diff options
Diffstat (limited to 'servers/rendering/renderer_rd')
25 files changed, 715 insertions, 132 deletions
diff --git a/servers/rendering/renderer_rd/cluster_builder_rd.cpp b/servers/rendering/renderer_rd/cluster_builder_rd.cpp index b952ecbff0..4a98cf0831 100644 --- a/servers/rendering/renderer_rd/cluster_builder_rd.cpp +++ b/servers/rendering/renderer_rd/cluster_builder_rd.cpp @@ -460,14 +460,6 @@ void ClusterBuilderRD::bake_cluster() { RD::get_singleton()->draw_list_set_push_constant(draw_list, &push_constant, sizeof(ClusterBuilderSharedDataRD::ClusterRender::PushConstant)); uint32_t instances = 1; -#if 0 - for (uint32_t j = i+1; j < element_count; j++) { - if (elements[i].type!=elements[j].type) { - break; - } - instances++; - } -#endif RD::get_singleton()->draw_list_draw(draw_list, true, instances); i += instances; } diff --git a/servers/rendering/renderer_rd/effects_rd.cpp b/servers/rendering/renderer_rd/effects_rd.cpp index fdd6939a8b..cf943901d4 100644 --- a/servers/rendering/renderer_rd/effects_rd.cpp +++ b/servers/rendering/renderer_rd/effects_rd.cpp @@ -237,6 +237,43 @@ RID EffectsRD::_get_compute_uniform_set_from_image_pair(RID p_texture1, RID p_te return uniform_set; } +void EffectsRD::fsr_upscale(RID p_source_rd_texture, RID p_secondary_texture, RID p_destination_texture, const Size2i &p_internal_size, const Size2i &p_size, float p_fsr_upscale_sharpness) { + memset(&FSR_upscale.push_constant, 0, sizeof(FSRUpscalePushConstant)); + + int dispatch_x = (p_size.x + 15) / 16; + int dispatch_y = (p_size.y + 15) / 16; + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, FSR_upscale.pipeline); + + FSR_upscale.push_constant.resolution_width = p_internal_size.width; + FSR_upscale.push_constant.resolution_height = p_internal_size.height; + FSR_upscale.push_constant.upscaled_width = p_size.width; + FSR_upscale.push_constant.upscaled_height = p_size.height; + FSR_upscale.push_constant.sharpness = p_fsr_upscale_sharpness; + + //FSR Easc + FSR_upscale.push_constant.pass = FSR_UPSCALE_PASS_EASU; + 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_secondary_texture), 1); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &FSR_upscale.push_constant, sizeof(FSRUpscalePushConstant)); + + RD::get_singleton()->compute_list_dispatch(compute_list, dispatch_x, dispatch_y, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + //FSR Rcas + FSR_upscale.push_constant.pass = FSR_UPSCALE_PASS_RCAS; + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_secondary_texture), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_destination_texture), 1); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &FSR_upscale.push_constant, sizeof(FSRUpscalePushConstant)); + + RD::get_singleton()->compute_list_dispatch(compute_list, dispatch_x, dispatch_y, 1); + + RD::get_singleton()->compute_list_end(compute_list); +} + void EffectsRD::copy_to_atlas_fb(RID p_source_rd_texture, RID p_dest_framebuffer, const Rect2 &p_uv_rect, RD::DrawListID p_draw_list, bool p_flip_y, bool p_panorama) { memset(©_to_fb.push_constant, 0, sizeof(CopyToFbPushConstant)); @@ -1888,6 +1925,27 @@ void EffectsRD::sort_buffer(RID p_uniform_set, int p_size) { } EffectsRD::EffectsRD(bool p_prefer_raster_effects) { + { + Vector<String> FSR_upscale_modes; + +#if defined(OSX_ENABLED) || defined(IPHONE_ENABLED) + // MoltenVK does not support some of the operations used by the normal mode of FSR. Fallback works just fine though. + FSR_upscale_modes.push_back("\n#define MODE_FSR_UPSCALE_FALLBACK\n"); +#else + // Everyone else can use normal mode when available. + if (RD::get_singleton()->get_device_capabilities()->supports_fsr_half_float) { + FSR_upscale_modes.push_back("\n#define MODE_FSR_UPSCALE_NORMAL\n"); + } else { + FSR_upscale_modes.push_back("\n#define MODE_FSR_UPSCALE_FALLBACK\n"); + } +#endif + + FSR_upscale.shader.initialize(FSR_upscale_modes); + + FSR_upscale.shader_version = FSR_upscale.shader.version_create(); + FSR_upscale.pipeline = RD::get_singleton()->compute_pipeline_create(FSR_upscale.shader.version_get_shader(FSR_upscale.shader_version, 0)); + } + prefer_raster_effects = p_prefer_raster_effects; if (prefer_raster_effects) { @@ -2523,6 +2581,7 @@ EffectsRD::~EffectsRD() { RD::get_singleton()->free(index_buffer); //array