diff options
Diffstat (limited to 'servers/rendering/renderer_rd')
28 files changed, 2786 insertions, 2146 deletions
diff --git a/servers/rendering/renderer_rd/cluster_builder_rd.cpp b/servers/rendering/renderer_rd/cluster_builder_rd.cpp index 8d9cff0f43..0fdd864d47 100644 --- a/servers/rendering/renderer_rd/cluster_builder_rd.cpp +++ b/servers/rendering/renderer_rd/cluster_builder_rd.cpp @@ -400,12 +400,14 @@ void ClusterBuilderRD::begin(const Transform &p_view_transform, const CameraMatr void ClusterBuilderRD::bake_cluster() { RENDER_TIMESTAMP(">Bake Cluster"); + RD::get_singleton()->draw_command_begin_label("Bake Light Cluster"); + //clear cluster buffer - RD::get_singleton()->buffer_clear(cluster_buffer, 0, cluster_buffer_size, true); + RD::get_singleton()->buffer_clear(cluster_buffer, 0, cluster_buffer_size, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); if (render_element_count > 0) { //clear render buffer - RD::get_singleton()->buffer_clear(cluster_render_buffer, 0, cluster_render_buffer_size, true); + RD::get_singleton()->buffer_clear(cluster_render_buffer, 0, cluster_render_buffer_size, RD::BARRIER_MASK_RASTER); { //fill state uniform @@ -420,12 +422,12 @@ void ClusterBuilderRD::bake_cluster() { state.cluster_depth_offset = (render_element_max / 32); state.cluster_data_size = state.cluster_depth_offset + render_element_max; - RD::get_singleton()->buffer_update(state_uniform, 0, sizeof(StateUniform), &state, true); + RD::get_singleton()->buffer_update(state_uniform, 0, sizeof(StateUniform), &state, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); } //update instances - RD::get_singleton()->buffer_update(element_buffer, 0, sizeof(RenderElementData) * render_element_count, render_elements, true); + RD::get_singleton()->buffer_update(element_buffer, 0, sizeof(RenderElementData) * render_element_count, render_elements, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); RENDER_TIMESTAMP("Render Elements"); @@ -469,7 +471,7 @@ void ClusterBuilderRD::bake_cluster() { RD::get_singleton()->draw_list_draw(draw_list, true, instances); i += instances; } - RD::get_singleton()->draw_list_end(); + RD::get_singleton()->draw_list_end(RD::BARRIER_MASK_COMPUTE); } //store elements RENDER_TIMESTAMP("Pack Elements"); @@ -491,12 +493,15 @@ void ClusterBuilderRD::bake_cluster() { RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ClusterBuilderSharedDataRD::ClusterStore::PushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, cluster_screen_size.x, cluster_screen_size.y, 1, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, cluster_screen_size.x, cluster_screen_size.y, 1); - RD::get_singleton()->compute_list_end(); + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); } + } else { + RD::get_singleton()->barrier(RD::BARRIER_MASK_TRANSFER, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); } RENDER_TIMESTAMP("<Bake Cluster"); + RD::get_singleton()->draw_command_end_label(); } void ClusterBuilderRD::debug(ElementType p_element) { @@ -519,7 +524,7 @@ void ClusterBuilderRD::debug(ElementType p_element) { RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ClusterBuilderSharedDataRD::ClusterDebug::PushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, screen_size.x, screen_size.y, 1, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, screen_size.x, screen_size.y, 1); RD::get_singleton()->compute_list_end(); } diff --git a/servers/rendering/renderer_rd/effects_rd.cpp b/servers/rendering/renderer_rd/effects_rd.cpp index 6e1d61ff94..bc304aedd8 100644 --- a/servers/rendering/renderer_rd/effects_rd.cpp +++ b/servers/rendering/renderer_rd/effects_rd.cpp @@ -299,15 +299,12 @@ void EffectsRD::copy_to_rect(RID p_source_rd_texture, RID p_dest_texture, const copy.push_constant.target[0] = p_rect.position.x; copy.push_constant.target[1] = p_rect.position.y; - int32_t x_groups = (p_rect.size.width - 1) / 8 + 1; - int32_t y_groups = (p_rect.size.height - 1) / 8 + 1; - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[p_8_bit_dst ? COPY_MODE_SIMPLY_COPY_8BIT : COPY_MODE_SIMPLY_COPY]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_rd_texture), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_texture), 3); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_rect.size.width, p_rect.size.height, 1); RD::get_singleton()->compute_list_end(); } @@ -322,15 +319,12 @@ void EffectsRD::copy_cubemap_to_panorama(RID p_source_cube, RID p_dest_panorama, copy.push_constant.target[1] = 0; copy.push_constant.camera_z_far = p_lod; - int32_t x_groups = (p_panorama_size.width - 1) / 8 + 1; - int32_t y_groups = (p_panorama_size.height - 1) / 8 + 1; - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[p_is_array ? COPY_MODE_CUBE_ARRAY_TO_PANORAMA : COPY_MODE_CUBE_TO_PANORAMA]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_cube), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_panorama), 3); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_panorama_size.width, p_panorama_size.height, 1); RD::get_singleton()->compute_list_end(); } @@ -349,15 +343,12 @@ void EffectsRD::copy_depth_to_rect_and_linearize(RID p_source_rd_texture, RID p_ copy.push_constant.camera_z_far = p_z_far; copy.push_constant.camera_z_near = p_z_near; - int32_t x_groups = (p_rect.size.width - 1) / 8 + 1; - int32_t y_groups = (p_rect.size.height - 1) / 8 + 1; - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[COPY_MODE_LINEARIZE_DEPTH]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_rd_texture), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_texture), 3); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_rect.size.width, p_rect.size.height, 1); RD::get_singleton()->compute_list_end(); } @@ -374,15 +365,12 @@ void EffectsRD::copy_depth_to_rect(RID p_source_rd_texture, RID p_dest_texture, copy.push_constant.target[0] = p_rect.position.x; copy.push_constant.target[1] = p_rect.position.y; - int32_t x_groups = (p_rect.size.width - 1) / 8 + 1; - int32_t y_groups = (p_rect.size.height - 1) / 8 + 1; - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[COPY_MODE_SIMPLY_COPY_DEPTH]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_rd_texture), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_texture), 3); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_rect.size.width, p_rect.size.height, 1); RD::get_singleton()->compute_list_end(); } @@ -400,14 +388,11 @@ void EffectsRD::set_color(RID p_dest_texture, const Color &p_color, const Rect2i copy.push_constant.set_color[2] = p_color.b; copy.push_constant.set_color[3] = p_color.a; - int32_t x_groups = (p_region.size.width - 1) / 8 + 1; - int32_t y_groups = (p_region.size.height - 1) / 8 + 1; - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[p_8bit_dst ? COPY_MODE_SET_COLOR_8BIT : COPY_MODE_SET_COLOR]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_texture), 3); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_region.size.width, p_region.size.height, 1); RD::get_singleton()->compute_list_end(); } @@ -420,8 +405,6 @@ void EffectsRD::gaussian_blur(RID p_source_rd_texture, RID p_texture, RID p_back copy.push_constant.section[2] = p_region.size.width; copy.push_constant.section[3] = p_region.size.height; - int32_t x_groups = (p_region.size.width - 1) / 8 + 1; - int32_t y_groups = (p_region.size.height - 1) / 8 + 1; //HORIZONTAL RD::DrawListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[p_8bit_dst ? COPY_MODE_GAUSSIAN_COPY_8BIT : COPY_MODE_GAUSSIAN_COPY]); @@ -431,7 +414,7 @@ void EffectsRD::gaussian_blur(RID p_source_rd_texture, RID p_texture, RID p_back copy.push_constant.flags = base_flags | COPY_FLAG_HORIZONTAL; RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_region.size.width, p_region.size.height, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); @@ -442,7 +425,7 @@ void EffectsRD::gaussian_blur(RID p_source_rd_texture, RID p_texture, RID p_back copy.push_constant.flags = base_flags; RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_region.size.width, p_region.size.height, 1); RD::get_singleton()->compute_list_end(); } @@ -452,9 +435,6 @@ void EffectsRD::gaussian_glow(RID p_source_rd_texture, RID p_back_texture, const CopyMode copy_mode = p_first_pass && p_auto_exposure.is_valid() ? COPY_MODE_GAUSSIAN_GLOW_AUTO_EXPOSURE : COPY_MODE_GAUSSIAN_GLOW; uint32_t base_flags = 0; - int32_t x_groups = (p_size.width + 7) / 8; - int32_t y_groups = (p_size.height + 7) / 8; - copy.push_constant.section[2] = p_size.x; copy.push_constant.section[3] = p_size.y; @@ -479,16 +459,13 @@ void EffectsRD::gaussian_glow(RID p_source_rd_texture, RID p_back_texture, const copy.push_constant.flags = base_flags | (p_first_pass ? COPY_FLAG_GLOW_FIRST_PASS : 0) | (p_high_quality ? COPY_FLAG_HIGH_QUALITY_GLOW : 0); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_size.width, p_size.height, 1); RD::get_singleton()->compute_list_end(); } void EffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal_roughness, RenderingServer::EnvironmentSSRRoughnessQuality p_roughness_quality, RID p_blur_radius, RID p_blur_radius2, RID p_metallic, const Color &p_metallic_mask, RID p_depth, RID p_scale_depth, RID p_scale_normal, RID p_output, RID p_output_blur, const Size2i &p_screen_size, int p_max_steps, float p_fade_in, float p_fade_out, float p_tolerance, const CameraMatrix &p_camera) { RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - int32_t x_groups = (p_screen_size.width - 1) / 8 + 1; - int32_t y_groups = (p_screen_size.height - 1) / 8 + 1; - { //scale color and depth to half ssr_scale.push_constant.camera_z_far = p_camera.get_z_far(); ssr_scale.push_constant.camera_z_near = p_camera.get_z_near(); @@ -506,7 +483,7 @@ void EffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal_roughness, R RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssr_scale.push_constant, sizeof(ScreenSpaceReflectionScalePushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); } @@ -547,7 +524,7 @@ void EffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal_roughness, R } RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_scale_normal), 2); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); } if (p_roughness_quality != RS::ENV_SSR_ROUGNESS_QUALITY_DISABLED) { @@ -585,7 +562,7 @@ void EffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal_roughness, R RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssr_filter.push_constant, sizeof(ScreenSpaceReflectionFilterPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); @@ -600,7 +577,7 @@ void EffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal_roughness, R RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssr_filter.push_constant, sizeof(ScreenSpaceReflectionFilterPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); } RD::get_singleton()->compute_list_end(); @@ -609,9 +586,6 @@ void EffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal_roughness, R void EffectsRD::sub_surface_scattering(RID p_diffuse, RID p_diffuse2, RID p_depth, const CameraMatrix &p_camera, const Size2i &p_screen_size, float p_scale, float p_depth_scale, RenderingServer::SubSurfaceScatteringQuality p_quality) { RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - int32_t x_groups = (p_screen_size.width - 1) / 8 + 1; - int32_t y_groups = (p_screen_size.height - 1) / 8 + 1; - Plane p = p_camera.xform4(Plane(1, 0, -1, 1)); p.normal /= p.d; float unit_size = p.normal.x; @@ -635,7 +609,7 @@ void EffectsRD::sub_surface_scattering(RID p_diffuse, RID p_diffuse2, RID p_dept RD::get_singleton()->compute_list_set_push_constant(compute_list, &sss.push_constant, sizeof(SubSurfaceScatteringPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); @@ -646,7 +620,7 @@ void EffectsRD::sub_surface_scattering(RID p_diffuse, RID p_diffuse2, RID p_dept sss.push_constant.vertical = true; RD::get_singleton()->compute_list_set_push_constant(compute_list, &sss.push_constant, sizeof(SubSurfaceScatteringPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); RD::get_singleton()->compute_list_end(); } @@ -690,39 +664,33 @@ void EffectsRD::make_mipmap(RID p_source_rd_texture, RID p_dest_texture, const S copy.push_constant.section[2] = p_size.width; copy.push_constant.section[3] = p_size.height; - int32_t x_groups = (p_size.width - 1) / 8 + 1; - int32_t y_groups = (p_size.height - 1) / 8 + 1; - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, copy.pipelines[COPY_MODE_MIPMAP]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_rd_texture), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_texture), 3); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©.push_constant, sizeof(CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_size.width, p_size.height, 1); RD::get_singleton()->compute_list_end(); } -void EffectsRD::copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dest_texture, const Rect2i &p_rect, float p_z_near, float p_z_far, float p_bias, bool p_dp_flip) { +void EffectsRD::copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dst_framebuffer, const Rect2 &p_rect, float p_z_near, float p_z_far, bool p_dp_flip) { CopyToDPPushConstant push_constant; - push_constant.screen_size[0] = p_rect.size.x; - push_constant.screen_size[1] = p_rect.size.y; - push_constant.dest_offset[0] = p_rect.position.x; - push_constant.dest_offset[1] = p_rect.position.y; - push_constant.bias = p_bias; + push_constant.screen_rect[0] = p_rect.position.x; + push_constant.screen_rect[1] = p_rect.position.y; + push_constant.screen_rect[2] = p_rect.size.width; + push_constant.screen_rect[3] = p_rect.size.height; push_constant.z_far = p_z_far; push_constant.z_near = p_z_near; push_constant.z_flip = p_dp_flip; - int32_t x_groups = (p_rect.size.width - 1) / 8 + 1; - int32_t y_groups = (p_rect.size.height - 1) / 8 + 1; + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dst_framebuffer, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, cube_to_dp.pipeline.get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dst_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_source_rd_texture), 0); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, cube_to_dp.pipeline); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_rd_texture), 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_texture), 1); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(CopyToDPPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); - RD::get_singleton()->compute_list_end(); + RD::get_singleton()->draw_list_set_push_constant(draw_list, &push_constant, sizeof(CopyToDPPushConstant)); + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_TRANSFER); } void EffectsRD::tonemapper(RID p_source_color, RID p_dst_framebuffer, const TonemapSettings &p_settings) { @@ -807,10 +775,7 @@ void EffectsRD::luminance_reduction(RID p_source_texture, const Size2i p_source_ RD::get_singleton()->compute_list_set_push_constant(compute_list, &luminance_reduce.push_constant, sizeof(LuminanceReducePushConstant)); - int32_t x_groups = (luminance_reduce.push_constant.source_size[0] - 1) / 8 + 1; - int32_t y_groups = (luminance_reduce.push_constant.source_size[1] - 1) / 8 + 1; - - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, luminance_reduce.push_constant.source_size[0], luminance_reduce.push_constant.source_size[1], 1); luminance_reduce.push_constant.source_size[0] = MAX(luminance_reduce.push_constant.source_size[0] / 8, 1); luminance_reduce.push_constant.source_size[1] = MAX(luminance_reduce.push_constant.source_size[1] / 8, 1); @@ -851,14 +816,12 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_base_texture), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_depth_texture), 1); - int32_t x_groups = (p_base_texture_size.x - 1) / 8 + 1; - int32_t y_groups = (p_base_texture_size.y - 1) / 8 + 1; bokeh.push_constant.size[0] = p_base_texture_size.x; bokeh.push_constant.size[1] = p_base_texture_size.y; RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_base_texture_size.x, p_base_texture_size.y, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); if (p_bokeh_shape == RS::DOF_BOKEH_BOX || p_bokeh_shape == RS::DOF_BOKEH_HEXAGON) { @@ -875,8 +838,6 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_halfsize_texture1), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_base_texture), 1); - x_groups = ((p_base_texture_size.x >> 1) - 1) / 8 + 1; - y_groups = ((p_base_texture_size.y >> 1) - 1) / 8 + 1; bokeh.push_constant.size[0] = p_base_texture_size.x >> 1; bokeh.push_constant.size[1] = p_base_texture_size.y >> 1; bokeh.push_constant.half_size = true; @@ -890,7 +851,7 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, bokeh.push_constant.size[0], bokeh.push_constant.size[1], 1); RD::get_singleton()->compute_list_add_barrier(compute_list); //third pass @@ -906,7 +867,7 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, bokeh.push_constant.size[0], bokeh.push_constant.size[1], 1); RD::get_singleton()->compute_list_add_barrier(compute_list); if (p_quality == RS::DOF_BLUR_QUALITY_VERY_LOW || p_quality == RS::DOF_BLUR_QUALITY_LOW) { @@ -917,8 +878,6 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_base_texture), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_halfsize_texture2), 1); - x_groups = (p_base_texture_size.x - 1) / 8 + 1; - y_groups = (p_base_texture_size.y - 1) / 8 + 1; bokeh.push_constant.size[0] = p_base_texture_size.x; bokeh.push_constant.size[1] = p_base_texture_size.y; bokeh.push_constant.half_size = false; @@ -926,7 +885,7 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_base_texture_size.x, p_base_texture_size.y, 1); } } else { //circle @@ -944,15 +903,13 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_halfsize_texture1), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_base_texture), 1); - x_groups = ((p_base_texture_size.x >> 1) - 1) / 8 + 1; - y_groups = ((p_base_texture_size.y >> 1) - 1) / 8 + 1; bokeh.push_constant.size[0] = p_base_texture_size.x >> 1; bokeh.push_constant.size[1] = p_base_texture_size.y >> 1; bokeh.push_constant.half_size = true; RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, bokeh.push_constant.size[0], bokeh.push_constant.size[1], 1); RD::get_singleton()->compute_list_add_barrier(compute_list); //circle is just one pass, then upscale @@ -964,8 +921,6 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_base_texture), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_halfsize_texture1), 1); - x_groups = (p_base_texture_size.x - 1) / 8 + 1; - y_groups = (p_base_texture_size.y - 1) / 8 + 1; bokeh.push_constant.size[0] = p_base_texture_size.x; bokeh.push_constant.size[1] = p_base_texture_size.y; bokeh.push_constant.half_size = false; @@ -973,7 +928,7 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_base_texture_size.x, p_base_texture_size.y, 1); } RD::get_singleton()->compute_list_end(); @@ -998,20 +953,20 @@ void EffectsRD::gather_ssao(RD::ComputeListID p_compute_list, const Vector<RID> RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, _get_uniform_set_from_image(p_ao_slices[i]), 2); RD::get_singleton()->compute_list_set_push_constant(p_compute_list, &ssao.gather_push_constant, sizeof(SSAOGatherPushConstant)); - int x_groups = ((p_settings.full_screen_size.x >> (p_settings.half_size ? 2 : 1)) - 1) / 8 + 1; - int y_groups = ((p_settings.full_screen_size.y >> (p_settings.half_size ? 2 : 1)) - 1) / 8 + 1; + Size2i size = Size2i(p_settings.full_screen_size.x >> (p_settings.half_size ? 2 : 1), p_settings.full_screen_size.y >> (p_settings.half_size ? 2 : 1)); - RD::get_singleton()->compute_list_dispatch(p_compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(p_compute_list, size.x, size.y, 1); } RD::get_singleton()->compute_list_add_barrier(p_compute_list); } void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_depth_mipmaps_texture, const Vector<RID> &p_depth_mipmaps, RID p_ao, const Vector<RID> p_ao_slices, RID p_ao_pong, const Vector<RID> p_ao_pong_slices, RID p_upscale_buffer, RID p_importance_map, RID p_importance_map_pong, const CameraMatrix &p_projection, const SSAOSettings &p_settings, bool p_invalidate_uniform_sets) { RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - + RD::get_singleton()->draw_command_begin_label("SSAO"); /* FIRST PASS */ // Downsample and deinterleave the depth buffer. { + RD::get_singleton()->draw_command_begin_label("Downsample Depth"); if (p_invalidate_uniform_sets) { Vector<RD::Uniform> uniforms; { @@ -1074,16 +1029,17 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep } RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.downsample_push_constant, sizeof(SSAODownsamplePushConstant)); - int x_groups = (MAX(1, p_settings.full_screen_size.x >> (p_settings.half_size ? 2 : 1)) - 1) / 8 + 1; - int y_groups = (MAX(1, p_settings.full_screen_size.y >> (p_settings.half_size ? 2 : 1)) - 1) / 8 + 1; + Size2i size(MAX(1, p_settings.full_screen_size.x >> (p_settings.half_size ? 2 : 1)), MAX(1, p_settings.full_screen_size.y >> (p_settings.half_size ? 2 : 1))); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, size.x, size.y, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); + RD::get_singleton()->draw_command_end_label(); // Downsample SSAO } /* SECOND PASS */ // Sample SSAO { + RD::get_singleton()->draw_command_begin_label("Gather Samples"); ssao.gather_push_constant.screen_size[0] = p_settings.full_screen_size.x; ssao.gather_push_constant.screen_size[1] = p_settings.full_screen_size.y; @@ -1184,6 +1140,7 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep } if (p_settings.quality == RS::ENV_SSAO_QUALITY_ULTRA) { + RD::get_singleton()->draw_command_begin_label("Generate Importance Map"); ssao.importance_map_push_constant.half_screen_pixel_size[0] = 1.0 / p_settings.half_screen_size.x; ssao.importance_map_push_constant.half_screen_pixel_size[1] = 1.0 / p_settings.half_screen_size.y; ssao.importance_map_push_constant.intensity = p_settings.intensity; @@ -1192,21 +1149,19 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GATHER_BASE]); gather_ssao(compute_list, p_ao_pong_slices, p_settings, true); //generate importance map - int x_groups = (p_settings.quarter_screen_size.x - 1) / 8 + 1; - int y_groups = (p_settings.quarter_screen_size.y - 1) / 8 + 1; RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GENERATE_IMPORTANCE_MAP]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_ao_pong), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_importance_map), 1); RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.importance_map_push_constant, sizeof(SSAOImportanceMapPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_settings.quarter_screen_size.x, p_settings.quarter_screen_size.y, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); //process importance map A RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_PROCESS_IMPORTANCE_MAPA]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_importance_map), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_importance_map_pong), 1); RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.importance_map_push_constant, sizeof(SSAOImportanceMapPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_settings.quarter_screen_size.x, p_settings.quarter_screen_size.y, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); //process Importance Map B RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_PROCESS_IMPORTANCE_MAPB]); @@ -1214,21 +1169,24 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_importance_map), 1); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, ssao.counter_uniform_set, 2); RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.importance_map_push_constant, sizeof(SSAOImportanceMapPushConstant)); - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_settings.quarter_screen_size.x, p_settings.quarter_screen_size.y, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GATHER_ADAPTIVE]); + RD::get_singleton()->draw_command_end_label(); // Importance Map } else { RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GATHER]); } gather_ssao(compute_list, p_ao_slices, p_settings, false); + RD::get_singleton()->draw_command_end_label(); // Gather SSAO } // /* THIRD PASS */ // // Blur // { + RD::get_singleton()->draw_command_begin_label("Edge Aware Blur"); ssao.blur_push_constant.edge_sharpness = 1.0 - p_settings.sharpness; ssao.blur_push_constant.half_screen_pixel_size[0] = 1.0 / p_settings.half_screen_size.x; ssao.blur_push_constant.half_screen_pixel_size[1] = 1.0 / p_settings.half_screen_size.y; @@ -1268,22 +1226,22 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep } RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.blur_push_constant, sizeof(SSAOBlurPushConstant)); - int x_groups = ((p_settings.full_screen_size.x >> (p_settings.half_size ? 2 : 1)) - 1) / 8 + 1; - int y_groups = ((p_settings.full_screen_size.y >> (p_settings.half_size ? 2 : 1)) - 1) / 8 + 1; - - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + Size2i size(p_settings.full_screen_size.x >> (p_settings.half_size ? 2 : 1), p_settings.full_screen_size.y >> (p_settings.half_size ? 2 : 1)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, size.x, size.y, 1); } if (p_settings.quality > RS::ENV_SSAO_QUALITY_VERY_LOW) { RD::get_singleton()->compute_list_add_barrier(compute_list); } } + RD::get_singleton()->draw_command_end_label(); // Blur } /* FOURTH PASS */ // Interleave buffers // back to full size { + RD::get_singleton()->draw_command_begin_label("Interleave Buffers"); ssao.interleave_push_constant.inv_sharpness = 1.0 - p_settings.sharpness; ssao.interleave_push_constant.pixel_size[0] = 1.0 / p_settings.full_screen_size.x; ssao.interleave_push_constant.pixel_size[1] = 1.0 / p_settings.full_screen_size.y; @@ -1307,17 +1265,15 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.interleave_push_constant, sizeof(SSAOInterleavePushConstant)); - int x_groups = (p_settings.full_screen_size.x - 1) / 8 + 1; - int y_groups = (p_settings.full_screen_size.y - 1) / 8 + 1; - - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_settings.full_screen_size.x, p_settings.full_screen_size.y, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); + RD::get_singleton()->draw_command_end_label(); // Interleave } - - RD::get_singleton()->compute_list_end(); + RD::get_singleton()->draw_command_end_label(); //SSAO + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_TRANSFER); //wait for upcoming transfer int zero[1] = { 0 }; - RD::get_singleton()->buffer_update(ssao.importance_map_load_counter, 0, sizeof(uint32_t), &zero, false); + RD::get_singleton()->buffer_update(ssao.importance_map_load_counter, 0, sizeof(uint32_t), &zero, 0); //no barrier } void EffectsRD::roughness_limit(RID p_source_normal, RID p_roughness, const Size2i &p_size, float p_curve) { @@ -1330,12 +1286,9 @@ void EffectsRD::roughness_limit(RID p_source_normal, RID p_roughness, const Size RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_normal), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_roughness), 1); - int x_groups = (p_size.x - 1) / 8 + 1; - int y_groups = (p_size.y - 1) / 8 + 1; - RD::get_singleton()->compute_list_set_push_constant(compute_list, &roughness_limiter.push_constant, sizeof(RoughnessLimiterPushConstant)); //not used but set anyway - RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_size.x, p_size.y, 1); RD::get_singleton()->compute_list_end(); } @@ -1448,7 +1401,7 @@ void EffectsRD::render_sky(RD::DrawListID p_list, float p_time, RID p_fb, RID p_ RD::get_singleton()->draw_list_draw(draw_list, true); } -void EffectsRD::resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RID p_source_giprobe, RID p_dest_depth, RID p_dest_normal_roughness, RID p_dest_giprobe, Vector2i p_screen_size, int p_samples) { +void EffectsRD::resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RID p_source_giprobe, RID p_dest_depth, RID p_dest_normal_roughness, RID p_dest_giprobe, Vector2i p_screen_size, int p_samples, uint32_t p_barrier) { ResolvePushConstant push_constant; push_constant.screen_size[0] = p_screen_size.x; push_constant.screen_size[1] = p_screen_size.y; @@ -1465,54 +1418,9 @@ void EffectsRD::resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RI RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ResolvePushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.x, p_screen_size.y, 1, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.x, p_screen_size.y, 1); - RD::get_singleton()->compute_list_end(); -} - -void EffectsRD::reduce_shadow(RID p_source_shadow, RID p_dest_shadow, const Size2i &p_source_size, const Rect2i &p_source_rect, int p_shrink_limit, RD::ComputeListID compute_list) { - uint32_t push_constant[8] = { (uint32_t)p_source_size.x, (uint32_t)p_source_size.y, (uint32_t)p_source_rect.position.x, (uint32_t)p_source_rect.position.y, (uint32_t)p_shrink_limit, 0, 0, 0 }; - - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, shadow_reduce.pipelines[SHADOW_REDUCE_REDUCE]); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_image_pair(p_source_shadow, p_dest_shadow), 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(uint32_t) * 8); - - RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_source_rect.size.width, p_source_rect.size.height, 1, 8, 8, 1); -} -void EffectsRD::filter_shadow(RID p_shadow, RID p_backing_shadow, const Size2i &p_source_size, const Rect2i &p_source_rect, RenderingServer::EnvVolumetricFogShadowFilter p_filter, RD::ComputeListID compute_list, bool p_vertical, bool p_horizontal) { - uint32_t push_constant[8] = { (uint32_t)p_source_size.x, (uint32_t)p_source_size.y, (uint32_t)p_source_rect.position.x, (uint32_t)p_source_rect.position.y, 0, 0, 0, 0 }; - - switch (p_filter) { - case RS::ENV_VOLUMETRIC_FOG_SHADOW_FILTER_DISABLED: - case RS::ENV_VOLUMETRIC_FOG_SHADOW_FILTER_LOW: { - push_constant[5] = 0; - } break; - case RS::ENV_VOLUMETRIC_FOG_SHADOW_FILTER_MEDIUM: { - push_constant[5] = 9; - } break; - case RS::ENV_VOLUMETRIC_FOG_SHADOW_FILTER_HIGH: { - push_constant[5] = 18; - } break; - } - - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, shadow_reduce.pipelines[SHADOW_REDUCE_FILTER]); - if (p_vertical) { - push_constant[6] = 1; - push_constant[7] = 0; - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_image_pair(p_shadow, p_backing_shadow), 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(uint32_t) * 8); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_source_rect.size.width, p_source_rect.size.height, 1, 8, 8, 1); - } - if (p_vertical && p_horizontal) { - RD::get_singleton()->compute_list_add_barrier(compute_list); - } - if (p_horizontal) { - push_constant[6] = 0; - push_constant[7] = 1; - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_image_pair(p_backing_shadow, p_shadow), 0); - RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(uint32_t) * 8); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_source_rect.size.width, p_source_rect.size.height, 1, 8, 8, 1); - } + RD::get_singleton()->compute_list_end(p_barrier); } void EffectsRD::sort_buffer(RID p_uniform_set, int p_size) { @@ -1678,8 +1586,12 @@ EffectsRD::EffectsRD() { cube_to_dp.shader.initialize(copy_modes); cube_to_dp.shader_version = cube_to_dp.shader.version_create(); - - cube_to_dp.pipeline = RD::get_singleton()->compute_pipeline_create(cube_to_dp.shader.version_get_shader(cube_to_dp.shader_version, 0)); + RID shader = cube_to_dp.shader.version_get_shader(cube_to_dp.shader_version, 0); + RD::PipelineDepthStencilState dss; + dss.enable_depth_test = true; + dss.depth_compare_operator = RD::COMPARE_OP_ALWAYS; + dss.enable_depth_write = true; + cube_to_dp.pipeline.setup(shader, RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), dss, RD::PipelineColorBlendState(), 0); } { @@ -1776,7 +1688,7 @@ EffectsRD::EffectsRD() { } } - RD::get_singleton()->buffer_update(ssao.gather_constants_buffer, 0, sizeof(SSAOGatherConstants), &gather_constants, false); + RD::get_singleton()->buffer_update(ssao.gather_constants_buffer, 0, sizeof(SSAOGatherConstants), &gather_constants); } { Vector<String> ssao_modes; @@ -1795,7 +1707,8 @@ EffectsRD::EffectsRD() { } ssao.importance_map_load_counter = RD::get_singleton()->storage_buffer_create(sizeof(uint32_t)); int zero[1] = { 0 }; - RD::get_singleton()->buffer_update(ssao.importance_map_load_counter, 0, sizeof(uint32_t), &zero, false); + RD::get_singleton()->buffer_update(ssao.importance_map_load_counter, 0, sizeof(uint32_t), &zero); + RD::get_singleton()->set_resource_name(ssao.importance_map_load_counter, "Importance Map Load Counter"); Vector<RD::Uniform> uniforms; { @@ -1806,6 +1719,7 @@ EffectsRD::EffectsRD() { uniforms.push_back(u); } ssao.counter_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssao.importance_map_shader.version_get_shader(ssao.importance_map_shader_version, 2), 2); + RD::get_singleton()->set_resource_name(ssao.counter_uniform_set, "Load Counter Uniform Set"); } { Vector<String> ssao_modes; @@ -1834,7 +1748,7 @@ EffectsRD::EffectsRD() { ssao.interleave_shader_version = ssao.interleave_shader.version_create(); for (int i = SSAO_INTERLEAVE; i <= SSAO_INTERLEAVE_HALF; i++) { ssao.pipelines[pipeline] = RD::get_singleton()->compute_pipeline_create(ssao.interleave_shader.version_get_shader(ssao.interleave_shader_version, i - SSAO_INTERLEAVE)); - + RD::get_singleton()->set_resource_name(ssao.pipelines[pipeline], "Interleave Pipeline " + itos(i)); pipeline++; } } @@ -1867,7 +1781,7 @@ EffectsRD::EffectsRD() { { // Initialize cubemap filter - filter.use_high_quality = GLOBAL_GET("rendering/quality/reflections/fast_filter_high_quality"); + filter.use_high_quality = GLOBAL_GET("rendering/reflections/sky_reflections/fast_filter_high_quality"); Vector<String> cubemap_filter_modes; cubemap_filter_modes.push_back("\n#define USE_HIGH_QUALITY\n"); @@ -1883,10 +1797,10 @@ EffectsRD::EffectsRD() { if (filter.use_high_quality) { filter.coefficient_buffer = RD::get_singleton()->storage_buffer_create(sizeof(high_quality_coeffs)); - RD::get_singleton()->buffer_update(filter.coefficient_buffer, 0, sizeof(high_quality_coeffs), &high_quality_coeffs[0], false); + RD::get_singleton()->buffer_update(filter.coefficient_buffer, 0, sizeof(high_quality_coeffs), &high_quality_coeffs[0]); } else { filter.coefficient_buffer = RD::get_singleton()->storage_buffer_create(sizeof(low_quality_coeffs)); - RD::get_singleton()->buffer_update(filter.coefficient_buffer, 0, sizeof(low_quality_coeffs), &low_quality_coeffs[0], false); + RD::get_singleton()->buffer_update(filter.coefficient_buffer, 0, sizeof(low_quality_coeffs), &low_quality_coeffs[0]); } Vector<RD::Uniform> uniforms; @@ -2005,20 +1919,6 @@ EffectsRD::EffectsRD() { } { - Vector<String> shadow_reduce_modes; - shadow_reduce_modes.push_back("\n#define MODE_REDUCE\n"); - shadow_reduce_modes.push_back("\n#define MODE_FILTER\n"); - - shadow_reduce.shader.initialize(shadow_reduce_modes); - - shadow_reduce.shader_version = shadow_reduce.shader.version_create(); - - for (int i = 0; i < SHADOW_REDUCE_MAX; i++) { - shadow_reduce.pipelines[i] = RD::get_singleton()->compute_pipeline_create(shadow_reduce.shader.version_get_shader(shadow_reduce.shader_version, i)); - } - } - - { Vector<String> sort_modes; sort_modes.push_back("\n#define MODE_SORT_BLOCK\n"); sort_modes.push_back("\n#define MODE_SORT_STEP\n"); @@ -2039,12 +1939,14 @@ EffectsRD::EffectsRD() { sampler.max_lod = 0; default_sampler = RD::get_singleton()->sampler_create(sampler); + RD::get_singleton()->set_resource_name(default_sampler, "Default Linear Sampler"); sampler.min_filter = RD::SAMPLER_FILTER_LINEAR; sampler.mip_filter = RD::SAMPLER_FILTER_LINEAR; sampler.max_lod = 1e20; default_mipmap_sampler = RD::get_singleton()->sampler_create(sampler); + RD::get_singleton()->set_resource_name(default_mipmap_sampler, "Default MipMap Sampler"); { //create index array for copy shaders Vector<uint8_t> pv; @@ -2104,5 +2006,4 @@ EffectsRD::~EffectsRD() { ssr_scale.shader.version_free(ssr_scale.shader_version); sss.shader.version_free(sss.shader_version); tonemap.shader.version_free(tonemap.shader_version); - shadow_reduce.shader.version_free(shadow_reduce.shader_version); } diff --git a/servers/rendering/renderer_rd/effects_rd.h b/servers/rendering/renderer_rd/effects_rd.h index e2cdd0c3d8..1ba25e301b 100644 --- a/servers/rendering/renderer_rd/effects_rd.h +++ b/servers/rendering/renderer_rd/effects_rd.h @@ -46,7 +46,6 @@ #include "servers/rendering/renderer_rd/shaders/screen_space_reflection.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/screen_space_reflection_filter.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/screen_space_reflection_scale.glsl.gen.h" -#include "servers/rendering/renderer_rd/shaders/shadow_reduce.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/sort.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/specular_merge.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/ssao.glsl.gen.h" @@ -234,18 +233,17 @@ class EffectsRD { } luminance_reduce; struct CopyToDPPushConstant { - int32_t screen_size[2]; - int32_t dest_offset[2]; - float bias; float z_far; float z_near; uint32_t z_flip; + uint32_t pad; + float screen_rect[4]; }; struct CoptToDP { CubeToDpShaderRD shader; RID shader_version; - RID pipeline; + PipelineCacheRD pipeline; } cube_to_dp; struct BokehPushConstant { @@ -598,18 +596,6 @@ class EffectsRD { RID pipelines[RESOLVE_MODE_MAX]; //3 quality levels } resolve; - enum ShadowReduceMode { - SHADOW_REDUCE_REDUCE, - SHADOW_REDUCE_FILTER, - SHADOW_REDUCE_MAX - }; - - struct ShadowReduce { - ShadowReduceShaderRD shader; - RID shader_version; - RID pipelines[SHADOW_REDUCE_MAX]; - } shadow_reduce; - enum SortMode { SORT_MODE_BLOCK, SORT_MODE_STEP, @@ -687,7 +673,7 @@ public: void cubemap_roughness(RID p_source_rd_texture, RID p_dest_framebuffer, uint32_t p_face_id, uint32_t p_sample_count, float p_roughness, float p_size); void make_mipmap(RID p_source_rd_texture, RID p_dest_texture, const Size2i &p_size); - void copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dest_texture, const Rect2i &p_rect, float p_z_near, float p_z_far, float p_bias, bool p_dp_flip); + void copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dest_texture, const Rect2 &p_rect, float p_z_near, float p_z_far, bool p_dp_flip); void luminance_reduction(RID p_source_texture, const Size2i p_source_size, const Vector<RID> p_reduce, RID p_prev_luminance, float p_min_luminance, float p_max_luminance, float p_adjust, bool p_set = false); void bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i &p_base_texture_size, RID p_secondary_texture, RID p_bokeh_texture1, RID p_bokeh_texture2, bool p_dof_far, float p_dof_far_begin, float p_dof_far_size, bool p_dof_near, float p_dof_near_begin, float p_dof_near_size, float p_bokeh_size, RS::DOFBokehShape p_bokeh_shape, RS::DOFBlurQuality p_quality, bool p_use_jitter, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal); @@ -764,10 +750,7 @@ public: void merge_specular(RID p_dest_framebuffer, RID p_specular, RID p_base, RID p_reflection); void sub_surface_scattering(RID p_diffuse, RID p_diffuse2, RID p_depth, const CameraMatrix &p_camera, const Size2i &p_screen_size, float p_scale, float p_depth_scale, RS::SubSurfaceScatteringQuality p_quality); - void resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RID p_source_giprobe, RID p_dest_depth, RID p_dest_normal_roughness, RID p_dest_giprobe, Vector2i p_screen_size, int p_samples); - - void reduce_shadow(RID p_source_shadow, RID p_dest_shadow, const Size2i &p_source_size, const Rect2i &p_source_rect, int p_shrink_limit, RenderingDevice::ComputeListID compute_list); - void filter_shadow(RID p_shadow, RID p_backing_shadow, const Size2i &p_source_size, const Rect2i &p_source_rect, RS::EnvVolumetricFogShadowFilter p_filter, RenderingDevice::ComputeListID compute_list, bool p_vertical = true, bool p_horizontal = true); + void resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RID p_source_giprobe, RID p_dest_depth, RID p_dest_normal_roughness, RID p_dest_giprobe, Vector2i p_screen_size, int p_samples, uint32_t p_barrier = RD::BARRIER_MASK_ALL); void sort_buffer(RID p_uniform_set, int p_size); diff --git a/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp b/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp index 792fcb0b59..7d6e2fa8e4 100644 --- a/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp @@ -1367,7 +1367,7 @@ void RendererCanvasRenderRD::canvas_render_items(RID p_to_render_target, Item *p } if (light_count > 0) { - RD::get_singleton()->buffer_update(state.lights_uniform_buffer, 0, sizeof(LightUniform) * light_count, &state.light_uniforms[0], true); + RD::get_singleton()->buffer_update(state.lights_uniform_buffer, 0, sizeof(LightUniform) * light_count, &state.light_uniforms[0]); } { @@ -1421,7 +1421,7 @@ void RendererCanvasRenderRD::canvas_render_items(RID p_to_render_target, Item *p //print_line("w: " + itos(ssize.width) + " s: " + rtos(canvas_scale)); state_buffer.tex_to_sdf = 1.0 / ((canvas_scale.x + canvas_scale.y) * 0.5); - RD::get_singleton()->buffer_update(state.canvas_state_buffer, 0, sizeof(State::Buffer), &state_buffer, true); + RD::get_singleton()->buffer_update(state.canvas_state_buffer, 0, sizeof(State::Buffer), &state_buffer); } { //default filter/repeat @@ -1622,7 +1622,7 @@ void RendererCanvasRenderRD::light_update_shadow(RID p_rid, int p_shadow_index, projection.set_frustum(xmin, xmax, ymin, ymax, nearp, farp); } - Vector3 cam_target = Basis(Vector3(0, 0, Math_PI * 2 * ((i + 3) / 4.0))).xform(Vector3(0, 1, 0)); + Vector3 cam_target = Basis(Vector3(0, 0, Math_TAU * ((i + 3) / 4.0))).xform(Vector3(0, 1, 0)); projection = projection * CameraMatrix(Transform().looking_at(cam_target, Vector3(0, 0, -1)).affine_inverse()); ShadowRenderPushConstant push_constant; @@ -2695,9 +2695,10 @@ RendererCanvasRenderRD::RendererCanvasRenderRD(RendererStorageRD *p_storage) { state.default_transforms_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, shader.default_version_rd_shader, TRANSFORMS_UNIFORM_SET); } - default_canvas_texture = storage->canvas_texture_create(); + default_canvas_texture = storage->canvas_texture_allocate(); + storage->canvas_texture_initialize(default_canvas_texture); - state.shadow_texture_size = GLOBAL_GET("rendering/quality/2d_shadow_atlas/size"); + state.shadow_texture_size = GLOBAL_GET("rendering/2d/shadow_atlas/size"); //create functions for shader and material storage->shader_set_data_request_function(RendererStorageRD::SHADER_TYPE_2D, _create_shader_funcs); @@ -2706,9 +2707,14 @@ RendererCanvasRenderRD::RendererCanvasRenderRD(RendererStorageRD *p_storage) { state.time = 0; { - default_canvas_group_shader = storage->shader_create(); + default_canvas_group_shader = storage->shader_allocate(); + storage->shader_initialize(default_canvas_group_shader); + storage->shader_set_code(default_canvas_group_shader, "shader_type canvas_item; \nvoid fragment() {\n\tvec4 c = textureLod(SCREEN_TEXTURE,SCREEN_UV,0.0); if (c.a > 0.0001) c.rgb/=c.a; COLOR *= c; \n}\n"); - default_canvas_group_material = storage->material_create(); + + default_canvas_group_material = storage->material_allocate(); + storage->material_initialize(default_canvas_group_material); + storage->material_set_shader(default_canvas_group_material, default_canvas_group_shader); } diff --git a/servers/rendering/renderer_rd/renderer_scene_render_forward.cpp b/servers/rendering/renderer_rd/renderer_scene_render_forward.cpp index f3b09399f9..a57dee7314 100644 --- a/servers/rendering/renderer_rd/renderer_scene_render_forward.cpp +++ b/servers/rendering/renderer_rd/renderer_scene_render_forward.cpp @@ -453,7 +453,7 @@ void RendererSceneRenderForward::MaterialData::update_parameters(const Map<Strin //check whether buffer changed if (p_uniform_dirty && ubo_data.size()) { update_uniform_buffer(shader_data->uniforms, shader_data->ubo_offsets.ptr(), p_parameters, ubo_data.ptrw(), ubo_data.size(), false); - RD::get_singleton()->buffer_update(uniform_buffer, 0, ubo_data.size(), ubo_data.ptrw()); + RD::get_singleton()->buffer_update(uniform_buffer, 0, ubo_data.size(), ubo_data.ptrw(), RD::BARRIER_MASK_RASTER); } uint32_t tex_uniform_count = shader_data->texture_uniforms.size(); @@ -583,19 +583,6 @@ void RendererSceneRenderForward::RenderBufferDataForward::ensure_specular() { } } -void RendererSceneRenderForward::RenderBufferDataForward::ensure_gi() { - if (!reflection_buffer.is_valid()) { - RD::TextureFormat tf; - tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; - tf.width = width; - tf.height = height; - tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; - - reflection_buffer = RD::get_singleton()->texture_create(tf, RD::TextureView()); - ambient_buffer = RD::get_singleton()->texture_create(tf, RD::TextureView()); - } -} - void RendererSceneRenderForward::RenderBufferDataForward::ensure_giprobe() { if (!giprobe_buffer.is_valid()) { RD::TextureFormat tf; @@ -633,16 +620,6 @@ void RendererSceneRenderForward::RenderBufferDataForward::ensure_giprobe() { } void RendererSceneRenderForward::RenderBufferDataForward::clear() { - if (ambient_buffer != RID() && ambient_buffer != color) { - RD::get_singleton()->free(ambient_buffer); - ambient_buffer = RID(); - } - - if (reflection_buffer != RID() && reflection_buffer != specular) { - RD::get_singleton()->free(reflection_buffer); - reflection_buffer = RID(); - } - if (giprobe_buffer != RID()) { RD::get_singleton()->free(giprobe_buffer); giprobe_buffer = RID(); @@ -833,16 +810,26 @@ void RendererSceneRenderForward::_render_list_template(RenderingDevice::DrawList bool shadow_pass = (p_params->pass_mode == PASS_MODE_SHADOW) || (p_params->pass_mode == PASS_MODE_SHADOW_DP); - float old_offset[2] = { 0, 0 }; + SceneState::PushConstant push_constant; + + if (p_params->pass_mode == PASS_MODE_DEPTH_MATERIAL) { + push_constant.uv_offset = Math::make_half_float(p_params->uv_offset.y) << 16; + push_constant.uv_offset |= Math::make_half_float(p_params->uv_offset.x); + } else { + push_constant.uv_offset = 0; + } for (uint32_t i = p_from_element; i < p_to_element; i++) { const GeometryInstanceSurfaceDataCache *surf = p_params->elements[i]; + const RenderElementInfo &element_info = p_params->element_info[i]; + + push_constant.base_index = i + p_params->element_offset; RID material_uniform_set; ShaderData *shader; void *mesh_surface; - if (shadow_pass) { + if (shadow_pass || p_params->pass_mode == PASS_MODE_DEPTH) { //regular depth pass can use these too material_uniform_set = surf->material_uniform_set_shadow; shader = surf->shader_shadow; mesh_surface = surf->surface_shadow; @@ -857,13 +844,6 @@ void RendererSceneRenderForward::_render_list_template(RenderingDevice::DrawList continue; } - if (p_params->pass_mode == PASS_MODE_DEPTH_MATERIAL) { - old_offset[0] = surf->owner->push_constant.lightmap_uv_scale[0]; - old_offset[1] = surf->owner->push_constant.lightmap_uv_scale[1]; - surf->owner->push_constant.lightmap_uv_scale[0] = p_params->uv_offset.x; - surf->owner->push_constant.lightmap_uv_scale[1] = p_params->uv_offset.y; - } - //find cull variant ShaderData::CullVariant cull_variant; @@ -885,16 +865,16 @@ void RendererSceneRenderForward::_render_list_template(RenderingDevice::DrawList switch (p_params->pass_mode) { case PASS_MODE_COLOR: case PASS_MODE_COLOR_TRANSPARENT: { - if (surf->sort.uses_lightmap) { + if (element_info.uses_lightmap) { shader_version = SHADER_VERSION_LIGHTMAP_COLOR_PASS; - } else if (surf->sort.uses_forward_gi) { + } else if (element_info.uses_forward_gi) { shader_version = SHADER_VERSION_COLOR_PASS_WITH_FORWARD_GI; } else { shader_version = SHADER_VERSION_COLOR_PASS; } } break; case PASS_MODE_COLOR_SPECULAR: { - if (surf->sort.uses_lightmap) { + if (element_info.uses_lightmap) { shader_version = SHADER_VERSION_LIGHTMAP_COLOR_PASS_WITH_SEPARATE_SPECULAR; } else { shader_version = SHADER_VERSION_COLOR_PASS_WITH_SEPARATE_SPECULAR; @@ -936,31 +916,7 @@ void RendererSceneRenderForward::_render_list_template(RenderingDevice::DrawList storage->mesh_surface_get_vertex_arrays_and_format(mesh_surface, pipeline->get_vertex_input_mask(), vertex_array_rd, vertex_format); } - if (p_params->screen_lod_threshold > 0.0 && storage->mesh_surface_has_lod(mesh_surface)) { - //lod - Vector3 support_min = surf->owner->transformed_aabb.get_support(-p_params->lod_plane.normal); - Vector3 support_max = surf->owner->transformed_aabb.get_support(p_params->lod_plane.normal); - - float distance_min = p_params->lod_plane.distance_to(support_min); - float distance_max = p_params->lod_plane.distance_to(support_max); - - float distance = 0.0; - - if (distance_min * distance_max < 0.0) { - //crossing plane - distance = 0.0; - } else if (distance_min >= 0.0) { - distance = distance_min; - } else if (distance_max <= 0.0) { - distance = -distance_max; - } - - index_array_rd = storage->mesh_surface_get_index_array_with_lod(mesh_surface, surf->owner->lod_model_scale * surf->owner->lod_bias, distance * p_params->lod_distance_multiplier, p_params->screen_lod_threshold); - - } else { - //no lod - index_array_rd = storage->mesh_surface_get_index_array(mesh_surface); - } + index_array_rd = storage->mesh_surface_get_index_array(mesh_surface, element_info.lod_index); if (prev_vertex_array_rd != vertex_array_rd) { RD::get_singleton()->draw_list_bind_vertex_array(draw_list, vertex_array_rd); @@ -997,14 +953,11 @@ void RendererSceneRenderForward::_render_list_template(RenderingDevice::DrawList prev_material_uniform_set = material_uniform_set; } - RD::get_singleton()->draw_list_set_push_constant(draw_list, &surf->owner->push_constant, sizeof(GeometryInstanceForward::PushConstant)); - - RD::get_singleton()->draw_list_draw(draw_list, index_array_rd.is_valid(), surf->owner->instance_count); + RD::get_singleton()->draw_list_set_push_constant(draw_list, &push_constant, sizeof(SceneState::PushConstant)); - if (p_params->pass_mode == PASS_MODE_DEPTH_MATERIAL) { - surf->owner->push_constant.lightmap_uv_scale[0] = old_offset[0]; - surf->owner->push_constant.lightmap_uv_scale[1] = old_offset[1]; - } + uint32_t instance_count = surf->owner->instance_count > 1 ? surf->owner->instance_count : element_info.repeat; + RD::get_singleton()->draw_list_draw(draw_list, index_array_rd.is_valid(), instance_count); + i += element_info.repeat - 1; //skip equal elements } } @@ -1062,16 +1015,16 @@ void RendererSceneRenderForward::_render_list_with_threads(RenderListParameters thread_draw_lists.resize(RendererThreadPool::singleton->thread_work_pool.get_thread_count()); RD::get_singleton()->draw_list_begin_split(p_framebuffer, thread_draw_lists.size(), thread_draw_lists.ptr(), p_initial_color_action, p_final_color_action, p_initial_depth_action, p_final_depth_action, p_clear_color_values, p_clear_depth, p_clear_stencil, p_region, p_storage_textures); RendererThreadPool::singleton->thread_work_pool.do_work(thread_draw_lists.size(), this, &RendererSceneRenderForward::_render_list_thread_function, p_params); - RD::get_singleton()->draw_list_end(); + RD::get_singleton()->draw_list_end(p_params->barrier); } else { //single threaded RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_framebuffer, p_initial_color_action, p_final_color_action, p_initial_depth_action, p_final_depth_action, p_clear_color_values, p_clear_depth, p_clear_stencil, p_region, p_storage_textures); _render_list(draw_list, fb_format, p_params, 0, p_params->element_count); - RD::get_singleton()->draw_list_end(); + RD::get_singleton()->draw_list_end(p_params->barrier); } } -void RendererSceneRenderForward::_setup_environment(RID p_environment, RID p_render_buffers, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_reflection_probe, bool p_no_fog, const Size2i &p_screen_size, uint32_t p_cluster_size, uint32_t p_max_cluster_elements, RID p_shadow_atlas, bool p_flip_y, const Color &p_default_bg_color, float p_znear, float p_zfar, bool p_opaque_render_buffers, bool p_pancake_shadows) { +void RendererSceneRenderForward::_setup_environment(RID p_environment, RID p_render_buffers, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_reflection_probe, bool p_no_fog, const Size2i &p_screen_size, uint32_t p_cluster_size, uint32_t p_max_cluster_elements, RID p_shadow_atlas, bool p_flip_y, const Color &p_default_bg_color, float p_znear, float p_zfar, bool p_opaque_render_buffers, bool p_pancake_shadows, int p_index) { //CameraMatrix projection = p_cam_projection; //projection.flip_y(); // Vulkan and modern APIs use Y-Down CameraMatrix correction; @@ -1310,22 +1263,120 @@ void RendererSceneRenderForward::_setup_environment(RID p_environment, RID p_ren scene_state.ubo.roughness_limiter_amount = screen_space_roughness_limiter_get_amount(); scene_state.ubo.roughness_limiter_limit = screen_space_roughness_limiter_get_limit(); - RD::get_singleton()->buffer_update(scene_state.uniform_buffer, 0, sizeof(SceneState::UBO), &scene_state.ubo, true); + if (p_index >= (int)scene_state.uniform_buffers.size()) { + uint32_t from = scene_state.uniform_buffers.size(); + scene_state.uniform_buffers.resize(p_index + 1); + render_pass_uniform_sets.resize(p_index + 1); + for (uint32_t i = from; i < scene_state.uniform_buffers.size(); i++) { + scene_state.uniform_buffers[i] = RD::get_singleton()->uniform_buffer_create(sizeof(SceneState::UBO)); + } + } + RD::get_singleton()->buffer_update(scene_state.uniform_buffers[p_index], 0, sizeof(SceneState::UBO), &scene_state.ubo, RD::BARRIER_MASK_RASTER); +} + +void RendererSceneRenderForward::_update_instance_data_buffer(RenderListType p_render_list) { + if (scene_state.instance_data[p_render_list].size() > 0) { + if (scene_state.instance_buffer[p_render_list] == RID() || scene_state.instance_buffer_size[p_render_list] < scene_state.instance_data[p_render_list].size()) { + if (scene_state.instance_buffer[p_render_list] != RID()) { + RD::get_singleton()->free(scene_state.instance_buffer[p_render_list]); + } + uint32_t new_size = nearest_power_of_2_templated(MAX(uint64_t(INSTANCE_DATA_BUFFER_MIN_SIZE), scene_state.instance_data[p_render_list].size())); + scene_state.instance_buffer[p_render_list] = RD::get_singleton()->storage_buffer_create(new_size * sizeof(SceneState::InstanceData)); + scene_state.instance_buffer_size[p_render_list] = new_size; + } + RD::get_singleton()->buffer_update(scene_state.instance_buffer[p_render_list], 0, sizeof(SceneState::InstanceData) * scene_state.instance_data[p_render_list].size(), scene_state.instance_data[p_render_list].ptr(), RD::BARRIER_MASK_RASTER); + } +} +void RendererSceneRenderForward::_fill_instance_data(RenderListType p_render_list, uint32_t p_offset, int32_t p_max_elements, bool p_update_buffer) { + RenderList *rl = &render_list[p_render_list]; + uint32_t element_total = p_max_elements >= 0 ? uint32_t(p_max_elements) : rl->elements.size(); + + scene_state.instance_data[p_render_list].resize(p_offset + element_total); + rl->element_info.resize(p_offset + element_total); + + uint32_t repeats = 0; + GeometryInstanceSurfaceDataCache *prev_surface = nullptr; + for (uint32_t i = 0; i < element_total; i++) { + GeometryInstanceSurfaceDataCache *surface = rl->elements[i + p_offset]; + GeometryInstanceForward *inst = surface->owner; + + SceneState::InstanceData &instance_data = scene_state.instance_data[p_render_list][i + p_offset]; + + if (inst->store_transform_cache) { + RendererStorageRD::store_transform(inst->transform, instance_data.transform); + } else { + RendererStorageRD::store_transform(Transform(), instance_data.transform); + } + + instance_data.flags = inst->flags_cache; + instance_data.gi_offset = inst->gi_offset_cache; + instance_data.layer_mask = inst->layer_mask; + instance_data.instance_uniforms_ofs = uint32_t(inst->shader_parameters_offset); + instance_data.lightmap_uv_scale[0] = inst->lightmap_uv_scale.position.x; + instance_data.lightmap_uv_scale[1] = inst->lightmap_uv_scale.position.y; + instance_data.lightmap_uv_scale[2] = inst->lightmap_uv_scale.size.x; + instance_data.lightmap_uv_scale[3] = inst->lightmap_uv_scale.size.y; + + bool cant_repeat = instance_data.flags & INSTANCE_DATA_FLAG_MULTIMESH || inst->mesh_instance.is_valid(); + + if (prev_surface != nullptr && !cant_repeat && prev_surface->sort.sort_key1 == surface->sort.sort_key1 && prev_surface->sort.sort_key2 == surface->sort.sort_key2) { + //this element is the same as the previous one, count repeats to draw it using instancing + repeats++; + } else { + if (repeats > 0) { + for (uint32_t j = 1; j <= repeats; j++) { + rl->element_info[p_offset + i - j].repeat = j; + } + } + repeats = 1; + } + + RenderElementInfo &element_info = rl->element_info[p_offset + i]; + + element_info.lod_index = surface->sort.lod_index; + element_info.uses_forward_gi = surface->sort.uses_forward_gi; + element_info.uses_lightmap = surface->sort.uses_lightmap; + + if (cant_repeat) { + prev_surface = nullptr; + } else { + prev_surface = surface; + } + } + + if (repeats > 0) { + for (uint32_t j = 1; j <= repeats; j++) { + rl->element_info[p_offset + element_total - j].repeat = j; + } + } + + if (p_update_buffer) { + _update_instance_data_buffer(p_render_list); + } } -void RendererSceneRenderForward::_fill_render_list(const PagedArray<GeometryInstance *> &p_instances, PassMode p_pass_mode, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, bool p_using_sdfgi, bool p_using_opaque_gi) { - scene_state.used_sss = false; - scene_state.used_screen_texture = false; - scene_state.used_normal_texture = false; - scene_state.used_depth_texture = false; +void RendererSceneRenderForward::_fill_render_list(RenderListType p_render_list, const PagedArray<GeometryInstance *> &p_instances, PassMode p_pass_mode, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, bool p_using_sdfgi, bool p_using_opaque_gi, const Plane &p_lod_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold, bool p_append) { + if (p_render_list == RENDER_LIST_OPAQUE) { + scene_state.used_sss = false; + scene_state.used_screen_texture = false; + scene_state.used_normal_texture = false; + scene_state.used_depth_texture = false; + } + uint32_t lightmap_captures_used = 0; Plane near_plane(p_cam_transform.origin, -p_cam_transform.basis.get_axis(Vector3::AXIS_Z)); near_plane.d += p_cam_projection.get_z_near(); float z_max = p_cam_projection.get_z_far() - p_cam_projection.get_z_near(); - uint32_t lightmap_captures_used = 0; + RenderList *rl = &render_list[p_render_list]; _update_dirty_geometry_instances(); - render_list.clear(); + + if (!p_append) { + rl->clear(); + if (p_render_list == RENDER_LIST_OPAQUE) { + render_list[RENDER_LIST_ALPHA].clear(); //opaque fills alpha too + } + } //fill list @@ -1341,7 +1392,7 @@ void RendererSceneRenderForward::_fill_render_list(const PagedArray<GeometryInst bool uses_lightmap = false; bool uses_gi = false; - if (p_pass_mode == PASS_MODE_COLOR) { + if (p_render_list == RENDER_LIST_OPAQUE) { //setup GI if (inst->lightmap_instance.is_valid()) { @@ -1353,15 +1404,15 @@ void RendererSceneRenderForward::_fill_render_list(const PagedArray<GeometryInst } } if (lightmap_cull_index >= 0) { - inst->push_constant.gi_offset &= 0xFFFF; - inst->push_constant.gi_offset |= lightmap_cull_index; + inst->gi_offset_cache = inst->lightmap_slice_index << 16; + inst->gi_offset_cache |= lightmap_cull_index; flags |= INSTANCE_DATA_FLAG_USE_LIGHTMAP; if (scene_state.lightmap_has_sh[lightmap_cull_index]) { flags |= INSTANCE_DATA_FLAG_USE_SH_LIGHTMAP; } uses_lightmap = true; } else { - inst->push_constant.gi_offset = 0xFFFFFFFF; + inst->gi_offset_cache = 0xFFFFFFFF; } } else if (inst->lightmap_sh) { @@ -1375,7 +1426,7 @@ void RendererSceneRenderForward::_fill_render_list(const PagedArray<GeometryInst lcd.sh[j * 4 + 3] = src_capture[j].a; } flags |= INSTANCE_DATA_FLAG_USE_LIGHTMAP_CAPTURE; - inst->push_constant.gi_offset = lightmap_captures_used; + inst->gi_offset_cache = lightmap_captures_used; lightmap_captures_used++; uses_lightmap = true; } @@ -1402,18 +1453,19 @@ void RendererSceneRenderForward::_fill_render_list(const PagedArray<GeometryInst SWAP(probe0_index, probe1_index); } - inst->push_constant.gi_offset = probe0_index | (probe1_index << 16); + inst->gi_offset_cache = probe0_index | (probe1_index << 16); + flags |= INSTANCE_DATA_FLAG_USE_GIPROBE; uses_gi = true; } else { if (p_using_sdfgi && inst->can_sdfgi) { flags |= INSTANCE_DATA_FLAG_USE_SDFGI; uses_gi = true; } - inst->push_constant.gi_offset = 0xFFFFFFFF; + inst->gi_offset_cache = 0xFFFFFFFF; } } } - inst->push_constant.flags = flags; + inst->flags_cache = flags; GeometryInstanceSurfaceDataCache *surf = inst->surface_caches; @@ -1421,12 +1473,39 @@ void RendererSceneRenderForward::_fill_render_list(const PagedArray<GeometryInst surf->sort.uses_forward_gi = 0; surf->sort.uses_lightmap = 0; + // LOD + + if (p_screen_lod_threshold > 0.0 && storage->mesh_surface_has_lod(surf->surface)) { + //lod + Vector3 lod_support_min = inst->transformed_aabb.get_support(-p_lod_plane.normal); + Vector3 lod_support_max = inst->transformed_aabb.get_support(p_lod_plane.normal); + + float distance_min = p_lod_plane.distance_to(lod_support_min); + float distance_max = p_lod_plane.distance_to(lod_support_max); + + float distance = 0.0; + + if (distance_min * distance_max < 0.0) { + //crossing plane + distance = 0.0; + } else if (distance_min >= 0.0) { + distance = distance_min; + } else if (distance_max <= 0.0) { + distance = -distance_max; + } + + surf->sort.lod_index = storage->mesh_surface_get_lod(surf->surface, inst->lod_model_scale * inst->lod_bias, distance * p_lod_distance_multiplier, p_screen_lod_threshold); + } else { + surf->sort.lod_index = 0; + } + + // ADD Element if (p_pass_mode == PASS_MODE_COLOR) { if (surf->flags & (GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH | GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE)) { - render_list.add_element(surf); + rl->add_element(surf); } if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_PASS_ALPHA) { - render_list.add_alpha_element(surf); + render_list[RENDER_LIST_ALPHA].add_element(surf); if (uses_gi) { surf->sort.uses_forward_gi = 1; } @@ -1451,11 +1530,11 @@ void RendererSceneRenderForward::_fill_render_list(const PagedArray<GeometryInst } else if (p_pass_mode == PASS_MODE_SHADOW || p_pass_mode == PASS_MODE_SHADOW_DP) { if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_PASS_SHADOW) { - render_list.add_element(surf); + rl->add_element(surf); } } else { if (surf->flags & (GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH | GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE)) { - render_list.add_element(surf); + rl->add_element(surf); } } @@ -1465,8 +1544,8 @@ void RendererSceneRenderForward::_fill_render_list(const PagedArray<GeometryInst } } - if (lightmap_captures_used) { - RD::get_singleton()->buffer_update(scene_state.lightmap_capture_buffer, 0, sizeof(LightmapCaptureData) * lightmap_captures_used, scene_state.lightmap_captures, true); + if (p_render_list == RENDER_LIST_OPAQUE && lightmap_captures_used) { + RD::get_singleton()->buffer_update(scene_state.lightmap_capture_buffer, 0, sizeof(LightmapCaptureData) * lightmap_captures_used, scene_state.lightmap_captures, RD::BARRIER_MASK_RASTER); } } @@ -1495,29 +1574,21 @@ void RendererSceneRenderForward::_setup_lightmaps(const PagedArray<RID> &p_light scene_state.lightmaps_used++; } if (scene_state.lightmaps_used > 0) { - RD::get_singleton()->buffer_update(scene_state.lightmap_buffer, 0, sizeof(LightmapData) * scene_state.lightmaps_used, scene_state.lightmaps, true); + RD::get_singleton()->buffer_update(scene_state.lightmap_buffer, 0, sizeof(LightmapData) * scene_state.lightmaps_used, scene_state.lightmaps, RD::BARRIER_MASK_RASTER); } } -void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, int p_directional_light_count, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_cluster_buffer, uint32_t p_cluster_size, uint32_t p_max_cluster_elements, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_bg_color, float p_screen_lod_threshold) { +void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_cluster_buffer, uint32_t p_cluster_size, uint32_t p_max_cluster_elements, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_bg_color, float p_screen_lod_threshold) { RenderBufferDataForward *render_buffer = nullptr; if (p_render_buffer.is_valid()) { render_buffer = (RenderBufferDataForward *)render_buffers_get_data(p_render_buffer); } //first of all, make a new render pass - render_pass++; - //fill up ubo RENDER_TIMESTAMP("Setup 3D Scene"); - if (p_reflection_probe.is_valid()) { - scene_state.ubo.reflection_multiplier = 0.0; - } else { - scene_state.ubo.reflection_multiplier = 1.0; - } - float lod_distance_multiplier = p_cam_projection.get_lod_multiplier(); Plane lod_camera_plane(p_cam_transform.get_origin(), -p_cam_transform.basis.get_axis(Vector3::AXIS_Z)); @@ -1530,7 +1601,7 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf Vector2 vp_he = p_cam_projection.get_viewport_half_extents(); scene_state.ubo.viewport_size[0] = vp_he.x; scene_state.ubo.viewport_size[1] = vp_he.y; - scene_state.ubo.directional_light_count = p_directional_light_count; + scene_state.ubo.directional_light_count = 0; Size2i screen_size; RID opaque_framebuffer; @@ -1553,7 +1624,6 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf if (!low_end && p_gi_probes.size() > 0) { using_giprobe = true; - render_buffer->ensure_gi(); } if (!p_environment.is_valid() && using_giprobe) { @@ -1563,7 +1633,6 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf if (environment_is_sdfgi_enabled(p_environment)) { depth_pass_mode = using_giprobe ? PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE : PASS_MODE_DEPTH_NORMAL_ROUGHNESS; // also giprobe using_sdfgi = true; - render_buffer->ensure_gi(); } else { depth_pass_mode = using_giprobe ? PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE : PASS_MODE_DEPTH_NORMAL_ROUGHNESS; } @@ -1616,13 +1685,21 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf ERR_FAIL(); //bug? } + RD::get_singleton()->draw_command_begin_label("Render Setup"); + _setup_lightmaps(p_lightmaps, p_cam_transform); _setup_giprobes(p_gi_probes); _setup_environment(p_environment, p_render_buffer, p_cam_projection, p_cam_transform, p_reflection_probe, p_reflection_probe.is_valid(), screen_size, p_cluster_size, p_max_cluster_elements, p_shadow_atlas, !p_reflection_probe.is_valid(), p_default_bg_color, p_cam_projection.get_z_near(), p_cam_projection.get_z_far(), false); _update_render_base_uniform_set(); //may have changed due to the above (light buffer enlarged, as an example) - _fill_render_list(p_instances, PASS_MODE_COLOR, p_cam_projection, p_cam_transform, using_sdfgi, using_sdfgi || using_giprobe); + _fill_render_list(RENDER_LIST_OPAQUE, p_instances, PASS_MODE_COLOR, p_cam_projection, p_cam_transform, using_sdfgi, using_sdfgi || using_giprobe, lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); + render_list[RENDER_LIST_OPAQUE].sort_by_key(); + render_list[RENDER_LIST_ALPHA].sort_by_depth(); + _fill_instance_data(RENDER_LIST_OPAQUE); + _fill_instance_data(RENDER_LIST_ALPHA); + + RD::get_singleton()->draw_command_end_label(); bool using_sss = !low_end && render_buffer && scene_state.used_sss && sub_surface_scattering_get_quality() != RS::SUB_SURFACE_SCATTERING_QUALITY_DISABLED; @@ -1682,6 +1759,7 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf // setup sky if used for ambient, reflections, or background if (draw_sky || draw_sky_fog_only || environment_get_reflection_source(p_environment) == RS::ENV_REFLECTION_SOURCE_SKY || environment_get_ambient_source(p_environment) == RS::ENV_AMBIENT_SOURCE_SKY) { RENDER_TIMESTAMP("Setup Sky"); + RD::get_singleton()->draw_command_begin_label("Setup Sky"); CameraMatrix projection = p_cam_projection; if (p_reflection_probe.is_valid()) { CameraMatrix correction; @@ -1699,13 +1777,12 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf // do not try to draw sky if invalid draw_sky = false; } + RD::get_singleton()->draw_command_end_label(); } } else { clear_color = p_default_bg_color; } - render_list.sort_by_key(false); - bool debug_giprobes = get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_ALBEDO || get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_LIGHTING || get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_EMISSION; bool debug_sdfgi_probes = get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_SDFGI_PROBES; bool depth_pre_pass = !low_end && depth_framebuffer.is_valid(); @@ -1713,40 +1790,64 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf bool using_ssao = depth_pre_pass && p_render_buffer.is_valid() && p_environment.is_valid() && environment_is_ssao_enabled(p_environment); bool continue_depth = false; if (depth_pre_pass) { //depth pre pass - RENDER_TIMESTAMP("Render Depth Pre-Pass"); - RID rp_uniform_set = _setup_render_pass_uniform_set(RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); + bool needs_pre_resolve = _needs_post_prepass_render(using_sdfgi || using_giprobe); + if (needs_pre_resolve) { + RENDER_TIMESTAMP("GI + Render Depth Pre-Pass (parallel)"); + } else { + RENDER_TIMESTAMP("Render Depth Pre-Pass"); + } + if (needs_pre_resolve) { + //pre clear the depth framebuffer, as AMD (and maybe others?) use compute for it, and barrier other compute shaders. + RD::get_singleton()->draw_list_begin(depth_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_CONTINUE, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_CONTINUE, depth_pass_clear); + RD::get_singleton()->draw_list_end(); + //start compute processes here, so they run at the same time as depth pre-pass + _post_prepass_render(using_sdfgi || using_giprobe); + } + + RD::get_singleton()->draw_command_begin_label("Render Depth Pre-Pass"); + + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_OPAQUE, RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); bool finish_depth = using_ssao || using_sdfgi || using_giprobe; - RenderListParameters render_list_params(render_list.elements, render_list.element_count, false, depth_pass_mode, render_buffer == nullptr, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); - _render_list_with_threads(&render_list_params, depth_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, finish_depth ? RD::FINAL_ACTION_READ : RD::FINAL_ACTION_CONTINUE, depth_pass_clear); + RenderListParameters render_list_params(render_list[RENDER_LIST_OPAQUE].elements.ptr(), render_list[RENDER_LIST_OPAQUE].element_info.ptr(), render_list[RENDER_LIST_OPAQUE].elements.size(), false, depth_pass_mode, render_buffer == nullptr, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); + _render_list_with_threads(&render_list_params, depth_framebuffer, needs_pre_resolve ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, needs_pre_resolve ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_CLEAR, finish_depth ? RD::FINAL_ACTION_READ : RD::FINAL_ACTION_CONTINUE, needs_pre_resolve ? Vector<Color>() : depth_pass_clear); + + RD::get_singleton()->draw_command_end_label(); + + if (needs_pre_resolve) { + _pre_resolve_render(using_sdfgi || using_giprobe); + } if (render_buffer && render_buffer->msaa != RS::VIEWPORT_MSAA_DISABLED) { RENDER_TIMESTAMP("Resolve Depth Pre-Pass"); + RD::get_singleton()->draw_command_begin_label("Resolve Depth Pre-Pass"); if (depth_pass_mode == PASS_MODE_DEPTH_NORMAL_ROUGHNESS || depth_pass_mode == PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE) { + if (needs_pre_resolve) { + RD::get_singleton()->barrier(RD::BARRIER_MASK_RASTER, RD::BARRIER_MASK_COMPUTE); + } static int texture_samples[RS::VIEWPORT_MSAA_MAX] = { 1, 2, 4, 8, 16 }; storage->get_effects()->resolve_gi(render_buffer->depth_msaa, render_buffer->normal_roughness_buffer_msaa, using_giprobe ? render_buffer->giprobe_buffer_msaa : RID(), render_buffer->depth, render_buffer->normal_roughness_buffer, using_giprobe ? render_buffer->giprobe_buffer : RID(), Vector2i(render_buffer->width, render_buffer->height), texture_samples[render_buffer->msaa]); } else if (finish_depth) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->depth_msaa, render_buffer->depth, true); + RD::get_singleton()->texture_resolve_multisample(render_buffer->depth_msaa, render_buffer->depth); } + RD::get_singleton()->draw_command_end_label(); } continue_depth = !finish_depth; } - if (using_ssao) { - _process_ssao(p_render_buffer, p_environment, render_buffer->normal_roughness_buffer, p_cam_projection); - } + _pre_opaque_render(using_ssao, using_sdfgi || using_giprobe, render_buffer ? render_buffer->normal_roughness_buffer : RID(), render_buffer ? render_buffer->giprobe_buffer : RID()); - if (using_sdfgi || using_giprobe) { - _process_gi(p_render_buffer, render_buffer->normal_roughness_buffer, render_buffer->ambient_buffer, render_buffer->reflection_buffer, render_buffer->giprobe_buffer, p_environment, p_cam_projection, p_cam_transform, p_gi_probes); - } + RD::get_singleton()->draw_command_begin_label("Render Opaque Pass"); + + scene_state.ubo.directional_light_count = _get_render_state_directional_light_count(); _setup_environment(p_environment, p_render_buffer, p_cam_projection, p_cam_transform, p_reflection_probe, p_reflection_probe.is_valid(), screen_size, p_cluster_size, p_max_cluster_elements, p_shadow_atlas, !p_reflection_probe.is_valid(), p_default_bg_color, p_cam_projection.get_z_near(), p_cam_projection.get_z_far(), p_render_buffer.is_valid()); RENDER_TIMESTAMP("Render Opaque Pass"); - RID rp_uniform_set = _setup_render_pass_uniform_set(p_render_buffer, radiance_texture, p_shadow_atlas, p_reflection_atlas, p_cluster_buffer, p_gi_probes, p_lightmaps); + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_OPAQUE, p_render_buffer, radiance_texture, p_shadow_atlas, p_reflection_atlas, p_cluster_buffer, p_gi_probes, p_lightmaps, true); bool can_continue_color = !scene_state.used_screen_texture && !using_ssr && !using_sss; bool can_continue_depth = !scene_state.used_depth_texture && !using_ssr && !using_sss; @@ -1767,10 +1868,8 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf } RID framebuffer = using_separate_specular ? opaque_specular_framebuffer : opaque_framebuffer; - RenderListParameters render_list_params(render_list.elements, render_list.element_count, false, using_separate_specular ? PASS_MODE_COLOR_SPECULAR : PASS_MODE_COLOR, render_buffer == nullptr, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); - - _render_list_with_threads(&render_list_params, framebuffer, keep_color ? RD::INITIAL_ACTION_KEEP : RD::INITIAL_ACTION_CLEAR, will_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, depth_pre_pass ? (continue_depth ? RD::INITIAL_ACTION_KEEP : RD::INITIAL_ACTION_CONTINUE) : RD::INITIAL_ACTION_CLEAR, will_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, c, 1.0, 0); - + RenderListParameters render_list_params(render_list[RENDER_LIST_OPAQUE].elements.ptr(), render_list[RENDER_LIST_OPAQUE].element_info.ptr(), render_list[RENDER_LIST_OPAQUE].elements.size(), false, using_separate_specular ? PASS_MODE_COLOR_SPECULAR : PASS_MODE_COLOR, render_buffer == nullptr, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); + _render_list_with_threads(&render_list_params, framebuffer, keep_color ? RD::INITIAL_ACTION_KEEP : RD::INITIAL_ACTION_CLEAR, will_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, depth_pre_pass ? (continue_depth ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP) : RD::INITIAL_ACTION_CLEAR, will_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, c, 1.0, 0); if (will_continue_color && using_separate_specular) { // close the specular framebuffer, as it's no longer used RD::get_singleton()->draw_list_begin(render_buffer->specular_only_fb, RD::INITIAL_ACTION_CONTINUE, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CONTINUE, RD::FINAL_ACTION_CONTINUE); @@ -1778,6 +1877,8 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf } } + RD::get_singleton()->draw_command_end_label(); + if (debug_giprobes) { //debug giprobes bool will_continue_color = (can_continue_color || draw_sky || draw_sky_fog_only); @@ -1787,9 +1888,11 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf dc.set_depth_correction(true); CameraMatrix cm = (dc * p_cam_projection) * CameraMatrix(p_cam_transform.affine_inverse()); RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(opaque_framebuffer, RD::INITIAL_ACTION_CONTINUE, will_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CONTINUE, will_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ); + RD::get_singleton()->draw_command_begin_label("Debug GIProbes"); for (int i = 0; i < (int)p_gi_probes.size(); i++) { _debug_giprobe(p_gi_probes[i], draw_list, opaque_framebuffer, cm, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_LIGHTING, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_EMISSION, 1.0); } + RD::get_singleton()->draw_command_end_label(); RD::get_singleton()->draw_list_end(); } @@ -1802,7 +1905,9 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf dc.set_depth_correction(true); CameraMatrix cm = (dc * p_cam_projection) * CameraMatrix(p_cam_transform.affine_inverse()); RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(opaque_framebuffer, RD::INITIAL_ACTION_CONTINUE, will_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CONTINUE, will_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ); + RD::get_singleton()->draw_command_begin_label("Debug SDFGI"); _debug_sdfgi_probes(p_render_buffer, draw_list, opaque_framebuffer, cm); + RD::get_singleton()->draw_command_end_label(); RD::get_singleton()->draw_list_end(); } @@ -1815,30 +1920,35 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf correction.set_depth_correction(true); projection = correction * p_cam_projection; } - + RD::get_singleton()->draw_command_begin_label("Draw Sky"); _draw_sky(can_continue_color, can_continue_depth, opaque_framebuffer, p_environment, projection, p_cam_transform); + RD::get_singleton()->draw_command_end_label(); } if (render_buffer && !can_continue_color && render_buffer->msaa != RS::VIEWPORT_MSAA_DISABLED) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->color_msaa, render_buffer->color, true); + RD::get_singleton()->texture_resolve_multisample(render_buffer->color_msaa, render_buffer->color); if (using_separate_specular) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->specular_msaa, render_buffer->specular, true); + RD::get_singleton()->texture_resolve_multisample(render_buffer->specular_msaa, render_buffer->specular); } } if (render_buffer && !can_continue_depth && render_buffer->msaa != RS::VIEWPORT_MSAA_DISABLED) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->depth_msaa, render_buffer->depth, true); + RD::get_singleton()->texture_resolve_multisample(render_buffer->depth_msaa, render_buffer->depth); } if (using_separate_specular) { if (using_sss) { RENDER_TIMESTAMP("Sub Surface Scattering"); + RD::get_singleton()->draw_command_begin_label("Process Sub Surface Scattering"); _process_sss(p_render_buffer, p_cam_projection); + RD::get_singleton()->draw_command_end_label(); } if (using_ssr) { RENDER_TIMESTAMP("Screen Space Reflection"); + RD::get_singleton()->draw_command_begin_label("Process Screen Space Reflections"); _process_ssr(p_render_buffer, render_buffer->color_fb, render_buffer->normal_roughness_buffer, render_buffer->specular, render_buffer->specular, Color(0, 0, 0, 1), p_environment, p_cam_projection, render_buffer->msaa == RS::VIEWPORT_MSAA_DISABLED); + RD::get_singleton()->draw_command_end_label(); } else { //just mix specular back RENDER_TIMESTAMP("Merge Specular"); @@ -1848,30 +1958,44 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf RENDER_TIMESTAMP("Render Transparent Pass"); - _setup_environment(p_environment, p_render_buffer, p_cam_projection, p_cam_transform, p_reflection_probe, p_reflection_probe.is_valid(), screen_size, p_cluster_size, p_max_cluster_elements, p_shadow_atlas, !p_reflection_probe.is_valid(), p_default_bg_color, p_cam_projection.get_z_near(), p_cam_projection.get_z_far(), false); + RD::get_singleton()->draw_command_begin_label("Render Transparent Pass"); + + rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_ALPHA, p_render_buffer, radiance_texture, p_shadow_atlas, p_reflection_atlas, p_cluster_buffer, p_gi_probes, p_lightmaps, true); - render_list.sort_by_reverse_depth_and_priority(true); + _setup_environment(p_environment, p_render_buffer, p_cam_projection, p_cam_transform, p_reflection_probe, p_reflection_probe.is_valid(), screen_size, p_cluster_size, p_max_cluster_elements, p_shadow_atlas, !p_reflection_probe.is_valid(), p_default_bg_color, p_cam_projection.get_z_near(), p_cam_projection.get_z_far(), false); { - RenderListParameters render_list_params(&render_list.elements[render_list.max_elements - render_list.alpha_element_count], render_list.alpha_element_count, false, PASS_MODE_COLOR, render_buffer == nullptr, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); + RenderListParameters render_list_params(render_list[RENDER_LIST_ALPHA].elements.ptr(), render_list[RENDER_LIST_ALPHA].element_info.ptr(), render_list[RENDER_LIST_ALPHA].elements.size(), false, PASS_MODE_COLOR, render_buffer == nullptr, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), lod_camera_plane, lod_distance_multiplier, p_screen_lod_threshold); _render_list_with_threads(&render_list_params, alpha_framebuffer, can_continue_color ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, can_continue_depth ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ); } + RD::get_singleton()->draw_command_end_label(); + + RD::get_singleton()->draw_command_begin_label("Resolve"); + if (render_buffer && render_buffer->msaa != RS::VIEWPORT_MSAA_DISABLED) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->color_msaa, render_buffer->color, true); + RD::get_singleton()->texture_resolve_multisample(render_buffer->color_msaa, render_buffer->color); } -} -void RendererSceneRenderForward::_render_shadow(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold) { - RENDER_TIMESTAMP("Setup Rendering Shadow"); + RD::get_singleton()->draw_command_end_label(); +} +void RendererSceneRenderForward::_render_shadow_begin() { + scene_state.shadow_passes.clear(); + RD::get_singleton()->draw_command_begin_label("Shadow Setup"); _update_render_base_uniform_set(); - render_pass++; + render_list[RENDER_LIST_SECONDARY].clear(); + scene_state.instance_data[RENDER_LIST_SECONDARY].clear(); +} +void RendererSceneRenderForward::_render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold, const Rect2i &p_rect, bool p_flip_y, bool p_clear_region, bool p_begin, bool p_end) { + uint32_t shadow_pass_index = scene_state.shadow_passes.size(); + + SceneState::ShadowPass shadow_pass; scene_state.ubo.dual_paraboloid_side = p_use_dp_flip ? -1 : 1; - _setup_environment(RID(), RID(), p_projection, p_transform, RID(), true, Vector2(1, 1), 1, 32, RID(), true, Color(), 0, p_zfar, false, p_use_pancake); + _setup_environment(RID(), RID(), p_projection, p_transform, RID(), true, Vector2(1, 1), 1, 32, RID(), !p_flip_y, Color(), 0, p_zfar, false, p_use_pancake, shadow_pass_index); if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_DISABLE_LOD) { p_screen_lod_threshold = 0.0; @@ -1879,72 +2003,116 @@ void RendererSceneRenderForward::_render_shadow(RID p_framebuffer, const PagedAr PassMode pass_mode = p_use_dp ? PASS_MODE_SHADOW_DP : PASS_MODE_SHADOW; - _fill_render_list(p_instances, pass_mode, p_projection, p_transform); + uint32_t render_list_from = render_list[RENDER_LIST_SECONDARY].elements.size(); + _fill_render_list(RENDER_LIST_SECONDARY, p_instances, pass_mode, p_projection, p_transform, false, false, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold, true); + uint32_t render_list_size = render_list[RENDER_LIST_SECONDARY].elements.size() - render_list_from; + render_list[RENDER_LIST_SECONDARY].sort_by_key_range(render_list_from, render_list_size); + _fill_instance_data(RENDER_LIST_SECONDARY, render_list_from, render_list_size, false); - RID rp_uniform_set = _setup_render_pass_uniform_set(RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); + { + //regular forward for now + bool flip_cull = p_use_dp_flip; + if (p_flip_y) { + flip_cull = !flip_cull; + } - RENDER_TIMESTAMP("Render Shadow"); + shadow_pass.element_from = render_list_from; + shadow_pass.element_count = render_list_size; + shadow_pass.flip_cull = flip_cull; + shadow_pass.pass_mode = pass_mode; - render_list.sort_by_key(false); + shadow_pass.rp_uniform_set = RID(); //will be filled later when instance buffer is complete + shadow_pass.camera_plane = p_camera_plane; + shadow_pass.screen_lod_threshold = p_screen_lod_threshold; + shadow_pass.lod_distance_multiplier = p_lod_distance_multiplier; - { - //regular forward for now - RenderListParameters render_list_params(render_list.elements, render_list.element_count, p_use_dp_flip, pass_mode, true, rp_uniform_set, false, Vector2(), p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold); - _render_list_with_threads(&render_list_params, p_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ); + shadow_pass.framebuffer = p_framebuffer; + shadow_pass.initial_depth_action = p_begin ? (p_clear_region ? RD::INITIAL_ACTION_CLEAR_REGION : RD::INITIAL_ACTION_CLEAR) : (p_clear_region ? RD::INITIAL_ACTION_CLEAR_REGION_CONTINUE : RD::INITIAL_ACTION_CONTINUE); + shadow_pass.final_depth_action = p_end ? RD::FINAL_ACTION_READ : RD::FINAL_ACTION_CONTINUE; + shadow_pass.rect = p_rect; + + scene_state.shadow_passes.push_back(shadow_pass); } } +void RendererSceneRenderForward::_render_shadow_process() { + _update_instance_data_buffer(RENDER_LIST_SECONDARY); + //render shadows one after the other, so this can be done un-barriered and the driver can optimize (as well as allow us to run compute at the same time) + + for (uint32_t i = 0; i < scene_state.shadow_passes.size(); i++) { + //render passes need to be configured after instance buffer is done, since they need the latest version + SceneState::ShadowPass &shadow_pass = scene_state.shadow_passes[i]; + shadow_pass.rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>(), false, i); + } + + RD::get_singleton()->draw_command_end_label(); +} +void RendererSceneRenderForward::_render_shadow_end(uint32_t p_barrier) { + RD::get_singleton()->draw_command_begin_label("Shadow Render"); + + for (uint32_t i = 0; i < scene_state.shadow_passes.size(); i++) { + SceneState::ShadowPass &shadow_pass = scene_state.shadow_passes[i]; + RenderListParameters render_list_parameters(render_list[RENDER_LIST_SECONDARY].elements.ptr() + shadow_pass.element_from, render_list[RENDER_LIST_SECONDARY].element_info.ptr() + shadow_pass.element_from, shadow_pass.element_count, shadow_pass.flip_cull, shadow_pass.pass_mode, true, shadow_pass.rp_uniform_set, false, Vector2(), shadow_pass.camera_plane, shadow_pass.lod_distance_multiplier, shadow_pass.screen_lod_threshold, shadow_pass.element_from, RD::BARRIER_MASK_NO_BARRIER); + _render_list_with_threads(&render_list_parameters, shadow_pass.framebuffer, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, shadow_pass.initial_depth_action, shadow_pass.final_depth_action, Vector<Color>(), 1.0, 0, shadow_pass.rect); + } + + if (p_barrier != RD::BARRIER_MASK_NO_BARRIER) { + RD::get_singleton()->barrier(RD::BARRIER_MASK_RASTER, p_barrier); + } + RD::get_singleton()->draw_command_end_label(); +} + void RendererSceneRenderForward::_render_particle_collider_heightfield(RID p_fb, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<GeometryInstance *> &p_instances) { RENDER_TIMESTAMP("Setup Render Collider Heightfield"); - _update_render_base_uniform_set(); - - render_pass++; + RD::get_singleton()->draw_command_begin_label("Render Collider Heightfield"); + _update_render_base_uniform_set(); scene_state.ubo.dual_paraboloid_side = 0; _setup_environment(RID(), RID(), p_cam_projection, p_cam_transform, RID(), true, Vector2(1, 1), 1, 32, RID(), true, Color(), 0, p_cam_projection.get_z_far(), false, false); PassMode pass_mode = PASS_MODE_SHADOW; - _fill_render_list(p_instances, pass_mode, p_cam_projection, p_cam_transform); + _fill_render_list(RENDER_LIST_SECONDARY, p_instances, pass_mode, p_cam_projection, p_cam_transform); + render_list[RENDER_LIST_SECONDARY].sort_by_key(); + _fill_instance_data(RENDER_LIST_SECONDARY); - RID rp_uniform_set = _setup_render_pass_uniform_set(RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); - RENDER_TIMESTAMP("Render Collider Heightield"); - - render_list.sort_by_key(false); + RENDER_TIMESTAMP("Render Collider Heightfield"); { //regular forward for now - RenderListParameters render_list_params(render_list.elements, render_list.element_count, false, pass_mode, true, rp_uniform_set); + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), false, pass_mode, true, rp_uniform_set); _render_list_with_threads(&render_list_params, p_fb, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ); } + RD::get_singleton()->draw_command_end_label(); } void RendererSceneRenderForward::_render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) { RENDER_TIMESTAMP("Setup Rendering Material"); - _update_render_base_uniform_set(); + RD::get_singleton()->draw_command_begin_label("Render Material"); - render_pass++; + _update_render_base_uniform_set(); scene_state.ubo.dual_paraboloid_side = 0; - scene_state.ubo.material_uv2_mode = true; + scene_state.ubo.material_uv2_mode = false; _setup_environment(RID(), RID(), p_cam_projection, p_cam_transform, RID(), true, Vector2(1, 1), 1, 32, RID(), false, Color(), 0, 0); PassMode pass_mode = PASS_MODE_DEPTH_MATERIAL; - _fill_render_list(p_instances, pass_mode, p_cam_projection, p_cam_transform); + _fill_render_list(RENDER_LIST_SECONDARY, p_instances, pass_mode, p_cam_projection, p_cam_transform); + render_list[RENDER_LIST_SECONDARY].sort_by_key(); + _fill_instance_data(RENDER_LIST_SECONDARY); - RID rp_uniform_set = _setup_render_pass_uniform_set(RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); RENDER_TIMESTAMP("Render Material"); - render_list.sort_by_key(false); - { - RenderListParameters render_list_params(render_list.elements, render_list.element_count, true, pass_mode, true, rp_uniform_set); + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, true, rp_uniform_set); //regular forward for now Vector<Color> clear; clear.push_back(Color(0, 0, 0, 0)); @@ -1956,14 +2124,16 @@ void RendererSceneRenderForward::_render_material(const Transform &p_cam_transfo _render_list(draw_list, RD::get_singleton()->framebuffer_get_format(p_framebuffer), &render_list_params, 0, render_list_params.element_count); RD::get_singleton()->draw_list_end(); } + + RD::get_singleton()->draw_command_end_label(); } void RendererSceneRenderForward::_render_uv2(const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) { RENDER_TIMESTAMP("Setup Rendering UV2"); - _update_render_base_uniform_set(); + RD::get_singleton()->draw_command_begin_label("Render UV2"); - render_pass++; + _update_render_base_uniform_set(); scene_state.ubo.dual_paraboloid_side = 0; scene_state.ubo.material_uv2_mode = true; @@ -1971,16 +2141,16 @@ void RendererSceneRenderForward::_render_uv2(const PagedArray<GeometryInstance * _setup_environment(RID(), RID(), CameraMatrix(), Transform(), RID(), true, Vector2(1, 1), 1, 32, RID(), false, Color(), 0, 0); PassMode pass_mode = PASS_MODE_DEPTH_MATERIAL; - _fill_render_list(p_instances, pass_mode, CameraMatrix(), Transform()); + _fill_render_list(RENDER_LIST_SECONDARY, p_instances, pass_mode, CameraMatrix(), Transform()); + render_list[RENDER_LIST_SECONDARY].sort_by_key(); + _fill_instance_data(RENDER_LIST_SECONDARY); - RID rp_uniform_set = _setup_render_pass_uniform_set(RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, RID(), RID(), RID(), RID(), RID(), PagedArray<RID>(), PagedArray<RID>()); RENDER_TIMESTAMP("Render Material"); - render_list.sort_by_key(false); - { - RenderListParameters render_list_params(render_list.elements, render_list.element_count, true, pass_mode, true, rp_uniform_set, true); + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, true, rp_uniform_set, true); //regular forward for now Vector<Color> clear; clear.push_back(Color(0, 0, 0, 0)); @@ -2016,23 +2186,24 @@ void RendererSceneRenderForward::_render_uv2(const PagedArray<GeometryInstance * RD::get_singleton()->draw_list_end(); } + + RD::get_singleton()->draw_command_end_label(); } void RendererSceneRenderForward::_render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<GeometryInstance *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture) { RENDER_TIMESTAMP("Render SDFGI"); + RD::get_singleton()->draw_command_begin_label("Render SDFGI Voxel"); + _update_render_base_uniform_set(); RenderBufferDataForward *render_buffer = (RenderBufferDataForward *)render_buffers_get_data(p_render_buffers); ERR_FAIL_COND(!render_buffer); - render_pass++; - PassMode pass_mode = PASS_MODE_SDF; - _fill_render_list(p_instances, pass_mode, CameraMatrix(), Transform()); - render_list.sort_by_key(false); - - RID rp_uniform_set = _setup_sdfgi_render_pass_uniform_set(p_albedo_texture, p_emission_texture, p_emission_aniso_texture, p_geom_facing_texture); + _fill_render_list(RENDER_LIST_SECONDARY, p_instances, pass_mode, CameraMatrix(), Transform()); + render_list[RENDER_LIST_SECONDARY].sort_by_key(); + _fill_instance_data(RENDER_LIST_SECONDARY); Vector3 half_extents = p_bounds.size * 0.5; Vector3 center = p_bounds.position + half_extents; @@ -2085,15 +2256,19 @@ void RendererSceneRenderForward::_render_sdfgi(RID p_render_buffers, const Vecto _setup_environment(RID(), RID(), camera_proj, cam_xform, RID(), true, Vector2(1, 1), 1, 32, RID(), false, Color(), 0, 0); + RID rp_uniform_set = _setup_sdfgi_render_pass_uniform_set(p_albedo_texture, p_emission_texture, p_emission_aniso_texture, p_geom_facing_texture); + Map<Size2i, RID>::Element *E = sdfgi_framebuffer_size_cache.find(fb_size); if (!E) { RID fb = RD::get_singleton()->framebuffer_create_empty(fb_size); E = sdfgi_framebuffer_size_cache.insert(fb_size, fb); } - RenderListParameters render_list_params(render_list.elements, render_list.element_count, true, pass_mode, true, rp_uniform_set, false); + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, true, rp_uniform_set, false); _render_list_with_threads(&render_list_params, E->get(), RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, Vector<Color>(), 1.0, 0, Rect2(), sbs); } + + RD::get_singleton()->draw_command_end_label(); } void RendererSceneRenderForward::_base_uniforms_changed() { @@ -2145,21 +2320,13 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() { { RD::Uniform u; u.binding = 3; - u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.ids.push_back(scene_state.uniform_buffer); - uniforms.push_back(u); - } - - { - RD::Uniform u; - u.binding = 5; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(get_omni_light_buffer()); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 6; + u.binding = 4; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(get_spot_light_buffer()); uniforms.push_back(u); @@ -2167,35 +2334,35 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() { { RD::Uniform u; - u.binding = 7; + u.binding = 5; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(get_reflection_probe_buffer()); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 8; + u.binding = 6; u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; u.ids.push_back(get_directional_light_buffer()); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 10; + u.binding = 7; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(scene_state.lightmap_buffer); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 11; + u.binding = 8; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(scene_state.lightmap_capture_buffer); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 12; + u.binding = 9; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID decal_atlas = storage->decal_atlas_get_texture(); u.ids.push_back(decal_atlas); @@ -2203,7 +2370,7 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() { } { RD::Uniform u; - u.binding = 13; + u.binding = 10; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID decal_atlas = storage->decal_atlas_get_texture_srgb(); u.ids.push_back(decal_atlas); @@ -2211,7 +2378,7 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() { } { RD::Uniform u; - u.binding = 14; + u.binding = 11; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(get_decal_buffer()); uniforms.push_back(u); @@ -2219,20 +2386,8 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() { { RD::Uniform u; - u.binding = 15; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - if (directional_shadow_get_texture().is_valid()) { - u.ids.push_back(directional_shadow_get_texture()); - } else { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE)); - } - uniforms.push_back(u); - } - - { - RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.binding = 16; + u.binding = 12; u.ids.push_back(storage->global_variables_get_storage_buffer()); uniforms.push_back(u); } @@ -2240,7 +2395,7 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() { if (!low_end) { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.binding = 17; + u.binding = 13; u.ids.push_back(sdfgi_get_ubo()); uniforms.push_back(u); } @@ -2249,10 +2404,9 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() { } } -RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buffers, RID p_radiance_texture, RID p_shadow_atlas, RID p_reflection_atlas, RID p_cluster_buffer, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps) { - if (render_pass_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(render_pass_uniform_set)) { - RD::get_singleton()->free(render_pass_uniform_set); - } +RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RenderListType p_render_list, RID p_render_buffers, RID p_radiance_texture, RID p_shadow_atlas, RID p_reflection_atlas, RID p_cluster_buffer, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, bool p_use_directional_shadow_atlas, int p_index) { + //there should always be enough uniform buffers for render passes, otherwise bugs + ERR_FAIL_INDEX_V(p_index, (int)scene_state.uniform_buffers.size(), RID()); RenderBufferDataForward *rb = nullptr; if (p_render_buffers.is_valid()) { @@ -2264,6 +2418,24 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff Vector<RD::Uniform> uniforms; { + RD::Uniform u; + u.binding = 0; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.ids.push_back(scene_state.uniform_buffers[p_index]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 1; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + RID instance_buffer = scene_state.instance_buffer[p_render_list]; + if (instance_buffer == RID()) { + instance_buffer = default_vec4_xform_buffer; // any buffer will do since its not used + } + u.ids.push_back(instance_buffer); + uniforms.push_back(u); + } + { RID radiance_texture; if (p_radiance_texture.is_valid()) { radiance_texture = p_radiance_texture; @@ -2271,7 +2443,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff radiance_texture = storage->texture_rd_get_default(is_using_radiance_cubemap_array() ? RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_BLACK : RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK); } RD::Uniform u; - u.binding = 0; + u.binding = 2; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.ids.push_back(radiance_texture); uniforms.push_back(u); @@ -2280,7 +2452,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff { RID ref_texture = p_reflection_atlas.is_valid() ? reflection_atlas_get_texture(p_reflection_atlas) : RID(); RD::Uniform u; - u.binding = 1; + u.binding = 3; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; if (ref_texture.is_valid()) { u.ids.push_back(ref_texture); @@ -2292,7 +2464,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff { RD::Uniform u; - u.binding = 2; + u.binding = 4; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID texture; if (p_shadow_atlas.is_valid()) { @@ -2306,7 +2478,18 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff } { RD::Uniform u; - u.binding = 3; + u.binding = 5; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + if (p_use_directional_shadow_atlas && directional_shadow_get_texture().is_valid()) { + u.ids.push_back(directional_shadow_get_texture()); + } else { + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE)); + } + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 6; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.ids.resize(scene_state.max_lightmaps); RID default_tex = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE); @@ -2325,7 +2508,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff } { RD::Uniform u; - u.binding = 4; + u.binding = 7; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.ids.resize(MAX_GI_PROBES); RID default_tex = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE); @@ -2346,7 +2529,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff { RD::Uniform u; - u.binding = 5; + u.binding = 8; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; RID cb = p_cluster_buffer.is_valid() ? p_cluster_buffer : default_vec4_xform_buffer; u.ids.push_back(cb); @@ -2355,7 +2538,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff { RD::Uniform u; - u.binding = 6; + u.binding = 9; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID texture = (false && rb && rb->depth.is_valid()) ? rb->depth : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE); u.ids.push_back(texture); @@ -2363,7 +2546,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff } { RD::Uniform u; - u.binding = 7; + u.binding = 10; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID bbt = rb ? render_buffers_get_back_buffer_texture(p_render_buffers) : RID(); RID texture = bbt.is_valid() ? bbt : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); @@ -2374,7 +2557,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff if (!low_end) { { RD::Uniform u; - u.binding = 8; + u.binding = 11; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID texture = rb && rb->normal_roughness_buffer.is_valid() ? rb->normal_roughness_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_NORMAL); u.ids.push_back(texture); @@ -2383,7 +2566,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff { RD::Uniform u; - u.binding = 9; + u.binding = 12; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID aot = rb ? render_buffers_get_ao_texture(p_render_buffers) : RID(); RID texture = aot.is_valid() ? aot : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); @@ -2393,24 +2576,26 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff { RD::Uniform u; - u.binding = 10; + u.binding = 13; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - RID texture = rb && rb->ambient_buffer.is_valid() ? rb->ambient_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); + RID ambient_buffer = p_render_buffers.is_valid() ? render_buffers_get_gi_ambient_texture(p_render_buffers) : RID(); + RID texture = ambient_buffer.is_valid() ? ambient_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); u.ids.push_back(texture); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 11; + u.binding = 14; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - RID texture = rb && rb->reflection_buffer.is_valid() ? rb->reflection_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); + RID reflection_buffer = p_render_buffers.is_valid() ? render_buffers_get_gi_reflection_texture(p_render_buffers) : RID(); + RID texture = reflection_buffer.is_valid() ? reflection_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); u.ids.push_back(texture); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 12; + u.binding = 15; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID t; if (rb && render_buffers_is_sdfgi_enabled(p_render_buffers)) { @@ -2423,7 +2608,7 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff } { RD::Uniform u; - u.binding = 13; + u.binding = 16; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; if (rb && render_buffers_is_sdfgi_enabled(p_render_buffers)) { u.ids.push_back(render_buffers_get_sdfgi_occlusion_texture(p_render_buffers)); @@ -2434,14 +2619,14 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff } { RD::Uniform u; - u.binding = 14; + u.binding = 17; u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; u.ids.push_back(rb ? render_buffers_get_gi_probe_buffer(p_render_buffers) : render_buffers_get_default_gi_probe_buffer()); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 15; + u.binding = 18; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID vfog = RID(); if (rb && render_buffers_has_volumetric_fog(p_render_buffers)) { @@ -2457,8 +2642,16 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff } } - render_pass_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, default_shader_rd, RENDER_PASS_UNIFORM_SET); - return render_pass_uniform_set; + if (p_index >= (int)render_pass_uniform_sets.size()) { + render_pass_uniform_sets.resize(p_index + 1); + } + + if (render_pass_uniform_sets[p_index].is_valid() && RD::get_singleton()->uniform_set_is_valid(render_pass_uniform_sets[p_index])) { + RD::get_singleton()->free(render_pass_uniform_sets[p_index]); + } + + render_pass_uniform_sets[p_index] = RD::get_singleton()->uniform_set_create(uniforms, default_shader_rd, RENDER_PASS_UNIFORM_SET); + return render_pass_uniform_sets[p_index]; } RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albedo_texture, RID p_emission_texture, RID p_emission_aniso_texture, RID p_geom_facing_texture) { @@ -2469,10 +2662,24 @@ RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albed Vector<RD::Uniform> uniforms; { + RD::Uniform u; + u.binding = 0; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.ids.push_back(scene_state.uniform_buffers[0]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.binding = 1; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.ids.push_back(scene_state.instance_buffer[RENDER_LIST_SECONDARY]); + uniforms.push_back(u); + } + { // No radiance texture. RID radiance_texture = storage->texture_rd_get_default(is_using_radiance_cubemap_array() ? RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_BLACK : RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK); RD::Uniform u; - u.binding = 0; + u.binding = 2; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.ids.push_back(radiance_texture); uniforms.push_back(u); @@ -2482,7 +2689,7 @@ RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albed // No reflection atlas. RID ref_texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_ARRAY_BLACK); RD::Uniform u; - u.binding = 1; + u.binding = 3; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.ids.push_back(ref_texture); uniforms.push_back(u); @@ -2491,7 +2698,17 @@ RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albed { // No shadow atlas. RD::Uniform u; - u.binding = 2; + u.binding = 4; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + RID texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE); + u.ids.push_back(texture); + uniforms.push_back(u); + } + + { + // No directional shadow atlas. + RD::Uniform u; + u.binding = 5; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE); u.ids.push_back(texture); @@ -2501,7 +2718,7 @@ RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albed { // No Lightmaps RD::Uniform u; - u.binding = 3; + u.binding = 6; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.ids.resize(scene_state.max_lightmaps); RID default_tex = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE); @@ -2515,7 +2732,7 @@ RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albed { // No GIProbes RD::Uniform u; - u.binding = 4; + u.binding = 7; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.ids.resize(MAX_GI_PROBES); RID default_tex = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE); @@ -2528,7 +2745,7 @@ RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albed { RD::Uniform u; - u.binding = 5; + u.binding = 8; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; RID cb = default_vec4_xform_buffer; u.ids.push_back(cb); @@ -2540,28 +2757,28 @@ RID RendererSceneRenderForward::_setup_sdfgi_render_pass_uniform_set(RID p_albed { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 6; + u.binding = 9; u.ids.push_back(p_albedo_texture); uniforms.push_back(u); } { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 7; + u.binding = 10; u.ids.push_back(p_emission_texture); uniforms.push_back(u); } { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 8; + u.binding = 11; u.ids.push_back(p_emission_aniso_texture); uniforms.push_back(u); } { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_IMAGE; - u.binding = 9; + u.binding = 12; u.ids.push_back(p_geom_facing_texture); uniforms.push_back(u); } @@ -2585,18 +2802,6 @@ RID RendererSceneRenderForward::_render_buffers_get_normal_texture(RID p_render_ return rb->normal_roughness_buffer; } -RID RendererSceneRenderForward::_render_buffers_get_ambient_texture(RID p_render_buffers) { - RenderBufferDataForward *rb = (RenderBufferDataForward *)render_buffers_get_data(p_render_buffers); - - return rb->ambient_buffer; -} - -RID RendererSceneRenderForward::_render_buffers_get_reflection_texture(RID p_render_buffers) { - RenderBufferDataForward *rb = (RenderBufferDataForward *)render_buffers_get_data(p_render_buffers); - - return rb->reflection_buffer; -} - RendererSceneRenderForward *RendererSceneRenderForward::singleton = nullptr; void RendererSceneRenderForward::set_time(double p_time, double p_step) { @@ -2666,10 +2871,17 @@ void RendererSceneRenderForward::_geometry_instance_add_surface_with_material(Ge } MaterialData *material_shadow = nullptr; - //void *surface_shadow = nullptr; + void *surface_shadow = nullptr; if (!p_material->shader_data->writes_modelview_or_projection && !p_material->shader_data->uses_vertex && !p_material->shader_data->uses_discard && !p_material->shader_data->uses_depth_pre_pass) { flags |= GeometryInstanceSurfaceDataCache::FLAG_USES_SHARED_SHADOW_MATERIAL; material_shadow = (MaterialData *)storage->material_get_data(default_material, RendererStorageRD::SHADER_TYPE_3D); + + RID shadow_mesh = storage->mesh_get_shadow_mesh(p_mesh); + + if (shadow_mesh.is_valid()) { + surface_shadow = storage->mesh_get_surface(shadow_mesh, p_surface); + } + } else { material_shadow = p_material; } @@ -2691,7 +2903,8 @@ void RendererSceneRenderForward::_geometry_instance_add_surface_with_material(Ge //shadow sdcache->shader_shadow = material_shadow->shader_data; sdcache->material_uniform_set_shadow = material_shadow->uniform_set; - sdcache->surface_shadow = sdcache->surface; //when adding special shadow meshes, will use this + + sdcache->surface_shadow = surface_shadow ? surface_shadow : sdcache->surface; sdcache->owner = ginstance; @@ -2703,10 +2916,11 @@ void RendererSceneRenderForward::_geometry_instance_add_surface_with_material(Ge sdcache->sort.sort_key1 = 0; sdcache->sort.sort_key2 = 0; - sdcache->sort.surface_type = ginstance->data->base_type; - sdcache->sort.material_id = p_material_id; + sdcache->sort.surface_index = p_surface; + sdcache->sort.material_id_low = p_material_id & 0x3FFF; + sdcache->sort.material_id_hi = p_material_id >> 14; sdcache->sort.shader_id = p_shader_id; - sdcache->sort.geometry_id = p_mesh.get_local_index(); + sdcache->sort.geometry_id = p_mesh.get_local_index(); //only meshes can repeat anyway sdcache->sort.uses_forward_gi = ginstance->can_sdfgi; sdcache->sort.priority = p_material->priority; } @@ -2836,11 +3050,6 @@ void RendererSceneRenderForward::_geometry_instance_update(GeometryInstance *p_g //Fill push constant - ginstance->push_constant.instance_uniforms_ofs = ginstance->data->shader_parameters_offset >= 0 ? ginstance->data->shader_parameters_offset : 0; - ginstance->push_constant.layer_mask = ginstance->data->layer_mask; - ginstance->push_constant.flags = 0; - ginstance->push_constant.gi_offset = 0xFFFFFFFF; //disabled - bool store_transform = true; if (ginstance->data->base_type == RS::INSTANCE_MULTIMESH) { @@ -2897,21 +3106,10 @@ void RendererSceneRenderForward::_geometry_instance_update(GeometryInstance *p_g } } - if (store_transform) { - RendererStorageRD::store_transform(ginstance->data->transform, ginstance->push_constant.transform); - } else { - RendererStorageRD::store_transform(Transform(), ginstance->push_constant.transform); - } - + ginstance->store_transform_cache = store_transform; ginstance->can_sdfgi = false; - if (lightmap_instance_is_valid(ginstance->lightmap_instance)) { - ginstance->push_constant.gi_offset = ginstance->data->lightmap_slice_index << 16; - ginstance->push_constant.lightmap_uv_scale[0] = ginstance->data->lightmap_uv_scale.position.x; - ginstance->push_constant.lightmap_uv_scale[1] = ginstance->data->lightmap_uv_scale.position.y; - ginstance->push_constant.lightmap_uv_scale[2] = ginstance->data->lightmap_uv_scale.size.width; - ginstance->push_constant.lightmap_uv_scale[3] = ginstance->data->lightmap_uv_scale.size.height; - } else if (!low_end) { + if (!lightmap_instance_is_valid(ginstance->lightmap_instance) && !low_end) { if (ginstance->gi_probes[0].is_null() && (ginstance->data->use_baked_light || ginstance->data->use_dynamic_gi)) { ginstance->can_sdfgi = true; } @@ -3001,8 +3199,7 @@ void RendererSceneRenderForward::geometry_instance_set_mesh_instance(GeometryIns void RendererSceneRenderForward::geometry_instance_set_transform(GeometryInstance *p_geometry_instance, const Transform &p_transform, const AABB &p_aabb, const AABB &p_transformed_aabb) { GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); ERR_FAIL_COND(!ginstance); - RendererStorageRD::store_transform(p_transform, ginstance->push_constant.transform); - ginstance->data->transform = p_transform; + ginstance->transform = p_transform; ginstance->mirror = p_transform.basis.determinant() < 0; ginstance->data->aabb = p_aabb; ginstance->transformed_aabb = p_transformed_aabb; @@ -3037,8 +3234,8 @@ void RendererSceneRenderForward::geometry_instance_set_use_lightmap(GeometryInst GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); ERR_FAIL_COND(!ginstance); ginstance->lightmap_instance = p_lightmap_instance; - ginstance->data->lightmap_uv_scale = p_lightmap_uv_scale; - ginstance->data->lightmap_slice_index = p_lightmap_slice_index; + ginstance->lightmap_uv_scale = p_lightmap_uv_scale; + ginstance->lightmap_slice_index = p_lightmap_slice_index; _geometry_instance_mark_dirty(ginstance); } void RendererSceneRenderForward::geometry_instance_set_lightmap_capture(GeometryInstance *p_geometry_instance, const Color *p_sh9) { @@ -3061,7 +3258,7 @@ void RendererSceneRenderForward::geometry_instance_set_lightmap_capture(Geometry void RendererSceneRenderForward::geometry_instance_set_instance_shader_parameters_offset(GeometryInstance *p_geometry_instance, int32_t p_offset) { GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); ERR_FAIL_COND(!ginstance); - ginstance->data->shader_parameters_offset = p_offset; + ginstance->shader_parameters_offset = p_offset; _geometry_instance_mark_dirty(ginstance); } void RendererSceneRenderForward::geometry_instance_set_cast_double_sided_shadows(GeometryInstance *p_geometry_instance, bool p_enable) { @@ -3075,8 +3272,7 @@ void RendererSceneRenderForward::geometry_instance_set_cast_double_sided_shadows void RendererSceneRenderForward::geometry_instance_set_layer_mask(GeometryInstance *p_geometry_instance, uint32_t p_layer_mask) { GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_geometry_instance); ERR_FAIL_COND(!ginstance); - ginstance->data->layer_mask = p_layer_mask; - ginstance->push_constant.layer_mask = p_layer_mask; + ginstance->layer_mask = p_layer_mask; } void RendererSceneRenderForward::geometry_instance_free(GeometryInstance *p_geometry_instance) { @@ -3108,7 +3304,7 @@ void RendererSceneRenderForward::geometry_instance_pair_decal_instances(Geometry Transform RendererSceneRenderForward::geometry_instance_get_transform(GeometryInstance *p_instance) { GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_instance); ERR_FAIL_COND_V(!ginstance, Transform()); - return ginstance->data->transform; + return ginstance->transform; } AABB RendererSceneRenderForward::geometry_instance_get_aabb(GeometryInstance *p_instance) { GeometryInstanceForward *ginstance = static_cast<GeometryInstanceForward *>(p_instance); @@ -3333,7 +3529,7 @@ RendererSceneRenderForward::RendererSceneRenderForward(RendererStorageRD *p_stor actions.render_mode_defines["cull_front"] = "#define DO_SIDE_CHECK\n"; actions.render_mode_defines["cull_disabled"] = "#define DO_SIDE_CHECK\n"; - bool force_lambert = GLOBAL_GET("rendering/quality/shading/force_lambert_over_burley"); + bool force_lambert = GLOBAL_GET("rendering/shading/overrides/force_lambert_over_burley"); if (!force_lambert) { actions.render_mode_defines["diffuse_burley"] = "#define DIFFUSE_BURLEY\n"; @@ -3345,7 +3541,7 @@ RendererSceneRenderForward::RendererSceneRenderForward(RendererStorageRD *p_stor actions.render_mode_defines["sss_mode_skin"] = "#define SSS_MODE_SKIN\n"; - bool force_blinn = GLOBAL_GET("rendering/quality/shading/force_blinn_over_ggx"); + bool force_blinn = GLOBAL_GET("rendering/shading/overrides/force_blinn_over_ggx"); if (!force_blinn) { actions.render_mode_defines["specular_schlick_ggx"] = "#define SPECULAR_SCHLICK_GGX\n"; @@ -3376,18 +3572,13 @@ RendererSceneRenderForward::RendererSceneRenderForward(RendererStorageRD *p_stor shader.compiler.initialize(actions); } - //render list - render_list.max_elements = GLOBAL_DEF_RST("rendering/limits/rendering/max_renderable_elements", (int)128000); - render_list.init(); - render_pass = 0; - - scene_state.uniform_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(SceneState::UBO)); - { //default material and shader - default_shader = storage->shader_create(); + default_shader = storage->shader_allocate(); + storage->shader_initialize(default_shader); storage->shader_set_code(default_shader, "shader_type spatial; void vertex() { ROUGHNESS = 0.8; } void fragment() { ALBEDO=vec3(0.6); ROUGHNESS=0.8; METALLIC=0.2; } \n"); - default_material = storage->material_create(); + default_material = storage->material_allocate(); + storage->material_initialize(default_material); storage->material_set_shader(default_material, default_shader); MaterialData *md = (MaterialData *)storage->material_get_data(default_material, RendererStorageRD::SHADER_TYPE_3D); @@ -3398,14 +3589,18 @@ RendererSceneRenderForward::RendererSceneRenderForward(RendererStorageRD *p_stor } { - overdraw_material_shader = storage->shader_create(); + overdraw_material_shader = storage->shader_allocate(); + storage->shader_initialize(overdraw_material_shader); storage->shader_set_code(overdraw_material_shader, "shader_type spatial;\nrender_mode blend_add,unshaded;\n void fragment() { ALBEDO=vec3(0.4,0.8,0.8); ALPHA=0.2; }"); - overdraw_material = storage->material_create(); + overdraw_material = storage->material_allocate(); + storage->material_initialize(overdraw_material); storage->material_set_shader(overdraw_material, overdraw_material_shader); - wireframe_material_shader = storage->shader_create(); + wireframe_material_shader = storage->shader_allocate(); + storage->shader_initialize(wireframe_material_shader); storage->shader_set_code(wireframe_material_shader, "shader_type spatial;\nrender_mode wireframe,unshaded;\n void fragment() { ALBEDO=vec3(0.0,0.0,0.0); }"); - wireframe_material = storage->material_create(); + wireframe_material = storage->material_allocate(); + storage->material_initialize(wireframe_material); storage->material_set_shader(wireframe_material, wireframe_material_shader); } @@ -3429,15 +3624,17 @@ RendererSceneRenderForward::RendererSceneRenderForward(RendererStorageRD *p_stor shadow_sampler = RD::get_singleton()->sampler_create(sampler); } - render_list_thread_threshold = GLOBAL_GET("rendering/forward_renderer/threaded_render_minimum_instances"); + render_list_thread_threshold = GLOBAL_GET("rendering/limits/forward_renderer/threaded_render_minimum_instances"); } RendererSceneRenderForward::~RendererSceneRenderForward() { directional_shadow_atlas_set_size(0); //clear base uniform set if still valid - if (render_pass_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(render_pass_uniform_set)) { - RD::get_singleton()->free(render_pass_uniform_set); + for (uint32_t i = 0; i < render_pass_uniform_sets.size(); i++) { + if (render_pass_uniform_sets[i].is_valid() && RD::get_singleton()->uniform_set_is_valid(render_pass_uniform_sets[i])) { + RD::get_singleton()->free(render_pass_uniform_sets[i]); + } } if (sdfgi_pass_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(sdfgi_pass_uniform_set)) { @@ -3456,9 +3653,16 @@ RendererSceneRenderForward::~RendererSceneRenderForward() { storage->free(default_material); { - RD::get_singleton()->free(scene_state.uniform_buffer); + for (uint32_t i = 0; i < scene_state.uniform_buffers.size(); i++) { + RD::get_singleton()->free(scene_state.uniform_buffers[i]); + } RD::get_singleton()->free(scene_state.lightmap_buffer); RD::get_singleton()->free(scene_state.lightmap_capture_buffer); + for (uint32_t i = 0; i < RENDER_LIST_MAX; i++) { + if (scene_state.instance_buffer[i] != RID()) { + RD::get_singleton()->free(scene_state.instance_buffer[i]); + } + } memdelete_arr(scene_state.lightmap_captures); } diff --git a/servers/rendering/renderer_rd/renderer_scene_render_forward.h b/servers/rendering/renderer_rd/renderer_scene_render_forward.h index d4a4c9a3a9..af78c50fda 100644 --- a/servers/rendering/renderer_rd/renderer_scene_render_forward.h +++ b/servers/rendering/renderer_rd/renderer_scene_render_forward.h @@ -50,6 +50,15 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { MAX_GI_PROBES = 8, MAX_LIGHTMAPS = 8, MAX_GI_PROBES_PER_INSTANCE = 2, + INSTANCE_DATA_BUFFER_MIN_SIZE = 4096 + }; + + enum RenderListType { + RENDER_LIST_OPAQUE, //used for opaque objects + RENDER_LIST_ALPHA, //used for transparent objects + RENDER_LIST_SECONDARY, //used for shadows and other objects + RENDER_LIST_MAX + }; /* Scene Shader */ @@ -213,9 +222,6 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { RID normal_roughness_buffer; RID giprobe_buffer; - RID ambient_buffer; - RID reflection_buffer; - RS::ViewportMSAA msaa; RD::TextureSamples texture_samples; @@ -236,7 +242,6 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { RID render_sdfgi_uniform_set; void ensure_specular(); - void ensure_gi(); void ensure_giprobe(); void clear(); virtual void configure(RID p_color_buffer, RID p_depth_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa); @@ -249,7 +254,7 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { RID shadow_sampler; RID render_base_uniform_set; - RID render_pass_uniform_set; + LocalVector<RID> render_pass_uniform_sets; RID sdfgi_pass_uniform_set; uint64_t lightmap_texture_array_version = 0xFFFFFFFF; @@ -258,12 +263,61 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { void _render_buffers_clear_uniform_set(RenderBufferDataForward *rb); virtual void _render_buffers_uniform_set_changed(RID p_render_buffers); virtual RID _render_buffers_get_normal_texture(RID p_render_buffers); - virtual RID _render_buffers_get_ambient_texture(RID p_render_buffers); - virtual RID _render_buffers_get_reflection_texture(RID p_render_buffers); void _update_render_base_uniform_set(); RID _setup_sdfgi_render_pass_uniform_set(RID p_albedo_texture, RID p_emission_texture, RID p_emission_aniso_texture, RID p_geom_facing_texture); - RID _setup_render_pass_uniform_set(RID p_render_buffers, RID p_radiance_texture, RID p_shadow_atlas, RID p_reflection_atlas, RID p_cluster_buffer, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps); + RID _setup_render_pass_uniform_set(RenderListType p_render_list, RID p_render_buffers, RID p_radiance_texture, RID p_shadow_atlas, RID p_reflection_atlas, RID p_cluster_buffer, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, bool p_use_directional_shadow_atlas = false, int p_index = 0); + + enum PassMode { + PASS_MODE_COLOR, + PASS_MODE_COLOR_SPECULAR, + PASS_MODE_COLOR_TRANSPARENT, + PASS_MODE_SHADOW, + PASS_MODE_SHADOW_DP, + PASS_MODE_DEPTH, + PASS_MODE_DEPTH_NORMAL_ROUGHNESS, + PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE, + PASS_MODE_DEPTH_MATERIAL, + PASS_MODE_SDF, + }; + + struct GeometryInstanceSurfaceDataCache; + struct RenderElementInfo; + + struct RenderListParameters { + GeometryInstanceSurfaceDataCache **elements = nullptr; + RenderElementInfo *element_info = nullptr; + int element_count = 0; + bool reverse_cull = false; + PassMode pass_mode = PASS_MODE_COLOR; + bool no_gi = false; + RID render_pass_uniform_set; + bool force_wireframe = false; + Vector2 uv_offset; + Plane lod_plane; + float lod_distance_multiplier = 0.0; + float screen_lod_threshold = 0.0; + RD::FramebufferFormatID framebuffer_format = 0; + uint32_t element_offset = 0; + uint32_t barrier = RD::BARRIER_MASK_ALL; + + RenderListParameters(GeometryInstanceSurfaceDataCache **p_elements, RenderElementInfo *p_element_info, int p_element_count, bool p_reverse_cull, PassMode p_pass_mode, bool p_no_gi, RID p_render_pass_uniform_set, bool p_force_wireframe = false, const Vector2 &p_uv_offset = Vector2(), const Plane &p_lod_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, uint32_t p_element_offset = 0, uint32_t p_barrier = RD::BARRIER_MASK_ALL) { + elements = p_elements; + element_info = p_element_info; + element_count = p_element_count; + reverse_cull = p_reverse_cull; + pass_mode = p_pass_mode; + no_gi = p_no_gi; + render_pass_uniform_set = p_render_pass_uniform_set; + force_wireframe = p_force_wireframe; + uv_offset = p_uv_offset; + lod_plane = p_lod_plane; + lod_distance_multiplier = p_lod_distance_multiplier; + screen_lod_threshold = p_screen_lod_threshold; + element_offset = p_element_offset; + barrier = p_barrier; + } + }; struct LightmapData { float normal_xform[12]; @@ -373,9 +427,24 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { uint32_t pancake_shadows; }; + struct PushConstant { + uint32_t base_index; // + uint32_t uv_offset; //packed + uint32_t pad[2]; + }; + + struct InstanceData { + float transform[16]; + uint32_t flags; + uint32_t instance_uniforms_ofs; //base offset in global buffer for instance variables + uint32_t gi_offset; //GI information when using lightmapping (VCT or lightmap index) + uint32_t layer_mask; + float lightmap_uv_scale[4]; + }; + UBO ubo; - RID uniform_buffer; + LocalVector<RID> uniform_buffers; LightmapData lightmaps[MAX_LIGHTMAPS]; RID lightmap_ids[MAX_LIGHTMAPS]; @@ -384,6 +453,10 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { uint32_t max_lightmaps; RID lightmap_buffer; + RID instance_buffer[RENDER_LIST_MAX]; + uint32_t instance_buffer_size[RENDER_LIST_MAX] = { 0, 0, 0 }; + LocalVector<InstanceData> instance_data[RENDER_LIST_MAX]; + LightmapCaptureData *lightmap_captures; uint32_t max_lightmap_captures; RID lightmap_capture_buffer; @@ -396,10 +469,29 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { bool used_depth_texture = false; bool used_sss = false; + struct ShadowPass { + uint32_t element_from; + uint32_t element_count; + bool flip_cull; + PassMode pass_mode; + + RID rp_uniform_set; + Plane camera_plane; + float lod_distance_multiplier; + float screen_lod_threshold; + + RID framebuffer; + RD::InitialAction initial_depth_action; + RD::FinalAction final_depth_action; + Rect2i rect; + }; + + LocalVector<ShadowPass> shadow_passes; + } scene_state; static RendererSceneRenderForward *singleton; - uint64_t render_pass; + double time; RID default_shader; RID default_material; @@ -413,51 +505,15 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { RID default_vec4_xform_buffer; RID default_vec4_xform_uniform_set; - enum PassMode { - PASS_MODE_COLOR, - PASS_MODE_COLOR_SPECULAR, - PASS_MODE_COLOR_TRANSPARENT, - PASS_MODE_SHADOW, - PASS_MODE_SHADOW_DP, - PASS_MODE_DEPTH, - PASS_MODE_DEPTH_NORMAL_ROUGHNESS, - PASS_MODE_DEPTH_NORMAL_ROUGHNESS_GIPROBE, - PASS_MODE_DEPTH_MATERIAL, - PASS_MODE_SDF, - }; - - void _setup_environment(RID p_environment, RID p_render_buffers, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_reflection_probe, bool p_no_fog, const Size2i &p_screen_size, uint32_t p_cluster_size, uint32_t p_max_cluster_elements, RID p_shadow_atlas, bool p_flip_y, const Color &p_default_bg_color, float p_znear, float p_zfar, bool p_opaque_render_buffers = false, bool p_pancake_shadows = false); + void _setup_environment(RID p_environment, RID p_render_buffers, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_reflection_probe, bool p_no_fog, const Size2i &p_screen_size, uint32_t p_cluster_size, uint32_t p_max_cluster_elements, RID p_shadow_atlas, bool p_flip_y, const Color &p_default_bg_color, float p_znear, float p_zfar, bool p_opaque_render_buffers = false, bool p_pancake_shadows = false, int p_index = 0); void _setup_giprobes(const PagedArray<RID> &p_giprobes); void _setup_lightmaps(const PagedArray<RID> &p_lightmaps, const Transform &p_cam_transform); - struct GeometryInstanceSurfaceDataCache; - - struct RenderListParameters { - GeometryInstanceSurfaceDataCache **elements = nullptr; - int element_count = 0; - bool reverse_cull = false; - PassMode pass_mode = PASS_MODE_COLOR; - bool no_gi = false; - RID render_pass_uniform_set; - bool force_wireframe = false; - Vector2 uv_offset; - Plane lod_plane; - float lod_distance_multiplier = 0.0; - float screen_lod_threshold = 0.0; - RD::FramebufferFormatID framebuffer_format = 0; - RenderListParameters(GeometryInstanceSurfaceDataCache **p_elements, int p_element_count, bool p_reverse_cull, PassMode p_pass_mode, bool p_no_gi, RID p_render_pass_uniform_set, bool p_force_wireframe = false, const Vector2 &p_uv_offset = Vector2(), const Plane &p_lod_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0) { - elements = p_elements; - element_count = p_element_count; - reverse_cull = p_reverse_cull; - pass_mode = p_pass_mode; - no_gi = p_no_gi; - render_pass_uniform_set = p_render_pass_uniform_set; - force_wireframe = p_force_wireframe; - uv_offset = p_uv_offset; - lod_plane = p_lod_plane; - lod_distance_multiplier = p_lod_distance_multiplier; - screen_lod_threshold = p_screen_lod_threshold; - } + struct RenderElementInfo { + uint32_t repeat : 22; + uint32_t uses_lightmap : 1; + uint32_t uses_forward_gi : 1; + uint32_t lod_index : 8; }; template <PassMode p_pass_mode> @@ -471,7 +527,9 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { uint32_t render_list_thread_threshold = 500; - void _fill_render_list(const PagedArray<GeometryInstance *> &p_instances, PassMode p_pass_mode, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, bool p_using_sdfgi = false, bool p_using_opaque_gi = false); + void _update_instance_data_buffer(RenderListType p_render_list); + void _fill_instance_data(RenderListType p_render_list, uint32_t p_offset = 0, int32_t p_max_elements = -1, bool p_update_buffer = true); + void _fill_render_list(RenderListType p_render_list, const PagedArray<GeometryInstance *> &p_instances, PassMode p_pass_mode, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, bool p_using_sdfgi = false, bool p_using_opaque_gi = false, const Plane &p_lod_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, bool p_append = false); Map<Size2i, RID> sdfgi_framebuffer_size_cache; @@ -499,14 +557,17 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { union { struct { - uint32_t geometry_id; - uint32_t material_id; - uint32_t shader_id; - uint32_t surface_type : 4; - uint32_t uses_forward_gi : 1; //set during addition - uint32_t uses_lightmap : 1; //set during addition - uint32_t depth_layer : 4; //set during addition - uint32_t priority : 8; + uint64_t lod_index : 8; + uint64_t surface_index : 10; + uint64_t geometry_id : 32; + uint64_t material_id_low : 14; + + uint64_t material_id_hi : 18; + uint64_t shader_id : 32; + uint64_t uses_forward_gi : 1; + uint64_t uses_lightmap : 1; + uint64_t depth_layer : 4; + uint64_t priority : 8; }; struct { uint64_t sort_key1; @@ -538,20 +599,20 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { float lod_model_scale = 1.0; AABB transformed_aabb; //needed for LOD float depth = 0; - struct PushConstant { - float transform[16]; - uint32_t flags; - uint32_t instance_uniforms_ofs; //base offset in global buffer for instance variables - uint32_t gi_offset; //GI information when using lightmapping (VCT or lightmap index) - uint32_t layer_mask; - float lightmap_uv_scale[4]; - } push_constant; + uint32_t gi_offset_cache = 0; + uint32_t flags_cache = 0; + bool store_transform_cache = true; + int32_t shader_parameters_offset = -1; + uint32_t lightmap_slice_index; + Rect2 lightmap_uv_scale; + uint32_t layer_mask = 1; RID transforms_uniform_set; uint32_t instance_count = 0; RID mesh_instance; bool can_sdfgi = false; //used during setup uint32_t base_flags = 0; + Transform transform; RID gi_probes[MAX_GI_PROBES_PER_INSTANCE]; RID lightmap_instance; GeometryInstanceLightmapSH *lightmap_sh = nullptr; @@ -564,21 +625,14 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { RS::InstanceType base_type; RID skeleton; - - uint32_t layer_mask = 1; - Vector<RID> surface_materials; RID material_override; - Transform transform; AABB aabb; - int32_t shader_parameters_offset = -1; bool use_dynamic_gi = false; bool use_baked_light = false; bool cast_double_sided_shaodows = false; bool mirror = false; - Rect2 lightmap_uv_scale; - uint32_t lightmap_slice_index = 0; bool dirty_dependencies = false; RendererStorage::DependencyTracker dependency_tracker; @@ -610,16 +664,12 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { /* Render List */ struct RenderList { - int max_elements; - - GeometryInstanceSurfaceDataCache **elements = nullptr; - - int element_count; - int alpha_element_count; + LocalVector<GeometryInstanceSurfaceDataCache *> elements; + LocalVector<RenderElementInfo> element_info; void clear() { - element_count = 0; - alpha_element_count = 0; + elements.clear(); + element_info.clear(); } //should eventually be replaced by radix @@ -630,13 +680,14 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { } }; - void sort_by_key(bool p_alpha) { + void sort_by_key() { SortArray<GeometryInstanceSurfaceDataCache *, SortByKey> sorter; - if (p_alpha) { - sorter.sort(&elements[max_elements - alpha_element_count], alpha_element_count); - } else { - sorter.sort(elements, element_count); - } + sorter.sort(elements.ptr(), elements.size()); + } + + void sort_by_key_range(uint32_t p_from, uint32_t p_size) { + SortArray<GeometryInstanceSurfaceDataCache *, SortByKey> sorter; + sorter.sort(elements.ptr() + p_from, p_size); } struct SortByDepth { @@ -645,14 +696,10 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { } }; - void sort_by_depth(bool p_alpha) { //used for shadows + void sort_by_depth() { //used for shadows SortArray<GeometryInstanceSurfaceDataCache *, SortByDepth> sorter; - if (p_alpha) { - sorter.sort(&elements[max_elements - alpha_element_count], alpha_element_count); - } else { - sorter.sort(elements, element_count); - } + sorter.sort(elements.ptr(), elements.size()); } struct SortByReverseDepthAndPriority { @@ -664,50 +711,24 @@ class RendererSceneRenderForward : public RendererSceneRenderRD { void sort_by_reverse_depth_and_priority(bool p_alpha) { //used for alpha SortArray<GeometryInstanceSurfaceDataCache *, SortByReverseDepthAndPriority> sorter; - if (p_alpha) { - sorter.sort(&elements[max_elements - alpha_element_count], alpha_element_count); - } else { - sorter.sort(elements, element_count); - } + sorter.sort(elements.ptr(), elements.size()); } _FORCE_INLINE_ void add_element(GeometryInstanceSurfaceDataCache *p_element) { - if (element_count + alpha_element_count >= max_elements) { - return; - } - elements[element_count] = p_element; - element_count++; - } - - _FORCE_INLINE_ void add_alpha_element(GeometryInstanceSurfaceDataCache *p_element) { - if (element_count + alpha_element_count >= max_elements) { - return; - } - int idx = max_elements - alpha_element_count - 1; - elements[idx] = p_element; - alpha_element_count++; - } - - void init() { - element_count = 0; - alpha_element_count = 0; - elements = memnew_arr(GeometryInstanceSurfaceDataCache *, max_elements); - } - - RenderList() { - max_elements = 0; - } - - ~RenderList() { - memdelete_arr(elements); + elements.push_back(p_element); } }; - RenderList render_list; + RenderList render_list[RENDER_LIST_MAX]; protected: - virtual void _render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, int p_directional_light_count, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_cluster_buffer, uint32_t p_cluster_size, uint32_t p_max_cluster_elements, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_bg_color, float p_lod_threshold); - virtual void _render_shadow(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0); + virtual void _render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_cluster_buffer, uint32_t p_cluster_size, uint32_t p_max_cluster_elements, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_bg_color, float p_lod_threshold); + + virtual void _render_shadow_begin(); + virtual void _render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, const Rect2i &p_rect = Rect2i(), bool p_flip_y = false, bool p_clear_region = true, bool p_begin = true, bool p_end = true); + virtual void _render_shadow_process(); + virtual void _render_shadow_end(uint32_t p_barrier = RD::BARRIER_MASK_ALL); + virtual void _render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region); virtual void _render_uv2(const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region); virtual void _render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<GeometryInstance *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture); diff --git a/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp b/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp index 2e457c2ce6..15e963f6e4 100644 --- a/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp @@ -183,13 +183,11 @@ void RendererSceneRenderRD::_create_reflection_importance_sample(ReflectionData void RendererSceneRenderRD::_update_reflection_mipmaps(ReflectionData &rd, int p_start, int p_end) { for (int i = p_start; i < p_end; i++) { - for (int j = 0; j < rd.layers[i].mipmaps.size() - 1; j++) { - for (int k = 0; k < 6; k++) { - RID view = rd.layers[i].mipmaps[j].views[k]; - RID texture = rd.layers[i].mipmaps[j + 1].views[k]; - Size2i size = rd.layers[i].mipmaps[j + 1].size; - storage->get_effects()->make_mipmap(view, texture, size); - } + for (int j = 0; j < rd.layers[i].views.size() - 1; j++) { + RID view = rd.layers[i].views[j]; + RID texture = rd.layers[i].views[j + 1]; + Size2i size = rd.layers[i].mipmaps[j + 1].size; + storage->get_effects()->cubemap_downsample(view, texture, size); } } } @@ -669,6 +667,13 @@ void RendererSceneRenderRD::sdfgi_update(RID p_render_buffers, RID p_environment u.ids.push_back(rb->sdfgi->lightprobe_texture); uniforms.push_back(u); } + { + RD::Uniform u; + u.binding = 11; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.ids.push_back(rb->sdfgi->occlusion_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); } @@ -951,7 +956,7 @@ void RendererSceneRenderRD::sdfgi_update(RID p_render_buffers, RID p_environment 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->bounce_feedback = env->sdfgi_bounce_feedback; sdfgi->energy = env->sdfgi_energy; sdfgi->normal_bias = env->sdfgi_normal_bias; sdfgi->probe_bias = env->sdfgi_probe_bias; @@ -964,7 +969,7 @@ void RendererSceneRenderRD::sdfgi_update(RID p_render_buffers, RID p_environment //check for updates - sdfgi->uses_multibounce = env->sdfgi_use_multibounce; + sdfgi->bounce_feedback = env->sdfgi_bounce_feedback; sdfgi->energy = env->sdfgi_energy; sdfgi->normal_bias = env->sdfgi_normal_bias; sdfgi->probe_bias = env->sdfgi_probe_bias; @@ -1150,150 +1155,72 @@ void RendererSceneRenderRD::_sdfgi_update_cascades(RID p_render_buffers) { 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); + RD::get_singleton()->buffer_update(rb->sdfgi->cascades_ubo, 0, sizeof(SDFGI::Cascade::UBO) * SDFGI::MAX_CASCADES, cascade_data, RD::BARRIER_MASK_COMPUTE); } -void RendererSceneRenderRD::sdfgi_update_probes(RID p_render_buffers, RID p_environment, const Vector<RID> &p_directional_lights, const RID *p_positional_light_instances, uint32_t p_positional_light_count) { +void RendererSceneRenderRD::_sdfgi_update_light(RID p_render_buffers, RID p_environment) { RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); ERR_FAIL_COND(rb == nullptr); if (rb->sdfgi == nullptr) { return; } - Environment *env = environment_owner.getornull(p_environment); - - RENDER_TIMESTAMP(">SDFGI Update Probes"); - - /* Update Cascades UBO */ - _sdfgi_update_cascades(p_render_buffers); - /* Update Dynamic Lights Buffer */ - - RENDER_TIMESTAMP("Update Lights"); - - /* Update dynamic lights */ - - { - int32_t cascade_light_count[SDFGI::MAX_CASCADES]; - - for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) { - SDFGI::Cascade &cascade = rb->sdfgi->cascades[i]; - - SDGIShader::Light lights[SDFGI::MAX_DYNAMIC_LIGHTS]; - uint32_t idx = 0; - for (uint32_t j = 0; j < (uint32_t)p_directional_lights.size(); j++) { - if (idx == SDFGI::MAX_DYNAMIC_LIGHTS) { - break; - } - - LightInstance *li = light_instance_owner.getornull(p_directional_lights[j]); - ERR_CONTINUE(!li); - - if (storage->light_directional_is_sky_only(li->light)) { - continue; - } - - Vector3 dir = -li->transform.basis.get_axis(Vector3::AXIS_Z); - dir.y *= rb->sdfgi->y_mult; - dir.normalize(); - lights[idx].direction[0] = dir.x; - lights[idx].direction[1] = dir.y; - lights[idx].direction[2] = dir.z; - Color color = storage->light_get_color(li->light); - color = color.to_linear(); - lights[idx].color[0] = color.r; - lights[idx].color[1] = color.g; - lights[idx].color[2] = color.b; - lights[idx].type = RS::LIGHT_DIRECTIONAL; - lights[idx].energy = storage->light_get_param(li->light, RS::LIGHT_PARAM_ENERGY); - lights[idx].has_shadow = storage->light_has_shadow(li->light); - idx++; - } - - AABB cascade_aabb; - cascade_aabb.position = Vector3((Vector3i(1, 1, 1) * -int32_t(rb->sdfgi->cascade_size >> 1) + cascade.position)) * cascade.cell_size; - cascade_aabb.size = Vector3(1, 1, 1) * rb->sdfgi->cascade_size * cascade.cell_size; + RD::get_singleton()->draw_command_begin_label("SDFGI Update dynamic Light"); - for (uint32_t j = 0; j < p_positional_light_count; j++) { - if (idx == SDFGI::MAX_DYNAMIC_LIGHTS) { - break; - } + /* Update dynamic light */ - LightInstance *li = light_instance_owner.getornull(p_positional_light_instances[j]); - ERR_CONTINUE(!li); + 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]); - uint32_t max_sdfgi_cascade = storage->light_get_max_sdfgi_cascade(li->light); - if (i > max_sdfgi_cascade) { - continue; - } + SDGIShader::DirectLightPushConstant push_constant; - if (!cascade_aabb.intersects(li->aabb)) { - continue; - } + 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.bounce_feedback = rb->sdfgi->bounce_feedback; + push_constant.y_mult = rb->sdfgi->y_mult; + push_constant.use_occlusion = rb->sdfgi->uses_occlusion; - 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); + for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) { + SDFGI::Cascade &cascade = rb->sdfgi->cascades[i]; + push_constant.light_count = rb->sdfgi->cascade_dynamic_light_count[i]; + push_constant.cascade = i; - idx++; - } + if (rb->sdfgi->cascades[i].all_dynamic_lights_dirty || sdfgi_frames_to_update_light == RS::ENV_SDFGI_UPDATE_LIGHT_IN_1_FRAME) { + push_constant.process_offset = 0; + push_constant.process_increment = 1; + } else { + static uint32_t frames_to_update_table[RS::ENV_SDFGI_UPDATE_LIGHT_MAX] = { + 1, 2, 4, 8, 16 + }; - if (idx > 0) { - RD::get_singleton()->buffer_update(cascade.lights_buffer, 0, idx * sizeof(SDGIShader::Light), lights, true); - } + uint32_t frames_to_update = frames_to_update_table[sdfgi_frames_to_update_light]; - cascade_light_count[i] = idx; + push_constant.process_offset = RSG::rasterizer->get_frame_number() % frames_to_update; + push_constant.process_increment = frames_to_update; } + rb->sdfgi->cascades[i].all_dynamic_lights_dirty = false; - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.direct_light_pipeline[SDGIShader::DIRECT_LIGHT_MODE_DYNAMIC]); - - SDGIShader::DirectLightPushConstant push_constant; - - push_constant.grid_size[0] = rb->sdfgi->cascade_size; - push_constant.grid_size[1] = rb->sdfgi->cascade_size; - push_constant.grid_size[2] = rb->sdfgi->cascade_size; - push_constant.max_cascades = rb->sdfgi->cascades.size(); - push_constant.probe_axis_size = rb->sdfgi->probe_axis_count; - push_constant.multibounce = rb->sdfgi->uses_multibounce; - push_constant.y_mult = rb->sdfgi->y_mult; - - push_constant.process_offset = 0; - push_constant.process_increment = 1; - - for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) { - SDFGI::Cascade &cascade = rb->sdfgi->cascades[i]; - push_constant.light_count = cascade_light_count[i]; - push_constant.cascade = i; + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, cascade.sdf_direct_light_uniform_set, 0); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::DirectLightPushConstant)); + RD::get_singleton()->compute_list_dispatch_indirect(compute_list, cascade.solid_cell_dispatch_buffer, 0); + } + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_COMPUTE); + RD::get_singleton()->draw_command_end_label(); +} - 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(); +void RendererSceneRenderRD::_sdfgi_update_probes(RID p_render_buffers, RID p_environment) { + RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); + ERR_FAIL_COND(rb == nullptr); + if (rb->sdfgi == nullptr) { + return; } - RENDER_TIMESTAMP("Raytrace"); + RD::get_singleton()->draw_command_begin_label("SDFGI