_mode, render_buffer == nullptr, radiance_uniform_set, RID(), get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME);
RD::get_singleton()->draw_list_end();
if (render_buffer && render_buffer->msaa != RS::VIEWPORT_MSAA_DISABLED) {
- if (finish_depth) {
+ if (depth_pass_mode == PASS_MODE_DEPTH_NORMAL_ROUGHNESS || depth_pass_mode == PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE) {
+ 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);
}
-
- if (depth_pass_mode == PASS_MODE_DEPTH_NORMAL || depth_pass_mode == PASS_MODE_DEPTH_NORMAL_ROUGHNESS) {
- RD::get_singleton()->texture_resolve_multisample(render_buffer->normal_buffer_msaa, render_buffer->normal_buffer, true);
- if (depth_pass_mode == PASS_MODE_DEPTH_NORMAL_ROUGHNESS) {
- RD::get_singleton()->texture_resolve_multisample(render_buffer->roughness_buffer_msaa, render_buffer->roughness_buffer, true);
- }
- }
}
+
+ continue_depth = !finish_depth;
}
if (using_ssao) {
- _process_ssao(p_render_buffer, p_environment, render_buffer->normal_buffer, p_cam_projection);
+ _process_ssao(p_render_buffer, p_environment, render_buffer->normal_roughness_buffer, p_cam_projection);
}
- if (p_render_buffer.is_valid() && screen_space_roughness_limiter_is_active()) {
- storage->get_effects()->roughness_limit(render_buffer->normal_buffer, render_buffer->roughness_buffer, Size2(render_buffer->width, render_buffer->height), screen_space_roughness_limiter_get_curve());
+ 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_probe_cull_result, p_gi_probe_cull_count);
}
if (p_render_buffer.is_valid()) {
@@ -2356,7 +2392,7 @@ void RasterizerSceneHighEndRD::_render_scene(RID p_render_buffer, const Transfor
render_buffers_uniform_set = render_buffer->uniform_set;
}
- _setup_environment(p_environment, 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());
+ _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());
RENDER_TIMESTAMP("Render Opaque Pass");
@@ -2364,8 +2400,8 @@ void RasterizerSceneHighEndRD::_render_scene(RID p_render_buffer, const Transfor
bool can_continue_depth = !scene_state.used_depth_texture && !using_ssr && !using_sss;
{
- bool will_continue_color = (can_continue_color || draw_sky || debug_giprobes);
- bool will_continue_depth = (can_continue_depth || draw_sky || debug_giprobes);
+ bool will_continue_color = (can_continue_color || draw_sky || debug_giprobes || debug_sdfgi_probes);
+ bool will_continue_depth = (can_continue_depth || draw_sky || debug_giprobes || debug_sdfgi_probes);
//regular forward for now
Vector<Color> c;
@@ -2379,7 +2415,7 @@ void RasterizerSceneHighEndRD::_render_scene(RID p_render_buffer, const Transfor
}
RID framebuffer = using_separate_specular ? opaque_specular_framebuffer : opaque_framebuffer;
- RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(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 ? (using_ssao ? 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);
+ RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(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);
_render_list(draw_list, RD::get_singleton()->framebuffer_get_format(framebuffer), render_list.elements, render_list.element_count, false, using_separate_specular ? PASS_MODE_COLOR_SPECULAR : PASS_MODE_COLOR, render_buffer == nullptr, radiance_uniform_set, render_buffers_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME);
RD::get_singleton()->draw_list_end();
@@ -2405,6 +2441,19 @@ void RasterizerSceneHighEndRD::_render_scene(RID p_render_buffer, const Transfor
RD::get_singleton()->draw_list_end();
}
+ if (debug_sdfgi_probes) {
+ //debug giprobes
+ bool will_continue_color = (can_continue_color || draw_sky);
+ bool will_continue_depth = (can_continue_depth || draw_sky);
+
+ CameraMatrix dc;
+ 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);
+ _debug_sdfgi_probes(p_render_buffer, draw_list, opaque_framebuffer, cm);
+ RD::get_singleton()->draw_list_end();
+ }
+
if (draw_sky) {
RENDER_TIMESTAMP("Render Sky");
@@ -2437,7 +2486,7 @@ void RasterizerSceneHighEndRD::_render_scene(RID p_render_buffer, const Transfor
if (using_ssr) {
RENDER_TIMESTAMP("Screen Space Reflection");
- _process_ssr(p_render_buffer, render_buffer->color_fb, render_buffer->normal_buffer, render_buffer->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);
+ _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);
} else {
//just mix specular back
RENDER_TIMESTAMP("Merge Specular");
@@ -2447,11 +2496,11 @@ void RasterizerSceneHighEndRD::_render_scene(RID p_render_buffer, const Transfor
RENDER_TIMESTAMP("Render Transparent Pass");
- _setup_environment(p_environment, 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_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);
render_list.sort_by_reverse_depth_and_priority(true);
- _fill_instances(&render_list.elements[render_list.max_elements - render_list.alpha_element_count], render_list.alpha_element_count, false);
+ _fill_instances(&render_list.elements[render_list.max_elements - render_list.alpha_element_count], render_list.alpha_element_count, false, using_sdfgi);
{
RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(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);
@@ -2473,15 +2522,15 @@ void RasterizerSceneHighEndRD::_render_shadow(RID p_framebuffer, InstanceBase **
scene_state.ubo.dual_paraboloid_side = p_use_dp_flip ? -1 : 1;
- _setup_environment(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), RID(), true, Color(), 0, p_zfar, false, p_use_pancake);
render_list.clear();
PassMode pass_mode = p_use_dp ? PASS_MODE_SHADOW_DP : PASS_MODE_SHADOW;
- _fill_render_list(p_cull_result, p_cull_count, pass_mode, true);
+ _fill_render_list(p_cull_result, p_cull_count, pass_mode);
- _setup_view_dependant_uniform_set(RID(), RID());
+ _setup_view_dependant_uniform_set(RID(), RID(), nullptr, 0);
RENDER_TIMESTAMP("Render Shadow");
@@ -2507,14 +2556,14 @@ void RasterizerSceneHighEndRD::_render_material(const Transform &p_cam_transform
scene_state.ubo.dual_paraboloid_side = 0;
scene_state.ubo.material_uv2_mode = true;
- _setup_environment(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), RID(), false, Color(), 0, 0);
render_list.clear();
PassMode pass_mode = PASS_MODE_DEPTH_MATERIAL;
- _fill_render_list(p_cull_result, p_cull_count, pass_mode, true);
+ _fill_render_list(p_cull_result, p_cull_count, pass_mode);
- _setup_view_dependant_uniform_set(RID(), RID());
+ _setup_view_dependant_uniform_set(RID(), RID(), nullptr, 0);
RENDER_TIMESTAMP("Render Material");
@@ -2546,14 +2595,14 @@ void RasterizerSceneHighEndRD::_render_uv2(InstanceBase **p_cull_result, int p_c
scene_state.ubo.dual_paraboloid_side = 0;
scene_state.ubo.material_uv2_mode = true;
- _setup_environment(RID(), CameraMatrix(), Transform(), RID(), true, Vector2(1, 1), RID(), false, Color(), 0, 0);
+ _setup_environment(RID(), RID(), CameraMatrix(), Transform(), RID(), true, Vector2(1, 1), RID(), false, Color(), 0, 0);
render_list.clear();
PassMode pass_mode = PASS_MODE_DEPTH_MATERIAL;
- _fill_render_list(p_cull_result, p_cull_count, pass_mode, true);
+ _fill_render_list(p_cull_result, p_cull_count, pass_mode);
- _setup_view_dependant_uniform_set(RID(), RID());
+ _setup_view_dependant_uniform_set(RID(), RID(), nullptr, 0);
RENDER_TIMESTAMP("Render Material");
@@ -2597,6 +2646,121 @@ void RasterizerSceneHighEndRD::_render_uv2(InstanceBase **p_cull_result, int p_c
}
}
+void RasterizerSceneHighEndRD::_render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, InstanceBase **p_cull_result, int p_cull_count, 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");
+
+ _update_render_base_uniform_set();
+
+ RenderBufferDataHighEnd *render_buffer = (RenderBufferDataHighEnd *)render_buffers_get_data(p_render_buffers);
+ ERR_FAIL_COND(!render_buffer);
+
+ render_pass++;
+ render_list.clear();
+
+ PassMode pass_mode = PASS_MODE_SDF;
+ _fill_render_list(p_cull_result, p_cull_count, pass_mode);
+ render_list.sort_by_key(false);
+ _fill_instances(render_list.elements, render_list.element_count, true);
+
+ _setup_view_dependant_uniform_set(RID(), RID(), nullptr, 0);
+
+ Vector3 half_extents = p_bounds.size * 0.5;
+ Vector3 center = p_bounds.position + half_extents;
+
+ if (render_buffer->render_sdfgi_uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(render_buffer->render_sdfgi_uniform_set)) {
+ Vector<RD::Uniform> uniforms;
+ {
+ RD::Uniform u;
+ u.type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 0;
+ u.ids.push_back(p_albedo_texture);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 1;
+ u.ids.push_back(p_emission_texture);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 2;
+ u.ids.push_back(p_emission_aniso_texture);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 3;
+ u.ids.push_back(p_geom_facing_texture);
+ uniforms.push_back(u);
+ }
+
+ render_buffer->render_sdfgi_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, default_shader_sdfgi_rd, RENDER_BUFFERS_UNIFORM_SET);
+ }
+
+ Vector<RID> sbs;
+ sbs.push_back(p_albedo_texture);
+ sbs.push_back(p_emission_texture);
+ sbs.push_back(p_emission_aniso_texture);
+ sbs.push_back(p_geom_facing_texture);
+
+ //print_line("re-render " + p_from + " - " + p_size + " bounds " + p_bounds);
+ for (int i = 0; i < 3; i++) {
+ scene_state.ubo.sdf_offset[i] = p_from[i];
+ scene_state.ubo.sdf_size[i] = p_size[i];
+ }
+
+ for (int i = 0; i < 3; i++) {
+ Vector3 axis;
+ axis[i] = 1.0;
+ Vector3 up, right;
+ int right_axis = (i + 1) % 3;
+ int up_axis = (i + 2) % 3;
+ up[up_axis] = 1.0;
+ right[right_axis] = 1.0;
+
+ Size2i fb_size;
+ fb_size.x = p_size[right_axis];
+ fb_size.y = p_size[up_axis];
+
+ Transform cam_xform;
+ cam_xform.origin = center + axis * half_extents;
+ cam_xform.basis.set_axis(0, right);
+ cam_xform.basis.set_axis(1, up);
+ cam_xform.basis.set_axis(2, axis);
+
+ //print_line("pass: " + itos(i) + " xform " + cam_xform);
+
+ float h_size = half_extents[right_axis];
+ float v_size = half_extents[up_axis];
+ float d_size = half_extents[i] * 2.0;
+ CameraMatrix camera_proj;
+ camera_proj.set_orthogonal(-h_size, h_size, -v_size, v_size, 0, d_size);
+ //print_line("pass: " + itos(i) + " cam hsize: " + rtos(h_size) + " vsize: " + rtos(v_size) + " dsize " + rtos(d_size));
+
+ Transform to_bounds;
+ to_bounds.origin = p_bounds.position;
+ to_bounds.basis.scale(p_bounds.size);
+
+ 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);
+
+ Map<Size2i, RID>::Element *E = sdfgi_framebuffer_size_cache.find(fb_size);
+ if (!E) {
+ RID fb = RD::get_singleton()->framebuffer_create_empty(fb_size);
+ E = sdfgi_framebuffer_size_cache.insert(fb_size, fb);
+ }
+
+ RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(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);
+ _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(E->get()), render_list.elements, render_list.element_count, true, pass_mode, true, RID(), render_buffer->render_sdfgi_uniform_set, false); //second regular triangles
+ RD::get_singleton()->draw_list_end();
+ }
+}
+
void RasterizerSceneHighEndRD::_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);
@@ -2682,49 +2846,6 @@ void RasterizerSceneHighEndRD::_update_render_base_uniform_set() {
}
{
RD::Uniform u;
- u.binding = 8;
- u.type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
- u.ids.push_back(scene_state.gi_probe_buffer);
- uniforms.push_back(u);
- }
- {
- RD::Uniform u;
- u.binding = 9;
- u.type = RD::UNIFORM_TYPE_TEXTURE;
- int slot_count = gi_probe_get_slots().size();
- if (gi_probe_is_anisotropic()) {
- u.ids.resize(slot_count * 3);
- } else {
- u.ids.resize(slot_count);
- }
-
- for (int i = 0; i < slot_count; i++) {
- RID probe = gi_probe_get_slots()[i];
-
- if (gi_probe_is_anisotropic()) {
- if (probe.is_null()) {
- RID empty_tex = storage->texture_rd_get_default(RasterizerStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE);
- u.ids.write[i * 3 + 0] = empty_tex;
- u.ids.write[i * 3 + 1] = empty_tex;
- u.ids.write[i * 3 + 2] = empty_tex;
- } else {
- u.ids.write[i * 3 + 0] = gi_probe_instance_get_texture(probe);
- u.ids.write[i * 3 + 1] = gi_probe_instance_get_aniso_texture(probe, 0);
- u.ids.write[i * 3 + 2] = gi_probe_instance_get_aniso_texture(probe, 1);
- }
- } else {
- if (probe.is_null()) {
- u.ids.write[i] = storage->texture_rd_get_default(RasterizerStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE);
- } else {
- u.ids.write[i] = gi_probe_instance_get_texture(probe);
- }
- }
- }
-
- uniforms.push_back(u);
- }
- {
- RD::Uniform u;
u.binding = 10;
u.type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
u.ids.push_back(scene_state.lightmap_buffer);
@@ -2803,11 +2924,19 @@ void RasterizerSceneHighEndRD::_update_render_base_uniform_set() {
uniforms.push_back(u);
}
+ {
+ RD::Uniform u;
+ u.type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
+ u.binding = 20;
+ u.ids.push_back(sdfgi_get_ubo());
+ uniforms.push_back(u);
+ }
+
render_base_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, default_shader_rd, SCENE_UNIFORM_SET);
}
}
-void RasterizerSceneHighEndRD::_setup_view_dependant_uniform_set(RID p_shadow_atlas, RID p_reflection_atlas) {
+void RasterizerSceneHighEndRD::_setup_view_dependant_uniform_set(RID p_shadow_atlas, RID p_reflection_atlas, RID *p_gi_probe_cull_result, int p_gi_probe_cull_count) {
if (view_dependant_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(view_dependant_uniform_set)) {
RD::get_singleton()->free(view_dependant_uniform_set);
}
@@ -2844,6 +2973,25 @@ void RasterizerSceneHighEndRD::_setup_view_dependant_uniform_set(RID p_shadow_at
uniforms.push_back(u);
}
+ {
+ RD::Uniform u;
+ u.binding = 2;
+ u.type = RD::UNIFORM_TYPE_TEXTURE;
+ RID default_tex = storage->texture_rd_get_default(RasterizerStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE);
+ for (int i = 0; i < MAX_GI_PROBES; i++) {
+ if (i < p_gi_probe_cull_count) {
+ RID tex = gi_probe_instance_get_texture(p_gi_probe_cull_result[i]);
+ if (!tex.is_valid()) {
+ tex = default_tex;
+ }
+ u.ids.push_back(tex);
+ } else {
+ u.ids.push_back(default_tex);
+ }
+ }
+
+ uniforms.push_back(u);
+ }
view_dependant_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, default_shader_rd, VIEW_DEPENDANT_UNIFORM_SET);
}
@@ -2860,16 +3008,22 @@ void RasterizerSceneHighEndRD::_render_buffers_uniform_set_changed(RID p_render_
_render_buffers_clear_uniform_set(rb);
}
-RID RasterizerSceneHighEndRD::_render_buffers_get_roughness_texture(RID p_render_buffers) {
+RID RasterizerSceneHighEndRD::_render_buffers_get_normal_texture(RID p_render_buffers) {
RenderBufferDataHighEnd *rb = (RenderBufferDataHighEnd *)render_buffers_get_data(p_render_buffers);
- return rb->roughness_buffer;
+ return rb->normal_roughness_buffer;
}
-RID RasterizerSceneHighEndRD::_render_buffers_get_normal_texture(RID p_render_buffers) {
+RID RasterizerSceneHighEndRD::_render_buffers_get_ambient_texture(RID p_render_buffers) {
RenderBufferDataHighEnd *rb = (RenderBufferDataHighEnd *)render_buffers_get_data(p_render_buffers);
- return rb->normal_buffer;
+ return rb->ambient_buffer;
+}
+
+RID RasterizerSceneHighEndRD::_render_buffers_get_reflection_texture(RID p_render_buffers) {
+ RenderBufferDataHighEnd *rb = (RenderBufferDataHighEnd *)render_buffers_get_data(p_render_buffers);
+
+ return rb->reflection_buffer;
}
void RasterizerSceneHighEndRD::_update_render_buffers_uniform_set(RID p_render_buffers) {
@@ -2898,30 +3052,70 @@ void RasterizerSceneHighEndRD::_update_render_buffers_uniform_set(RID p_render_b
RD::Uniform u;
u.binding = 2;
u.type = RD::UNIFORM_TYPE_TEXTURE;
- RID texture = rb->normal_buffer.is_valid() ? rb->normal_buffer : storage->texture_rd_get_default(RasterizerStorageRD::DEFAULT_RD_TEXTURE_NORMAL);
+ RID texture = rb->normal_roughness_buffer.is_valid() ? rb->normal_roughness_buffer : storage->texture_rd_get_default(RasterizerStorageRD::DEFAULT_RD_TEXTURE_NORMAL);
u.ids.push_back(texture);
uniforms.push_back(u);
}
{
RD::Uniform u;
- u.binding = 3;
+ u.binding = 4;
u.type = RD::UNIFORM_TYPE_TEXTURE;
- RID texture = rb->roughness_buffer.is_valid() ? rb->roughness_buffer : storage->texture_rd_get_default(RasterizerStorageRD::DEFAULT_RD_TEXTURE_BLACK);
+ RID aot = render_buffers_get_ao_texture(p_render_buffers);
+ RID texture = aot.is_valid() ? aot : storage->texture_rd_get_default(RasterizerStorageRD::DEFAULT_RD_TEXTURE_BLACK);
u.ids.push_back(texture);
uniforms.push_back(u);
}
{
RD::Uniform u;
- u.binding = 4;
+ u.binding = 5;
u.type = RD::UNIFORM_TYPE_TEXTURE;
- RID aot = render_buffers_get_ao_texture(p_render_buffers);
- RID texture = aot.is_valid() ? aot : storage->texture_rd_get_default(RasterizerStorageRD::DEFAULT_RD_TEXTURE_BLACK);
+ RID texture = rb->ambient_buffer.is_valid() ? rb->ambient_buffer : storage->texture_rd_get_default(RasterizerStorageRD::DEFAULT_RD_TEXTURE_BLACK);
u.ids.push_back(texture);
uniforms.push_back(u);
}
+ {
+ RD::Uniform u;
+ u.binding = 6;
+ u.type = RD::UNIFORM_TYPE_TEXTURE;
+ RID texture = rb->reflection_buffer.is_valid() ? rb->reflection_buffer : storage->texture_rd_get_default(RasterizerStorageRD::DEFAULT_RD_TEXTURE_BLACK);
+ u.ids.push_back(texture);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.binding = 7;
+ u.type = RD::UNIFORM_TYPE_TEXTURE;
+ RID t;
+ if (render_buffers_is_sdfgi_enabled(p_render_buffers)) {
+ t = render_buffers_get_sdfgi_irradiance_probes(p_render_buffers);
+ } else {
+ t = storage->texture_rd_get_default(RasterizerStorageRD::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE);
+ }
+ u.ids.push_back(t);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.binding = 8;
+ u.type = RD::UNIFORM_TYPE_TEXTURE;
+ if (render_buffers_is_sdfgi_enabled(p_render_buffers)) {
+ u.ids.push_back(render_buffers_get_sdfgi_occlusion_texture(p_render_buffers));
+ } else {
+ u.ids.push_back(storage->texture_rd_get_default(RasterizerStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
+ }
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.binding = 9;
+ u.type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
+ u.ids.push_back(render_buffers_get_gi_probe_buffer(p_render_buffers));
+ uniforms.push_back(u);
+ }
+
rb->uniform_set = RD::get_singleton()->uniform_set_create(uniforms, default_shader_rd, RENDER_BUFFERS_UNIFORM_SET);
}
}
@@ -2946,6 +3140,7 @@ RasterizerSceneHighEndRD::RasterizerSceneHighEndRD(RasterizerStorageRD *p_storag
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";
uint32_t uniform_max_size = RD::get_singleton()->limit_get(RD::LIMIT_MAX_UNIFORM_BUFFER_SIZE);
@@ -2977,36 +3172,6 @@ RasterizerSceneHighEndRD::RasterizerSceneHighEndRD(RasterizerStorageRD *p_storag
scene_state.directional_light_buffer = RD::get_singleton()->uniform_buffer_create(directional_light_buffer_size);
defines += "\n#define MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS " + itos(scene_state.max_directional_lights) + "\n";
}
-
- { //giprobes
- int slot_count = gi_probe_get_slots().size();
- if (gi_probe_is_anisotropic()) {
- slot_count *= 3;
- defines += "\n#define GI_PROBE_USE_ANISOTROPY\n";
- }
-
- if (gi_probe_get_quality() == GIPROBE_QUALITY_ULTRA_LOW) {
- defines += "\n#define GI_PROBE_LOW_QUALITY\n";
- } else if (gi_probe_get_quality() == GIPROBE_QUALITY_HIGH) {
- defines += "\n#define GI_PROBE_HIGH_QUALITY\n";
- }
-
- defines += "\n#define MAX_GI_PROBE_TEXTURES " + itos(slot_count) + "\n";
-
- uint32_t giprobe_buffer_size;
- if (uniform_max_size < 65536) {
- //Yes, you guessed right, ARM again
- giprobe_buffer_size = uniform_max_size;
- } else {
- giprobe_buffer_size = 65536;
- }
-
- giprobe_buffer_size = MIN(sizeof(GIProbeData) * gi_probe_get_slots().size(), giprobe_buffer_size);
- scene_state.max_gi_probes = giprobe_buffer_size / sizeof(GIProbeData);
- scene_state.gi_probes = memnew_arr(GIProbeData, scene_state.max_gi_probes);
- scene_state.gi_probe_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(GIProbeData) * scene_state.max_gi_probes);
- defines += "\n#define MAX_GI_PROBES " + itos(scene_state.max_gi_probes) + "\n";
- }
{
//lightmaps
scene_state.max_lightmaps = storage->lightmap_array_get_size();
@@ -3036,13 +3201,13 @@ RasterizerSceneHighEndRD::RasterizerSceneHighEndRD(RasterizerStorageRD *p_storag
Vector<String> shader_versions;
shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n");
shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_DUAL_PARABOLOID\n");
- shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_RENDER_NORMAL\n");
- shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_RENDER_NORMAL\n#define MODE_RENDER_ROUGHNESS\n");
+ shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_RENDER_NORMAL_ROUGHNESS\n");
+ shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_RENDER_NORMAL_ROUGHNESS\n#define MODE_RENDER_GIPROBE\n");
shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_RENDER_MATERIAL\n");
+ shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_RENDER_SDF\n");
shader_versions.push_back("");
+ shader_versions.push_back("\n#define USE_FORWARD_GI\n");
shader_versions.push_back("\n#define MODE_MULTIPLE_RENDER_TARGETS\n");
- shader_versions.push_back("\n#define USE_VOXEL_CONE_TRACING\n");
- shader_versions.push_back("\n#define MODE_MULTIPLE_RENDER_TARGETS\n#define USE_VOXEL_CONE_TRACING\n");
shader_versions.push_back("\n#define USE_LIGHTMAP\n");
shader_versions.push_back("\n#define MODE_MULTIPLE_RENDER_TARGETS\n#define USE_LIGHTMAP\n");
shader.scene_shader.initialize(shader_versions, defines);
@@ -3109,7 +3274,7 @@ RasterizerSceneHighEndRD::RasterizerSceneHighEndRD(RasterizerStorageRD *p_storag
actions.renames["SCREEN_UV"] = "screen_uv";
actions.renames["SCREEN_TEXTURE"] = "color_buffer";
actions.renames["DEPTH_TEXTURE"] = "depth_buffer";
- actions.renames["NORMAL_TEXTURE"] = "normal_buffer";
+ actions.renames["NORMAL_ROUGHNESS_TEXTURE"] = "normal_roughness_buffer";
actions.renames["DEPTH"] = "gl_FragDepth";
actions.renames["OUTPUT_IS_SRGB"] = "true";
@@ -3219,6 +3384,7 @@ RasterizerSceneHighEndRD::RasterizerSceneHighEndRD(RasterizerStorageRD *p_storag
MaterialData *md = (MaterialData *)storage->material_get_data(default_material, RasterizerStorageRD::SHADER_TYPE_3D);
default_shader_rd = shader.scene_shader.version_get_shader(md->shader_data->version, SHADER_VERSION_COLOR_PASS);
+ default_shader_sdfgi_rd = shader.scene_shader.version_get_shader(md->shader_data->version, SHADER_VERSION_DEPTH_PASS_WITH_SDF);
}
{
@@ -3268,7 +3434,7 @@ RasterizerSceneHighEndRD::RasterizerSceneHighEndRD(RasterizerStorageRD *p_storag
{ //render buffers
Vector<RD::Uniform> uniforms;
- for (int i = 0; i < 5; i++) {
+ for (int i = 0; i < 7; i++) {
RD::Uniform u;
u.binding = i;
u.type = RD::UNIFORM_TYPE_TEXTURE;
@@ -3276,6 +3442,28 @@ RasterizerSceneHighEndRD::RasterizerSceneHighEndRD(RasterizerStorageRD *p_storag
u.ids.push_back(texture);
uniforms.push_back(u);
}
+ {
+ RD::Uniform u;
+ u.binding = 7;
+ u.type = RD::UNIFORM_TYPE_TEXTURE;
+ RID texture = storage->texture_rd_get_default(RasterizerStorageRD::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE);
+ u.ids.push_back(texture);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.binding = 8;
+ u.type = RD::UNIFORM_TYPE_TEXTURE;
+ u.ids.push_back(storage->texture_rd_get_default(RasterizerStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.binding = 9;
+ u.type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
+ u.ids.push_back(render_buffers_get_default_gi_probe_buffer());
+ uniforms.push_back(u);
+ }
default_render_buffers_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, default_shader_rd, RENDER_BUFFERS_UNIFORM_SET);
}
@@ -3307,7 +3495,6 @@ RasterizerSceneHighEndRD::~RasterizerSceneHighEndRD() {
{
RD::get_singleton()->free(scene_state.uniform_buffer);
RD::get_singleton()->free(scene_state.instance_buffer);
- RD::get_singleton()->free(scene_state.gi_probe_buffer);
RD::get_singleton()->free(scene_state.directional_light_buffer);
RD::get_singleton()->free(scene_state.light_buffer);
RD::get_singleton()->free(scene_state.lightmap_buffer);
@@ -3315,7 +3502,6 @@ RasterizerSceneHighEndRD::~RasterizerSceneHighEndRD() {
RD::get_singleton()->free(scene_state.reflection_buffer);
RD::get_singleton()->free(scene_state.decal_buffer);
memdelete_arr(scene_state.instances);
- memdelete_arr(scene_state.gi_probes);
memdelete_arr(scene_state.directional_lights);
memdelete_arr(scene_state.lights);
memdelete_arr(scene_state.lightmaps);
@@ -3323,4 +3509,9 @@ RasterizerSceneHighEndRD::~RasterizerSceneHighEndRD() {
memdelete_arr(scene_state.reflections);
memdelete_arr(scene_state.decals);
}
+
+ while (sdfgi_framebuffer_size_cache.front()) {
+ RD::get_singleton()->free(sdfgi_framebuffer_size_cache.front()->get());
+ sdfgi_framebuffer_size_cache.erase(sdfgi_framebuffer_size_cache.front());
+ }
}
diff --git a/servers/rendering/rasterizer_rd/rasterizer_scene_high_end_rd.h b/servers/rendering/rasterizer_rd/rasterizer_scene_high_end_rd.h
index 8438a4f730..cb03da48c1 100644
--- a/servers/rendering/rasterizer_rd/rasterizer_scene_high_end_rd.h
+++ b/servers/rendering/rasterizer_rd/rasterizer_scene_high_end_rd.h
@@ -47,18 +47,23 @@ class RasterizerSceneHighEndRD : public RasterizerSceneRD {
MATERIAL_UNIFORM_SET = 5
};
+ enum {
+ SDFGI_MAX_CASCADES = 8,
+ MAX_GI_PROBES = 8
+ };
+
/* Scene Shader */
enum ShaderVersion {
SHADER_VERSION_DEPTH_PASS,
SHADER_VERSION_DEPTH_PASS_DP,
- SHADER_VERSION_DEPTH_PASS_WITH_NORMAL,
SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS,
+ SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_GIPROBE,
SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL,
+ SHADER_VERSION_DEPTH_PASS_WITH_SDF,
SHADER_VERSION_COLOR_PASS,
+ SHADER_VERSION_COLOR_PASS_WITH_FORWARD_GI,
SHADER_VERSION_COLOR_PASS_WITH_SEPARATE_SPECULAR,
- SHADER_VERSION_VCT_COLOR_PASS,
- SHADER_VERSION_VCT_COLOR_PASS_WITH_SEPARATE_SPECULAR,
SHADER_VERSION_LIGHTMAP_COLOR_PASS,
SHADER_VERSION_LIGHTMAP_COLOR_PASS_WITH_SEPARATE_SPECULAR,
SHADER_VERSION_MAX
@@ -203,8 +208,11 @@ class RasterizerSceneHighEndRD : public RasterizerSceneRD {
RID color;
RID depth;
RID specular;
- RID normal_buffer;
- RID roughness_buffer;
+ RID normal_roughness_buffer;
+ RID giprobe_buffer;
+
+ RID ambient_buffer;
+ RID reflection_buffer;
RS::ViewportMSAA msaa;
RD::TextureSamples texture_samples;
@@ -212,18 +220,22 @@ class RasterizerSceneHighEndRD : public RasterizerSceneRD {
RID color_msaa;
RID depth_msaa;
RID specular_msaa;
- RID normal_buffer_msaa;
+ RID normal_roughness_buffer_msaa;
RID roughness_buffer_msaa;
+ RID giprobe_buffer_msaa;
RID depth_fb;
- RID depth_normal_fb;
RID depth_normal_roughness_fb;
+ RID depth_normal_roughness_giprobe_fb;
RID color_fb;
RID color_specular_fb;
RID specular_only_fb;
int width, height;
+ 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);
@@ -233,8 +245,7 @@ class RasterizerSceneHighEndRD : public RasterizerSceneRD {
};
virtual RenderBufferData *_create_render_buffer_data();
- void _allocate_normal_texture(RenderBufferDataHighEnd *rb);
- void _allocate_roughness_texture(RenderBufferDataHighEnd *rb);
+ void _allocate_normal_roughness_texture(RenderBufferDataHighEnd *rb);
RID shadow_sampler;
RID render_base_uniform_set;
@@ -245,11 +256,12 @@ class RasterizerSceneHighEndRD : public RasterizerSceneRD {
virtual void _base_uniforms_changed();
void _render_buffers_clear_uniform_set(RenderBufferDataHighEnd *rb);
virtual void _render_buffers_uniform_set_changed(RID p_render_buffers);
- virtual RID _render_buffers_get_roughness_texture(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();
- void _setup_view_dependant_uniform_set(RID p_shadow_atlas, RID p_reflection_atlas);
+ void _setup_view_dependant_uniform_set(RID p_shadow_atlas, RID p_reflection_atlas, RID *p_gi_probe_cull_result, int p_gi_probe_cull_count);
void _update_render_buffers_uniform_set(RID p_render_buffers);
/* Scene State UBO */
@@ -260,7 +272,8 @@ class RasterizerSceneHighEndRD : public RasterizerSceneRD {
float box_offset[3];
uint32_t mask;
float params[4]; // intensity, 0, interior , boxproject
- float ambient[4]; // ambient color, energy
+ float ambient[3]; // ambient color,
+ uint32_t ambient_mode;
float local_matrix[16]; // up to here for spot and omni, rest is for directional
};
@@ -315,22 +328,6 @@ class RasterizerSceneHighEndRD : public RasterizerSceneRD {
float uv_scale4[2];
};
- 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 pad[1];
- };
-
struct LightmapData {
float normal_xform[12];
};
@@ -358,6 +355,8 @@ class RasterizerSceneHighEndRD : public RasterizerSceneRD {
};
enum {
+ INSTANCE_DATA_FLAG_USE_GI_BUFFERS = 1 << 6,
+ INSTANCE_DATA_FLAG_USE_SDFGI = 1 << 7,
INSTANCE_DATA_FLAG_USE_LIGHTMAP_CAPTURE = 1 << 8,
INSTANCE_DATA_FLAG_USE_LIGHTMAP = 1 << 9,
INSTANCE_DATA_FLAG_USE_SH_LIGHTMAP = 1 << 10,
@@ -430,10 +429,19 @@ class RasterizerSceneHighEndRD : public RasterizerSceneRD {
float ssao_ao_affect;
uint32_t roughness_limiter_enabled;
+ float roughness_limiter_amount;
+ float roughness_limiter_limit;
+ uint32_t roughness_limiter_pad[2];
+
float ao_color[4];
+ float sdf_to_bounds[16];
+
+ int32_t sdf_offset[3];
uint32_t material_uv2_mode;
- uint32_t pad_material[3];
+
+ int32_t sdf_size[3];
+ uint32_t gi_upscale_for_msaa;
};
UBO ubo;
@@ -445,11 +453,6 @@ class RasterizerSceneHighEndRD : public RasterizerSceneRD {
RID reflection_buffer;
uint32_t max_reflection_probes_per_instance;
- GIProbeData *gi_probes;
- uint32_t max_gi_probes;
- RID gi_probe_buffer;
- uint32_t max_gi_probe_probes_per_instance;
-
LightmapData *lightmaps;
uint32_t max_lightmaps;
RID lightmap_buffer;
@@ -498,7 +501,7 @@ class RasterizerSceneHighEndRD : public RasterizerSceneRD {
uint64_t material_index : 15;
uint64_t shader_index : 12;
uint64_t uses_instancing : 1;
- uint64_t uses_vct : 1;
+ uint64_t uses_forward_gi : 1;
uint64_t uses_lightmap : 1;
uint64_t depth_layer : 4;
uint64_t priority : 8;
@@ -625,6 +628,7 @@ class RasterizerSceneHighEndRD : public RasterizerSceneRD {
RID wireframe_material_shader;
RID wireframe_material;
RID default_shader_rd;
+ RID default_shader_sdfgi_rd;
RID default_radiance_uniform_set;
RID default_render_buffers_uniform_set;
@@ -640,30 +644,33 @@ class RasterizerSceneHighEndRD : public RasterizerSceneRD {
PASS_MODE_SHADOW,
PASS_MODE_SHADOW_DP,
PASS_MODE_DEPTH,
- PASS_MODE_DEPTH_NORMAL,
PASS_MODE_DEPTH_NORMAL_ROUGHNESS,
+ PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE,
PASS_MODE_DEPTH_MATERIAL,
+ PASS_MODE_SDF,
};
- void _setup_environment(RID p_environment, 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 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_lights(RID *p_light_cull_result, int p_light_cull_count, const Transform &p_camera_inverse_transform, RID p_shadow_atlas, bool p_using_shadows);
void _setup_decals(const RID *p_decal_instances, int p_decal_count, const Transform &p_camera_inverse_xform);
void _setup_reflections(RID *p_reflection_probe_cull_result, int p_reflection_probe_cull_count, const Transform &p_camera_inverse_transform, RID p_environment);
- void _setup_gi_probes(RID *p_gi_probe_probe_cull_result, int p_gi_probe_probe_cull_count, const Transform &p_camera_transform);
void _setup_lightmaps(InstanceBase **p_lightmap_cull_result, int p_lightmap_cull_count, const Transform &p_cam_transform);
- void _fill_instances(RenderList::Element **p_elements, int p_element_count, bool p_for_depth);
+ void _fill_instances(RenderList::Element **p_elements, int p_element_count, bool p_for_depth, bool p_has_sdfgi = false, bool p_has_opaque_gi = false);
void _render_list(RenderingDevice::DrawListID p_draw_list, RenderingDevice::FramebufferFormatID p_framebuffer_Format, RenderList::Element **p_elements, int p_element_count, bool p_reverse_cull, PassMode p_pass_mode, bool p_no_gi, RID p_radiance_uniform_set, RID p_render_buffers_uniform_set, bool p_force_wireframe = false, const Vector2 &p_uv_offset = Vector2());
- _FORCE_INLINE_ void _add_geometry(InstanceBase *p_instance, uint32_t p_surface, RID p_material, PassMode p_pass_mode, uint32_t p_geometry_index);
- _FORCE_INLINE_ void _add_geometry_with_material(InstanceBase *p_instance, uint32_t p_surface, MaterialData *p_material, RID p_material_rid, PassMode p_pass_mode, uint32_t p_geometry_index);
+ _FORCE_INLINE_ void _add_geometry(InstanceBase *p_instance, uint32_t p_surface, RID p_material, PassMode p_pass_mode, uint32_t p_geometry_index, bool p_using_sdfgi = false);