gets freed as dependency RD::get_singleton()->free(filter.coefficient_buffer); + FSR_upscale.shader.version_free(FSR_upscale.shader_version); if (prefer_raster_effects) { blur_raster.shader.version_free(blur_raster.shader_version); bokeh.raster_shader.version_free(blur_raster.shader_version); diff --git a/servers/rendering/renderer_rd/effects_rd.h b/servers/rendering/renderer_rd/effects_rd.h index 551e50ed25..6037127e82 100644 --- a/servers/rendering/renderer_rd/effects_rd.h +++ b/servers/rendering/renderer_rd/effects_rd.h @@ -45,6 +45,7 @@ #include "servers/rendering/renderer_rd/shaders/cubemap_filter_raster.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/cubemap_roughness_raster.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/fsr_upscale.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/luminance_reduce.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/luminance_reduce_raster.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/resolve.glsl.gen.h" @@ -69,6 +70,28 @@ class EffectsRD { private: bool prefer_raster_effects; + enum FSRUpscalePass { + FSR_UPSCALE_PASS_EASU = 0, + FSR_UPSCALE_PASS_RCAS = 1 + }; + + struct FSRUpscalePushConstant { + float resolution_width; + float resolution_height; + float upscaled_width; + float upscaled_height; + float sharpness; + int pass; + int _unused0, _unused1; + }; + + struct FSRUpscale { + FSRUpscalePushConstant push_constant; + FsrUpscaleShaderRD shader; + RID shader_version; + RID pipeline; + } FSR_upscale; + enum BlurRasterMode { BLUR_MIPMAP, @@ -754,6 +777,7 @@ private: public: bool get_prefer_raster_effects(); + void fsr_upscale(RID p_source_rd_texture, RID p_secondary_texture, RID p_destination_texture, const Size2i &p_internal_size, const Size2i &p_size, float p_fsr_upscale_sharpness); void copy_to_fb_rect(RID p_source_rd_texture, RID p_dest_framebuffer, const Rect2i &p_rect, bool p_flip_y = false, bool p_force_luminance = false, bool p_alpha_to_zero = false, bool p_srgb = false, RID p_secondary = RID()); void copy_to_rect(RID p_source_rd_texture, RID p_dest_texture, const Rect2i &p_rect, bool p_flip_y = false, bool p_force_luminance = false, bool p_all_source = false, bool p_8_bit_dst = false, bool p_alpha_to_one = false); void copy_cubemap_to_panorama(RID p_source_cube, RID p_dest_panorama, const Size2i &p_panorama_size, float p_lod, bool p_is_array); diff --git a/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp index 22e3bbdb00..03ce2690bf 100644 --- a/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp +++ b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp @@ -334,6 +334,10 @@ void RenderForwardClustered::_render_list_template(RenderingDevice::DrawListID p const GeometryInstanceSurfaceDataCache *surf = p_params->elements[i]; const RenderElementInfo &element_info = p_params->element_info[i]; + if ((p_pass_mode == PASS_MODE_COLOR || p_pass_mode == PASS_MODE_COLOR_SPECULAR) && !(surf->flags & GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE)) { + continue; // Objects with "Depth-prepass" transparency are included in both render lists, but should only be rendered in the transparent pass + } + if (surf->owner->instance_count == 0) { continue; } @@ -958,20 +962,20 @@ void RenderForwardClustered::_fill_render_list(RenderListType p_render_list, con bool uses_gi = false; float fade_alpha = 1.0; - if (p_render_list == RENDER_LIST_OPAQUE) { - if (inst->fade_near || inst->fade_far) { - float fade_dist = inst->transform.origin.distance_to(p_render_data->cam_transform.origin); - if (inst->fade_far && fade_dist > inst->fade_far_begin) { - fade_alpha = MAX(0.0, 1.0 - (fade_dist - inst->fade_far_begin) / (inst->fade_far_end - inst->fade_far_begin)); - } else if (inst->fade_near && fade_dist < inst->fade_near_end) { - fade_alpha = MAX(0.0, (fade_dist - inst->fade_near_begin) / (inst->fade_near_end - inst->fade_near_begin)); - } + if (inst->fade_near || inst->fade_far) { + float fade_dist = inst->transform.origin.distance_to(p_render_data->cam_transform.origin); + if (inst->fade_far && fade_dist > inst->fade_far_begin) { + fade_alpha = MAX(0.0, 1.0 - (fade_dist - inst->fade_far_begin) / (inst->fade_far_end - inst->fade_far_begin)); + } else if (inst->fade_near && fade_dist < inst->fade_near_end) { + fade_alpha = MAX(0.0, (fade_dist - inst->fade_near_begin) / (inst->fade_near_end - inst->fade_near_begin)); } + } - fade_alpha *= inst->force_alpha * inst->parent_fade_alpha; + fade_alpha *= inst->force_alpha * inst->parent_fade_alpha; - flags = (flags & ~INSTANCE_DATA_FLAGS_FADE_MASK) | (uint32_t(fade_alpha * 255.0) << INSTANCE_DATA_FLAGS_FADE_SHIFT); + flags = (flags & ~INSTANCE_DATA_FLAGS_FADE_MASK) | (uint32_t(fade_alpha * 255.0) << INSTANCE_DATA_FLAGS_FADE_SHIFT); + if (p_render_list == RENDER_LIST_OPAQUE) { // Setup GI if (inst->lightmap_instance.is_valid()) { int32_t lightmap_cull_index = -1; @@ -1206,6 +1210,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co scene_state.ubo.viewport_size[0] = vp_he.x; scene_state.ubo.viewport_size[1] = vp_he.y; scene_state.ubo.directional_light_count = 0; + scene_state.ubo.opaque_prepass_threshold = 0.99f; Size2i screen_size; RID opaque_framebuffer; @@ -1449,6 +1454,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co RD::get_singleton()->draw_command_begin_label("Render Opaque Pass"); scene_state.ubo.directional_light_count = p_render_data->directional_light_count; + scene_state.ubo.opaque_prepass_threshold = 0.0f; _setup_environment(p_render_data, p_render_data->reflection_probe.is_valid(), screen_size, !p_render_data->reflection_probe.is_valid(), p_default_bg_color, p_render_data->render_buffers.is_valid()); @@ -1632,6 +1638,7 @@ void RenderForwardClustered::_render_shadow_append(RID p_framebuffer, const Page render_data.render_info = p_render_info; scene_state.ubo.dual_paraboloid_side = p_use_dp_flip ? -1 : 1; + scene_state.ubo.opaque_prepass_threshold = 0.1f; _setup_environment(&render_data, true, Vector2(1, 1), !p_flip_y, Color(), false, p_use_pancake, shadow_pass_index); @@ -1718,6 +1725,7 @@ void RenderForwardClustered::_render_particle_collider_heightfield(RID p_fb, con _update_render_base_uniform_set(); scene_state.ubo.dual_paraboloid_side = 0; + scene_state.ubo.opaque_prepass_threshold = 0.0; _setup_environment(&render_data, true, Vector2(1, 1), true, Color(), false, false); @@ -1755,6 +1763,7 @@ void RenderForwardClustered::_render_material(const Transform3D &p_cam_transform scene_state.ubo.dual_paraboloid_side = 0; scene_state.ubo.material_uv2_mode = false; + scene_state.ubo.opaque_prepass_threshold = 0.0f; _setup_environment(&render_data, true, Vector2(1, 1), false, Color()); @@ -1798,6 +1807,7 @@ void RenderForwardClustered::_render_uv2(const PagedArray<GeometryInstance *> &p scene_state.ubo.dual_paraboloid_side = 0; scene_state.ubo.material_uv2_mode = true; + scene_state.ubo.opaque_prepass_threshold = 0.0; _setup_environment(&render_data, true, Vector2(1, 1), false, Color()); @@ -1944,7 +1954,9 @@ void RenderForwardClustered::_base_uniforms_changed() { } void RenderForwardClustered::_update_render_base_uniform_set() { - if (render_base_uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(render_base_uniform_set) || (lightmap_texture_array_version != storage->lightmap_array_get_version())) { + if (render_base_uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(render_base_uniform_set) || (lightmap_texture_array_version != storage->lightmap_array_get_version()) || base_uniform_set_updated) { + base_uniform_set_updated = false; + if (render_base_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(render_base_uniform_set)) { RD::get_singleton()->free(render_base_uniform_set); } @@ -1959,18 +1971,18 @@ void RenderForwardClustered::_update_render_base_uniform_set() { u.binding = 1; u.ids.resize(12); RID *ids_ptr = u.ids.ptrw(); - ids_ptr[0] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); - ids_ptr[1] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); - ids_ptr[2] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); - ids_ptr[3] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); - ids_ptr[4] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); - ids_ptr[5] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); - ids_ptr[6] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); - ids_ptr[7] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); - ids_ptr[8] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); - ids_ptr[9] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); - ids_ptr[10] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); - ids_ptr[11] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[0] = storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[1] = storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[2] = storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[3] = storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[4] = storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[5] = storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + ids_ptr[6] = storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[7] = storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[8] = storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[9] = storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[10] = storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); + ids_ptr[11] = storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED); uniforms.push_back(u); } @@ -1989,19 +2001,19 @@ void RenderForwardClustered::_update_render_base_uniform_set() { RID sampler; switch (decals_get_filter()) { case RS::DECAL_FILTER_NEAREST: { - sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + sampler = storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); } break; case RS::DECAL_FILTER_NEAREST_MIPMAPS: { - sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + sampler = storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); } break; case RS::DECAL_FILTER_LINEAR: { - sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + sampler = storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); } break; case RS::DECAL_FILTER_LINEAR_MIPMAPS: { - sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + sampler = storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); } break; case RS::DECAL_FILTER_LINEAR_MIPMAPS_ANISOTROPIC: { - sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + sampler = storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); } break; } @@ -2016,19 +2028,19 @@ void RenderForwardClustered::_update_render_base_uniform_set() { RID sampler; switch (light_projectors_get_filter()) { case RS::LIGHT_PROJECTOR_FILTER_NEAREST: { - sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + sampler = storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); } break; case RS::LIGHT_PROJECTOR_FILTER_NEAREST_MIPMAPS: { - sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + sampler = storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); } break; case RS::LIGHT_PROJECTOR_FILTER_LINEAR: { - sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + sampler = storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); } break; case RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS: { - sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + sampler = storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); } break; case RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS_ANISOTROPIC: { - sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + sampler = storage->sampler_rd_get_custom(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); } break; } diff --git a/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h index 2ba9558128..d6ab4d1db2 100644 --- a/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h +++ b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h @@ -129,6 +129,7 @@ class RenderForwardClustered : public RendererSceneRenderRD { virtual void _base_uniforms_changed() override; virtual RID _render_buffers_get_normal_texture(RID p_render_buffers) override; + bool base_uniform_set_updated = false; void _update_render_base_uniform_set(); RID _setup_sdfgi_render_pass_uniform_set(RID p_albedo_texture, RID p_emission_texture, RID p_emission_aniso_texture, RID p_geom_facing_texture); RID _setup_render_pass_uniform_set(RenderListType p_render_list, const RenderDataRD *p_render_data, RID p_radiance_texture, bool p_use_directional_shadow_atlas = false, int p_index = 0); @@ -257,7 +258,8 @@ class RenderForwardClustered : public RendererSceneRenderRD { float roughness_limiter_amount; float roughness_limiter_limit; - uint32_t roughness_limiter_pad[2]; + float opaque_prepass_threshold; + uint32_t