Update Probes"); + + Environment *env = environment_owner.getornull(p_environment); SDGIShader::IntegratePushConstant push_constant; push_constant.grid_size[1] = rb->sdfgi->cascade_size; @@ -1303,7 +1230,7 @@ void RendererSceneRenderRD::sdfgi_update_probes(RID p_render_buffers, RID p_envi 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 }; + static const uint32_t ray_count[RS::ENV_SDFGI_RAY_COUNT_MAX] = { 4, 8, 16, 32, 64, 96, 128 }; push_constant.ray_count = ray_count[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; @@ -1362,7 +1289,7 @@ void RendererSceneRenderRD::sdfgi_update_probes(RID p_render_buffers, RID p_envi rb->sdfgi->render_pass++; - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(true); 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; @@ -1376,14 +1303,47 @@ void RendererSceneRenderRD::sdfgi_update_probes(RID p_render_buffers, RID p_envi 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_dispatch_threads(compute_list, rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count, rb->sdfgi->probe_axis_count, 1); + } + + //end later after raster to avoid barriering on layout changes + //RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_NO_BARRIER); + + RD::get_singleton()->draw_command_end_label(); +} + +void RendererSceneRenderRD::_sdfgi_store_probes(RID p_render_buffers) { + RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); + ERR_FAIL_COND(rb == nullptr); + if (rb->sdfgi == nullptr) { + return; } - RD::get_singleton()->compute_list_add_barrier(compute_list); //wait until done + RD::get_singleton()->barrier(RD::BARRIER_MASK_COMPUTE, RD::BARRIER_MASK_COMPUTE); + RD::get_singleton()->draw_command_begin_label("SDFGI Store Probes"); + + 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] = { 4, 8, 16, 32, 64, 96, 128 }; + push_constant.ray_count = ray_count[sdfgi_ray_count]; + push_constant.ray_bias = rb->sdfgi->probe_bias; + push_constant.image_size[0] = rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count; + push_constant.image_size[1] = rb->sdfgi->probe_axis_count; + push_constant.store_ambient_texture = false; + + push_constant.sky_mode = 0; + push_constant.y_mult = rb->sdfgi->y_mult; // Then store values into the lightprobe texture. Separating these steps has a small performance hit, but it allows for multiple bounces RENDER_TIMESTAMP("Average Probes"); + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.integrate_pipeline[SDGIShader::INTEGRATE_MODE_STORE]); //convert to octahedral to store @@ -1393,20 +1353,22 @@ void RendererSceneRenderRD::sdfgi_update_probes(RID p_render_buffers, RID p_envi 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_bind_uniform_set(compute_list, sdfgi_shader.integrate_default_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 * SDFGI::LIGHTPROBE_OCT_SIZE, rb->sdfgi->probe_axis_count * SDFGI::LIGHTPROBE_OCT_SIZE, 1, 8, 8, 1); + 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); } - RD::get_singleton()->compute_list_end(); + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_COMPUTE); - RENDER_TIMESTAMP("<SDFGI Update Probes"); + RD::get_singleton()->draw_command_end_label(); } - void RendererSceneRenderRD::_setup_giprobes(RID p_render_buffers, const Transform &p_transform, const PagedArray<RID> &p_gi_probes, uint32_t &r_gi_probes_used) { r_gi_probes_used = 0; RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); ERR_FAIL_COND(rb == nullptr); + RD::get_singleton()->draw_command_begin_label("GIProbes Setup"); + RID gi_probe_buffer = render_buffers_get_gi_probe_buffer(p_render_buffers); GI::GIProbeData gi_probe_data[RenderBuffers::MAX_GIPROBES]; @@ -1490,58 +1452,25 @@ void RendererSceneRenderRD::_setup_giprobes(RID p_render_buffers, const Transfor } if (p_gi_probes.size() > 0) { - RD::get_singleton()->buffer_update(gi_probe_buffer, 0, sizeof(GI::GIProbeData) * MIN((uint64_t)RenderBuffers::MAX_GIPROBES, p_gi_probes.size()), gi_probe_data, true); + RD::get_singleton()->buffer_update(gi_probe_buffer, 0, sizeof(GI::GIProbeData) * MIN((uint64_t)RenderBuffers::MAX_GIPROBES, p_gi_probes.size()), gi_probe_data, RD::BARRIER_MASK_COMPUTE); } -} -void RendererSceneRenderRD::_process_gi(RID p_render_buffers, RID p_normal_roughness_buffer, RID p_ambient_buffer, RID p_reflection_buffer, RID p_gi_probe_buffer, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform, const PagedArray<RID> &p_gi_probes) { - RENDER_TIMESTAMP("Render GI"); + RD::get_singleton()->draw_command_end_label(); +} +void RendererSceneRenderRD::_pre_process_gi(RID p_render_buffers, const Transform &p_transform) { + // Do the required buffer transfers and setup before the depth-pre pass, this way GI can + // run in parallel during depth-pre pass and shadow rendering. RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); ERR_FAIL_COND(rb == nullptr); - Environment *env = environment_owner.getornull(p_environment); - - GI::PushConstant push_constant; - - push_constant.screen_size[0] = rb->width; - push_constant.screen_size[1] = rb->height; - push_constant.z_near = p_projection.get_z_near(); - push_constant.z_far = p_projection.get_z_far(); - push_constant.orthogonal = p_projection.is_orthogonal(); - push_constant.proj_info[0] = -2.0f / (rb->width * p_projection.matrix[0][0]); - push_constant.proj_info[1] = -2.0f / (rb->height * p_projection.matrix[1][1]); - push_constant.proj_info[2] = (1.0f - p_projection.matrix[0][2]) / p_projection.matrix[0][0]; - push_constant.proj_info[3] = (1.0f + p_projection.matrix[1][2]) / p_projection.matrix[1][1]; - push_constant.max_giprobes = MIN((uint64_t)RenderBuffers::MAX_GIPROBES, p_gi_probes.size()); - push_constant.high_quality_vct = gi_probe_quality == RS::GI_PROBE_QUALITY_HIGH; - bool use_sdfgi = rb->sdfgi != nullptr; - bool use_giprobes = push_constant.max_giprobes > 0; + /* Update Cascades UBO */ - 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; - } + if (rb->sdfgi) { + /* Update general SDFGI Buffer */ - 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; + _sdfgi_update_cascades(p_render_buffers); - if (rb->sdfgi) { GI::SDFGIData sdfgi_data; sdfgi_data.grid_size[0] = rb->sdfgi->cascade_size; @@ -1608,8 +1537,171 @@ void RendererSceneRenderRD::_process_gi(RID p_render_buffers, RID p_normal_rough 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); + RD::get_singleton()->buffer_update(gi.sdfgi_ubo, 0, sizeof(GI::SDFGIData), &sdfgi_data, RD::BARRIER_MASK_COMPUTE); + + /* Update dynamic lights in SDFGI cascades */ + + for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) { + SDFGI::Cascade &cascade = rb->sdfgi->cascades[i]; + + SDGIShader::Light lights[SDFGI::MAX_DYNAMIC_LIGHTS]; + uint32_t idx = 0; + for (uint32_t j = 0; j < (uint32_t)render_state.sdfgi_update_data->directional_lights->size(); j++) { + if (idx == SDFGI::MAX_DYNAMIC_LIGHTS) { + break; + } + + LightInstance *li = light_instance_owner.getornull(render_state.sdfgi_update_data->directional_lights->get(j)); + ERR_CONTINUE(!li); + + if (storage->light_directional_is_sky_only(li->light)) { + continue; + } + + Vector3 dir = -li->transform.basis.get_axis(Vector3::AXIS_Z); + dir.y *= 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 < render_state.sdfgi_update_data->positional_light_count; j++) { + if (idx == SDFGI::MAX_DYNAMIC_LIGHTS) { + break; + } + + LightInstance *li = light_instance_owner.getornull(render_state.sdfgi_update_data->positional_light_instances[j]); + ERR_CONTINUE(!li); + + uint32_t max_sdfgi_cascade = storage->light_get_max_sdfgi_cascade(li->light); + if (i > max_sdfgi_cascade) { + continue; + } + + if (!cascade_aabb.intersects(li->aabb)) { + continue; + } + + Vector3 dir = -li->transform.basis.get_axis(Vector3::AXIS_Z); + //faster to not do this here + //dir.y *= 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].cos_spot_angle = Math::cos(Math::deg2rad(storage->light_get_param(li->light, RS::LIGHT_PARAM_SPOT_ANGLE))); + lights[idx].inv_spot_attenuation = 1.0f / storage->light_get_param(li->light, RS::LIGHT_PARAM_SPOT_ATTENUATION); + + idx++; + } + + if (idx > 0) { + RD::get_singleton()->buffer_update(cascade.lights_buffer, 0, idx * sizeof(SDGIShader::Light), lights, RD::BARRIER_MASK_COMPUTE); + } + + rb->sdfgi->cascade_dynamic_light_count[i] = idx; + } } +} + +void RendererSceneRenderRD::_process_gi(RID p_render_buffers, RID p_normal_roughness_buffer, RID p_gi_probe_buffer, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform, const PagedArray<RID> &p_gi_probes) { + RD::get_singleton()->draw_command_begin_label("GI Render"); + + RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); + ERR_FAIL_COND(rb == nullptr); + Environment *env = environment_owner.getornull(p_environment); + + if (rb->ambient_buffer.is_null() || rb->using_half_size_gi != gi.half_resolution) { + if (rb->ambient_buffer.is_valid()) { + RD::get_singleton()->free(rb->ambient_buffer); + RD::get_singleton()->free(rb->reflection_buffer); + } + + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.width = rb->width; + tf.height = rb->height; + if (gi.half_resolution) { + tf.width >>= 1; + tf.height >>= 1; + } + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + rb->reflection_buffer = RD::get_singleton()->texture_create(tf, RD::TextureView()); + rb->ambient_buffer = RD::get_singleton()->texture_create(tf, RD::TextureView()); + rb->using_half_size_gi = gi.half_resolution; + + _render_buffers_uniform_set_changed(p_render_buffers); + } + + GI::PushConstant push_constant; + + push_constant.screen_size[0] = rb->width; + push_constant.screen_size[1] = rb->height; + push_constant.z_near = p_projection.get_z_near(); + push_constant.z_far = p_projection.get_z_far(); + push_constant.orthogonal = p_projection.is_orthogonal(); + push_constant.proj_info[0] = -2.0f / (rb->width * p_projection.matrix[0][0]); + push_constant.proj_info[1] = -2.0f / (rb->height * p_projection.matrix[1][1]); + push_constant.proj_info[2] = (1.0f - p_projection.matrix[0][2]) / p_projection.matrix[0][0]; + push_constant.proj_info[3] = (1.0f + p_projection.matrix[1][2]) / p_projection.matrix[1][1]; + push_constant.max_giprobes = MIN((uint64_t)RenderBuffers::MAX_GIPROBES, p_gi_probes.size()); + push_constant.high_quality_vct = gi_probe_quality == RS::GI_PROBE_QUALITY_HIGH; + + bool use_sdfgi = rb->sdfgi != nullptr; + bool use_giprobes = push_constant.max_giprobes > 0; + + if (env) { + push_constant.ao_color[0] = env->ao_color.r; + push_constant.ao_color[1] = env->ao_color.g; + push_constant.ao_color[2] = env->ao_color.b; + } else { + push_constant.ao_color[0] = 0; + push_constant.ao_color[1] = 0; + push_constant.ao_color[2] = 0; + } + + push_constant.cam_rotation[0] = p_transform.basis[0][0]; + push_constant.cam_rotation[1] = p_transform.basis[1][0]; + push_constant.cam_rotation[2] = p_transform.basis[2][0]; + push_constant.cam_rotation[3] = 0; + push_constant.cam_rotation[4] = p_transform.basis[0][1]; + push_constant.cam_rotation[5] = p_transform.basis[1][1]; + push_constant.cam_rotation[6] = p_transform.basis[2][1]; + push_constant.cam_rotation[7] = 0; + push_constant.cam_rotation[8] = p_transform.basis[0][2]; + push_constant.cam_rotation[9] = p_transform.basis[1][2]; + push_constant.cam_rotation[10] = p_transform.basis[2][2]; + push_constant.cam_rotation[11] = 0; if (rb->gi_uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(rb->gi_uniform_set)) { Vector<RD::Uniform> uniforms; @@ -1695,7 +1787,7 @@ void RendererSceneRenderRD::_process_gi(RID p_render_buffers, RID p_normal_rough RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_IMAGE; u.binding = 9; - u.ids.push_back(p_ambient_buffer); + u.ids.push_back(rb->ambient_buffer); uniforms.push_back(u); } @@ -1703,7 +1795,7 @@ void RendererSceneRenderRD::_process_gi(RID p_render_buffers, RID p_normal_rough RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_IMAGE; u.binding = 10; - u.ids.push_back(p_reflection_buffer); + u.ids.push_back(rb->reflection_buffer); uniforms.push_back(u); } @@ -1767,17 +1859,33 @@ void RendererSceneRenderRD::_process_gi(RID p_render_buffers, RID p_normal_rough rb->gi_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi.shader.version_get_shader(gi.shader_version, 0), 0); } - GI::Mode mode = (use_sdfgi && use_giprobes) ? GI::MODE_COMBINED : (use_sdfgi ? GI::MODE_SDFGI : GI::MODE_GIPROBE); - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + GI::Mode mode; + + if (rb->using_half_size_gi) { + mode = (use_sdfgi && use_giprobes) ? GI::MODE_HALF_RES_COMBINED : (use_sdfgi ? GI::MODE_HALF_RES_SDFGI : GI::MODE_HALF_RES_GIPROBE); + } else { + mode = (use_sdfgi && use_giprobes) ? GI::MODE_COMBINED : (use_sdfgi ? GI::MODE_SDFGI : GI::MODE_GIPROBE); + } + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(true); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi.pipelines[mode]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->gi_uniform_set, 0); RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(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(); + + if (rb->using_half_size_gi) { + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->width >> 1, rb->height >> 1, 1); + } else { + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->width, rb->height, 1); + } + //do barrier later to allow oeverlap + //RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_NO_BARRIER); //no barriers, let other compute, raster and transfer happen at the same time + RD::get_singleton()->draw_command_end_label(); } -RID RendererSceneRenderRD::sky_create() { - return sky_owner.make_rid(Sky()); +RID RendererSceneRenderRD::sky_allocate() { + return sky_owner.allocate_rid(); +} +void RendererSceneRenderRD::sky_initialize(RID p_rid) { + sky_owner.initialize_rid(p_rid, Sky()); } void RendererSceneRenderRD::_sky_invalidate(Sky *p_sky) { @@ -2291,7 +2399,7 @@ void RendererSceneRenderRD::_setup_sky(RID p_environment, RID p_render_buffers, } if (light_data_dirty) { - RD::get_singleton()->buffer_update(sky_scene_state.directional_light_buffer, 0, sizeof(SkyDirectionalLightData) * sky_scene_state.max_directional_lights, sky_scene_state.directional_lights, true); + RD::get_singleton()->buffer_update(sky_scene_state.directional_light_buffer, 0, sizeof(SkyDirectionalLightData) * sky_scene_state.max_directional_lights, sky_scene_state.directional_lights); RendererSceneRenderRD::SkyDirectionalLightData *temp = sky_scene_state.last_frame_directional_lights; sky_scene_state.last_frame_directional_lights = sky_scene_state.directional_lights; @@ -2343,7 +2451,7 @@ void RendererSceneRenderRD::_setup_sky(RID p_environment, RID p_render_buffers, sky_scene_state.ubo.fog_light_color[2] = fog_color.b * fog_energy; sky_scene_state.ubo.fog_sun_scatter = environment_get_fog_sun_scatter(p_environment); - RD::get_singleton()->buffer_update(sky_scene_state.uniform_buffer, 0, sizeof(SkySceneState::UBO), &sky_scene_state.ubo, true); + RD::get_singleton()->buffer_update(sky_scene_state.uniform_buffer, 0, sizeof(SkySceneState::UBO), &sky_scene_state.ubo); } void RendererSceneRenderRD::_update_sky(RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform) { @@ -2801,8 +2909,11 @@ RendererStorageRD::MaterialData *RendererSceneRenderRD::_create_sky_material_fun return material_data; } -RID RendererSceneRenderRD::environment_create() { - return environment_owner.make_rid(Environment()); +RID RendererSceneRenderRD::environment_allocate() { + return environment_owner.allocate_rid(); +} +void RendererSceneRenderRD::environment_initialize(RID p_rid) { + environment_owner.initialize_rid(p_rid, Environment()); } void RendererSceneRenderRD::environment_set_background(RID p_env, RS::EnvironmentBG p_bg) { @@ -2976,7 +3087,7 @@ void RendererSceneRenderRD::environment_glow_set_use_high_quality(bool p_enable) glow_high_quality = p_enable; } -void RendererSceneRenderRD::environment_set_sdfgi(RID p_env, bool p_enable, RS::EnvironmentSDFGICascades p_cascades, float p_min_cell_size, RS::EnvironmentSDFGIYScale p_y_scale, bool p_use_occlusion, bool p_use_multibounce, bool p_read_sky, float p_energy, float p_normal_bias, float p_probe_bias) { +void RendererSceneRenderRD::environment_set_sdfgi(RID p_env, bool p_enable, RS::EnvironmentSDFGICascades p_cascades, float p_min_cell_size, RS::EnvironmentSDFGIYScale p_y_scale, bool p_use_occlusion, float p_bounce_feedback, bool p_read_sky, float p_energy, float p_normal_bias, float p_probe_bias) { Environment *env = environment_owner.getornull(p_env); ERR_FAIL_COND(!env); @@ -2988,7 +3099,7 @@ void RendererSceneRenderRD::environment_set_sdfgi(RID p_env, bool p_enable, RS:: 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_bounce_feedback = p_bounce_feedback; env->sdfgi_read_sky_light = p_read_sky; env->sdfgi_energy = p_energy; env->sdfgi_normal_bias = p_normal_bias; @@ -3054,7 +3165,7 @@ float RendererSceneRenderRD::environment_get_fog_aerial_perspective(RID p_env) c return env->fog_aerial_perspective; } -void RendererSceneRenderRD::environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, RenderingServer::EnvVolumetricFogShadowFilter p_shadow_filter) { +void RendererSceneRenderRD::environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, bool p_temporal_reprojection, float p_temporal_reprojection_amount) { Environment *env = environment_owner.getornull(p_env); ERR_FAIL_COND(!env); @@ -3068,8 +3179,9 @@ void RendererSceneRenderRD::environment_set_volumetric_fog(RID p_env, bool p_ena env->volumetric_fog_light_energy = p_light_energy; env->volumetric_fog_length = p_length; env->volumetric_fog_detail_spread = p_detail_spread; - env->volumetric_fog_shadow_filter = p_shadow_filter; env->volumetric_fog_gi_inject = p_gi_inject; + env->volumetric_fog_temporal_reprojection = p_temporal_reprojection; + env->volumetric_fog_temporal_reprojection_amount = p_temporal_reprojection_amount; } void RendererSceneRenderRD::environment_set_volumetric_fog_volume_size(int p_size, int p_depth) { @@ -3080,25 +3192,6 @@ void RendererSceneRenderRD::environment_set_volumetric_fog_volume_size(int p_siz void RendererSceneRenderRD::environment_set_volumetric_fog_filter_active(bool p_enable) { volumetric_fog_filter_active = p_enable; } -void RendererSceneRenderRD::environment_set_volumetric_fog_directional_shadow_shrink_size(int p_shrink_size) { - p_shrink_size = nearest_power_of_2_templated(p_shrink_size); - if (volumetric_fog_directional_shadow_shrink == (uint32_t)p_shrink_size) { - return; - } - - _clear_shadow_shrink_stages(directional_shadow.shrink_stages); -} -void RendererSceneRenderRD::environment_set_volumetric_fog_positional_shadow_shrink_size(int p_shrink_size) { - p_shrink_size = nearest_power_of_2_templated(p_shrink_size); - if (volumetric_fog_positional_shadow_shrink == (uint32_t)p_shrink_size) { - return; - } - - for (uint32_t i = 0; i < shadow_atlas_owner.get_rid_count(); i++) { - ShadowAtlas *sa = shadow_atlas_owner.get_ptr_by_index(i); - _clear_shadow_shrink_stages(sa->shrink_stages); - } -} void RendererSceneRenderRD::environment_set_sdfgi_ray_count(RS::EnvironmentSDFGIRayCount p_ray_count) { sdfgi_ray_count = p_ray_count; @@ -3107,6 +3200,9 @@ void RendererSceneRenderRD::environment_set_sdfgi_ray_count(RS::EnvironmentSDFGI void RendererSceneRenderRD::environment_set_sdfgi_frames_to_converge(RS::EnvironmentSDFGIFramesToConverge p_frames) { sdfgi_frames_to_converge = p_frames; } +void RendererSceneRenderRD::environment_set_sdfgi_frames_to_update_light(RS::EnvironmentSDFGIFramesToUpdateLight p_update) { + sdfgi_frames_to_update_light = p_update; +} void RendererSceneRenderRD::environment_set_ssr(RID p_env, bool p_enable, int p_max_steps, float p_fade_int, float p_fade_out, float p_depth_tolerance) { Environment *env = environment_owner.getornull(p_env); @@ -3233,8 +3329,8 @@ Ref<Image> RendererSceneRenderRD::environment_bake_panorama(RID p_env, bool p_ba RID RendererSceneRenderRD::reflection_atlas_create() { ReflectionAtlas ra; - ra.count = GLOBAL_GET("rendering/quality/reflection_atlas/reflection_count"); - ra.size = GLOBAL_GET("rendering/quality/reflection_atlas/reflection_size"); + ra.count = GLOBAL_GET("rendering/reflections/reflection_atlas/reflection_count"); + ra.size = GLOBAL_GET("rendering/reflections/reflection_atlas/reflection_size"); ra.cluster_builder = memnew(ClusterBuilderRD); ra.cluster_builder->set_shared(&cluster_builder_shared); @@ -3518,13 +3614,28 @@ RID RendererSceneRenderRD::shadow_atlas_create() { return shadow_atlas_owner.make_rid(ShadowAtlas()); } -void RendererSceneRenderRD::shadow_atlas_set_size(RID p_atlas, int p_size) { +void RendererSceneRenderRD::_update_shadow_atlas(ShadowAtlas *shadow_atlas) { + if (shadow_atlas->size > 0 && shadow_atlas->depth.is_null()) { + RD::TextureFormat tf; + tf.format = shadow_atlas->use_16_bits ? RD::DATA_FORMAT_D16_UNORM : RD::DATA_FORMAT_D32_SFLOAT; + tf.width = shadow_atlas->size; + tf.height = shadow_atlas->size; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT; + + shadow_atlas->depth = RD::get_singleton()->texture_create(tf, RD::TextureView()); + Vector<RID> fb_tex; + fb_tex.push_back(shadow_atlas->depth); + shadow_atlas->fb = RD::get_singleton()->framebuffer_create(fb_tex); + } +} + +void RendererSceneRenderRD::shadow_atlas_set_size(RID p_atlas, int p_size, bool p_16_bits) { ShadowAtlas *shadow_atlas = shadow_atlas_owner.getornull(p_atlas); ERR_FAIL_COND(!shadow_atlas); ERR_FAIL_COND(p_size < 0); p_size = next_power_of_2(p_size); - if (p_size == shadow_atlas->size) { + if (p_size == shadow_atlas->size && p_16_bits == shadow_atlas->use_16_bits) { return; } @@ -3532,7 +3643,6 @@ void RendererSceneRenderRD::shadow_atlas_set_size(RID p_atlas, int p_size) { if (shadow_atlas->depth.is_valid()) { RD::get_singleton()->free(shadow_atlas->depth); shadow_atlas->depth = RID(); - _clear_shadow_shrink_stages(shadow_atlas->shrink_stages); } for (int i = 0; i < 4; i++) { //clear subdivisions @@ -3551,16 +3661,7 @@ void RendererSceneRenderRD::shadow_atlas_set_size(RID p_atlas, int p_size) { shadow_atlas->shadow_owners.clear(); shadow_atlas->size = p_size; - - if (shadow_atlas->size) { - RD::TextureFormat tf; - tf.format = RD::DATA_FORMAT_R32_SFLOAT; - tf.width = shadow_atlas->size; - tf.height = shadow_atlas->size; - tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; - - shadow_atlas->depth = RD::get_singleton()->texture_create(tf, RD::TextureView()); - } + shadow_atlas->use_16_bits = p_size; } void RendererSceneRenderRD::shadow_atlas_set_quadrant_subdivision(RID p_atlas, int p_quadrant, int p_subdivision) { @@ -3815,10 +3916,24 @@ bool RendererSceneRenderRD::shadow_atlas_update_light(RID p_atlas, RID p_light_i return false; } -void RendererSceneRenderRD::directional_shadow_atlas_set_size(int p_size) { +void RendererSceneRenderRD::_update_directional_shadow_atlas() { + if (directional_shadow.depth.is_null() && directional_shadow.size > 0) { + RD::TextureFormat tf; + tf.format = directional_shadow.use_16_bits ? RD::DATA_FORMAT_D16_UNORM : RD::DATA_FORMAT_D32_SFLOAT; + tf.width = directional_shadow.size; + tf.height = directional_shadow.size; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT; + + directional_shadow.depth = RD::get_singleton()->texture_create(tf, RD::TextureView()); + Vector<RID> fb_tex; + fb_tex.push_back(directional_shadow.depth); + directional_shadow.fb = RD::get_singleton()->framebuffer_create(fb_tex); + } +} +void RendererSceneRenderRD::directional_shadow_atlas_set_size(int p_size, bool p_16_bits) { p_size = nearest_power_of_2_templated(p_size); - if (directional_shadow.size == p_size) { + if (directional_shadow.size == p_size && directional_shadow.use_16_bits == p_16_bits) { return; } @@ -3826,21 +3941,9 @@ void RendererSceneRenderRD::directional_shadow_atlas_set_size(int p_size) { if (directional_shadow.depth.is_valid()) { RD::get_singleton()->free(directional_shadow.depth); - _clear_shadow_shrink_stages(directional_shadow.shrink_stages); directional_shadow.depth = RID(); + _base_uniforms_changed(); } - - if (p_size > 0) { - RD::TextureFormat tf; - tf.format = RD::DATA_FORMAT_R32_SFLOAT; - tf.width = p_size; - tf.height = p_size; - tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; - - directional_shadow.depth = RD::get_singleton()->texture_create(tf, RD::TextureView()); - } - - _base_uniforms_changed(); } void RendererSceneRenderRD::set_directional_shadow_count(int p_count) { @@ -3894,8 +3997,11 @@ int RendererSceneRenderRD::get_directional_light_shadow_size(RID p_light_intance ////////////////////////////////////////////////// -RID RendererSceneRenderRD::camera_effects_create() { - return camera_effects_owner.make_rid(CameraEffects()); +RID RendererSceneRenderRD::camera_effects_allocate() { + return camera_effects_owner.allocate_rid(); +} +void RendererSceneRenderRD::camera_effects_initialize(RID p_rid) { + camera_effects_owner.initialize_rid(p_rid, CameraEffects()); } void RendererSceneRenderRD::camera_effects_set_dof_blur_quality(RS::DOFBlurQuality p_quality, bool p_use_jitter) { @@ -3960,11 +4066,7 @@ void RendererSceneRenderRD::light_instance_set_shadow_transform(RID p_light_inst LightInstance *light_instance = light_instance_owner.getornull(p_light_instance); ERR_FAIL_COND(!light_instance); - if (storage->light_get_type(light_instance->light) != RS::LIGHT_DIRECTIONAL) { - p_pass = 0; - } - - ERR_FAIL_INDEX(p_pass, 4); + ERR_FAIL_INDEX(p_pass, 6); light_instance->shadow_transform[p_pass].camera = p_projection; light_instance->shadow_transform[p_pass].transform = p_transform; @@ -4010,29 +4112,6 @@ RendererSceneRenderRD::ShadowCubemap *RendererSceneRenderRD::_get_shadow_cubemap return &shadow_cubemaps[p_size]; } -RendererSceneRenderRD::ShadowMap *RendererSceneRenderRD::_get_shadow_map(const Size2i &p_size) { - if (!shadow_maps.has(p_size)) { - ShadowMap sm; - { - RD::TextureFormat tf; - tf.format = RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_D32_SFLOAT, RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) ? RD::DATA_FORMAT_D32_SFLOAT : RD::DATA_FORMAT_X8_D24_UNORM_PACK32; - tf.width = p_size.width; - tf.height = p_size.height; - tf.usage_bits = RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; - - sm.depth = RD::get_singleton()->texture_create(tf, RD::TextureView()); - } - - Vector<RID> fbtex; - fbtex.push_back(sm.depth); - sm.fb = RD::get_singleton()->framebuffer_create(fbtex); - - shadow_maps[p_size] = sm; - } - - return &shadow_maps[p_size]; -} - ////////////////////////// RID RendererSceneRenderRD::decal_instance_create(RID p_decal) { @@ -4133,7 +4212,7 @@ void RendererSceneRenderRD::gi_probe_update(RID p_probe, bool p_update_light_ins gi_probe->texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); - RD::get_singleton()->texture_clear(gi_probe->texture, Color(0, 0, 0, 0), 0, levels.size(), 0, 1, false); + RD::get_singleton()->texture_clear(gi_probe->texture, Color(0, 0, 0, 0), 0, levels.size(), 0, 1); { int total_elements = 0; @@ -4445,7 +4524,7 @@ void RendererSceneRenderRD::gi_probe_update(RID p_probe, bool p_update_light_ins 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); + RD::get_singleton()->texture_clear(gi_probe->texture, Color(0, 0, 0, 0), 0, gi_probe->mipmaps.size(), 0, 1); } uint32_t light_count = 0; @@ -4477,8 +4556,8 @@ void RendererSceneRenderRD::gi_probe_update(RID p_probe, bool p_update_light_ins l.color[1] = color.g; l.color[2] = color.b; - l.spot_angle_radians = Math::deg2rad(storage->light_get_param(light, RS::LIGHT_PARAM_SPOT_ANGLE)); - l.spot_attenuation = storage->light_get_param(light, RS::LIGHT_PARAM_SPOT_ATTENUATION); + l.cos_spot_angle = Math::cos(Math::deg2rad(storage->light_get_param(light, RS::LIGHT_PARAM_SPOT_ANGLE))); + l.inv_spot_attenuation = 1.0f / storage->light_get_param(light, RS::LIGHT_PARAM_SPOT_ATTENUATION); Transform xform = light_instance_get_base_transform(light_instance); @@ -4496,7 +4575,7 @@ void RendererSceneRenderRD::gi_probe_update(RID p_probe, bool p_update_light_ins l.has_shadow = storage->light_has_shadow(light); } - RD::get_singleton()->buffer_update(gi_probe_lights_uniform, 0, sizeof(GIProbeLight) * light_count, gi_probe_lights, true); + RD::get_singleton()->buffer_update(gi_probe_lights_uniform, 0, sizeof(GIProbeLight) * light_count, gi_probe_lights); } } @@ -4897,7 +4976,7 @@ void RendererSceneRenderRD::_debug_sdfgi_probes(RID p_render_buffers, RD::DrawLi 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.section_arc = Math_TAU / 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; @@ -5118,9 +5197,6 @@ void RendererSceneRenderRD::_free_render_buffer_data(RenderBuffers *rb) { RD::get_singleton()->free(rb->luminance.reduce[i]); } - for (int i = 0; i < rb->luminance.reduce.size(); i++) { - RD::get_singleton()->free(rb->luminance.reduce[i]); - } rb->luminance.reduce.clear(); if (rb->luminance.current.is_valid()) { @@ -5161,6 +5237,13 @@ void RendererSceneRenderRD::_free_render_buffer_data(RenderBuffers *rb) { RD::get_singleton()->free(rb->ssr.normal_scaled); rb->ssr.normal_scaled = RID(); } + + if (rb->ambient_buffer.is_valid()) { + RD::get_singleton()->free(rb->ambient_buffer); + RD::get_singleton()->free(rb->reflection_buffer); + rb->ambient_buffer = RID(); + rb->reflection_buffer = RID(); + } } void RendererSceneRenderRD::_process_sss(RID p_render_buffers, const CameraMatrix &p_camera) { @@ -5293,9 +5376,11 @@ void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environmen tf.array_layers = 4; tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; rb->ssao.depth = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(rb->ssao.depth, "SSAO Depth"); for (uint32_t i = 0; i < tf.mipmaps; i++) { RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), rb->ssao.depth, 0, i, RD::TEXTURE_SLICE_2D_ARRAY); rb->ssao.depth_slices.push_back(slice); + RD::get_singleton()->set_resource_name(rb->ssao.depth_slices[i], "SSAO Depth Mip " + itos(i) + " "); } } @@ -5308,9 +5393,11 @@ void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environmen tf.array_layers = 4; tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; rb->ssao.ao_deinterleaved = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(rb->ssao.ao_deinterleaved, "SSAO De-interleaved Array"); for (uint32_t i = 0; i < 4; i++) { RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), rb->ssao.ao_deinterleaved, i, 0); rb->ssao.ao_deinterleaved_slices.push_back(slice); + RD::get_singleton()->set_resource_name(rb->ssao.ao_deinterleaved_slices[i], "SSAO De-interleaved Array Layer " + itos(i) + " "); } } @@ -5323,9 +5410,11 @@ void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environmen tf.array_layers = 4; tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; rb->ssao.ao_pong = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(rb->ssao.ao_pong, "SSAO De-interleaved Array Pong"); for (uint32_t i = 0; i < 4; i++) { RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), rb->ssao.ao_pong, i, 0); rb->ssao.ao_pong_slices.push_back(slice); + RD::get_singleton()->set_resource_name(rb->ssao.ao_deinterleaved_slices[i], "SSAO De-interleaved Array Layer " + itos(i) + " Pong"); } } @@ -5336,7 +5425,9 @@ void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environmen tf.height = half_height; tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; rb->ssao.importance_map[0] = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(rb->ssao.importance_map[0], "SSAO Importance Map"); rb->ssao.importance_map[1] = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(rb->ssao.importance_map[1], "SSAO Importance Map Pong"); } { RD::TextureFormat tf; @@ -5345,6 +5436,7 @@ void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environmen tf.height = rb->height; tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; rb->ssao.ao_final = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(rb->ssao.ao_final, "SSAO Final"); _render_buffers_uniform_set_changed(p_render_buffers); } ssao_using_half_size = ssao_half_size; @@ -5560,10 +5652,10 @@ void RendererSceneRenderRD::_render_buffers_debug_draw(RID p_render_buffers, RID effects->copy_to_fb_rect(_render_buffers_get_normal_texture(p_render_buffers), storage->render_target_get_rd_framebuffer(rb->render_target), Rect2(Vector2(), rtsize), false, false); } - if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_GI_BUFFER && _render_buffers_get_ambient_texture(p_render_buffers).is_valid()) { + if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_GI_BUFFER && rb->ambient_buffer.is_valid()) { Size2 rtsize = storage->render_target_get_size(rb->render_target); - RID ambient_texture = _render_buffers_get_ambient_texture(p_render_buffers); - RID reflection_texture = _render_buffers_get_reflection_texture(p_render_buffers); + RID ambient_texture = rb->ambient_buffer; + RID reflection_texture = rb->reflection_buffer; effects->copy_to_fb_rect(ambient_texture, storage->render_target_get_rd_framebuffer(rb->render_target), Rect2(Vector2(), rtsize), false, false, false, true, reflection_texture); } } @@ -5719,7 +5811,7 @@ void RendererSceneRenderRD::_sdfgi_debug_draw(RID p_render_buffers, const Camera 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_dispatch_threads(compute_list, rb->width, rb->height, 1); RD::get_singleton()->compute_list_end(); Size2 rtsize = storage->render_target_get_size(rb->render_target); @@ -5755,6 +5847,17 @@ RID RendererSceneRenderRD::render_buffers_get_default_gi_probe_buffer() { return default_giprobe_buffer; } +RID RendererSceneRenderRD::render_buffers_get_gi_ambient_texture(RID p_render_buffers) { + RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); + ERR_FAIL_COND_V(!rb, RID()); + return rb->ambient_buffer; +} +RID RendererSceneRenderRD::render_buffers_get_gi_reflection_texture(RID p_render_buffers) { + RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); + ERR_FAIL_COND_V(!rb, RID()); + return rb->reflection_buffer; +} + uint32_t RendererSceneRenderRD::render_buffers_get_sdfgi_cascade_count(RID p_render_buffers) const { const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); ERR_FAIL_COND_V(!rb, 0); @@ -5941,6 +6044,10 @@ void RendererSceneRenderRD::render_buffers_configure(RID p_render_buffers, RID p rb->cluster_builder->setup(Size2i(p_width, p_height), max_cluster_elements, rb->depth_texture, storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED), rb->texture); } +void RendererSceneRenderRD::gi_set_use_half_resolution(bool p_enable) { + gi.half_resolution = p_enable; +} + void RendererSceneRenderRD::sub_surface_scattering_set_quality(RS::SubSurfaceScatteringQuality p_quality) { sss_quality = p_quality; } @@ -6116,7 +6223,7 @@ void RendererSceneRenderRD::_setup_reflections(const PagedArray<RID> &p_reflecti } if (cluster.reflection_count) { - RD::get_singleton()->buffer_update(cluster.reflection_buffer, 0, cluster.reflection_count * sizeof(ReflectionData), cluster.reflections, true); + RD::get_singleton()->buffer_update(cluster.reflection_buffer, 0, cluster.reflection_count * sizeof(ReflectionData), cluster.reflections, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); } } @@ -6368,8 +6475,6 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const LightInstance *li = (i < cluster.omni_light_count) ? cluster.omni_light_sort[index].instance : cluster.spot_light_sort[index].instance; RID base = li->light; - cluster.lights_instances[i] = li->self; - Transform light_transform = li->transform; float sign = storage->light_is_negative(base) ? -1 : 1; @@ -6403,9 +6508,9 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const light_data.size = size; - light_data.cone_attenuation = storage->light_get_param(base, RS::LIGHT_PARAM_SPOT_ATTENUATION); + light_data.inv_spot_attenuation = 1.0f / storage->light_get_param(base, RS::LIGHT_PARAM_SPOT_ATTENUATION); float spot_angle = storage->light_get_param(base, RS::LIGHT_PARAM_SPOT_ANGLE); - light_data.cone_angle = Math::cos(Math::deg2rad(spot_angle)); + light_data.cos_spot_angle = Math::cos(Math::deg2rad(spot_angle)); light_data.mask = storage->light_get_cull_mask(base); @@ -6508,16 +6613,17 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const r_positional_light_count++; } + //update without barriers if (cluster.omni_light_count) { - RD::get_singleton()->buffer_update(cluster.omni_light_buffer, 0, sizeof(Cluster::LightData) * cluster.omni_light_count, cluster.omni_lights, true); + RD::get_singleton()->buffer_update(cluster.omni_light_buffer, 0, sizeof(Cluster::LightData) * cluster.omni_light_count, cluster.omni_lights, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); } if (cluster.spot_light_count) { - RD::get_singleton()->buffer_update(cluster.spot_light_buffer, 0, sizeof(Cluster::LightData) * cluster.spot_light_count, cluster.spot_lights, true); + RD::get_singleton()->buffer_update(cluster.spot_light_buffer, 0, sizeof(Cluster::LightData) * cluster.spot_light_count, cluster.spot_lights, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); } if (r_directional_light_count) { - RD::get_singleton()->buffer_update(cluster.directional_light_buffer, 0, sizeof(Cluster::DirectionalLightData) * r_directional_light_count, cluster.directional_lights, true); + RD::get_singleton()->buffer_update(cluster.directional_light_buffer, 0, sizeof(Cluster::DirectionalLightData) * r_directional_light_count, cluster.directional_lights, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); } } @@ -6678,13 +6784,14 @@ void RendererSceneRenderRD::_setup_decals(const PagedArray<RID> &p_decals, const } if (cluster.decal_count > 0) { - RD::get_singleton()->buffer_update(cluster.decal_buffer, 0, sizeof(Cluster::DecalData) * cluster.decal_count, cluster.decals, true); + RD::get_singleton()->buffer_update(cluster.decal_buffer, 0, sizeof(Cluster::DecalData) * cluster.decal_count, cluster.decals, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); } } void RendererSceneRenderRD::_volumetric_fog_erase(RenderBuffers *rb) { ERR_FAIL_COND(!rb->volumetric_fog); + RD::get_singleton()->free(rb->volumetric_fog->prev_light_density_map); RD::get_singleton()->free(rb->volumetric_fog->light_density_map); RD::get_singleton()->free(rb->volumetric_fog->fog_map); @@ -6706,49 +6813,6 @@ void RendererSceneRenderRD::_volumetric_fog_erase(RenderBuffers *rb) { rb->volumetric_fog = nullptr; } -void RendererSceneRenderRD::_allocate_shadow_shrink_stages(RID p_base, int p_base_size, Vector<ShadowShrinkStage> &shrink_stages, uint32_t p_target_size) { - //create fog mipmaps - uint32_t fog_texture_size = p_target_size; - uint32_t base_texture_size = p_base_size; - - ShadowShrinkStage first; - first.size = base_texture_size; - first.texture = p_base; - shrink_stages.push_back(first); //put depth first in case we dont find smaller ones - - while (fog_texture_size < base_texture_size) { - base_texture_size = MAX(base_texture_size / 8, fog_texture_size); - - ShadowShrinkStage s; - s.size = base_texture_size; - - RD::TextureFormat tf; - tf.format = RD::DATA_FORMAT_R32_SFLOAT; - tf.width = base_texture_size; - tf.height = base_texture_size; - tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT; - - if (base_texture_size == fog_texture_size) { - s.filter_texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); - tf.usage_bits |= RD::TEXTURE_USAGE_SAMPLING_BIT; - } - - s.texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); - - shrink_stages.push_back(s); - } -} - -void RendererSceneRenderRD::_clear_shadow_shrink_stages(Vector<ShadowShrinkStage> &shrink_stages) { - for (int i = 1; i < shrink_stages.size(); i++) { - RD::get_singleton()->free(shrink_stages[i].texture); - if (shrink_stages[i].filter_texture.is_valid()) { - RD::get_singleton()->free(shrink_stages[i].filter_texture); - } - } - shrink_stages.clear(); -} - void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_environment, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_shadow_atlas, int p_directional_light_count, bool p_use_directional_shadows, int p_positional_light_count, int p_gi_probe_count) { RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); ERR_FAIL_COND(!rb); @@ -6771,6 +6835,8 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e return; } + RENDER_TIMESTAMP(">Volumetric Fog"); + if (env && env->volumetric_fog_enabled && !rb->volumetric_fog) { //required volumetric fog but not existing, create rb->volumetric_fog = memnew(VolumetricFog); @@ -6784,11 +6850,16 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e tf.height = target_height; tf.depth = volumetric_fog_depth; tf.texture_type = RD::TEXTURE_TYPE_3D; - tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT; + tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT; rb->volumetric_fog->light_density_map = RD::get_singleton()->texture_create(tf, RD::TextureView()); - tf.usage_bits |= RD::TEXTURE_USAGE_SAMPLING_BIT; + tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; + + rb->volumetric_fog->prev_light_density_map = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->texture_clear(rb->volumetric_fog->prev_light_density_map, Color(0, 0, 0, 0), 0, 1, 0, 1); + + tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; rb->volumetric_fog->fog_map = RD::get_singleton()->texture_create(tf, RD::TextureView()); _render_buffers_uniform_set_changed(p_render_buffers); @@ -6805,164 +6876,6 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e rb->volumetric_fog->sky_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sky_shader.default_shader_rd, SKY_SET_FOG); } - //update directional shadow - - if (p_use_directional_shadows) { - if (directional_shadow.shrink_stages.is_empty()) { - if (rb->volumetric_fog->uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(rb->volumetric_fog->uniform_set)) { - //invalidate uniform set, we will need a new one - RD::get_singleton()->free(rb->volumetric_fog->uniform_set); - rb->volumetric_fog->uniform_set = RID(); - } - _allocate_shadow_shrink_stages(directional_shadow.depth, directional_shadow.size, directional_shadow.shrink_stages, volumetric_fog_directional_shadow_shrink); - } - - if (directional_shadow.shrink_stages.size() > 1) { - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - for (int i = 1; i < directional_shadow.shrink_stages.size(); i++) { - int32_t src_size = directional_shadow.shrink_stages[i - 1].size; - int32_t dst_size = directional_shadow.shrink_stages[i].size; - Rect2i r(0, 0, src_size, src_size); - int32_t shrink_limit = 8 / (src_size / dst_size); - - storage->get_effects()->reduce_shadow(directional_shadow.shrink_stages[i - 1].texture, directional_shadow.shrink_stages[i].texture, Size2i(src_size, src_size), r, shrink_limit, compute_list); - RD::get_singleton()->compute_list_add_barrier(compute_list); - if (env->volumetric_fog_shadow_filter != RS::ENV_VOLUMETRIC_FOG_SHADOW_FILTER_DISABLED && directional_shadow.shrink_stages[i].filter_texture.is_valid()) { - Rect2i rf(0, 0, dst_size, dst_size); - storage->get_effects()->filter_shadow(directional_shadow.shrink_stages[i].texture, directional_shadow.shrink_stages[i].filter_texture, Size2i(dst_size, dst_size), rf, env->volumetric_fog_shadow_filter, compute_list); - } - } - RD::get_singleton()->compute_list_end(); - } - } - - ShadowAtlas *shadow_atlas = shadow_atlas_owner.getornull(p_shadow_atlas); - - if (shadow_atlas) { - //shrink shadows that need to be shrunk - - bool force_shrink_shadows = false; - - if (shadow_atlas->shrink_stages.is_empty()) { - if (rb->volumetric_fog->uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(rb->volumetric_fog->uniform_set)) { - //invalidate uniform set, we will need a new one - RD::get_singleton()->free(rb->volumetric_fog->uniform_set); - rb->volumetric_fog->uniform_set = RID(); - } - _allocate_shadow_shrink_stages(shadow_atlas->depth, shadow_atlas->size, shadow_atlas->shrink_stages, volumetric_fog_positional_shadow_shrink); - force_shrink_shadows = true; - } - - if (rb->volumetric_fog->last_shadow_filter != env->volumetric_fog_shadow_filter) { - //if shadow filter changed, invalidate caches - rb->volumetric_fog->last_shadow_filter = env->volumetric_fog_shadow_filter; - force_shrink_shadows = true; - } - - cluster.lights_shadow_rect_cache_count = 0; - - for (uint32_t i = 0; i < cluster.omni_light_count + cluster.spot_light_count; i++) { - Cluster::LightData &ld = i < cluster.omni_light_count ? cluster.omni_lights[i] : cluster.spot_lights[i - cluster.omni_light_count]; - - if (ld.shadow_enabled != 0) { - RID li = cluster.lights_instances[i]; - - ERR_CONTINUE(!shadow_atlas->shadow_owners.has(li)); - - uint32_t key = shadow_atlas->shadow_owners[li]; - - uint32_t quadrant = (key >> ShadowAtlas::QUADRANT_SHIFT) & 0x3; - uint32_t shadow = key & ShadowAtlas::SHADOW_INDEX_MASK; - - ERR_CONTINUE((int)shadow >= shadow_atlas->quadrants[quadrant].shadows.size()); - - ShadowAtlas::Quadrant::Shadow &s = shadow_atlas->quadrants[quadrant].shadows.write[shadow]; - - if (!force_shrink_shadows && s.fog_version == s.version) { - continue; //do not update, no need - } - - s.fog_version = s.version; - - uint32_t quadrant_size = shadow_atlas->size >> 1; - - Rect2i atlas_rect; - - atlas_rect.position.x = (quadrant & 1) * quadrant_size; - atlas_rect.position.y = (quadrant >> 1) * quadrant_size; - - uint32_t shadow_size = (quadrant_size / shadow_atlas->quadrants[quadrant].subdivision); - atlas_rect.position.x += (shadow % shadow_atlas->quadrants[quadrant].subdivision) * shadow_size; - atlas_rect.position.y += (shadow / shadow_atlas->quadrants[quadrant].subdivision) * shadow_size; - - atlas_rect.size.x = shadow_size; - atlas_rect.size.y = shadow_size; - - cluster.lights_shadow_rect_cache[cluster.lights_shadow_rect_cache_count] = atlas_rect; - - cluster.lights_shadow_rect_cache_count++; - - if (cluster.lights_shadow_rect_cache_count == cluster.max_lights * 2) { - break; //light limit reached - } - } - } - - if (cluster.lights_shadow_rect_cache_count > 0) { - //there are shadows to be shrunk, try to do them in parallel - RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - - for (int i = 1; i < shadow_atlas->shrink_stages.size(); i++) { - int32_t base_size = shadow_atlas->shrink_stages[0].size; - int32_t src_size = shadow_atlas->shrink_stages[i - 1].size; - int32_t dst_size = shadow_atlas->shrink_stages[i].size; - - uint32_t rect_divisor = base_size / src_size; - - int32_t shrink_limit = 8 / (src_size / dst_size); - - //shrink in parallel for more performance - for (uint32_t j = 0; j < cluster.lights_shadow_rect_cache_count; j++) { - Rect2i src_rect = cluster.lights_shadow_rect_cache[j]; - - src_rect.position /= rect_divisor; - src_rect.size /= rect_divisor; - - storage->get_effects()->reduce_shadow(shadow_atlas->shrink_stages[i - 1].texture, shadow_atlas->shrink_stages[i].texture, Size2i(src_size, src_size), src_rect, shrink_limit, compute_list); - } - - RD::get_singleton()->compute_list_add_barrier(compute_list); - - if (env->volumetric_fog_shadow_filter != RS::ENV_VOLUMETRIC_FOG_SHADOW_FILTER_DISABLED && shadow_atlas->shrink_stages[i].filter_texture.is_valid()) { - uint32_t filter_divisor = base_size / dst_size; - - //filter in parallel for more performance - for (uint32_t j = 0; j < cluster.lights_shadow_rect_cache_count; j++) { - Rect2i dst_rect = cluster.lights_shadow_rect_cache[j]; - - dst_rect.position /= filter_divisor; - dst_rect.size /= filter_divisor; - - storage->get_effects()->filter_shadow(shadow_atlas->shrink_stages[i].texture, shadow_atlas->shrink_stages[i].filter_texture, Size2i(dst_size, dst_size), dst_rect, env->volumetric_fog_shadow_filter, compute_list, true, false); - } - - RD::get_singleton()->compute_list_add_barrier(compute_list); - - for (uint32_t j = 0; j < cluster.lights_shadow_rect_cache_count; j++) { - Rect2i dst_rect = cluster.lights_shadow_rect_cache[j]; - - dst_rect.position /= filter_divisor; - dst_rect.size /= filter_divisor; - - storage->get_effects()->filter_shadow(shadow_atlas->shrink_stages[i].texture, shadow_atlas->shrink_stages[i].filter_texture, Size2i(dst_size, dst_size), dst_rect, env->volumetric_fog_shadow_filter, compute_list, false, true); - } - } - } - - RD::get_singleton()->compute_list_end(); - } - } - //update volumetric fog if (rb->volumetric_fog->uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(rb->volumetric_fog->uniform_set)) { @@ -6974,10 +6887,11 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.binding = 1; - if (shadow_atlas == nullptr || shadow_atlas->shrink_stages.size() == 0) { + ShadowAtlas *shadow_atlas = shadow_atlas_owner.getornull(p_shadow_atlas); + if (shadow_atlas == nullptr || shadow_atlas->depth.is_null()) { u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK)); } else { - u.ids.push_back(shadow_atlas->shrink_stages[shadow_atlas->shrink_stages.size() - 1].texture); + u.ids.push_back(shadow_atlas->depth); } uniforms.push_back(u); @@ -6987,10 +6901,10 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.binding = 2; - if (directional_shadow.shrink_stages.size() == 0) { - u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK)); + if (directional_shadow.depth.is_valid()) { + u.ids.push_back(directional_shadow.depth); } else { - u.ids.push_back(directional_shadow.shrink_stages[directional_shadow.shrink_stages.size() - 1].texture); + u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK)); } uniforms.push_back(u); } @@ -7089,6 +7003,13 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e u.ids.push_back(volumetric_fog.params_ubo); uniforms.push_back(u); } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 15; + u.ids.push_back(rb->volumetric_fog->prev_light_density_map); + uniforms.push_back(u); + } rb->volumetric_fog->uniform_set = RD::get_singleton()->uniform_set_create(uniforms, volumetric_fog.shader.version_get_shader(volumetric_fog.shader_version, 0), 0); @@ -7190,6 +7111,13 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e params.cam_rotation[11] = 0; params.filter_axis = 0; params.max_gi_probes = env->volumetric_fog_gi_inject > 0.001 ? p_gi_probe_count : 0; + params.temporal_frame = RSG::rasterizer->get_frame_number() % VolumetricFog::MAX_TEMPORAL_FRAMES; + + Transform to_prev_cam_view = rb->volumetric_fog->prev_cam_transform.affine_inverse() * p_cam_transform; + storage->store_transform(to_prev_cam_view, params.to_prev_view); + + params.use_temporal_reprojection = env->volumetric_fog_temporal_reprojection; + params.temporal_blend = env->volumetric_fog_temporal_reprojection_amount; { uint32_t cluster_size = rb->cluster_builder->get_cluster_size(); @@ -7210,10 +7138,10 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e push_constant.directional_shadow_pixel_size[1] = 1.0 / dssize.y; */ - RENDER_TIMESTAMP(">Volumetric Fog"); + RD::get_singleton()->draw_command_begin_label("Render Volumetric Fog"); RENDER_TIMESTAMP("Render Fog"); - RD::get_singleton()->buffer_update(volumetric_fog.params_ubo, 0, sizeof(VolumetricFogShader::ParamsUBO), ¶ms, true); + RD::get_singleton()->buffer_update(volumetric_fog.params_ubo, 0, sizeof(VolumetricFogShader::ParamsUBO), ¶ms, RD::BARRIER_MASK_COMPUTE); RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); @@ -7226,23 +7154,31 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e if (using_sdfgi) { RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->volumetric_fog->sdfgi_uniform_set, 1); } - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth, 4, 4, 4); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth); + + RD::get_singleton()->draw_command_end_label(); + + RD::get_singleton()->compute_list_end(); + + RD::get_singleton()->texture_copy(rb->volumetric_fog->light_density_map, rb->volumetric_fog->prev_light_density_map, Vector3(0, 0, 0), Vector3(0, 0, 0), Vector3(rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth), 0, 0, 0, 0); - RD::get_singleton()->compute_list_add_barrier(compute_list); + compute_list = RD::get_singleton()->compute_list_begin(); if (use_filter) { + RD::get_singleton()->draw_command_begin_label("Filter Fog"); + RENDER_TIMESTAMP("Filter Fog"); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, volumetric_fog.pipelines[VOLUMETRIC_FOG_SHADER_FILTER]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->volumetric_fog->uniform_set, 0); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth); RD::get_singleton()->compute_list_end(); //need restart for buffer update params.filter_axis = 1; - RD::get_singleton()->buffer_update(volumetric_fog.params_ubo, 0, sizeof(VolumetricFogShader::ParamsUBO), ¶ms, true); + RD::get_singleton()->buffer_update(volumetric_fog.params_ubo, 0, sizeof(VolumetricFogShader::ParamsUBO), ¶ms); compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, volumetric_fog.pipelines[VOLUMETRIC_FOG_SHADER_FILTER]); @@ -7250,95 +7186,176 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e if (using_sdfgi) { RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->volumetric_fog->sdfgi_uniform_set, 1); } - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth); RD::get_singleton()->compute_list_add_barrier(compute_list); + RD::get_singleton()->draw_command_end_label(); } RENDER_TIMESTAMP("Integrate Fog"); + RD::get_singleton()->draw_command_begin_label("Integrate Fog"); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, volumetric_fog.pipelines[VOLUMETRIC_FOG_SHADER_FOG]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->volumetric_fog->uniform_set, 0); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, 1, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, 1); - RD::get_singleton()->compute_list_end(); + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_RASTER); RENDER_TIMESTAMP("<Volumetric Fog"); + RD::get_singleton()->draw_command_end_label(); + + rb->volumetric_fog->prev_cam_transform = p_cam_transform; } -void RendererSceneRenderRD::render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold) { - Color clear_color; - if (p_render_buffers.is_valid()) { - RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); - ERR_FAIL_COND(!rb); - clear_color = storage->render_target_get_clear_request_color(rb->render_target); - } else { - clear_color = storage->get_default_clear_color(); +uint32_t RendererSceneRenderRD::_get_render_state_directional_light_count() const { + return render_state.directional_light_count; +} + +bool RendererSceneRenderRD::_needs_post_prepass_render(bool p_use_gi) { + if (render_state.render_buffers.is_valid()) { + RenderBuffers *rb = render_buffers_owner.getornull(render_state.render_buffers); + if (rb->sdfgi != nullptr) { + return true; + } } + return false; +} - //assign render indices to giprobes - for (uint32_t i = 0; i < (uint32_t)p_gi_probes.size(); i++) { - GIProbeInstance *giprobe_inst = gi_probe_instance_owner.getornull(p_gi_probes[i]); - if (giprobe_inst) { - giprobe_inst->render_index = i; +void RendererSceneRenderRD::_post_prepass_render(bool p_use_gi) { + if (render_state.render_buffers.is_valid()) { + if (p_use_gi) { + _sdfgi_update_probes(render_state.render_buffers, render_state.environment); } } +} - const PagedArray<RID> *lights = &p_lights; - const PagedArray<RID> *reflections = &p_reflection_probes; - const PagedArray<RID> *gi_probes = &p_gi_probes; +void RendererSceneRenderRD::_pre_resolve_render(bool p_use_gi) { + if (render_state.render_buffers.is_valid()) { + if (p_use_gi) { + RD::get_singleton()->compute_list_end(); + } + } +} - PagedArray<RID> empty; +void RendererSceneRenderRD::_pre_opaque_render(bool p_use_ssao, bool p_use_gi, RID p_normal_roughness_buffer, RID p_gi_probe_buffer) { + // Render shadows while GI is rendering, due to how barriers are handled, this should happen at the same time - if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_UNSHADED) { - lights = ∅ - reflections = ∅ - gi_probes = ∅ + if (render_state.render_buffers.is_valid() && p_use_gi) { + _sdfgi_store_probes(render_state.render_buffers); } - if (render_buffers_owner.owns(p_render_buffers)) { - RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); - current_cluster_builder = rb->cluster_builder; - } else if (reflection_probe_instance_owner.owns(p_reflection_probe)) { - ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_reflection_probe); - ReflectionAtlas *ra = reflection_atlas_owner.getornull(rpi->atlas); - if (!ra) { - ERR_PRINT("reflection probe has no reflection atlas! Bug?"); - current_cluster_builder = nullptr; - } else { - current_cluster_builder = ra->cluster_builder; + render_state.cube_shadows.clear(); + render_state.shadows.clear(); + render_state.directional_shadows.clear(); + + Plane camera_plane(render_state.cam_transform.origin, -render_state.cam_transform.basis.get_axis(Vector3::AXIS_Z)); + float lod_distance_multiplier = render_state.cam_projection.get_lod_multiplier(); + + { + for (int i = 0; i < render_state.render_shadow_count; i++) { + LightInstance *li = light_instance_owner.getornull(render_state.render_shadows[i].light); + + if (storage->light_get_type(li->light) == RS::LIGHT_DIRECTIONAL) { + render_state.directional_shadows.push_back(i); + } else if (storage->light_get_type(li->light) == RS::LIGHT_OMNI && storage->light_omni_get_shadow_mode(li->light) == RS::LIGHT_OMNI_SHADOW_CUBE) { + render_state.cube_shadows.push_back(i); + } else { + render_state.shadows.push_back(i); + } } - } else { - ERR_PRINT("No cluster builder, bug"); //should never happen, will crash - current_cluster_builder = nullptr; + + //cube shadows are rendered in their own way + for (uint32_t i = 0; i < render_state.cube_shadows.size(); i++) { + _render_shadow_pass(render_state.render_shadows[render_state.cube_shadows[i]].light, render_state.shadow_atlas, render_state.render_shadows[render_state.cube_shadows[i]].pass, render_state.render_shadows[render_state.cube_shadows[i]].instances, camera_plane, lod_distance_multiplier, render_state.screen_lod_threshold, true, true, true); + } + + if (render_state.directional_shadows.size()) { + //open the pass for directional shadows + _update_directional_shadow_atlas(); + RD::get_singleton()->draw_list_begin(directional_shadow.fb, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_CONTINUE); + RD::get_singleton()->draw_list_end(); + } + } + + // Render GI + + bool render_shadows = render_state.directional_shadows.size() || render_state.shadows.size(); + bool render_gi = render_state.render_buffers.is_valid() && p_use_gi; + + if (render_shadows && render_gi) { + RENDER_TIMESTAMP("Render GI + Render Shadows (parallel)"); + } else if (render_shadows) { + RENDER_TIMESTAMP("Render Shadows"); + } else if (render_gi) { + RENDER_TIMESTAMP("Render GI"); } - current_cluster_builder->begin(p_cam_transform, p_cam_projection, !p_reflection_probe.is_valid()); + //prepare shadow rendering + if (render_shadows) { + _render_shadow_begin(); + + //render directional shadows + for (uint32_t i = 0; i < render_state.directional_shadows.size(); i++) { + _render_shadow_pass(render_state.render_shadows[render_state.directional_shadows[i]].light, render_state.shadow_atlas, render_state.render_shadows[render_state.directional_shadows[i]].pass, render_state.render_shadows[render_state.directional_shadows[i]].instances, camera_plane, lod_distance_multiplier, render_state.screen_lod_threshold, false, i == render_state.directional_shadows.size() - 1, false); + } + //render positional shadows + for (uint32_t i = 0; i < render_state.shadows.size(); i++) { + _render_shadow_pass(render_state.render_shadows[render_state.shadows[i]].light, render_state.shadow_atlas, render_state.render_shadows[render_state.shadows[i]].pass, render_state.render_shadows[render_state.shadows[i]].instances, camera_plane, lod_distance_multiplier, render_state.screen_lod_threshold, i == 0, i == render_state.shadows.size() - 1, true); + } + + _render_shadow_process(); + } + + //start GI + if (render_gi) { + _process_gi(render_state.render_buffers, p_normal_roughness_buffer, p_gi_probe_buffer, render_state.environment, render_state.cam_projection, render_state.cam_transform, *render_state.gi_probes); + } + + //Do shadow rendering (in parallel with GI) + if (render_shadows) { + _render_shadow_end(RD::BARRIER_MASK_NO_BARRIER); + } + + if (render_gi) { + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_NO_BARRIER); //use a later barrier + } + + if (render_state.render_buffers.is_valid()) { + if (p_use_ssao) { + _process_ssao(render_state.render_buffers, render_state.environment, p_normal_roughness_buffer, render_state.cam_projection); + } + } + + //full barrier here, we need raster, transfer and compute and it depends from the previous work + RD::get_singleton()->barrier(RD::BARRIER_MASK_ALL, RD::BARRIER_MASK_ALL); + + if (current_cluster_builder) { + current_cluster_builder->begin(render_state.cam_transform, render_state.cam_projection, !render_state.reflection_probe.is_valid()); + } bool using_shadows = true; - if (p_reflection_probe.is_valid()) { - if (!storage->reflection_probe_renders_shadows(reflection_probe_instance_get_probe(p_reflection_probe))) { + if (render_state.reflection_probe.is_valid()) { + if (!storage->reflection_probe_renders_shadows(reflection_probe_instance_get_probe(render_state.reflection_probe))) { using_shadows = false; } } else { //do not render reflections when rendering a reflection probe - _setup_reflections(*reflections, p_cam_transform.affine_inverse(), p_environment); + _setup_reflections(*render_state.reflection_probes, render_state.cam_transform.affine_inverse(), render_state.environment); } uint32_t directional_light_count = 0; uint32_t positional_light_count = 0; - _setup_lights(*lights, p_cam_transform, p_shadow_atlas, using_shadows, directional_light_count, positional_light_count); - _setup_decals(p_decals, p_cam_transform.affine_inverse()); + _setup_lights(*render_state.lights, render_state.cam_transform, render_state.shadow_atlas, using_shadows, directional_light_count, positional_light_count); + _setup_decals(*render_state.decals, render_state.cam_transform.affine_inverse()); - current_cluster_builder->bake_cluster(); + render_state.directional_light_count = directional_light_count; - uint32_t gi_probe_count = 0; - if (p_render_buffers.is_valid()) { - _setup_giprobes(p_render_buffers, p_cam_transform, *gi_probes, gi_probe_count); + if (current_cluster_builder) { + current_cluster_builder->bake_cluster(); } - if (p_render_buffers.is_valid()) { + if (render_state.render_buffers.is_valid()) { bool directional_shadows = false; for (uint32_t i = 0; i < directional_light_count; i++) { if (cluster.directional_lights[i].shadow_enabled) { @@ -7346,10 +7363,103 @@ void RendererSceneRenderRD::render_scene(RID p_render_buffers, const Transform & break; } } - _update_volumetric_fog(p_render_buffers, p_environment, p_cam_projection, p_cam_transform, p_shadow_atlas, directional_light_count, directional_shadows, positional_light_count, gi_probe_count); + _update_volumetric_fog(render_state.render_buffers, render_state.environment, render_state.cam_projection, render_state.cam_transform, render_state.shadow_atlas, directional_light_count, directional_shadows, positional_light_count, render_state.gi_probe_count); + } +} + +void RendererSceneRenderRD::render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold, const RenderShadowData *p_render_shadows, int p_render_shadow_count, const RenderSDFGIData *p_render_sdfgi_regions, int p_render_sdfgi_region_count, const RenderSDFGIUpdateData *p_sdfgi_update_data) { + //assign render data + { + render_state.render_buffers = p_render_buffers; + render_state.cam_transform = p_cam_transform; + render_state.cam_projection = p_cam_projection; + render_state.cam_ortogonal = p_cam_projection.is_orthogonal(); + render_state.instances = &p_instances; + render_state.lights = &p_lights; + render_state.reflection_probes = &p_reflection_probes; + render_state.gi_probes = &p_gi_probes; + render_state.decals = &p_decals; + render_state.lightmaps = &p_lightmaps; + render_state.environment = p_environment; + render_state.camera_effects = p_camera_effects; + render_state.shadow_atlas = p_shadow_atlas; + render_state.reflection_atlas = p_reflection_atlas; + render_state.reflection_probe = p_reflection_probe; + render_state.reflection_probe_pass = p_reflection_probe_pass; + render_state.screen_lod_threshold = p_screen_lod_threshold; + + render_state.render_shadows = p_render_shadows; + render_state.render_shadow_count = p_render_shadow_count; + render_state.render_sdfgi_regions = p_render_sdfgi_regions; + render_state.render_sdfgi_region_count = p_render_sdfgi_region_count; + render_state.sdfgi_update_data = p_sdfgi_update_data; + } + + PagedArray<RID> empty; + + if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_UNSHADED) { + render_state.lights = ∅ + render_state.reflection_probes = ∅ + render_state.gi_probes = ∅ + } + + //sdfgi first + if (p_render_buffers.is_valid()) { + for (int i = 0; i < render_state.render_sdfgi_region_count; i++) { + _render_sdfgi_region(p_render_buffers, render_state.render_sdfgi_regions[i].region, render_state.render_sdfgi_regions[i].instances); + } + if (render_state.sdfgi_update_data->update_static) { + _render_sdfgi_static_lights(p_render_buffers, render_state.sdfgi_update_data->static_cascade_count, p_sdfgi_update_data->static_cascade_indices, render_state.sdfgi_update_data->static_positional_lights); + } + } + + Color clear_color; + if (p_render_buffers.is_valid()) { + RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); + ERR_FAIL_COND(!rb); + clear_color = storage->render_target_get_clear_request_color(rb->render_target); + } else { + clear_color = storage->get_default_clear_color(); } - _render_scene(p_render_buffers, p_cam_transform, p_cam_projection, p_cam_ortogonal, p_instances, directional_light_count, *gi_probes, p_lightmaps, p_environment, current_cluster_builder->get_cluster_buffer(), current_cluster_builder->get_cluster_size(), current_cluster_builder->get_max_cluster_elements(), p_camera_effects, p_shadow_atlas, p_reflection_atlas, p_reflection_probe, p_reflection_probe_pass, clear_color, p_screen_lod_threshold); + //assign render indices to giprobes + for (uint32_t i = 0; i < (uint32_t)p_gi_probes.size(); i++) { + GIProbeInstance *giprobe_inst = gi_probe_instance_owner.getornull(p_gi_probes[i]); + if (giprobe_inst) { + giprobe_inst->render_index = i; + } + } + + if (render_buffers_owner.owns(render_state.render_buffers)) { + RenderBuffers *rb = render_buffers_owner.getornull(render_state.render_buffers); + current_cluster_builder = rb->cluster_builder; + } else if (reflection_probe_instance_owner.owns(render_state.reflection_probe)) { + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(render_state.reflection_probe); + ReflectionAtlas *ra = reflection_atlas_owner.getornull(rpi->atlas); + if (!ra) { + ERR_PRINT("reflection probe has no reflection atlas! Bug?"); + current_cluster_builder = nullptr; + } else { + current_cluster_builder = ra->cluster_builder; + } + } else { + ERR_PRINT("No cluster builder, bug"); //should never happen, will crash + current_cluster_builder = nullptr; + } + + if (p_render_buffers.is_valid()) { + _pre_process_gi(p_render_buffers, p_cam_transform); + } + + render_state.gi_probe_count = 0; + if (render_state.render_buffers.is_valid()) { + _setup_giprobes(render_state.render_buffers, render_state.cam_transform, *render_state.gi_probes, render_state.gi_probe_count); + _sdfgi_update_light(render_state.render_buffers, render_state.environment); + } + + render_state.depth_prepass_used = false; + //calls _pre_opaque_render between depth pre-pass and opaque pass + _render_scene(p_render_buffers, p_cam_transform, p_cam_projection, p_cam_ortogonal, p_instances, *render_state.gi_probes, p_lightmaps, p_environment, current_cluster_builder->get_cluster_buffer(), current_cluster_builder->get_cluster_size(), current_cluster_builder->get_max_cluster_elements(), p_camera_effects, p_shadow_atlas, p_reflection_atlas, p_reflection_probe, p_reflection_probe_pass, clear_color, p_screen_lod_threshold); if (p_render_buffers.is_valid()) { if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_OMNI_LIGHTS || debug_draw == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_SPOT_LIGHTS || debug_draw == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_DECALS || debug_draw == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_REFLECTION_PROBES) { @@ -7383,27 +7493,26 @@ void RendererSceneRenderRD::render_scene(RID p_render_buffers, const Transform & } } -void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<GeometryInstance *> &p_instances, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold) { +void RendererSceneRenderRD::_render_shadow_pass(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<GeometryInstance *> &p_instances, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold, bool p_open_pass, bool p_close_pass, bool p_clear_region) { LightInstance *light_instance = light_instance_owner.getornull(p_light); ERR_FAIL_COND(!light_instance); Rect2i atlas_rect; - RID atlas_texture; + uint32_t atlas_size; + RID atlas_fb; bool using_dual_paraboloid = false; bool using_dual_paraboloid_flip = false; - float znear = 0; - float zfar = 0; RID render_fb; RID render_texture; - float bias = 0; - float normal_bias = 0; + float zfar; bool use_pancake = false; - bool use_linear_depth = false; bool render_cubemap = false; bool finalize_cubemap = false; + bool flip_y = false; + CameraMatrix light_projection; Transform light_transform; @@ -7436,7 +7545,6 @@ void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p atlas_rect.position.x += atlas_rect.size.width; atlas_rect.position.y += atlas_rect.size.height; } - } else if (storage->light_directional_get_shadow_mode(light_instance->light) == RS::LIGHT_DIRECTIONAL_SHADOW_PARALLEL_2_SPLITS) { atlas_rect.size.height /= 2; @@ -7451,15 +7559,11 @@ void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p light_instance->shadow_transform[p_pass].atlas_rect.position /= directional_shadow.size; light_instance->shadow_transform[p_pass].atlas_rect.size /= directional_shadow.size; - float bias_mult = light_instance->shadow_transform[p_pass].bias_scale; zfar = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_RANGE); - bias = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_SHADOW_BIAS) * bias_mult; - normal_bias = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * bias_mult; - ShadowMap *shadow_map = _get_shadow_map(atlas_rect.size); - render_fb = shadow_map->fb; - render_texture = shadow_map->depth; - atlas_texture = directional_shadow.depth; + render_fb = directional_shadow.fb; + render_texture = RID(); + flip_y = true; } else { //set from shadow atlas @@ -7468,6 +7572,8 @@ void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p ERR_FAIL_COND(!shadow_atlas); ERR_FAIL_COND(!shadow_atlas->shadow_owners.has(p_light)); + _update_shadow_atlas(shadow_atlas); + uint32_t key = shadow_atlas->shadow_owners[p_light]; uint32_t quadrant = (key >> ShadowAtlas::QUADRANT_SHIFT) & 0x3; @@ -7486,11 +7592,8 @@ void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p atlas_rect.size.width = shadow_size; atlas_rect.size.height = shadow_size; - atlas_texture = shadow_atlas->depth; zfar = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_RANGE); - bias = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_SHADOW_BIAS); - normal_bias = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS); if (storage->light_get_type(light_instance->light) == RS::LIGHT_OMNI) { if (storage->light_omni_get_shadow_mode(light_instance->light) == RS::LIGHT_OMNI_SHADOW_CUBE) { @@ -7499,10 +7602,17 @@ void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p render_fb = cubemap->side_fb[p_pass]; render_texture = cubemap->cubemap; - light_projection = light_instance->shadow_transform[0].camera; - light_transform = light_instance->shadow_transform[0].transform; + light_projection = light_instance->shadow_transform[p_pass].camera; + light_transform = light_instance->shadow_transform[p_pass].transform; render_cubemap = true; finalize_cubemap = p_pass == 5; + atlas_fb = shadow_atlas->fb; + + atlas_size = shadow_atlas->size; + + if (p_pass == 0) { + _render_shadow_begin(); + } } else { light_projection = light_instance->shadow_transform[0].camera; @@ -7513,49 +7623,44 @@ void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p using_dual_paraboloid = true; using_dual_paraboloid_flip = p_pass == 1; - - ShadowMap *shadow_map = _get_shadow_map(atlas_rect.size); - render_fb = shadow_map->fb; - render_texture = shadow_map->depth; + render_fb = shadow_atlas->fb; + flip_y = true; } } else if (storage->light_get_type(light_instance->light) == RS::LIGHT_SPOT) { light_projection = light_instance->shadow_transform[0].camera; light_transform = light_instance->shadow_transform[0].transform; - ShadowMap *shadow_map = _get_shadow_map(atlas_rect.size); - render_fb = shadow_map->fb; - render_texture = shadow_map->depth; + render_fb = shadow_atlas->fb; - znear = light_instance->shadow_transform[0].camera.get_z_near(); - use_linear_depth = true; + flip_y = true; } } if (render_cubemap) { //rendering to cubemap - _render_shadow(render_fb, p_instances, light_projection, light_transform, zfar, 0, 0, false, false, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold); + _render_shadow_append(render_fb, p_instances, light_projection, light_transform, zfar, 0, 0, false, false, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold, Rect2(), false, true, true, true); if (finalize_cubemap) { + _render_shadow_process(); + _render_shadow_end(); //reblit - atlas_rect.size.height /= 2; - storage->get_effects()->copy_cubemap_to_dp(render_texture, atlas_texture, atlas_rect, light_projection.get_z_near(), light_projection.get_z_far(), 0.0, false); - atlas_rect.position.y += atlas_rect.size.height; - storage->get_effects()->copy_cubemap_to_dp(render_texture, atlas_texture, atlas_rect, light_projection.get_z_near(), light_projection.get_z_far(), 0.0, true); - } - } else { - //render shadow - - _render_shadow(render_fb, p_instances, light_projection, light_transform, zfar, bias, normal_bias, using_dual_paraboloid, using_dual_paraboloid_flip, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold); + Rect2 atlas_rect_norm = atlas_rect; + atlas_rect_norm.position.x /= float(atlas_size); + atlas_rect_norm.position.y /= float(atlas_size); + atlas_rect_norm.size.x /= float(atlas_size); + atlas_rect_norm.size.y /= float(atlas_size); + atlas_rect_norm.size.height /= 2; + storage->get_effects()->copy_cubemap_to_dp(render_texture, atlas_fb, atlas_rect_norm, light_projection.get_z_near(), light_projection.get_z_far(), false); + atlas_rect_norm.position.y += atlas_rect_norm.size.height; + storage->get_effects()->copy_cubemap_to_dp(render_texture, atlas_fb, atlas_rect_norm, light_projection.get_z_near(), light_projection.get_z_far(), true); - //copy to atlas - if (use_linear_depth) { - storage->get_effects()->copy_depth_to_rect_and_linearize(render_texture, atlas_texture, atlas_rect, true, znear, zfar); - } else { - storage->get_effects()->copy_depth_to_rect(render_texture, atlas_texture, atlas_rect, true); + //restore transform so it can be properly used + light_instance_set_shadow_transform(p_light, CameraMatrix(), light_instance->transform, zfar, 0, 0, 0); } - //does not work from depth to color - //RD::get_singleton()->texture_copy(render_texture, atlas_texture, Vector3(0, 0, 0), Vector3(atlas_rect.position.x, atlas_rect.position.y, 0), Vector3(atlas_rect.size.x, atlas_rect.size.y, 1), 0, 0, 0, 0, true); + } else { + //render shadow + _render_shadow_append(render_fb, p_instances, light_projection, light_transform, zfar, 0, 0, using_dual_paraboloid, using_dual_paraboloid_flip, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold, atlas_rect, flip_y, p_clear_region, p_open_pass, p_close_pass); } } @@ -7563,7 +7668,7 @@ void RendererSceneRenderRD::render_material(const Transform &p_cam_transform, co _render_material(p_cam_transform, p_cam_projection, p_cam_ortogonal, p_instances, p_framebuffer, p_region); } -void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, const PagedArray<GeometryInstance *> &p_instances) { +void RendererSceneRenderRD::_render_sdfgi_region(RID p_render_buffers, int p_region, const PagedArray<GeometryInstance *> &p_instances) { //print_line("rendering region " + itos(p_region)); RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); ERR_FAIL_COND(!rb); @@ -7579,16 +7684,18 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con 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); + RD::get_singleton()->texture_clear(rb->sdfgi->render_albedo, Color(0, 0, 0, 0), 0, 1, 0, 1); + RD::get_singleton()->texture_clear(rb->sdfgi->render_emission, Color(0, 0, 0, 0), 0, 1, 0, 1); + RD::get_singleton()->texture_clear(rb->sdfgi->render_emission_aniso, Color(0, 0, 0, 0), 0, 1, 0, 1); + RD::get_singleton()->texture_clear(rb->sdfgi->render_geom_facing, Color(0, 0, 0, 0), 0, 1, 0, 1); } //print_line("rendering cascade " + itos(p_region) + " objects: " + itos(p_cull_count) + " bounds: " + bounds + " from: " + from + " size: " + size + " cell size: " + rtos(rb->sdfgi->cascades[cascade].cell_size)); _render_sdfgi(p_render_buffers, from, size, bounds, p_instances, rb->sdfgi->render_albedo, rb->sdfgi->render_emission, rb->sdfgi->render_emission_aniso, rb->sdfgi->render_geom_facing); if (cascade_next != cascade) { + RD::get_singleton()->draw_command_begin_label("SDFGI Pre-Process Cascade"); + RENDER_TIMESTAMP(">SDFGI Update SDF"); //done rendering! must update SDF //clear dispatch indirect data @@ -7611,6 +7718,9 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con push_constant.scroll[1] = 0; push_constant.scroll[2] = 0; } + + rb->sdfgi->cascades[cascade].all_dynamic_lights_dirty = true; + push_constant.grid_size = rb->sdfgi->cascade_size; push_constant.cascade = cascade; @@ -7635,7 +7745,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con 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); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, groups.x, groups.y, groups.z); //no barrier, continue together @@ -7677,7 +7787,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con 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_dispatch_threads(compute_list, rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count, rb->sdfgi->probe_axis_count, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); @@ -7685,7 +7795,24 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con 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_dispatch_threads(compute_list, rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count, rb->sdfgi->probe_axis_count, 1); + + RD::get_singleton()->compute_list_add_barrier(compute_list); + + if (rb->sdfgi->bounce_feedback > 0.0) { + //multibounce requires this to be stored so direct light can read from it + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.integrate_pipeline[SDGIShader::INTEGRATE_MODE_STORE]); + + //convert to octahedral to store + ipush_constant.image_size[0] *= SDFGI::LIGHTPROBE_OCT_SIZE; + ipush_constant.image_size[1] *= SDFGI::LIGHTPROBE_OCT_SIZE; + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, 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 * SDFGI::LIGHTPROBE_OCT_SIZE, rb->sdfgi->probe_axis_count * SDFGI::LIGHTPROBE_OCT_SIZE, 1); + } } //ok finally barrier @@ -7694,7 +7821,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con //clear dispatch indirect data uint32_t dispatch_indirct_data[4] = { 0, 0, 0, 0 }; - RD::get_singleton()->buffer_update(rb->sdfgi->cascades[cascade].solid_cell_dispatch_buffer, 0, sizeof(uint32_t) * 4, dispatch_indirct_data, true); + RD::get_singleton()->buffer_update(rb->sdfgi->cascades[cascade].solid_cell_dispatch_buffer, 0, sizeof(uint32_t) * 4, dispatch_indirct_data); RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); @@ -7708,7 +7835,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con 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_dispatch_threads(compute_list, cascade_half_size, cascade_half_size, cascade_half_size); RD::get_singleton()->compute_list_add_barrier(compute_list); //must start with regular jumpflood @@ -7728,7 +7855,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con 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_dispatch_threads(compute_list, cascade_half_size, cascade_half_size, cascade_half_size); RD::get_singleton()->compute_list_add_barrier(compute_list); jf_us = jf_us == 0 ? 1 : 0; @@ -7746,7 +7873,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con 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_dispatch_threads(compute_list, cascade_half_size, cascade_half_size, cascade_half_size); RD::get_singleton()->compute_list_add_barrier(compute_list); jf_us = jf_us == 0 ? 1 : 0; } @@ -7758,7 +7885,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con 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_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size); RD::get_singleton()->compute_list_add_barrier(compute_list); //run one pass of fullsize jumpflood to fix up half size arctifacts @@ -7768,7 +7895,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con 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_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size); RD::get_singleton()->compute_list_add_barrier(compute_list); } else { @@ -7778,7 +7905,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con 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_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size); RD::get_singleton()->compute_list_add_barrier(compute_list); @@ -7795,7 +7922,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con 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_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size); RD::get_singleton()->compute_list_add_barrier(compute_list); jf_us = jf_us == 0 ? 1 : 0; @@ -7813,7 +7940,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con 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_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size); RD::get_singleton()->compute_list_add_barrier(compute_list); jf_us = jf_us == 0 ? 1 : 0; } @@ -7860,14 +7987,14 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con 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_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size); RD::get_singleton()->compute_list_end(); //clear these textures, as they will have previous garbage on next draw - RD::get_singleton()->texture_clear(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); + RD::get_singleton()->texture_clear(rb->sdfgi->cascades[cascade].light_tex, Color(0, 0, 0, 0), 0, 1, 0, 1); + RD::get_singleton()->texture_clear(rb->sdfgi->cascades[cascade].light_aniso_0_tex, Color(0, 0, 0, 0), 0, 1, 0, 1); + RD::get_singleton()->texture_clear(rb->sdfgi->cascades[cascade].light_aniso_1_tex, Color(0, 0, 0, 0), 0, 1, 0, 1); #if 0 Vector<uint8_t> data = RD::get_singleton()->texture_get_data(rb->sdfgi->cascades[cascade].sdf, 0); @@ -7897,6 +8024,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, con #endif RENDER_TIMESTAMP("<SDFGI Update SDF"); + RD::get_singleton()->draw_command_end_label(); } } @@ -7917,32 +8045,17 @@ void RendererSceneRenderRD::render_particle_collider_heightfield(RID p_collider, _render_particle_collider_heightfield(fb, cam_xform, cm, p_instances); } -void RendererSceneRenderRD::render_sdfgi_static_lights(RID p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const PagedArray<RID> *p_positional_light_cull_result) { +void RendererSceneRenderRD::_render_sdfgi_static_lights(RID p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const PagedArray<RID> *p_positional_light_cull_result) { RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); ERR_FAIL_COND(!rb); ERR_FAIL_COND(!rb->sdfgi); - _sdfgi_update_cascades(p_render_buffers); //need cascades updated for this + RD::get_singleton()->draw_command_begin_label("SDFGI Render Static Lighs"); - 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; + _sdfgi_update_cascades(p_render_buffers); //need cascades updated for this SDGIShader::Light lights[SDFGI::MAX_STATIC_LIGHTS]; + uint32_t light_count[SDFGI::MAX_STATIC_LIGHTS]; for (uint32_t i = 0; i < p_cascade_count; i++) { ERR_CONTINUE(p_cascade_indices[i] >= rb->sdfgi->cascades.size()); @@ -7998,18 +8111,46 @@ void RendererSceneRenderRD::render_sdfgi_static_lights(RID p_render_buffers, uin 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); + lights[idx].cos_spot_angle = Math::cos(Math::deg2rad(storage->light_get_param(li->light, RS::LIGHT_PARAM_SPOT_ANGLE))); + lights[idx].inv_spot_attenuation = 1.0f / storage->light_get_param(li->light, RS::LIGHT_PARAM_SPOT_ATTENUATION); idx++; } if (idx > 0) { - RD::get_singleton()->buffer_update(cc.lights_buffer, 0, idx * sizeof(SDGIShader::Light), lights, true); + RD::get_singleton()->buffer_update(cc.lights_buffer, 0, idx * sizeof(SDGIShader::Light), lights); } - dl_push_constant.light_count = idx; + + light_count[i] = idx; } + } + /* Static Lights */ + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, 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.bounce_feedback = 0.0; // this is static light, do not multibounce yet + dl_push_constant.y_mult = rb->sdfgi->y_mult; + dl_push_constant.use_occlusion = rb->sdfgi->uses_occlusion; + + //all must be processed + dl_push_constant.process_offset = 0; + dl_push_constant.process_increment = 1; + + for (uint32_t i = 0; i < p_cascade_count; i++) { + ERR_CONTINUE(p_cascade_indices[i] >= rb->sdfgi->cascades.size()); + + SDFGI::Cascade &cc = rb->sdfgi->cascades[p_cascade_indices[i]]; + + dl_push_constant.light_count = light_count[i]; dl_push_constant.cascade = p_cascade_indices[i]; if (dl_push_constant.light_count > 0) { @@ -8020,6 +8161,8 @@ void RendererSceneRenderRD::render_sdfgi_static_lights(RID p_render_buffers, uin } RD::get_singleton()->compute_list_end(); + + RD::get_singleton()->draw_command_end_label(); } bool RendererSceneRenderRD::free(RID p_rid) { @@ -8293,19 +8436,25 @@ bool RendererSceneRenderRD::is_low_end() const { } RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { - max_cluster_elements = GLOBAL_GET("rendering/cluster_builder/max_clustered_elements"); + max_cluster_elements = GLOBAL_GET("rendering/limits/cluster_builder/max_clustered_elements"); storage = p_storage; singleton = this; - roughness_layers = GLOBAL_GET("rendering/quality/reflections/roughness_layers"); - sky_ggx_samples_quality = GLOBAL_GET("rendering/quality/reflections/ggx_samples"); - sky_use_cubemap_array = GLOBAL_GET("rendering/quality/reflections/texture_array_reflections"); - // sky_use_cubemap_array = false; + roughness_layers = GLOBAL_GET("rendering/reflections/sky_reflections/roughness_layers"); + sky_ggx_samples_quality = GLOBAL_GET("rendering/reflections/sky_reflections/ggx_samples"); + sky_use_cubemap_array = GLOBAL_GET("rendering/reflections/sky_reflections/texture_array_reflections"); + + sdfgi_ray_count = RS::EnvironmentSDFGIRayCount(CLAMP(int32_t(GLOBAL_GET("rendering/global_illumination/sdfgi/probe_ray_count")), 0, int32_t(RS::ENV_SDFGI_RAY_COUNT_MAX - 1))); + sdfgi_frames_to_converge = RS::EnvironmentSDFGIFramesToConverge(CLAMP(int32_t(GLOBAL_GET("rendering/global_illumination/sdfgi/frames_to_converge")), 0, int32_t(RS::ENV_SDFGI_CONVERGE_MAX - 1))); + sdfgi_frames_to_update_light = RS::EnvironmentSDFGIFramesToUpdateLight(CLAMP(int32_t(GLOBAL_GET("rendering/global_illumination/sdfgi/frames_to_update_lights")), 0, int32_t(RS::ENV_SDFGI_UPDATE_LIGHT_MAX - 1))); + + directional_shadow.size = GLOBAL_GET("rendering/shadows/directional_shadow/size"); + directional_shadow.use_16_bits = GLOBAL_GET("rendering/shadows/directional_shadow/16_bits"); uint32_t textures_per_stage = RD::get_singleton()->limit_get(RD::LIMIT_MAX_TEXTURES_PER_SHADER_STAGE); - low_end = GLOBAL_GET("rendering/quality/rd_renderer/use_low_end_renderer"); + low_end = GLOBAL_GET("rendering/driver/rd_renderer/use_low_end_renderer"); if (textures_per_stage < 48) { low_end = true; @@ -8318,7 +8467,7 @@ RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *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_quality = RS::GIProbeQuality(CLAMP(int(GLOBAL_GET("rendering/quality/gi_probes/quality")), 0, 1)); + gi_probe_quality = RS::GIProbeQuality(CLAMP(int(GLOBAL_GET("rendering/global_illumination/gi_probes/quality")), 0, 1)); String defines = "\n#define MAX_LIGHTS " + itos(gi_probe_max_lights) + "\n"; @@ -8449,9 +8598,14 @@ RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { { // default material and shader for sky shader - sky_shader.default_shader = storage->shader_create(); + sky_shader.default_shader = storage->shader_allocate(); + storage->shader_initialize(sky_shader.default_shader); + storage->shader_set_code(sky_shader.default_shader, "shader_type sky; void fragment() { COLOR = vec3(0.0); } \n"); - sky_shader.default_material = storage->material_create(); + + sky_shader.default_material = storage->material_allocate(); + storage->material_initialize(sky_shader.default_material); + storage->material_set_shader(sky_shader.default_material, sky_shader.default_shader); SkyMaterialData *md = (SkyMaterialData *)storage->material_get_data(sky_shader.default_material, RendererStorageRD::SHADER_TYPE_SKY); @@ -8525,9 +8679,13 @@ RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { { // Need defaults for using fog with clear color - sky_scene_state.fog_shader = storage->shader_create(); + sky_scene_state.fog_shader = storage->shader_allocate(); + storage->shader_initialize(sky_scene_state.fog_shader); + storage->shader_set_code(sky_scene_state.fog_shader, "shader_type sky; uniform vec4 clear_color; void fragment() { COLOR = clear_color.rgb; } \n"); - sky_scene_state.fog_material = storage->material_create(); + sky_scene_state.fog_material = storage->material_allocate(); + storage->material_initialize(sky_scene_state.fog_material); + storage->material_set_shader(sky_scene_state.fog_material, sky_scene_state.fog_shader); Vector<RD::Uniform> uniforms; @@ -8595,6 +8753,9 @@ RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { //calculate tables String defines = "\n#define OCT_SIZE " + itos(SDFGI::LIGHTPROBE_OCT_SIZE) + "\n"; defines += "\n#define SH_SIZE " + itos(SDFGI::SH_SIZE) + "\n"; + if (sky_use_cubemap_array) { + defines += "\n#define USE_CUBEMAP_ARRAY\n"; + } Vector<String> integrate_modes; integrate_modes.push_back("\n#define MODE_PROCESS\n"); @@ -8634,9 +8795,12 @@ RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { //calculate tables String defines = "\n#define SDFGI_OCT_SIZE " + itos(SDFGI::LIGHTPROBE_OCT_SIZE) + "\n"; Vector<String> gi_modes; - gi_modes.push_back("\n#define USE_GIPROBE\n"); + gi_modes.push_back("\n#define USE_GIPROBES\n"); gi_modes.push_back("\n#define USE_SDFGI\n"); - gi_modes.push_back("\n#define USE_SDFGI\n\n#define USE_GIPROBE\n"); + gi_modes.push_back("\n#define USE_SDFGI\n\n#define USE_GIPROBES\n"); + gi_modes.push_back("\n#define MODE_HALF_RES\n#define USE_GIPROBES\n"); + gi_modes.push_back("\n#define MODE_HALF_RES\n#define USE_SDFGI\n"); + gi_modes.push_back("\n#define MODE_HALF_RES\n#define USE_SDFGI\n\n#define USE_GIPROBES\n"); gi.shader.initialize(gi_modes, defines); gi.shader_version = gi.shader.version_create(); @@ -8681,11 +8845,19 @@ RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { default_giprobe_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(GI::GIProbeData) * RenderBuffers::MAX_GIPROBES); } + { //decals + cluster.max_decals = max_cluster_elements; + uint32_t decal_buffer_size = cluster.max_decals * sizeof(Cluster::DecalData); + cluster.decals = memnew_arr(Cluster::DecalData, cluster.max_decals); + cluster.decal_sort = memnew_arr(Cluster::InstanceSort<DecalInstance>, cluster.max_decals); + cluster.decal_buffer = RD::get_singleton()->storage_buffer_create(decal_buffer_size); + } + { //reflections cluster.max_reflections = max_cluster_elements; cluster.reflections = memnew_arr(Cluster::ReflectionData, cluster.max_reflections); - cluster.reflection_sort = memnew_arr(Cluster::InstanceSort<ReflectionProbeInstance>, cluster.max_decals); + cluster.reflection_sort = memnew_arr(Cluster::InstanceSort<ReflectionProbeInstance>, cluster.max_reflections); cluster.reflection_buffer = RD::get_singleton()->storage_buffer_create(sizeof(Cluster::ReflectionData) * cluster.max_reflections); } @@ -8701,24 +8873,12 @@ RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { cluster.spot_light_sort = memnew_arr(Cluster::InstanceSort<LightInstance>, cluster.max_lights); //defines += "\n#define MAX_LIGHT_DATA_STRUCTS " + itos(cluster.max_lights) + "\n"; - //used for volumetric fog shrinking - cluster.lights_instances = memnew_arr(RID, cluster.max_lights * 2); - cluster.lights_shadow_rect_cache = memnew_arr(Rect2i, cluster.max_lights * 2); - cluster.max_directional_lights = MAX_DIRECTIONAL_LIGHTS; uint32_t directional_light_buffer_size = cluster.max_directional_lights * sizeof(Cluster::DirectionalLightData); cluster.directional_lights = memnew_arr(Cluster::DirectionalLightData, cluster.max_directional_lights); cluster.directional_light_buffer = RD::get_singleton()->uniform_buffer_create(directional_light_buffer_size); } - { //decals - cluster.max_decals = max_cluster_elements; - uint32_t decal_buffer_size = cluster.max_decals * sizeof(Cluster::DecalData); - cluster.decals = memnew_arr(Cluster::DecalData, cluster.max_decals); - cluster.decal_sort = memnew_arr(Cluster::InstanceSort<DecalInstance>, cluster.max_decals); - cluster.decal_buffer = RD::get_singleton()->storage_buffer_create(decal_buffer_size); - } - if (!low_end) { String defines = "\n#define MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS " + itos(cluster.max_directional_lights) + "\n"; Vector<String> volumetric_fog_modes; @@ -8743,37 +8903,34 @@ RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { shadow_sampler = RD::get_singleton()->sampler_create(sampler); } - camera_effects_set_dof_blur_bokeh_shape(RS::DOFBokehShape(int(GLOBAL_GET("rendering/quality/depth_of_field/depth_of_field_bokeh_shape")))); - camera_effects_set_dof_blur_quality(RS::DOFBlurQuality(int(GLOBAL_GET("rendering/quality/depth_of_field/depth_of_field_bokeh_quality"))), GLOBAL_GET("rendering/quality/depth_of_field/depth_of_field_use_jitter")); - environment_set_ssao_quality(RS::EnvironmentSSAOQuality(int(GLOBAL_GET("rendering/quality/ssao/quality"))), GLOBAL_GET("rendering/quality/ssao/half_size"), GLOBAL_GET("rendering/quality/ssao/adaptive_target"), GLOBAL_GET("rendering/quality/ssao/blur_passes"), GLOBAL_GET("rendering/quality/ssao/fadeout_from"), GLOBAL_GET("rendering/quality/ssao/fadeout_to")); - screen_space_roughness_limiter = GLOBAL_GET("rendering/quality/screen_filters/screen_space_roughness_limiter_enabled"); - screen_space_roughness_limiter_amount = GLOBAL_GET("rendering/quality/screen_filters/screen_space_roughness_limiter_amount"); - screen_space_roughness_limiter_limit = GLOBAL_GET("rendering/quality/screen_filters/screen_space_roughness_limiter_limit"); - glow_bicubic_upscale = int(GLOBAL_GET("rendering/quality/glow/upscale_mode")) > 0; - glow_high_quality = GLOBAL_GET("rendering/quality/glow/use_high_quality"); - ssr_roughness_quality = RS::EnvironmentSSRRoughnessQuality(int(GLOBAL_GET("rendering/quality/screen_space_reflection/roughness_quality"))); - sss_quality = RS::SubSurfaceScatteringQuality(int(GLOBAL_GET("rendering/quality/subsurface_scattering/subsurface_scattering_quality"))); - sss_scale = GLOBAL_GET("rendering/quality/subsurface_scattering/subsurface_scattering_scale"); - sss_depth_scale = GLOBAL_GET("rendering/quality/subsurface_scattering/subsurface_scattering_depth_scale"); + camera_effects_set_dof_blur_bokeh_shape(RS::DOFBokehShape(int(GLOBAL_GET("rendering/camera/depth_of_field/depth_of_field_bokeh_shape")))); + camera_effects_set_dof_blur_quality(RS::DOFBlurQuality(int(GLOBAL_GET("rendering/camera/depth_of_field/depth_of_field_bokeh_quality"))), GLOBAL_GET("rendering/camera/depth_of_field/depth_of_field_use_jitter")); + environment_set_ssao_quality(RS::EnvironmentSSAOQuality(int(GLOBAL_GET("rendering/environment/ssao/quality"))), GLOBAL_GET("rendering/environment/ssao/half_size"), GLOBAL_GET("rendering/environment/ssao/adaptive_target"), GLOBAL_GET("rendering/environment/ssao/blur_passes"), GLOBAL_GET("rendering/environment/ssao/fadeout_from"), GLOBAL_GET("rendering/environment/ssao/fadeout_to")); + screen_space_roughness_limiter = GLOBAL_GET("rendering/anti_aliasing/screen_space_roughness_limiter/enabled"); + screen_space_roughness_limiter_amount = GLOBAL_GET("rendering/anti_aliasing/screen_space_roughness_limiter/amount"); + screen_space_roughness_limiter_limit = GLOBAL_GET("rendering/anti_aliasing/screen_space_roughness_limiter/limit"); + glow_bicubic_upscale = int(GLOBAL_GET("rendering/environment/glow/upscale_mode")) > 0; + glow_high_quality = GLOBAL_GET("rendering/environment/glow/use_high_quality"); + ssr_roughness_quality = RS::EnvironmentSSRRoughnessQuality(int(GLOBAL_GET("rendering/environment/screen_space_reflection/roughness_quality"))); + sss_quality = RS::SubSurfaceScatteringQuality(int(GLOBAL_GET("rendering/environment/subsurface_scattering/subsurface_scattering_quality"))); + sss_scale = GLOBAL_GET("rendering/environment/subsurface_scattering/subsurface_scattering_scale"); + sss_depth_scale = GLOBAL_GET("rendering/environment/subsurface_scattering/subsurface_scattering_depth_scale"); directional_penumbra_shadow_kernel = memnew_arr(float, 128); directional_soft_shadow_kernel = memnew_arr(float, 128); penumbra_shadow_kernel = memnew_arr(float, 128); soft_shadow_kernel = memnew_arr(float, 128); - shadows_quality_set(RS::ShadowQuality(int(GLOBAL_GET("rendering/quality/shadows/soft_shadow_quality")))); - directional_shadow_quality_set(RS::ShadowQuality(int(GLOBAL_GET("rendering/quality/directional_shadow/soft_shadow_quality")))); + shadows_quality_set(RS::ShadowQuality(int(GLOBAL_GET("rendering/shadows/shadows/soft_shadow_quality")))); + directional_shadow_quality_set(RS::ShadowQuality(int(GLOBAL_GET("rendering/shadows/directional_shadow/soft_shadow_quality")))); - environment_set_volumetric_fog_volume_size(GLOBAL_GET("rendering/volumetric_fog/volume_size"), GLOBAL_GET("rendering/volumetric_fog/volume_depth")); - environment_set_volumetric_fog_filter_active(GLOBAL_GET("rendering/volumetric_fog/use_filter")); - environment_set_volumetric_fog_directional_shadow_shrink_size(GLOBAL_GET("rendering/volumetric_fog/directional_shadow_shrink")); - environment_set_volumetric_fog_positional_shadow_shrink_size(GLOBAL_GET("rendering/volumetric_fog/positional_shadow_shrink")); + environment_set_volumetric_fog_volume_size(GLOBAL_GET("rendering/environment/volumetric_fog/volume_size"), GLOBAL_GET("rendering/environment/volumetric_fog/volume_depth")); + environment_set_volumetric_fog_filter_active(GLOBAL_GET("rendering/environment/volumetric_fog/use_filter")); cull_argument.set_page_pool(&cull_argument_pool); + + gi.half_resolution = GLOBAL_GET("rendering/global_illumination/gi/use_half_resolution"); } RendererSceneRenderRD::~RendererSceneRenderRD() { - for (Map<Vector2i, ShadowMap>::Element *E = shadow_maps.front(); E; E = E->next()) { - RD::get_singleton()->free(E->get().depth); - } for (Map<int, ShadowCubemap>::Element *E = shadow_cubemaps.front(); E; E = E->next()) { RD::get_singleton()->free(E->get().cubemap); } @@ -8828,8 +8985,6 @@ RendererSceneRenderRD::~RendererSceneRenderRD() { memdelete_arr(cluster.spot_lights); memdelete_arr(cluster.omni_light_sort); memdelete_arr(cluster.spot_light_sort); - memdelete_arr(cluster.lights_shadow_rect_cache); - memdelete_arr(cluster.lights_instances); memdelete_arr(cluster.reflections); memdelete_arr(cluster.reflection_sort); memdelete_arr(cluster.decals); diff --git a/servers/rendering/renderer_rd/renderer_scene_render_rd.h b/servers/rendering/renderer_rd/renderer_scene_render_rd.h index 3e69335225..e4eaa93212 100644 --- a/servers/rendering/renderer_rd/renderer_scene_render_rd.h +++ b/servers/rendering/renderer_rd/renderer_scene_render_rd.h @@ -109,8 +109,13 @@ protected: void _setup_reflections(const PagedArray<RID> &p_reflections, const Transform &p_camera_inverse_transform, RID p_environment); void _setup_giprobes(RID p_render_buffers, const Transform &p_transform, const PagedArray<RID> &p_gi_probes, uint32_t &r_gi_probes_used); - virtual void _render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, int p_directional_light_count, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_cluster_buffer, uint32_t p_cluster_size, uint32_t p_cluster_max_elements, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_color, float p_screen_lod_threshold) = 0; - virtual void _render_shadow(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0) = 0; + virtual void _render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_cluster_buffer, uint32_t p_cluster_size, uint32_t p_cluster_max_elements, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_color, float p_screen_lod_threshold) = 0; + + virtual void _render_shadow_begin() = 0; + virtual void _render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, const Rect2i &p_rect = Rect2i(), bool p_flip_y = false, bool p_clear_region = true, bool p_begin = true, bool p_end = true) = 0; + virtual void _render_shadow_process() = 0; + virtual void _render_shadow_end(uint32_t p_barrier = RD::BARRIER_MASK_ALL) = 0; + virtual void _render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) = 0; virtual void _render_uv2(const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) = 0; virtual void _render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<GeometryInstance *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture) = 0; @@ -124,8 +129,6 @@ protected: virtual void _base_uniforms_changed() = 0; virtual void _render_buffers_uniform_set_changed(RID p_render_buffers) = 0; virtual RID _render_buffers_get_normal_texture(RID p_render_buffers) = 0; - virtual RID _render_buffers_get_ambient_texture(RID p_render_buffers) = 0; - virtual RID _render_buffers_get_reflection_texture(RID p_render_buffers) = 0; void _process_ssao(RID p_render_buffers, RID p_environment, RID p_normal_buffer, const CameraMatrix &p_projection); void _process_ssr(RID p_render_buffers, RID p_dest_framebuffer, RID p_normal_buffer, RID p_specular_buffer, RID p_metallic, const Color &p_metallic_mask, RID p_environment, const CameraMatrix &p_projection, bool p_use_additive); @@ -134,7 +137,15 @@ protected: void _setup_sky(RID p_environment, RID p_render_buffers, const CameraMatrix &p_projection, const Transform &p_transform, const Size2i p_screen_size); void _update_sky(RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform); void _draw_sky(bool p_can_continue_color, bool p_can_continue_depth, RID p_fb, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform); - void _process_gi(RID p_render_buffers, RID p_normal_roughness_buffer, RID p_ambient_buffer, RID p_reflection_buffer, RID p_gi_probe_buffer, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform, const PagedArray<RID> &p_gi_probes); + void _pre_process_gi(RID p_render_buffers, const Transform &p_transform); + void _process_gi(RID p_render_buffers, RID p_normal_roughness_buffer, RID p_gi_probe_buffer, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform, const PagedArray<RID> &p_gi_probes); + + bool _needs_post_prepass_render(bool p_use_gi); + void _post_prepass_render(bool p_use_gi); + void _pre_resolve_render(bool p_use_gi); + + void _pre_opaque_render(bool p_use_ssao, bool p_use_gi, RID p_normal_roughness_buffer, RID p_gi_probe_buffer); + uint32_t _get_render_state_directional_light_count() const; // needed for a single argument calls (material and uv2) PagedArrayPool<GeometryInstance *> cull_argument_pool; @@ -322,7 +333,7 @@ private: uint32_t sky_ggx_samples_quality; bool sky_use_cubemap_array; - mutable RID_Owner<Sky> sky_owner; + mutable RID_Owner<Sky, true> sky_owner; /* REFLECTION ATLAS */ @@ -395,10 +406,10 @@ private: float attenuation; float color[3]; - float spot_angle_radians; + float cos_spot_angle; float position[3]; - float spot_attenuation; + float inv_spot_attenuation; float direction[3]; uint32_t has_shadow; @@ -574,17 +585,18 @@ private: uint32_t smallest_subdiv = 0; int size = 0; + bool use_16_bits = false; RID depth; RID fb; //for copying Map<RID, uint32_t> shadow_owners; - - Vector<ShadowShrinkStage> shrink_stages; }; RID_Owner<ShadowAtlas> shadow_atlas_owner; + void _update_shadow_atlas(ShadowAtlas *shadow_atlas); + bool _shadow_atlas_find_shadow(ShadowAtlas *shadow_atlas, int *p_in_quadrants, int p_quadrant_count, int p_current_subdiv, uint64_t p_tick, int &r_quadrant, int &r_shadow); RS::ShadowQuality shadows_quality = RS::SHADOW_QUALITY_MAX; //So it always updates when first set @@ -605,17 +617,16 @@ private: struct DirectionalShadow { RID depth; + RID fb; //when renderign direct int light_count = 0; int size = 0; + bool use_16_bits = false; int current_light = 0; - Vector<ShadowShrinkStage> shrink_stages; - } directional_shadow; - void _allocate_shadow_shrink_stages(RID p_base, int p_base_size, Vector<ShadowShrinkStage> &shrink_stages, uint32_t p_target_size); - void _clear_shadow_shrink_stages(Vector<ShadowShrinkStage> &shrink_stages); + void _update_directional_shadow_atlas(); /* SHADOW CUBEMAPS */ @@ -627,14 +638,6 @@ private: Map<int, ShadowCubemap> shadow_cubemaps; ShadowCubemap *_get_shadow_cubemap(int p_size); - struct ShadowMap { - RID depth; - RID fb; - }; - - Map<Vector2i, ShadowMap> shadow_maps; - ShadowMap *_get_shadow_map(const Size2i &p_size); - void _create_shadow_cubemaps(); /* LIGHT INSTANCE */ @@ -654,7 +657,7 @@ private: RS::LightType light_type = RS::LIGHT_DIRECTIONAL; - ShadowTransform shadow_transform[4]; + ShadowTransform shadow_transform[6]; AABB aabb; RID self; @@ -733,8 +736,9 @@ private: float volumetric_fog_light_energy = 0.0; float volumetric_fog_length = 64.0; float volumetric_fog_detail_spread = 2.0; - RS::EnvVolumetricFogShadowFilter volumetric_fog_shadow_filter = RS::ENV_VOLUMETRIC_FOG_SHADOW_FILTER_LOW; float volumetric_fog_gi_inject = 0.0; + bool volumetric_fog_temporal_reprojection = true; + float volumetric_fog_temporal_reprojection_amount = 0.9; /// Glow @@ -774,7 +778,7 @@ private: RS::EnvironmentSDFGICascades sdfgi_cascades; float sdfgi_min_cell_size = 0.2; bool sdfgi_use_occlusion = false; - bool sdfgi_use_multibounce = false; + float sdfgi_bounce_feedback = 0.0; bool sdfgi_read_sky_light = false; float sdfgi_energy = 1.0; float sdfgi_normal_bias = 1.1; @@ -805,7 +809,7 @@ private: static uint64_t auto_exposure_counter; - mutable RID_Owner<Environment> environment_owner; + mutable RID_Owner<Environment, true> environment_owner; /* CAMERA EFFECTS */ @@ -831,7 +835,7 @@ private: float sss_scale = 0.05; float sss_depth_scale = 0.01; - mutable RID_Owner<CameraEffects> camera_effects_owner; + mutable RID_Owner<CameraEffects, true> camera_effects_owner; /* RENDER BUFFERS */ @@ -904,6 +908,16 @@ private: RID giprobe_textures[MAX_GIPROBES]; RID giprobe_buffer; + + RID ambient_buffer; + RID reflection_buffer; + bool using_half_size_gi = false; + + struct GI { + RID full_buffer; + RID full_dispatch; + RID full_mask; + } gi; }; RID default_giprobe_buffer; @@ -965,6 +979,8 @@ private: RID scroll_occlusion_uniform_set; RID integrate_uniform_set; RID lights_buffer; + + bool all_dynamic_lights_dirty = true; }; //used for rendering (voxelization) @@ -1012,7 +1028,7 @@ private: RID cascades_ubo; bool uses_occlusion = false; - bool uses_multibounce = false; + float bounce_feedback = 0.0; bool reads_sky = false; float energy = 1.0; float normal_bias = 1.1; @@ -1022,10 +1038,18 @@ private: float y_mult = 1.0; uint32_t render_pass = 0; + + int32_t cascade_dynamic_light_count[SDFGI::MAX_CASCADES]; //used dynamically }; + void _sdfgi_update_light(RID p_render_buffers, RID p_environment); + void _sdfgi_update_probes(RID p_render_buffers, RID p_environment); + void _sdfgi_store_probes(RID p_render_buffers); + RS::EnvironmentSDFGIRayCount sdfgi_ray_count = RS::ENV_SDFGI_RAY_COUNT_16; RS::EnvironmentSDFGIFramesToConverge sdfgi_frames_to_converge = RS::ENV_SDFGI_CONVERGE_IN_10_FRAMES; + RS::EnvironmentSDFGIFramesToUpdateLight sdfgi_frames_to_update_light = RS::ENV_SDFGI_UPDATE_LIGHT_IN_4_FRAMES; + float sdfgi_solid_cell_ratio = 0.25; Vector3 sdfgi_debug_probe_pos; Vector3 sdfgi_debug_probe_dir; @@ -1122,8 +1146,8 @@ private: float attenuation; uint32_t type; - float spot_angle; - float spot_attenuation; + float cos_spot_angle; + float inv_spot_attenuation; float radius; float shadow_color[4]; @@ -1139,9 +1163,9 @@ private: uint32_t process_increment; int32_t probe_axis_size; - uint32_t multibounce; + float bounce_feedback; float y_mult; - uint32_t pad; + uint32_t use_occlusion; }; enum { @@ -1263,14 +1287,12 @@ private: float z_far; float proj_info[4]; - + float ao_color[3]; uint32_t max_giprobes; + uint32_t high_quality_vct; - uint32_t pad2; uint32_t orthogonal; - - float ao_color[3]; - uint32_t pad; + uint32_t pad[2]; float cam_rotation[12]; }; @@ -1280,9 +1302,13 @@ private: MODE_GIPROBE, MODE_SDFGI, MODE_COMBINED, + MODE_HALF_RES_GIPROBE, + MODE_HALF_RES_SDFGI, + MODE_HALF_RES_COMBINED, MODE_MAX }; + bool half_resolution = false; GiShaderRD shader; RID shader_version; RID pipelines[MODE_MAX]; @@ -1336,8 +1362,8 @@ private: float color[3]; float attenuation; - float cone_attenuation; - float cone_angle; + float inv_spot_attenuation; + float cos_spot_angle; float specular_amount; uint32_t shadow_enabled; @@ -1437,17 +1463,52 @@ private: uint32_t omni_light_count = 0; uint32_t spot_light_count = 0; - RID *lights_instances; - Rect2i *lights_shadow_rect_cache; - uint32_t lights_shadow_rect_cache_count = 0; - DirectionalLightData *directional_lights; uint32_t max_directional_lights; RID directional_light_buffer; } cluster; + struct RenderState { + RID render_buffers; + Transform cam_transform; + CameraMatrix cam_projection; + bool cam_ortogonal = false; + const PagedArray<GeometryInstance *> *instances = nullptr; + const PagedArray<RID> *lights = nullptr; + const PagedArray<RID> *reflection_probes = nullptr; + const PagedArray<RID> *gi_probes = nullptr; + const PagedArray<RID> *decals = nullptr; + const PagedArray<RID> *lightmaps = nullptr; + RID environment; + RID camera_effects; + RID shadow_atlas; + RID reflection_atlas; + RID reflection_probe; + int reflection_probe_pass = 0; + float screen_lod_threshold = 0.0; + + const RenderShadowData *render_shadows = nullptr; + int render_shadow_count = 0; + const RenderSDFGIData *render_sdfgi_regions = nullptr; + int render_sdfgi_region_count = 0; + const RenderSDFGIUpdateData *sdfgi_update_data = nullptr; + + uint32_t directional_light_count = 0; + uint32_t gi_probe_count = 0; + + LocalVector<int> cube_shadows; + LocalVector<int> shadows; + LocalVector<int> directional_shadows; + + bool depth_prepass_used; + } render_state; + struct VolumetricFog { + enum { + MAX_TEMPORAL_FRAMES = 16 + }; + uint32_t width = 0; uint32_t height = 0; uint32_t depth = 0; @@ -1456,6 +1517,8 @@ private: float spread; RID light_density_map; + RID prev_light_density_map; + RID fog_map; RID uniform_set; RID uniform_set2; @@ -1463,6 +1526,8 @@ private: RID sky_uniform_set; int last_shadow_filter = -1; + + Transform prev_cam_transform; }; enum { @@ -1498,10 +1563,13 @@ private: uint32_t cluster_shift; uint32_t cluster_width; - uint32_t cluster_pad[3]; uint32_t max_cluster_element_count_div_32; + uint32_t use_temporal_reprojection; + uint32_t temporal_frame; + float temporal_blend; float cam_rotation[12]; + float to_prev_view[16]; }; VolumetricFogShaderRD shader; @@ -1514,9 +1582,7 @@ private: uint32_t volumetric_fog_depth = 128; uint32_t volumetric_fog_size = 128; - bool volumetric_fog_filter_active = false; - uint32_t volumetric_fog_directional_shadow_shrink = 512; - uint32_t volumetric_fog_positional_shadow_shrink = 512; + bool volumetric_fog_filter_active = true; void _volumetric_fog_erase(RenderBuffers *rb); void _update_volumetric_fog(RID p_render_buffers, RID p_environment, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_shadow_atlas, int p_directional_light_count, bool p_use_directional_shadows, int p_positional_light_count, int p_gi_probe_count); @@ -1534,6 +1600,10 @@ private: uint32_t max_cluster_elements = 512; bool low_end = false; + void _render_shadow_pass(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<GeometryInstance *> &p_instances, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0, float p_screen_lod_threshold = 0.0, bool p_open_pass = true, bool p_close_pass = true, bool p_clear_region = true); + void _render_sdfgi_region(RID p_render_buffers, int p_region, const PagedArray<GeometryInstance *> &p_instances); + void _render_sdfgi_static_lights(RID p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const PagedArray<RID> *p_positional_light_cull_result); + public: virtual Transform geometry_instance_get_transform(GeometryInstance *p_instance) = 0; virtual AABB geometry_instance_get_aabb(GeometryInstance *p_instance) = 0; @@ -1541,7 +1611,7 @@ public: /* SHADOW ATLAS API */ RID shadow_atlas_create(); - void shadow_atlas_set_size(RID p_atlas, int p_size); + void shadow_atlas_set_size(RID p_atlas, int p_size, bool p_16_bits = false); void shadow_atlas_set_quadrant_subdivision(RID p_atlas, int p_quadrant, int p_subdivision); bool shadow_atlas_update_light(RID p_atlas, RID p_light_intance, float p_coverage, uint64_t p_light_version); _FORCE_INLINE_ bool shadow_atlas_owns_light_instance(RID p_atlas, RID p_light_intance) { @@ -1562,7 +1632,7 @@ public: return Size2(atlas->size, atlas->size); } - void directional_shadow_atlas_set_size(int p_size); + void directional_shadow_atlas_set_size(int p_size, bool p_16_bits = false); int get_directional_light_shadow_size(RID p_light_intance); void set_directional_shadow_count(int p_count); @@ -1581,11 +1651,12 @@ public: virtual int sdfgi_get_pending_region_count(RID p_render_buffers) const; virtual AABB sdfgi_get_pending_region_bounds(RID p_render_buffers, int p_region) const; virtual uint32_t sdfgi_get_pending_region_cascade(RID p_render_buffers, int p_region) const; - virtual void sdfgi_update_probes(RID p_render_buffers, RID p_environment, const Vector<RID> &p_directional_lights, const RID *p_positional_light_instances, uint32_t p_positional_light_count); RID sdfgi_get_ubo() const { return gi.sdfgi_ubo; } /* SKY API */ - RID sky_create(); + virtual RID sky_allocate(); + virtual void sky_initialize(RID p_rid); + void sky_set_radiance_size(RID p_sky, int p_radiance_size); void sky_set_mode(RID p_sky, RS::SkyMode p_mode); void sky_set_material(RID p_sky, RID p_material); @@ -1597,7 +1668,8 @@ public: /* ENVIRONMENT API */ - RID environment_create(); + virtual RID environment_allocate(); + virtual void environment_initialize(RID p_rid); void environment_set_background(RID p_env, RS::EnvironmentBG p_bg); void environment_set_sky(RID p_env, RID p_sky); @@ -1638,12 +1710,10 @@ public: float environment_get_fog_height_density(RID p_env) const; float environment_get_fog_aerial_perspective(RID p_env) const; - void environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, RS::EnvVolumetricFogShadowFilter p_shadow_filter); + void environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, bool p_temporal_reprojection, float p_temporal_reprojection_amount); virtual void environment_set_volumetric_fog_volume_size(int p_size, int p_depth); virtual void environment_set_volumetric_fog_filter_active(bool p_enable); - virtual void environment_set_volumetric_fog_directional_shadow_shrink_size(int p_shrink_size); - virtual void environment_set_volumetric_fog_positional_shadow_shrink_size(int p_shrink_size); void environment_set_ssr(RID p_env, bool p_enable, int p_max_steps, float p_fade_int, float p_fade_out, float p_depth_tolerance); void environment_set_ssao(RID p_env, bool p_enable, float p_radius, float p_intensity, float p_power, float p_detail, float p_horizon, float p_sharpness, float p_light_affect, float p_ao_channel_affect); @@ -1654,9 +1724,10 @@ public: bool environment_is_ssr_enabled(RID p_env) const; bool environment_is_sdfgi_enabled(RID p_env) const; - virtual void environment_set_sdfgi(RID p_env, bool p_enable, RS::EnvironmentSDFGICascades p_cascades, float p_min_cell_size, RS::EnvironmentSDFGIYScale p_y_scale, bool p_use_occlusion, bool p_use_multibounce, bool p_read_sky, float p_energy, float p_normal_bias, float p_probe_bias); + virtual void environment_set_sdfgi(RID p_env, bool p_enable, RS::EnvironmentSDFGICascades p_cascades, float p_min_cell_size, RS::EnvironmentSDFGIYScale p_y_scale, bool p_use_occlusion, float p_bounce_feedback, bool p_read_sky, float p_energy, float p_normal_bias, float p_probe_bias); virtual void environment_set_sdfgi_ray_count(RS::EnvironmentSDFGIRayCount p_ray_count); virtual void environment_set_sdfgi_frames_to_converge(RS::EnvironmentSDFGIFramesToConverge p_frames); + virtual void environment_set_sdfgi_frames_to_update_light(RS::EnvironmentSDFGIFramesToUpdateLight p_update); void environment_set_ssr_roughness_quality(RS::EnvironmentSSRRoughnessQuality p_quality); RS::EnvironmentSSRRoughnessQuality environment_get_ssr_roughness_quality() const; @@ -1666,7 +1737,8 @@ public: virtual Ref<Image> environment_bake_panorama(RID p_env, bool p_bake_irradiance, const Size2i &p_size); - virtual RID camera_effects_create(); + virtual RID camera_effects_allocate(); + virtual void camera_effects_initialize(RID p_rid); virtual void camera_effects_set_dof_blur_quality(RS::DOFBlurQuality p_quality, bool p_use_jitter); virtual void camera_effects_set_dof_blur_bokeh_shape(RS::DOFBokehShape p_shape); @@ -1955,11 +2027,14 @@ public: */ RID render_buffers_create(); void render_buffers_configure(RID p_render_buffers, RID p_render_target, int p_width, int p_height, RS::ViewportMSAA p_msaa, RS::ViewportScreenSpaceAA p_screen_space_aa, bool p_use_debanding); + void gi_set_use_half_resolution(bool p_enable); RID render_buffers_get_ao_texture(RID p_render_buffers); RID render_buffers_get_back_buffer_texture(RID p_render_buffers); RID render_buffers_get_gi_probe_buffer(RID p_render_buffers); RID render_buffers_get_default_gi_probe_buffer(); + RID render_buffers_get_gi_ambient_texture(RID p_render_buffers); + RID render_buffers_get_gi_reflection_texture(RID p_render_buffers); uint32_t render_buffers_get_sdfgi_cascade_count(RID p_render_buffers) const; bool render_buffers_is_sdfgi_enabled(RID p_render_buffers) const; @@ -1980,15 +2055,10 @@ public: float render_buffers_get_volumetric_fog_end(RID p_render_buffers); float render_buffers_get_volumetric_fog_detail_spread(RID p_render_buffers); - void render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold); - - void render_shadow(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<GeometryInstance *> &p_instances, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0, float p_screen_lod_threshold = 0.0); + void render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold, const RenderShadowData *p_render_shadows, int p_render_shadow_count, const RenderSDFGIData *p_render_sdfgi_regions, int p_render_sdfgi_region_count, const RenderSDFGIUpdateData *p_sdfgi_update_data = nullptr); void render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region); - void render_sdfgi(RID p_render_buffers, int p_region, const PagedArray<GeometryInstance *> &p_instances); - void render_sdfgi_static_lights(RID p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const PagedArray<RID> *p_positional_light_cull_result); - void render_particle_collider_heightfield(RID p_collider, const Transform &p_transform, const PagedArray<GeometryInstance *> &p_instances); virtual void set_scene_pass(uint64_t p_pass) { diff --git a/servers/rendering/renderer_rd/renderer_storage_rd.cpp b/servers/rendering/renderer_rd/renderer_storage_rd.cpp index 6203f3ba64..2a34049675 100644 --- a/servers/rendering/renderer_rd/renderer_storage_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_storage_rd.cpp @@ -36,6 +36,10 @@ #include "renderer_compositor_rd.h" #include "servers/rendering/shader_language.h" +bool RendererStorageRD::can_create_resources_async() const { + return true; +} + Ref<Image> RendererStorageRD::_validate_texture_format(const Ref<Image> &p_image, TextureToRDFormat &r_format) { Ref<Image> image = p_image->duplicate(); @@ -535,9 +539,13 @@ Ref<Image> RendererStorageRD::_validate_texture_format(const Ref<Image> &p_image return image; } -RID RendererStorageRD::texture_2d_create(const Ref<Image> &p_image) { - ERR_FAIL_COND_V(p_image.is_null(), RID()); - ERR_FAIL_COND_V(p_image->is_empty(), RID()); +RID RendererStorageRD::texture_allocate() { + return texture_owner.allocate_rid(); +} + +void RendererStorageRD::texture_2d_initialize(RID p_texture, const Ref<Image> &p_image) { + ERR_FAIL_COND(p_image.is_null()); + ERR_FAIL_COND(p_image->is_empty()); TextureToRDFormat ret_format; Ref<Image> image = _validate_texture_format(p_image, ret_format); @@ -585,13 +593,13 @@ RID RendererStorageRD::texture_2d_create(const Ref<Image> &p_image) { Vector<Vector<uint8_t>> data_slices; data_slices.push_back(data); texture.rd_texture = RD::get_singleton()->texture_create(rd_format, rd_view, data_slices); - ERR_FAIL_COND_V(texture.rd_texture.is_null(), RID()); + ERR_FAIL_COND(texture.rd_texture.is_null()); if (texture.rd_format_srgb != RD::DATA_FORMAT_MAX) { rd_view.format_override = texture.rd_format_srgb; texture.rd_texture_srgb = RD::get_singleton()->texture_create_shared(rd_view, texture.rd_texture); if (texture.rd_texture_srgb.is_null()) { RD::get_singleton()->free(texture.rd_texture); - ERR_FAIL_COND_V(texture.rd_texture_srgb.is_null(), RID()); + ERR_FAIL_COND(texture.rd_texture_srgb.is_null()); } } @@ -602,14 +610,14 @@ RID RendererStorageRD::texture_2d_create(const Ref<Image> &p_image) { texture.rd_view = rd_view; texture.is_proxy = false; - return texture_owner.make_rid(texture); + texture_owner.initialize_rid(p_texture, texture); } -RID RendererStorageRD::texture_2d_layered_create(const Vector<Ref<Image>> &p_layers, RS::TextureLayeredType p_layered_type) { - ERR_FAIL_COND_V(p_layers.size() == 0, RID()); +void RendererStorageRD::texture_2d_layered_initialize(RID p_texture, const Vector<Ref<Image>> &p_layers, RS::TextureLayeredType p_layered_type) { + ERR_FAIL_COND(p_layers.size() == 0); - ERR_FAIL_COND_V(p_layered_type == RS::TEXTURE_LAYERED_CUBEMAP && p_layers.size() != 6, RID()); - ERR_FAIL_COND_V(p_layered_type == RS::TEXTURE_LAYERED_CUBEMAP_ARRAY && (p_layers.size() < 6 || (p_layers.size() % 6) != 0), RID()); + ERR_FAIL_COND(p_layered_type == RS::TEXTURE_LAYERED_CUBEMAP && p_layers.size() != 6); + ERR_FAIL_COND(p_layered_type == RS::TEXTURE_LAYERED_CUBEMAP_ARRAY && (p_layers.size() < 6 || (p_layers.size() % 6) != 0)); TextureToRDFormat ret_format; Vector<Ref<Image>> images; @@ -620,7 +628,7 @@ RID RendererStorageRD::texture_2d_layered_create(const Vector<Ref<Image>> &p_lay Image::Format valid_format = Image::FORMAT_MAX; for (int i = 0; i < p_layers.size(); i++) { - ERR_FAIL_COND_V(p_layers[i]->is_empty(), RID()); + ERR_FAIL_COND(p_layers[i]->is_empty()); if (i == 0) { valid_width = p_layers[i]->get_width(); @@ -628,10 +636,10 @@ RID RendererStorageRD::texture_2d_layered_create(const Vector<Ref<Image>> &p_lay valid_format = p_layers[i]->get_format(); valid_mipmaps = p_layers[i]->has_mipmaps(); } else { - ERR_FAIL_COND_V(p_layers[i]->get_width() != valid_width, RID()); - ERR_FAIL_COND_V(p_layers[i]->get_height() != valid_height, RID()); - ERR_FAIL_COND_V(p_layers[i]->get_format() != valid_format, RID()); - ERR_FAIL_COND_V(p_layers[i]->has_mipmaps() != valid_mipmaps, RID()); + ERR_FAIL_COND(p_layers[i]->get_width() != valid_width); + ERR_FAIL_COND(p_layers[i]->get_height() != valid_height); + ERR_FAIL_COND(p_layers[i]->get_format() != valid_format); + ERR_FAIL_COND(p_layers[i]->has_mipmaps() != valid_mipmaps); } images.push_back(_validate_texture_format(p_layers[i], ret_format)); @@ -695,13 +703,13 @@ RID RendererStorageRD::texture_2d_layered_create(const Vector<Ref<Image>> &p_lay data_slices.push_back(data); } texture.rd_texture = RD::get_singleton()->texture_create(rd_format, rd_view, data_slices); - ERR_FAIL_COND_V(texture.rd_texture.is_null(), RID()); + ERR_FAIL_COND(texture.rd_texture.is_null()); if (texture.rd_format_srgb != RD::DATA_FORMAT_MAX) { rd_view.format_override = texture.rd_format_srgb; texture.rd_texture_srgb = RD::get_singleton()->texture_create_shared(rd_view, texture.rd_texture); if (texture.rd_texture_srgb.is_null()) { RD::get_singleton()->free(texture.rd_texture); - ERR_FAIL_COND_V(texture.rd_texture_srgb.is_null(), RID()); + ERR_FAIL_COND(texture.rd_texture_srgb.is_null()); } } @@ -712,14 +720,14 @@ RID RendererStorageRD::texture_2d_layered_create(const Vector<Ref<Image>> &p_lay texture.rd_view = rd_view; texture.is_proxy = false; - return texture_owner.make_rid(texture); + texture_owner.initialize_rid(p_texture, texture); } -RID RendererStorageRD::texture_3d_create(Image::Format p_format, int p_width, int p_height, int p_depth, bool p_mipmaps, const Vector<Ref<Image>> &p_data) { - ERR_FAIL_COND_V(p_data.size() == 0, RID()); +void RendererStorageRD::texture_3d_initialize(RID p_texture, Image::Format p_format, int p_width, int p_height, int p_depth, bool p_mipmaps, const Vector<Ref<Image>> &p_data) { + ERR_FAIL_COND(p_data.size() == 0); Image::Image3DValidateError verr = Image::validate_3d_image(p_format, p_width, p_height, p_depth, p_mipmaps, p_data); if (verr != Image::VALIDATE_3D_OK) { - ERR_FAIL_V_MSG(RID(), Image::get_3d_image_validation_error_text(verr)); + ERR_FAIL_MSG(Image::get_3d_image_validation_error_text(verr)); } TextureToRDFormat ret_format; @@ -811,13 +819,13 @@ RID RendererStorageRD::texture_3d_create(Image::Format p_format, int p_width, in data_slices.push_back(all_data); //one slice texture.rd_texture = RD::get_singleton()->texture_create(rd_format, rd_view, data_slices); - ERR_FAIL_COND_V(texture.rd_texture.is_null(), RID()); + ERR_FAIL_COND(texture.rd_texture.is_null()); if (texture.rd_format_srgb != RD::DATA_FORMAT_MAX) { rd_view.format_override = texture.rd_format_srgb; texture.rd_texture_srgb = RD::get_singleton()->texture_create_shared(rd_view, texture.rd_texture); if (texture.rd_texture_srgb.is_null()) { RD::get_singleton()->free(texture.rd_texture); - ERR_FAIL_COND_V(texture.rd_texture_srgb.is_null(), RID()); + ERR_FAIL_COND(texture.rd_texture_srgb.is_null()); } } @@ -828,12 +836,12 @@ RID RendererStorageRD::texture_3d_create(Image::Format p_format, int p_width, in texture.rd_view = rd_view; texture.is_proxy = false; - return texture_owner.make_rid(texture); + texture_owner.initialize_rid(p_texture, texture); } -RID RendererStorageRD::texture_proxy_create(RID p_base) { +void RendererStorageRD::texture_proxy_initialize(RID p_texture, RID p_base) { Texture *tex = texture_owner.getornull(p_base); - ERR_FAIL_COND_V(!tex, RID()); + ERR_FAIL_COND(!tex); Texture proxy_tex = *tex; proxy_tex.rd_view.format_override = tex->rd_format; @@ -847,11 +855,9 @@ RID RendererStorageRD::texture_proxy_create(RID p_base) { proxy_tex.is_proxy = true; proxy_tex.proxies.clear(); - RID rid = texture_owner.make_rid(proxy_tex); - - tex->proxies.push_back(rid); + texture_owner.initialize_rid(p_texture, proxy_tex); - return rid; + tex->proxies.push_back(p_texture); } void RendererStorageRD::_texture_2d_update(RID p_texture, const Ref<Image> &p_image, int p_layer, bool p_immediate) { @@ -873,7 +879,7 @@ void RendererStorageRD::_texture_2d_update(RID p_texture, const Ref<Image> &p_im TextureToRDFormat f; Ref<Image> validated = _validate_texture_format(p_image, f); - RD::get_singleton()->texture_update(tex->rd_texture, p_layer, validated->get_data(), !p_immediate); + RD::get_singleton()->texture_update(tex->rd_texture, p_layer, validated->get_data()); } void RendererStorageRD::texture_2d_update_immediate(RID p_texture, const Ref<Image> &p_image, int p_layer) { @@ -918,7 +924,7 @@ void RendererStorageRD::texture_3d_update(RID p_texture, const Vector<Ref<Image> } } - RD::get_singleton()->texture_update(tex->rd_texture, 0, all_data, true); + RD::get_singleton()->texture_update(tex->rd_texture, 0, all_data); } void RendererStorageRD::texture_proxy_update(RID p_texture, RID p_proxy_to) { @@ -961,7 +967,7 @@ void RendererStorageRD::texture_proxy_update(RID p_texture, RID p_proxy_to) { } //these two APIs can be used together or in combination with the others. -RID RendererStorageRD::texture_2d_placeholder_create() { +void RendererStorageRD::texture_2d_placeholder_initialize(RID p_texture) { //this could be better optimized to reuse an existing image , done this way //for now to get it working Ref<Image> image; @@ -974,10 +980,10 @@ RID RendererStorageRD::texture_2d_placeholder_create() { } } - return texture_2d_create(image); + texture_2d_initialize(p_texture, image); } -RID RendererStorageRD::texture_2d_layered_placeholder_create(RS::TextureLayeredType p_layered_type) { +void RendererStorageRD::texture_2d_layered_placeholder_initialize(RID p_texture, RS::TextureLayeredType p_layered_type) { //this could be better optimized to reuse an existing image , done this way //for now to get it working Ref<Image> image; @@ -1000,10 +1006,10 @@ RID RendererStorageRD::texture_2d_layered_placeholder_create(RS::TextureLayeredT } } - return texture_2d_layered_create(images, p_layered_type); + texture_2d_layered_initialize(p_texture, images, p_layered_type); } -RID RendererStorageRD::texture_3d_placeholder_create() { +void RendererStorageRD::texture_3d_placeholder_initialize(RID p_texture) { //this could be better optimized to reuse an existing image , done this way //for now to get it working Ref<Image> image; @@ -1022,7 +1028,7 @@ RID RendererStorageRD::texture_3d_placeholder_create() { images.push_back(image); } - return texture_3d_create(Image::FORMAT_RGBA8, 4, 4, 4, false, images); + texture_3d_initialize(p_texture, Image::FORMAT_RGBA8, 4, 4, 4, false, images); } Ref<Image> RendererStorageRD::texture_2d_get(RID p_texture) const { @@ -1223,8 +1229,11 @@ RendererStorageRD::CanvasTexture::~CanvasTexture() { clear_sets(); } -RID RendererStorageRD::canvas_texture_create() { - return canvas_texture_owner.make_rid(memnew(CanvasTexture)); +RID RendererStorageRD::canvas_texture_allocate() { + return canvas_texture_owner.allocate_rid(); +} +void RendererStorageRD::canvas_texture_initialize(RID p_rid) { + canvas_texture_owner.initialize_rid(p_rid, memnew(CanvasTexture)); } void RendererStorageRD::canvas_texture_set_channel(RID p_canvas_texture, RS::CanvasTextureChannel p_channel, RID p_texture) { @@ -1365,12 +1374,15 @@ bool RendererStorageRD::canvas_texture_get_uniform_set(RID p_texture, RS::Canvas /* SHADER API */ -RID RendererStorageRD::shader_create() { +RID RendererStorageRD::shader_allocate() { + return shader_owner.allocate_rid(); +} +void RendererStorageRD::shader_initialize(RID p_rid) { Shader shader; shader.data = nullptr; shader.type = SHADER_TYPE_MAX; - return shader_owner.make_rid(shader); + shader_owner.initialize_rid(p_rid, shader); } void RendererStorageRD::shader_set_code(RID p_shader, const String &p_code) { @@ -1510,7 +1522,10 @@ RS::ShaderNativeSourceCode RendererStorageRD::shader_get_native_source_code(RID /* COMMON MATERIAL API */ -RID RendererStorageRD::material_create() { +RID RendererStorageRD::material_allocate() { + return material_owner.allocate_rid(); +} +void RendererStorageRD::material_initialize(RID p_rid) { Material material; material.data = nullptr; material.shader = nullptr; @@ -1520,12 +1535,8 @@ RID RendererStorageRD::material_create() { material.uniform_dirty = false; material.texture_dirty = false; material.priority = 0; - RID id = material_owner.make_rid(material); - { - Material *material_ptr = material_owner.getornull(id); - material_ptr->self = id; - } - return id; + material.self = p_rid; + material_owner.initialize_rid(p_rid, material); } void RendererStorageRD::_material_queue_update(Material *material, bool p_uniform, bool p_texture) { @@ -2399,8 +2410,11 @@ void RendererStorageRD::_update_queued_materials() { /* MESH API */ -RID RendererStorageRD::mesh_create() { - return mesh_owner.make_rid(Mesh()); +RID RendererStorageRD::mesh_allocate() { + return mesh_owner.allocate_rid(); +} +void RendererStorageRD::mesh_initialize(RID p_rid) { + mesh_owner.initialize_rid(p_rid, Mesh()); } void RendererStorageRD::mesh_set_blend_shape_count(RID p_mesh, int p_blend_shape_count) { @@ -2609,6 +2623,12 @@ void RendererStorageRD::mesh_add_surface(RID p_mesh, const RS::SurfaceData &p_su mesh->dependency.changed_notify(DEPENDENCY_CHANGED_MESH); + for (Set<Mesh *>::Element *E = mesh->shadow_owners.front(); E; E = E->next()) { + Mesh *shadow_owner = E->get(); + shadow_owner->shadow_mesh = RID(); + shadow_owner->dependency.changed_notify(DEPENDENCY_CHANGED_MESH); + } + mesh->material_cache.clear(); } @@ -2824,6 +2844,25 @@ AABB RendererStorageRD::mesh_get_aabb(RID p_mesh, RID p_skeleton) { return aabb; } +void RendererStorageRD::mesh_set_shadow_mesh(RID p_mesh, RID p_shadow_mesh) { + Mesh *mesh = mesh_owner.getornull(p_mesh); + ERR_FAIL_COND(!mesh); + + Mesh *shadow_mesh = mesh_owner.getornull(mesh->shadow_mesh); + if (shadow_mesh) { + shadow_mesh->shadow_owners.erase(mesh); + } + mesh->shadow_mesh = p_shadow_mesh; + + shadow_mesh = mesh_owner.getornull(mesh->shadow_mesh); + + if (shadow_mesh) { + shadow_mesh->shadow_owners.insert(mesh); + } + + mesh->dependency.changed_notify(DEPENDENCY_CHANGED_MESH); +} + void RendererStorageRD::mesh_clear(RID p_mesh) { Mesh *mesh = mesh_owner.getornull(p_mesh); ERR_FAIL_COND(!mesh); @@ -2871,6 +2910,12 @@ void RendererStorageRD::mesh_clear(RID p_mesh) { } mesh->has_bone_weights = false; mesh->dependency.changed_notify(DEPENDENCY_CHANGED_MESH); + + for (Set<Mesh *>::Element *E = mesh->shadow_owners.front(); E; E = E->next()) { + Mesh *shadow_owner = E->get(); + shadow_owner->shadow_mesh = RID(); + shadow_owner->dependency.changed_notify(DEPENDENCY_CHANGED_MESH); + } } bool RendererStorageRD::mesh_needs_instance(RID p_mesh, bool p_has_skeleton) { @@ -3013,7 +3058,7 @@ void RendererStorageRD::update_mesh_instances() { MeshInstance *mi = dirty_mesh_instance_weights.first()->self(); if (mi->blend_weights_buffer.is_valid()) { - RD::get_singleton()->buffer_update(mi->blend_weights_buffer, 0, mi->blend_weights.size() * sizeof(float), mi->blend_weights.ptr(), true); + RD::get_singleton()->buffer_update(mi->blend_weights_buffer, 0, mi->blend_weights.size() * sizeof(float), mi->blend_weights.ptr()); } dirty_mesh_instance_weights.remove(&mi->weight_update_list); mi->weights_dirty = false; @@ -3067,7 +3112,7 @@ void RendererStorageRD::update_mesh_instances() { RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SkeletonShader::PushConstant)); //dispatch without barrier, so all is done at the same time - RD::get_singleton()->compute_list_dispatch_threads(compute_list, push_constant.vertex_count, 1, 1, 64, 1, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, push_constant.vertex_count, 1, 1); } mi->dirty = false; @@ -3267,11 +3312,14 @@ void RendererStorageRD::_mesh_surface_generate_version_for_input_mask(Mesh::Surf ////////////////// MULTIMESH -RID RendererStorageRD::multimesh_create() { - return multimesh_owner.make_rid(MultiMesh()); +RID RendererStorageRD::multimesh_allocate() { + return multimesh_owner.allocate_rid(); +} +void RendererStorageRD::multimesh_initialize(RID p_rid) { + multimesh_owner.initialize_rid(p_rid, MultiMesh()); } -void RendererStorageRD::multimesh_allocate(RID p_multimesh, int p_instances, RS::MultimeshTransformFormat p_transform_format, bool p_use_colors, bool p_use_custom_data) { +void RendererStorageRD::multimesh_allocate_data(RID p_multimesh, int p_instances, RS::MultimeshTransformFormat p_transform_format, bool p_use_colors, bool p_use_custom_data) { MultiMesh *multimesh = multimesh_owner.getornull(p_multimesh); ERR_FAIL_COND(!multimesh); @@ -3681,7 +3729,7 @@ void RendererStorageRD::multimesh_set_buffer(RID p_multimesh, const Vector<float { const float *r = p_buffer.ptr(); - RD::get_singleton()->buffer_update(multimesh->buffer, 0, p_buffer.size() * sizeof(float), r, false); + RD::get_singleton()->buffer_update(multimesh->buffer, 0, p_buffer.size() * sizeof(float), r); multimesh->buffer_set = true; } @@ -3780,14 +3828,14 @@ void RendererStorageRD::_update_dirty_multimeshes() { if (multimesh->data_cache_used_dirty_regions > 32 || multimesh->data_cache_used_dirty_regions > visible_region_count / 2) { //if there too many dirty regions, or represent the majority of regions, just copy all, else transfer cost piles up too much - RD::get_singleton()->buffer_update(multimesh->buffer, 0, MIN(visible_region_count * region_size, multimesh->instances * multimesh->stride_cache * sizeof(float)), data, false); + RD::get_singleton()->buffer_update(multimesh->buffer, 0, MIN(visible_region_count * region_size, multimesh->instances * multimesh->stride_cache * sizeof(float)), data); } else { //not that many regions? update them all for (uint32_t i = 0; i < visible_region_count; i++) { if (multimesh->data_cache_dirty_regions[i]) { uint64_t offset = i * region_size; uint64_t size = multimesh->stride_cache * multimesh->instances * sizeof(float); - RD::get_singleton()->buffer_update(multimesh->buffer, offset, MIN(region_size, size - offset), &data[i * region_size], false); + RD::get_singleton()->buffer_update(multimesh->buffer, offset, MIN(region_size, size - offset), &data[i * region_size]); } } } @@ -3818,8 +3866,11 @@ void RendererStorageRD::_update_dirty_multimeshes() { /* PARTICLES */ -RID RendererStorageRD::particles_create() { - return particles_owner.make_rid(Particles()); +RID RendererStorageRD::particles_allocate() { + return particles_owner.allocate_rid(); +} +void RendererStorageRD::particles_initialize(RID p_rid) { + particles_owner.initialize_rid(p_rid, Particles()); } void RendererStorageRD::particles_set_emitting(RID p_particles, bool p_emitting) { @@ -4478,7 +4529,7 @@ void RendererStorageRD::_particles_process(Particles *p_particles, float p_delta if (sub_emitter && sub_emitter->emission_storage_buffer.is_valid()) { // print_line("updating subemitter buffer"); int32_t zero[4] = { 0, sub_emitter->amount, 0, 0 }; - RD::get_singleton()->buffer_update(sub_emitter->emission_storage_buffer, 0, sizeof(uint32_t) * 4, zero, true); + RD::get_singleton()->buffer_update(sub_emitter->emission_storage_buffer, 0, sizeof(uint32_t) * 4, zero); push_constant.can_emit = true; if (sub_emitter->emitting) { @@ -4496,13 +4547,13 @@ void RendererStorageRD::_particles_process(Particles *p_particles, float p_delta } if (p_particles->emission_buffer && p_particles->emission_buffer->particle_count) { - RD::get_singleton()->buffer_update(p_particles->emission_storage_buffer, 0, sizeof(uint32_t) * 4 + sizeof(ParticleEmissionBuffer::Data) * p_particles->emission_buffer->particle_count, p_particles->emission_buffer, true); + RD::get_singleton()->buffer_update(p_particles->emission_storage_buffer, 0, sizeof(uint32_t) * 4 + sizeof(ParticleEmissionBuffer::Data) * p_particles->emission_buffer->particle_count, p_particles->emission_buffer); p_particles->emission_buffer->particle_count = 0; } p_particles->clear = false; - RD::get_singleton()->buffer_update(p_particles->frame_params_buffer, 0, sizeof(ParticlesFrameParams), &frame_params, true); + RD::get_singleton()->buffer_update(p_particles->frame_params_buffer, 0, sizeof(ParticlesFrameParams), &frame_params); ParticlesMaterialData *m = (ParticlesMaterialData *)material_get_data(p_particles->process_material, SHADER_TYPE_PARTICLES); if (!m) { @@ -4524,7 +4575,7 @@ void RendererStorageRD::_particles_process(Particles *p_particles, float p_delta RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ParticlesShader::PushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_particles->amount, 1, 1, 64, 1, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_particles->amount, 1, 1); RD::get_singleton()->compute_list_end(); } @@ -4578,7 +4629,7 @@ void RendererStorageRD::particles_set_view_axis(RID p_particles, const Vector3 & RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->particles_sort_uniform_set, 1); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©_push_constant, sizeof(ParticlesShader::CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, particles->amount, 1, 1, 64, 1, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, particles->amount, 1, 1); RD::get_singleton()->compute_list_end(); @@ -4590,7 +4641,7 @@ void RendererStorageRD::particles_set_view_axis(RID p_particles, const Vector3 & RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->particles_sort_uniform_set, 1); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©_push_constant, sizeof(ParticlesShader::CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, particles->amount, 1, 1, 64, 1, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, particles->amount, 1, 1); RD::get_singleton()->compute_list_end(); } @@ -4697,7 +4748,7 @@ void RendererStorageRD::update_particles() { RD::get_singleton()->compute_list_bind_uniform_set(compute_list, particles->particles_copy_uniform_set, 0); RD::get_singleton()->compute_list_set_push_constant(compute_list, ©_push_constant, sizeof(ParticlesShader::CopyPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, particles->amount, 1, 1, 64, 1, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, particles->amount, 1, 1); RD::get_singleton()->compute_list_end(); } @@ -4953,8 +5004,11 @@ RendererStorageRD::MaterialData *RendererStorageRD::_create_particles_material_f /* PARTICLES COLLISION API */ -RID RendererStorageRD::particles_collision_create() { - return particles_collision_owner.make_rid(ParticlesCollision()); +RID RendererStorageRD::particles_collision_allocate() { + return particles_collision_owner.allocate_rid(); +} +void RendererStorageRD::particles_collision_initialize(RID p_rid) { + particles_collision_owner.initialize_rid(p_rid, ParticlesCollision()); } RID RendererStorageRD::particles_collision_get_heightfield_framebuffer(RID p_particles_collision) const { @@ -5133,8 +5187,11 @@ void RendererStorageRD::particles_collision_instance_set_active(RID p_collision_ /* SKELETON API */ -RID RendererStorageRD::skeleton_create() { - return skeleton_owner.make_rid(Skeleton()); +RID RendererStorageRD::skeleton_allocate() { + return skeleton_owner.allocate_rid(); +} +void RendererStorageRD::skeleton_initialize(RID p_rid) { + skeleton_owner.initialize_rid(p_rid, Skeleton()); } void RendererStorageRD::_skeleton_make_dirty(Skeleton *skeleton) { @@ -5145,7 +5202,7 @@ void RendererStorageRD::_skeleton_make_dirty(Skeleton *skeleton) { } } -void RendererStorageRD::skeleton_allocate(RID p_skeleton, int p_bones, bool p_2d_skeleton) { +void RendererStorageRD::skeleton_allocate_data(RID p_skeleton, int p_bones, bool p_2d_skeleton) { Skeleton *skeleton = skeleton_owner.getornull(p_skeleton); ERR_FAIL_COND(!skeleton); ERR_FAIL_COND(p_bones < 0); @@ -5301,7 +5358,7 @@ void RendererStorageRD::_update_dirty_skeletons() { Skeleton *skeleton = skeleton_dirty_list; if (skeleton->size) { - RD::get_singleton()->buffer_update(skeleton->buffer, 0, skeleton->data.size() * sizeof(float), skeleton->data.ptr(), false); + RD::get_singleton()->buffer_update(skeleton->buffer, 0, skeleton->data.size() * sizeof(float), skeleton->data.ptr()); } skeleton_dirty_list = skeleton->dirty_list; @@ -5319,7 +5376,7 @@ void RendererStorageRD::_update_dirty_skeletons() { /* LIGHT */ -RID RendererStorageRD::light_create(RS::LightType p_type) { +void RendererStorageRD::_light_initialize(RID p_light, RS::LightType p_type) { Light light; light.type = p_type; @@ -5340,10 +5397,31 @@ RID RendererStorageRD::light_create(RS::LightType p_type) { light.param[RS::LIGHT_PARAM_SHADOW_BIAS] = 0.02; light.param[RS::LIGHT_PARAM_SHADOW_BLUR] = 0; light.param[RS::LIGHT_PARAM_SHADOW_PANCAKE_SIZE] = 20.0; - light.param[RS::LIGHT_PARAM_SHADOW_VOLUMETRIC_FOG_FADE] = 1.0; + light.param[RS::LIGHT_PARAM_SHADOW_VOLUMETRIC_FOG_FADE] = 0.1; light.param[RS::LIGHT_PARAM_TRANSMITTANCE_BIAS] = 0.05; - return light_owner.make_rid(light); + light_owner.initialize_rid(p_light, light); +} + +RID RendererStorageRD::directional_light_allocate() { + return light_owner.allocate_rid(); +} +void RendererStorageRD::directional_light_initialize(RID p_light) { + _light_initialize(p_light, RS::LIGHT_DIRECTIONAL); +} + +RID RendererStorageRD::omni_light_allocate() { + return light_owner.allocate_rid(); +} +void RendererStorageRD::omni_light_initialize(RID p_light) { + _light_initialize(p_light, RS::LIGHT_OMNI); +} + +RID RendererStorageRD::spot_light_allocate() { + return light_owner.allocate_rid(); +} +void RendererStorageRD::spot_light_initialize(RID p_light) { + _light_initialize(p_light, RS::LIGHT_SPOT); } void RendererStorageRD::light_set_color(RID p_light, const Color &p_color) { @@ -5581,8 +5659,11 @@ AABB RendererStorageRD::light_get_aabb(RID p_light) const { /* REFLECTION PROBE */ -RID RendererStorageRD::reflection_probe_create() { - return reflection_probe_owner.make_rid(ReflectionProbe()); +RID RendererStorageRD::reflection_probe_allocate() { + return reflection_probe_owner.allocate_rid(); +} +void RendererStorageRD::reflection_probe_initialize(RID p_reflection_probe) { + reflection_probe_owner.initialize_rid(p_reflection_probe, ReflectionProbe()); } void RendererStorageRD::reflection_probe_set_update_mode(RID p_probe, RS::ReflectionProbeUpdateMode p_mode) { @@ -5804,8 +5885,11 @@ float RendererStorageRD::reflection_probe_get_ambient_color_energy(RID p_probe) return reflection_probe->ambient_color_energy; } -RID RendererStorageRD::decal_create() { - return decal_owner.make_rid(Decal()); +RID RendererStorageRD::decal_allocate() { + return decal_owner.allocate_rid(); +} +void RendererStorageRD::decal_initialize(RID p_decal) { + decal_owner.initialize_rid(p_decal, Decal()); } void RendererStorageRD::decal_set_extents(RID p_decal, const Vector3 &p_extents) { @@ -5892,11 +5976,14 @@ AABB RendererStorageRD::decal_get_aabb(RID p_decal) const { return AABB(-decal->extents, decal->extents * 2.0); } -RID RendererStorageRD::gi_probe_create() { - return gi_probe_owner.make_rid(GIProbe()); +RID RendererStorageRD::gi_probe_allocate() { + return gi_probe_owner.allocate_rid(); +} +void RendererStorageRD::gi_probe_initialize(RID p_gi_probe) { + gi_probe_owner.initialize_rid(p_gi_probe, GIProbe()); } -void RendererStorageRD::gi_probe_allocate(RID p_gi_probe, const Transform &p_to_cell_xform, const AABB &p_aabb, const Vector3i &p_octree_size, const Vector<uint8_t> &p_octree_cells, const Vector<uint8_t> &p_data_cells, const Vector<uint8_t> &p_distance_field, const Vector<int> &p_level_counts) { +void RendererStorageRD::gi_probe_allocate_data(RID p_gi_probe, const Transform &p_to_cell_xform, const AABB &p_aabb, const Vector3i &p_octree_size, const Vector<uint8_t> &p_octree_cells, const Vector<uint8_t> &p_data_cells, const Vector<uint8_t> &p_distance_field, const Vector<int> &p_level_counts) { GIProbe *gi_probe = gi_probe_owner.getornull(p_gi_probe); ERR_FAIL_COND(!gi_probe); @@ -6245,8 +6332,12 @@ RID RendererStorageRD::gi_probe_get_sdf_texture(RID p_gi_probe) { /* LIGHTMAP API */ -RID RendererStorageRD::lightmap_create() { - return lightmap_owner.make_rid(Lightmap()); +RID RendererStorageRD::lightmap_allocate() { + return lightmap_owner.allocate_rid(); +} + +void RendererStorageRD::lightmap_initialize(RID p_lightmap) { + lightmap_owner.initialize_rid(p_lightmap, Lightmap()); } void RendererStorageRD::lightmap_set_textures(RID p_lightmap, RID p_light, bool p_uses_spherical_haromics) { @@ -6449,7 +6540,8 @@ void RendererStorageRD::_clear_render_target(RenderTarget *rt) { void RendererStorageRD::_update_render_target(RenderTarget *rt) { if (rt->texture.is_null()) { //create a placeholder until updated - rt->texture = texture_2d_placeholder_create(); + rt->texture = texture_allocate(); + texture_2d_placeholder_initialize(rt->texture); Texture *tex = texture_owner.getornull(rt->texture); tex->is_render_target = true; } @@ -6949,7 +7041,7 @@ void RendererStorageRD::render_target_sdf_process(RID p_render_target) { RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rt->sdf_buffer_process_uniform_sets[1], 0); //fill [0] RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(RenderTargetSDF::PushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, push_constant.size[0], push_constant.size[1], 1, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, push_constant.size[0], push_constant.size[1], 1); /* Process */ @@ -6965,7 +7057,7 @@ void RendererStorageRD::render_target_sdf_process(RID p_render_target) { RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rt->sdf_buffer_process_uniform_sets[swap ? 