+ _FORCE_INLINE_ void _add_geometry_with_material(InstanceBase *p_instance, uint32_t p_surface, MaterialData *p_material, RID p_material_rid, PassMode p_pass_mode, uint32_t p_geometry_index, bool p_using_sdfgi = false);
+
+ void _fill_render_list(InstanceBase **p_cull_result, int p_cull_count, PassMode p_pass_mode, bool p_using_sdfgi = false);
- void _fill_render_list(InstanceBase **p_cull_result, int p_cull_count, PassMode p_pass_mode, bool p_no_gi);
+ Map<Size2i, RID> sdfgi_framebuffer_size_cache;
protected:
virtual void _render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, InstanceBase **p_cull_result, int p_cull_count, RID *p_light_cull_result, int p_light_cull_count, RID *p_reflection_probe_cull_result, int p_reflection_probe_cull_count, RID *p_gi_probe_cull_result, int p_gi_probe_cull_count, RID *p_decal_cull_result, int p_decal_cull_count, InstanceBase **p_lightmap_cull_result, int p_lightmap_cull_count, 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);
virtual void _render_shadow(RID p_framebuffer, InstanceBase **p_cull_result, int p_cull_count, 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);
virtual void _render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, InstanceBase **p_cull_result, int p_cull_count, RID p_framebuffer, const Rect2i &p_region);
virtual void _render_uv2(InstanceBase **p_cull_result, int p_cull_count, 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, InstanceBase **p_cull_result, int p_cull_count, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture);
public:
virtual void set_time(double p_time, double p_step);
diff --git a/servers/rendering/rasterizer_rd/rasterizer_scene_rd.cpp b/servers/rendering/rasterizer_rd/rasterizer_scene_rd.cpp
index 689552be2f..898cb39204 100644
--- a/servers/rendering/rasterizer_rd/rasterizer_scene_rd.cpp
+++ b/servers/rendering/rasterizer_rd/rasterizer_scene_rd.cpp
@@ -32,8 +32,8 @@
#include "core/os/os.h"
#include "core/project_settings.h"
+#include "rasterizer_rd.h"
#include "servers/rendering/rendering_server_raster.h"
-
uint64_t RasterizerSceneRD::auto_exposure_counter = 2;
void get_vogel_disk(float *r_kernel, int p_sample_count) {
@@ -195,6 +195,1541 @@ void RasterizerSceneRD::_update_reflection_mipmaps(ReflectionData &rd) {
}
}
+void RasterizerSceneRD::_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->cascades_ubo);
+
+ memdelete(rb->sdfgi);
+
+ rb->sdfgi = nullptr;
+}
+
+const Vector3i RasterizerSceneRD::SDFGI::Cascade::DIRTY_ALL = Vector3i(0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF);
+
+void RasterizerSceneRD::sdfgi_update(RID p_render_buffers, RID p_environment, const Vector3 &p_world_position) {
+ Environent *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);
+ _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];
+
+ 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);
+ }
+
+ 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.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.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.type = RD::TEXTURE_TYPE_2D_ARRAY;
+ tf_probe_average.array_layers = 1;
+
+ 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);
+ }
+
+ 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.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.type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 2;
+ u.ids.push_back(sdfgi->render_albedo);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.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.type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 4;
+ u.ids.push_back(sdfgi->render_emission);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 5;
+ u.ids.push_back(sdfgi->render_emission_aniso);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 6;
+ u.ids.push_back(sdfgi->render_geom_facing);
+ uniforms.push_back(u);
+ }
+
+ {
+ RD::Uniform u;
+ u.type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 7;
+ u.ids.push_back(cascade.sdf_tex);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 8;
+ u.ids.push_back(sdfgi->occlusion_data);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.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.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.type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 1;
+ u.ids.push_back(sdfgi->render_albedo);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 2;
+ u.ids.push_back(sdfgi->render_geom_facing);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 3;
+ u.ids.push_back(sdfgi->render_emission);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 4;
+ u.ids.push_back(sdfgi->render_emission_aniso);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.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.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.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.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.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(RasterizerStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
+ }
+ }
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.binding = 2;
+ u.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.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.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.type = RD::UNIFORM_TYPE_IMAGE;
+ u.ids.push_back(cascade.light_data);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.binding = 6;
+ u.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.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.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.type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
+ u.ids.push_back(cascade.lights_buffer);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.binding = 10;
+ u.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.type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 1;
+ u.ids.push_back(sdfgi->render_albedo);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.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.type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 1;
+ u.ids.push_back(sdfgi->render_albedo);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.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.type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 1;
+ u.ids.push_back(sdfgi->render_sdf[0]);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.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.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.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.type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 1;
+ u.ids.push_back(sdfgi->render_albedo);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.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.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.type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 1;
+ u.ids.push_back(sdfgi->render_albedo);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.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.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.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(RasterizerStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
+ }
+ }
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.binding = 2;
+ u.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(RasterizerStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
+ }
+ }
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.binding = 3;
+ u.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(RasterizerStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
+ }
+ }
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.binding = 4;
+ u.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(RasterizerStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
+ }
+ }
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.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.type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
+ u.binding = 7;
+ u.ids.push_back(sdfgi->cascades_ubo);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 8;
+ u.ids.push_back(sdfgi->lightprobe_data);
+ uniforms.push_back(u);
+ }
+
+ {
+ RD::Uniform u;
+ u.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.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.type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 11;
+ u.ids.push_back(sdfgi->lightprobe_history_scroll);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 12;
+ u.ids.push_back(sdfgi->lightprobe_average_scroll);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.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 wont be used
+ }
+ u.ids.push_back(parent_average);
+ 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;
+
+ _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;
+ }
+ }
+ }
+}
+
+int RasterizerSceneRD::sdfgi_get_pending_region_count(RID p_render_buffers) const {
+ RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
+
+ ERR_FAIL_COND_V(rb == nullptr, 0);
+
+ if (rb->sdfgi == nullptr) {
+ return 0;
+ }
+
+ 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) {
+ dirty_count++;
+ } else {
+ for (int j = 0; j < 3; j++) {
+ if (c.dirty_regions[j] != 0) {
+ dirty_count++;
+ }
+ }
+ }
+ }
+
+ return dirty_count;
+}
+
+int RasterizerSceneRD::_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 RasterizerSceneRD::sdfgi_get_pending_region_bounds(RID p_render_buffers, int p_region) const {
+ AABB bounds;
+ Vector3i from;
+ Vector3i size;
+
+ int c = _sdfgi_get_pending_region_data(p_render_buffers, p_region, from, size, bounds);
+ ERR_FAIL_COND_V(c == -1, AABB());
+ return bounds;
+}
+
+uint32_t RasterizerSceneRD::sdfgi_get_pending_region_cascade(RID p_render_buffers, int p_region) const {
+ AABB bounds;
+ Vector3i from;
+ Vector3i size;
+
+ return _sdfgi_get_pending_region_data(p_render_buffers, p_region, from, size, bounds);
+}
+
+void RasterizerSceneRD::_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 RasterizerSceneRD::sdfgi_update_probes(RID p_render_buffers, RID p_environment, const RID *p_directional_light_instances, uint32_t p_directional_light_count, 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;
+ }
+ Environent *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 */
+
+ {
+ 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];
+
+ { //fill light buffer
+
+ SDGIShader::Light lights[SDFGI::MAX_DYNAMIC_LIGHTS];
+ uint32_t idx = 0;
+ for (uint32_t j = 0; j < p_directional_light_count; j++) {
+ if (idx == SDFGI::MAX_DYNAMIC_LIGHTS) {
+ break;
+ }
+
+ LightInstance *li = light_instance_owner.getornull(p_directional_light_instances[j]);
+ ERR_CONTINUE(!li);
+ 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);
+ }
+ push_constant.light_count = idx;
+ }
+
+ 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;
+
+ 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.type = RD::UNIFORM_TYPE_TEXTURE;
+ u.binding = 0;
+ u.ids.push_back(sky->radiance);
+ uniforms.push_back(u);
+ }
+
+ {
+ RD::Uniform u;
+ u.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 RasterizerSceneRD::_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, RID *p_gi_probe_cull_result, int p_gi_probe_cull_count) {
+ RENDER_TIMESTAMP("Render GI");
+
+ RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
+ ERR_FAIL_COND(rb == nullptr);
+ Environent *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(RenderBuffers::MAX_GIPROBES, p_gi_probe_cull_count);
+ 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);
+ }
+
+ {
+ 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 < p_gi_probe_cull_count) {
+ GIProbeInstance *gipi = gi_probe_instance_owner.getornull(p_gi_probe_cull_result[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);
+ }
+ }
+
+ if (texture == RID()) {
+ texture = storage->texture_rd_get_default(RasterizerStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE);
+ }
+
+ if (texture != rb->giprobe_textures[i]) {
+ giprobes_changed = true;
+ rb->giprobe_textures[i] = texture;
+ }
+ }
+
+ if (giprobes_changed) {
+ RD::get_singleton()->free(rb->gi_uniform_set);
+ rb->gi_uniform_set = RID();
+ }
+
+ if (p_gi_probe_cull_count > 0) {
+ RD::get_singleton()->buffer_update(gi_probe_buffer, 0, sizeof(GI::GIProbeData) * MIN(RenderBuffers::MAX_GIPROBES, p_gi_probe_cull_count), gi_probe_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.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(RasterizerStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
+ }
+ }
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.binding = 2;
+ u.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(RasterizerStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
+ }
+ }
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.binding = 3;
+ u.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(RasterizerStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
+ }
+ }
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.binding = 4;
+ u.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(RasterizerStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
+ }
+ }
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.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(RasterizerStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
+ }
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.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.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.type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 9;
+ u.ids.push_back(p_ambient_buffer);
+ uniforms.push_back(u);
+ }
+
+ {
+ RD::Uniform u;
+ u.type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 10;
+ u.ids.push_back(p_reflection_buffer);
+ uniforms.push_back(u);
+ }
+
+ {
+ RD::Uniform u;
+ u.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(RasterizerStorageRD::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE));
+ }
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.type = RD::UNIFORM_TYPE_TEXTURE;
+ u.binding = 12;
+ u.ids.push_back(rb->depth_texture);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.type = RD::UNIFORM_TYPE_TEXTURE;
+ u.binding = 13;
+ u.ids.push_back(p_normal_roughness_buffer);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.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(RasterizerStorageRD::DEFAULT_RD_TEXTURE_BLACK);
+ u.ids.push_back(buffer);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
+ u.binding = 15;
+ u.ids.push_back(gi.sdfgi_ubo);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
+ u.binding = 16;
+ u.ids.push_back(rb->giprobe_buffer);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.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);
+ }
+
+ 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();
+}
+
RID RasterizerSceneRD::sky_create() {
return sky_owner.make_rid(Sky());
}
@@ -1291,6 +2826,31 @@ void RasterizerSceneRD::environment_glow_set_use_bicubic_upscale(bool p_enable)
glow_bicubic_upscale = p_enable;
}
+void RasterizerSceneRD::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, bool p_enhance_ssr, float p_energy, float p_normal_bias, float p_probe_bias) {
+ Environent *env = environment_owner.getornull(p_env);
+ ERR_FAIL_COND(!env);
+
+ 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_enhance_ssr = p_enhance_ssr;
+ 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;
+}
+
+void RasterizerSceneRD::environment_set_sdfgi_ray_count(RS::EnvironmentSDFGIRayCount p_ray_count) {
+ sdfgi_ray_count = p_ray_count;
+}
+
+void RasterizerSceneRD::environment_set_sdfgi_frames_to_converge(RS::EnvironmentSDFGIFramesToConverge p_frames) {
+ sdfgi_frames_to_converge = p_frames;
+}
+
void RasterizerSceneRD::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) {
Environent *env = environment_owner.getornull(p_env);
ERR_FAIL_COND(!env);
@@ -1351,6 +2911,11 @@ bool RasterizerSceneRD::environment_is_ssr_enabled(RID p_env) const {
ERR_FAIL_COND_V(!env, false);
return env->ssr_enabled;
}
+bool RasterizerSceneRD::environment_is_sdfgi_enabled(RID p_env) const {
+ Environent *env = environment_owner.getornull(p_env);
+ ERR_FAIL_COND_V(!env, false);
+ return env->sdfgi_enabled;
+}
bool RasterizerSceneRD::is_environment(RID p_env) const {
return environment_owner.owns(p_env);
@@ -2099,6 +3664,13 @@ void RasterizerSceneRD::light_instance_set_transform(RID p_light_instance, const
light_instance->transform = p_transform;
}
+void RasterizerSceneRD::light_instance_set_aabb(RID p_light_instance, const AABB &p_aabb) {
+ LightInstance *light_instance = light_instance_owner.getornull(p_light_instance);
+ ERR_FAIL_COND(!light_instance);
+
+ light_instance->aabb = p_aabb;
+}
+
void RasterizerSceneRD::light_instance_set_shadow_transform(RID p_light_instance, const CameraMatrix &p_projection, const Transform &p_transform, float p_far, float p_split, int p_pass, float p_shadow_texel_size, float p_bias_scale, float p_range_begin, const Vector2 &p_uv_scale) {
LightInstance *light_instance = light_instance_owner.getornull(p_light_instance);
ERR_FAIL_COND(!light_instance);
@@ -2193,23 +3765,9 @@ void RasterizerSceneRD::decal_instance_set_transform(RID p_decal, const Transfor
/////////////////////////////////
RID RasterizerSceneRD::gi_probe_instance_create(RID p_base) {
- //find a free slot
- int index = -1;
- for (int i = 0; i < gi_probe_slots.size(); i++) {
- if (gi_probe_slots[i] == RID()) {
- index = i;
- break;
- }
- }
-
- ERR_FAIL_COND_V(index == -1, RID());
-
GIProbeInstance gi_probe;
- gi_probe.slot = index;
gi_probe.probe = p_base;
RID rid = gi_probe_instance_owner.make_rid(gi_probe);
- gi_probe_slots.write[index] = rid;
-
return rid;
}
@@ -2240,10 +3798,6 @@ void RasterizerSceneRD::gi_probe_update(RID p_probe, bool p_update_light_instanc
//need to re-create everything
if (gi_probe->texture.is_valid()) {
RD::get_singleton()->free(gi_probe->texture);
- if (gi_probe_use_anisotropy) {
- RD::get_singleton()->free(gi_probe->anisotropy_r16[0]);
- RD::get_singleton()->free(gi_probe->anisotropy_r16[1]);
- }
RD::get_singleton()->free(gi_probe->write_buffer);
gi_probe->mipmaps.clear();
}
@@ -2275,47 +3829,18 @@ void RasterizerSceneRD::gi_probe_update(RID p_probe, bool p_update_light_instanc
RD::get_singleton()->texture_clear(gi_probe->texture, Color(0, 0, 0, 0), 0, levels.size(), 0, 1, false);
- if (gi_probe_use_anisotropy) {
- tf.format = RD::DATA_FORMAT_R16_UINT;
- tf.shareable_formats.push_back(RD::DATA_FORMAT_R16_UINT);
- tf.shareable_formats.push_back(RD::DATA_FORMAT_R5G6B5_UNORM_PACK16);
-
- //need to create R16 first, else driver does not like the storage bit for compute..
- gi_probe->anisotropy_r16[0] = RD::get_singleton()->texture_create(tf, RD::TextureView());
- gi_probe->anisotropy_r16[1] = RD::get_singleton()->texture_create(tf, RD::TextureView());
-
- RD::TextureView tv;
- tv.format_override = RD::DATA_FORMAT_R5G6B5_UNORM_PACK16;
- gi_probe->anisotropy[0] = RD::get_singleton()->texture_create_shared(tv, gi_probe->anisotropy_r16[0]);
- gi_probe->anisotropy[1] = RD::get_singleton()->texture_create_shared(tv, gi_probe->anisotropy_r16[1]);
-
- RD::get_singleton()->texture_clear(gi_probe->anisotropy[0], Color(0, 0, 0, 0), 0, levels.size(), 0, 1, false);
- RD::get_singleton()->texture_clear(gi_probe->anisotropy[1], 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];
}
- if (gi_probe_use_anisotropy) {
- total_elements *= 6;
- }
-
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);
- if (gi_probe_use_anisotropy) {
- RD::TextureView tv;
- tv.format_override = RD::DATA_FORMAT_R16_UINT;
- mipmap.anisotropy[0] = RD::get_singleton()->texture_create_shared_from_slice(tv, gi_probe->anisotropy[0], 0, i, RD::TEXTURE_SLICE_3D);
- mipmap.anisotropy[1] = RD::get_singleton()->texture_create_shared_from_slice(tv, gi_probe->anisotropy[1], 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++) {
@@ -2383,24 +3908,6 @@ void RasterizerSceneRD::gi_probe_update(RID p_probe, bool p_update_light_instanc
u.ids.push_back(gi_probe->texture);
copy_uniforms.push_back(u);
}
-
- if (gi_probe_use_anisotropy) {
- {
- RD::Uniform u;
- u.type = RD::UNIFORM_TYPE_TEXTURE;
- u.binding = 7;
- u.ids.push_back(gi_probe->anisotropy[0]);
- copy_uniforms.push_back(u);
- }
- {
- RD::Uniform u;
- u.type = RD::UNIFORM_TYPE_TEXTURE;
- u.binding = 8;
- u.ids.push_back(gi_probe->anisotropy[1]);
- 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);
@@ -2415,23 +3922,6 @@ void RasterizerSceneRD::gi_probe_update(RID p_probe, bool p_update_light_instanc
uniforms.push_back(u);
}
- if (gi_probe_use_anisotropy) {
- {
- RD::Uniform u;
- u.type = RD::UNIFORM_TYPE_IMAGE;
- u.binding = 6;
- u.ids.push_back(mipmap.anisotropy[0]);
- uniforms.push_back(u);
- }
- {
- RD::Uniform u;
- u.type = RD::UNIFORM_TYPE_IMAGE;
- u.binding = 7;
- u.ids.push_back(mipmap.anisotropy[1]);
- 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);
@@ -2626,22 +4116,6 @@ void RasterizerSceneRD::gi_probe_update(RID p_probe, bool p_update_light_instanc
u.ids.push_back(gi_probe->mipmaps[dmap.mipmap].texture);
uniforms.push_back(u);
}
- if (gi_probe_is_anisotropic()) {
- {
- RD::Uniform u;
- u.type = RD::UNIFORM_TYPE_IMAGE;
- u.binding = 12;
- u.ids.push_back(gi_probe->mipmaps[dmap.mipmap].anisotropy[0]);
- uniforms.push_back(u);
- }
- {
- RD::Uniform u;
- u.type = RD::UNIFORM_TYPE_IMAGE;
- u.binding = 13;
- u.ids.push_back(gi_probe->mipmaps[dmap.mipmap].anisotropy[1]);
- 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);
@@ -2663,10 +4137,6 @@ void RasterizerSceneRD::gi_probe_update(RID p_probe, bool p_update_light_instanc
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);
- if (gi_probe_is_anisotropic()) {
- RD::get_singleton()->texture_clear(gi_probe->anisotropy[0], Color(0, 0, 0, 0), 0, gi_probe->mipmaps.size(), 0, 1, true);
- RD::get_singleton()->texture_clear(gi_probe->anisotropy[1], Color(0, 0, 0, 0), 0, gi_probe->mipmaps.size(), 0, 1, true);
- }
}
uint32_t light_count = 0;
@@ -2733,7 +4203,7 @@ void RasterizerSceneRD::gi_probe_update(RID p_probe, bool p_update_light_instanc
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 = storage->gi_probe_get_anisotropy_strength(gi_probe->probe);
+ push_constant.aniso_strength = 0;
/* print_line("probe update to version " + itos(gi_probe->last_probe_version));
print_line("propagation " + rtos(push_constant.propagation));
@@ -3067,23 +4537,6 @@ void RasterizerSceneRD::_debug_giprobe(RID p_gi_probe, RD::DrawListID p_draw_lis
uniforms.push_back(u);
}
- if (gi_probe_use_anisotropy) {
- {
- RD::Uniform u;
- u.type = RD::UNIFORM_TYPE_TEXTURE;
- u.binding = 4;
- u.ids.push_back(gi_probe->anisotropy[0]);
- uniforms.push_back(u);
- }
- {
- RD::Uniform u;
- u.type = RD::UNIFORM_TYPE_TEXTURE;
- u.binding = 5;
- u.ids.push_back(gi_probe->anisotropy[1]);
- 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];
@@ -3098,12 +4551,140 @@ void RasterizerSceneRD::_debug_giprobe(RID p_gi_probe, RD::DrawListID p_draw_lis
RD::get_singleton()->draw_list_draw(p_draw_list, false, cell_count, 36);
}
-const Vector<RID> &RasterizerSceneRD::gi_probe_get_slots() const {
- return gi_probe_slots;
-}
+void RasterizerSceneRD::_debug_sdfgi_probes(RID p_render_buffers, RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform) {
+ RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
+ ERR_FAIL_COND(!rb);
+
+ if (!rb->sdfgi) {
+ 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.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.type = RD::UNIFORM_TYPE_TEXTURE;
+ u.ids.push_back(rb->sdfgi->lightprobe_texture);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.binding = 3;
+ u.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.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;
-RasterizerSceneRD::GIProbeQuality RasterizerSceneRD::gi_probe_get_quality() const {
- return gi_probe_quality;
+ 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);
+ }
}
////////////////////////////////
@@ -3273,7 +4854,7 @@ void RasterizerSceneRD::_process_sss(RID p_render_buffers, const CameraMatrix &p
storage->get_effects()->sub_surface_scattering(rb->texture, rb->blur[0].mipmaps[0].texture, rb->depth_texture, p_camera, Size2i(rb->width, rb->height), sss_scale, sss_depth_scale, sss_quality);
}
-void RasterizerSceneRD::_process_ssr(RID p_render_buffers, RID p_dest_framebuffer, RID p_normal_buffer, RID p_roughness_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 RasterizerSceneRD::_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) {
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND(!rb);
@@ -3322,7 +4903,7 @@ void RasterizerSceneRD::_process_ssr(RID p_render_buffers, RID p_dest_framebuffe
_render_buffers_uniform_set_changed(p_render_buffers);
}
- storage->get_effects()->screen_space_reflection(rb->texture, p_normal_buffer, ssr_roughness_quality, p_roughness_buffer, rb->ssr.blur_radius[0], rb->ssr.blur_radius[1], p_metallic, p_metallic_mask, rb->depth_texture, rb->ssr.depth_scaled, rb->ssr.normal_scaled, rb->blur[0].mipmaps[1].texture, rb->blur[1].mipmaps[0].texture, Size2i(rb->width / 2, rb->height / 2), env->ssr_max_steps, env->ssr_fade_in, env->ssr_fade_out, env->ssr_depth_tolerance, p_projection);
+ storage->get_effects()->screen_space_reflection(rb->texture, p_normal_buffer, ssr_roughness_quality, rb->ssr.blur_radius[0], rb->ssr.blur_radius[1], p_metallic, p_metallic_mask, rb->depth_texture, rb->ssr.depth_scaled, rb->ssr.normal_scaled, rb->blur[0].mipmaps[1].texture, rb->blur[1].mipmaps[0].texture, Size2i(rb->width / 2, rb->height / 2), env->ssr_max_steps, env->ssr_fade_in, env->ssr_fade_out, env->ssr_depth_tolerance, p_projection);
storage->get_effects()->merge_specular(p_dest_framebuffer, p_specular_buffer, p_use_additive ? RID() : rb->texture, rb->blur[0].mipmaps[1].texture);
}
@@ -3563,17 +5144,165 @@ void RasterizerSceneRD::_render_buffers_debug_draw(RID p_render_buffers, RID p_s
effects->copy_to_fb_rect(ao_buf, storage->render_target_get_rd_framebuffer(rb->render_target), Rect2(Vector2(), rtsize), false, true);
}
- if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_ROUGHNESS_LIMITER && _render_buffers_get_roughness_texture(p_render_buffers).is_valid()) {
+ if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_NORMAL_BUFFER && _render_buffers_get_normal_texture(p_render_buffers).is_valid()) {
Size2 rtsize = storage->render_target_get_size(rb->render_target);
- effects->copy_to_fb_rect(_render_buffers_get_roughness_texture(p_render_buffers), storage->render_target_get_rd_framebuffer(rb->render_target), Rect2(Vector2(), rtsize), false, true);
+ 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_NORMAL_BUFFER && _render_buffers_get_normal_texture(p_render_buffers).is_valid()) {
+ if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_GI_BUFFER && _render_buffers_get_ambient_texture(p_render_buffers).is_valid()) {
Size2 rtsize = storage->render_target_get_size(rb->render_target);
- 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);
+ RID ambient_texture = _render_buffers_get_ambient_texture(p_render_buffers);
+ RID reflection_texture = _render_buffers_get_reflection_texture(p_render_buffers);
+ 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 RasterizerSceneRD::_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.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(RasterizerStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
+ }
+ }
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.binding = 2;
+ u.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(RasterizerStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
+ }
+ }
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.binding = 3;
+ u.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(RasterizerStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
+ }
+ }
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.binding = 4;
+ u.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(RasterizerStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
+ }
+ }
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.binding = 5;
+ u.type = RD::UNIFORM_TYPE_TEXTURE;
+ u.ids.push_back(rb->sdfgi->occlusion_texture);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.binding = 8;
+ u.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.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.type = RD::UNIFORM_TYPE_IMAGE;
+ u.ids.push_back(rb->texture);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.binding = 11;
+ u.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 RasterizerSceneRD::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());
@@ -3590,6 +5319,113 @@ RID RasterizerSceneRD::render_buffers_get_ao_texture(RID p_render_buffers) {
return rb->ssao.ao_full.is_valid() ? rb->ssao.ao_full : rb->ssao.ao[0];
}
+RID RasterizerSceneRD::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);
+ }
+ return rb->giprobe_buffer;
+}
+
+RID RasterizerSceneRD::render_buffers_get_default_gi_probe_buffer() {
+ return default_giprobe_buffer;
+}
+
+uint32_t RasterizerSceneRD::render_buffers_get_sdfgi_cascade_count(RID p_render_buffers) const {
+ const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
+ ERR_FAIL_COND_V(!rb, 0);
+ ERR_FAIL_COND_V(!rb->sdfgi, 0);
+
+ return rb->sdfgi->cascades.size();
+}
+bool RasterizerSceneRD::render_buffers_is_sdfgi_enabled(RID p_render_buffers) const {
+ const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
+ ERR_FAIL_COND_V(!rb, false);
+
+ return rb->sdfgi != nullptr;
+}
+RID RasterizerSceneRD::render_buffers_get_sdfgi_irradiance_probes(RID p_render_buffers) const {
+ const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
+ ERR_FAIL_COND_V(!rb, RID());
+ ERR_FAIL_COND_V(!rb->sdfgi, RID());
+
+ return rb->sdfgi->lightprobe_texture;
+}
+
+Vector3 RasterizerSceneRD::render_buffers_get_sdfgi_cascade_offset(RID p_render_buffers, uint32_t p_cascade) const {
+ const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
+ ERR_FAIL_COND_V(!rb, Vector3());
+ ERR_FAIL_COND_V(!rb->sdfgi, Vector3());
+ ERR_FAIL_UNSIGNED_INDEX_V(p_cascade, rb->sdfgi->cascades.size(), Vector3());
+
+ return Vector3((Vector3i(1, 1, 1) * -int32_t(rb->sdfgi->cascade_size >> 1) + rb->sdfgi->cascades[p_cascade].position)) * rb->sdfgi->cascades[p_cascade].cell_size;
+}
+
+Vector3i RasterizerSceneRD::render_buffers_get_sdfgi_cascade_probe_offset(RID p_render_buffers, uint32_t p_cascade) const {
+ const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
+ 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;
+
+ return rb->sdfgi->cascades[p_cascade].position / probe_divisor;
+}
+
+float RasterizerSceneRD::render_buffers_get_sdfgi_normal_bias(RID p_render_buffers) const {
+ const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
+ ERR_FAIL_COND_V(!rb, 0);
+ ERR_FAIL_COND_V(!rb->sdfgi, 0);
+
+ return rb->sdfgi->normal_bias;
+}
+float RasterizerSceneRD::render_buffers_get_sdfgi_cascade_probe_size(RID p_render_buffers, uint32_t p_cascade) const {
+ const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
+ ERR_FAIL_COND_V(!rb, 0);
+ ERR_FAIL_COND_V(!rb->sdfgi, 0);
+ ERR_FAIL_UNSIGNED_INDEX_V(p_cascade, rb->sdfgi->cascades.size(), 0);
+
+ return float(rb->sdfgi->cascade_size) * rb->sdfgi->cascades[p_cascade].cell_size / float(rb->sdfgi->probe_axis_count - 1);
+}
+uint32_t RasterizerSceneRD::render_buffers_get_sdfgi_cascade_probe_count(RID p_render_buffers) const {
+ const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
+ ERR_FAIL_COND_V(!rb, 0);
+ ERR_FAIL_COND_V(!rb->sdfgi, 0);
+
+ return rb->sdfgi->probe_axis_count;
+}
+
+uint32_t RasterizerSceneRD::render_buffers_get_sdfgi_cascade_size(RID p_render_buffers) const {
+ const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
+ ERR_FAIL_COND_V(!rb, 0);
+ ERR_FAIL_COND_V(!rb->sdfgi, 0);
+
+ return rb->sdfgi->cascade_size;
+}
+
+bool RasterizerSceneRD::render_buffers_is_sdfgi_using_occlusion(RID p_render_buffers) const {
+ const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
+ ERR_FAIL_COND_V(!rb, false);
+ ERR_FAIL_COND_V(!rb->sdfgi, false);
+
+ return rb->sdfgi->uses_occlusion;
+}
+
+float RasterizerSceneRD::render_buffers_get_sdfgi_energy(RID p_render_buffers) const {
+ const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
+ ERR_FAIL_COND_V(!rb, 0);
+ ERR_FAIL_COND_V(!rb->sdfgi, false);
+
+ return rb->sdfgi->energy;
+}