roughness_limiter_pad; float sdf_to_bounds[16]; @@ -602,6 +604,11 @@ protected: virtual void _render_particle_collider_heightfield(RID p_fb, const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<GeometryInstance *> &p_instances) override; public: + _FORCE_INLINE_ virtual void update_uniform_sets() override { + base_uniform_set_updated = true; + _update_render_base_uniform_set(); + } + virtual GeometryInstance *geometry_instance_create(RID p_base) override; virtual void geometry_instance_set_skeleton(GeometryInstance *p_geometry_instance, RID p_skeleton) override; virtual void geometry_instance_set_material_override(GeometryInstance *p_geometry_instance, RID p_override) override; diff --git a/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.cpp b/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.cpp index 61f2031a70..768bd1de9d 100644 --- a/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.cpp +++ b/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.cpp @@ -650,6 +650,7 @@ void SceneShaderForwardClustered::init(RendererStorageRD *p_storage, const Strin actions.render_mode_defines["cull_front"] = "#define DO_SIDE_CHECK\n"; actions.render_mode_defines["cull_disabled"] = "#define DO_SIDE_CHECK\n"; actions.render_mode_defines["particle_trails"] = "#define USE_PARTICLE_TRAILS\n"; + actions.render_mode_defines["depth_draw_opaque"] = "#define USE_OPAQUE_PREPASS\n"; bool force_lambert = GLOBAL_GET("rendering/shading/overrides/force_lambert_over_burley"); diff --git a/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.cpp b/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.cpp index f23a386acf..445623fb86 100644 --- a/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.cpp +++ b/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.cpp @@ -483,6 +483,7 @@ void RenderForwardMobile::_render_scene(RenderDataRD *p_render_data, const Color scene_state.ubo.viewport_size[0] = vp_he.x; scene_state.ubo.viewport_size[1] = vp_he.y; scene_state.ubo.directional_light_count = 0; + scene_state.ubo.opaque_prepass_threshold = 0.0; // We can only use our full subpass approach if we're: // - not reading from SCREEN_TEXTURE/DEPTH_TEXTURE @@ -793,7 +794,7 @@ void RenderForwardMobile::_render_scene(RenderDataRD *p_render_data, const Color if (using_subpass_transparent) { RD::FramebufferFormatID fb_format = RD::get_singleton()->framebuffer_get_format(framebuffer); - RenderListParameters render_list_params(render_list[RENDER_LIST_ALPHA].elements.ptr(), render_list[RENDER_LIST_ALPHA].element_info.ptr(), render_list[RENDER_LIST_ALPHA].elements.size(), reverse_cull, PASS_MODE_COLOR, rp_uniform_set, spec_constant_base_flags, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->lod_camera_plane, p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold, p_render_data->view_count); + RenderListParameters render_list_params(render_list[RENDER_LIST_ALPHA].elements.ptr(), render_list[RENDER_LIST_ALPHA].element_info.ptr(), render_list[RENDER_LIST_ALPHA].elements.size(), reverse_cull, PASS_MODE_COLOR_TRANSPARENT, rp_uniform_set, spec_constant_base_flags, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->lod_camera_plane, p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold, p_render_data->view_count); render_list_params.framebuffer_format = fb_format; if ((uint32_t)render_list_params.element_count > render_list_thread_threshold && false) { // secondary command buffers need more testing at this time @@ -897,6 +898,7 @@ void RenderForwardMobile::_render_shadow_append(RID p_framebuffer, const PagedAr render_data.lod_distance_multiplier = p_lod_distance_multiplier; scene_state.ubo.dual_paraboloid_side = p_use_dp_flip ? -1 : 1; + scene_state.ubo.opaque_prepass_threshold = 0.1; _setup_environment(&render_data, true, Vector2(1, 1), !p_flip_y, Color(), false, p_use_pancake, shadow_pass_index); @@ -978,6 +980,7 @@ void RenderForwardMobile::_render_material(const Transform3D &p_cam_transform, c scene_state.ubo.dual_paraboloid_side = 0; scene_state.ubo.material_uv2_mode = false; + scene_state.ubo.opaque_prepass_threshold = 0.0f; RenderDataRD render_data; render_data.cam_projection = p_cam_projection; @@ -1089,6 +1092,7 @@ void RenderForwardMobile::_render_particle_collider_heightfield(RID