1 : 0], 0); push_constant.stride = stride; RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(RenderTargetSDF::PushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, push_constant.size[0], push_constant.size[1], 1, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, push_constant.size[0], push_constant.size[1], 1); stride /= 2; swap = !swap; RD::get_singleton()->compute_list_add_barrier(compute_list); @@ -6976,7 +7068,7 @@ void RendererStorageRD::render_target_sdf_process(RID p_render_target) { RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, rt_sdf.pipelines[shrink ? RenderTargetSDF::SHADER_STORE_SHRINK : RenderTargetSDF::SHADER_STORE]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rt->sdf_buffer_process_uniform_sets[swap ? 1 : 0], 0); RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(RenderTargetSDF::PushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, push_constant.size[0], push_constant.size[1], 1, 8, 8, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, push_constant.size[0], push_constant.size[1], 1); RD::get_singleton()->compute_list_end(); } @@ -7340,7 +7432,7 @@ void RendererStorageRD::_update_decal_atlas() { tformat.shareable_formats.push_back(RD::DATA_FORMAT_R8G8B8A8_SRGB); decal_atlas.texture = RD::get_singleton()->texture_create(tformat, RD::TextureView()); - RD::get_singleton()->texture_clear(decal_atlas.texture, Color(0, 0, 0, 0), 0, decal_atlas.mipmaps, 0, 1, true); + RD::get_singleton()->texture_clear(decal_atlas.texture, Color(0, 0, 0, 0), 0, decal_atlas.mipmaps, 0, 1); { //create the framebuffer @@ -7395,7 +7487,7 @@ void RendererStorageRD::_update_decal_atlas() { prev_texture = mm.texture; } } else { - RD::get_singleton()->texture_clear(mm.texture, clear_color, 0, 1, 0, 1, false); + RD::get_singleton()->texture_clear(mm.texture, clear_color, 0, 1, 0, 1); } } } @@ -8161,29 +8253,38 @@ bool RendererStorageRD::free(RID p_rid) { material_owner.free(p_rid); } else if (mesh_owner.owns(p_rid)) { mesh_clear(p_rid); + mesh_set_shadow_mesh(p_rid, RID()); Mesh *mesh = mesh_owner.getornull(p_rid); mesh->dependency.deleted_notify(p_rid); if (mesh->instances.size()) { ERR_PRINT("deleting mesh with active instances"); } + if (mesh->shadow_owners.size()) { + for (Set<Mesh *>::Element *E = mesh->shadow_owners.front(); E; E = E->next()) { + Mesh *shadow_owner = E->get(); + shadow_owner->shadow_mesh = RID(); + shadow_owner->dependency.changed_notify(DEPENDENCY_CHANGED_MESH); + } + } mesh_owner.free(p_rid); } else if (mesh_instance_owner.owns(p_rid)) { MeshInstance *mi = mesh_instance_owner.getornull(p_rid); _mesh_instance_clear(mi); mi->mesh->instances.erase(mi->I); mi->I = nullptr; + mesh_instance_owner.free(p_rid); memdelete(mi); } else if (multimesh_owner.owns(p_rid)) { _update_dirty_multimeshes(); - multimesh_allocate(p_rid, 0, RS::MULTIMESH_TRANSFORM_2D); + multimesh_allocate_data(p_rid, 0, RS::MULTIMESH_TRANSFORM_2D); MultiMesh *multimesh = multimesh_owner.getornull(p_rid); multimesh->dependency.deleted_notify(p_rid); multimesh_owner.free(p_rid); } else if (skeleton_owner.owns(p_rid)) { _update_dirty_skeletons(); - skeleton_allocate(p_rid, 0); + skeleton_allocate_data(p_rid, 0); Skeleton *skeleton = skeleton_owner.getornull(p_rid); skeleton->dependency.deleted_notify(p_rid); skeleton_owner.free(p_rid); @@ -8201,7 +8302,7 @@ bool RendererStorageRD::free(RID p_rid) { decal->dependency.deleted_notify(p_rid); decal_owner.free(p_rid); } else if (gi_probe_owner.owns(p_rid)) { - gi_probe_allocate(p_rid, Transform(), AABB(), Vector3i(), Vector<uint8_t>(), Vector<uint8_t>(), Vector<uint8_t>(), Vector<int>()); //deallocate + gi_probe_allocate_data(p_rid, Transform(), AABB(), Vector3i(), Vector<uint8_t>(), Vector<uint8_t>(), Vector<uint8_t>(), Vector<int>()); //deallocate GIProbe *gi_probe = gi_probe_owner.getornull(p_rid); gi_probe->dependency.deleted_notify(p_rid); gi_probe_owner.free(p_rid); @@ -8257,11 +8358,11 @@ EffectsRD *RendererStorageRD::get_effects() { } void RendererStorageRD::capture_timestamps_begin() { - RD::get_singleton()->capture_timestamp("Frame Begin", false); + RD::get_singleton()->capture_timestamp("Frame Begin"); } void RendererStorageRD::capture_timestamp(const String &p_name) { - RD::get_singleton()->capture_timestamp(p_name, true); + RD::get_singleton()->capture_timestamp(p_name); } uint32_t RendererStorageRD::get_captured_timestamps_count() const { @@ -8295,7 +8396,7 @@ RendererStorageRD::RendererStorageRD() { static_assert(sizeof(GlobalVariables::Value) == 16); - global_variables.buffer_size = GLOBAL_GET("rendering/high_end/global_shader_variables_buffer_size"); + global_variables.buffer_size = GLOBAL_GET("rendering/limits/global_shader_variables/buffer_size"); global_variables.buffer_size = MAX(4096, global_variables.buffer_size); global_variables.buffer_values = memnew_arr(GlobalVariables::Value, global_variables.buffer_size); zeromem(global_variables.buffer_values, sizeof(GlobalVariables::Value) * global_variables.buffer_size); @@ -8563,14 +8664,14 @@ RendererStorageRD::RendererStorageRD() { sampler_state.min_filter = RD::SAMPLER_FILTER_LINEAR; sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; sampler_state.use_anisotropy = true; - sampler_state.anisotropy_max = 1 << int(GLOBAL_GET("rendering/quality/texture_filters/anisotropic_filtering_level")); + sampler_state.anisotropy_max = 1 << int(GLOBAL_GET("rendering/textures/default_filters/anisotropic_filtering_level")); } break; case RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC: { sampler_state.mag_filter = RD::SAMPLER_FILTER_LINEAR; sampler_state.min_filter = RD::SAMPLER_FILTER_LINEAR; sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; sampler_state.use_anisotropy = true; - sampler_state.anisotropy_max = 1 << int(GLOBAL_GET("rendering/quality/texture_filters/anisotropic_filtering_level")); + sampler_state.anisotropy_max = 1 << int(GLOBAL_GET("rendering/textures/default_filters/anisotropic_filtering_level")); } break; default: { @@ -8740,7 +8841,7 @@ RendererStorageRD::RendererStorageRD() { } } - lightmap_probe_capture_update_speed = GLOBAL_GET("rendering/lightmapper/probe_capture_update_speed"); + lightmap_probe_capture_update_speed = GLOBAL_GET("rendering/lightmapping/probe_capture/update_speed"); /* Particles */ @@ -8807,9 +8908,11 @@ RendererStorageRD::RendererStorageRD() { { // default material and shader for particles shader - particles_shader.default_shader = shader_create(); + particles_shader.default_shader = shader_allocate(); + shader_initialize(particles_shader.default_shader); shader_set_code(particles_shader.default_shader, "shader_type particles; void compute() { COLOR = vec4(1.0); } \n"); - particles_shader.default_material = material_create(); + particles_shader.default_material = material_allocate(); + material_initialize(particles_shader.default_material); material_set_shader(particles_shader.default_material, particles_shader.default_shader); ParticlesMaterialData *md = (ParticlesMaterialData *)material_get_data(particles_shader.default_material, RendererStorageRD::SHADER_TYPE_PARTICLES); diff --git a/servers/rendering/renderer_rd/renderer_storage_rd.h b/servers/rendering/renderer_rd/renderer_storage_rd.h index 2fb66ac573..68256dc155 100644 --- a/servers/rendering/renderer_rd/renderer_storage_rd.h +++ b/servers/rendering/renderer_rd/renderer_storage_rd.h @@ -221,7 +221,7 @@ private: ~CanvasTexture(); }; - RID_PtrOwner<CanvasTexture> canvas_texture_owner; + RID_PtrOwner<CanvasTexture, true> canvas_texture_owner; /* TEXTURE API */ struct Texture { @@ -367,7 +367,7 @@ private: }; ShaderDataRequestFunction shader_data_request_func[SHADER_TYPE_MAX]; - mutable RID_Owner<Shader> shader_owner; + mutable RID_Owner<Shader, true> shader_owner; /* Material */ @@ -389,7 +389,7 @@ private: }; MaterialDataRequestFunction material_data_request_func[SHADER_TYPE_MAX]; - mutable RID_Owner<Material> material_owner; + mutable RID_Owner<Material, true> material_owner; Material *material_update_list; void _material_queue_update(Material *material, bool p_uniform, bool p_texture); @@ -478,10 +478,13 @@ private: List<MeshInstance *> instances; + RID shadow_mesh; + Set<Mesh *> shadow_owners; + Dependency dependency; }; - mutable RID_Owner<Mesh> mesh_owner; + mutable RID_Owner<Mesh, true> mesh_owner; struct MeshInstance { Mesh *mesh; @@ -584,7 +587,7 @@ private: Dependency dependency; }; - mutable RID_Owner<MultiMesh> multimesh_owner; + mutable RID_Owner<MultiMesh, true> multimesh_owner; MultiMesh *multimesh_dirty_list = nullptr; @@ -890,7 +893,7 @@ private: void update_particles(); - mutable RID_Owner<Particles> particles_owner; + mutable RID_Owner<Particles, true> particles_owner; /* Particles Collision */ @@ -912,7 +915,7 @@ private: Dependency dependency; }; - mutable RID_Owner<ParticlesCollision> particles_collision_owner; + mutable RID_Owner<ParticlesCollision, true> particles_collision_owner; struct ParticlesCollisionInstance { RID collision; @@ -942,7 +945,7 @@ private: Dependency dependency; }; - mutable RID_Owner<Skeleton> skeleton_owner; + mutable RID_Owner<Skeleton, true> skeleton_owner; _FORCE_INLINE_ void _skeleton_make_dirty(Skeleton *skeleton); @@ -974,7 +977,7 @@ private: Dependency dependency; }; - mutable RID_Owner<Light> light_owner; + mutable RID_Owner<Light, true> light_owner; /* REFLECTION PROBE */ @@ -997,7 +1000,7 @@ private: Dependency dependency; }; - mutable RID_Owner<ReflectionProbe> reflection_probe_owner; + mutable RID_Owner<ReflectionProbe, true> reflection_probe_owner; /* DECAL */ @@ -1018,7 +1021,7 @@ private: Dependency dependency; }; - mutable RID_Owner<Decal> decal_owner; + mutable RID_Owner<Decal, true> decal_owner; /* GI PROBE */ @@ -1061,7 +1064,7 @@ private: RID giprobe_sdf_shader_version_shader; RID giprobe_sdf_shader_pipeline; - mutable RID_Owner<GIProbe> gi_probe_owner; + mutable RID_Owner<GIProbe, true> gi_probe_owner; /* REFLECTION PROBE */ @@ -1092,7 +1095,7 @@ private: uint64_t lightmap_array_version = 0; - mutable RID_Owner<Lightmap> lightmap_owner; + mutable RID_Owner<Lightmap, true> lightmap_owner; float lightmap_probe_capture_update_speed = 4; @@ -1246,12 +1249,16 @@ private: EffectsRD effects; public: + virtual bool can_create_resources_async() const; + /* TEXTURE API */ - virtual RID texture_2d_create(const Ref<Image> &p_image); - virtual RID texture_2d_layered_create(const Vector<Ref<Image>> &p_layers, RS::TextureLayeredType p_layered_type); - virtual RID texture_3d_create(Image::Format p_format, int p_width, int p_height, int p_depth, bool p_mipmaps, const Vector<Ref<Image>> &p_data); //all slices, then all the mipmaps, must be coherent - virtual RID texture_proxy_create(RID p_base); + virtual RID texture_allocate(); + + virtual void texture_2d_initialize(RID p_texture, const Ref<Image> &p_image); + virtual void texture_2d_layered_initialize(RID p_texture, const Vector<Ref<Image>> &p_layers, RS::TextureLayeredType p_layered_type); + virtual void texture_3d_initialize(RID p_texture, Image::Format p_format, int p_width, int p_height, int p_depth, bool p_mipmaps, const Vector<Ref<Image>> &p_data); //all slices, then all the mipmaps, must be coherent + virtual void texture_proxy_initialize(RID p_texture, RID p_base); virtual void _texture_2d_update(RID p_texture, const Ref<Image> &p_image, int p_layer, bool p_immediate); @@ -1261,9 +1268,9 @@ public: virtual void texture_proxy_update(RID p_texture, RID p_proxy_to); //these two APIs can be used together or in combination with the others. - virtual RID texture_2d_placeholder_create(); - virtual RID texture_2d_layered_placeholder_create(RenderingServer::TextureLayeredType p_layered_type); - virtual RID texture_3d_placeholder_create(); + virtual void texture_2d_placeholder_initialize(RID p_texture); + virtual void texture_2d_layered_placeholder_initialize(RID p_texture, RenderingServer::TextureLayeredType p_layered_type); + virtual void texture_3d_placeholder_initialize(RID p_texture); virtual Ref<Image> texture_2d_get(RID p_texture) const; virtual Ref<Image> texture_2d_layer_get(RID p_texture, int p_layer) const; @@ -1335,7 +1342,8 @@ public: /* CANVAS TEXTURE API */ - virtual RID canvas_texture_create(); + RID canvas_texture_allocate(); + void canvas_texture_initialize(RID p_canvas_texture); virtual void canvas_texture_set_channel(RID p_canvas_texture, RS::CanvasTextureChannel p_channel, RID p_texture); virtual void canvas_texture_set_shading_parameters(RID p_canvas_texture, const Color &p_specular_color, float p_shininess); @@ -1347,7 +1355,8 @@ public: /* SHADER API */ - RID shader_create(); + RID shader_allocate(); + void shader_initialize(RID p_shader); void shader_set_code(RID p_shader, const String &p_code); String shader_get_code(RID p_shader) const; @@ -1362,7 +1371,8 @@ public: /* COMMON MATERIAL API */ - RID material_create(); + RID material_allocate(); + void material_initialize(RID p_material); void material_set_shader(RID p_material, RID p_shader); @@ -1398,7 +1408,8 @@ public: /* MESH API */ - virtual RID mesh_create(); + RID mesh_allocate(); + void mesh_initialize(RID p_mesh); virtual void mesh_set_blend_shape_count(RID p_mesh, int p_blend_shape_count); @@ -1423,6 +1434,7 @@ public: virtual AABB mesh_get_custom_aabb(RID p_mesh) const; virtual AABB mesh_get_aabb(RID p_mesh, RID p_skeleton = RID()); + virtual void mesh_set_shadow_mesh(RID p_mesh, RID p_shadow_mesh); virtual void mesh_clear(RID p_mesh); @@ -1461,6 +1473,13 @@ public: return mesh->surfaces[p_surface_index]; } + _FORCE_INLINE_ RID mesh_get_shadow_mesh(RID p_mesh) { + Mesh *mesh = mesh_owner.getornull(p_mesh); + ERR_FAIL_COND_V(!mesh, RID()); + + return mesh->shadow_mesh; + } + _FORCE_INLINE_ RS::PrimitiveType mesh_surface_get_primitive(void *p_surface) { Mesh::Surface *surface = reinterpret_cast<Mesh::Surface *>(p_surface); return surface->primitive; @@ -1471,13 +1490,7 @@ public: return s->lod_count > 0; } - _FORCE_INLINE_ RID mesh_surface_get_index_array(void *p_surface) const { - Mesh::Surface *s = reinterpret_cast<Mesh::Surface *>(p_surface); - - return s->index_array; - } - - _FORCE_INLINE_ RID mesh_surface_get_index_array_with_lod(void *p_surface, float p_model_scale, float p_distance_threshold, float p_lod_threshold) const { + _FORCE_INLINE_ uint32_t mesh_surface_get_lod(void *p_surface, float p_model_scale, float p_distance_threshold, float p_lod_threshold) const { Mesh::Surface *s = reinterpret_cast<Mesh::Surface *>(p_surface); int32_t current_lod = -1; @@ -1489,9 +1502,19 @@ public: current_lod = i; } if (current_lod == -1) { + return 0; + } else { + return current_lod + 1; + } + } + + _FORCE_INLINE_ RID mesh_surface_get_index_array(void *p_surface, uint32_t p_lod) const { + Mesh::Surface *s = reinterpret_cast<Mesh::Surface *>(p_surface); + + if (p_lod == 0) { return s->index_array; } else { - return s->lods[current_lod].index_array; + return s->lods[p_lod - 1].index_array; } } @@ -1607,9 +1630,10 @@ public: /* MULTIMESH API */ - RID multimesh_create(); + RID multimesh_allocate(); + void multimesh_initialize(RID p_multimesh); - void multimesh_allocate(RID p_multimesh, int p_instances, RS::MultimeshTransformFormat p_transform_format, bool p_use_colors = false, bool p_use_custom_data = false); + void multimesh_allocate_data(RID p_multimesh, int p_instances, RS::MultimeshTransformFormat p_transform_format, bool p_use_colors = false, bool p_use_custom_data = false); int multimesh_get_instance_count(RID p_multimesh) const; void multimesh_set_mesh(RID p_multimesh, RID p_mesh); @@ -1673,24 +1697,28 @@ public: /* IMMEDIATE API */ - RID immediate_create() { return RID(); } - void immediate_begin(RID p_immediate, RS::PrimitiveType p_rimitive, RID p_texture = RID()) {} - void immediate_vertex(RID p_immediate, const Vector3 &p_vertex) {} - void immediate_normal(RID p_immediate, const Vector3 &p_normal) {} - void immediate_tangent(RID p_immediate, const Plane &p_tangent) {} - void immediate_color(RID p_immediate, const Color &p_color) {} - void immediate_uv(RID p_immediate, const Vector2 &tex_uv) {} - void immediate_uv2(RID p_immediate, const Vector2 &tex_uv) {} - void immediate_end(RID p_immediate) {} - void immediate_clear(RID p_immediate) {} - void immediate_set_material(RID p_immediate, RID p_material) {} - RID immediate_get_material(RID p_immediate) const { return RID(); } - AABB immediate_get_aabb(RID p_immediate) const { return AABB(); } + RID immediate_allocate() { return RID(); } + void immediate_initialize(RID p_immediate) {} + + virtual void immediate_begin(RID p_immediate, RS::PrimitiveType p_rimitive, RID p_texture = RID()) {} + virtual void immediate_vertex(RID p_immediate, const Vector3 &p_vertex) {} + virtual void immediate_normal(RID p_immediate, const Vector3 &p_normal) {} + virtual void immediate_tangent(RID p_immediate, const Plane &p_tangent) {} + virtual void immediate_color(RID p_immediate, const Color &p_color) {} + virtual void immediate_uv(RID p_immediate, const Vector2 &tex_uv) {} + virtual void immediate_uv2(RID p_immediate, const Vector2 &tex_uv) {} + virtual void immediate_end(RID p_immediate) {} + virtual void immediate_clear(RID p_immediate) {} + virtual void immediate_set_material(RID p_immediate, RID p_material) {} + virtual RID immediate_get_material(RID p_immediate) const { return RID(); } + virtual AABB immediate_get_aabb(RID p_immediate) const { return AABB(); } /* SKELETON API */ - RID skeleton_create(); - void skeleton_allocate(RID p_skeleton, int p_bones, bool p_2d_skeleton = false); + RID skeleton_allocate(); + void skeleton_initialize(RID p_skeleton); + + void skeleton_allocate_data(RID p_skeleton, int p_bones, bool p_2d_skeleton = false); void skeleton_set_base_transform_2d(RID p_skeleton, const Transform2D &p_base_transform); void skeleton_set_world_transform(RID p_skeleton, bool p_enable, const Transform &p_world_transform); int skeleton_get_bone_count(RID p_skeleton) const; @@ -1724,11 +1752,16 @@ public: } /* Light API */ - RID light_create(RS::LightType p_type); + void _light_initialize(RID p_rid, RS::LightType p_type); + + RID directional_light_allocate(); + void directional_light_initialize(RID p_light); + + RID omni_light_allocate(); + void omni_light_initialize(RID p_light); - RID directional_light_create() { return light_create(RS::LIGHT_DIRECTIONAL); } - RID omni_light_create() { return light_create(RS::LIGHT_OMNI); } - RID spot_light_create() { return light_create(RS::LIGHT_SPOT); } + RID spot_light_allocate(); + void spot_light_initialize(RID p_light); void light_set_color(RID p_light, const Color &p_color); void light_set_param(RID p_light, RS::LightParam p_param, float p_value); @@ -1831,7 +1864,8 @@ public: /* PROBE API */ - RID reflection_probe_create(); + RID reflection_probe_allocate(); + void reflection_probe_initialize(RID p_reflection_probe); void reflection_probe_set_update_mode(RID p_probe, RS::ReflectionProbeUpdateMode p_mode); void reflection_probe_set_intensity(RID p_probe, float p_intensity); @@ -1871,7 +1905,9 @@ public: /* DECAL API */ - virtual RID decal_create(); + RID decal_allocate(); + void decal_initialize(RID p_decal); + virtual void decal_set_extents(RID p_decal, const Vector3 &p_extents); virtual void decal_set_texture(RID p_decal, RS::DecalTexture p_type, RID p_texture); virtual void decal_set_emission_energy(RID p_decal, float p_energy); @@ -1946,9 +1982,10 @@ public: /* GI PROBE API */ - RID gi_probe_create(); + RID gi_probe_allocate(); + void gi_probe_initialize(RID p_gi_probe); - void gi_probe_allocate(RID p_gi_probe, const Transform &p_to_cell_xform, const AABB &p_aabb, const Vector3i &p_octree_size, const Vector<uint8_t> &p_octree_cells, const Vector<uint8_t> &p_data_cells, const Vector<uint8_t> &p_distance_field, const Vector<int> &p_level_counts); + void gi_probe_allocate_data(RID p_gi_probe, const Transform &p_to_cell_xform, const AABB &p_aabb, const Vector3i &p_octree_size, const Vector<uint8_t> &p_octree_cells, const Vector<uint8_t> &p_data_cells, const Vector<uint8_t> &p_distance_field, const Vector<int> &p_level_counts); AABB gi_probe_get_bounds(RID p_gi_probe) const; Vector3i gi_probe_get_octree_size(RID p_gi_probe) const; @@ -1999,7 +2036,8 @@ public: /* LIGHTMAP CAPTURE */ - virtual RID lightmap_create(); + RID lightmap_allocate(); + void lightmap_initialize(RID p_lightmap); virtual void lightmap_set_textures(RID p_lightmap, RID p_light, bool p_uses_spherical_haromics); virtual void lightmap_set_probe_bounds(RID p_lightmap, const AABB &p_bounds); @@ -2048,7 +2086,8 @@ public: /* PARTICLES */ - RID particles_create(); + RID particles_allocate(); + void particles_initialize(RID p_particles_collision); void particles_set_emitting(RID p_particles, bool p_emitting); void particles_set_amount(RID p_particles, int p_amount); @@ -2126,7 +2165,9 @@ public: /* PARTICLES COLLISION */ - virtual RID particles_collision_create(); + RID particles_collision_allocate(); + void particles_collision_initialize(RID p_particles_collision); + virtual void particles_collision_set_collision_type(RID p_particles_collision, RS::ParticlesCollisionType p_type); virtual void particles_collision_set_cull_mask(RID p_particles_collision, uint32_t p_cull_mask); virtual void particles_collision_set_sphere_radius(RID p_particles_collision, float p_radius); //for spheres diff --git a/servers/rendering/renderer_rd/shader_compiler_rd.cpp b/servers/rendering/renderer_rd/shader_compiler_rd.cpp index e77141b26c..8135d388e1 100644 --- a/servers/rendering/renderer_rd/shader_compiler_rd.cpp +++ b/servers/rendering/renderer_rd/shader_compiler_rd.cpp @@ -687,7 +687,15 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge uint32_t index = p_default_actions.base_varying_index; + List<Pair<StringName, SL::ShaderNode::Varying>> var_frag_to_light; + for (Map<StringName, SL::ShaderNode::Varying>::Element *E = pnode->varyings.front(); E; E = E->next()) { + if (E->get().stage == SL::ShaderNode::Varying::STAGE_FRAGMENT_TO_LIGHT || E->get().stage == SL::ShaderNode::Varying::STAGE_FRAGMENT) { + var_frag_to_light.push_back(Pair<StringName, SL::ShaderNode::Varying>(E->key(), E->get())); + fragment_varyings.insert(E->key()); + continue; + } + String vcode; String interp_mode = _interpstr(E->get().interpolation); vcode += _prestr(E->get().precision); @@ -705,6 +713,21 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge index++; } + if (var_frag_to_light.size() > 0) { + String gcode = "\n\nstruct {\n"; + for (List<Pair<StringName, SL::ShaderNode::Varying>>::Element *E = var_frag_to_light.front(); E; E = E->next()) { + gcode += "\t" + _prestr(E->get().second.precision) + _typestr(E->get().second.type) + " " + _mkid(E->get().first); + if (E->get().second.array_size > 0) { + gcode += "["; + gcode += itos(E->get().second.array_size); + gcode += "]"; + } + gcode += ";\n"; + } + gcode += "} frag_to_light;\n"; + r_gen_code.fragment_global += gcode; + } + for (int i = 0; i < pnode->vconstants.size(); i++) { const SL::ShaderNode::Constant &cnode = pnode->vconstants[i]; String gcode; @@ -833,6 +856,19 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge } break; case SL::Node::TYPE_VARIABLE: { SL::VariableNode *vnode = (SL::VariableNode *)p_node; + bool use_fragment_varying = false; + + if (current_func_name != vertex_name) { + if (p_assigning) { + if (shader->varyings.has(vnode->name)) { + use_fragment_varying = true; + } + } else { + if (fragment_varyings.has(vnode->name)) { + use_fragment_varying = true; + } + } + } if (p_assigning && p_actions.write_flag_pointers.has(vnode->name)) { *p_actions.write_flag_pointers[vnode->name] = true; @@ -877,7 +913,10 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge } } else { - code = _mkid(vnode->name); //its something else (local var most likely) use as is + if (use_fragment_varying) { + code = "frag_to_light."; + } + code += _mkid(vnode->name); //its something else (local var most likely) use as is } } @@ -962,6 +1001,23 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge } break; case SL::Node::TYPE_ARRAY: { SL::ArrayNode *anode = (SL::ArrayNode *)p_node; + bool use_fragment_varying = false; + + if (current_func_name != vertex_name) { + if (anode->assign_expression != nullptr) { + use_fragment_varying = true; + } else { + if (p_assigning) { + if (shader->varyings.has(anode->name)) { + use_fragment_varying = true; + } + } else { + if (fragment_varyings.has(anode->name)) { + use_fragment_varying = true; + } + } + } + } if (p_assigning && p_actions.write_flag_pointers.has(anode->name)) { *p_actions.write_flag_pointers[anode->name] = true; @@ -984,7 +1040,10 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge if (p_default_actions.renames.has(anode->name)) { code = p_default_actions.renames[anode->name]; } else { - code = _mkid(anode->name); + if (use_fragment_varying) { + code = "frag_to_light."; + } + code += _mkid(anode->name); } if (anode->call_expression != nullptr) { @@ -1277,6 +1336,7 @@ Error ShaderCompilerRD::compile(RS::ShaderMode p_mode, const String &p_code, Ide used_name_defines.clear(); used_rmode_defines.clear(); used_flag_pointers.clear(); + fragment_varyings.clear(); shader = parser.get_shader(); function = nullptr; @@ -1408,7 +1468,7 @@ ShaderCompilerRD::ShaderCompilerRD() { actions[RS::SHADER_SPATIAL].render_mode_defines["cull_front"] = "#define DO_SIDE_CHECK\n"; actions[RS::SHADER_SPATIAL].render_mode_defines["cull_disabled"] = "#define DO_SIDE_CHECK\n"; - bool force_lambert = GLOBAL_GET("rendering/quality/shading/force_lambert_over_burley"); + bool force_lambert = GLOBAL_GET("rendering/shading/overrides/force_lambert_over_burley"); if (!force_lambert) { actions[RS::SHADER_SPATIAL].render_mode_defines["diffuse_burley"] = "#define DIFFUSE_BURLEY\n"; @@ -1418,7 +1478,7 @@ ShaderCompilerRD::ShaderCompilerRD() { actions[RS::SHADER_SPATIAL].render_mode_defines["diffuse_lambert_wrap"] = "#define DIFFUSE_LAMBERT_WRAP\n"; actions[RS::SHADER_SPATIAL].render_mode_defines["diffuse_toon"] = "#define DIFFUSE_TOON\n"; - bool force_blinn = GLOBAL_GET("rendering/quality/shading/force_blinn_over_ggx"); + bool force_blinn = GLOBAL_GET("rendering/shading/overrides/force_blinn_over_ggx"); if (!force_blinn) { actions[RS::SHADER_SPATIAL].render_mode_defines["specular_schlick_ggx"] = "#define SPECULAR_SCHLICK_GGX\n"; diff --git a/servers/rendering/renderer_rd/shader_compiler_rd.h b/servers/rendering/renderer_rd/shader_compiler_rd.h index d127d8e01c..6575829e73 100644 --- a/servers/rendering/renderer_rd/shader_compiler_rd.h +++ b/servers/rendering/renderer_rd/shader_compiler_rd.h @@ -114,6 +114,7 @@ private: Set<StringName> used_flag_pointers; Set<StringName> used_rmode_defines; Set<StringName> internal_functions; + Set<StringName> fragment_varyings; DefaultIdentifierActions actions; diff --git a/servers/rendering/renderer_rd/shader_rd.cpp b/servers/rendering/renderer_rd/shader_rd.cpp index 2ae22a8a38..e4a39ff813 100644 --- a/servers/rendering/renderer_rd/shader_rd.cpp +++ b/servers/rendering/renderer_rd/shader_rd.cpp @@ -301,6 +301,7 @@ void ShaderRD::_compile_variant(uint32_t p_variant, Version *p_version) { builder.append(compute_codev.get_data()); // version info (if exists) builder.append("\n"); //make sure defines begin at newline + builder.append(base_compute_defines.get_data()); builder.append(general_defines.get_data()); builder.append(variant_defines[p_variant].get_data()); @@ -401,7 +402,6 @@ RS::ShaderNativeSourceCode ShaderRD::version_get_native_source_code(RID p_versio builder.append(fragment_codev.get_data()); // version info (if exists) builder.append("\n"); //make sure defines begin at newline - builder.append(general_defines.get_data()); builder.append(variant_defines[i].get_data()); for (int j = 0; j < version->custom_defines.size(); j++) { @@ -440,6 +440,7 @@ RS::ShaderNativeSourceCode ShaderRD::version_get_native_source_code(RID p_versio builder.append(compute_codev.get_data()); // version info (if exists) builder.append("\n"); //make sure defines begin at newline + builder.append(base_compute_defines.get_data()); builder.append(general_defines.get_data()); builder.append(variant_defines[i].get_data()); @@ -596,6 +597,22 @@ bool ShaderRD::is_variant_enabled(int p_variant) const { return variants_enabled[p_variant]; } +ShaderRD::ShaderRD() { + // Do not feel forced to use this, in most cases it makes little to no difference. + bool use_32_threads = false; + if (RD::get_singleton()->get_device_vendor_name() == "NVIDIA") { + use_32_threads = true; + } + String base_compute_define_text; + if (use_32_threads) { + base_compute_define_text = "\n#define NATIVE_LOCAL_GROUP_SIZE 32\n#define NATIVE_LOCAL_SIZE_2D_X 8\n#define NATIVE_LOCAL_SIZE_2D_Y 4\n"; + } else { + base_compute_define_text = "\n#define NATIVE_LOCAL_GROUP_SIZE 64\n#define NATIVE_LOCAL_SIZE_2D_X 8\n#define NATIVE_LOCAL_SIZE_2D_Y 8\n"; + } + + base_compute_defines = base_compute_define_text.ascii(); +} + void ShaderRD::initialize(const Vector<String> &p_variant_defines, const String &p_general_defines) { ERR_FAIL_COND(variant_defines.size()); ERR_FAIL_COND(p_variant_defines.size() == 0); diff --git a/servers/rendering/renderer_rd/shader_rd.h b/servers/rendering/renderer_rd/shader_rd.h index a3474c6f93..e0f4dcf2d0 100644 --- a/servers/rendering/renderer_rd/shader_rd.h +++ b/servers/rendering/renderer_rd/shader_rd.h @@ -99,8 +99,10 @@ class ShaderRD { const char *name; + CharString base_compute_defines; + protected: - ShaderRD() {} + ShaderRD(); void setup(const char *p_vertex_code, const char *p_fragment_code, const char *p_compute_code, const char *p_name); public: diff --git a/servers/rendering/renderer_rd/shaders/SCsub b/servers/rendering/renderer_rd/shaders/SCsub index 1b0197c1c1..c192574ff2 100644 --- a/servers/rendering/renderer_rd/shaders/SCsub +++ b/servers/rendering/renderer_rd/shaders/SCsub @@ -39,7 +39,6 @@ if "RD_GLSL" in env["BUILDERS"]: env.RD_GLSL("sdfgi_debug.glsl") env.RD_GLSL("sdfgi_debug_probes.glsl") env.RD_GLSL("volumetric_fog.glsl") - env.RD_GLSL("shadow_reduce.glsl") env.RD_GLSL("particles.glsl") env.RD_GLSL("particles_copy.glsl") env.RD_GLSL("sort.glsl") diff --git a/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl b/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl index 54d67db6c6..c3ac0bee57 100644 --- a/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl +++ b/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl @@ -1,33 +1,48 @@ -#[compute] +#[vertex] #version 450 VERSION_DEFINES -layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; +layout(push_constant, binding = 1, std430) uniform Params { + float z_far; + float z_near; + bool z_flip; + uint pad; + vec4 screen_rect; +} +params; + +layout(location = 0) out vec2 uv_interp; + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + vec2 screen_pos = uv_interp * params.screen_rect.zw + params.screen_rect.xy; + gl_Position = vec4(screen_pos * 2.0 - 1.0, 0.0, 1.0); +} + +#[fragment] + +#version 450 + +VERSION_DEFINES + +layout(location = 0) in vec2 uv_interp; layout(set = 0, binding = 0) uniform samplerCube source_cube; layout(push_constant, binding = 1, std430) uniform Params { - ivec2 screen_size; - ivec2 offset; - float bias; float z_far; float z_near; bool z_flip; + uint pad; + vec4 screen_rect; } params; -layout(r32f, set = 1, binding = 0) uniform restrict writeonly image2D depth_buffer; - void main() { - ivec2 pos = ivec2(gl_GlobalInvocationID.xy); - if (any(greaterThan(pos, params.screen_size))) { //too large, do nothing - return; - } - - vec2 pixel_size = 1.0 / vec2(params.screen_size); - vec2 uv = (vec2(pos) + 0.5) * pixel_size; + vec2 uv = uv_interp; vec3 normal = vec3(uv * 2.0 - 1.0, 0.0); @@ -65,5 +80,5 @@ void main() { float linear_depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - depth * (params.z_far - params.z_near)); depth = (linear_depth * depth_fix) / params.z_far; - imageStore(depth_buffer, pos + params.offset, vec4(depth)); + gl_FragDepth = depth; } diff --git a/servers/rendering/renderer_rd/shaders/gi.glsl b/servers/rendering/renderer_rd/shaders/gi.glsl index c2965f9874..92a5682572 100644 --- a/servers/rendering/renderer_rd/shaders/gi.glsl +++ b/servers/rendering/renderer_rd/shaders/gi.glsl @@ -97,13 +97,12 @@ layout(push_constant, binding = 0, std430) uniform Params { vec4 proj_info; + vec3 ao_color; uint max_giprobes; + bool high_quality_vct; - uint pad2; bool orthogonal; - - vec3 ao_color; - uint pad; + uint pad[2]; mat3x4 cam_rotation; } @@ -364,57 +363,62 @@ void sdfgi_process(vec3 vertex, vec3 normal, vec3 reflection, float roughness, o ray_pos += (ray_dir * 1.0 / max(abs_ray_dir.x, max(abs_ray_dir.y, abs_ray_dir.z)) + cam_normal * 1.4) * bias / sdfgi.cascades[cascade].to_cell; } float softness = 0.2 + min(1.0, roughness * 5.0) * 4.0; //approximation to roughness so it does not seem like a hard fade - while (length(ray_pos) < max_distance) { - for (uint i = 0; i < sdfgi.max_cascades; i++) { - if (i >= cascade && length(ray_pos) < radius_sizes[i]) { - cascade = max(i, cascade); //never go down - - vec3 pos = ray_pos - sdfgi.cascades[i].position; - pos *= sdfgi.cascades[i].to_cell * pos_to_uvw; - - float distance = texture(sampler3D(sdf_cascades[i], linear_sampler), pos).r * 255.0 - 1.1; - - vec4 hit_light = vec4(0.0); - if (distance < softness) { - hit_light.rgb = texture(sampler3D(light_cascades[i], linear_sampler), pos).rgb; - hit_light.rgb *= 0.5; //approximation given value read is actually meant for anisotropy - hit_light.a = clamp(1.0 - (distance / softness), 0.0, 1.0); - hit_light.rgb *= hit_light.a; - } + uint i = 0; + bool found = false; + while (true) { + if (length(ray_pos) >= max_distance || light_accum.a > 0.99) { + break; + } + if (!found && i >= cascade && length(ray_pos) < radius_sizes[i]) { + uint next_i = min(i + 1, sdfgi.max_cascades - 1); + cascade = max(i, cascade); //never go down - distance /= sdfgi.cascades[i].to_cell; + vec3 pos = ray_pos - sdfgi.cascades[i].position; + pos *= sdfgi.cascades[i].to_cell * pos_to_uvw; - if (i < (sdfgi.max_cascades - 1)) { - pos = ray_pos - sdfgi.cascades[i + 1].position; - pos *= sdfgi.cascades[i + 1].to_cell * pos_to_uvw; + float fdistance = textureLod(sampler3D(sdf_cascades[i], linear_sampler), pos, 0.0).r * 255.0 - 1.1; - float distance2 = texture(sampler3D(sdf_cascades[i + 1], linear_sampler), pos).r * 255.0 - 1.1; + vec4 hit_light = vec4(0.0); + if (fdistance < softness) { + hit_light.rgb = textureLod(sampler3D(light_cascades[i], linear_sampler), pos, 0.0).rgb; + hit_light.rgb *= 0.5; //approximation given value read is actually meant for anisotropy + hit_light.a = clamp(1.0 - (fdistance / softness), 0.0, 1.0); + hit_light.rgb *= hit_light.a; + } - vec4 hit_light2 = vec4(0.0); - if (distance2 < softness) { - hit_light2.rgb = texture(sampler3D(light_cascades[i + 1], linear_sampler), pos).rgb; - hit_light2.rgb *= 0.5; //approximation given value read is actually meant for anisotropy - hit_light2.a = clamp(1.0 - (distance2 / softness), 0.0, 1.0); - hit_light2.rgb *= hit_light2.a; - } + fdistance /= sdfgi.cascades[i].to_cell; - float prev_radius = i == 0 ? 0.0 : radius_sizes[i - 1]; - float blend = clamp((length(ray_pos) - prev_radius) / (radius_sizes[i] - prev_radius), 0.0, 1.0); + if (i < (sdfgi.max_cascades - 1)) { + pos = ray_pos - sdfgi.cascades[next_i].position; + pos *= sdfgi.cascades[next_i].to_cell * pos_to_uvw; - distance2 /= sdfgi.cascades[i + 1].to_cell; + float fdistance2 = textureLod(sampler3D(sdf_cascades[next_i], linear_sampler), pos, 0.0).r * 255.0 - 1.1; - hit_light = mix(hit_light, hit_light2, blend); - distance = mix(distance, distance2, blend); + vec4 hit_light2 = vec4(0.0); + if (fdistance2 < softness) { + hit_light2.rgb = textureLod(sampler3D(light_cascades[next_i], linear_sampler), pos, 0.0).rgb; + hit_light2.rgb *= 0.5; //approximation given value read is actually meant for anisotropy + hit_light2.a = clamp(1.0 - (fdistance2 / softness), 0.0, 1.0); + hit_light2.rgb *= hit_light2.a; } - light_accum += hit_light; - ray_pos += ray_dir * distance; - break; + float prev_radius = i == 0 ? 0.0 : radius_sizes[max(0, i - 1)]; + float blend = clamp((length(ray_pos) - prev_radius) / (radius_sizes[i] - prev_radius), 0.0, 1.0); + + fdistance2 /= sdfgi.cascades[next_i].to_cell; + + hit_light = mix(hit_light, hit_light2, blend); + fdistance = mix(fdistance, fdistance2, blend); } - } - if (light_accum.a > 0.99) { - break; + light_accum += hit_light; + ray_pos += ray_dir * fdistance; + found = true; + } + i++; + if (i == sdfgi.max_cascades) { + i = 0; + found = false; } } @@ -594,28 +598,16 @@ vec4 fetch_normal_and_roughness(ivec2 pos) { return normal_roughness; } -void main() { - // Pixel being shaded - ivec2 pos = ivec2(gl_GlobalInvocationID.xy); - if (any(greaterThanEqual(pos, params.screen_size))) { //too large, do nothing - return; - } - - vec3 vertex = reconstruct_position(pos); - vertex.y = -vertex.y; - +void process_gi(ivec2 pos, vec3 vertex, inout vec4 ambient_light, inout vec4 reflection_light) { vec4 normal_roughness = fetch_normal_and_roughness(pos); - vec3 normal = normal_roughness.xyz; - vec4 ambient_light = vec4(0.0), reflection_light = vec4(0.0); + vec3 normal = normal_roughness.xyz; if (normal.length() > 0.5) { //valid normal, can do GI float roughness = normal_roughness.w; - vertex = mat3(params.cam_rotation) * vertex; normal = normalize(mat3(params.cam_rotation) * normal); - vec3 reflection = normalize(reflect(normalize(vertex), normal)); #ifdef USE_SDFGI @@ -646,16 +638,39 @@ void main() { spec_accum /= blend_accum; } - if (params.use_sdfgi) { - reflection_light = blend_color(spec_accum, reflection_light); - ambient_light = blend_color(amb_accum, ambient_light); - } else { - reflection_light = spec_accum; - ambient_light = amb_accum; - } +#ifdef USE_SDFGI + reflection_light = blend_color(spec_accum, reflection_light); + ambient_light = blend_color(amb_accum, ambient_light); +#else + reflection_light = spec_accum; + ambient_light = amb_accum; +#endif } #endif } +} + +void main() { + ivec2 pos = ivec2(gl_GlobalInvocationID.xy); + +#ifdef MODE_HALF_RES + pos <<= 1; +#endif + if (any(greaterThanEqual(pos, params.screen_size))) { //too large, do nothing + return; + } + + vec4 ambient_light = vec4(0.0); + vec4 reflection_light = vec4(0.0); + + vec3 vertex = reconstruct_position(pos); + vertex.y = -vertex.y; + + process_gi(pos, vertex, ambient_light, reflection_light); + +#ifdef MODE_HALF_RES + pos >>= 1; +#endif imageStore(ambient_buffer, pos, ambient_light); imageStore(reflection_buffer, pos, reflection_light); diff --git a/servers/rendering/renderer_rd/shaders/giprobe.glsl b/servers/rendering/renderer_rd/shaders/giprobe.glsl index 4f4753d147..b931461b31 100644 --- a/servers/rendering/renderer_rd/shaders/giprobe.glsl +++ b/servers/rendering/renderer_rd/shaders/giprobe.glsl @@ -51,10 +51,10 @@ struct Light { float attenuation; vec3 color; - float spot_angle_radians; + float cos_spot_angle; vec3 position; - float spot_attenuation; + float inv_spot_attenuation; vec3 direction; bool has_shadow; @@ -233,13 +233,15 @@ bool compute_light_vector(uint light, vec3 pos, out float attenuation, out vec3 if (lights.data[light].type == LIGHT_TYPE_SPOT) { vec3 rel = normalize(pos - light_pos); - float angle = acos(dot(rel, lights.data[light].direction)); - if (angle > lights.data[light].spot_angle_radians) { + float cos_spot_angle = lights.data[light].cos_spot_angle; + float cos_angle = dot(rel, lights.data[light].direction); + if (cos_angle < cos_spot_angle) { return false; } - float d = clamp(angle / lights.data[light].spot_angle_radians, 0, 1); - attenuation *= pow(1.0 - d, lights.data[light].spot_attenuation); + float scos = max(cos_angle, cos_spot_angle); + float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - cos_spot_angle)); + attenuation *= 1.0 - pow(spot_rim, lights.data[light].inv_spot_attenuation); } } diff --git a/servers/rendering/renderer_rd/shaders/giprobe_write.glsl b/servers/rendering/renderer_rd/shaders/giprobe_write.glsl index 9c794f1bcc..56b3b7ccb4 100644 --- a/servers/rendering/renderer_rd/shaders/giprobe_write.glsl +++ b/servers/rendering/renderer_rd/shaders/giprobe_write.glsl @@ -43,10 +43,10 @@ struct Light { float attenuation; vec3 color; - float spot_angle_radians; + float cos_spot_angle; vec3 position; - float spot_attenuation; + float inv_spot_attenuation; vec3 direction; bool has_shadow; @@ -146,13 +146,15 @@ bool compute_light_vector(uint light, uint cell, vec3 pos, out float attenuation if (lights.data[light].type == LIGHT_TYPE_SPOT) { vec3 rel = normalize(pos - light_pos); - float angle = acos(dot(rel, lights.data[light].direction)); - if (angle > lights.data[light].spot_angle_radians) { + float cos_spot_angle = lights.data[light].cos_spot_angle; + float cos_angle = dot(rel, lights.data[light].direction); + if (cos_angle < cos_spot_angle) { return false; } - float d = clamp(angle / lights.data[light].spot_angle_radians, 0, 1); - attenuation *= pow(1.0 - d, lights.data[light].spot_attenuation); + float scos = max(cos_angle, cos_spot_angle); + float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - cos_spot_angle)); + attenuation *= 1.0 - pow(spot_rim, lights.data[light].inv_spot_attenuation); } } diff --git a/servers/rendering/renderer_rd/shaders/resolve.glsl b/servers/rendering/renderer_rd/shaders/resolve.glsl index 9429a66dc9..e83c4ca93b 100644 --- a/servers/rendering/renderer_rd/shaders/resolve.glsl +++ b/servers/rendering/renderer_rd/shaders/resolve.glsl @@ -58,6 +58,116 @@ void main() { #else +#if 1 + + vec4 group1; + vec4 group2; + vec4 group3; + vec4 group4; + int best_index = 0; + + //2X + group1.x = texelFetch(source_depth, pos, 0).r; + group1.y = texelFetch(source_depth, pos, 1).r; + + //4X + if (params.sample_count >= 4) { + group1.z = texelFetch(source_depth, pos, 2).r; + group1.w = texelFetch(source_depth, pos, 3).r; + } + //8X + if (params.sample_count >= 8) { + group2.x = texelFetch(source_depth, pos, 4).r; + group2.y = texelFetch(source_depth, pos, 5).r; + group2.z = texelFetch(source_depth, pos, 6).r; + group2.w = texelFetch(source_depth, pos, 7).r; + } + //16X + if (params.sample_count >= 16) { + group3.x = texelFetch(source_depth, pos, 8).r; + group3.y = texelFetch(source_depth, pos, 9).r; + group3.z = texelFetch(source_depth, pos, 10).r; + group3.w = texelFetch(source_depth, pos, 11).r; + + group4.x = texelFetch(source_depth, pos, 12).r; + group4.y = texelFetch(source_depth, pos, 13).r; + group4.z = texelFetch(source_depth, pos, 14).r; + group4.w = texelFetch(source_depth, pos, 15).r; + } + + if (params.sample_count == 2) { + best_index = (pos.x & 1) ^ ((pos.y >> 1) & 1); //not much can be done here + } else if (params.sample_count == 4) { + vec4 freq = vec4(equal(group1, vec4(group1.x))); + freq += vec4(equal(group1, vec4(group1.y))); + freq += vec4(equal(group1, vec4(group1.z))); + freq += vec4(equal(group1, vec4(group1.w))); + + float min_f = freq.x; + best_index = 0; + if (freq.y < min_f) { + best_index = 1; + min_f = freq.y; + } + if (freq.z < min_f) { + best_index = 2; + min_f = freq.z; + } + if (freq.w < min_f) { + best_index = 3; + } + } else if (params.sample_count == 8) { + vec4 freq0 = vec4(equal(group1, vec4(group1.x))); + vec4 freq1 = vec4(equal(group2, vec4(group1.x))); + freq0 += vec4(equal(group1, vec4(group1.y))); + freq1 += vec4(equal(group2, vec4(group1.y))); + freq0 += vec4(equal(group1, vec4(group1.z))); + freq1 += vec4(equal(group2, vec4(group1.z))); + freq0 += vec4(equal(group1, vec4(group1.w))); + freq1 += vec4(equal(group2, vec4(group1.w))); + freq0 += vec4(equal(group1, vec4(group2.x))); + freq1 += vec4(equal(group2, vec4(group2.x))); + freq0 += vec4(equal(group1, vec4(group2.y))); + freq1 += vec4(equal(group2, vec4(group2.y))); + freq0 += vec4(equal(group1, vec4(group2.z))); + freq1 += vec4(equal(group2, vec4(group2.z))); + freq0 += vec4(equal(group1, vec4(group2.w))); + freq1 += vec4(equal(group2, vec4(group2.w))); + + float min_f0 = freq0.x; + int best_index0 = 0; + if (freq0.y < min_f0) { + best_index0 = 1; + min_f0 = freq0.y; + } + if (freq0.z < min_f0) { + best_index0 = 2; + min_f0 = freq0.z; + } + if (freq0.w < min_f0) { + best_index0 = 3; + min_f0 = freq0.w; + } + + float min_f1 = freq1.x; + int best_index1 = 4; + if (freq1.y < min_f1) { + best_index1 = 5; + min_f1 = freq1.y; + } + if (freq1.z < min_f1) { + best_index1 = 6; + min_f1 = freq1.z; + } + if (freq1.w < min_f1) { + best_index1 = 7; + min_f1 = freq1.w; + } + + best_index = mix(best_index0, best_index1, min_f0 < min_f1); + } + +#else float depths[16]; int depth_indices[16]; int depth_amount[16]; @@ -91,7 +201,7 @@ void main() { depth_least = depth_amount[j]; } } - +#endif best_depth = texelFetch(source_depth, pos, best_index).r; best_normal_roughness = texelFetch(source_normal_roughness, pos, best_index); #ifdef GIPROBE_RESOLVE diff --git a/servers/rendering/renderer_rd/shaders/scene_forward.glsl b/servers/rendering/renderer_rd/shaders/scene_forward.glsl index 8d08fae348..1cea9bf8db 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward.glsl @@ -89,12 +89,6 @@ MATERIAL_UNIFORMS } material; #endif -/* clang-format off */ - -VERTEX_SHADER_GLOBALS - -/* clang-format on */ - invariant gl_Position; #ifdef MODE_DUAL_PARABOLOID @@ -103,28 +97,43 @@ layout(location = 8) out float dp_clip; #endif +layout(location = 9) out flat uint instance_index; + +/* clang-format off */ + +VERTEX_SHADER_GLOBALS + +/* clang-format on */ + void main() { vec4 instance_custom = vec4(0.0); #if defined(COLOR_USED) color_interp = color_attrib; #endif - mat4 world_matrix = draw_call.transform; + instance_index = draw_call.instance_index; + + bool is_multimesh = bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH); + if (!is_multimesh) { + instance_index += gl_InstanceIndex; + } + + mat4 world_matrix = instances.data[instance_index].transform; mat3 world_normal_matrix; - if (bool(draw_call.flags & INSTANCE_FLAGS_NON_UNIFORM_SCALE)) { + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_NON_UNIFORM_SCALE)) { world_normal_matrix = inverse(mat3(world_matrix)); } else { world_normal_matrix = mat3(world_matrix); } - if (bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH)) { + if (is_multimesh) { //multimesh, instances are for it - uint offset = (draw_call.flags >> INSTANCE_FLAGS_MULTIMESH_STRIDE_SHIFT) & INSTANCE_FLAGS_MULTIMESH_STRIDE_MASK; + uint offset = (instances.data[instance_index].flags >> INSTANCE_FLAGS_MULTIMESH_STRIDE_SHIFT) & INSTANCE_FLAGS_MULTIMESH_STRIDE_MASK; offset *= gl_InstanceIndex; mat4 matrix; - if (bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH_FORMAT_2D)) { + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH_FORMAT_2D)) { matrix = mat4(transforms.data[offset + 0], transforms.data[offset + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0)); offset += 2; } else { @@ -132,14 +141,14 @@ void main() { offset += 3; } - if (bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH_HAS_COLOR)) { + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH_HAS_COLOR)) { #ifdef COLOR_USED color_interp *= transforms.data[offset]; #endif offset += 1; } - if (bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH_HAS_CUSTOM_DATA)) { + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH_HAS_CUSTOM_DATA)) { instance_custom = transforms.data[offset]; } @@ -161,7 +170,7 @@ void main() { #endif #if 0 - if (bool(draw_call.flags & INSTANCE_FLAGS_SKELETON)) { + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_SKELETON)) { //multimesh, instances are for it uvec2 bones_01 = uvec2(bone_attrib.x & 0xFFFF, bone_attrib.x >> 16) * 3; @@ -194,7 +203,7 @@ void main() { uv2_interp = uv2_attrib; #endif -#ifdef USE_OVERRIDE_POSITION +#ifdef OVERRIDE_POSITION vec4 position; #endif @@ -289,7 +298,7 @@ VERTEX_SHADER_CODE #endif //MODE_RENDER_DEPTH -#ifdef USE_OVERRIDE_POSITION +#ifdef OVERRIDE_POSITION gl_Position = position; #else gl_Position = projection_matrix * vec4(vertex_interp, 1.0); @@ -304,7 +313,8 @@ VERTEX_SHADER_CODE #endif #ifdef MODE_RENDER_MATERIAL if (scene_data.material_uv2_mode) { - gl_Position.xy = (uv2_attrib.xy + draw_call.lightmap_uv_scale.xy) * 2.0 - 1.0; + vec2 uv_offset = unpackHalf2x16(draw_call.uv_offset); + gl_Position.xy = (uv2_attrib.xy + uv_offset) * 2.0 - 1.0; gl_Position.z = 0.00001; gl_Position.w = 1.0; } @@ -350,9 +360,11 @@ layout(location = 8) in float dp_clip; #endif +layout(location = 9) in flat uint instance_index; + //defines to keep compatibility with vertex -#define world_matrix draw_call.transform +#define world_matrix instances.data[instance_index].transform #define projection_matrix scene_data.projection_matrix #if defined(ENABLE_SSS) && defined(ENABLE_TRANSMITTANCE) @@ -1280,7 +1292,7 @@ float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { } else { #endif //hard shadow - vec4 shadow_uv = vec4(splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy, z_norm, 1.0); + vec4 shadow_uv = vec4(splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy, splane.z, 1.0); shadow = sample_pcf_shadow(shadow_atlas, spot_lights.data[idx].soft_shadow_scale * scene_data.shadow_atlas_pixel_size, shadow_uv); #ifdef USE_SOFT_SHADOWS @@ -1768,7 +1780,7 @@ vec4 fog_process(vec3 vertex) { } } - float fog_amount = 1.0 - exp(vertex.z * scene_data.fog_density); + float fog_amount = 1.0 - exp(min(0.0, vertex.z * scene_data.fog_density)); if (abs(scene_data.fog_height_density) > 0.001) { float y = (scene_data.camera_matrix * vec4(vertex, 1.0)).y; @@ -2081,7 +2093,7 @@ FRAGMENT_SHADER_CODE #endif uint decal_index = 32 * i + bit; - if (!bool(decals.data[decal_index].mask & draw_call.layer_mask)) { + if (!bool(decals.data[decal_index].mask & instances.data[instance_index].layer_mask)) { continue; //not masked } @@ -2208,8 +2220,8 @@ FRAGMENT_SHADER_CODE #ifdef USE_LIGHTMAP //lightmap - if (bool(draw_call.flags & INSTANCE_FLAGS_USE_LIGHTMAP_CAPTURE)) { //has lightmap capture - uint index = draw_call.gi_offset; + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_LIGHTMAP_CAPTURE)) { //has lightmap capture + uint index = instances.data[instance_index].gi_offset; vec3 wnormal = mat3(scene_data.camera_matrix) * normal; const float c1 = 0.429043; @@ -2228,12 +2240,12 @@ FRAGMENT_SHADER_CODE 2.0 * c2 * lightmap_captures.data[index].sh[1].rgb * wnormal.y + 2.0 * c2 * lightmap_captures.data[index].sh[2].rgb * wnormal.z); - } else if (bool(draw_call.flags & INSTANCE_FLAGS_USE_LIGHTMAP)) { // has actual lightmap - bool uses_sh = bool(draw_call.flags & INSTANCE_FLAGS_USE_SH_LIGHTMAP); - uint ofs = draw_call.gi_offset & 0xFFFF; + } else if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_LIGHTMAP)) { // has actual lightmap + bool uses_sh = bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_SH_LIGHTMAP); + uint ofs = instances.data[instance_index].gi_offset & 0xFFFF; vec3 uvw; - uvw.xy = uv2 * draw_call.lightmap_uv_scale.zw + draw_call.lightmap_uv_scale.xy; - uvw.z = float((draw_call.gi_offset >> 16) & 0xFFFF); + uvw.xy = uv2 * instances.data[instance_index].lightmap_uv_scale.zw + instances.data[instance_index].lightmap_uv_scale.xy; + uvw.z = float((instances.data[instance_index].gi_offset >> 16) & 0xFFFF); if (uses_sh) { uvw.z *= 4.0; //SH textures use 4 times more data @@ -2242,7 +2254,7 @@ FRAGMENT_SHADER_CODE vec3 lm_light_l1_0 = textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw + vec3(0.0, 0.0, 2.0), 0.0).rgb; vec3 lm_light_l1p1 = textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw + vec3(0.0, 0.0, 3.0), 0.0).rgb; - uint idx = draw_call.gi_offset >> 20; + uint idx = instances.data[instance_index].gi_offset >> 20; vec3 n = normalize(lightmaps.data[idx].normal_xform * normal); ambient_light += lm_light_l0 * 0.282095f; @@ -2262,7 +2274,7 @@ FRAGMENT_SHADER_CODE } #elif defined(USE_FORWARD_GI) - if (bool(draw_call.flags & INSTANCE_FLAGS_USE_SDFGI)) { //has lightmap capture + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_SDFGI)) { //has lightmap capture //make vertex orientation the world one, but still align to camera vec3 cam_pos = mat3(scene_data.camera_matrix) * vertex; @@ -2334,9 +2346,9 @@ FRAGMENT_SHADER_CODE } } - if (bool(draw_call.flags & INSTANCE_FLAGS_USE_GIPROBE)) { // process giprobes + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_GIPROBE)) { // process giprobes - uint index1 = draw_call.gi_offset & 0xFFFF; + uint index1 = instances.data[instance_index].gi_offset & 0xFFFF; vec3 ref_vec = normalize(reflect(normalize(vertex), normal)); //find arbitrary tangent and bitangent, then build a matrix vec3 v0 = abs(normal.