+RID RasterizerSceneRD::render_buffers_get_sdfgi_occlusion_texture(RID p_render_buffers) const {
+ const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
+ ERR_FAIL_COND_V(!rb, RID());
+ ERR_FAIL_COND_V(!rb->sdfgi, RID());
+
+ return rb->sdfgi->occlusion_texture;
+}
+
void RasterizerSceneRD::render_buffers_configure(RID p_render_buffers, RID p_render_target, int p_width, int p_height, RS::ViewportMSAA p_msaa, RenderingServer::ViewportScreenSpaceAA p_screen_space_aa) {
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
rb->width = p_width;
@@ -3606,7 +5442,7 @@ void RasterizerSceneRD::render_buffers_configure(RID p_render_buffers, RID p_ren
tf.height = rb->height;
tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT;
if (rb->msaa != RS::VIEWPORT_MSAA_DISABLED) {
- tf.usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT;
+ tf.usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_STORAGE_BIT;
} else {
tf.usage_bits |= RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT;
}
@@ -3616,12 +5452,20 @@ void RasterizerSceneRD::render_buffers_configure(RID p_render_buffers, RID p_ren
{
RD::TextureFormat tf;
- tf.format = RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_D24_UNORM_S8_UINT, RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) ? RD::DATA_FORMAT_D24_UNORM_S8_UINT : RD::DATA_FORMAT_D32_SFLOAT_S8_UINT;
+ if (rb->msaa == RS::VIEWPORT_MSAA_DISABLED) {
+ tf.format = RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_D24_UNORM_S8_UINT, RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) ? RD::DATA_FORMAT_D24_UNORM_S8_UINT : RD::DATA_FORMAT_D32_SFLOAT_S8_UINT;
+ } else {
+ tf.format = RD::DATA_FORMAT_R32_SFLOAT;
+ }
+
tf.width = p_width;
tf.height = p_height;
- tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT;
+ tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT;
+
if (rb->msaa != RS::VIEWPORT_MSAA_DISABLED) {
- tf.usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT;
+ tf.usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_STORAGE_BIT;
+ } else {
+ tf.usage_bits |= RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT;
}
rb->depth_texture = RD::get_singleton()->texture_create(tf, RD::TextureView());
@@ -3748,6 +5592,14 @@ void RasterizerSceneRD::render_scene(RID p_render_buffers, const Transform &p_ca
clear_color = storage->get_default_clear_color();
}
+ //assign render indices to giprobes
+ for (int i = 0; i < p_gi_probe_cull_count; i++) {
+ GIProbeInstance *giprobe_inst = gi_probe_instance_owner.getornull(p_gi_probe_cull_result[i]);
+ if (giprobe_inst) {
+ giprobe_inst->render_index = i;
+ }
+ }
+
_render_scene(p_render_buffers, p_cam_transform, p_cam_projection, p_cam_ortogonal, p_cull_result, p_cull_count, p_light_cull_result, p_light_cull_count, p_reflection_probe_cull_result, p_reflection_probe_cull_count, p_gi_probe_cull_result, p_gi_probe_cull_count, p_decal_cull_result, p_decal_cull_count, p_lightmap_cull_result, p_lightmap_cull_count, p_environment, p_camera_effects, p_shadow_atlas, p_reflection_atlas, p_reflection_probe, p_reflection_probe_pass, clear_color);
if (p_render_buffers.is_valid()) {
@@ -3755,6 +5607,9 @@ void RasterizerSceneRD::render_scene(RID p_render_buffers, const Transform &p_ca
_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);
+ }
}
}
@@ -3938,11 +5793,455 @@ void RasterizerSceneRD::render_material(const Transform &p_cam_transform, const
_render_material(p_cam_transform, p_cam_projection, p_cam_ortogonal, p_cull_result, p_cull_count, p_framebuffer, p_region);
}
+void RasterizerSceneRD::render_sdfgi(RID p_render_buffers, int p_region, InstanceBase **p_cull_result, int p_cull_count) {
+ //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_cull_result, p_cull_count, 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;
+
+ RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin();
+ if (rb->sdfgi->cascades[cascade].dirty_regions != SDFGI::Cascade::DIRTY_ALL) {
+ //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.image_size[0] = rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count;
+ ipush_constant.image_size[1] = 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_add_barrier(compute_list);
+ }
+
+ //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);
+
+ bool half_size = true; //much faster, very little differnce
+ 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 offseted, its 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 RasterizerSceneRD::render_sdfgi_static_lights(RID p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const RID **p_positional_light_cull_result, const uint32_t *p_positional_light_cull_count) {
+ RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
+ ERR_FAIL_COND(!rb);
+ ERR_FAIL_COND(!rb->sdfgi);
+
+ ERR_FAIL_COND(p_positional_light_cull_count == 0);
+
+ _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 < p_positional_light_cull_count[i]; 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 RasterizerSceneRD::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);
+ }
render_buffers_owner.free(p_rid);
} else if (environment_owner.owns(p_rid)) {
//not much to delete, just free it
@@ -3966,18 +6265,12 @@ bool RasterizerSceneRD::free(RID p_rid) {
RD::get_singleton()->free(gi_probe->texture);
RD::get_singleton()->free(gi_probe->write_buffer);
}
- if (gi_probe->anisotropy[0].is_valid()) {
- RD::get_singleton()->free(gi_probe->anisotropy[0]);
- RD::get_singleton()->free(gi_probe->anisotropy[1]);
- }
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_slots.write[gi_probe->slot] = RID();
-
gi_probe_instance_owner.free(p_rid);
} else if (sky_owner.owns(p_rid)) {
_update_dirty_skys();
@@ -4050,17 +6343,22 @@ void RasterizerSceneRD::set_time(double p_time, double p_step) {
time_step = p_step;
}
-void RasterizerSceneRD::screen_space_roughness_limiter_set_active(bool p_enable, float p_curve) {
+void RasterizerSceneRD::screen_space_roughness_limiter_set_active(bool p_enable, float p_amount, float p_limit) {
screen_space_roughness_limiter = p_enable;
- screen_space_roughness_limiter_curve = p_curve;
+ screen_space_roughness_limiter_amount = p_amount;
+ screen_space_roughness_limiter_limit = p_limit;
}
bool RasterizerSceneRD::screen_space_roughness_limiter_is_active() const {
return screen_space_roughness_limiter;
}
-float RasterizerSceneRD::screen_space_roughness_limiter_get_curve() const {
- return screen_space_roughness_limiter_curve;
+float RasterizerSceneRD::screen_space_roughness_limiter_get_amount() const {
+ return screen_space_roughness_limiter_amount;
+}
+
+float RasterizerSceneRD::screen_space_roughness_limiter_get_limit() const {
+ return screen_space_roughness_limiter_limit;
}
TypedArray<Image> RasterizerSceneRD::bake_render_uv2(RID p_base, const Vector<RID> &p_material_overrides, const Size2i &p_image_size) {
@@ -4154,6 +6452,11 @@ TypedArray<Image> RasterizerSceneRD::bake_render_uv2(RID p_base, const Vector<RI
return ret;
}
+void RasterizerSceneRD::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;
+}
+
RasterizerSceneRD *RasterizerSceneRD::singleton = nullptr;
RasterizerSceneRD::RasterizerSceneRD(RasterizerStorageRD *p_storage) {
@@ -4165,7 +6468,7 @@ RasterizerSceneRD::RasterizerSceneRD(RasterizerStorageRD *p_storage) {
sky_use_cubemap_array = GLOBAL_GET("rendering/quality/reflections/texture_array_reflections");
// sky_use_cubemap_array = false;
- uint32_t textures_per_stage = RD::get_singleton()->limit_get(RD::LIMIT_MAX_TEXTURES_PER_SHADER_STAGE);
+ //uint32_t textures_per_stage = RD::get_singleton()->limit_get(RD::LIMIT_MAX_TEXTURES_PER_SHADER_STAGE);
{
//kinda complicated to compute the amount of slots, we try to use as many as we can
@@ -4174,34 +6477,9 @@ RasterizerSceneRD::RasterizerSceneRD(RasterizerStorageRD *p_storage) {
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_use_anisotropy = GLOBAL_GET("rendering/quality/gi_probes/anisotropic");
- gi_probe_quality = GIProbeQuality(CLAMP(int(GLOBAL_GET("rendering/quality/gi_probes/quality")), 0, 2));
-
- if (textures_per_stage <= 16) {
- gi_probe_slots.resize(2); //thats all you can get
- gi_probe_use_anisotropy = false;
- } else if (textures_per_stage <= 31) {
- gi_probe_slots.resize(4); //thats all you can get, iOS
- gi_probe_use_anisotropy = false;
- } else if (textures_per_stage <= 128) {
- gi_probe_slots.resize(32); //old intel
- gi_probe_use_anisotropy = false;
- } else if (textures_per_stage <= 256) {
- gi_probe_slots.resize(64); //old intel too
- gi_probe_use_anisotropy = false;
- } else {
- if (gi_probe_use_anisotropy) {
- gi_probe_slots.resize(1024 / 3); //needs 3 textures
- } else {
- gi_probe_slots.resize(1024); //modern intel, nvidia, 8192 or greater
- }
- }
+ 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";
- if (gi_probe_use_anisotropy) {
- defines += "\n#define MODE_ANISOTROPIC\n";
- }
Vector<String> versions;
versions.push_back("\n#define MODE_COMPUTE_LIGHT\n");
@@ -4223,9 +6501,6 @@ RasterizerSceneRD::RasterizerSceneRD(RasterizerStorageRD *p_storage) {
{
String defines;
- if (gi_probe_use_anisotropy) {
- defines += "\n#define USE_ANISOTROPY\n";
- }
Vector<String> versions;
versions.push_back("\n#define MODE_DEBUG_COLOR\n");
versions.push_back("\n#define MODE_DEBUG_LIGHT\n");
@@ -4373,11 +6648,131 @@ RasterizerSceneRD::RasterizerSceneRD(RasterizerStorageRD *p_storage) {
sky_scene_state.sampler_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sky_shader.default_shader_rd, SKY_SET_SAMPLERS);
}
+ {
+ 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.type = RD::UNIFORM_TYPE_TEXTURE;
+ u.binding = 0;
+ u.ids.push_back(storage->texture_rd_get_default(RasterizerStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.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);
+
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"));
- screen_space_roughness_limiter = GLOBAL_GET("rendering/quality/screen_filters/screen_space_roughness_limiter");
- screen_space_roughness_limiter_curve = GLOBAL_GET("rendering/quality/screen_filters/screen_space_roughness_limiter_curve");
+ 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;
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")));
@@ -4406,6 +6801,8 @@ RasterizerSceneRD::~RasterizerSceneRD() {
RD::get_singleton()->free(sky_scene_state.light_uniform_set);
}
+ RD::get_singleton()->free(default_giprobe_buffer);
+
RD::get_singleton()->free(gi_probe_lights_uniform);
giprobe_debug_shader.version_free(giprobe_debug_shader_version);
giprobe_shader.version_free(giprobe_lighting_shader_version);
diff --git a/servers/rendering/rasterizer_rd/rasterizer_scene_rd.h b/servers/rendering/rasterizer_rd/rasterizer_scene_rd.h
index 781dbd50cc..88c2f5a5e6 100644
--- a/servers/rendering/rasterizer_rd/rasterizer_scene_rd.h
+++ b/servers/rendering/rasterizer_rd/rasterizer_scene_rd.h
@@ -31,22 +31,22 @@
#ifndef RASTERIZER_SCENE_RD_H
#define RASTERIZER_SCENE_RD_H
+#include "core/local_vector.h"
#include "core/rid_owner.h"
#include "servers/rendering/rasterizer.h"
#include "servers/rendering/rasterizer_rd/rasterizer_storage_rd.h"
+#include "servers/rendering/rasterizer_rd/shaders/gi.glsl.gen.h"
#include "servers/rendering/rasterizer_rd/shaders/giprobe.glsl.gen.h"
#include "servers/rendering/rasterizer_rd/shaders/giprobe_debug.glsl.gen.h"
+#include "servers/rendering/rasterizer_rd/shaders/sdfgi_debug.glsl.gen.h"
+#include "servers/rendering/rasterizer_rd/shaders/sdfgi_debug_probes.glsl.gen.h"
+#include "servers/rendering/rasterizer_rd/shaders/sdfgi_direct_light.glsl.gen.h"
+#include "servers/rendering/rasterizer_rd/shaders/sdfgi_integrate.glsl.gen.h"
+#include "servers/rendering/rasterizer_rd/shaders/sdfgi_preprocess.glsl.gen.h"
#include "servers/rendering/rasterizer_rd/shaders/sky.glsl.gen.h"
#include "servers/rendering/rendering_device.h"
class RasterizerSceneRD : public RasterizerScene {
-public:
- enum GIProbeQuality {
- GIPROBE_QUALITY_ULTRA_LOW,
- GIPROBE_QUALITY_MEDIUM,
- GIPROBE_QUALITY_HIGH,
- };
-
protected:
double time;
@@ -81,23 +81,27 @@ protected:
virtual void _render_shadow(RID p_framebuffer, InstanceBase **p_cull_result, int p_cull_count, 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) = 0;
virtual void _render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, InstanceBase **p_cull_result, int p_cull_count, RID p_framebuffer, const Rect2i &p_region) = 0;
virtual void _render_uv2(InstanceBase **p_cull_result, int p_cull_count, 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, InstanceBase **p_cull_result, int p_cull_count, 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 _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);
virtual void _base_uniforms_changed() = 0;
virtual void _render_buffers_uniform_set_changed(RID p_render_buffers) = 0;
- virtual RID _render_buffers_get_roughness_texture(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_roughness_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_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, const Vector3 &p_position, 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, RID *p_gi_probe_cull_result, int p_gi_probe_cull_count);
private:
RS::ViewportDebugDraw debug_draw = RS::VIEWPORT_DEBUG_DRAW_DISABLED;
@@ -267,6 +271,8 @@ private:
SkyMaterialData *prev_material;
Vector3 prev_position;
float prev_time;
+
+ RID sdfgi_integrate_sky_uniform_set;
};
Sky *dirty_sky_list = nullptr;
@@ -388,13 +394,10 @@ private:
struct GIProbeInstance {
RID probe;
RID texture;
- RID anisotropy[2]; //only if anisotropy is used
- RID anisotropy_r16[2]; //only if anisotropy is used
RID write_buffer;
struct Mipmap {
RID texture;
- RID anisotropy[2]; //only if anisotropy is used
RID uniform_set;
RID second_bounce_uniform_set;
RID write_uniform_set;
@@ -423,7 +426,7 @@ private:
uint32_t last_probe_version = 0;
uint32_t last_probe_data_version = 0;
- uint64_t last_pass = 0;
+ //uint64_t last_pass = 0;
uint32_t render_index = 0;
bool has_dynamic_object_data = false;
@@ -435,11 +438,6 @@ private:
uint32_t gi_probe_max_lights;
RID gi_probe_lights_uniform;
- bool gi_probe_use_anisotropy = false;
- GIProbeQuality gi_probe_quality = GIPROBE_QUALITY_MEDIUM;
-
- Vector<RID> gi_probe_slots;
-
enum {
GI_PROBE_SHADER_VERSION_COMPUTE_LIGHT,
GI_PROBE_SHADER_VERSION_COMPUTE_SECOND_BOUNCE,
@@ -458,6 +456,8 @@ private:
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,
@@ -591,6 +591,7 @@ private:
ShadowTransform shadow_transform[4];
+ AABB aabb;
RID self;
RID light;
Transform transform;
@@ -680,6 +681,19 @@ private:
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;
+ bool sdfgi_enhance_ssr = 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;
};
RS::EnvironmentSSAOQuality ssao_quality = RS::ENV_SSAO_QUALITY_MEDIUM;
@@ -719,7 +733,13 @@ private:
/* RENDER BUFFERS */
+ struct SDFGI;
+
struct RenderBuffers {
+ enum {
+ MAX_GIPROBES = 8
+ };
+
RenderBufferData *data = nullptr;
int width = 0, height = 0;
RS::ViewportMSAA msaa = RS::VIEWPORT_MSAA_DISABLED;
@@ -732,6 +752,9 @@ private:
RID texture; //main texture for rendering to, must be filled after done rendering
RID depth_texture; //main depth texture
+ RID gi_uniform_set;
+ SDFGI *sdfgi = nullptr;
+
//built-in textures used for ping pong image processing and blurring
struct Blur {
RID texture;
@@ -764,10 +787,389 @@ private:
RID depth_scaled;
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;
+ };
+
+ //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 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;
+ };
+
+ 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;
+
+ RenderPipelineVertexFormatCacheRD 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;
+ };
+
+ 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 pad[1];
+ };
+
+ 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_curve = 1.0;
+ float screen_space_roughness_limiter_amount = 0.25;
+ float screen_space_roughness_limiter_limit = 0.18;
mutable RID_Owner<RenderBuffers> render_buffers_owner;
@@ -777,10 +1179,16 @@ 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);
uint64_t scene_pass = 0;
uint64_t shadow_atlas_realloc_tolerance_msec = 500;
+ struct SDFGICosineNeighbour {
+ uint32_t neighbour;
+ float weight;
+ };
+
public:
/* SHADOW ATLAS API */
@@ -818,6 +1226,15 @@ public:
return Size2i(directional_shadow.size, directional_shadow.size);
}
+ /* 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 RID *p_directional_light_instances, uint32_t p_directional_light_count, 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();
@@ -871,6 +1288,11 @@ public:
float environment_get_ssao_ao_affect(RID p_env) const;
float environment_get_ssao_light_affect(RID p_env) const;
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, bool p_enhance_ssr, 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);
void environment_set_ssr_roughness_quality(RS::EnvironmentSSRRoughnessQuality p_quality);
RS::EnvironmentSSRRoughnessQuality environment_get_ssr_roughness_quality() const;
@@ -894,6 +1316,7 @@ public:
RID light_instance_create(RID p_light);
void light_instance_set_transform(RID p_light_instance, const Transform &p_transform);
+ void light_instance_set_aabb(RID p_light_instance, const AABB &p_aabb);
void light_instance_set_shadow_transform(RID p_light_instance, const CameraMatrix &p_projection, const Transform &p_transform, float p_far, float p_split, int p_pass, float p_shadow_texel_size, float p_bias_scale = 1.0, float p_range_begin = 0, const Vector2 &p_uv_scale = Vector2());
void light_instance_mark_visible(RID p_light_instance);
@@ -1107,6 +1530,8 @@ public:
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, int p_dynamic_object_count, InstanceBase **p_dynamic_objects);
+ 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;
@@ -1124,10 +1549,6 @@ public:
GIProbeInstance *gi_probe = gi_probe_instance_owner.getornull(p_probe);
return gi_probe->texture;
}
- _FORCE_INLINE_ RID gi_probe_instance_get_aniso_texture(RID p_probe, int p_index) {
- GIProbeInstance *gi_probe = gi_probe_instance_owner.getornull(p_probe);
- return gi_probe->anisotropy[p_index];
- }
_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);
@@ -1141,7 +1562,7 @@ public:
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);
@@ -1154,18 +1575,27 @@ public:
return g_probe->last_pass;
}
-
- const Vector<RID> &gi_probe_get_slots() const;
- _FORCE_INLINE_ bool gi_probe_is_anisotropic() const {
- return gi_probe_use_anisotropy;
- }
- GIProbeQuality gi_probe_get_quality() const;
-
+*/
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);
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();
+
+ uint32_t render_buffers_get_sdfgi_cascade_count(RID p_render_buffers) const;
+ bool render_buffers_is_sdfgi_enabled(RID p_render_buffers) const;
+ RID render_buffers_get_sdfgi_irradiance_probes(RID p_render_buffers) const;
+ Vector3 render_buffers_get_sdfgi_cascade_offset(RID p_render_buffers, uint32_t p_cascade) const;
+ Vector3i render_buffers_get_sdfgi_cascade_probe_offset(RID p_render_buffers, uint32_t p_cascade) const;
+ float render_buffers_get_sdfgi_cascade_probe_size(RID p_render_buffers, uint32_t p_cascade) const;
+ float render_buffers_get_sdfgi_normal_bias(RID p_render_buffers) const;
+ uint32_t render_buffers_get_sdfgi_cascade_probe_count(RID p_render_buffers) const;
+ uint32_t render_buffers_get_sdfgi_cascade_size(RID p_render_buffers) const;
+ bool render_buffers_is_sdfgi_using_occlusion(RID p_render_buffers) const;
+ float render_buffers_get_sdfgi_energy(RID p_render_buffers) const;
+ RID render_buffers_get_sdfgi_occlusion_texture(RID p_render_buffers) const;
void render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, InstanceBase **p_cull_result, int p_cull_count, RID *p_light_cull_result, int p_light_cull_count, RID *p_reflection_probe_cull_result, int p_reflection_probe_cull_count, RID *p_gi_probe_cull_result, int p_gi_probe_cull_count, RID *p_decal_cull_result, int p_decal_cull_count, InstanceBase **p_lightmap_cull_result, int p_lightmap_cull_count, RID p_environment, RID p_shadow_atlas, RID p_camera_effects, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass);
@@ -1173,6 +1603,9 @@ public:
void render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, InstanceBase **p_cull_result, int p_cull_count, RID p_framebuffer, const Rect2i &p_region);
+ void render_sdfgi(RID p_render_buffers, int p_region, InstanceBase **p_cull_result, int p_cull_count);
+ void render_sdfgi_static_lights(RID p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const RID **p_positional_light_cull_result, const uint32_t *p_positional_light_cull_count);
+
virtual void set_scene_pass(uint64_t p_pass) {
scene_pass = p_pass;
}
@@ -1180,9 +1613,10 @@ public:
return scene_pass;
}
- virtual void screen_space_roughness_limiter_set_active(bool p_enable, float p_curve);
+ virtual void screen_space_roughness_limiter_set_active(bool p_enable, float p_amount, float p_limit);
virtual bool screen_space_roughness_limiter_is_active() const;
- virtual float screen_space_roughness_limiter_get_curve() const;
+ virtual float screen_space_roughness_limiter_get_amount() const;
+ virtual float screen_space_roughness_limiter_get_limit() const;
virtual void sub_surface_scattering_set_quality(RS::SubSurfaceScatteringQuality p_quality);
RS::SubSurfaceScatteringQuality sub_surface_scattering_get_quality() const;
@@ -1221,6 +1655,8 @@ public:
virtual void set_time(double p_time, double p_step);
+ void sdfgi_set_debug_probe_select(const Vector3 &p_position, const Vector3 &p_dir);
+
RasterizerSceneRD(RasterizerStorageRD *p_storage);
~RasterizerSceneRD();
};
diff --git a/servers/rendering/rasterizer_rd/rasterizer_storage_rd.cpp b/servers/rendering/rasterizer_rd/rasterizer_storage_rd.cpp
index a5151d1ff8..8f3e2c25f9 100644
--- a/servers/rendering/rasterizer_rd/rasterizer_storage_rd.cpp
+++ b/servers/rendering/rasterizer_rd/rasterizer_storage_rd.cpp
@@ -3387,11 +3387,21 @@ void RasterizerStorageRD::light_set_reverse_cull_face_mode(RID p_light, bool p_e
light->instance_dependency.instance_notify_changed(true, false);
}
-void RasterizerStorageRD::light_set_use_gi(RID p_light, bool p_enabled) {
+void RasterizerStorageRD::light_set_bake_mode(RID p_light, RS::LightBakeMode p_bake_mode) {
Light *light = light_owner.getornull(p_light);
ERR_FAIL_COND(!light);
- light->use_gi = p_enabled;
+ light->bake_mode = p_bake_mode;
+
+ light->version++;
+ light->instance_dependency.instance_notify_changed(true, false);
+}
+
+void RasterizerStorageRD::light_set_max_sdfgi_cascade(RID p_light, uint32_t p_cascade) {
+ Light *light = light_owner.getornull(p_light);
+ ERR_FAIL_COND(!light);
+
+ light->max_sdfgi_cascade = p_cascade;
light->version++;
light->instance_dependency.instance_notify_changed(true, false);
@@ -3460,11 +3470,18 @@ RS::LightDirectionalShadowDepthRangeMode RasterizerStorageRD::light_directional_
return light->directional_range_mode;
}
-bool RasterizerStorageRD::light_get_use_gi(RID p_light) {
- Light *light = light_owner.getornull(p_light);
- ERR_FAIL_COND_V(!light, false);
+uint32_t RasterizerStorageRD::light_get_max_sdfgi_cascade(RID p_light) {
+ const Light *light = light_owner.getornull(p_light);
+ ERR_FAIL_COND_V(!light, 0);
+
+ return light->max_sdfgi_cascade;
+}
+
+RS::LightBakeMode RasterizerStorageRD::light_get_bake_mode(RID p_light) {
+ const Light *light = light_owner.getornull(p_light);
+ ERR_FAIL_COND_V(!light, RS::LIGHT_BAKE_DISABLED);
- return light->use_gi;
+ return light->bake_mode;
}
uint64_t RasterizerStorageRD::light_get_version(RID p_light) const {
@@ -3517,25 +3534,25 @@ void RasterizerStorageRD::reflection_probe_set_intensity(RID p_probe, float p_in
reflection_probe->intensity = p_intensity;
}
-void RasterizerStorageRD::reflection_probe_set_interior_ambient(RID p_probe, const Color &p_ambient) {
+void RasterizerStorageRD::reflection_probe_set_ambient_mode(RID p_probe, RS::ReflectionProbeAmbientMode p_mode) {
ReflectionProbe *reflection_probe = reflection_probe_owner.getornull(p_probe);
ERR_FAIL_COND(!reflection_probe);
- reflection_probe->interior_ambient = p_ambient;
+ reflection_probe->ambient_mode = p_mode;
}
-void RasterizerStorageRD::reflection_probe_set_interior_ambient_energy(RID p_probe, float p_energy) {
+void RasterizerStorageRD::reflection_probe_set_ambient_color(RID p_probe, const Color &p_color) {
ReflectionProbe *reflection_probe = reflection_probe_owner.getornull(p_probe);
ERR_FAIL_COND(!reflection_probe);
- reflection_probe->interior_ambient_energy = p_energy;
+ reflection_probe->ambient_color = p_color;
}
-void RasterizerStorageRD::reflection_probe_set_interior_ambient_probe_contribution(RID p_probe, float p_contrib) {
+void RasterizerStorageRD::reflection_probe_set_ambient_energy(RID p_probe, float p_energy) {
ReflectionProbe *reflection_probe = reflection_probe_owner.getornull(p_probe);
ERR_FAIL_COND(!reflection_probe);
- reflection_probe->interior_ambient_probe_contrib = p_contrib;
+ reflection_probe->ambient_color_energy = p_energy;
}
void RasterizerStorageRD::reflection_probe_set_max_distance(RID p_probe, float p_distance) {
@@ -3683,25 +3700,23 @@ bool RasterizerStorageRD::reflection_probe_is_box_projection(RID p_probe) const
return reflection_probe->box_projection;
}
-Color RasterizerStorageRD::reflection_probe_get_interior_ambient(RID p_probe) const {
+RS::ReflectionProbeAmbientMode RasterizerStorageRD::reflection_probe_get_ambient_mode(RID p_probe) const {
const ReflectionProbe *reflection_probe = reflection_probe_owner.getornull(p_probe);
- ERR_FAIL_COND_V(!reflection_probe, Color());
-
- return reflection_probe->interior_ambient;
+ ERR_FAIL_COND_V(!reflection_probe, RS::REFLECTION_PROBE_AMBIENT_DISABLED);
+ return reflection_probe->ambient_mode;
}
-float RasterizerStorageRD::reflection_probe_get_interior_ambient_energy(RID p_probe) const {
+Color RasterizerStorageRD::reflection_probe_get_ambient_color(RID p_probe) const {
const ReflectionProbe *reflection_probe = reflection_probe_owner.getornull(p_probe);
- ERR_FAIL_COND_V(!reflection_probe, 0);
+ ERR_FAIL_COND_V(!reflection_probe, Color());
- return reflection_probe->interior_ambient_energy;
+ return reflection_probe->ambient_color;
}
-
-float RasterizerStorageRD::reflection_probe_get_interior_ambient_probe_contribution(RID p_probe) const {
+float RasterizerStorageRD::reflection_probe_get_ambient_color_energy(RID p_probe) const {
const ReflectionProbe *reflection_probe = reflection_probe_owner.getornull(p_probe);
ERR_FAIL_COND_V(!reflection_probe, 0);
- return reflection_probe->interior_ambient_probe_contrib;
+ return reflection_probe->ambient_color_energy;
}
RID RasterizerStorageRD::decal_create() {
@@ -5878,6 +5893,20 @@ RasterizerStorageRD::RasterizerStorageRD() {
}
default_rd_textures[DEFAULT_RD_TEXTURE_MULTIMESH_BUFFER] = RD::get_singleton()->texture_buffer_create(16, RD::DATA_FORMAT_R8G8B8A8_UNORM, pv);
+
+ for (int i = 0; i < 16; i++) {
+ pv.set(i * 4 + 0, 0);
+ pv.set(i * 4 + 1, 0);
+ pv.set(i * 4 + 2, 0);
+ pv.set(i * 4 + 3, 0);
+ }
+
+ {
+ tformat.format = RD::DATA_FORMAT_R8G8B8A8_UINT;
+ Vector<Vector<uint8_t>> vpv;
+ vpv.push_back(pv);
+ default_rd_textures[DEFAULT_RD_TEXTURE_2D_UINT] = RD::get_singleton()->texture_create(tformat, RD::TextureView(), vpv);
+ }
}
{ //create default cubemap
diff --git a/servers/rendering/rasterizer_rd/rasterizer_storage_rd.h b/servers/rendering/rasterizer_rd/rasterizer_storage_rd.h
index fe9377192e..b1146f1386 100644
--- a/servers/rendering/rasterizer_rd/rasterizer_storage_rd.h
+++ b/servers/rendering/rasterizer_rd/rasterizer_storage_rd.h
@@ -92,6 +92,7 @@ public:
DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_BLACK,
DEFAULT_RD_TEXTURE_3D_WHITE,
DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE,
+ DEFAULT_RD_TEXTURE_2D_UINT,
DEFAULT_RD_TEXTURE_MAX
};
@@ -420,7 +421,8 @@ private:
bool shadow = false;
bool negative = false;
bool reverse_cull = false;
- bool use_gi = true;
+ RS::LightBakeMode bake_mode = RS::LIGHT_BAKE_DYNAMIC;
+ uint32_t max_sdfgi_cascade = 2;
uint32_t cull_mask = 0xFFFFFFFF;
RS::LightOmniShadowMode omni_shadow_mode = RS::LIGHT_OMNI_SHADOW_DUAL_PARABOLOID;
RS::LightDirectionalShadowMode directional_shadow_mode = RS::LIGHT_DIRECTIONAL_SHADOW_ORTHOGONAL;
@@ -439,9 +441,9 @@ private:
RS::ReflectionProbeUpdateMode update_mode = RS::REFLECTION_PROBE_UPDATE_ONCE;
int resolution = 256;
float intensity = 1.0;
- Color interior_ambient;
- float interior_ambient_energy = 1.0;
- float interior_ambient_probe_contrib = 0.0;
+ RS::ReflectionProbeAmbientMode ambient_mode = RS::REFLECTION_PROBE_AMBIENT_ENVIRONMENT;
+ Color ambient_color;
+ float ambient_color_energy = 1.0;
float max_distance = 0;
Vector3 extents = Vector3(1, 1, 1);
Vector3 origin_offset;
@@ -1041,7 +1043,8 @@ public:
void light_set_negative(RID p_light, bool p_enable);
void light_set_cull_mask(RID p_light, uint32_t p_mask);
void light_set_reverse_cull_face_mode(RID p_light, bool p_enabled);
- void light_set_use_gi(RID p_light, bool p_enabled);
+ void light_set_bake_mode(RID p_light, RS::LightBakeMode p_bake_mode);
+ void light_set_max_sdfgi_cascade(RID p_light, uint32_t p_cascade);
void light_omni_set_shadow_mode(RID p_light, RS::LightOmniShadowMode p_mode);
@@ -1118,7 +1121,8 @@ public:
return light->param[RS::LIGHT_PARAM_TRANSMITTANCE_BIAS];
}
- bool light_get_use_gi(RID p_light);
+ RS::LightBakeMode light_get_bake_mode(RID p_light);
+ uint32_t light_get_max_sdfgi_cascade(RID p_light);
uint64_t light_get_version(RID p_light) const;
/* PROBE API */
@@ -1127,9 +1131,9 @@ public:
void reflection_probe_set_update_mode(RID p_probe, RS::ReflectionProbeUpdateMode p_mode);
void reflection_probe_set_intensity(RID p_probe, float p_intensity);
- void reflection_probe_set_interior_ambient(RID p_probe, const Color &p_ambient);
- void reflection_probe_set_interior_ambient_energy(RID p_probe, float p_energy);
- void reflection_probe_set_interior_ambient_probe_contribution(RID p_probe, float p_contrib);
+ void reflection_probe_set_ambient_mode(RID p_probe, RS::ReflectionProbeAmbientMode p_mode);
+ void reflection_probe_set_ambient_color(RID p_probe, const Color &p_color);
+ void reflection_probe_set_ambient_energy(RID p_probe, float p_energy);
void reflection_probe_set_max_distance(RID p_probe, float p_distance);
void reflection_probe_set_extents(RID p_probe, const Vector3 &p_extents);
void reflection_probe_set_origin_offset(RID p_probe, const Vector3 &p_offset);
@@ -1151,9 +1155,9 @@ public:
float reflection_probe_get_intensity(RID p_probe) const;
bool reflection_probe_is_interior(RID p_probe) const;
bool reflection_probe_is_box_projection(RID p_probe) const;
- Color reflection_probe_get_interior_ambient(RID p_probe) const;
- float reflection_probe_get_interior_ambient_energy(RID p_probe) const;
- float reflection_probe_get_interior_ambient_probe_contribution(RID p_probe) const;
+ RS::ReflectionProbeAmbientMode reflection_probe_get_ambient_mode(RID p_probe) const;
+ Color reflection_probe_get_ambient_color(RID p_probe) const;
+ float reflection_probe_get_ambient_color_energy(RID p_probe) const;
void base_update_dependency(RID p_base, RasterizerScene::InstanceBase *p_instance);
void skeleton_update_dependency(RID p_skeleton, RasterizerScene::InstanceBase *p_instance);
diff --git a/servers/rendering/rasterizer_rd/shaders/SCsub b/servers/rendering/rasterizer_rd/shaders/SCsub
index a454d144aa..67f4edc626 100644
--- a/servers/rendering/rasterizer_rd/shaders/SCsub
+++ b/servers/rendering/rasterizer_rd/shaders/SCsub
@@ -28,3 +28,10 @@ if "RD_GLSL" in env["BUILDERS"]:
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")
diff --git a/servers/rendering/rasterizer_rd/shaders/copy_to_fb.glsl b/servers/rendering/rasterizer_rd/shaders/copy_to_fb.glsl
index b1cfe1e91e..9751e13b4e 100644
--- a/servers/rendering/rasterizer_rd/shaders/copy_to_fb.glsl
+++ b/servers/rendering/rasterizer_rd/shaders/copy_to_fb.glsl
@@ -47,16 +47,26 @@ layout(push_constant, binding = 1, std430) uniform Params {
bool force_luminance;
bool alpha_to_zero;
- uint pad[2];
+ bool srgb;
+ uint pad;
}
params;
layout(location = 0) in vec2 uv_interp;
layout(set = 0, binding = 0) uniform sampler2D source_color;
-
+#ifdef MODE_TWO_SOURCES
+layout(set = 1, binding = 0) uniform sampler2D source_color2;
+#endif
layout(location = 0) out vec4 frag_color;
+vec3 linear_to_srgb(vec3 color) {
+ //if going to srgb, clamp from 0 to 1.
+ color = clamp(color, vec3(0.0), vec3(1.0));
+ const vec3 a = vec3(0.055f);
+ return mix((vec3(1.0f) + a) * pow(color.rgb, vec3(1.0f / 2.4f)) - a, 12.92f * color.rgb, lessThan(color.rgb, vec3(0.0031308f)));
+}
+
void main() {
vec2 uv = uv_interp;
@@ -89,11 +99,17 @@ void main() {
}
#endif
vec4 color = textureLod(source_color, uv, 0.0);
+#ifdef MODE_TWO_SOURCES
+ color += textureLod(source_color2, uv, 0.0);
+#endif
if (params.force_luminance) {
color.rgb = vec3(max(max(color.r, color.g), color.b));
}
if (params.alpha_to_zero) {
color.rgb *= color.a;
}
+ if (params.srgb) {
+ color.rgb = linear_to_srgb(color.rgb);
+ }
frag_color = color;
}
diff --git a/servers/rendering/rasterizer_rd/shaders/gi.glsl b/servers/rendering/rasterizer_rd/shaders/gi.glsl
new file mode 100644
index 0000000000..a1939f75ad
--- /dev/null
+++ b/servers/rendering/rasterizer_rd/shaders/gi.glsl
@@ -0,0 +1,663 @@
+#[compute]
+
+#version 450
+
+VERSION_DEFINES
+
+layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in;
+
+#define M_PI 3.141592
+
+#define SDFGI_MAX_CASCADES 8
+
+//set 0 for SDFGI and render buffers
+
+layout(set = 0, binding = 1) uniform texture3D sdf_cascades[SDFGI_MAX_CASCADES];
+layout(set = 0, binding = 2) uniform texture3D light_cascades[SDFGI_MAX_CASCADES];
+layout(set = 0, binding = 3) uniform texture3D aniso0_cascades[SDFGI_MAX_CASCADES];
+layout(set = 0, binding = 4) uniform texture3D aniso1_cascades[SDFGI_MAX_CASCADES];
+layout(set = 0, binding = 5) uniform texture3D occlusion_texture;
+
+layout(set = 0, binding = 6) uniform sampler linear_sampler;
+layout(set = 0, binding = 7) uniform sampler linear_sampler_with_mipmaps;
+
+struct ProbeCascadeData {
+ vec3 position;
+ float to_probe;
+ ivec3 probe_world_offset;
+ float to_cell; // 1/bounds * grid_size
+};
+
+layout(rgba16f, set = 0, binding = 9) uniform restrict writeonly image2D ambient_buffer;
+layout(rgba16f, set = 0, binding = 10) uniform restrict writeonly image2D reflection_buffer;
+
+layout(set = 0, binding = 11) uniform texture2DArray lightprobe_texture;
+
+layout(set = 0, binding = 12) uniform texture2D depth_buffer;
+layout(set = 0, binding = 13) uniform texture2D normal_roughness_buffer;
+layout(set = 0, binding = 14) uniform utexture2D giprobe_buffer;
+
+layout(set = 0, binding = 15, std140) uniform SDFGI {
+ vec3 grid_size;
+ uint max_cascades;
+
+ bool use_occlusion;
+ int probe_axis_size;
+ float probe_to_uvw;
+ float normal_bias;
+
+ vec3 lightprobe_tex_pixel_size;
+ float energy;
+
+ vec3 lightprobe_uv_offset;
+ float y_mult;
+
+ vec3 occlusion_clamp;
+ uint pad3;
+
+ vec3 occlusion_renormalize;
+ uint pad4;
+
+ vec3 cascade_probe_size;
+ uint pad5;
+
+ ProbeCascadeData cascades[SDFGI_MAX_CASCADES];
+}
+sdfgi;
+
+#define MAX_GI_PROBES 8
+
+struct GIProbeData {
+ mat4 xform;
+ vec3 bounds;
+ float dynamic_range;
+
+ float bias;
+ float normal_bias;
+ bool blend_ambient;
+ uint texture_slot;
+
+ float anisotropy_strength;
+ float ambient_occlusion;
+ float ambient_occlusion_size;
+ uint pad2;
+};
+
+layout(set = 0, binding = 16, std140) uniform GIProbes {
+ GIProbeData data[MAX_GI_PROBES];
+}
+gi_probes;
+
+layout(set = 0, binding = 17) uniform texture3D gi_probe_textures[MAX_GI_PROBES];
+
+layout(push_constant, binding = 0, std430) uniform Params {
+ ivec2 screen_size;
+ float z_near;
+ float z_far;
+
+ vec4 proj_info;
+
+ uint max_giprobes;
+ bool high_quality_vct;
+ bool use_sdfgi;
+ bool orthogonal;
+
+ vec3 ao_color;
+ uint pad;
+
+ mat3x4 cam_rotation;
+}
+params;
+
+vec2 octahedron_wrap(vec2 v) {
+ vec2 signVal;
+ signVal.x = v.x >= 0.0 ? 1.0 : -1.0;
+ signVal.y = v.y >= 0.0 ? 1.0 : -1.0;
+ return (1.0 - abs(v.yx)) * signVal;
+}
+
+vec2 octahedron_encode(vec3 n) {
+ // https://twitter.com/Stubbesaurus/status/937994790553227264
+ n /= (abs(n.x) + abs(n.y) + abs(n.z));
+ n.xy = n.z >= 0.0 ? n.xy : octahedron_wrap(n.xy);
+ n.xy = n.xy * 0.5 + 0.5;
+ return n.xy;
+}
+
+vec4 blend_color(vec4 src, vec4 dst) {
+ vec4 res;
+ float sa = 1.0 - src.a;
+ res.a = dst.a * sa + src.a;
+ if (res.a == 0.0) {
+ res.rgb = vec3(0);
+ } else {
+ res.rgb = (dst.rgb * dst.a * sa + src.rgb * src.a) / res.a;
+ }
+ return res;
+}
+
+vec3 reconstruct_position(ivec2 screen_pos) {
+ vec3 pos;
+ pos.z = texelFetch(sampler2D(depth_buffer, linear_sampler), screen_pos, 0).r;
+
+ pos.z = pos.z * 2.0 - 1.0;
+ if (params.orthogonal) {
+ pos.z = ((pos.z + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0;
+ } else {
+ pos.z = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - pos.z * (params.z_far - params.z_near));
+ }
+ pos.z = -pos.z;
+
+ pos.xy = vec2(screen_pos) * params.proj_info.xy + params.proj_info.zw;
+ if (!params.orthogonal) {
+ pos.xy *= pos.z;
+ }
+
+ return pos;
+}
+
+void sdfgi_probe_process(uint cascade, vec3 cascade_pos, vec3 cam_pos, vec3 cam_normal, vec3 cam_specular_normal, float roughness, out vec3 diffuse_light, out vec3 specular_light) {
+ cascade_pos += cam_normal * sdfgi.normal_bias;
+
+ vec3 base_pos = floor(cascade_pos);
+ //cascade_pos += mix(vec3(0.0),vec3(0.01),lessThan(abs(cascade_pos-base_pos),vec3(0.01))) * cam_normal;
+ ivec3 probe_base_pos = ivec3(base_pos);
+
+ vec4 diffuse_accum = vec4(0.0);
+ vec3 specular_accum;
+
+ ivec3 tex_pos = ivec3(probe_base_pos.xy, int(cascade));
+ tex_pos.x += probe_base_pos.z * sdfgi.probe_axis_size;
+ tex_pos.xy = tex_pos.xy * (SDFGI_OCT_SIZE + 2) + ivec2(1);
+
+ vec3 diffuse_posf = (vec3(tex_pos) + vec3(octahedron_encode(cam_normal) * float(SDFGI_OCT_SIZE), 0.0)) * sdfgi.lightprobe_tex_pixel_size;
+
+ vec3 specular_posf = (vec3(tex_pos) + vec3(octahedron_encode(cam_specular_normal) * float(SDFGI_OCT_SIZE), 0.0)) * sdfgi.lightprobe_tex_pixel_size;
+
+ specular_accum = vec3(0.0);
+
+ vec4 light_accum = vec4(0.0);
+ float weight_accum = 0.0;
+
+ for (uint j = 0; j < 8; j++) {
+ ivec3 offset = (ivec3(j) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1);
+ ivec3 probe_posi = probe_base_pos;
+ probe_posi += offset;
+
+ // Compute weight
+
+ vec3 probe_pos = vec3(probe_posi);
+ vec3 probe_to_pos = cascade_pos - probe_pos;
+ vec3 probe_dir = normalize(-probe_to_pos);
+
+ vec3 trilinear = vec3(1.0) - abs(probe_to_pos);
+ float weight = trilinear.x * trilinear.y * trilinear.z * max(0.005, dot(cam_normal, probe_dir));
+
+ // Compute lightprobe occlusion
+
+ if (sdfgi.use_occlusion) {
+ ivec3 occ_indexv = abs((sdfgi.cascades[cascade].probe_world_offset + probe_posi) & ivec3(1, 1, 1)) * ivec3(1, 2, 4);
+ vec4 occ_mask = mix(vec4(0.0), vec4(1.0), equal(ivec4(occ_indexv.x | occ_indexv.y), ivec4(0, 1, 2, 3)));
+
+ vec3 occ_pos = clamp(cascade_pos, probe_pos - sdfgi.occlusion_clamp, probe_pos + sdfgi.occlusion_clamp) * sdfgi.probe_to_uvw;
+ occ_pos.z += float(cascade);
+ if (occ_indexv.z != 0) { //z bit is on, means index is >=4, so make it switch to the other half of textures
+ occ_pos.x += 1.0;
+ }
+
+ occ_pos *= sdfgi.occlusion_renormalize;
+ float occlusion = dot(textureLod(sampler3D(occlusion_texture, linear_sampler), occ_pos, 0.0), occ_mask);
+
+ weight *= max(occlusion, 0.01);
+ }
+
+ // Compute lightprobe texture position
+
+ vec3 diffuse;
+ vec3 pos_uvw = diffuse_posf;
+ pos_uvw.xy += vec2(offset.xy) * sdfgi.lightprobe_uv_offset.xy;
+ pos_uvw.x += float(offset.z) * sdfgi.lightprobe_uv_offset.z;
+ diffuse = textureLod(sampler2DArray(lightprobe_texture, linear_sampler), pos_uvw, 0.0).rgb;
+
+ diffuse_accum += vec4(diffuse * weight, weight);
+
+ {
+ vec3 specular = vec3(0.0);
+ vec3 pos_uvw = specular_posf;
+ pos_uvw.xy += vec2(offset.xy) * sdfgi.lightprobe_uv_offset.xy;
+ pos_uvw.x += float(offset.z) * sdfgi.lightprobe_uv_offset.z;
+ if (roughness < 0.99) {
+ specular = textureLod(sampler2DArray(lightprobe_texture, linear_sampler), pos_uvw + vec3(0, 0, float(sdfgi.max_cascades)), 0.0).rgb;
+ }
+ if (roughness > 0.2) {
+ specular = mix(specular, textureLod(sampler2DArray(lightprobe_texture, linear_sampler), pos_uvw, 0.0).rgb, (roughness - 0.2) * 1.25);
+ }
+
+ specular_accum += specular * weight;
+ }
+ }
+
+ if (diffuse_accum.a > 0.0) {
+ diffuse_accum.rgb /= diffuse_accum.a;
+ }
+
+ diffuse_light = diffuse_accum.rgb;
+
+ if (diffuse_accum.a > 0.0) {
+ specular_accum /= diffuse_accum.a;
+ }
+
+ specular_light = specular_accum;
+}
+
+void sdfgi_process(vec3 vertex, vec3 normal, vec3 reflection, float roughness, out vec4 ambient_light, out vec4 reflection_light) {
+ //make vertex orientation the world one, but still align to camera
+ vertex.y *= sdfgi.y_mult;
+ normal.y *= sdfgi.y_mult;
+ reflection.y *= sdfgi.y_mult;
+
+ //renormalize
+ normal = normalize(normal);
+ reflection = normalize(reflection);
+
+ vec3 cam_pos = vertex;
+ vec3 cam_normal = normal;
+
+ vec4 light_accum = vec4(0.0);
+ float weight_accum = 0.0;
+
+ vec4 light_blend_accum = vec4(0.0);
+ float weight_blend_accum = 0.0;
+
+ float blend = -1.0;
+
+ // helper constants, compute once
+
+ uint cascade = 0xFFFFFFFF;
+ vec3 cascade_pos;
+ vec3 cascade_normal;
+
+ for (uint i = 0; i < sdfgi.max_cascades; i++) {
+ cascade_pos = (cam_pos - sdfgi.cascades[i].position) * sdfgi.cascades[i].to_probe;
+
+ if (any(lessThan(cascade_pos, vec3(0.0))) || any(greaterThanEqual(cascade_pos, sdfgi.cascade_probe_size))) {
+ continue; //skip cascade
+ }
+
+ cascade = i;
+ break;
+ }
+
+ if (cascade < SDFGI_MAX_CASCADES) {
+ ambient_light = vec4(0, 0, 0, 1);
+ reflection_light = vec4(0, 0, 0, 1);
+
+ float blend;
+ vec3 diffuse, specular;
+ sdfgi_probe_process(cascade, cascade_pos, cam_pos, cam_normal, reflection, roughness, diffuse, specular);
+
+ {
+ //process blend
+ float blend_from = (float(sdfgi.probe_axis_size - 1) / 2.0) - 2.5;
+ float blend_to = blend_from + 2.0;
+
+ vec3 inner_pos = cam_pos * sdfgi.cascades[cascade].to_probe;
+
+ float len = length(inner_pos);
+
+ inner_pos = abs(normalize(inner_pos));
+ len *= max(inner_pos.x, max(inner_pos.y, inner_pos.z));
+
+ if (len >= blend_from) {
+ blend = smoothstep(blend_from, blend_to, len);
+ } else {
+ blend = 0.0;
+ }
+ }
+
+ if (blend > 0.0) {
+ //blend
+ if (cascade == sdfgi.max_cascades - 1) {
+ ambient_light.a = 1.0 - blend;
+ reflection_light.a = 1.0 - blend;
+
+ } else {
+ vec3 diffuse2, specular2;
+ cascade_pos = (cam_pos - sdfgi.cascades[cascade + 1].position) * sdfgi.cascades[cascade + 1].to_probe;
+ sdfgi_probe_process(cascade + 1, cascade_pos, cam_pos, cam_normal, reflection, roughness, diffuse2, specular2);
+ diffuse = mix(diffuse, diffuse2, blend);
+ specular = mix(specular, specular2, blend);
+ }
+ }
+
+ ambient_light.rgb = diffuse;
+#if 1
+ if (roughness < 0.2) {
+ vec3 pos_to_uvw = 1.0 / sdfgi.grid_size;
+ vec4 light_accum = vec4(0.0);
+
+ float blend_size = (sdfgi.grid_size.x / float(sdfgi.probe_axis_size - 1)) * 0.5;
+
+ float radius_sizes[SDFGI_MAX_CASCADES];
+ cascade = 0xFFFF;
+
+ float base_distance = length(cam_pos);
+ for (uint i = 0; i < sdfgi.max_cascades; i++) {
+ radius_sizes[i] = (1.0 / sdfgi.cascades[i].to_cell) * (sdfgi.grid_size.x * 0.5 - blend_size);
+ if (cascade == 0xFFFF && base_distance < radius_sizes[i]) {
+ cascade = i;
+ }
+ }
+
+ cascade = min(cascade, sdfgi.max_cascades - 1);
+
+ float max_distance = radius_sizes[sdfgi.max_cascades - 1];
+ vec3 ray_pos = cam_pos;
+ vec3 ray_dir = reflection;
+
+ {
+ float prev_radius = cascade > 0 ? radius_sizes[cascade - 1] : 0.0;
+ float base_blend = (base_distance - prev_radius) / (radius_sizes[cascade] - prev_radius);
+ float bias = (1.0 + base_blend) * 1.1;
+ vec3 abs_ray_dir = abs(ray_dir);
+ //ray_pos += ray_dir * (bias / sdfgi.cascades[cascade].to_cell); //bias to avoid self occlusion
+ ray_pos += (ray_dir * 1.0 / max(abs_ray_dir.x, max(abs_ray_dir.y, abs_ray_dir.z)) + cam_normal * 1.4) * bias / sdfgi.cascades[cascade].to_cell;
+ }
+
+ float softness = 0.2 + min(1.0, roughness * 5.0) * 4.0; //approximation to roughness so it does not seem like a hard fade
+ 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;
+ }
+
+ distance /= sdfgi.cascades[i].to_cell;
+
+ 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 distance2 = texture(sampler3D(sdf_cascades[i + 1], linear_sampler), pos).r * 255.0 - 1.1;
+
+ 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;
+ }
+
+ 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);
+
+ distance2 /= sdfgi.cascades[i + 1].to_cell;
+
+ hit_light = mix(hit_light, hit_light2, blend);
+ distance = mix(distance, distance2, blend);
+ }
+
+ light_accum += hit_light;
+ ray_pos += ray_dir * distance;
+ break;
+ }
+ }
+
+ if (light_accum.a > 0.99) {
+ break;
+ }
+ }
+
+ vec3 light = light_accum.rgb / max(light_accum.a, 0.00001);
+ float alpha = min(1.0, light_accum.a);
+
+ float b = min(1.0, roughness * 5.0);
+
+ float sa = 1.0 - b;
+
+ reflection_light.a = alpha * sa + b;
+ if (reflection_light.a == 0) {
+ specular = vec3(0.0);
+ } else {
+ specular = (light * alpha * sa + specular * b) / reflection_light.a;
+ }
+ }
+
+#endif
+
+ reflection_light.rgb = specular;
+
+ ambient_light.rgb *= sdfgi.energy;
+ reflection_light.rgb *= sdfgi.energy;
+ } else {
+ ambient_light = vec4(0);
+ reflection_light = vec4(0);
+ }
+}
+
+//standard voxel cone trace
+vec4 voxel_cone_trace(texture3D probe, vec3 cell_size, vec3 pos, vec3 direction, float tan_half_angle, float max_distance, float p_bias) {
+ float dist = p_bias;
+ vec4 color = vec4(0.0);
+
+ while (dist < max_distance && color.a < 0.95) {
+ float diameter = max(1.0, 2.0 * tan_half_angle * dist);
+ vec3 uvw_pos = (pos + dist * direction) * cell_size;
+ float half_diameter = diameter * 0.5;
+ //check if outside, then break
+ if (any(greaterThan(abs(uvw_pos - 0.5), vec3(0.5f + half_diameter * cell_size)))) {
+ break;
+ }
+ vec4 scolor = textureLod(sampler3D(probe, linear_sampler_with_mipmaps), uvw_pos, log2(diameter));
+ float a = (1.0 - color.a);
+ color += a * scolor;
+ dist += half_diameter;
+ }
+
+ return color;
+}
+
+vec4 voxel_cone_trace_45_degrees(texture3D probe, vec3 cell_size, vec3 pos, vec3 direction, float max_distance, float p_bias) {
+ float dist = p_bias;
+ vec4 color = vec4(0.0);
+ float radius = max(0.5, dist);
+ float lod_level = log2(radius * 2.0);
+
+ while (dist < max_distance && color.a < 0.95) {
+ vec3 uvw_pos = (pos + dist * direction) * cell_size;
+
+ //check if outside, then break
+ if (any(greaterThan(abs(uvw_pos - 0.5), vec3(0.5f + radius * cell_size)))) {
+ break;
+ }
+ vec4 scolor = textureLod(sampler3D(probe, linear_sampler_with_mipmaps), uvw_pos, lod_level);
+ lod_level += 1.0;
+
+ float a = (1.0 - color.a);
+ scolor *= a;
+ color += scolor;
+ dist += radius;
+ radius = max(0.5, dist);
+ }
+ return color;
+}
+
+void gi_probe_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3 normal_xform, float roughness, inout vec4 out_spec, inout vec4 out_diff, inout float out_blend) {
+ position = (gi_probes.data[index].xform * vec4(position, 1.0)).xyz;
+ ref_vec = normalize((gi_probes.data[index].xform * vec4(ref_vec, 0.0)).xyz);
+ normal = normalize((gi_probes.data[index].xform * vec4(normal, 0.0)).xyz);
+
+ position += normal * gi_probes.data[index].normal_bias;
+
+ //this causes corrupted pixels, i have no idea why..
+ if (any(bvec2(any(lessThan(position, vec3(0.0))), any(greaterThan(position, gi_probes.data[index].bounds))))) {
+ return;
+ }
+
+ mat3 dir_xform = mat3(gi_probes.data[index].xform) * normal_xform;
+
+ vec3 blendv = abs(position / gi_probes.data[index].bounds * 2.0 - 1.0);
+ float blend = clamp(1.0 - max(blendv.x, max(blendv.y, blendv.z)), 0.0, 1.0);
+ //float blend=1.0;
+
+ float max_distance = length(gi_probes.data[index].bounds);
+ vec3 cell_size = 1.0 / gi_probes.data[index].bounds;
+
+ //irradiance
+
+ vec4 light = vec4(0.0);
+
+ if (params.high_quality_vct) {
+ const uint cone_dir_count = 6;
+ vec3 cone_dirs[cone_dir_count] = vec3[](
+ vec3(0.0, 0.0, 1.0),
+ vec3(0.866025, 0.0, 0.5),
+ vec3(0.267617, 0.823639, 0.5),
+ vec3(-0.700629, 0.509037, 0.5),
+ vec3(-0.700629, -0.509037, 0.5),
+ vec3(0.267617, -0.823639, 0.5));
+
+ float cone_weights[cone_dir_count] = float[](0.25, 0.15, 0.15, 0.15, 0.15, 0.15);
+ float cone_angle_tan = 0.577;
+
+ for (uint i = 0; i < cone_dir_count; i++) {
+ vec3 dir = normalize(dir_xform * cone_dirs[i]);
+ light += cone_weights[i] * voxel_cone_trace(gi_probe_textures[index], cell_size, position, dir, cone_angle_tan, max_distance, gi_probes.data[index].bias);
+ }
+ } else {
+ const uint cone_dir_count = 4;
+ vec3 cone_dirs[cone_dir_count] = vec3[](
+ vec3(0.707107, 0.0, 0.707107),
+ vec3(0.0, 0.707107, 0.707107),
+ vec3(-0.707107, 0.0, 0.707107),
+ vec3(0.0, -0.707107, 0.707107));
+
+ float cone_weights[cone_dir_count] = float[](0.25, 0.25, 0.25, 0.25);
+ for (int i = 0; i < cone_dir_count; i++) {
+ vec3 dir = normalize(dir_xform * cone_dirs[i]);
+ light += cone_weights[i] * voxel_cone_trace_45_degrees(gi_probe_textures[index], cell_size, position, dir, max_distance, gi_probes.data[index].bias);
+ }
+ }
+
+ if (gi_probes.data[index].ambient_occlusion > 0.001) {
+ float size = 1.0 + gi_probes.data[index].ambient_occlusion_size * 7.0;
+
+ float taps, blend;
+ blend = modf(size, taps);
+ float ao = 0.0;
+ for (float i = 1.0; i <= taps; i++) {
+ vec3 ofs = (position + normal * (i * 0.5 + 1.0)) * cell_size;
+ ao += textureLod(sampler3D(gi_probe_textures[index], linear_sampler_with_mipmaps), ofs, i - 1.0).a * i;
+ }
+
+ if (blend > 0.001) {
+ vec3 ofs = (position + normal * ((taps + 1.0) * 0.5 + 1.0)) * cell_size;
+ ao += textureLod(sampler3D(gi_probe_textures[index], linear_sampler_with_mipmaps), ofs, taps).a * (taps + 1.0) * blend;
+ }
+
+ ao = 1.0 - min(1.0, ao);
+
+ light.rgb = mix(params.ao_color, light.rgb, mix(1.0, ao, gi_probes.data[index].ambient_occlusion));
+ }
+
+ light.rgb *= gi_probes.data[index].dynamic_range;
+ if (!gi_probes.data[index].blend_ambient) {
+ light.a = 1.0;
+ }
+
+ out_diff += light * blend;
+
+ //radiance
+ vec4 irr_light = voxel_cone_trace(gi_probe_textures[index], cell_size, position, ref_vec, tan(roughness * 0.5 * M_PI * 0.99), max_distance, gi_probes.data[index].bias);
+ irr_light.rgb *= gi_probes.data[index].dynamic_range;
+ if (!gi_probes.data[index].blend_ambient) {
+ irr_light.a = 1.0;
+ }
+
+ out_spec += irr_light * blend;
+
+ out_blend += blend;
+}
+
+vec4 fetch_normal_and_roughness(ivec2 pos) {
+ vec4 normal_roughness = texelFetch(sampler2D(normal_roughness_buffer, linear_sampler), pos, 0);
+
+ normal_roughness.xyz = normalize(normal_roughness.xyz * 2.0 - 1.0);
+ return normal_roughness;
+}
+
+void 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;
+
+ vec4 normal_roughness = fetch_normal_and_roughness(pos);
+ vec3 normal = normal_roughness.xyz;
+
+ vec4 ambient_light = vec4(0.0), reflection_light = vec4(0.0);
+
+ 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);
+ }
+
+ if (params.max_giprobes > 0) {
+ uvec2 giprobe_tex = texelFetch(usampler2D(giprobe_buffer, linear_sampler), pos, 0).rg;
+ roughness *= roughness;
+ //find arbitrary tangent and bitangent, then build a matrix
+ vec3 v0 = abs(normal.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 1.0, 0.0);
+ vec3 tangent = normalize(cross(v0, normal));
+ vec3 bitangent = normalize(cross(tangent, normal));
+ mat3 normal_mat = mat3(tangent, bitangent, normal);
+
+ vec4 amb_accum = vec4(0.0);
+ vec4 spec_accum = vec4(0.0);
+ float blend_accum = 0.0;
+
+ for (uint i = 0; i < params.max_giprobes; i++) {
+ if (any(equal(uvec2(i), giprobe_tex))) {
+ gi_probe_compute(i, vertex, normal, reflection, normal_mat, roughness, spec_accum, amb_accum, blend_accum);
+ }
+ }
+ if (blend_accum > 0.0) {
+ amb_accum /= blend_accum;
+ spec_accum /= blend_accum;
+ }
+
+ 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;
+ }
+ }
+ }
+
+ imageStore(ambient_buffer, pos, ambient_light);
+ imageStore(reflection_buffer, pos, reflection_light);
+}
diff --git a/servers/rendering/rasterizer_rd/shaders/resolve.glsl b/servers/rendering/rasterizer_rd/shaders/resolve.glsl
new file mode 100644
index 0000000000..9429a66dc9
--- /dev/null
+++ b/servers/rendering/rasterizer_rd/shaders/resolve.glsl
@@ -0,0 +1,110 @@
+#[compute]
+
+#version 450
+
+VERSION_DEFINES
+
+layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in;
+
+#ifdef MODE_RESOLVE_GI
+layout(set = 0, binding = 0) uniform sampler2DMS source_depth;
+layout(set = 0, binding = 1) uniform sampler2DMS source_normal_roughness;
+
+layout(r32f, set = 1, binding = 0) uniform restrict writeonly image2D dest_depth;
+layout(rgba8, set = 1, binding = 1) uniform restrict writeonly image2D dest_normal_roughness;
+
+#ifdef GIPROBE_RESOLVE
+layout(set = 2, binding = 0) uniform usampler2DMS source_giprobe;
+layout(rg8ui, set = 3, binding = 0) uniform restrict writeonly uimage2D dest_giprobe;
+#endif
+
+#endif
+
+layout(push_constant, binding = 16, std430) uniform Params {
+ ivec2 screen_size;
+ int sample_count;
+ uint pad;
+}
+params;
+
+void main() {
+ // Pixel being shaded
+ ivec2 pos = ivec2(gl_GlobalInvocationID.xy);
+ if (any(greaterThanEqual(pos, params.screen_size))) { //too large, do nothing
+ return;
+ }
+
+#ifdef MODE_RESOLVE_GI
+
+ float best_depth = 1e20;
+ vec4 best_normal_roughness = vec4(0.0);
+#ifdef GIPROBE_RESOLVE
+ uvec2 best_giprobe;
+#endif
+
+#if 0
+
+ for(int i=0;i<params.sample_count;i++) {
+ float depth = texelFetch(source_depth,pos,i).r;
+ if (depth < best_depth) { //use the depth closest to camera
+ best_depth = depth;
+ best_normal_roughness = texelFetch(source_normal_roughness,pos,i);
+
+#ifdef GIPROBE_RESOLVE
+ best_giprobe = texelFetch(source_giprobe,pos,i).rg;
+#endif
+ }
+ }
+
+#else
+
+ float depths[16];
+ int depth_indices[16];
+ int depth_amount[16];
+ int depth_count = 0;
+
+ for (int i = 0; i < params.sample_count; i++) {
+ float depth = texelFetch(source_depth, pos, i).r;
+ int depth_index = -1;
+ for (int j = 0; j < depth_count; j++) {
+ if (abs(depths[j] - depth) < 0.000001) {
+ depth_index = j;
+ break;
+ }
+ }
+
+ if (depth_index == -1) {
+ depths[depth_count] = depth;
+ depth_indices[depth_count] = i;
+ depth_amount[depth_count] = 1;
+ depth_count += 1;
+ } else {
+ depth_amount[depth_index] += 1;
+ }
+ }
+
+ int depth_least = 0xFFFF;
+ int best_index = 0;
+ for (int j = 0; j < depth_count; j++) {
+ if (depth_amount[j] < depth_least) {
+ best_index = depth_indices[j];
+ depth_least = depth_amount[j];
+ }
+ }
+
+ best_depth = texelFetch(source_depth, pos, best_index).r;
+ best_normal_roughness = texelFetch(source_normal_roughness, pos, best_index);
+#ifdef GIPROBE_RESOLVE
+ best_giprobe = texelFetch(source_giprobe, pos, best_index).rg;
+#endif
+
+#endif
+
+ imageStore(dest_depth, pos, vec4(best_depth));
+ imageStore(dest_normal_roughness, pos, vec4(best_normal_roughness));
+#ifdef GIPROBE_RESOLVE
+ imageStore(dest_giprobe, pos, uvec4(best_giprobe, 0, 0));
+#endif
+
+#endif
+}
diff --git a/servers/rendering/rasterizer_rd/shaders/scene_high_end.glsl b/servers/rendering/rasterizer_rd/shaders/scene_high_end.glsl
index 9f42b0f814..9a94a7914e 100644
--- a/servers/rendering/rasterizer_rd/shaders/scene_high_end.glsl
+++ b/servers/rendering/rasterizer_rd/shaders/scene_high_end.glsl
@@ -258,7 +258,6 @@ VERTEX_SHADER_CODE
}
}
#endif
-
#ifdef MODE_RENDER_MATERIAL
if (scene_data.material_uv2_mode) {
gl_Position.xy = (uv2_attrib.xy + draw_call.bake_uv2_offset) * 2.0 - 1.0;
@@ -341,11 +340,13 @@ layout(location = 4) out float depth_output_buffer;
#endif
-#ifdef MODE_RENDER_NORMAL
-layout(location = 0) out vec4 normal_output_buffer;
-#ifdef MODE_RENDER_ROUGHNESS
-layout(location = 1) out float roughness_output_buffer;
-#endif //MODE_RENDER_ROUGHNESS
+#ifdef MODE_RENDER_NORMAL_ROUGHNESS
+layout(location = 0) out vec4 normal_roughness_output_buffer;
+
+#ifdef MODE_RENDER_GIPROBE
+layout(location = 1) out uvec2 giprobe_buffer;
+#endif
+
#endif //MODE_RENDER_NORMAL
#else // RENDER DEPTH
@@ -1321,37 +1322,39 @@ void reflection_process(uint ref_index, vec3 vertex, vec3 normal, float roughnes
reflection_accum += reflection;
}
-#if !defined(USE_LIGHTMAP) && !defined(USE_VOXEL_CONE_TRACING)
- if (reflections.data[ref_index].ambient.a > 0.0) { //compute ambient using skybox
+ switch (reflections.data[ref_index].ambient_mode) {
+ case REFLECTION_AMBIENT_DISABLED: {
+ //do nothing
+ } break;
+ case REFLECTION_AMBIENT_ENVIRONMENT: {
+ //do nothing
+ vec3 local_amb_vec = (reflections.data[ref_index].local_matrix * vec4(normal, 0.0)).xyz;
- vec3 local_amb_vec = (reflections.data[ref_index].local_matrix * vec4(normal, 0.0)).xyz;
+ vec4 ambient_out;
- vec4 ambient_out;
-
- ambient_out.rgb = textureLod(samplerCubeArray(reflection_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(local_amb_vec, reflections.data[ref_index].index), MAX_ROUGHNESS_LOD).rgb;
-
- ambient_out.a = blend;
- ambient_out.rgb = mix(reflections.data[ref_index].ambient.rgb, ambient_out.rgb, reflections.data[ref_index].ambient.a);
- if (reflections.data[ref_index].params.z < 0.5) {
- ambient_out.rgb = mix(ambient_light, ambient_out.rgb, blend);
- }
+ 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
+ ambient_out.rgb = mix(ambient_light, ambient_out.rgb, blend);
+ }
- ambient_out.rgb *= ambient_out.a;
- ambient_accum += ambient_out;
- } else {
- vec4 ambient_out;
- ambient_out.a = blend;
- ambient_out.rgb = reflections.data[ref_index].ambient.rgb;
- if (reflections.data[ref_index].params.z < 0.5) {
- ambient_out.rgb = mix(ambient_light, ambient_out.rgb, blend);
- }
- ambient_out.rgb *= ambient_out.a;
- ambient_accum += ambient_out;
+ ambient_out.rgb *= ambient_out.a;
+ ambient_accum += ambient_out;
+ } break;
+ case REFLECTION_AMBIENT_COLOR: {
+ vec4 ambient_out;
+ ambient_out.a = blend;
+ ambient_out.rgb = reflections.data[ref_index].ambient;
+ if (reflections.data[ref_index].params.z < 0.5) {
+ ambient_out.rgb = mix(ambient_light, ambient_out.rgb, blend);
+ }
+ ambient_out.rgb *= ambient_out.a;
+ ambient_accum += ambient_out;
+ } break;
}
-#endif //USE_LIGHTMAP or VCT
}
-#ifdef USE_VOXEL_CONE_TRACING
+#ifdef USE_FORWARD_GI
//standard voxel cone trace
vec4 voxel_cone_trace(texture3D probe, vec3 cell_size, vec3 pos, vec3 direction, float tan_half_angle, float max_distance, float p_bias) {
@@ -1375,42 +1378,6 @@ vec4 voxel_cone_trace(texture3D probe, vec3 cell_size, vec3 pos, vec3 direction,
return color;
}
-#ifndef GI_PROBE_HIGH_QUALITY
-//faster version for 45 degrees
-
-#ifdef GI_PROBE_USE_ANISOTROPY
-
-vec4 voxel_cone_trace_anisotropic_45_degrees(texture3D probe, texture3D aniso_pos, texture3D aniso_neg, vec3 normal, vec3 cell_size, vec3 pos, vec3 direction, float tan_half_angle, float max_distance, float p_bias) {
- float dist = p_bias;
- vec4 color = vec4(0.0);
- float radius = max(0.5, tan_half_angle * dist);
- float lod_level = log2(radius * 2.0);
-
- while (dist < max_distance && color.a < 0.95) {
- vec3 uvw_pos = (pos + dist * direction) * cell_size;
- //check if outside, then break
- if (any(greaterThan(abs(uvw_pos - 0.5), vec3(0.5f + radius * cell_size)))) {
- break;
- }
-
- vec4 scolor = textureLod(sampler3D(probe, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uvw_pos, lod_level);
- vec3 aniso_neg = textureLod(sampler3D(aniso_neg, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uvw_pos, lod_level).rgb;
- vec3 aniso_pos = textureLod(sampler3D(aniso_pos, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uvw_pos, lod_level).rgb;
-
- scolor.rgb *= dot(max(vec3(0.0), (normal * aniso_pos)), vec3(1.0)) + dot(max(vec3(0.0), (-normal * aniso_neg)), vec3(1.0));
- lod_level += 1.0;
-
- float a = (1.0 - color.a);
- scolor *= a;
- color += scolor;
- dist += radius;
- radius = max(0.5, tan_half_angle * dist);
- }
-
- return color;
-}
-#else
-
vec4 voxel_cone_trace_45_degrees(texture3D probe, vec3 cell_size, vec3 pos, vec3 direction, float tan_half_angle, float max_distance, float p_bias) {
float dist = p_bias;
vec4 color = vec4(0.0);
@@ -1437,41 +1404,6 @@ vec4 voxel_cone_trace_45_degrees(texture3D probe, vec3 cell_size, vec3 pos, vec3
return color;
}
-#endif
-
-#elif defined(GI_PROBE_USE_ANISOTROPY)
-
-//standard voxel cone trace
-vec4 voxel_cone_trace_anisotropic(texture3D probe, texture3D aniso_pos, texture3D aniso_neg, vec3 normal, vec3 cell_size, vec3 pos, vec3 direction, float tan_half_angle, float max_distance, float p_bias) {
- float dist = p_bias;
- vec4 color = vec4(0.0);
-
- while (dist < max_distance && color.a < 0.95) {
- float diameter = max(1.0, 2.0 * tan_half_angle * dist);
- vec3 uvw_pos = (pos + dist * direction) * cell_size;
- float half_diameter = diameter * 0.5;
- //check if outside, then break