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 1.0, 0.0); @@ -2348,7 +2360,7 @@ FRAGMENT_SHADER_CODE vec4 spec_accum = vec4(0.0); gi_probe_compute(index1, vertex, normal, ref_vec, normal_mat, roughness * roughness, ambient_light, specular_light, spec_accum, amb_accum); - uint index2 = draw_call.gi_offset >> 16; + uint index2 = instances.data[instance_index].gi_offset >> 16; if (index2 != 0xFFFF) { gi_probe_compute(index2, vertex, normal, ref_vec, normal_mat, roughness * roughness, ambient_light, specular_light, spec_accum, amb_accum); @@ -2367,19 +2379,19 @@ FRAGMENT_SHADER_CODE } #elif !defined(LOW_END_MODE) - if (bool(draw_call.flags & INSTANCE_FLAGS_USE_GI_BUFFERS)) { //use GI buffers + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_GI_BUFFERS)) { //use GI buffers - ivec2 coord; + vec2 coord; if (scene_data.gi_upscale_for_msaa) { - ivec2 base_coord = ivec2(gl_FragCoord.xy); - ivec2 closest_coord = base_coord; - float closest_ang = dot(normal, texelFetch(sampler2D(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), base_coord, 0).xyz * 2.0 - 1.0); + vec2 base_coord = screen_uv; + vec2 closest_coord = base_coord; + float closest_ang = dot(normal, textureLod(sampler2D(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), base_coord, 0.0).xyz * 2.0 - 1.0); for (int i = 0; i < 4; i++) { - const ivec2 neighbours[4] = ivec2[](ivec2(-1, 0), ivec2(1, 0), ivec2(0, -1), ivec2(0, 1)); - ivec2 neighbour_coord = base_coord + neighbours[i]; - float neighbour_ang = dot(normal, texelFetch(sampler2D(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), neighbour_coord, 0).xyz * 2.0 - 1.0); + const vec2 neighbours[4] = vec2[](vec2(-1, 0), vec2(1, 0), vec2(0, -1), vec2(0, 1)); + vec2 neighbour_coord = base_coord + neighbours[i] * scene_data.screen_pixel_size; + float neighbour_ang = dot(normal, textureLod(sampler2D(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), neighbour_coord, 0.0).xyz * 2.0 - 1.0); if (neighbour_ang > closest_ang) { closest_ang = neighbour_ang; closest_coord = neighbour_coord; @@ -2389,11 +2401,11 @@ FRAGMENT_SHADER_CODE coord = closest_coord; } else { - coord = ivec2(gl_FragCoord.xy); + coord = screen_uv; } - vec4 buffer_ambient = texelFetch(sampler2D(ambient_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), coord, 0); - vec4 buffer_reflection = texelFetch(sampler2D(reflection_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), coord, 0); + vec4 buffer_ambient = textureLod(sampler2D(ambient_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), coord, 0.0); + vec4 buffer_reflection = textureLod(sampler2D(reflection_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), coord, 0.0); ambient_light = mix(ambient_light, buffer_ambient.rgb, buffer_ambient.a); specular_light = mix(specular_light, buffer_reflection.rgb, buffer_reflection.a); @@ -2446,7 +2458,7 @@ FRAGMENT_SHADER_CODE #endif uint reflection_index = 32 * i + bit; - if (!bool(reflections.data[reflection_index].mask & draw_call.layer_mask)) { + if (!bool(reflections.data[reflection_index].mask & instances.data[instance_index].layer_mask)) { continue; //not masked } @@ -2517,7 +2529,7 @@ FRAGMENT_SHADER_CODE break; } - if (!bool(directional_lights.data[i].mask & draw_call.layer_mask)) { + if (!bool(directional_lights.data[i].mask & instances.data[instance_index].layer_mask)) { continue; //not masked } @@ -2836,7 +2848,7 @@ FRAGMENT_SHADER_CODE break; } - if (!bool(directional_lights.data[i].mask & draw_call.layer_mask)) { + if (!bool(directional_lights.data[i].mask & instances.data[instance_index].layer_mask)) { continue; //not masked } @@ -2966,7 +2978,7 @@ FRAGMENT_SHADER_CODE #endif uint light_index = 32 * i + bit; - if (!bool(omni_lights.data[light_index].mask & draw_call.layer_mask)) { + if (!bool(omni_lights.data[light_index].mask & instances.data[instance_index].layer_mask)) { continue; //not masked } @@ -3039,7 +3051,7 @@ FRAGMENT_SHADER_CODE uint light_index = 32 * i + bit; - if (!bool(spot_lights.data[light_index].mask & draw_call.layer_mask)) { + if (!bool(spot_lights.data[light_index].mask & instances.data[instance_index].layer_mask)) { continue; //not masked } @@ -3212,9 +3224,9 @@ FRAGMENT_SHADER_CODE normal_roughness_output_buffer = vec4(normal * 0.5 + 0.5, roughness); #ifdef MODE_RENDER_GIPROBE - if (bool(draw_call.flags & INSTANCE_FLAGS_USE_GIPROBE)) { // process giprobes - uint index1 = draw_call.gi_offset & 0xFFFF; - uint index2 = draw_call.gi_offset >> 16; + 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 { @@ -3273,6 +3285,7 @@ FRAGMENT_SHADER_CODE // Draw "fixed" fog before volumetric fog to ensure volumetric fog can appear in front of the sky. frag_color.rgb = mix(frag_color.rgb, fog.rgb, fog.a); + ; #endif //MODE_MULTIPLE_RENDER_TARGETS diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_inc.glsl index a37e32e1fc..d78890fa9e 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_inc.glsl @@ -21,12 +21,10 @@ #endif layout(push_constant, binding = 0, std430) uniform DrawCall { - mat4 transform; - uint flags; - uint instance_uniforms_ofs; //base offset in global buffer for instance variables - uint gi_offset; //GI information when using lightmapping (VCT or lightmap index) - uint layer_mask; - vec4 lightmap_uv_scale; + uint instance_index; + uint uv_offset; + uint pad0; + uint pad1; } draw_call; @@ -45,96 +43,13 @@ draw_call; #define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_REPEAT 10 #define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_REPEAT 11 -layout(set = 0, binding = 1) uniform sampler material_samplers[12]; - -layout(set = 0, binding = 2) uniform sampler shadow_sampler; - #define SDFGI_MAX_CASCADES 8 -layout(set = 0, binding = 3, std140) uniform SceneData { - mat4 projection_matrix; - mat4 inv_projection_matrix; - - mat4 camera_matrix; - mat4 inv_camera_matrix; - - vec2 viewport_size; - vec2 screen_pixel_size; - - uint cluster_shift; - uint cluster_width; - uint cluster_type_size; - uint max_cluster_element_count_div_32; - - //use vec4s because std140 doesnt play nice with vec2s, z and w are wasted - vec4 directional_penumbra_shadow_kernel[32]; - vec4 directional_soft_shadow_kernel[32]; - vec4 penumbra_shadow_kernel[32]; - vec4 soft_shadow_kernel[32]; - - uint directional_penumbra_shadow_samples; - uint directional_soft_shadow_samples; - uint penumbra_shadow_samples; - uint soft_shadow_samples; - - vec4 ambient_light_color_energy; +/* Set 1: Base Pass (never changes) */ - float ambient_color_sky_mix; - bool use_ambient_light; - bool use_ambient_cubemap; - bool use_reflection_cubemap; - - mat3 radiance_inverse_xform; - - vec2 shadow_atlas_pixel_size; - vec2 directional_shadow_pixel_size; - - uint directional_light_count; - float dual_paraboloid_side; - float z_far; - float z_near; - - bool ssao_enabled; - float ssao_light_affect; - float ssao_ao_affect; - bool roughness_limiter_enabled; - - float roughness_limiter_amount; - float roughness_limiter_limit; - uvec2 roughness_limiter_pad; - - vec4 ao_color; - - mat4 sdf_to_bounds; - - ivec3 sdf_offset; - bool material_uv2_mode; - - ivec3 sdf_size; - bool gi_upscale_for_msaa; - - bool volumetric_fog_enabled; - float volumetric_fog_inv_length; - float volumetric_fog_detail_spread; - uint volumetric_fog_pad; - - bool fog_enabled; - float fog_density; - float fog_height; - float fog_height_density; - - vec3 fog_light_color; - float fog_sun_scatter; - - float fog_aerial_perspective; - - float time; - float reflection_multiplier; // one normally, zero when rendering reflections - - bool pancake_shadows; -} +layout(set = 0, binding = 1) uniform sampler material_samplers[12]; -scene_data; +layout(set = 0, binding = 2) uniform sampler shadow_sampler; #define INSTANCE_FLAGS_USE_GI_BUFFERS (1 << 6) #define INSTANCE_FLAGS_USE_SDFGI (1 << 7) @@ -153,22 +68,22 @@ scene_data; #define INSTANCE_FLAGS_SKELETON (1 << 19) #define INSTANCE_FLAGS_NON_UNIFORM_SCALE (1 << 20) -layout(set = 0, binding = 5, std430) restrict readonly buffer OmniLights { +layout(set = 0, binding = 3, std430) restrict readonly buffer OmniLights { LightData data[]; } omni_lights; -layout(set = 0, binding = 6, std430) restrict readonly buffer SpotLights { +layout(set = 0, binding = 4, std430) restrict readonly buffer SpotLights { LightData data[]; } spot_lights; -layout(set = 0, binding = 7) buffer restrict readonly ReflectionProbeData { +layout(set = 0, binding = 5) buffer restrict readonly ReflectionProbeData { ReflectionData data[]; } reflections; -layout(set = 0, binding = 8, std140) uniform DirectionalLights { +layout(set = 0, binding = 6, std140) uniform DirectionalLights { DirectionalLightData data[MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS]; } directional_lights; @@ -180,7 +95,7 @@ struct Lightmap { mat3 normal_xform; }; -layout(set = 0, binding = 10, std140) restrict readonly buffer Lightmaps { +layout(set = 0, binding = 7, std140) restrict readonly buffer Lightmaps { Lightmap data[]; } lightmaps; @@ -189,22 +104,20 @@ struct LightmapCapture { vec4 sh[9]; }; -layout(set = 0, binding = 11, std140) restrict readonly buffer LightmapCaptures { +layout(set = 0, binding = 8, std140) restrict readonly buffer LightmapCaptures { LightmapCapture data[]; } lightmap_captures; -layout(set = 0, binding = 12) uniform texture2D decal_atlas; -layout(set = 0, binding = 13) uniform texture2D decal_atlas_srgb; +layout(set = 0, binding = 9) uniform texture2D decal_atlas; +layout(set = 0, binding = 10) uniform texture2D decal_atlas_srgb; -layout(set = 0, binding = 14, std430) restrict readonly buffer Decals { +layout(set = 0, binding = 11, std430) restrict readonly buffer Decals { DecalData data[]; } decals; -layout(set = 0, binding = 15) uniform texture2D directional_shadow_atlas; - -layout(set = 0, binding = 16, std430) restrict readonly buffer GlobalVariableData { +layout(set = 0, binding = 12, std430) restrict readonly buffer GlobalVariableData { vec4 data[]; } global_variables; @@ -218,7 +131,7 @@ struct SDFGIProbeCascadeData { float to_cell; // 1/bounds * grid_size }; -layout(set = 0, binding = 17, std140) uniform SDFGI { +layout(set = 0, binding = 13, std140) uniform SDFGI { vec3 grid_size; uint max_cascades; @@ -248,45 +161,140 @@ sdfgi; #endif //LOW_END_MODE -// decal atlas +/* Set 2: Render Pass (changes per render pass) */ -/* Set 1, Radiance */ +layout(set = 1, binding = 0, std140) uniform SceneData { + mat4 projection_matrix; + mat4 inv_projection_matrix; + + mat4 camera_matrix; + mat4 inv_camera_matrix; + + vec2 viewport_size; + vec2 screen_pixel_size; + + uint cluster_shift; + uint cluster_width; + uint cluster_type_size; + uint max_cluster_element_count_div_32; + + //use vec4s because std140 doesnt play nice with vec2s, z and w are wasted + vec4 directional_penumbra_shadow_kernel[32]; + vec4 directional_soft_shadow_kernel[32]; + vec4 penumbra_shadow_kernel[32]; + vec4 soft_shadow_kernel[32]; + + uint directional_penumbra_shadow_samples; + uint directional_soft_shadow_samples; + uint penumbra_shadow_samples; + uint soft_shadow_samples; + + vec4 ambient_light_color_energy; + + float ambient_color_sky_mix; + bool use_ambient_light; + bool use_ambient_cubemap; + bool use_reflection_cubemap; + + mat3 radiance_inverse_xform; + + vec2 shadow_atlas_pixel_size; + vec2 directional_shadow_pixel_size; + + uint directional_light_count; + float dual_paraboloid_side; + float z_far; + float z_near; + + bool ssao_enabled; + float ssao_light_affect; + float ssao_ao_affect; + bool roughness_limiter_enabled; + + float roughness_limiter_amount; + float roughness_limiter_limit; + uvec2 roughness_limiter_pad; + + vec4 ao_color; + + mat4 sdf_to_bounds; + + ivec3 sdf_offset; + bool material_uv2_mode; + + ivec3 sdf_size; + bool gi_upscale_for_msaa; + + bool volumetric_fog_enabled; + float volumetric_fog_inv_length; + float volumetric_fog_detail_spread; + uint volumetric_fog_pad; + + bool fog_enabled; + float fog_density; + float fog_height; + float fog_height_density; + + vec3 fog_light_color; + float fog_sun_scatter; + + float fog_aerial_perspective; + + float time; + float reflection_multiplier; // one normally, zero when rendering reflections + + bool pancake_shadows; +} + +scene_data; + +struct InstanceData { + mat4 transform; + uint flags; + uint instance_uniforms_ofs; //base offset in global buffer for instance variables + uint gi_offset; //GI information when using lightmapping (VCT or lightmap index) + uint layer_mask; + vec4 lightmap_uv_scale; +}; + +layout(set = 1, binding = 1, std430) buffer restrict readonly InstanceDataBuffer { + InstanceData data[]; +} +instances; #ifdef USE_RADIANCE_CUBEMAP_ARRAY -layout(set = 1, binding = 0) uniform textureCubeArray radiance_cubemap; +layout(set = 1, binding = 2) uniform textureCubeArray radiance_cubemap; #else -layout(set = 1, binding = 0) uniform textureCube radiance_cubemap; +layout(set = 1, binding = 2) uniform textureCube radiance_cubemap; #endif -/* Set 2, Reflection and Shadow Atlases (view dependent) */ +layout(set = 1, binding = 3) uniform textureCubeArray reflection_atlas; -layout(set = 1, binding = 1) uniform textureCubeArray reflection_atlas; +layout(set = 1, binding = 4) uniform texture2D shadow_atlas; -layout(set = 1, binding = 2) uniform texture2D shadow_atlas; +layout(set = 1, binding = 5) uniform texture2D directional_shadow_atlas; -layout(set = 1, binding = 3) uniform texture2DArray lightmap_textures[MAX_LIGHTMAP_TEXTURES]; +layout(set = 1, binding = 6) uniform texture2DArray lightmap_textures[MAX_LIGHTMAP_TEXTURES]; -#ifndef LOW_END_MODE -layout(set = 1, binding = 4) uniform texture3D gi_probe_textures[MAX_GI_PROBES]; +#ifndef LOW_END_MOD +layout(set = 1, binding = 7) uniform texture3D gi_probe_textures[MAX_GI_PROBES]; #endif -layout(set = 1, binding = 5, std430) buffer restrict readonly ClusterBuffer { +layout(set = 1, binding = 8, std430) buffer restrict readonly ClusterBuffer { uint data[]; } cluster_buffer; -/* Set 3, Render Buffers */ - #ifdef MODE_RENDER_SDF -layout(r16ui, set = 1, binding = 6) uniform restrict writeonly uimage3D albedo_volume_grid; -layout(r32ui, set = 1, binding = 7) uniform restrict writeonly uimage3D emission_grid; -layout(r32ui, set = 1, binding = 8) uniform restrict writeonly uimage3D emission_aniso_grid; -layout(r32ui, set = 1, binding = 9) uniform restrict uimage3D geom_facing_grid; +layout(r16ui, set = 1, binding = 9) uniform restrict writeonly uimage3D albedo_volume_grid; +layout(r32ui, set = 1, binding = 10) uniform restrict writeonly uimage3D emission_grid; +layout(r32ui, set = 1, binding = 11) uniform restrict writeonly uimage3D emission_aniso_grid; +layout(r32ui, set = 1, binding = 12) uniform restrict uimage3D geom_facing_grid; //still need to be present for shaders that use it, so remap them to something #define depth_buffer shadow_atlas @@ -295,17 +303,17 @@ layout(r32ui, set = 1, binding = 9) uniform restrict uimage3D geom_facing_grid; #else -layout(set = 1, binding = 6) uniform texture2D depth_buffer; -layout(set = 1, binding = 7) uniform texture2D color_buffer; +layout(set = 1, binding = 9) uniform texture2D depth_buffer; +layout(set = 1, binding = 10) uniform texture2D color_buffer; #ifndef LOW_END_MODE -layout(set = 1, binding = 8) uniform texture2D normal_roughness_buffer; -layout(set = 1, binding = 9) uniform texture2D ao_buffer; -layout(set = 1, binding = 10) uniform texture2D ambient_buffer; -layout(set = 1, binding = 11) uniform texture2D reflection_buffer; -layout(set = 1, binding = 12) uniform texture2DArray sdfgi_lightprobe_texture; -layout(set = 1, binding = 13) uniform texture3D sdfgi_occlusion_cascades; +layout(set = 1, binding = 11) uniform texture2D normal_roughness_buffer; +layout(set = 1, binding = 12) uniform texture2D ao_buffer; +layout(set = 1, binding = 13) uniform texture2D ambient_buffer; +layout(set = 1, binding = 14) uniform texture2D reflection_buffer; +layout(set = 1, binding = 15) uniform texture2DArray sdfgi_lightprobe_texture; +layout(set = 1, binding = 16) uniform texture3D sdfgi_occlusion_cascades; struct GIProbeData { mat4 xform; @@ -323,22 +331,22 @@ struct GIProbeData { uint mipmaps; }; -layout(set = 1, binding = 14, std140) uniform GIProbes { +layout(set = 1, binding = 17, std140) uniform GIProbes { GIProbeData data[MAX_GI_PROBES]; } gi_probes; -layout(set = 1, binding = 15) uniform texture3D volumetric_fog_texture; +layout(set = 1, binding = 18) uniform texture3D volumetric_fog_texture; #endif // LOW_END_MODE #endif -/* Set 4 Skeleton & Instancing (Multimesh) */ +/* Set 2 Skeleton & Instancing (can change per item) */ layout(set = 2, binding = 0, std430) restrict readonly buffer Transforms { vec4 data[]; } transforms; -/* Set 5 User Material */ +/* Set 3 User Material */ diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl index 813ea29fa1..e4c3f3a84b 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl @@ -97,6 +97,8 @@ void main() { float blend = 0.0; #if 1 + // No interpolation + vec3 inv_dir = 1.0 / ray_dir; float rough = 0.5; @@ -161,114 +163,11 @@ void main() { hit_light *= (dot(max(vec3(0.0), (hit_normal * hit_aniso0)), vec3(1.0)) + dot(max(vec3(0.0), (-hit_normal * hit_aniso1)), vec3(1.0))); - if (blend > 0.0) { - light = mix(light, hit_light, blend); - blend = 0.0; - } else { - light = hit_light; - - //process blend - float blend_from = (float(params.probe_axis_size - 1) / 2.0) - 2.5; - float blend_to = blend_from + 2.0; - - vec3 cam_pos = params.cam_transform[3].xyz - cascades.data[i].offset; - cam_pos *= cascades.data[i].to_cell; - - pos += ray_dir * min(advance, max_advance); - vec3 inner_pos = pos - cam_pos; - - inner_pos = inner_pos * float(params.probe_axis_size - 1) / params.grid_size.x; - - float len = length(inner_pos); - - inner_pos = abs(normalize(inner_pos)); - len *= max(inner_pos.x, max(inner_pos.y, inner_pos.z)); - - if (len >= blend_from) { - blend = smoothstep(blend_from, blend_to, len); - - pos /= cascades.data[i].to_cell; - pos += cascades.data[i].offset; - ray_pos = pos; - hit = false; //continue trace for blend - - continue; - } - } + light = hit_light; break; } - light = mix(light, vec3(0.0), blend); - -#else - - vec3 inv_dir = 1.0 / ray_dir; - - bool hit = false; - vec4 light_accum = vec4(0.0); - - float blend_size = (params.grid_size.x / float(params.probe_axis_size - 1)) * 0.5; - - float radius_sizes[MAX_CASCADES]; - for (uint i = 0; i < params.max_cascades; i++) { - radius_sizes[i] = (1.0 / cascades.data[i].to_cell) * (params.grid_size.x * 0.5 - blend_size); - } - - float max_distance = radius_sizes[params.max_cascades - 1]; - float advance = 0; - while (advance < max_distance) { - for (uint i = 0; i < params.max_cascades; i++) { - if (advance < radius_sizes[i]) { - vec3 pos = (ray_pos + ray_dir * advance) - cascades.data[i].offset; - pos *= cascades.data[i].to_cell * pos_to_uvw; - - float distance = texture(sampler3D(sdf_cascades[i], linear_sampler), pos).r * 255.0 - 1.0; - - vec4 hit_light = vec4(0.0); - if (distance < 1.0) { - hit_light.a = max(0.0, 1.0 - distance); - hit_light.rgb = texture(sampler3D(light_cascades[i], linear_sampler), pos).rgb; - hit_light.rgb *= hit_light.a; - } - - distance /= cascades.data[i].to_cell; - - if (i < (params.max_cascades - 1)) { - pos = (ray_pos + ray_dir * advance) - cascades.data[i + 1].offset; - pos *= cascades.data[i + 1].to_cell * pos_to_uvw; - - float distance2 = texture(sampler3D(sdf_cascades[i + 1], linear_sampler), pos).r * 255.0 - 1.0; - - vec4 hit_light2 = vec4(0.0); - if (distance2 < 1.0) { - hit_light2.a = max(0.0, 1.0 - distance2); - hit_light2.rgb = texture(sampler3D(light_cascades[i + 1], linear_sampler), pos).rgb; - hit_light2.rgb *= hit_light2.a; - } - - float prev_radius = i == 0 ? 0.0 : radius_sizes[i - 1]; - float blend = (advance - prev_radius) / (radius_sizes[i] - prev_radius); - - distance2 /= cascades.data[i + 1].to_cell; - - hit_light = mix(hit_light, hit_light2, blend); - distance = mix(distance, distance2, blend); - } - - light_accum += hit_light; - advance += distance; - break; - } - } - - if (light_accum.a > 0.98) { - break; - } - } - - light = light_accum.rgb / light_accum.a; - #endif imageStore(screen_buffer, screen_pos, vec4(linear_to_srgb(light), 1.0)); diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl index ed0a8a4b86..dc7238abed 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl @@ -67,8 +67,8 @@ struct Light { float attenuation; uint type; - float spot_angle; - float spot_attenuation; + float cos_spot_angle; + float inv_spot_attenuation; float radius; vec4 shadow_color; @@ -80,6 +80,7 @@ layout(set = 0, binding = 9, std140) buffer restrict readonly Lights { lights; layout(set = 0, binding = 10) uniform texture2DArray lightprobe_texture; +layout(set = 0, binding = 11) uniform texture3D occlusion_texture; layout(push_constant, binding = 0, std430) uniform Params { vec3 grid_size; @@ -91,9 +92,9 @@ layout(push_constant, binding = 0, std430) uniform Params { uint process_increment; int probe_axis_size; - bool multibounce; + float bounce_feedback; float y_mult; - uint pad; + bool use_occlusion; } params; @@ -125,7 +126,10 @@ void main() { uint voxel_index = uint(gl_GlobalInvocationID.x); //used for skipping voxels every N frames - voxel_index = params.process_offset + voxel_index * params.process_increment; + if (params.process_increment > 1) { + voxel_index *= params.process_increment; + voxel_index += params.process_offset; + } if (voxel_index >= dispatch_data.total_count) { return; @@ -156,7 +160,8 @@ void main() { // Add indirect light first, in order to save computation resources #ifdef MODE_PROCESS_DYNAMIC - if (params.multibounce) { + if (params.bounce_feedback > 0.001) { + vec3 feedback = (params.bounce_feedback < 1.0) ? (albedo * params.bounce_feedback) : mix(albedo, vec3(1.0), params.bounce_feedback - 1.0); vec3 pos = (vec3(positioni) + vec3(0.5)) * float(params.probe_axis_size - 1) / params.grid_size; ivec3 probe_base_pos = ivec3(pos); @@ -169,7 +174,7 @@ void main() { 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; + vec3 probe_uv_offset = vec3(ivec3(OCT_SIZE + 2, OCT_SIZE + 2, (OCT_SIZE + 2) * params.probe_axis_size)) * tex_pixel_size.xyx; for (uint j = 0; j < 8; j++) { ivec3 offset = (ivec3(j) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1); @@ -189,18 +194,35 @@ void main() { 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; - vec3 indirect_light = textureLod(sampler2DArray(lightprobe_texture, linear_sampler), pos_uvw, 0.0).rgb; - - light_accum[k] += indirect_light * weight; - weight_accum[k] += weight; + float weight = trilinear.x * trilinear.y * trilinear.z * max(0, dot(n, probe_dir)); + + if (weight > 0.0 && params.use_occlusion) { + ivec3 occ_indexv = abs((cascades.data[params.cascade].probe_world_offset + probe_posi) & ivec3(1, 1, 1)) * ivec3(1, 2, 4); + vec4 occ_mask = mix(vec4(0.0), vec4(1.0), equal(ivec4(occ_indexv.x | occ_indexv.y), ivec4(0, 1, 2, 3))); + + vec3 occ_pos = (vec3(positioni) + aniso_dir[k] + vec3(0.5)) / params.grid_size; + occ_pos.z += float(params.cascade); + if (occ_indexv.z != 0) { //z bit is on, means index is >=4, so make it switch to the other half of textures + occ_pos.x += 1.0; + } + occ_pos *= vec3(0.5, 1.0, 1.0 / float(params.max_cascades)); //renormalize + float occlusion = dot(textureLod(sampler3D(occlusion_texture, linear_sampler), occ_pos, 0.0), occ_mask); + + weight *= occlusion; + } + + if (weight > 0.0) { + vec3 tex_posf = base_tex_posf + vec3(octahedron_encode(n) * float(OCT_SIZE), 0.0); + tex_posf.xy *= tex_pixel_size; + + vec3 pos_uvw = tex_posf; + pos_uvw.xy += vec2(offset.xy) * probe_uv_offset.xy; + pos_uvw.x += float(offset.z) * probe_uv_offset.z; + vec3 indirect_light = textureLod(sampler2DArray(lightprobe_texture, linear_sampler), pos_uvw, 0.0).rgb; + + light_accum[k] += indirect_light * weight; + weight_accum[k] += weight; + } } } } @@ -208,7 +230,7 @@ void main() { for (uint k = 0; k < 6; k++) { if (weight_accum[k] > 0.0) { light_accum[k] /= weight_accum[k]; - light_accum[k] *= albedo; + light_accum[k] *= feedback; } } } @@ -263,13 +285,16 @@ void main() { rel_vec.y /= params.y_mult; attenuation = get_omni_attenuation(light_distance, 1.0 / lights.data[i].radius, lights.data[i].attenuation); - float angle = acos(dot(normalize(rel_vec), -lights.data[i].direction)); - if (angle > lights.data[i].spot_angle) { - attenuation = 0.0; - } else { - float d = clamp(angle / lights.data[i].spot_angle, 0, 1); - attenuation *= pow(1.0 - d, lights.data[i].spot_attenuation); + float cos_spot_angle = lights.data[i].cos_spot_angle; + float cos_angle = dot(-direction, lights.data[i].direction); + + if (cos_angle < cos_spot_angle) { + continue; } + + float scos = max(cos_angle, cos_spot_angle); + float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - cos_spot_angle)); + attenuation *= 1.0 - pow(spot_rim, lights.data[i].inv_spot_attenuation); } break; } diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl index 67630a3aa1..007e4c113a 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl @@ -39,8 +39,11 @@ layout(rgba32i, set = 0, binding = 13) uniform restrict iimage2D lightprobe_aver layout(rgba16f, set = 0, binding = 14) uniform restrict writeonly image2DArray lightprobe_ambient_texture; +#ifdef USE_CUBEMAP_ARRAY +layout(set = 1, binding = 0) uniform textureCubeArray sky_irradiance; +#else layout(set = 1, binding = 0) uniform textureCube sky_irradiance; - +#endif layout(set = 1, binding = 1) uniform sampler linear_sampler_mipmaps; #define HISTORY_BITS 10 @@ -189,14 +192,12 @@ void main() { vec3 inv_dir = 1.0 / ray_dir; bool hit = false; - vec3 hit_normal; - vec3 hit_light; - vec3 hit_aniso0; - vec3 hit_aniso1; + uint hit_cascade; float bias = params.ray_bias; vec3 abs_ray_dir = abs(ray_dir); ray_pos += ray_dir * 1.0 / max(abs_ray_dir.x, max(abs_ray_dir.y, abs_ray_dir.z)) * bias / cascades.data[params.cascade].to_cell; + vec3 uvw; for (uint j = params.cascade; j < params.max_cascades; j++) { //convert to local bounds @@ -215,14 +216,12 @@ void main() { float advance = 0.0; - vec3 uvw; - while (advance < max_advance) { //read how much to advance from SDF uvw = (pos + ray_dir * advance) * pos_to_uvw; float distance = texture(sampler3D(sdf_cascades[j], linear_sampler), uvw).r * 255.0 - 1.0; - if (distance < 0.001) { + if (distance < 0.05) { //consider hit hit = true; break; @@ -232,17 +231,7 @@ void main() { } if (hit) { - const float EPSILON = 0.001; - hit_normal = normalize(vec3( - texture(sampler3D(sdf_cascades[j], linear_sampler), uvw + vec3(EPSILON, 0.0, 0.0)).r - texture(sampler3D(sdf_cascades[j], linear_sampler), uvw - vec3(EPSILON, 0.0, 0.0)).r, - texture(sampler3D(sdf_cascades[j], linear_sampler), uvw + vec3(0.0, EPSILON, 0.0)).r - texture(sampler3D(sdf_cascades[j], linear_sampler), uvw - vec3(0.0, EPSILON, 0.0)).r, - texture(sampler3D(sdf_cascades[j], linear_sampler), uvw + vec3(0.0, 0.0, EPSILON)).r - texture(sampler3D(sdf_cascades[j], linear_sampler), uvw - vec3(0.0, 0.0, EPSILON)).r)); - - hit_light = texture(sampler3D(light_cascades[j], linear_sampler), uvw).rgb; - vec4 aniso0 = texture(sampler3D(aniso0_cascades[j], linear_sampler), uvw); - hit_aniso0 = aniso0.rgb; - hit_aniso1 = vec3(aniso0.a, texture(sampler3D(aniso1_cascades[j], linear_sampler), uvw).rg); - + hit_cascade = j; break; } @@ -255,11 +244,32 @@ void main() { vec4 light; if (hit) { - //one liner magic - light.rgb = hit_light * (dot(max(vec3(0.0), (hit_normal * hit_aniso0)), vec3(1.0)) + dot(max(vec3(0.0), (-hit_normal * hit_aniso1)), vec3(1.0))); - light.a = 1.0; + //avoid reading different texture from different threads + for (uint j = params.cascade; j < params.max_cascades; j++) { + if (j == hit_cascade) { + const float EPSILON = 0.001; + vec3 hit_normal = normalize(vec3( + texture(sampler3D(sdf_cascades[hit_cascade], linear_sampler), uvw + vec3(EPSILON, 0.0, 0.0)).r - texture(sampler3D(sdf_cascades[hit_cascade], linear_sampler), uvw - vec3(EPSILON, 0.0, 0.0)).r, + texture(sampler3D(sdf_cascades[hit_cascade], linear_sampler), uvw + vec3(0.0, EPSILON, 0.0)).r - texture(sampler3D(sdf_cascades[hit_cascade], linear_sampler), uvw - vec3(0.0, EPSILON, 0.0)).r, + texture(sampler3D(sdf_cascades[hit_cascade], linear_sampler), uvw + vec3(0.0, 0.0, EPSILON)).r - texture(sampler3D(sdf_cascades[hit_cascade], linear_sampler), uvw - vec3(0.0, 0.0, EPSILON)).r)); + + vec3 hit_light = texture(sampler3D(light_cascades[hit_cascade], linear_sampler), uvw).rgb; + vec4 aniso0 = texture(sampler3D(aniso0_cascades[hit_cascade], linear_sampler), uvw); + vec3 hit_aniso0 = aniso0.rgb; + vec3 hit_aniso1 = vec3(aniso0.a, texture(sampler3D(aniso1_cascades[hit_cascade], linear_sampler), uvw).rg); + + //one liner magic + light.rgb = hit_light * (dot(max(vec3(0.0), (hit_normal * hit_aniso0)), vec3(1.0)) + dot(max(vec3(0.0), (-hit_normal * hit_aniso1)), vec3(1.0))); + light.a = 1.0; + } + } + } else if (params.sky_mode == SKY_MODE_SKY) { +#ifdef USE_CUBEMAP_ARRAY + light.rgb = textureLod(samplerCubeArray(sky_irradiance, linear_sampler_mipmaps), vec4(ray_dir, 0.0), 2.0).rgb; //use second mipmap because we dont usually throw a lot of rays, so this compensates +#else light.rgb = textureLod(samplerCube(sky_irradiance, linear_sampler_mipmaps), ray_dir, 2.0).rgb; //use second mipmap because we dont usually throw a lot of rays, so this compensates +#endif light.rgb *= params.sky_energy; light.a = 0.0; @@ -490,13 +500,15 @@ void main() { //can't scroll, must look for position in parent cascade //to global coords - float probe_cell_size = float(params.grid_size.x / float(params.probe_axis_size - 1)) / cascades.data[params.cascade].to_cell; + float cell_to_probe = float(params.grid_size.x / float(params.probe_axis_size - 1)); + + float probe_cell_size = cell_to_probe / cascades.data[params.cascade].to_cell; vec3 probe_pos = cascades.data[params.cascade].offset + vec3(probe_cell) * probe_cell_size; //to parent local coords + float probe_cell_size_next = cell_to_probe / cascades.data[params.cascade + 1].to_cell; probe_pos -= cascades.data[params.cascade + 1].offset; - probe_pos *= cascades.data[params.cascade + 1].to_cell; - probe_pos = probe_pos * float(params.probe_axis_size - 1) / float(params.grid_size.x); + probe_pos /= probe_cell_size_next; ivec3 probe_posi = ivec3(probe_pos); //add up all light, no need to use occlusion here, since occlusion will do its work afterwards @@ -549,20 +561,28 @@ void main() { } } else { - // clear and let it re-raytrace, only for the last cascade, which happens very un-often - //scroll + //scroll at the edge of the highest cascade, just copy what is there, + //since its the closest we have anyway + for (uint j = 0; j < params.history_size; j++) { + ivec2 tex_pos; + tex_pos = probe_cell.xy; + tex_pos.x += probe_cell.z * int(params.probe_axis_size); + for (int i = 0; i < SH_SIZE; i++) { // copy from history texture + ivec3 src_pos = ivec3(tex_pos.x, tex_pos.y * SH_SIZE + i, int(j)); ivec3 dst_pos = ivec3(pos.x, pos.y * SH_SIZE + i, int(j)); - imageStore(lightprobe_history_scroll_texture, dst_pos, ivec4(0)); + ivec4 value = imageLoad(lightprobe_history_texture, dst_pos); + imageStore(lightprobe_history_scroll_texture, dst_pos, value); } } for (int i = 0; i < SH_SIZE; i++) { // copy from average texture - ivec2 dst_pos = ivec2(pos.x, pos.y * SH_SIZE + i); - imageStore(lightprobe_average_scroll_texture, dst_pos, ivec4(0)); + ivec2 spos = ivec2(pos.x, pos.y * SH_SIZE + i); + ivec4 average = imageLoad(lightprobe_average_texture, spos); + imageStore(lightprobe_average_scroll_texture, spos, average); } } diff --git a/servers/rendering/renderer_rd/shaders/shadow_reduce.glsl b/servers/rendering/renderer_rd/shaders/shadow_reduce.glsl deleted file mode 100644 index 29443ae7db..0000000000 --- a/servers/rendering/renderer_rd/shaders/shadow_reduce.glsl +++ /dev/null @@ -1,105 +0,0 @@ -#[compute] - -#version 450 - -VERSION_DEFINES - -#define BLOCK_SIZE 8 - -layout(local_size_x = BLOCK_SIZE, local_size_y = BLOCK_SIZE, local_size_z = 1) in; - -#ifdef MODE_REDUCE - -shared float tmp_data[BLOCK_SIZE * BLOCK_SIZE]; -const uint swizzle_table[BLOCK_SIZE] = uint[](0, 4, 2, 6, 1, 5, 3, 7); -const uint unswizzle_table[BLOCK_SIZE] = uint[](0, 0, 0, 1, 0, 2, 1, 3); - -#endif - -layout(r32f, set = 0, binding = 0) uniform restrict readonly image2D source_depth; -layout(r32f, set = 0, binding = 1) uniform restrict writeonly image2D dst_depth; - -layout(push_constant, binding = 1, std430) uniform Params { - ivec2 source_size; - ivec2 source_offset; - uint min_size; - uint gaussian_kernel_version; - ivec2 filter_dir; -} -params; - -void main() { -#ifdef MODE_REDUCE - - uvec2 pos = gl_LocalInvocationID.xy; - - ivec2 image_offset = params.source_offset; - ivec2 image_pos = image_offset + ivec2(gl_GlobalInvocationID.xy); - uint dst_t = swizzle_table[pos.y] * BLOCK_SIZE + swizzle_table[pos.x]; - tmp_data[dst_t] = imageLoad(source_depth, min(image_pos, params.source_size - ivec2(1))).r; - ivec2 image_size = params.source_size; - - uint t = pos.y * BLOCK_SIZE + pos.x; - - //neighbours - uint size = BLOCK_SIZE; - - do { - groupMemoryBarrier(); - barrier(); - - size >>= 1; - image_size >>= 1; - image_offset >>= 1; - - if (all(lessThan(pos, uvec2(size)))) { - uint nx = t + size; - uint ny = t + (BLOCK_SIZE * size); - uint nxy = ny + size; - - tmp_data[t] += tmp_data[nx]; - tmp_data[t] += tmp_data[ny]; - tmp_data[t] += tmp_data[nxy]; - tmp_data[t] /= 4.0; - } - - } while (size > params.min_size); - - if (all(lessThan(pos, uvec2(size)))) { - image_pos = ivec2(unswizzle_table[size + pos.x], unswizzle_table[size + pos.y]); - image_pos += image_offset + ivec2(gl_WorkGroupID.xy) * int(size); - - image_size = max(ivec2(1), image_size); //in case image size became 0 - - if (all(lessThan(image_pos, uvec2(image_size)))) { - imageStore(dst_depth, image_pos, vec4(tmp_data[t])); - } - } -#endif - -#ifdef MODE_FILTER - - ivec2 image_pos = params.source_offset + ivec2(gl_GlobalInvocationID.xy); - if (any(greaterThanEqual(image_pos, params.source_size))) { - return; - } - - ivec2 clamp_min = ivec2(params.source_offset); - ivec2 clamp_max = ivec2(params.source_size) - 1; - - //gaussian kernel, size 9, sigma 4 - const int kernel_size = 9; - const float gaussian_kernel[kernel_size * 3] = float[]( - 0.000229, 0.005977, 0.060598, 0.241732, 0.382928, 0.241732, 0.060598, 0.005977, 0.000229, - 0.028532, 0.067234, 0.124009, 0.179044, 0.20236, 0.179044, 0.124009, 0.067234, 0.028532, - 0.081812, 0.101701, 0.118804, 0.130417, 0.134535, 0.130417, 0.118804, 0.101701, 0.081812); - float accum = 0.0; - for (int i = 0; i < kernel_size; i++) { - ivec2 ofs = clamp(image_pos + params.filter_dir * (i - kernel_size / 2), clamp_min, clamp_max); - accum += imageLoad(source_depth, ofs).r * gaussian_kernel[params.gaussian_kernel_version + i]; - } - - imageStore(dst_depth, image_pos, vec4(accum)); - -#endif -} diff --git a/servers/rendering/renderer_rd/shaders/skeleton.glsl b/servers/rendering/renderer_rd/shaders/skeleton.glsl index b19f5a9ad3..680d1045cd 100644 --- a/servers/rendering/renderer_rd/shaders/skeleton.glsl +++ b/servers/rendering/renderer_rd/shaders/skeleton.glsl @@ -100,7 +100,7 @@ void main() { for (uint i = 0; i < params.blend_shape_count; i++) { float w = blend_shape_weights.data[i]; - if (w > 0.0001) { + if (abs(w) > 0.0001) { uint base_offset = (params.vertex_count * i + index) * params.vertex_stride; blend_vertex += uintBitsToFloat(uvec3(src_blend_shapes.data[base_offset + 0], src_blend_shapes.data[base_offset + 1], src_blend_shapes.data[base_offset + 2])) * w; diff --git a/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl b/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl index aa32809a06..e7ba8feb80 100644 --- a/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl +++ b/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl @@ -168,13 +168,18 @@ layout(set = 0, binding = 14, std140) uniform Params { uint cluster_shift; uint cluster_width; - uvec3 cluster_pad; uint max_cluster_element_count_div_32; + bool use_temporal_reprojection; + uint temporal_frame; + float temporal_blend; mat3x4 cam_rotation; + mat4 to_prev_view; } params; +layout(set = 0, binding = 15) uniform texture3D prev_density_texture; + float get_depth_at_pos(float cell_depth_size, int z) { float d = float(z) * cell_depth_size + cell_depth_size * 0.5; //center of voxels d = pow(d, params.detail_spread); @@ -213,6 +218,26 @@ uint cluster_get_range_clip_mask(uint i, uint z_min, uint z_max) { return bitfieldInsert(uint(0), uint(0xFFFFFFFF), local_min, mask_width); } +#define TEMPORAL_FRAMES 16 + +const vec3 halton_map[TEMPORAL_FRAMES] = vec3[]( + vec3(0.5, 0.33333333, 0.2), + vec3(0.25, 0.66666667, 0.4), + vec3(0.75, 0.11111111, 0.6), + vec3(0.125, 0.44444444, 0.8), + vec3(0.625, 0.77777778, 0.04), + vec3(0.375, 0.22222222, 0.24), + vec3(0.875, 0.55555556, 0.44), + vec3(0.0625, 0.88888889, 0.64), + vec3(0.5625, 0.03703704, 0.84), + vec3(0.3125, 0.37037037, 0.08), + vec3(0.8125, 0.7037037, 0.28), + vec3(0.1875, 0.14814815, 0.48), + vec3(0.6875, 0.48148148, 0.68), + vec3(0.4375, 0.81481481, 0.88), + vec3(0.9375, 0.25925926, 0.12), + vec3(0.03125, 0.59259259, 0.32)); + void main() { vec3 fog_cell_size = 1.0 / vec3(params.fog_volume_size); @@ -241,6 +266,45 @@ void main() { view_pos.z = -params.fog_frustum_end * fog_unit_pos.z; view_pos.y = -view_pos.y; + vec4 reprojected_density = vec4(0.0); + float reproject_amount = 0.0; + + if (params.use_temporal_reprojection) { + vec3 prev_view = (params.to_prev_view * vec4(view_pos, 1.0)).xyz; + //undo transform into prev view + prev_view.y = -prev_view.y; + //z back to unit size + prev_view.z /= -params.fog_frustum_end; + //xy back to unit size + prev_view.xy /= mix(params.fog_frustum_size_begin, params.fog_frustum_size_end, vec2(prev_view.z)); + prev_view.xy = prev_view.xy * 0.5 + 0.5; + //z back to unspread value + prev_view.z = pow(prev_view.z, 1.0 / params.detail_spread); + + if (all(greaterThan(prev_view, vec3(0.0))) && all(lessThan(prev_view, vec3(1.0)))) { + //reprojectinon fits + + reprojected_density = textureLod(sampler3D(prev_density_texture, linear_sampler), prev_view, 0.0); + reproject_amount = params.temporal_blend; + + // Since we can reproject, now we must jitter the current view pos. + // This is done here because cells that can't reproject should not jitter. + + fog_unit_pos = posf * fog_cell_size + fog_cell_size * halton_map[params.temporal_frame]; //center of voxels, offset by halton table + + screen_pos = uvec2(fog_unit_pos.xy * params.screen_size); + cluster_pos = screen_pos >> params.cluster_shift; + cluster_offset = (params.cluster_width * cluster_pos.y + cluster_pos.x) * (params.max_cluster_element_count_div_32 + 32); + //positions in screen are too spread apart, no hopes for optimizing with subgroups + + fog_unit_pos.z = pow(fog_unit_pos.z, params.detail_spread); + + view_pos.xy = (fog_unit_pos.xy * 2.0 - 1.0) * mix(params.fog_frustum_size_begin, params.fog_frustum_size_end, vec2(fog_unit_pos.z)); + view_pos.z = -params.fog_frustum_end * fog_unit_pos.z; + view_pos.y = -view_pos.y; + } + } + uint cluster_z = uint(clamp((abs(view_pos.z) / params.z_far) * 32.0, 0.0, 31.0)); vec3 total_light = params.light_color; @@ -433,31 +497,31 @@ void main() { uint light_index = 32 * i + bit; - vec3 light_pos = omni_lights.data[light_index].position; - vec3 light_rel_vec = omni_lights.data[light_index].position - view_pos; + vec3 light_pos = spot_lights.data[light_index].position; + vec3 light_rel_vec = spot_lights.data[light_index].position - view_pos; float d = length(light_rel_vec); float shadow_attenuation = 1.0; - if (d * omni_lights.data[light_index].inv_radius < 1.0) { - float attenuation = get_omni_attenuation(d, omni_lights.data[light_index].inv_radius, omni_lights.data[light_index].attenuation); + if (d * spot_lights.data[light_index].inv_radius < 1.0) { + float attenuation = get_omni_attenuation(d, spot_lights.data[light_index].inv_radius, spot_lights.data[light_index].attenuation); - vec3 spot_dir = omni_lights.data[light_index].direction; - float scos = max(dot(-normalize(light_rel_vec), spot_dir), omni_lights.data[light_index].cone_angle); - float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - omni_lights.data[light_index].cone_angle)); - attenuation *= 1.0 - pow(spot_rim, omni_lights.data[light_index].cone_attenuation); + vec3 spot_dir = spot_lights.data[light_index].direction; + float scos = max(dot(-normalize(light_rel_vec), spot_dir), spot_lights.data[light_index].cone_angle); + float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - spot_lights.data[light_index].cone_angle)); + attenuation *= 1.0 - pow(spot_rim, spot_lights.data[light_index].cone_attenuation); - vec3 light = omni_lights.data[light_index].color / M_PI; + vec3 light = spot_lights.data[light_index].color / M_PI; - if (omni_lights.data[light_index].shadow_enabled) { + if (spot_lights.data[light_index].shadow_enabled) { //has shadow vec4 v = vec4(view_pos, 1.0); - vec4 splane = (omni_lights.data[light_index].shadow_matrix * v); + vec4 splane = (spot_lights.data[light_index].shadow_matrix * v); splane /= splane.w; float depth = texture(sampler2D(shadow_atlas, linear_sampler), splane.xy).r; - shadow_attenuation = exp(min(0.0, (depth - splane.z)) / omni_lights.data[light_index].inv_radius * omni_lights.data[light_index].shadow_volumetric_fog_fade); + shadow_attenuation = exp(min(0.0, (depth - splane.z)) / spot_lights.data[light_index].inv_radius * spot_lights.data[light_index].shadow_volumetric_fog_fade); } total_light += light * attenuation * shadow_attenuation; @@ -565,7 +629,11 @@ void main() { #endif - imageStore(density_map, pos, vec4(total_light, total_density)); + vec4 final_density = vec4(total_light, total_density); + + final_density = mix(final_density, reprojected_density, reproject_amount); + + imageStore(density_map, pos, final_density); #endif #ifdef MODE_FOG |