- if (any(greaterThan(abs(uvw_pos - 0.5), vec3(0.5f + half_diameter * cell_size)))) {
- break;
- }
- float log2_diameter = log2(diameter);
- vec4 scolor = textureLod(sampler3D(probe, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uvw_pos, log2_diameter);
- vec3 aniso_neg = textureLod(sampler3D(aniso_neg, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uvw_pos, log2_diameter).rgb;
- vec3 aniso_pos = textureLod(sampler3D(aniso_pos, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uvw_pos, log2_diameter).rgb;
-
- scolor.rgb *= dot(max(vec3(0.0), (normal * aniso_pos)), vec3(1.0)) + dot(max(vec3(0.0), (-normal * aniso_neg)), vec3(1.0));
-
- float a = (1.0 - color.a);
- scolor *= a;
- color += scolor;
- dist += half_diameter;
- }
-
- return color;
-}
-
-#endif
-
void gi_probe_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3 normal_xform, float roughness, vec3 ambient, vec3 environment, inout vec4 out_spec, inout vec4 out_diff) {
position = (gi_probes.data[index].xform * vec4(position, 1.0)).xyz;
ref_vec = normalize((gi_probes.data[index].xform * vec4(ref_vec, 0.0)).xyz);
@@ -1493,31 +1425,6 @@ void gi_probe_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3
//radiance
-#ifdef GI_PROBE_HIGH_QUALITY
-
-#define MAX_CONE_DIRS 6
- vec3 cone_dirs[MAX_CONE_DIRS] = vec3[](
- vec3(0.0, 0.0, 1.0),
- vec3(0.866025, 0.0, 0.5),
- vec3(0.267617, 0.823639, 0.5),
- vec3(-0.700629, 0.509037, 0.5),
- vec3(-0.700629, -0.509037, 0.5),
- vec3(0.267617, -0.823639, 0.5));
-
- float cone_weights[MAX_CONE_DIRS] = float[](0.25, 0.15, 0.15, 0.15, 0.15, 0.15);
- float cone_angle_tan = 0.577;
-
-#elif defined(GI_PROBE_LOW_QUALITY)
-
-#define MAX_CONE_DIRS 1
-
- vec3 cone_dirs[MAX_CONE_DIRS] = vec3[](
- vec3(0.0, 0.0, 1.0));
-
- float cone_weights[MAX_CONE_DIRS] = float[](1.0);
- float cone_angle_tan = 4; //~76 degrees
-#else // MEDIUM QUALITY
-
#define MAX_CONE_DIRS 4
vec3 cone_dirs[MAX_CONE_DIRS] = vec3[](
@@ -1529,31 +1436,13 @@ void gi_probe_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3
float cone_weights[MAX_CONE_DIRS] = float[](0.25, 0.25, 0.25, 0.25);
float cone_angle_tan = 0.98269;
-#endif
vec3 light = vec3(0.0);
for (int i = 0; i < MAX_CONE_DIRS; i++) {
vec3 dir = normalize((gi_probes.data[index].xform * vec4(normal_xform * cone_dirs[i], 0.0)).xyz);
-#if defined(GI_PROBE_HIGH_QUALITY) || defined(GI_PROBE_LOW_QUALITY)
-
-#ifdef GI_PROBE_USE_ANISOTROPY
- vec4 cone_light = voxel_cone_trace_anisotropic(gi_probe_textures[gi_probes.data[index].texture_slot], gi_probe_textures[gi_probes.data[index].texture_slot + 1], gi_probe_textures[gi_probes.data[index].texture_slot + 2], normalize(mix(dir, normal, gi_probes.data[index].anisotropy_strength)), cell_size, position, dir, cone_angle_tan, max_distance, gi_probes.data[index].bias);
-#else
-
- vec4 cone_light = voxel_cone_trace(gi_probe_textures[gi_probes.data[index].texture_slot], cell_size, position, dir, cone_angle_tan, max_distance, gi_probes.data[index].bias);
-
-#endif // GI_PROBE_USE_ANISOTROPY
-
-#else
-
-#ifdef GI_PROBE_USE_ANISOTROPY
- vec4 cone_light = voxel_cone_trace_anisotropic_45_degrees(gi_probe_textures[gi_probes.data[index].texture_slot], gi_probe_textures[gi_probes.data[index].texture_slot + 1], gi_probe_textures[gi_probes.data[index].texture_slot + 2], normalize(mix(dir, normal, gi_probes.data[index].anisotropy_strength)), cell_size, position, dir, cone_angle_tan, max_distance, gi_probes.data[index].bias);
-#else
- vec4 cone_light = voxel_cone_trace_45_degrees(gi_probe_textures[gi_probes.data[index].texture_slot], cell_size, position, dir, cone_angle_tan, max_distance, gi_probes.data[index].bias);
-#endif // GI_PROBE_USE_ANISOTROPY
+ vec4 cone_light = voxel_cone_trace_45_degrees(gi_probe_textures[index], cell_size, position, dir, cone_angle_tan, max_distance, gi_probes.data[index].bias);
-#endif
if (gi_probes.data[index].blend_ambient) {
cone_light.rgb = mix(ambient, cone_light.rgb, min(1.0, cone_light.a / 0.95));
}
@@ -1562,33 +1451,10 @@ void gi_probe_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3
}
light *= gi_probes.data[index].dynamic_range;
-
- if (gi_probes.data[index].ambient_occlusion > 0.001) {
- float size = 1.0 + gi_probes.data[index].ambient_occlusion_size * 7.0;
-
- float taps, blend;
- blend = modf(size, taps);
- float ao = 0.0;
- for (float i = 1.0; i <= taps; i++) {
- vec3 ofs = (position + normal * (i * 0.5 + 1.0)) * cell_size;
- ao += textureLod(sampler3D(gi_probe_textures[gi_probes.data[index].texture_slot], material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), ofs, i - 1.0).a * i;
- }
-
- if (blend > 0.001) {
- vec3 ofs = (position + normal * ((taps + 1.0) * 0.5 + 1.0)) * cell_size;
- ao += textureLod(sampler3D(gi_probe_textures[gi_probes.data[index].texture_slot], material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), ofs, taps).a * (taps + 1.0) * blend;
- }
-
- ao = 1.0 - min(1.0, ao);
-
- light = mix(scene_data.ao_color.rgb, light, mix(1.0, ao, gi_probes.data[index].ambient_occlusion));
- }
-
out_diff += vec4(light * blend, blend);
//irradiance
-#ifndef GI_PROBE_LOW_QUALITY
- vec4 irr_light = voxel_cone_trace(gi_probe_textures[gi_probes.data[index].texture_slot], cell_size, position, ref_vec, tan(roughness * 0.5 * M_PI * 0.99), max_distance, gi_probes.data[index].bias);
+ vec4 irr_light = voxel_cone_trace(gi_probe_textures[index], cell_size, position, ref_vec, tan(roughness * 0.5 * M_PI * 0.99), max_distance, gi_probes.data[index].bias);
if (gi_probes.data[index].blend_ambient) {
irr_light.rgb = mix(environment, irr_light.rgb, min(1.0, irr_light.a / 0.95));
}
@@ -1596,10 +1462,142 @@ void gi_probe_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3
//irr_light=vec3(0.0);
out_spec += vec4(irr_light.rgb * blend, blend);
-#endif
}
-#endif //USE_VOXEL_CONE_TRACING
+#endif //USE_FORWARD_GI
+
+vec2 octahedron_wrap(vec2 v) {
+ vec2 signVal;
+ signVal.x = v.x >= 0.0 ? 1.0 : -1.0;
+ signVal.y = v.y >= 0.0 ? 1.0 : -1.0;
+ return (1.0 - abs(v.yx)) * signVal;
+}
+
+vec2 octahedron_encode(vec3 n) {
+ // https://twitter.com/Stubbesaurus/status/937994790553227264
+ n /= (abs(n.x) + abs(n.y) + abs(n.z));
+ n.xy = n.z >= 0.0 ? n.xy : octahedron_wrap(n.xy);
+ n.xy = n.xy * 0.5 + 0.5;
+ return n.xy;
+}
+
+void sdfgi_process(uint cascade, vec3 cascade_pos, vec3 cam_pos, vec3 cam_normal, vec3 cam_specular_normal, bool use_specular, float roughness, out vec3 diffuse_light, out vec3 specular_light, out float blend) {
+ cascade_pos += cam_normal * sdfgi.normal_bias;
+
+ vec3 base_pos = floor(cascade_pos);
+ //cascade_pos += mix(vec3(0.0),vec3(0.01),lessThan(abs(cascade_pos-base_pos),vec3(0.01))) * cam_normal;
+ ivec3 probe_base_pos = ivec3(base_pos);
+
+ vec4 diffuse_accum = vec4(0.0);
+ vec3 specular_accum;
+
+ ivec3 tex_pos = ivec3(probe_base_pos.xy, int(cascade));
+ tex_pos.x += probe_base_pos.z * sdfgi.probe_axis_size;
+ tex_pos.xy = tex_pos.xy * (SDFGI_OCT_SIZE + 2) + ivec2(1);
+
+ vec3 diffuse_posf = (vec3(tex_pos) + vec3(octahedron_encode(cam_normal) * float(SDFGI_OCT_SIZE), 0.0)) * sdfgi.lightprobe_tex_pixel_size;
+
+ vec3 specular_posf;
+
+ if (use_specular) {
+ specular_accum = vec3(0.0);
+ specular_posf = (vec3(tex_pos) + vec3(octahedron_encode(cam_specular_normal) * float(SDFGI_OCT_SIZE), 0.0)) * sdfgi.lightprobe_tex_pixel_size;
+ }
+
+ vec4 light_accum = vec4(0.0);
+ float weight_accum = 0.0;
+
+ for (uint j = 0; j < 8; j++) {
+ ivec3 offset = (ivec3(j) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1);
+ ivec3 probe_posi = probe_base_pos;
+ probe_posi += offset;
+
+ // Compute weight
+
+ vec3 probe_pos = vec3(probe_posi);
+ vec3 probe_to_pos = cascade_pos - probe_pos;
+ vec3 probe_dir = normalize(-probe_to_pos);
+
+ vec3 trilinear = vec3(1.0) - abs(probe_to_pos);
+ float weight = trilinear.x * trilinear.y * trilinear.z * max(0.005, dot(cam_normal, probe_dir));
+
+ // Compute lightprobe occlusion
+
+ if (sdfgi.use_occlusion) {
+ ivec3 occ_indexv = abs((sdfgi.cascades[cascade].probe_world_offset + probe_posi) & ivec3(1, 1, 1)) * ivec3(1, 2, 4);
+ vec4 occ_mask = mix(vec4(0.0), vec4(1.0), equal(ivec4(occ_indexv.x | occ_indexv.y), ivec4(0, 1, 2, 3)));
+
+ vec3 occ_pos = clamp(cascade_pos, probe_pos - sdfgi.occlusion_clamp, probe_pos + sdfgi.occlusion_clamp) * sdfgi.probe_to_uvw;
+ occ_pos.z += float(cascade);
+ if (occ_indexv.z != 0) { //z bit is on, means index is >=4, so make it switch to the other half of textures
+ occ_pos.x += 1.0;
+ }
+
+ occ_pos *= sdfgi.occlusion_renormalize;
+ float occlusion = dot(textureLod(sampler3D(sdfgi_occlusion_cascades, material_samplers[SAMPLER_LINEAR_CLAMP]), occ_pos, 0.0), occ_mask);
+
+ weight *= max(occlusion, 0.01);
+ }
+
+ // Compute lightprobe texture position
+
+ vec3 diffuse;
+ vec3 pos_uvw = diffuse_posf;
+ pos_uvw.xy += vec2(offset.xy) * sdfgi.lightprobe_uv_offset.xy;
+ pos_uvw.x += float(offset.z) * sdfgi.lightprobe_uv_offset.z;
+ diffuse = textureLod(sampler2DArray(sdfgi_lightprobe_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), pos_uvw, 0.0).rgb;
+
+ diffuse_accum += vec4(diffuse * weight, weight);
+
+ if (use_specular) {
+ vec3 specular = vec3(0.0);
+ vec3 pos_uvw = specular_posf;
+ pos_uvw.xy += vec2(offset.xy) * sdfgi.lightprobe_uv_offset.xy;
+ pos_uvw.x += float(offset.z) * sdfgi.lightprobe_uv_offset.z;
+ if (roughness < 0.99) {
+ specular = textureLod(sampler2DArray(sdfgi_lightprobe_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), pos_uvw + vec3(0, 0, float(sdfgi.max_cascades)), 0.0).rgb;
+ }
+ if (roughness > 0.5) {
+ specular = mix(specular, textureLod(sampler2DArray(sdfgi_lightprobe_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), pos_uvw, 0.0).rgb, (roughness - 0.5) * 2.0);
+ }
+
+ specular_accum += specular * weight;
+ }
+ }
+
+ if (diffuse_accum.a > 0.0) {
+ diffuse_accum.rgb /= diffuse_accum.a;
+ }
+
+ diffuse_light = diffuse_accum.rgb;
+
+ if (use_specular) {
+ if (diffuse_accum.a > 0.0) {
+ specular_accum /= diffuse_accum.a;
+ }
+
+ specular_light = specular_accum;
+ }
+
+ {
+ //process blend
+ float blend_from = (float(sdfgi.probe_axis_size - 1) / 2.0) - 2.5;
+ float blend_to = blend_from + 2.0;
+
+ vec3 inner_pos = cam_pos * sdfgi.cascades[cascade].to_probe;
+
+ float len = length(inner_pos);
+
+ inner_pos = abs(normalize(inner_pos));
+ len *= max(inner_pos.x, max(inner_pos.y, inner_pos.z));
+
+ if (len >= blend_from) {
+ blend = smoothstep(blend_from, blend_to, len);
+ } else {
+ blend = 0.0;
+ }
+ }
+}
#endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED)
@@ -1812,6 +1810,15 @@ FRAGMENT_SHADER_CODE
#endif //not render depth
/////////////////////// LIGHTING //////////////////////////////
+ if (scene_data.roughness_limiter_enabled) {
+ //http://www.jp.square-enix.com/tech/library/pdf/ImprovedGeometricSpecularAA.pdf
+ float roughness2 = roughness * roughness;
+ vec3 dndu = dFdx(normal), dndv = dFdx(normal);
+ float variance = scene_data.roughness_limiter_amount * (dot(dndu, dndu) + dot(dndv, dndv));
+ float kernelRoughness2 = min(2.0 * variance, scene_data.roughness_limiter_limit); //limit effect
+ float filteredRoughness2 = min(1.0, roughness2 + kernelRoughness2);
+ roughness = sqrt(filteredRoughness2);
+ }
//apply energy conservation
vec3 specular_light = vec3(0.0, 0.0, 0.0);
@@ -1820,11 +1827,6 @@ FRAGMENT_SHADER_CODE
#if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED)
- if (scene_data.roughness_limiter_enabled) {
- float limit = texelFetch(sampler2D(roughness_buffer, material_samplers[SAMPLER_NEAREST_CLAMP]), ivec2(gl_FragCoord.xy), 0).r;
- roughness = max(roughness, limit);
- }
-
if (scene_data.use_reflection_cubemap) {
vec3 ref_vec = reflect(-view, normal);
ref_vec = scene_data.radiance_inverse_xform * ref_vec;
@@ -1871,7 +1873,6 @@ FRAGMENT_SHADER_CODE
#endif
#if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED)
- //gi probes
#ifdef USE_LIGHTMAP
@@ -1928,10 +1929,80 @@ FRAGMENT_SHADER_CODE
ambient_light += textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw, 0.0).rgb;
}
}
-#endif
- //lightmap capture
+#elif defined(USE_FORWARD_GI)
+
+ 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;
+ vec3 cam_normal = mat3(scene_data.camera_matrix) * normal;
+ vec3 cam_reflection = mat3(scene_data.camera_matrix) * reflect(-view, normal);
+
+ //apply y-mult
+ cam_pos.y *= sdfgi.y_mult;
+ cam_normal.y *= sdfgi.y_mult;
+ cam_normal = normalize(cam_normal);
+ cam_reflection.y *= sdfgi.y_mult;
+ cam_normal = normalize(cam_normal);
+ cam_reflection = normalize(cam_reflection);
+
+ vec4 light_accum = vec4(0.0);
+ float weight_accum = 0.0;
+
+ vec4 light_blend_accum = vec4(0.0);
+ float weight_blend_accum = 0.0;
+
+ float blend = -1.0;
+
+ // helper constants, compute once
+
+ uint cascade = 0xFFFFFFFF;
+ vec3 cascade_pos;
+ vec3 cascade_normal;
+
+ for (uint i = 0; i < sdfgi.max_cascades; i++) {
+ cascade_pos = (cam_pos - sdfgi.cascades[i].position) * sdfgi.cascades[i].to_probe;
+
+ if (any(lessThan(cascade_pos, vec3(0.0))) || any(greaterThanEqual(cascade_pos, sdfgi.cascade_probe_size))) {
+ continue; //skip cascade
+ }
+
+ cascade = i;
+ break;
+ }
+
+ if (cascade < SDFGI_MAX_CASCADES) {
+ bool use_specular = true;
+ float blend;
+ vec3 diffuse, specular;
+ sdfgi_process(cascade, cascade_pos, cam_pos, cam_normal, cam_reflection, use_specular, roughness, diffuse, specular, blend);
+
+ if (blend > 0.0) {
+ //blend
+ if (cascade == sdfgi.max_cascades - 1) {
+ diffuse = mix(diffuse, ambient_light, blend);
+ if (use_specular) {
+ specular = mix(specular, specular_light, blend);
+ }
+ } else {
+ vec3 diffuse2, specular2;
+ float blend2;
+ cascade_pos = (cam_pos - sdfgi.cascades[cascade + 1].position) * sdfgi.cascades[cascade + 1].to_probe;
+ sdfgi_process(cascade + 1, cascade_pos, cam_pos, cam_normal, cam_reflection, use_specular, roughness, diffuse2, specular2, blend2);
+ diffuse = mix(diffuse, diffuse2, blend);
+ if (use_specular) {
+ specular = mix(specular, specular2, blend);
+ }
+ }
+ }
+
+ ambient_light = diffuse;
+ if (use_specular) {
+ specular_light = specular;
+ }
+ }
+ }
-#ifdef USE_VOXEL_CONE_TRACING
if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_GIPROBE)) { // process giprobes
uint index1 = instances.data[instance_index].gi_offset & 0xFFFF;
@@ -1963,6 +2034,56 @@ FRAGMENT_SHADER_CODE
specular_light = spec_accum.rgb;
ambient_light = amb_accum.rgb;
}
+#else
+ if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_GI_BUFFERS)) { //use GI buffers
+
+ ivec2 coord;
+
+ if (scene_data.gi_upscale_for_msaa) {
+ /*
+ //find the closest depth to upscale from, based on neighbours
+ ivec2 base_coord = ivec2(gl_FragCoord.xy);
+ float z_dist = gl_FragCoord.z;
+ ivec2 closest_coord = base_coord;
+ float closest_z_dist = abs(texelFetch(sampler2D(depth_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), base_coord,0).r-z_dist);
+
+ 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_z_dist = abs(texelFetch(sampler2D(depth_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), neighbour_coord,0).r-z_dist);
+ if (neighbour_z_dist < closest_z_dist) {
+ closest_z_dist = neighbour_z_dist;
+ closest_coord = neighbour_coord;
+ }
+ }
+
+*/
+ 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);
+
+ 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);
+ if (neighbour_ang > closest_ang) {
+ closest_ang = neighbour_ang;
+ closest_coord = neighbour_coord;
+ }
+ }
+
+ coord = closest_coord;
+
+ } else {
+ coord = ivec2(gl_FragCoord.xy);
+ }
+
+ 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);
+
+ ambient_light = mix(ambient_light, buffer_ambient.rgb, buffer_ambient.a);
+ specular_light = mix(specular_light, buffer_reflection.rgb, buffer_reflection.a);
+ }
#endif
{ // process reflections
@@ -2376,6 +2497,93 @@ FRAGMENT_SHADER_CODE
#ifdef MODE_RENDER_DEPTH
+#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) * normal;
+
+ for (uint i = 0; i < 6; i++) {
+ if (dot(cam_normal, aniso_dir[i]) > 0.001) {
+ facing_bits |= (1 << i);
+ }
+ }
+
+ imageAtomicOr(geom_facing_grid, grid_pos, facing_bits); //store facing bits
+
+ if (length(emission) > 0.001) {
+ float lumas[6];
+ vec3 light_total = vec3(0);
+
+ for (int i = 0; i < 6; i++) {
+ float strength = max(0.0, dot(cam_normal, aniso_dir[i]));
+ vec3 light = emission * strength;
+ light_total += light;
+ lumas[i] = max(light.r, max(light.g, light.b));
+ }
+
+ float luma_total = max(light_total.r, max(light_total.g, light_total.b));
+
+ uint light_aniso = 0;
+
+ for (int i = 0; i < 6; i++) {
+ light_aniso |= min(31, uint((lumas[i] / luma_total) * 31.0)) << (i * 5);
+ }
+
+ //compress to RGBE9995 to save space
+
+ const float pow2to9 = 512.0f;
+ const float B = 15.0f;
+ const float N = 9.0f;
+ const float LN2 = 0.6931471805599453094172321215;
+
+ float cRed = clamp(light_total.r, 0.0, 65408.0);
+ float cGreen = clamp(light_total.g, 0.0, 65408.0);
+ float cBlue = clamp(light_total.b, 0.0, 65408.0);
+
+ float cMax = max(cRed, max(cGreen, cBlue));
+
+ float expp = max(-B - 1.0f, floor(log(cMax) / LN2)) + 1.0f + B;
+
+ float sMax = floor((cMax / pow(2.0f, expp - B - N)) + 0.5f);
+
+ float exps = expp + 1.0f;
+
+ if (0.0 <= sMax && sMax < pow2to9) {
+ exps = expp;
+ }
+
+ float sRed = floor((cRed / pow(2.0f, exps - B - N)) + 0.5f);
+ float sGreen = floor((cGreen / pow(2.0f, exps - B - N)) + 0.5f);
+ float sBlue = floor((cBlue / pow(2.0f, exps - B - N)) + 0.5f);
+ //store as 8985 to have 2 extra neighbour bits
+ uint light_rgbe = ((uint(sRed) & 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));
+ }
+ }
+
+#endif
+
#ifdef MODE_RENDER_MATERIAL
albedo_output_buffer.rgb = albedo;
@@ -2398,11 +2606,21 @@ FRAGMENT_SHADER_CODE
emission_output_buffer.a = 0.0;
#endif
-#ifdef MODE_RENDER_NORMAL
- normal_output_buffer = vec4(normal * 0.5 + 0.5, 0.0);
-#ifdef MODE_RENDER_ROUGHNESS
- roughness_output_buffer = roughness;
-#endif //MODE_RENDER_ROUGHNESS
+#ifdef MODE_RENDER_NORMAL_ROUGHNESS
+ normal_roughness_output_buffer = vec4(normal * 0.5 + 0.5, roughness);
+
+#ifdef MODE_RENDER_GIPROBE
+ 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
//nothing happens, so a tree-ssa optimizer will result in no fragment shader :)
@@ -2455,7 +2673,6 @@ FRAGMENT_SHADER_CODE
#endif
diffuse_buffer = vec4(emission + diffuse_light + ambient_light, sss_strength);
specular_buffer = vec4(specular_light, metallic);
-
#endif
#else //MODE_MULTIPLE_RENDER_TARGETS
diff --git a/servers/rendering/rasterizer_rd/shaders/scene_high_end_inc.glsl b/servers/rendering/rasterizer_rd/shaders/scene_high_end_inc.glsl
index 1cac12406a..1244599097 100644
--- a/servers/rendering/rasterizer_rd/shaders/scene_high_end_inc.glsl
+++ b/servers/rendering/rasterizer_rd/shaders/scene_high_end_inc.glsl
@@ -1,6 +1,8 @@
#define M_PI 3.14159265359
#define ROUGHNESS_MAX_LOD 5
+#define MAX_GI_PROBES 8
+
layout(push_constant, binding = 0, std430) uniform DrawCall {
uint instance_index;
uint pad; //16 bits minimum size
@@ -27,6 +29,8 @@ 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;
@@ -76,11 +80,19 @@ layout(set = 0, binding = 3, std140) uniform SceneData {
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;
- uint pad_material0;
- uint pad_material1;
- uint pad_material2;
+
+ ivec3 sdf_size;
+ bool gi_upscale_for_msaa;
#if 0
vec4 ambient_light_color;
@@ -120,6 +132,8 @@ layout(set = 0, binding = 3, std140) uniform SceneData {
scene_data;
+#define INSTANCE_FLAGS_USE_GI_BUFFERS (1 << 6)
+#define INSTANCE_FLAGS_USE_SDFGI (1 << 7)
#define INSTANCE_FLAGS_USE_LIGHTMAP_CAPTURE (1 << 8)
#define INSTANCE_FLAGS_USE_LIGHTMAP (1 << 9)
#define INSTANCE_FLAGS_USE_SH_LIGHTMAP (1 << 10)
@@ -175,13 +189,18 @@ layout(set = 0, binding = 5, std430) restrict readonly buffer Lights {
}
lights;
+#define REFLECTION_AMBIENT_DISABLED 0
+#define REFLECTION_AMBIENT_ENVIRONMENT 1
+#define REFLECTION_AMBIENT_COLOR 2
+
struct ReflectionData {
vec3 box_extents;
float index;
vec3 box_offset;
uint mask;
vec4 params; // intensity, 0, interior , boxproject
- vec4 ambient; // ambient color, energy
+ vec3 ambient; // ambient color
+ uint 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
};
@@ -229,29 +248,6 @@ layout(set = 0, binding = 7, std140) uniform DirectionalLights {
}
directional_lights;
-struct GIProbeData {
- mat4 xform;
- vec3 bounds;
- float dynamic_range;
-
- float bias;
- float normal_bias;
- bool blend_ambient;
- uint texture_slot;
-
- float anisotropy_strength;
- float ambient_occlusion;
- float ambient_occlusion_size;
- uint pad2;
-};
-
-layout(set = 0, binding = 8, std140) uniform GIProbes {
- GIProbeData data[MAX_GI_PROBES];
-}
-gi_probes;
-
-layout(set = 0, binding = 9) uniform texture3D gi_probe_textures[MAX_GI_PROBE_TEXTURES];
-
#define LIGHTMAP_FLAG_USE_DIRECTION 1
#define LIGHTMAP_FLAG_USE_SPECULAR_DIRECTION 2
@@ -319,6 +315,41 @@ layout(set = 0, binding = 19, std430) restrict readonly buffer GlobalVariableDat
}
global_variables;
+struct SDFGIProbeCascadeData {
+ vec3 position;
+ float to_probe;
+ ivec3 probe_world_offset;
+ float to_cell; // 1/bounds * grid_size
+};
+
+layout(set = 0, binding = 20, std140) uniform SDFGI {
+ vec3 grid_size;
+ uint max_cascades;
+
+ bool use_occlusion;
+ int probe_axis_size;
+ float probe_to_uvw;
+ float normal_bias;
+
+ vec3 lightprobe_tex_pixel_size;
+ float energy;
+
+ vec3 lightprobe_uv_offset;
+ float y_mult;
+
+ vec3 occlusion_clamp;
+ uint pad3;
+
+ vec3 occlusion_renormalize;
+ uint pad4;
+
+ vec3 cascade_probe_size;
+ uint pad5;
+
+ SDFGIProbeCascadeData cascades[SDFGI_MAX_CASCADES];
+}
+sdfgi;
+
// decal atlas
/* Set 1, Radiance */
@@ -339,13 +370,57 @@ layout(set = 2, binding = 0) uniform textureCubeArray reflection_atlas;
layout(set = 2, binding = 1) uniform texture2D shadow_atlas;
+layout(set = 2, binding = 2) uniform texture3D gi_probe_textures[MAX_GI_PROBES];
+
/* Set 3, Render Buffers */
+#ifdef MODE_RENDER_SDF
+
+layout(r16ui, set = 3, binding = 0) uniform restrict writeonly uimage3D albedo_volume_grid;
+layout(r32ui, set = 3, binding = 1) uniform restrict writeonly uimage3D emission_grid;
+layout(r32ui, set = 3, binding = 2) uniform restrict writeonly uimage3D emission_aniso_grid;
+layout(r32ui, set = 3, binding = 3) uniform restrict uimage3D geom_facing_grid;
+
+//still need to be present for shaders that use it, so remap them to something
+#define depth_buffer shadow_atlas
+#define color_buffer shadow_atlas
+#define normal_roughness_buffer shadow_atlas
+
+#else
+
layout(set = 3, binding = 0) uniform texture2D depth_buffer;
layout(set = 3, binding = 1) uniform texture2D color_buffer;
-layout(set = 3, binding = 2) uniform texture2D normal_buffer;
-layout(set = 3, binding = 3) uniform texture2D roughness_buffer;
+layout(set = 3, binding = 2) uniform texture2D normal_roughness_buffer;
layout(set = 3, binding = 4) uniform texture2D ao_buffer;
+layout(set = 3, binding = 5) uniform texture2D ambient_buffer;
+layout(set = 3, binding = 6) uniform texture2D reflection_buffer;
+
+layout(set = 3, binding = 7) uniform texture2DArray sdfgi_lightprobe_texture;
+
+layout(set = 3, binding = 8) uniform texture3D sdfgi_occlusion_cascades;
+
+struct GIProbeData {
+ mat4 xform;
+ vec3 bounds;
+ float dynamic_range;
+
+ float bias;
+ float normal_bias;
+ bool blend_ambient;
+ uint texture_slot;
+
+ float anisotropy_strength;
+ float ambient_occlusion;
+ float ambient_occlusion_size;
+ uint pad2;
+};
+
+layout(set = 3, binding = 9, std140) uniform GIProbes {
+ GIProbeData data[MAX_GI_PROBES];
+}
+gi_probes;
+
+#endif
/* Set 4 Skeleton & Instancing (Multimesh) */
diff --git a/servers/rendering/rasterizer_rd/shaders/screen_space_reflection.glsl b/servers/rendering/rasterizer_rd/shaders/screen_space_reflection.glsl
index 084f28d932..a8ee33a664 100644
--- a/servers/rendering/rasterizer_rd/shaders/screen_space_reflection.glsl
+++ b/servers/rendering/rasterizer_rd/shaders/screen_space_reflection.glsl
@@ -12,11 +12,8 @@ layout(rgba16f, set = 1, binding = 0) uniform restrict writeonly image2D ssr_ima
#ifdef MODE_ROUGH
layout(r8, set = 1, binding = 1) uniform restrict writeonly image2D blur_radius_image;
#endif
-layout(rgba8, set = 2, binding = 0) uniform restrict readonly image2D source_normal;
+layout(rgba8, set = 2, binding = 0) uniform restrict readonly image2D source_normal_roughness;
layout(set = 3, binding = 0) uniform sampler2D source_metallic;
-#ifdef MODE_ROUGH
-layout(set = 3, binding = 1) uniform sampler2D source_roughness;
-#endif
layout(push_constant, binding = 2, std430) uniform Params {
vec4 proj_info;
@@ -75,7 +72,8 @@ void main() {
// World space point being shaded
vec3 vertex = reconstructCSPosition(uv * vec2(params.screen_size), base_depth);
- vec3 normal = imageLoad(source_normal, ssC).xyz * 2.0 - 1.0;
+ vec4 normal_roughness = imageLoad(source_normal_roughness, ssC);
+ vec3 normal = normal_roughness.xyz * 2.0 - 1.0;
normal = normalize(normal);
normal.y = -normal.y; //because this code reads flipped
@@ -208,7 +206,7 @@ void main() {
// if roughness is enabled, do screen space cone tracing
float blur_radius = 0.0;
- float roughness = texelFetch(source_roughness, ssC << 1, 0).r;
+ float roughness = normal_roughness.w;
if (roughness > 0.001) {
float cone_angle = min(roughness, 0.999) * M_PI * 0.5;
diff --git a/servers/rendering/rasterizer_rd/shaders/sdfgi_debug.glsl b/servers/rendering/rasterizer_rd/shaders/sdfgi_debug.glsl
new file mode 100644
index 0000000000..813ea29fa1
--- /dev/null
+++ b/servers/rendering/rasterizer_rd/shaders/sdfgi_debug.glsl
@@ -0,0 +1,275 @@
+#[compute]
+
+#version 450
+
+VERSION_DEFINES
+
+layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in;
+
+#define MAX_CASCADES 8
+
+layout(set = 0, binding = 1) uniform texture3D sdf_cascades[MAX_CASCADES];
+layout(set = 0, binding = 2) uniform texture3D light_cascades[MAX_CASCADES];
+layout(set = 0, binding = 3) uniform texture3D aniso0_cascades[MAX_CASCADES];
+layout(set = 0, binding = 4) uniform texture3D aniso1_cascades[MAX_CASCADES];
+layout(set = 0, binding = 5) uniform texture3D occlusion_texture;
+
+layout(set = 0, binding = 8) uniform sampler linear_sampler;
+
+struct CascadeData {
+ vec3 offset; //offset of (0,0,0) in world coordinates
+ float to_cell; // 1/bounds * grid_size
+ ivec3 probe_world_offset;
+ uint pad;
+};
+
+layout(set = 0, binding = 9, std140) uniform Cascades {
+ CascadeData data[MAX_CASCADES];
+}
+cascades;
+
+layout(rgba16f, set = 0, binding = 10) uniform restrict writeonly image2D screen_buffer;
+
+layout(set = 0, binding = 11) uniform texture2DArray lightprobe_texture;
+
+layout(push_constant, binding = 0, std430) uniform Params {
+ vec3 grid_size;
+ uint max_cascades;
+
+ ivec2 screen_size;
+ bool use_occlusion;
+ float y_mult;
+
+ vec3 cam_extent;
+ int probe_axis_size;
+
+ mat4 cam_transform;
+}
+params;
+
+vec3 linear_to_srgb(vec3 color) {
+ //if going to srgb, clamp from 0 to 1.
+ color = clamp(color, vec3(0.0), vec3(1.0));
+ const vec3 a = vec3(0.055f);
+ return mix((vec3(1.0f) + a) * pow(color.rgb, vec3(1.0f / 2.4f)) - a, 12.92f * color.rgb, lessThan(color.rgb, vec3(0.0031308f)));
+}
+
+vec2 octahedron_wrap(vec2 v) {
+ vec2 signVal;
+ signVal.x = v.x >= 0.0 ? 1.0 : -1.0;
+ signVal.y = v.y >= 0.0 ? 1.0 : -1.0;
+ return (1.0 - abs(v.yx)) * signVal;
+}
+
+vec2 octahedron_encode(vec3 n) {
+ // https://twitter.com/Stubbesaurus/status/937994790553227264
+ n /= (abs(n.x) + abs(n.y) + abs(n.z));
+ n.xy = n.z >= 0.0 ? n.xy : octahedron_wrap(n.xy);
+ n.xy = n.xy * 0.5 + 0.5;
+ return n.xy;
+}
+
+void main() {
+ // Pixel being shaded
+ ivec2 screen_pos = ivec2(gl_GlobalInvocationID.xy);
+ if (any(greaterThanEqual(screen_pos, params.screen_size))) { //too large, do nothing
+ return;
+ }
+
+ vec3 ray_pos;
+ vec3 ray_dir;
+ {
+ ray_pos = params.cam_transform[3].xyz;
+
+ ray_dir.xy = params.cam_extent.xy * ((vec2(screen_pos) / vec2(params.screen_size)) * 2.0 - 1.0);
+ ray_dir.z = params.cam_extent.z;
+
+ ray_dir = normalize(mat3(params.cam_transform) * ray_dir);
+ }
+
+ ray_pos.y *= params.y_mult;
+ ray_dir.y *= params.y_mult;
+ ray_dir = normalize(ray_dir);
+
+ vec3 pos_to_uvw = 1.0 / params.grid_size;
+
+ vec3 light = vec3(0.0);
+ float blend = 0.0;
+
+#if 1
+ vec3 inv_dir = 1.0 / ray_dir;
+
+ float rough = 0.5;
+ bool hit = false;
+
+ for (uint i = 0; i < params.max_cascades; i++) {
+ //convert to local bounds
+ vec3 pos = ray_pos - cascades.data[i].offset;
+ pos *= cascades.data[i].to_cell;
+
+ // Should never happen for debug, since we start mostly at the bounds center,
+ // but add anyway.
+ //if (any(lessThan(pos,vec3(0.0))) || any(greaterThanEqual(pos,params.grid_size))) {
+ // continue; //already past bounds for this cascade, goto next
+ //}
+
+ //find maximum advance distance (until reaching bounds)
+ vec3 t0 = -pos * inv_dir;
+ vec3 t1 = (params.grid_size - pos) * inv_dir;
+ vec3 tmax = max(t0, t1);
+ float max_advance = min(tmax.x, min(tmax.y, tmax.z));
+
+ float advance = 0.0;
+ vec3 uvw;
+ hit = false;
+
+ while (advance < max_advance) {
+ //read how much to advance from SDF
+ uvw = (pos + ray_dir * advance) * pos_to_uvw;
+
+ float distance = texture(sampler3D(sdf_cascades[i], linear_sampler), uvw).r * 255.0 - 1.7;
+
+ if (distance < 0.001) {
+ //consider hit
+ hit = true;
+ break;
+ }
+
+ advance += distance;
+ }
+
+ if (!hit) {
+ pos += ray_dir * min(advance, max_advance);
+ pos /= cascades.data[i].to_cell;
+ pos += cascades.data[i].offset;
+ ray_pos = pos;
+ continue;
+ }
+
+ //compute albedo, emission and normal at hit point
+
+ const float EPSILON = 0.001;
+ vec3 hit_normal = normalize(vec3(
+ texture(sampler3D(sdf_cascades[i], linear_sampler), uvw + vec3(EPSILON, 0.0, 0.0)).r - texture(sampler3D(sdf_cascades[i], linear_sampler), uvw - vec3(EPSILON, 0.0, 0.0)).r,
+ texture(sampler3D(sdf_cascades[i], linear_sampler), uvw + vec3(0.0, EPSILON, 0.0)).r - texture(sampler3D(sdf_cascades[i], linear_sampler), uvw - vec3(0.0, EPSILON, 0.0)).r,
+ texture(sampler3D(sdf_cascades[i], linear_sampler), uvw + vec3(0.0, 0.0, EPSILON)).r - texture(sampler3D(sdf_cascades[i], linear_sampler), uvw - vec3(0.0, 0.0, EPSILON)).r));
+
+ vec3 hit_light = texture(sampler3D(light_cascades[i], linear_sampler), uvw).rgb;
+ vec4 aniso0 = texture(sampler3D(aniso0_cascades[i], linear_sampler), uvw);
+ vec3 hit_aniso0 = aniso0.rgb;
+ vec3 hit_aniso1 = vec3(aniso0.a, texture(sampler3D(aniso1_cascades[i], linear_sampler), uvw).rg);
+
+ hit_light *= (dot(max(vec3(0.0), (hit_normal * hit_aniso0)), vec3(1.0)) + dot(max(vec3(0.0), (-hit_normal * hit_aniso1)), vec3(1.0)));
+
+ 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;
+ }
+ }
+
+ 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/rasterizer_rd/shaders/sdfgi_debug_probes.glsl b/servers/rendering/rasterizer_rd/shaders/sdfgi_debug_probes.glsl
new file mode 100644
index 0000000000..08da283dad
--- /dev/null
+++ b/servers/rendering/rasterizer_rd/shaders/sdfgi_debug_probes.glsl
@@ -0,0 +1,231 @@
+#[vertex]
+
+#version 450
+
+VERSION_DEFINES
+
+#define MAX_CASCADES 8
+
+layout(push_constant, binding = 0, std430) uniform Params {
+ mat4 projection;
+
+ uint band_power;
+ uint sections_in_band;
+ uint band_mask;
+ float section_arc;
+
+ vec3 grid_size;
+ uint cascade;
+
+ uint pad;
+ float y_mult;
+ uint probe_debug_index;
+ int probe_axis_size;
+}
+params;
+
+// http://in4k.untergrund.net/html_articles/hugi_27_-_coding_corner_polaris_sphere_tessellation_101.htm
+
+vec3 get_sphere_vertex(uint p_vertex_id) {
+ float x_angle = float(p_vertex_id & 1u) + (p_vertex_id >> params.band_power);
+
+ float y_angle =
+ float((p_vertex_id & params.band_mask) >> 1) + ((p_vertex_id >> params.band_power) * params.sections_in_band);
+
+ x_angle *= params.section_arc * 0.5f; // remember - 180AA x rot not 360
+ y_angle *= -params.section_arc;
+
+ vec3 point = vec3(sin(x_angle) * sin(y_angle), cos(x_angle), sin(x_angle) * cos(y_angle));
+
+ return point;
+}
+
+#ifdef MODE_PROBES
+
+layout(location = 0) out vec3 normal_interp;
+layout(location = 1) out flat uint probe_index;
+
+#endif
+
+#ifdef MODE_VISIBILITY
+
+layout(location = 0) out float visibility;
+
+#endif
+
+struct CascadeData {
+ vec3 offset; //offset of (0,0,0) in world coordinates
+ float to_cell; // 1/bounds * grid_size
+ ivec3 probe_world_offset;
+ uint pad;
+};
+
+layout(set = 0, binding = 1, std140) uniform Cascades {
+ CascadeData data[MAX_CASCADES];
+}
+cascades;
+
+layout(set = 0, binding = 4) uniform texture3D occlusion_texture;
+layout(set = 0, binding = 3) uniform sampler linear_sampler;
+
+void main() {
+#ifdef MODE_PROBES
+ probe_index = gl_InstanceIndex;
+
+ normal_interp = get_sphere_vertex(gl_VertexIndex);
+
+ vec3 vertex = normal_interp * 0.2;
+
+ float probe_cell_size = float(params.grid_size / float(params.probe_axis_size - 1)) / cascades.data[params.cascade].to_cell;
+
+ ivec3 probe_cell;
+ probe_cell.x = int(probe_index % params.probe_axis_size);
+ probe_cell.y = int(probe_index / (params.probe_axis_size * params.probe_axis_size));
+ probe_cell.z = int((probe_index / params.probe_axis_size) % params.probe_axis_size);
+
+ vertex += (cascades.data[params.cascade].offset + vec3(probe_cell) * probe_cell_size) / vec3(1.0, params.y_mult, 1.0);
+
+ gl_Position = params.projection * vec4(vertex, 1.0);
+#endif
+
+#ifdef MODE_VISIBILITY
+
+ int probe_index = int(params.probe_debug_index);
+
+ vec3 vertex = get_sphere_vertex(gl_VertexIndex) * 0.01;
+
+ float probe_cell_size = float(params.grid_size / float(params.probe_axis_size - 1)) / cascades.data[params.cascade].to_cell;
+
+ ivec3 probe_cell;
+ probe_cell.x = int(probe_index % params.probe_axis_size);
+ probe_cell.y = int((probe_index % (params.probe_axis_size * params.probe_axis_size)) / params.probe_axis_size);
+ probe_cell.z = int(probe_index / (params.probe_axis_size * params.probe_axis_size));
+
+ vertex += (cascades.data[params.cascade].offset + vec3(probe_cell) * probe_cell_size) / vec3(1.0, params.y_mult, 1.0);
+
+ int probe_voxels = int(params.grid_size.x) / int(params.probe_axis_size - 1);
+ int occluder_index = int(gl_InstanceIndex);
+
+ int diameter = probe_voxels * 2;
+ ivec3 occluder_pos;
+ occluder_pos.x = int(occluder_index % diameter);
+ occluder_pos.y = int(occluder_index / (diameter * diameter));
+ occluder_pos.z = int((occluder_index / diameter) % diameter);
+
+ float cell_size = 1.0 / cascades.data[params.cascade].to_cell;
+
+ ivec3 occluder_offset = occluder_pos - ivec3(diameter / 2);
+ vertex += ((vec3(occluder_offset) + vec3(0.5)) * cell_size) / vec3(1.0, params.y_mult, 1.0);
+
+ ivec3 global_cell = probe_cell + cascades.data[params.cascade].probe_world_offset;
+ uint occlusion_layer = 0;
+ if ((global_cell.x & 1) != 0) {
+ occlusion_layer |= 1;
+ }
+ if ((global_cell.y & 1) != 0) {
+ occlusion_layer |= 2;
+ }
+ if ((global_cell.z & 1) != 0) {
+ occlusion_layer |= 4;
+ }
+ ivec3 tex_pos = probe_cell * probe_voxels + occluder_offset;
+
+ const vec4 layer_axis[4] = vec4[](
+ vec4(1, 0, 0, 0),
+ vec4(0, 1, 0, 0),
+ vec4(0, 0, 1, 0),
+ vec4(0, 0, 0, 1));
+
+ tex_pos.z += int(params.cascade) * int(params.grid_size);
+ if (occlusion_layer >= 4) {
+ tex_pos.x += int(params.grid_size.x);
+ occlusion_layer &= 3;
+ }
+
+ visibility = dot(texelFetch(sampler3D(occlusion_texture, linear_sampler), tex_pos, 0), layer_axis[occlusion_layer]);
+
+ gl_Position = params.projection * vec4(vertex, 1.0);
+
+#endif
+}
+
+#[fragment]
+
+#version 450
+
+VERSION_DEFINES
+
+layout(location = 0) out vec4 frag_color;
+
+layout(set = 0, binding = 2) uniform texture2DArray lightprobe_texture;
+layout(set = 0, binding = 3) uniform sampler linear_sampler;
+
+layout(push_constant, binding = 0, std430) uniform Params {
+ mat4 projection;
+
+ uint band_power;
+ uint sections_in_band;
+ uint band_mask;
+ float section_arc;
+
+ vec3 grid_size;
+ uint cascade;
+
+ uint pad;
+ float y_mult;
+ uint probe_debug_index;
+ int probe_axis_size;
+}
+params;
+
+#ifdef MODE_PROBES
+
+layout(location = 0) in vec3 normal_interp;
+layout(location = 1) in flat uint probe_index;
+
+#endif
+
+#ifdef MODE_VISIBILITY
+layout(location = 0) in float visibility;
+#endif
+
+vec2 octahedron_wrap(vec2 v) {
+ vec2 signVal;
+ signVal.x = v.x >= 0.0 ? 1.0 : -1.0;
+ signVal.y = v.y >= 0.0 ? 1.0 : -1.0;
+ return (1.0 - abs(v.yx)) * signVal;
+}
+
+vec2 octahedron_encode(vec3 n) {
+ // https://twitter.com/Stubbesaurus/status/937994790553227264
+ n /= (abs(n.x) + abs(n.y) + abs(n.z));
+ n.xy = n.z >= 0.0 ? n.xy : octahedron_wrap(n.xy);
+ n.xy = n.xy * 0.5 + 0.5;
+ return n.xy;
+}
+
+void main() {
+#ifdef MODE_PROBES
+
+ ivec3 tex_pos;
+ tex_pos.x = int(probe_index) % params.probe_axis_size; //x
+ tex_pos.y = int(probe_index) / (params.probe_axis_size * params.probe_axis_size);
+ tex_pos.x += params.probe_axis_size * ((int(probe_index) / params.probe_axis_size) % params.probe_axis_size); //z
+ tex_pos.z = int(params.cascade);
+
+ vec3 tex_pos_ofs = vec3(octahedron_encode(normal_interp) * float(OCT_SIZE), 0.0);
+ vec3 tex_posf = vec3(vec2(tex_pos.xy * (OCT_SIZE + 2) + ivec2(1)), float(tex_pos.z)) + tex_pos_ofs;
+
+ tex_posf.xy /= vec2(ivec2(params.probe_axis_size * params.probe_axis_size * (OCT_SIZE + 2), params.probe_axis_size * (OCT_SIZE + 2)));
+
+ vec4 indirect_light = textureLod(sampler2DArray(lightprobe_texture, linear_sampler), tex_posf, 0.0);
+
+ frag_color = indirect_light;
+
+#endif
+
+#ifdef MODE_VISIBILITY
+
+ frag_color = vec4(vec3(1, visibility, visibility), 1.0);
+#endif
+}
diff --git a/servers/rendering/rasterizer_rd/shaders/sdfgi_direct_light.glsl b/servers/rendering/rasterizer_rd/shaders/sdfgi_direct_light.glsl
new file mode 100644
index 0000000000..c4b29216d5
--- /dev/null
+++ b/servers/rendering/rasterizer_rd/shaders/sdfgi_direct_light.glsl
@@ -0,0 +1,472 @@
+#[compute]
+
+#version 450
+
+VERSION_DEFINES
+
+layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in;
+
+#define MAX_CASCADES 8
+
+layout(set = 0, binding = 1) uniform texture3D sdf_cascades[MAX_CASCADES];
+layout(set = 0, binding = 2) uniform sampler linear_sampler;
+
+layout(set = 0, binding = 3, std430) restrict readonly buffer DispatchData {
+ uint x;
+ uint y;
+ uint z;
+ uint total_count;
+}
+dispatch_data;
+
+struct ProcessVoxel {
+ uint position; //xyz 7 bit packed, extra 11 bits for neigbours
+ uint albedo; //rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neibhbours
+ uint light; //rgbe8985 encoded total saved light, extra 2 bits for neighbous
+ uint light_aniso; //55555 light anisotropy, extra 2 bits for neighbours
+ //total neighbours: 26
+};
+
+#ifdef MODE_PROCESS_STATIC
+layout(set = 0, binding = 4, std430) restrict buffer ProcessVoxels {
+#else
+layout(set = 0, binding = 4, std430) restrict buffer readonly ProcessVoxels {
+#endif
+ ProcessVoxel data[];
+}
+process_voxels;
+
+layout(r32ui, set = 0, binding = 5) uniform restrict uimage3D dst_light;
+layout(rgba8, set = 0, binding = 6) uniform restrict image3D dst_aniso0;
+layout(rg8, set = 0, binding = 7) uniform restrict image3D dst_aniso1;
+
+struct CascadeData {
+ vec3 offset; //offset of (0,0,0) in world coordinates
+ float to_cell; // 1/bounds * grid_size
+ ivec3 probe_world_offset;
+ uint pad;
+};
+
+layout(set = 0, binding = 8, std140) uniform Cascades {
+ CascadeData data[MAX_CASCADES];
+}
+cascades;
+
+#define LIGHT_TYPE_DIRECTIONAL 0
+#define LIGHT_TYPE_OMNI 1
+#define LIGHT_TYPE_SPOT 2
+
+struct Light {
+ vec3 color;
+ float energy;
+
+ vec3 direction;
+ bool has_shadow;
+
+ vec3 position;
+ float attenuation;
+
+ uint type;
+ float spot_angle;
+ float spot_attenuation;
+ float radius;
+
+ vec4 shadow_color;
+};
+
+layout(set = 0, binding = 9, std140) buffer restrict readonly Lights {
+ Light data[];
+}
+lights;
+
+layout(set = 0, binding = 10) uniform texture2DArray lightprobe_texture;
+
+layout(push_constant, binding = 0, std430) uniform Params {
+ vec3 grid_size;
+ uint max_cascades;
+
+ uint cascade;
+ uint light_count;
+ uint process_offset;
+ uint process_increment;
+
+ int probe_axis_size;
+ bool multibounce;
+ float y_mult;
+ uint pad;
+}
+params;
+
+vec2 octahedron_wrap(vec2 v) {
+ vec2 signVal;
+ signVal.x = v.x >= 0.0 ? 1.0 : -1.0;
+ signVal.y = v.y >= 0.0 ? 1.0 : -1.0;
+ return (1.0 - abs(v.yx)) * signVal;
+}
+
+vec2 octahedron_encode(vec3 n) {
+ // https://twitter.com/Stubbesaurus/status/937994790553227264
+ n /= (abs(n.x) + abs(n.y) + abs(n.z));
+ n.xy = n.z >= 0.0 ? n.xy : octahedron_wrap(n.xy);
+ n.xy = n.xy * 0.5 + 0.5;
+ return n.xy;
+}
+
+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 (voxel_index >= dispatch_data.total_count) {
+ return;
+ }
+
+ uint voxel_position = process_voxels.data[voxel_index].position;
+
+ //keep for storing to texture
+ ivec3 positioni = ivec3((uvec3(voxel_position, voxel_position, voxel_position) >> uvec3(0, 7, 14)) & uvec3(0x7F));
+
+ vec3 position = vec3(positioni) + vec3(0.5);
+ position /= cascades.data[params.cascade].to_cell;
+ position += cascades.data[params.cascade].offset;
+
+ uint voxel_albedo = process_voxels.data[voxel_index].albedo;
+
+ vec3 albedo = vec3(uvec3(voxel_albedo >> 10, voxel_albedo >> 5, voxel_albedo) & uvec3(0x1F)) / float(0x1F);
+ vec3 light_accum[6];
+
+ uint valid_aniso = (voxel_albedo >> 15) & 0x3F;
+
+ {
+ uint rgbe = process_voxels.data[voxel_index].light;
+
+ //read rgbe8985
+ float r = float((rgbe & 0xff) << 1);
+ float g = float((rgbe >> 8) & 0x1ff);
+ float b = float(((rgbe >> 17) & 0xff) << 1);
+ float e = float((rgbe >> 25) & 0x1F);
+ float m = pow(2.0, e - 15.0 - 9.0);
+
+ vec3 l = vec3(r, g, b) * m;
+
+ uint aniso = process_voxels.data[voxel_index].light_aniso;
+ for (uint i = 0; i < 6; i++) {
+ float strength = ((aniso >> (i * 5)) & 0x1F) / float(0x1F);
+ light_accum[i] = l * strength;
+ }
+ }
+
+ 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;
+ vec3 uvw_ofs = pos_to_uvw * 0.5;
+
+ for (uint i = 0; i < params.light_count; i++) {
+ float attenuation = 1.0;
+ vec3 direction;
+ float light_distance = 1e20;
+
+ switch (lights.data[i].type) {
+ case LIGHT_TYPE_DIRECTIONAL: {
+ direction = -lights.data[i].direction;
+ } break;
+ case LIGHT_TYPE_OMNI: {
+ vec3 rel_vec = lights.data[i].position - position;
+ direction = normalize(rel_vec);
+ light_distance = length(rel_vec);
+ rel_vec.y /= params.y_mult;
+ attenuation = pow(clamp(1.0 - length(rel_vec) / lights.data[i].radius, 0.0, 1.0), lights.data[i].attenuation);
+ } break;
+ case LIGHT_TYPE_SPOT: {
+ vec3 rel_vec = lights.data[i].position - position;
+ direction = normalize(rel_vec);
+ light_distance = length(rel_vec);
+ rel_vec.y /= params.y_mult;
+ attenuation = pow(clamp(1.0 - length(rel_vec) / lights.data[i].radius, 0.0, 1.0), 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);
+ }
+ } break;
+ }
+
+ if (attenuation < 0.001) {
+ continue;
+ }
+
+ bool hit = false;
+
+ vec3 ray_pos = position;
+ vec3 ray_dir = direction;
+ vec3 inv_dir = 1.0 / ray_dir;
+
+ //this is how to properly bias outgoing rays
+ float cell_size = 1.0 / cascades.data[params.cascade].to_cell;
+ ray_pos += sign(direction) * cell_size * 0.48; // go almost to the box edge but remain inside
+ ray_pos += ray_dir * 0.4 * cell_size; //apply a small bias from there
+
+ for (uint j = params.cascade; j < params.max_cascades; j++) {
+ //convert to local bounds
+ vec3 pos = ray_pos - cascades.data[j].offset;
+ pos *= cascades.data[j].to_cell;
+ float local_distance = light_distance * cascades.data[j].to_cell;
+
+ if (any(lessThan(pos, vec3(0.0))) || any(greaterThanEqual(pos, params.grid_size))) {
+ continue; //already past bounds for this cascade, goto next
+ }
+
+ //find maximum advance distance (until reaching bounds)
+ vec3 t0 = -pos * inv_dir;
+ vec3 t1 = (params.grid_size - pos) * inv_dir;
+ vec3 tmax = max(t0, t1);
+ float max_advance = min(tmax.x, min(tmax.y, tmax.z));
+
+ max_advance = min(local_distance, max_advance);
+
+ float advance = 0.0;
+ float occlusion = 1.0;
+
+ while (advance < max_advance) {
+ //read how much to advance from SDF
+ vec3 uvw = (pos + ray_dir * advance) * pos_to_uvw;
+
+ float distance = texture(sampler3D(sdf_cascades[j], linear_sampler), uvw).r * 255.0 - 1.0;
+ if (distance < 0.001) {
+ //consider hit
+ hit = true;
+ break;
+ }
+
+ occlusion = min(occlusion, distance);
+
+ advance += distance;
+ }
+
+ if (hit) {
+ attenuation *= occlusion;
+ break;
+ }
+
+ if (advance >= local_distance) {
+ break; //past light distance, abandon search
+ }
+ //change ray origin to collision with bounds
+ pos += ray_dir * max_advance;
+ pos /= cascades.data[j].to_cell;
+ pos += cascades.data[j].offset;
+ light_distance -= max_advance / cascades.data[j].to_cell;
+ ray_pos = pos;
+ }
+
+ if (!hit) {
+ vec3 light = albedo * lights.data[i].color.rgb * lights.data[i].energy * attenuation;
+
+ for (int j = 0; j < 6; j++) {
+ if (bool(valid_aniso & (1 << j))) {
+ light_accum[j] += max(0.0, dot(aniso_dir[j], direction)) * light;
+ }
+ }
+ }
+ }
+
+ // 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];
+ vec3 light_total = vec3(0);
+
+ for (int i = 0; i < 6; i++) {
+ light_total += light_accum[i];
+ lumas[i] = max(light_accum[i].r, max(light_accum[i].g, light_accum[i].b));
+ }
+
+ float luma_total = max(light_total.r, max(light_total.g, light_total.b));
+
+ uint light_total_rgbe;
+
+ {
+ //compress to RGBE9995 to save space
+
+ const float pow2to9 = 512.0f;
+ const float B = 15.0f;
+ const float N = 9.0f;
+ const float LN2 = 0.6931471805599453094172321215;
+
+ float cRed = clamp(light_total.r, 0.0, 65408.0);
+ float cGreen = clamp(light_total.g, 0.0, 65408.0);
+ float cBlue = clamp(light_total.b, 0.0, 65408.0);
+
+ float cMax = max(cRed, max(cGreen, cBlue));
+
+ float expp = max(-B - 1.0f, floor(log(cMax) / LN2)) + 1.0f + B;
+
+ float sMax = floor((cMax / pow(2.0f, expp - B - N)) + 0.5f);
+
+ float exps = expp + 1.0f;
+
+ if (0.0 <= sMax && sMax < pow2to9) {
+ exps = expp;
+ }
+
+ float sRed = floor((cRed / pow(2.0f, exps - B - N)) + 0.5f);
+ float sGreen = floor((cGreen / pow(2.0f, exps - B - N)) + 0.5f);
+ float sBlue = floor((cBlue / pow(2.0f, exps - B - N)) + 0.5f);
+#ifdef MODE_PROCESS_STATIC
+ //since its self-save, use RGBE8985
+ light_total_rgbe = ((uint(sRed) & 0x1FF) >> 1) | ((uint(sGreen) & 0x1FF) << 8) | (((uint(sBlue) & 0x1FF) >> 1) << 17) | ((uint(exps) & 0x1F) << 25);
+
+#else
+ light_total_rgbe = (uint(sRed) & 0x1FF) | ((uint(sGreen) & 0x1FF) << 9) | ((uint(sBlue) & 0x1FF) << 18) | ((uint(exps) & 0x1F) << 27);
+#endif
+ }
+
+#ifdef MODE_PROCESS_DYNAMIC
+
+ vec4 aniso0;
+ aniso0.r = lumas[0] / luma_total;
+ aniso0.g = lumas[1] / luma_total;
+ aniso0.b = lumas[2] / luma_total;
+ aniso0.a = lumas[3] / luma_total;
+
+ vec2 aniso1;
+ aniso1.r = lumas[4] / luma_total;
+ aniso1.g = lumas[5] / luma_total;
+
+ //save to 3D textures
+ imageStore(dst_aniso0, positioni, aniso0);
+ imageStore(dst_aniso1, positioni, vec4(aniso1, 0.0, 0.0));
+ imageStore(dst_light, positioni, uvec4(light_total_rgbe));
+
+ //also fill neighbours, so light interpolation during the indirect pass works
+
+ //recover the neighbour list from the leftover bits
+ uint neighbours = (voxel_albedo >> 21) | ((voxel_position >> 21) << 11) | ((process_voxels.data[voxel_index].light >> 30) << 22) | ((process_voxels.data[voxel_index].light_aniso >> 30) << 24);
+
+ const uint max_neighbours = 26;
+ const ivec3 neighbour_positions[max_neighbours] = ivec3[](
+ ivec3(-1, -1, -1),
+ ivec3(-1, -1, 0),
+ ivec3(-1, -1, 1),
+ ivec3(-1, 0, -1),
+ ivec3(-1, 0, 0),
+ ivec3(-1, 0, 1),
+ ivec3(-1, 1, -1),
+ ivec3(-1, 1, 0),
+ ivec3(-1, 1, 1),
+ ivec3(0, -1, -1),
+ ivec3(0, -1, 0),
+ ivec3(0, -1, 1),
+ ivec3(0, 0, -1),
+ ivec3(0, 0, 1),
+ ivec3(0, 1, -1),
+ ivec3(0, 1, 0),
+ ivec3(0, 1, 1),
+ ivec3(1, -1, -1),
+ ivec3(1, -1, 0),
+ ivec3(1, -1, 1),
+ ivec3(1, 0, -1),
+ ivec3(1, 0, 0),
+ ivec3(1, 0, 1),
+ ivec3(1, 1, -1),
+ ivec3(1, 1, 0),
+ ivec3(1, 1, 1));
+
+ for (uint i = 0; i < max_neighbours; i++) {
+ if (bool(neighbours & (1 << i))) {
+ ivec3 neighbour_pos = positioni + neighbour_positions[i];
+ imageStore(dst_light, neighbour_pos, uvec4(light_total_rgbe));
+ imageStore(dst_aniso0, neighbour_pos, aniso0);
+ imageStore(dst_aniso1, neighbour_pos, vec4(aniso1, 0.0, 0.0));
+ }
+ }
+
+#endif
+
+#ifdef MODE_PROCESS_STATIC
+
+ //save back the anisotropic
+
+ uint light = process_voxels.data[voxel_index].light & (3 << 30);
+ light |= light_total_rgbe;
+ process_voxels.data[voxel_index].light = light; //replace
+
+ uint light_aniso = process_voxels.data[voxel_index].light_aniso & (3 << 30);
+ for (int i = 0; i < 6; i++) {
+ light_aniso |= min(31, uint((lumas[i] / luma_total) * 31.0)) << (i * 5);
+ }
+
+ process_voxels.data[voxel_index].light_aniso = light_aniso;
+
+#endif
+}
diff --git a/servers/rendering/rasterizer_rd/shaders/sdfgi_fields.glsl b/servers/rendering/rasterizer_rd/shaders/sdfgi_fields.glsl
new file mode 100644
index 0000000000..eec0a90c0d
--- /dev/null
+++ b/servers/rendering/rasterizer_rd/shaders/sdfgi_fields.glsl
@@ -0,0 +1,182 @@
+/* clang-format off */
+[compute]
+
+#version 450
+
+VERSION_DEFINES
+
+layout(local_size_x = OCT_RES, local_size_y = OCT_RES, local_size_z = 1) in;
+
+/* clang-format on */
+
+#define MAX_CASCADES 8
+
+layout(rgba16f, set = 0, binding = 1) uniform restrict image2DArray irradiance_texture;
+layout(rg16f, set = 0, binding = 2) uniform restrict image2DArray depth_texture;
+
+ayout(rgba32ui, set = 0, binding = 3) uniform restrict uimage2DArray irradiance_history_texture;
+layout(rg32ui, set = 0, binding = 4) uniform restrict uimage2DArray depth_history_texture;
+
+struct CascadeData {
+ vec3 offset; //offset of (0,0,0) in world coordinates
+ float to_cell; // 1/bounds * grid_size
+};
+
+layout(set = 0, binding = 5, std140) uniform Cascades {
+ CascadeData data[MAX_CASCADES];
+}
+cascades;
+
+#define DEPTH_HISTORY_BITS 24
+#define IRRADIANCE_HISTORY_BITS 16
+
+layout(push_constant, binding = 0, std430) uniform Params {
+ vec3 grid_size;
+ uint max_cascades;
+
+ uint probe_axis_size;
+ uint cascade;
+ uint history_size;
+ uint pad0;
+
+ ivec3 scroll; //scroll in probes
+ uint pad1;
+}
+params;
+
+void main() {
+ ivec2 local = ivec2(gl_LocalInvocationID.xy);
+ ivec2 probe = ivec2(gl_WorkGroupID.xy);
+
+ ivec3 probe_cell;
+ probe_cell.x = probe.x % int(params.probe_axis_size);
+ probe_cell.y = probe.y;
+ probe_cell.z = probe.x / int(params.probe_axis_size);
+
+#ifdef MODE_SCROLL_BEGIN
+
+ ivec3 read_cell = probe_cell - params.scroll;
+
+ uint src_layer = (params.history_size + 1) * params.cascade;
+ uint dst_layer = (params.history_size + 1) * params.max_cascades;
+
+ for (uint i = 0; i <= params.history_size; i++) {
+ ivec3 write_pos = ivec3(probe * OCT_RES + local, int(i));
+
+ if (any(lessThan(read_pos, ivec3(0))) || any(greaterThanEqual(read_pos, ivec3(params.probe_axis_size)))) {
+ // nowhere to read from for scrolling, try finding the value from upper probes
+
+#ifdef MODE_IRRADIANCE
+ imageStore(irradiance_history_texture, write_pos, uvec4(0));
+#endif
+#ifdef MODE_DEPTH
+ imageStore(depth_history_texture, write_pos, uvec4(0));
+#endif
+ } else {
+ ivec3 read_pos;
+ read_pos.xy = read_cell.xy;
+ read_pos.x += read_cell.z * params.probe_axis_size;
+ read_pos.xy = read_pos.xy * OCT_RES + local;
+ read_pos.z = int(i);
+
+#ifdef MODE_IRRADIANCE
+ uvec4 value = imageLoad(irradiance_history_texture, read_pos);
+ imageStore(irradiance_history_texture, write_pos, value);
+#endif
+#ifdef MODE_DEPTH
+ uvec2 value = imageLoad(depth_history_texture, read_pos);
+ imageStore(depth_history_texture, write_pos, value);
+#endif
+ }
+ }
+
+#endif // MODE_SCROLL_BEGIN
+
+#ifdef MODE_SCROLL_END
+
+ uint src_layer = (params.history_size + 1) * params.max_cascades;
+ uint dst_layer = (params.history_size + 1) * params.cascade;
+
+ for (uint i = 0; i <= params.history_size; i++) {
+ ivec3 pos = ivec3(probe * OCT_RES + local, int(i));
+
+#ifdef MODE_IRRADIANCE
+ uvec4 value = imageLoad(irradiance_history_texture, read_pos);
+ imageStore(irradiance_history_texture, write_pos, value);
+#endif
+#ifdef MODE_DEPTH
+ uvec2 value = imageLoad(depth_history_texture, read_pos);
+ imageStore(depth_history_texture, write_pos, value);
+#endif
+ }
+
+#endif //MODE_SCROLL_END
+
+#ifdef MODE_STORE
+
+ uint src_layer = (params.history_size + 1) * params.cascade + params.history_size;
+ ivec3 read_pos = ivec3(probe * OCT_RES + local, int(src_layer));
+
+ ivec3 write_pos = ivec3(probe * (OCT_RES + 2) + ivec2(1), int(params.cascade));
+
+ ivec3 copy_to[4] = ivec3[](write_pos, ivec3(-2, -2, -2), ivec3(-2, -2, -2), ivec3(-2, -2, -2));
+
+#ifdef MODE_IRRADIANCE
+ uvec4 average = imageLoad(irradiance_history_texture, read_pos);
+ vec4 light_accum = vec4(average / params.history_size) / float(1 << IRRADIANCE_HISTORY_BITS);
+
+#endif
+#ifdef MODE_DEPTH
+ uvec2 value = imageLoad(depth_history_texture, read_pos);
+ vec2 depth_accum = vec4(average / params.history_size) / float(1 << IRRADIANCE_HISTORY_BITS);
+
+ float probe_cell_size = float(params.grid_size / float(params.probe_axis_size - 1)) / cascades.data[params.cascade].to_cell;
+ float max_depth = length(params.grid_size / cascades.data[params.max_cascades - 1].to_cell);
+ max_depth /= probe_cell_size;
+
+ depth_value = (vec2(average / params.history_size) / float(1 << DEPTH_HISTORY_BITS)) * vec2(max_depth, max_depth * max_depth);
+
+#endif
+
+ /* Fill the border if required */
+
+ if (local == ivec2(0, 0)) {
+ copy_to[1] = texture_pos + ivec3(OCT_RES - 1, -1, 0);
+ copy_to[2] = texture_pos + ivec3(-1, OCT_RES - 1, 0);
+ copy_to[3] = texture_pos + ivec3(OCT_RES, OCT_RES, 0);
+ } else if (local == ivec2(OCT_RES - 1, 0)) {
+ copy_to[1] = texture_pos + ivec3(0, -1, 0);
+ copy_to[2] = texture_pos + ivec3(OCT_RES, OCT_RES - 1, 0);
+ copy_to[3] = texture_pos + ivec3(-1, OCT_RES, 0);
+ } else if (local == ivec2(0, OCT_RES - 1)) {
+ copy_to[1] = texture_pos + ivec3(-1, 0, 0);
+ copy_to[2] = texture_pos + ivec3(OCT_RES - 1, OCT_RES, 0);
+ copy_to[3] = texture_pos + ivec3(OCT_RES, -1, 0);
+ } else if (local == ivec2(OCT_RES - 1, OCT_RES - 1)) {
+ copy_to[1] = texture_pos + ivec3(0, OCT_RES, 0);
+ copy_to[2] = texture_pos + ivec3(OCT_RES, 0, 0);
+ copy_to[3] = texture_pos + ivec3(-1, -1, 0);
+ } else if (local.y == 0) {
+ copy_to[1] = texture_pos + ivec3(OCT_RES - local.x - 1, local.y - 1, 0);
+ } else if (local.x == 0) {
+ copy_to[1] = texture_pos + ivec3(local.x - 1, OCT_RES - local.y - 1, 0);
+ } else if (local.y == OCT_RES - 1) {
+ copy_to[1] = texture_pos + ivec3(OCT_RES - local.x - 1, local.y + 1, 0);
+ } else if (local.x == OCT_RES - 1) {
+ copy_to[1] = texture_pos + ivec3(local.x + 1, OCT_RES - local.y - 1, 0);
+ }
+
+ for (int i = 0; i < 4; i++) {
+ if (copy_to[i] == ivec3(-2, -2, -2)) {
+ continue;
+ }
+#ifdef MODE_IRRADIANCE
+ imageStore(irradiance_texture, copy_to[i], light_accum);
+#endif
+#ifdef MODE_DEPTH
+ imageStore(depth_texture, copy_to[i], vec4(depth_value, 0.0, 0.0));
+#endif
+ }
+
+#endif // MODE_STORE
+}
diff --git a/servers/rendering/rasterizer_rd/shaders/sdfgi_integrate.glsl b/servers/rendering/rasterizer_rd/shaders/sdfgi_integrate.glsl
new file mode 100644
index 0000000000..e4779aafaf
--- /dev/null
+++ b/servers/rendering/rasterizer_rd/shaders/sdfgi_integrate.glsl
@@ -0,0 +1,605 @@
+#[compute]
+
+#version 450
+
+VERSION_DEFINES
+
+layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in;
+
+#define MAX_CASCADES 8
+
+layout(set = 0, binding = 1) uniform texture3D sdf_cascades[MAX_CASCADES];
+layout(set = 0, binding = 2) uniform texture3D light_cascades[MAX_CASCADES];
+layout(set = 0, binding = 3) uniform texture3D aniso0_cascades[MAX_CASCADES];
+layout(set = 0, binding = 4) uniform texture3D aniso1_cascades[MAX_CASCADES];
+
+layout(set = 0, binding = 6) uniform sampler linear_sampler;
+
+struct CascadeData {
+ vec3 offset; //offset of (0,0,0) in world coordinates
+ float to_cell; // 1/bounds * grid_size
+ ivec3 probe_world_offset;
+ uint pad;
+};
+
+layout(set = 0, binding = 7, std140) uniform Cascades {
+ CascadeData data[MAX_CASCADES];
+}
+cascades;
+
+layout(r32ui, set = 0, binding = 8) uniform restrict uimage2DArray lightprobe_texture_data;
+layout(rgba16i, set = 0, binding = 9) uniform restrict iimage2DArray lightprobe_history_texture;
+layout(rgba32i, set = 0, binding = 10) uniform restrict iimage2D lightprobe_average_texture;
+
+//used for scrolling
+layout(rgba16i, set = 0, binding = 11) uniform restrict iimage2DArray lightprobe_history_scroll_texture;
+layout(rgba32i, set = 0, binding = 12) uniform restrict iimage2D lightprobe_average_scroll_texture;
+
+layout(rgba32i, set = 0, binding = 13) uniform restrict iimage2D lightprobe_average_parent_texture;
+
+layout(set = 1, binding = 0) uniform textureCube sky_irradiance;
+
+layout(set = 1, binding = 1) uniform sampler linear_sampler_mipmaps;
+
+#define HISTORY_BITS 10
+
+#define SKY_MODE_DISABLED 0
+#define SKY_MODE_COLOR 1
+#define SKY_MODE_SKY 2
+
+layout(push_constant, binding = 0, std430) uniform Params {
+ vec3 grid_size;
+ uint max_cascades;
+
+ uint probe_axis_size;
+ uint cascade;
+ uint history_index;
+ uint history_size;
+
+ uint ray_count;
+ float ray_bias;
+ ivec2 image_size;
+
+ ivec3 world_offset;
+ uint sky_mode;
+
+ ivec3 scroll;
+ float sky_energy;
+
+ vec3 sky_color;
+ float y_mult;
+}
+params;
+
+const float PI = 3.14159265f;
+const float GOLDEN_ANGLE = PI * (3.0 - sqrt(5.0));
+
+vec3 vogel_hemisphere(uint p_index, uint p_count, float p_offset) {
+ float r = sqrt(float(p_index) + 0.5f) / sqrt(float(p_count));
+ float theta = float(p_index) * GOLDEN_ANGLE + p_offset;
+ float y = cos(r * PI * 0.5);
+ float l = sin(r * PI * 0.5);
+ return vec3(l * cos(theta), l * sin(theta), y * (float(p_index & 1) * 2.0 - 1.0));
+}
+
+uvec3 hash3(uvec3 x) {
+ x = ((x >> 16) ^ x) * 0x45d9f3b;
+ x = ((x >> 16) ^ x) * 0x45d9f3b;
+ x = (x >> 16) ^ x;
+ return x;
+}
+
+float hashf3(vec3 co) {
+ return fract(sin(dot(co, vec3(12.9898, 78.233, 137.13451))) * 43758.5453);
+}
+
+vec3 octahedron_encode(vec2 f) {
+ // https://twitter.com/Stubbesaurus/status/937994790553227264
+ f = f * 2.0 - 1.0;
+ vec3 n = vec3(f.x, f.y, 1.0f - abs(f.x) - abs(f.y));
+ float t = clamp(-n.z, 0.0, 1.0);
+ n.x += n.x >= 0 ? -t : t;
+ n.y += n.y >= 0 ? -t : t;
+ return normalize(n);
+}
+
+uint rgbe_encode(vec3 color) {
+ const float pow2to9 = 512.0f;
+ const float B = 15.0f;
+ const float N = 9.0f;
+ const float LN2 = 0.6931471805599453094172321215;
+
+ float cRed = clamp(color.r, 0.0, 65408.0);
+ float cGreen = clamp(color.g, 0.0, 65408.0);
+ float cBlue = clamp(color.b, 0.0, 65408.0);
+
+ float cMax = max(cRed, max(cGreen, cBlue));
+
+ float expp = max(-B - 1.0f, floor(log(cMax) / LN2)) + 1.0f + B;
+
+ float sMax = floor((cMax / pow(2.0f, expp - B - N)) + 0.5f);
+
+ float exps = expp + 1.0f;
+
+ if (0.0 <= sMax && sMax < pow2to9) {
+ exps = expp;
+ }
+
+ float sRed = floor((cRed / pow(2.0f, exps - B - N)) + 0.5f);
+ float sGreen = floor((cGreen / pow(2.0f, exps - B - N)) + 0.5f);
+ float sBlue = floor((cBlue / pow(2.0f, exps - B - N)) + 0.5f);
+ return (uint(sRed) & 0x1FF) | ((uint(sGreen) & 0x1FF) << 9) | ((uint(sBlue) & 0x1FF) << 18) | ((uint(exps) & 0x1F) << 27);
+}
+
+void main() {
+ ivec2 pos = ivec2(gl_GlobalInvocationID.xy);
+ if (any(greaterThanEqual(pos, params.image_size))) { //too large, do nothing
+ return;
+ }
+
+#ifdef MODE_PROCESS
+
+ float probe_cell_size = float(params.grid_size.x / float(params.probe_axis_size - 1)) / cascades.data[params.cascade].to_cell;
+
+ ivec3 probe_cell;
+ probe_cell.x = pos.x % int(params.probe_axis_size);
+ probe_cell.y = pos.y;
+ probe_cell.z = pos.x / int(params.probe_axis_size);
+
+ vec3 probe_pos = cascades.data[params.cascade].offset + vec3(probe_cell) * probe_cell_size;
+ vec3 pos_to_uvw = 1.0 / params.grid_size;
+
+ 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
+ );
+
+ // quickly ensure each probe has a different "offset" for the vogel function, based on integer world position
+ uvec3 h3 = hash3(uvec3(params.world_offset + probe_cell));
+ float offset = hashf3(vec3(h3 & uvec3(0xFFFFF)));
+
+ //for a more homogeneous hemisphere, alternate based on history frames
+ uint ray_offset = params.history_index;
+ uint ray_mult = params.history_size;
+ uint ray_total = ray_mult * params.ray_count;
+
+ for (uint i = 0; i < params.ray_count; i++) {
+ vec3 ray_dir = vogel_hemisphere(ray_offset + i * ray_mult, ray_total, offset);
+ ray_dir.y *= params.y_mult;
+ ray_dir = normalize(ray_dir);
+
+ //needs to be visible
+ vec3 ray_pos = probe_pos;
+ vec3 inv_dir = 1.0 / ray_dir;
+
+ bool hit = false;
+ vec3 hit_normal;
+ vec3 hit_light;
+ vec3 hit_aniso0;
+ vec3 hit_aniso1;
+
+ 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;
+
+ for (uint j = params.cascade; j < params.max_cascades; j++) {
+ //convert to local bounds
+ vec3 pos = ray_pos - cascades.data[j].offset;
+ pos *= cascades.data[j].to_cell;
+
+ if (any(lessThan(pos, vec3(0.0))) || any(greaterThanEqual(pos, params.grid_size))) {
+ continue; //already past bounds for this cascade, goto next
+ }
+
+ //find maximum advance distance (until reaching bounds)
+ vec3 t0 = -pos * inv_dir;
+ vec3 t1 = (params.grid_size - pos) * inv_dir;
+ vec3 tmax = max(t0, t1);
+ float max_advance = min(tmax.x, min(tmax.y, tmax.z));
+
+ float advance = 0.0;
+
+ 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) {
+ //consider hit
+ hit = true;
+ break;
+ }
+
+ advance += distance;
+ }
+
+ 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);
+
+ break;
+ }
+
+ //change ray origin to collision with bounds
+ pos += ray_dir * max_advance;
+ pos /= cascades.data[j].to_cell;
+ pos += cascades.data[j].offset;
+ ray_pos = pos;
+ }
+
+ vec4 light;
+ if (hit) {
+ //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) {
+ 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
+ light.rgb *= params.sky_energy;
+ light.a = 0.0;
+
+ } else if (params.sky_mode == SKY_MODE_COLOR) {
+ light.rgb = params.sky_color;
+ light.rgb *= params.sky_energy;
+ light.a = 0.0;
+ } else {
+ light = vec4(0, 0, 0, 0);
+ }
+
+ vec3 ray_dir2 = ray_dir * ray_dir;
+ 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
+#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)
+
+#endif
+ );
+
+ for (uint j = 0; j < SH_SIZE; j++) {
+ probe_sh_accum[j] += light * c[j];
+ }
+ }
+
+ for (uint i = 0; i < SH_SIZE; i++) {
+ // store in history texture
+ ivec3 prev_pos = ivec3(pos.x, pos.y * SH_SIZE + i, int(params.history_index));
+ ivec2 average_pos = prev_pos.xy;
+
+ vec4 value = probe_sh_accum[i] * 4.0 / float(params.ray_count);
+
+ ivec4 ivalue = clamp(ivec4(value * float(1 << HISTORY_BITS)), -32768, 32767); //clamp to 16 bits, so higher values don't break average
+
+ ivec4 prev_value = imageLoad(lightprobe_history_texture, prev_pos);
+ ivec4 average = imageLoad(lightprobe_average_texture, average_pos);
+
+ average -= prev_value;
+ average += ivalue;
+
+ imageStore(lightprobe_history_texture, prev_pos, ivalue);
+ imageStore(lightprobe_average_texture, average_pos, average);
+ }
+#endif // MODE PROCESS
+
+#ifdef MODE_STORE
+
+ // converting to octahedral in this step is requiered because
+ // octahedral is much faster to read from the screen than spherical harmonics,
+ // despite the very slight quality loss
+
+ ivec2 sh_pos = (pos / OCT_SIZE) * ivec2(1, SH_SIZE);
+ ivec2 oct_pos = (pos / OCT_SIZE) * (OCT_SIZE + 2) + ivec2(1);
+ ivec2 local_pos = pos % OCT_SIZE;
+
+ //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[](
+
+ 0.282095, //l0
+ 0.488603 * normal.y, //l1n1
+ 0.488603 * normal.z, //l1n0
+ 0.488603 * normal.x, //l1p1
+ 1.092548 * normal.x * normal.y, //l2n2
+ 1.092548 * normal.y * normal.z, //l2n1
+ 0.315392 * (3.0 * normal2.z - 1.0), //l20
+ 1.092548 * normal.x * normal.z, //l2p1
+ 0.546274 * (normal2.x - normal2.y) //l2p2
+#if (SH_SIZE == 16)
+ ,
+ 0.590043 * normal.y * (3.0f * normal2.x - normal2.y),
+ 2.890611 * normal.y * normal.x * normal.z,
+ 0.646360 * normal.y * (-1.0f + 5.0f * normal2.z),
+ 0.373176 * (5.0f * normal2.z * normal.z - 3.0f * normal.z),
+ 0.457045 * normal.x * (-1.0f + 5.0f * normal2.z),
+ 1.445305 * (normal2.x - normal2.y) * normal.z,
+ 0.590043 * normal.x * (normal2.x - 3.0f * normal2.y)
+
+#endif
+ );
+
+ const float l_mult[SH_SIZE] = float[](
+ 1.0,
+ 2.0 / 3.0,
+ 2.0 / 3.0,
+ 2.0 / 3.0,
+ 1.0 / 4.0,
+ 1.0 / 4.0,
+ 1.0 / 4.0,
+ 1.0 / 4.0,
+ 1.0 / 4.0
+#if (SH_SIZE == 16)
+ , // l4 does not contribute to irradiance
+ 0.0,
+ 0.0,
+ 0.0,
+ 0.0,
+ 0.0,
+ 0.0,
+ 0.0
+#endif
+ );
+
+ vec3 irradiance = vec3(0.0);
+ vec3 radiance = vec3(0.0);
+
+ for (uint i = 0; i < SH_SIZE; i++) {
+ vec3 m = sh[i].rgb * c[i] * 4.0;
+ irradiance += m * l_mult[i];
+ radiance += m;
+ }
+
+ //encode RGBE9995 for the final texture
+
+ uint irradiance_rgbe = rgbe_encode(irradiance);
+ uint radiance_rgbe = rgbe_encode(radiance);
+
+ //store in octahedral map
+
+ ivec3 texture_pos = ivec3(oct_pos, int(params.cascade));
+ ivec3 copy_to[4] = ivec3[](ivec3(-2, -2, -2), ivec3(-2, -2, -2), ivec3(-2, -2, -2), ivec3(-2, -2, -2));
+ copy_to[0] = texture_pos + ivec3(local_pos, 0);
+
+ if (local_pos == ivec2(0, 0)) {
+ copy_to[1] = texture_pos + ivec3(OCT_SIZE - 1, -1, 0);
+ copy_to[2] = texture_pos + ivec3(-1, OCT_SIZE - 1, 0);
+ copy_to[3] = texture_pos + ivec3(OCT_SIZE, OCT_SIZE, 0);
+ } else if (local_pos == ivec2(OCT_SIZE - 1, 0)) {
+ copy_to[1] = texture_pos + ivec3(0, -1, 0);
+ copy_to[2] = texture_pos + ivec3(OCT_SIZE, OCT_SIZE - 1, 0);
+ copy_to[3] = texture_pos + ivec3(-1, OCT_SIZE, 0);
+ } else if (local_pos == ivec2(0, OCT_SIZE - 1)) {
+ copy_to[1] = texture_pos + ivec3(-1, 0, 0);
+ copy_to[2] = texture_pos + ivec3(OCT_SIZE - 1, OCT_SIZE, 0);
+ copy_to[3] = texture_pos + ivec3(OCT_SIZE, -1, 0);
+ } else if (local_pos == ivec2(OCT_SIZE - 1, OCT_SIZE - 1)) {
+ copy_to[1] = texture_pos + ivec3(0, OCT_SIZE, 0);
+ copy_to[2] = texture_pos + ivec3(OCT_SIZE, 0, 0);
+ copy_to[3] = texture_pos + ivec3(-1, -1, 0);
+ } else if (local_pos.y == 0) {
+ copy_to[1] = texture_pos + ivec3(OCT_SIZE - local_pos.x - 1, local_pos.y - 1, 0);
+ } else if (local_pos.x == 0) {
+ copy_to[1] = texture_pos + ivec3(local_pos.x - 1, OCT_SIZE - local_pos.y - 1, 0);
+ } else if (local_pos.y == OCT_SIZE - 1) {
+ copy_to[1] = texture_pos + ivec3(OCT_SIZE - local_pos.x - 1, local_pos.y + 1, 0);
+ } else if (local_pos.x == OCT_SIZE - 1) {
+ copy_to[1] = texture_pos + ivec3(local_pos.x + 1, OCT_SIZE - local_pos.y - 1, 0);
+ }
+
+ for (int i = 0; i < 4; i++) {
+ if (copy_to[i] == ivec3(-2, -2, -2)) {
+ continue;
+ }
+ imageStore(lightprobe_texture_data, copy_to[i], uvec4(irradiance_rgbe));
+ imageStore(lightprobe_texture_data, copy_to[i] + ivec3(0, 0, int(params.max_cascades)), uvec4(radiance_rgbe));
+ }
+
+#endif
+
+#ifdef MODE_SCROLL
+
+ ivec3 probe_cell;
+ probe_cell.x = pos.x % int(params.probe_axis_size);
+ probe_cell.y = pos.y;
+ probe_cell.z = pos.x / int(params.probe_axis_size);
+
+ ivec3 read_probe = probe_cell - params.scroll;
+
+ if (all(greaterThanEqual(read_probe, ivec3(0))) && all(lessThan(read_probe, ivec3(params.probe_axis_size)))) {
+ // can scroll
+ ivec2 tex_pos;
+ tex_pos = read_probe.xy;
+ tex_pos.x += read_probe.z * int(params.probe_axis_size);
+
+ //scroll
+ for (uint j = 0; j < params.history_size; j++) {
+ for (int i = 0; i < SH_SIZE; i++) {
+ // copy from history texture
+ ivec3 src_pos = ivec3(tex_pos.x, tex_pos.y * SH_SIZE + i, int(j));
+ ivec3 dst_pos = ivec3(pos.x, pos.y * SH_SIZE + i, int(j));
+ ivec4 value = imageLoad(lightprobe_history_texture, src_pos);
+ imageStore(lightprobe_history_scroll_texture, dst_pos, value);
+ }
+ }
+
+ for (int i = 0; i < SH_SIZE; i++) {
+ // copy from average texture
+ ivec2 src_pos = ivec2(tex_pos.x, tex_pos.y * SH_SIZE + i);
+ ivec2 dst_pos = ivec2(pos.x, pos.y * SH_SIZE + i);
+ ivec4 value = imageLoad(lightprobe_average_texture, src_pos);
+ imageStore(lightprobe_average_scroll_texture, dst_pos, value);
+ }
+ } else if (params.cascade < params.max_cascades - 1) {
+ //cant 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;
+ vec3 probe_pos = cascades.data[params.cascade].offset + vec3(probe_cell) * probe_cell_size;
+
+ //to parent local coords
+ 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);
+
+ ivec3 probe_posi = ivec3(probe_pos);
+ //add up all light, no need to use occlusion here, since occlusion will do its work afterwards
+
+ vec4 average_light[SH_SIZE] = vec4[](vec4(0), vec4(0), vec4(0), vec4(0), vec4(0), vec4(0), vec4(0), vec4(0), vec4(0)
+#if (SH_SIZE == 16)
+ ,
+ vec4(0), vec4(0), vec4(0), vec4(0), vec4(0), vec4(0), vec4(0)
+#endif
+ );
+ float total_weight = 0.0;
+
+ for (int i = 0; i < 8; i++) {
+ ivec3 offset = probe_posi + ((ivec3(i) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1));
+
+ vec3 trilinear = vec3(1.0) - abs(probe_pos - vec3(offset));
+ float weight = trilinear.x * trilinear.y * trilinear.z;
+
+ ivec2 tex_pos;
+ tex_pos = offset.xy;
+ tex_pos.x += offset.z * int(params.probe_axis_size);
+
+ for (int j = 0; j < SH_SIZE; j++) {
+ // copy from history texture
+ ivec2 src_pos = ivec2(tex_pos.x, tex_pos.y * SH_SIZE + j);
+ ivec4 average = imageLoad(lightprobe_average_parent_texture, src_pos);
+ vec4 value = (vec4(average) / float(params.history_size)) / float(1 << HISTORY_BITS);
+ average_light[j] += value * weight;
+ }
+
+ total_weight += weight;
+ }
+
+ if (total_weight > 0.0) {
+ total_weight = 1.0 / total_weight;
+ }
+ //store the averaged values everywhere
+
+ for (int i = 0; i < SH_SIZE; i++) {
+ ivec4 ivalue = clamp(ivec4(average_light[i] * total_weight * float(1 << HISTORY_BITS)), ivec4(-32768), ivec4(32767)); //clamp to 16 bits, so higher values don't break average
+ // copy from history texture
+ ivec3 dst_pos = ivec3(pos.x, pos.y * SH_SIZE + i, 0);
+ for (uint j = 0; j < params.history_size; j++) {
+ dst_pos.z = int(j);
+ imageStore(lightprobe_history_scroll_texture, dst_pos, ivalue);
+ }
+
+ ivalue *= int(params.history_size); //average needs to have all history added up
+ imageStore(lightprobe_average_scroll_texture, dst_pos.xy, ivalue);
+ }
+
+ } else {
+ // clear and let it re-raytrace, only for the last cascade, which happens very un-often
+ //scroll
+ for (uint j = 0; j < params.history_size; j++) {
+ for (int i = 0; i < SH_SIZE; i++) {
+ // copy from history texture
+ ivec3 dst_pos = ivec3(pos.x, pos.y * SH_SIZE + i, int(j));
+ imageStore(lightprobe_history_scroll_texture, dst_pos, ivec4(0));
+ }
+ }
+
+ 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));
+ }
+ }
+
+#endif
+
+#ifdef MODE_SCROLL_STORE
+
+ //do not update probe texture, as these will be updated later
+
+ for (uint j = 0; j < params.history_size; j++) {
+ for (int i = 0; i < SH_SIZE; i++) {
+ // copy from history texture
+ ivec3 spos = ivec3(pos.x, pos.y * SH_SIZE + i, int(j));
+ ivec4 value = imageLoad(lightprobe_history_scroll_texture, spos);
+ imageStore(lightprobe_history_texture, spos, value);
+ }
+ }
+
+ for (int i = 0; i < SH_SIZE; i++) {
+ // copy from average texture
+ ivec2 spos = ivec2(pos.x, pos.y * SH_SIZE + i);
+ ivec4 average = imageLoad(lightprobe_average_scroll_texture, spos);
+ imageStore(lightprobe_average_texture, spos, average);
+ }
+
+#endif
+}
diff --git a/servers/rendering/rasterizer_rd/shaders/sdfgi_preprocess.glsl b/servers/rendering/rasterizer_rd/shaders/sdfgi_preprocess.glsl
new file mode 100644
index 0000000000..7a523d3a3c
--- /dev/null
+++ b/servers/rendering/rasterizer_rd/shaders/sdfgi_preprocess.glsl
@@ -0,0 +1,1155 @@
+#[compute]
+
+#version 450
+
+VERSION_DEFINES
+
+#ifdef MODE_JUMPFLOOD_OPTIMIZED
+#define GROUP_SIZE 8
+
+layout(local_size_x = GROUP_SIZE, local_size_y = GROUP_SIZE, local_size_z = GROUP_SIZE) in;
+
+#elif defined(MODE_OCCLUSION) || defined(MODE_SCROLL)
+//buffer layout
+layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in;
+
+#else
+//grid layout
+layout(local_size_x = 4, local_size_y = 4, local_size_z = 4) in;
+
+#endif
+
+#if defined(MODE_INITIALIZE_JUMP_FLOOD) || defined(MODE_INITIALIZE_JUMP_FLOOD_HALF)
+layout(r16ui, set = 0, binding = 1) uniform restrict readonly uimage3D src_color;
+layout(rgba8ui, set = 0, binding = 2) uniform restrict writeonly uimage3D dst_positions;
+#endif
+
+#ifdef MODE_UPSCALE_JUMP_FLOOD
+layout(r16ui, set = 0, binding = 1) uniform restrict readonly uimage3D src_color;
+layout(rgba8ui, set = 0, binding = 2) uniform restrict readonly uimage3D src_positions_half;
+layout(rgba8ui, set = 0, binding = 3) uniform restrict writeonly uimage3D dst_positions;
+#endif
+
+#if defined(MODE_JUMPFLOOD) || defined(MODE_JUMPFLOOD_OPTIMIZED)
+layout(rgba8ui, set = 0, binding = 1) uniform restrict readonly uimage3D src_positions;
+layout(rgba8ui, set = 0, binding = 2) uniform restrict writeonly uimage3D dst_positions;
+#endif
+
+#ifdef MODE_JUMPFLOOD_OPTIMIZED
+
+shared uvec4 group_positions[(GROUP_SIZE + 2) * (GROUP_SIZE + 2) * (GROUP_SIZE + 2)]; //4x4x4 with margins
+
+void group_store(ivec3 p_pos, uvec4 p_value) {
+ uint offset = uint(p_pos.z * (GROUP_SIZE + 2) * (GROUP_SIZE + 2) + p_pos.y * (GROUP_SIZE + 2) + p_pos.x);
+ group_positions[offset] = p_value;
+}
+
+uvec4 group_load(ivec3 p_pos) {
+ uint offset = uint(p_pos.z * (GROUP_SIZE + 2) * (GROUP_SIZE + 2) + p_pos.y * (GROUP_SIZE + 2) + p_pos.x);
+ return group_positions[offset];
+}
+
+#endif
+
+#ifdef MODE_OCCLUSION
+
+layout(r16ui, set = 0, binding = 1) uniform restrict readonly uimage3D src_color;
+layout(r8, set = 0, binding = 2) uniform restrict image3D dst_occlusion[8];
+layout(r32ui, set = 0, binding = 3) uniform restrict readonly uimage3D src_facing;
+
+const uvec2 group_size_offset[11] = uvec2[](uvec2(1, 0), uvec2(3, 1), uvec2(6, 4), uvec2(10, 10), uvec2(15, 20), uvec2(21, 35), uvec2(28, 56), uvec2(36, 84), uvec2(42, 120), uvec2(46, 162), uvec2(48, 208));
+const uint group_pos[256] = uint[](0,
+ 65536, 256, 1,
+ 131072, 65792, 512, 65537, 257, 2,
+ 196608, 131328, 66048, 768, 131073, 65793, 513, 65538, 258, 3,
+ 262144, 196864, 131584, 66304, 1024, 196609, 131329, 66049, 769, 131074, 65794, 514, 65539, 259, 4,
+ 327680, 262400, 197120, 131840, 66560, 1280, 262145, 196865, 131585, 66305, 1025, 196610, 131330, 66050, 770, 131075, 65795, 515, 65540, 260, 5,
+ 393216, 327936, 262656, 197376, 132096, 66816, 1536, 327681, 262401, 197121, 131841, 66561, 1281, 262146, 196866, 131586, 66306, 1026, 196611, 131331, 66051, 771, 131076, 65796, 516, 65541, 261, 6,
+ 458752, 393472, 328192, 262912, 197632, 132352, 67072, 1792, 393217, 327937, 262657, 197377, 132097, 66817, 1537, 327682, 262402, 197122, 131842, 66562, 1282, 262147, 196867, 131587, 66307, 1027, 196612, 131332, 66052, 772, 131077, 65797, 517, 65542, 262, 7,
+ 459008, 393728, 328448, 263168, 197888, 132608, 67328, 458753, 393473, 328193, 262913, 197633, 132353, 67073, 1793, 393218, 327938, 262658, 197378, 132098, 66818, 1538, 327683, 262403, 197123, 131843, 66563, 1283, 262148, 196868, 131588, 66308, 1028, 196613, 131333, 66053, 773, 131078, 65798, 518, 65543, 263,
+ 459264, 393984, 328704, 263424, 198144, 132864, 459009, 393729, 328449, 263169, 197889, 132609, 67329, 458754, 393474, 328194, 262914, 197634, 132354, 67074, 1794, 393219, 327939, 262659, 197379, 132099, 66819, 1539, 327684, 262404, 197124, 131844, 66564, 1284, 262149, 196869, 131589, 66309, 1029, 196614, 131334, 66054, 774, 131079, 65799, 519,
+ 459520, 394240, 328960, 263680, 198400, 459265, 393985, 328705, 263425, 198145, 132865, 459010, 393730, 328450, 263170, 197890, 132610, 67330, 458755, 393475, 328195, 262915, 197635, 132355, 67075, 1795, 393220, 327940, 262660, 197380, 132100, 66820, 1540, 327685, 262405, 197125, 131845, 66565, 1285, 262150, 196870, 131590, 66310, 1030, 196615, 131335, 66055, 775);
+
+shared uint occlusion_facing[((OCCLUSION_SIZE * 2) * (OCCLUSION_SIZE * 2) * (OCCLUSION_SIZE * 2)) / 4];
+
+uint get_facing(ivec3 p_pos) {
+ uint ofs = uint(p_pos.z * OCCLUSION_SIZE * 2 * OCCLUSION_SIZE * 2 + p_pos.y * OCCLUSION_SIZE * 2 + p_pos.x);
+ uint v = occlusion_facing[ofs / 4];
+ return (v >> ((ofs % 4) * 8)) & 0xFF;
+}
+
+#endif
+
+#ifdef MODE_STORE
+
+layout(rgba8ui, set = 0, binding = 1) uniform restrict readonly uimage3D src_positions;
+layout(r16ui, set = 0, binding = 2) uniform restrict readonly uimage3D src_albedo;
+layout(r8, set = 0, binding = 3) uniform restrict readonly image3D src_occlusion[8];
+layout(r32ui, set = 0, binding = 4) uniform restrict readonly uimage3D src_light;
+layout(r32ui, set = 0, binding = 5) uniform restrict readonly uimage3D src_light_aniso;
+layout(r32ui, set = 0, binding = 6) uniform restrict readonly uimage3D src_facing;
+
+layout(r8, set = 0, binding = 7) uniform restrict writeonly image3D dst_sdf;
+layout(r16ui, set = 0, binding = 8) uniform restrict writeonly uimage3D dst_occlusion;
+
+layout(set = 0, binding = 10, std430) restrict buffer DispatchData {
+ uint x;
+ uint y;
+ uint z;
+ uint total_count;
+}
+dispatch_data;
+
+struct ProcessVoxel {
+ uint position; //xyz 7 bit packed, extra 11 bits for neigbours
+ uint albedo; //rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neibhbours
+ uint light; //rgbe8985 encoded total saved light, extra 2 bits for neighbous
+ uint light_aniso; //55555 light anisotropy, extra 2 bits for neighbours
+ //total neighbours: 26
+};
+
+layout(set = 0, binding = 11, std430) restrict buffer writeonly ProcessVoxels {
+ ProcessVoxel data[];
+}
+dst_process_voxels;
+
+shared ProcessVoxel store_positions[4 * 4 * 4];
+shared uint store_position_count;
+shared uint store_from_index;
+#endif
+
+#ifdef MODE_SCROLL
+
+layout(r16ui, set = 0, binding = 1) uniform restrict writeonly uimage3D dst_albedo;
+layout(r32ui, set = 0, binding = 2) uniform restrict writeonly uimage3D dst_facing;
+layout(r32ui, set = 0, binding = 3) uniform restrict writeonly uimage3D dst_light;
+layout(r32ui, set = 0, binding = 4) uniform restrict writeonly uimage3D dst_light_aniso;
+
+layout(set = 0, binding = 5, std430) restrict buffer readonly DispatchData {
+ uint x;
+ uint y;
+ uint z;
+ uint total_count;
+}
+dispatch_data;
+
+struct ProcessVoxel {
+ uint position; //xyz 7 bit packed, extra 11 bits for neigbours
+ uint albedo; //rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neibhbours
+ uint light; //rgbe8985 encoded total saved light, extra 2 bits for neighbous
+ uint light_aniso; //55555 light anisotropy, extra 2 bits for neighbours
+ //total neighbours: 26
+};
+
+layout(set = 0, binding = 6, std430) restrict buffer readonly ProcessVoxels {
+ ProcessVoxel data[];
+}
+src_process_voxels;
+
+#endif
+
+#ifdef MODE_SCROLL_OCCLUSION
+
+layout(r8, set = 0, binding = 1) uniform restrict image3D dst_occlusion[8];
+layout(r16ui, set = 0, binding = 2) uniform restrict readonly uimage3D src_occlusion;
+
+#endif
+
+layout(push_constant, binding = 0, std430) uniform Params {
+ ivec3 scroll;
+
+ int grid_size;
+
+ ivec3 probe_offset;
+ int step_size;
+
+ bool half_size;
+ uint occlusion_index;
+ int cascade;
+ uint pad;
+}
+params;
+
+void main() {
+#ifdef MODE_SCROLL
+
+ // Pixel being shaded
+ int index = int(gl_GlobalInvocationID.x);
+ if (index >= dispatch_data.total_count) { //too big
+ return;
+ }
+
+ ivec3 read_pos = (ivec3(src_process_voxels.data[index].position) >> ivec3(0, 7, 14)) & ivec3(0x7F);
+ ivec3 write_pos = read_pos + params.scroll;
+
+ if (any(lessThan(write_pos, ivec3(0))) || any(greaterThanEqual(write_pos, ivec3(params.grid_size)))) {
+ return; //fits outside the 3D texture, dont do anything
+ }
+
+ uint albedo = ((src_process_voxels.data[index].albedo & 0x7FFF) << 1) | 1; //add solid bit
+ imageStore(dst_albedo, write_pos, uvec4(albedo));
+
+ uint facing = (src_process_voxels.data[index].albedo >> 15) & 0x3F; //6 anisotropic facing bits
+ imageStore(dst_facing, write_pos, uvec4(facing));
+
+ uint light = src_process_voxels.data[index].light & 0x3fffffff; //30 bits of RGBE8985
+ imageStore(dst_light, write_pos, uvec4(light));
+
+ uint light_aniso = src_process_voxels.data[index].light_aniso & 0x3fffffff; //30 bits of 6 anisotropic 5 bits values
+ imageStore(dst_light_aniso, write_pos, uvec4(light_aniso));
+
+#endif
+
+#ifdef MODE_SCROLL_OCCLUSION
+
+ ivec3 pos = ivec3(gl_GlobalInvocationID.xyz);
+ if (any(greaterThanEqual(pos, ivec3(params.grid_size) - abs(params.scroll)))) { //too large, do nothing
+ return;
+ }
+
+ ivec3 read_pos = pos + max(ivec3(0), -params.scroll);
+ ivec3 write_pos = pos + max(ivec3(0), params.scroll);
+
+ read_pos.z += params.cascade * params.grid_size;
+ uint occlusion = imageLoad(src_occlusion, read_pos).r;
+ read_pos.x += params.grid_size;
+ occlusion |= imageLoad(src_occlusion, read_pos).r << 16;
+
+ const uint occlusion_shift[8] = uint[](12, 8, 4, 0, 28, 24, 20, 16);
+
+ for (uint i = 0; i < 8; i++) {
+ float o = float((occlusion >> occlusion_shift[i]) & 0xF) / 15.0;
+ imageStore(dst_occlusion[i], write_pos, vec4(o));
+ }
+
+#endif
+
+#ifdef MODE_INITIALIZE_JUMP_FLOOD
+
+ ivec3 pos = ivec3(gl_GlobalInvocationID.xyz);
+
+ uint c = imageLoad(src_color, pos).r;
+ uvec4 v;
+ if (bool(c & 0x1)) {
+ //bit set means this is solid
+ v.xyz = uvec3(pos);
+ v.w = 255; //not zero means used
+ } else {
+ v.xyz = uvec3(0);
+ v.w = 0; // zero means unused
+ }
+
+ imageStore(dst_positions, pos, v);
+#endif
+
+#ifdef MODE_INITIALIZE_JUMP_FLOOD_HALF
+
+ ivec3 pos = ivec3(gl_GlobalInvocationID.xyz);
+ ivec3 base_pos = pos * 2;
+
+ //since we store in half size, lets kind of randomize what we store, so
+ //the half size jump flood has a bit better chance to find something
+ uvec4 closest[8];
+ int closest_count = 0;
+
+ for (uint i = 0; i < 8; i++) {
+ ivec3 src_pos = base_pos + ((ivec3(i) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1));
+ uint c = imageLoad(src_color, src_pos).r;
+ if (bool(c & 1)) {
+ uvec4 v = uvec4(uvec3(src_pos), 255);
+ closest[closest_count] = v;
+ closest_count++;
+ }
+ }
+
+ if (closest_count == 0) {
+ imageStore(dst_positions, pos, uvec4(0));
+ } else {
+ ivec3 indexv = (pos & ivec3(1, 1, 1)) * ivec3(1, 2, 4);
+ int index = (indexv.x | indexv.y | indexv.z) % closest_count;
+ imageStore(dst_positions, pos, closest[index]);
+ }
+
+#endif
+
+#ifdef MODE_JUMPFLOOD
+
+ //regular jumpflood, efficent for large steps, inefficient for small steps
+ ivec3 pos = ivec3(gl_GlobalInvocationID.xyz);
+
+ vec3 posf = vec3(pos);
+
+ if (params.half_size) {
+ posf = posf * 2.0 + 0.5;
+ }
+
+ uvec4 p = imageLoad(src_positions, pos);
+
+ if (!params.half_size && p == uvec4(uvec3(pos), 255)) {
+ imageStore(dst_positions, pos, p);
+ return; //points to itself and valid, nothing better can be done, just pass
+ }
+
+ float p_dist;
+
+ if (p.w != 0) {
+ p_dist = distance(posf, vec3(p.xyz));
+ } else {
+ p_dist = 0.0; //should not matter
+ }
+
+ const uint offset_count = 26;
+ const ivec3 offsets[offset_count] = ivec3[](
+ ivec3(-1, -1, -1),
+ ivec3(-1, -1, 0),
+ ivec3(-1, -1, 1),
+ ivec3(-1, 0, -1),
+ ivec3(-1, 0, 0),
+ ivec3(-1, 0, 1),
+ ivec3(-1, 1, -1),
+ ivec3(-1, 1, 0),
+ ivec3(-1, 1, 1),
+ ivec3(0, -1, -1),
+ ivec3(0, -1, 0),
+ ivec3(0, -1, 1),
+ ivec3(0, 0, -1),
+ ivec3(0, 0, 1),
+ ivec3(0, 1, -1),
+ ivec3(0, 1, 0),
+ ivec3(0, 1, 1),
+ ivec3(1, -1, -1),
+ ivec3(1, -1, 0),
+ ivec3(1, -1, 1),
+ ivec3(1, 0, -1),
+ ivec3(1, 0, 0),
+ ivec3(1, 0, 1),
+ ivec3(1, 1, -1),
+ ivec3(1, 1, 0),
+ ivec3(1, 1, 1));
+
+ for (uint i = 0; i < offset_count; i++) {
+ ivec3 ofs = pos + offsets[i] * params.step_size;
+ if (any(lessThan(ofs, ivec3(0))) || any(greaterThanEqual(ofs, ivec3(params.grid_size)))) {
+ continue;
+ }
+ uvec4 q = imageLoad(src_positions, ofs);
+
+ if (q.w == 0) {
+ continue; //was not initialized yet, ignore
+ }
+
+ float q_dist = distance(posf, vec3(p.xyz));
+ if (p.w == 0 || q_dist < p_dist) {
+ p = q; //just replace because current is unused
+ p_dist = q_dist;
+ }
+ }
+
+ imageStore(dst_positions, pos, p);
+#endif
+
+#ifdef MODE_JUMPFLOOD_OPTIMIZED
+ //optimized version using shared compute memory
+
+ ivec3 group_offset = ivec3(gl_WorkGroupID.xyz) % params.step_size;
+ ivec3 group_pos = group_offset + (ivec3(gl_WorkGroupID.xyz) / params.step_size) * ivec3(GROUP_SIZE * params.step_size);
+
+ //load data into local group memory
+
+ if (all(lessThan(ivec3(gl_LocalInvocationID.xyz), ivec3((GROUP_SIZE + 2) / 2)))) {
+ //use this thread for loading, this method uses less threads for this but its simpler and less divergent
+ ivec3 base_pos = ivec3(gl_LocalInvocationID.xyz) * 2;
+ for (uint i = 0; i < 8; i++) {
+ ivec3 load_pos = base_pos + ((ivec3(i) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1));
+ ivec3 load_global_pos = group_pos + (load_pos - ivec3(1)) * params.step_size;
+ uvec4 q;
+ if (all(greaterThanEqual(load_global_pos, ivec3(0))) && all(lessThan(load_global_pos, ivec3(params.grid_size)))) {
+ q = imageLoad(src_positions, load_global_pos);
+ } else {
+ q = uvec4(0); //unused
+ }
+
+ group_store(load_pos, q);
+ }
+ }
+
+ ivec3 global_pos = group_pos + ivec3(gl_LocalInvocationID.xyz) * params.step_size;
+
+ if (any(lessThan(global_pos, ivec3(0))) || any(greaterThanEqual(global_pos, ivec3(params.grid_size)))) {
+ return; //do nothing else, end here because outside range
+ }
+
+ //sync
+ groupMemoryBarrier();
+ barrier();
+
+ ivec3 local_pos = ivec3(gl_LocalInvocationID.xyz) + ivec3(1);
+
+ const uint offset_count = 27;
+ const ivec3 offsets[offset_count] = ivec3[](
+ ivec3(-1, -1, -1),
+ ivec3(-1, -1, 0),
+ ivec3(-1, -1, 1),
+ ivec3(-1, 0, -1),
+ ivec3(-1, 0, 0),
+ ivec3(-1, 0, 1),
+ ivec3(-1, 1, -1),
+ ivec3(-1, 1, 0),
+ ivec3(-1, 1, 1),
+ ivec3(0, -1, -1),
+ ivec3(0, -1, 0),
+ ivec3(0, -1, 1),
+ ivec3(0, 0, -1),
+ ivec3(0, 0, 0),
+ ivec3(0, 0, 1),
+ ivec3(0, 1, -1),
+ ivec3(0, 1, 0),
+ ivec3(0, 1, 1),
+ ivec3(1, -1, -1),
+ ivec3(1, -1, 0),
+ ivec3(1, -1, 1),
+ ivec3(1, 0, -1),
+ ivec3(1, 0, 0),
+ ivec3(1, 0, 1),
+ ivec3(1, 1, -1),
+ ivec3(1, 1, 0),
+ ivec3(1, 1, 1));
+
+ //only makes sense if point is inside screen
+ uvec4 closest = uvec4(0);
+ float closest_dist = 0.0;
+
+ vec3 posf = vec3(global_pos);
+
+ if (params.half_size) {
+ posf = posf * 2.0 + 0.5;
+ }
+
+ for (uint i = 0; i < offset_count; i++) {
+ uvec4 point = group_load(local_pos + offsets[i]);
+
+ if (point.w == 0) {
+ continue; //was not initialized yet, ignore
+ }
+
+ float dist = distance(posf, vec3(point.xyz));
+ if (closest.w == 0 || dist < closest_dist) {
+ closest = point;
+ closest_dist = dist;
+ }
+ }
+
+ imageStore(dst_positions, global_pos, closest);
+
+#endif
+
+#ifdef MODE_UPSCALE_JUMP_FLOOD
+
+ ivec3 pos = ivec3(gl_GlobalInvocationID.xyz);
+
+ uint c = imageLoad(src_color, pos).r;
+ uvec4 v;
+ if (bool(c & 1)) {
+ //bit set means this is solid
+ v.xyz = uvec3(pos);
+ v.w = 255; //not zero means used
+ } else {
+ v = imageLoad(src_positions_half, pos >> 1);
+ float d = length(vec3(ivec3(v.xyz) - pos));
+
+ ivec3 vbase = ivec3(v.xyz - (v.xyz & uvec3(1)));
+
+ //search around if there is a better candidate from the same block
+ for (int i = 0; i < 8; i++) {
+ ivec3 bits = ((ivec3(i) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1));
+ ivec3 p = vbase + bits;
+
+ float d2 = length(vec3(p - pos));
+ if (d2 < d) { //check valid distance before test so we avoid a read
+ uint c2 = imageLoad(src_color, p).r;
+ if (bool(c2 & 1)) {
+ v.xyz = uvec3(p);
+ d = d2;
+ }
+ }
+ }
+
+ //could validate better position..
+ }
+
+ imageStore(dst_positions, pos, v);
+
+#endif
+
+#ifdef MODE_OCCLUSION
+
+ uint invocation_idx = uint(gl_LocalInvocationID.x);
+ ivec3 region = ivec3(gl_WorkGroupID);
+
+ ivec3 region_offset = -ivec3(OCCLUSION_SIZE);
+ region_offset += region * OCCLUSION_SIZE * 2;
+ region_offset += params.probe_offset * OCCLUSION_SIZE;
+
+ if (params.scroll != ivec3(0)) {
+ //validate scroll region
+ ivec3 region_offset_to = region_offset + ivec3(OCCLUSION_SIZE * 2);
+ uvec3 scroll_mask = uvec3(notEqual(params.scroll, ivec3(0))); //save which axes acre scrolling
+ ivec3 scroll_from = mix(ivec3(0), ivec3(params.grid_size) + params.scroll, lessThan(params.scroll, ivec3(0)));
+ ivec3 scroll_to = mix(ivec3(params.grid_size), params.scroll, greaterThan(params.scroll, ivec3(0)));
+
+ if ((uvec3(lessThanEqual(region_offset_to, scroll_from)) | uvec3(greaterThanEqual(region_offset, scroll_to))) * scroll_mask == scroll_mask) { //all axes that scroll are out, exit
+ return; //region outside scroll bounds, quit
+ }
+ }
+
+#define OCC_HALF_SIZE (OCCLUSION_SIZE / 2)
+
+ ivec3 local_ofs = ivec3(uvec3(invocation_idx % OCC_HALF_SIZE, (invocation_idx % (OCC_HALF_SIZE * OCC_HALF_SIZE)) / OCC_HALF_SIZE, invocation_idx / (OCC_HALF_SIZE * OCC_HALF_SIZE))) * 4;
+
+ /* for(int i=0;i<64;i++) {
+ ivec3 offset = region_offset + local_ofs + ((ivec3(i) >> ivec3(0,2,4)) & ivec3(3,3,3));
+ uint facig =
+ if (all(greaterThanEqual(offset,ivec3(0))) && all(lessThan(offset,ivec3(params.grid_size)))) {*/
+
+ for (int i = 0; i < 16; i++) { //skip x, so it can be packed
+
+ ivec3 offset = local_ofs + ((ivec3(i * 4) >> ivec3(0, 2, 4)) & ivec3(3, 3, 3));
+
+ uint facing_pack = 0;
+ for (int j = 0; j < 4; j++) {
+ ivec3 foffset = region_offset + offset + ivec3(j, 0, 0);
+ if (all(greaterThanEqual(foffset, ivec3(0))) && all(lessThan(foffset, ivec3(params.grid_size)))) {
+ uint f = imageLoad(src_facing, foffset).r;
+ facing_pack |= f << (j * 8);
+ }
+ }
+
+ occlusion_facing[(offset.z * (OCCLUSION_SIZE * 2 * OCCLUSION_SIZE * 2) + offset.y * (OCCLUSION_SIZE * 2) + offset.x) / 4] = facing_pack;
+ }
+
+ //sync occlusion saved
+ groupMemoryBarrier();
+ barrier();
+
+ //process occlusion
+
+#define OCC_STEPS (OCCLUSION_SIZE * 3 - 2)
+#define OCC_HALF_STEPS (OCC_STEPS / 2)
+
+ for (int step = 0; step < OCC_STEPS; step++) {
+ bool shrink = step >= OCC_HALF_STEPS;
+ int occ_step = shrink ? OCC_HALF_STEPS - (step - OCC_HALF_STEPS) - 1 : step;
+
+ if (invocation_idx < group_size_offset[occ_step].x) {
+ uint pv = group_pos[group_size_offset[occ_step].y + invocation_idx];
+ ivec3 proc_abs = (ivec3(int(pv)) >> ivec3(0, 8, 16)) & ivec3(0xFF);
+
+ if (shrink) {
+ proc_abs = ivec3(OCCLUSION_SIZE) - proc_abs - ivec3(1);
+ }
+
+ for (int i = 0; i < 8; i++) {
+ ivec3 bits = ((ivec3(i) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1));
+ ivec3 proc_sign = bits * 2 - 1;
+ ivec3 local_offset = ivec3(OCCLUSION_SIZE) + proc_abs * proc_sign - (ivec3(1) - bits);
+ ivec3 offset = local_offset + region_offset;
+ if (all(greaterThanEqual(offset, ivec3(0))) && all(lessThan(offset, ivec3(params.grid_size)))) {
+ float occ;
+
+ uint facing = get_facing(local_offset);
+
+ if (facing != 0) { //solid
+ occ = 0.0;
+ } else if (step == 0) {
+#if 0
+ occ = 0.0;
+ if (get_facing(local_offset - ivec3(proc_sign.x,0,0))==0) {
+ occ+=1.0;
+ }
+ if (get_facing(local_offset - ivec3(0,proc_sign.y,0))==0) {
+ occ+=1.0;
+ }
+ if (get_facing(local_offset - ivec3(0,0,proc_sign.z))==0) {
+ occ+=1.0;
+ }
+ /*
+ if (get_facing(local_offset - proc_sign)==0) {
+ occ+=1.0;
+ }*/
+
+ occ/=3.0;
+#endif
+ occ = 1.0;
+
+ } else {
+ ivec3 read_dir = -proc_sign;
+
+ ivec3 major_axis;
+ if (proc_abs.x < proc_abs.y) {
+ if (proc_abs.z < proc_abs.y) {
+ major_axis = ivec3(0, 1, 0);
+ } else {
+ major_axis = ivec3(0, 0, 1);
+ }
+ } else {
+ if (proc_abs.z < proc_abs.x) {
+ major_axis = ivec3(1, 0, 0);
+ } else {
+ major_axis = ivec3(0, 0, 1);
+ }
+ }
+
+ float avg = 0.0;
+ occ = 0.0;
+
+ ivec3 read_x = offset + ivec3(read_dir.x, 0, 0) + (proc_abs.x == 0 ? major_axis * read_dir : ivec3(0));
+ ivec3 read_y = offset + ivec3(0, read_dir.y, 0) + (proc_abs.y == 0 ? major_axis * read_dir : ivec3(0));
+ ivec3 read_z = offset + ivec3(0, 0, read_dir.z) + (proc_abs.z == 0 ? major_axis * read_dir : ivec3(0));
+
+ uint facing_x = get_facing(read_x - region_offset);
+ if (facing_x == 0) {
+ if (all(greaterThanEqual(read_x, ivec3(0))) && all(lessThan(read_x, ivec3(params.grid_size)))) {
+ occ += imageLoad(dst_occlusion[params.occlusion_index], read_x).r;
+ avg += 1.0;
+ }
+ } else {
+ if (proc_abs.x != 0) { //do not occlude from voxels in the opposite octant
+ avg += 1.0;
+ }
+ }
+
+ uint facing_y = get_facing(read_y - region_offset);
+ if (facing_y == 0) {
+ if (all(greaterThanEqual(read_y, ivec3(0))) && all(lessThan(read_y, ivec3(params.grid_size)))) {
+ occ += imageLoad(dst_occlusion[params.occlusion_index], read_y).r;
+ avg += 1.0;
+ }
+ } else {
+ if (proc_abs.y != 0) {
+ avg += 1.0;
+ }
+ }
+
+ uint facing_z = get_facing(read_z - region_offset);
+ if (facing_z == 0) {
+ if (all(greaterThanEqual(read_z, ivec3(0))) && all(lessThan(read_z, ivec3(params.grid_size)))) {
+ occ += imageLoad(dst_occlusion[params.occlusion_index], read_z).r;
+ avg += 1.0;
+ }
+ } else {
+ if (proc_abs.z != 0) {
+ avg += 1.0;
+ }
+ }
+
+ if (avg > 0.0) {
+ occ /= avg;
+ }
+ }
+
+ imageStore(dst_occlusion[params.occlusion_index], offset, vec4(occ));
+ }
+ }
+ }
+
+ groupMemoryBarrier();
+ barrier();
+ }
+ /*
+ for(int s=0;s<8;s++) {
+ ivec3 local_pos = pos * 2 + ((ivec3(s) >> ivec3(2,1,0)) & ivec3(1,1,1));
+ for(int i=0;i<8;i++) {
+ ivec3 offset = local_pos + ((ivec3(i) >> ivec3(2,1,0)) & ivec3(1,1,1)) * OCCLUSION_SIZE;
+
+ ivec3 global_offset = region_offset + offset;
+
+ if (any(lessThan(global_offset,ivec3(0))) || any(greaterThanEqual(global_offset,ivec3(params.grid_size)))) {
+ continue; // do not process
+ }
+
+ ivec3 proc_pos = offset - ivec3(OCCLUSION_SIZE);
+ proc_pos += mix(ivec3(0),ivec3(1),greaterThanEqual(proc_pos,ivec3(0)));
+
+ ivec3 proc_abs = abs(proc_pos);
+ ivec3 proc_pass = proc_abs - ivec3(1);
+
+ if (proc_pass.x+proc_pass.y+proc_pass.z != step) {
+ continue;
+ }
+
+ offset = global_offset;
+
+ float occ;
+
+ uint facing = imageLoad(src_facing,offset).r; //any bit present in facing means this is solid
+
+ if (facing != 0) { //solid
+ occ = 0.0;
+ } else if (step==0) {
+ occ = 1.0; //first step
+ } else {
+ ivec3 read_dir = -sign(proc_pos);
+
+ ivec3 major_axis;
+ if (proc_pass.x < proc_pass.y) {
+ if (proc_pass.z < proc_pass.y) {
+ major_axis = ivec3(0,1,0);
+ } else {
+ major_axis = ivec3(0,0,1);
+ }
+ } else {
+ if (proc_pass.z < proc_pass.x) {
+ major_axis = ivec3(1,0,0);
+ } else {
+ major_axis = ivec3(0,0,1);
+ }
+ }
+
+ float avg = 0.0;
+ occ = 0.0;
+
+ ivec3 read_x = offset + ivec3(read_dir.x,0,0) + (proc_pass.x == 0 ? major_axis * read_dir : ivec3(0));
+ ivec3 read_y = offset + ivec3(0,read_dir.y,0) + (proc_pass.y == 0 ? major_axis * read_dir : ivec3(0));
+ ivec3 read_z = offset + ivec3(0,0,read_dir.z) + (proc_pass.z == 0 ? major_axis * read_dir : ivec3(0));
+
+ if (all(greaterThanEqual(read_x,ivec3(0))) && all(lessThan(read_x,ivec3(params.grid_size)))) {
+
+ uint f = imageLoad(src_facing,read_x).r;
+ if (f==0) { //non solid
+ occ += imageLoad(dst_occlusion[params.occlusion_index],read_x).r;
+ avg+=1.0;
+ }
+ }
+
+ if (all(greaterThanEqual(read_y,ivec3(0))) && all(lessThan(read_y,ivec3(params.grid_size)))) {
+
+ uint f = imageLoad(src_facing,read_y).r;
+ if (f==0) {//non solid
+ occ += imageLoad(dst_occlusion[params.occlusion_index],read_y).r;
+ avg+=1.0;
+ }
+ }
+
+ if (all(greaterThanEqual(read_z,ivec3(0))) && all(lessThan(read_z,ivec3(params.grid_size)))) {
+
+ uint f = imageLoad(src_facing,read_z).r;
+ if (f==0) {//non solid
+ occ += imageLoad(dst_occlusion[params.occlusion_index],read_z).r;
+ avg+=1.0;
+ }
+ }
+
+ if (avg > 0.0) {
+ occ/=avg;
+ }
+ }
+
+ imageStore(dst_occlusion[params.occlusion_index],offset,vec4(occ));
+
+ }
+ }
+
+ groupMemoryBarrier();
+ barrier();
+
+ }
+*/
+#if 1
+ //bias solid voxels away
+
+ for (int i = 0; i < 64; i++) {
+ ivec3 local_offset = local_ofs + ((ivec3(i) >> ivec3(0, 2, 4)) & ivec3(3, 3, 3));
+ ivec3 offset = region_offset + local_offset;
+
+ if (all(greaterThanEqual(offset, ivec3(0))) && all(lessThan(offset, ivec3(params.grid_size)))) {
+ uint facing = get_facing(local_offset);
+
+ if (facing != 0) {
+ //only work on solids
+
+ ivec3 proc_pos = local_offset - ivec3(OCCLUSION_SIZE);
+ proc_pos += mix(ivec3(0), ivec3(1), greaterThanEqual(proc_pos, ivec3(0)));
+
+ float avg = 0.0;
+ float occ = 0.0;
+
+ ivec3 read_dir = -sign(proc_pos);
+ ivec3 read_dir_x = ivec3(read_dir.x, 0, 0);
+ ivec3 read_dir_y = ivec3(0, read_dir.y, 0);
+ ivec3 read_dir_z = ivec3(0, 0, read_dir.z);
+ //solid
+#if 0
+
+ uvec3 facing_pos_base = (uvec3(facing) >> uvec3(0,1,2)) & uvec3(1,1,1);
+ uvec3 facing_neg_base = (uvec3(facing) >> uvec3(3,4,5)) & uvec3(1,1,1);
+ uvec3 facing_pos= facing_pos_base &((~facing_neg_base)&uvec3(1,1,1));
+ uvec3 facing_neg= facing_neg_base &((~facing_pos_base)&uvec3(1,1,1));
+#else
+ uvec3 facing_pos = (uvec3(facing) >> uvec3(0, 1, 2)) & uvec3(1, 1, 1);
+ uvec3 facing_neg = (uvec3(facing) >> uvec3(3, 4, 5)) & uvec3(1, 1, 1);
+#endif
+ bvec3 read_valid = bvec3(mix(facing_neg, facing_pos, greaterThan(read_dir, ivec3(0))));
+
+ //sides
+ if (read_valid.x) {
+ ivec3 read_offset = local_offset + read_dir_x;
+ uint f = get_facing(read_offset);
+ if (f == 0) {
+ read_offset += region_offset;
+ if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) {
+ occ += imageLoad(dst_occlusion[params.occlusion_index], read_offset).r;
+ avg += 1.0;
+ }
+ }
+ }
+
+ if (read_valid.y) {
+ ivec3 read_offset = local_offset + read_dir_y;
+ uint f = get_facing(read_offset);
+ if (f == 0) {
+ read_offset += region_offset;
+ if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) {
+ occ += imageLoad(dst_occlusion[params.occlusion_index], read_offset).r;
+ avg += 1.0;
+ }
+ }
+ }
+
+ if (read_valid.z) {
+ ivec3 read_offset = local_offset + read_dir_z;
+ uint f = get_facing(read_offset);
+ if (f == 0) {
+ read_offset += region_offset;
+ if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) {
+ occ += imageLoad(dst_occlusion[params.occlusion_index], read_offset).r;
+ avg += 1.0;
+ }
+ }
+ }
+
+ //adjacents
+
+ if (all(read_valid.yz)) {
+ ivec3 read_offset = local_offset + read_dir_y + read_dir_z;
+ uint f = get_facing(read_offset);
+ if (f == 0) {
+ read_offset += region_offset;
+ if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) {
+ occ += imageLoad(dst_occlusion[params.occlusion_index], read_offset).r;
+ avg += 1.0;
+ }
+ }
+ }
+
+ if (all(read_valid.xz)) {
+ ivec3 read_offset = local_offset + read_dir_x + read_dir_z;
+ uint f = get_facing(read_offset);
+ if (f == 0) {
+ read_offset += region_offset;
+ if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) {
+ occ += imageLoad(dst_occlusion[params.occlusion_index], read_offset).r;
+ avg += 1.0;
+ }
+ }
+ }
+
+ if (all(read_valid.xy)) {
+ ivec3 read_offset = local_offset + read_dir_x + read_dir_y;
+ uint f = get_facing(read_offset);
+ if (f == 0) {
+ read_offset += region_offset;
+ if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) {
+ occ += imageLoad(dst_occlusion[params.occlusion_index], read_offset).r;
+ avg += 1.0;
+ }
+ }
+ }
+
+ //diagonal
+
+ if (all(read_valid)) {
+ ivec3 read_offset = local_offset + read_dir;
+ uint f = get_facing(read_offset);
+ if (f == 0) {
+ read_offset += region_offset;
+ if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) {
+ occ += imageLoad(dst_occlusion[params.occlusion_index], read_offset).r;
+ avg += 1.0;
+ }
+ }
+ }
+
+ if (avg > 0.0) {
+ occ /= avg;
+ }
+
+ imageStore(dst_occlusion[params.occlusion_index], offset, vec4(occ));
+ }
+ }
+ }
+
+#endif
+
+#if 1
+ groupMemoryBarrier();
+ barrier();
+
+ for (int i = 0; i < 64; i++) {
+ ivec3 local_offset = local_ofs + ((ivec3(i) >> ivec3(0, 2, 4)) & ivec3(3, 3, 3));
+ ivec3 offset = region_offset + local_offset;
+
+ if (all(greaterThanEqual(offset, ivec3(0))) && all(lessThan(offset, ivec3(params.grid_size)))) {
+ uint facing = get_facing(local_offset);
+
+ if (facing == 0) {
+ ivec3 proc_pos = local_offset - ivec3(OCCLUSION_SIZE);
+ proc_pos += mix(ivec3(0), ivec3(1), greaterThanEqual(proc_pos, ivec3(0)));
+
+ ivec3 proc_abs = abs(proc_pos);
+
+ ivec3 read_dir = sign(proc_pos); //opposite direction
+ ivec3 read_dir_x = ivec3(read_dir.x, 0, 0);
+ ivec3 read_dir_y = ivec3(0, read_dir.y, 0);
+ ivec3 read_dir_z = ivec3(0, 0, read_dir.z);
+ //solid
+ uvec3 read_mask = mix(uvec3(1, 2, 4), uvec3(8, 16, 32), greaterThan(read_dir, ivec3(0))); //match positive with negative normals
+ uvec3 block_mask = mix(uvec3(1, 2, 4), uvec3(8, 16, 32), lessThan(read_dir, ivec3(0))); //match positive with negative normals
+
+ block_mask = uvec3(0);
+
+ float visible = 0.0;
+ float occlude_total = 0.0;
+
+ if (proc_abs.x < OCCLUSION_SIZE) {
+ ivec3 read_offset = local_offset + read_dir_x;
+ uint x_mask = get_facing(read_offset);
+ if (x_mask != 0) {
+ read_offset += region_offset;
+ if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) {
+ occlude_total += 1.0;
+ if (bool(x_mask & read_mask.x) && !bool(x_mask & block_mask.x)) {
+ visible += 1.0;
+ }
+ }
+ }
+ }
+
+ if (proc_abs.y < OCCLUSION_SIZE) {
+ ivec3 read_offset = local_offset + read_dir_y;
+ uint y_mask = get_facing(read_offset);
+ if (y_mask != 0) {
+ read_offset += region_offset;
+ if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) {
+ occlude_total += 1.0;
+ if (bool(y_mask & read_mask.y) && !bool(y_mask & block_mask.y)) {
+ visible += 1.0;
+ }
+ }
+ }
+ }
+
+ if (proc_abs.z < OCCLUSION_SIZE) {
+ ivec3 read_offset = local_offset + read_dir_z;
+ uint z_mask = get_facing(read_offset);
+ if (z_mask != 0) {
+ read_offset += region_offset;
+ if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) {
+ occlude_total += 1.0;
+ if (bool(z_mask & read_mask.z) && !bool(z_mask & block_mask.z)) {
+ visible += 1.0;
+ }
+ }
+ }
+ }
+
+ //if near the cartesian plane, test in opposite direction too
+
+ read_mask = mix(uvec3(1, 2, 4), uvec3(8, 16, 32), lessThan(read_dir, ivec3(0))); //match negative with positive normals
+ block_mask = mix(uvec3(1, 2, 4), uvec3(8, 16, 32), greaterThan(read_dir, ivec3(0))); //match negative with positive normals
+ block_mask = uvec3(0);
+
+ if (proc_abs.x == 1) {
+ ivec3 read_offset = local_offset - read_dir_x;
+ uint x_mask = get_facing(read_offset);
+ if (x_mask != 0) {
+ read_offset += region_offset;
+ if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) {
+ occlude_total += 1.0;
+ if (bool(x_mask & read_mask.x) && !bool(x_mask & block_mask.x)) {
+ visible += 1.0;
+ }
+ }
+ }
+ }
+
+ if (proc_abs.y == 1) {
+ ivec3 read_offset = local_offset - read_dir_y;
+ uint y_mask = get_facing(read_offset);
+ if (y_mask != 0) {
+ read_offset += region_offset;
+ if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) {
+ occlude_total += 1.0;
+ if (bool(y_mask & read_mask.y) && !bool(y_mask & block_mask.y)) {
+ visible += 1.0;
+ }
+ }
+ }
+ }
+
+ if (proc_abs.z == 1) {
+ ivec3 read_offset = local_offset - read_dir_z;
+ uint z_mask = get_facing(read_offset);
+ if (z_mask != 0) {
+ read_offset += region_offset;
+ if (all(greaterThanEqual(read_offset, ivec3(0))) && all(lessThan(read_offset, ivec3(params.grid_size)))) {
+ occlude_total += 1.0;
+ if (bool(z_mask & read_mask.z) && !bool(z_mask & block_mask.z)) {
+ visible += 1.0;
+ }
+ }
+ }
+ }
+
+ if (occlude_total > 0.0) {
+ float occ = imageLoad(dst_occlusion[params.occlusion_index], offset).r;
+ occ *= visible / occlude_total;
+ imageStore(dst_occlusion[params.occlusion_index], offset, vec4(occ));
+ }
+ }
+ }
+ }
+
+#endif
+
+ /*
+ for(int i=0;i<8;i++) {
+ ivec3 local_offset = local_pos + ((ivec3(i) >> ivec3(2,1,0)) & ivec3(1,1,1)) * OCCLUSION_SIZE;
+ ivec3 offset = local_offset - ivec3(OCCLUSION_SIZE); //looking around probe, so starts negative
+ offset += region * OCCLUSION_SIZE * 2; //offset by region
+ offset += params.probe_offset * OCCLUSION_SIZE; // offset by probe offset
+ if (all(greaterThanEqual(offset,ivec3(0))) && all(lessThan(offset,ivec3(params.grid_size)))) {
+ imageStore(dst_occlusion[params.occlusion_index],offset,vec4( occlusion_data[ to_linear(local_offset) ] ));
+ //imageStore(dst_occlusion[params.occlusion_index],offset,vec4( occlusion_solid[ to_linear(local_offset) ] ));
+ }
+ }
+*/
+
+#endif
+
+#ifdef MODE_STORE
+
+ ivec3 local = ivec3(gl_LocalInvocationID.xyz);
+ ivec3 pos = ivec3(gl_GlobalInvocationID.xyz);
+ // store SDF
+ uvec4 p = imageLoad(src_positions, pos);
+
+ bool solid = false;
+ float d;
+ if (ivec3(p.xyz) == pos) {
+ //solid block
+ d = 0;
+ solid = true;
+ } else {
+ //distance block
+ d = 1.0 + length(vec3(p.xyz) - vec3(pos));
+ }
+
+ d /= 255.0;
+
+ imageStore(dst_sdf, pos, vec4(d));
+
+ // STORE OCCLUSION
+
+ uint occlusion = 0;
+ const uint occlusion_shift[8] = uint[](12, 8, 4, 0, 28, 24, 20, 16);
+ for (int i = 0; i < 8; i++) {
+ float occ = imageLoad(src_occlusion[i], pos).r;
+ occlusion |= uint(clamp(occ * 15.0, 0.0, 15.0)) << occlusion_shift[i];
+ }
+ {
+ ivec3 occ_pos = pos;
+ occ_pos.z += params.cascade * params.grid_size;
+ imageStore(dst_occlusion, occ_pos, uvec4(occlusion & 0xFFFF));
+ occ_pos.x += params.grid_size;
+ imageStore(dst_occlusion, occ_pos, uvec4(occlusion >> 16));
+ }
+
+ // STORE POSITIONS
+
+ if (local == ivec3(0)) {
+ store_position_count = 0; //base one stores as zero, the others wait
+ }
+
+ groupMemoryBarrier();
+ barrier();
+
+ if (solid) {
+ uint index = atomicAdd(store_position_count, 1);
+ // At least do the conversion work in parallel
+ store_positions[index].position = uint(pos.x | (pos.y << 7) | (pos.z << 14));
+
+ //see around which voxels point to this one, add them to the list
+ uint bit_index = 0;
+ uint neighbour_bits = 0;
+ for (int i = -1; i <= 1; i++) {
+ for (int j = -1; j <= 1; j++) {
+ for (int k = -1; k <= 1; k++) {
+ if (i == 0 && j == 0 && k == 0) {
+ continue;
+ }
+ ivec3 npos = pos + ivec3(i, j, k);
+ if (all(greaterThanEqual(npos, ivec3(0))) && all(lessThan(npos, ivec3(params.grid_size)))) {
+ p = imageLoad(src_positions, npos);
+ if (ivec3(p.xyz) == pos) {
+ neighbour_bits |= (1 << bit_index);
+ }
+ }
+ bit_index++;
+ }
+ }
+ }
+
+ uint rgb = imageLoad(src_albedo, pos).r;
+ uint facing = imageLoad(src_facing, pos).r;
+
+ store_positions[index].albedo = rgb >> 1; //store as it comes (555) to avoid precision loss (and move away the alpha bit)
+ store_positions[index].albedo |= (facing & 0x3F) << 15; // store facing in bits 15-21
+
+ store_positions[index].albedo |= neighbour_bits << 21; //store lower 11 bits of neighbours with remaining albedo
+ store_positions[index].position |= (neighbour_bits >> 11) << 21; //store 11 bits more of neighbours with position
+
+ store_positions[index].light = imageLoad(src_light, pos).r;
+ store_positions[index].light_aniso = imageLoad(src_light_aniso, pos).r;
+ //add neighbours
+ store_positions[index].light |= (neighbour_bits >> 22) << 30; //store 2 bits more of neighbours with light
+ store_positions[index].light_aniso |= (neighbour_bits >> 24) << 30; //store 2 bits more of neighbours with aniso
+ }
+
+ groupMemoryBarrier();
+ barrier();
+
+ // global increment only once per group, to reduce pressure
+
+ if (local == ivec3(0) && store_position_count > 0) {
+ store_from_index = atomicAdd(dispatch_data.total_count, store_position_count);
+ uint group_count = (store_from_index + store_position_count - 1) / 64 + 1;
+ atomicMax(dispatch_data.x, group_count);
+ }
+
+ groupMemoryBarrier();
+ barrier();
+
+ uint read_index = uint(local.z * 4 * 4 + local.y * 4 + local.x);
+ uint write_index = store_from_index + read_index;
+
+ if (read_index < store_position_count) {
+ dst_process_voxels.data[write_index] = store_positions[read_index];
+ }
+
+ if (pos == ivec3(0)) {
+ //this thread clears y and z
+ dispatch_data.y = 1;
+ dispatch_data.z = 1;
+ }
+#endif
+}