p_fb, const _update_render_base_uniform_set(); scene_state.ubo.dual_paraboloid_side = 0; + scene_state.ubo.opaque_prepass_threshold = 0.0; RenderDataRD render_data; render_data.cam_projection = p_cam_projection; @@ -1447,13 +1451,13 @@ void RenderForwardMobile::_fill_render_list(RenderListType p_render_list, const } // ADD Element - if (p_pass_mode == PASS_MODE_COLOR) { + if (p_pass_mode == PASS_MODE_COLOR || p_pass_mode == PASS_MODE_COLOR_TRANSPARENT) { #ifdef DEBUG_ENABLED bool force_alpha = unlikely(get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_OVERDRAW); #else bool force_alpha = false; #endif - if (!force_alpha && (surf->flags & (GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH | GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE))) { + if (!force_alpha && (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE)) { rl->add_element(surf); } if (force_alpha || (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_PASS_ALPHA)) { diff --git a/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.h b/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.h index 74bfe16557..485d08b589 100644 --- a/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.h +++ b/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.h @@ -292,7 +292,8 @@ protected: float roughness_limiter_amount; float roughness_limiter_limit; - uint32_t roughness_limiter_pad[2]; + float opaque_prepass_threshold; + uint32_t roughness_limiter_pad; // Fog uint32_t fog_enabled; diff --git a/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.cpp b/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.cpp index 5d5cb7a6b7..e6d9a60f94 100644 --- a/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.cpp +++ b/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.cpp @@ -637,6 +637,7 @@ void SceneShaderForwardMobile::init(RendererStorageRD *p_storage, const String p actions.render_mode_defines["cull_front"] = "#define DO_SIDE_CHECK\n"; actions.render_mode_defines["cull_disabled"] = "#define DO_SIDE_CHECK\n"; actions.render_mode_defines["particle_trails"] = "#define USE_PARTICLE_TRAILS\n"; + actions.render_mode_defines["depth_draw_opaque"] = "#define USE_OPAQUE_PREPASS\n"; bool force_lambert = GLOBAL_GET("rendering/shading/overrides/force_lambert_over_burley"); if (!force_lambert) { diff --git a/servers/rendering/renderer_rd/renderer_compositor_rd.cpp b/servers/rendering/renderer_rd/renderer_compositor_rd.cpp index 559e6d5ad7..522a8e8112 100644 --- a/servers/rendering/renderer_rd/renderer_compositor_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_compositor_rd.cpp @@ -281,12 +281,12 @@ RendererCompositorRD::RendererCompositorRD() { storage = memnew(RendererStorageRD); canvas = memnew(RendererCanvasRenderRD(storage)); - uint32_t back_end = GLOBAL_GET("rendering/vulkan/rendering/back_end"); + back_end = (bool)(int)GLOBAL_GET("rendering/vulkan/rendering/back_end"); uint32_t textures_per_stage = RD::get_singleton()->limit_get(RD::LIMIT_MAX_TEXTURES_PER_SHADER_STAGE); - if (back_end == 1 || textures_per_stage < 48) { + if (back_end || textures_per_stage < 48) { scene = memnew(RendererSceneRenderImplementation::RenderForwardMobile(storage)); - } else { // back_end == 0 + } else { // back_end == false // default to our high end renderer scene = memnew(RendererSceneRenderImplementation::RenderForwardClustered(storage)); } diff --git a/servers/rendering/renderer_rd/renderer_compositor_rd.h b/servers/rendering/renderer_rd/renderer_compositor_rd.h index 0230c46800..f69e40e0ff 100644 --- a/servers/rendering/renderer_rd/renderer_compositor_rd.h +++ b/servers/rendering/renderer_rd/renderer_compositor_rd.h @@ -116,8 +116,6 @@ public: _create_func = _create_current; } - virtual bool is_low_end() const { return false; } - static RendererCompositorRD *singleton; RendererCompositorRD(); ~RendererCompositorRD(); diff --git a/servers/rendering/renderer_rd/renderer_scene_gi_rd.cpp b/servers/rendering/renderer_rd/renderer_scene_gi_rd.cpp index 807af00c8e..b6b5c90b39 100644 --- a/servers/rendering/renderer_rd/renderer_scene_gi_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_scene_gi_rd.cpp @@ -3132,8 +3132,8 @@ void RendererSceneGIRD::process_gi(RID p_render_buffers, RID p_normal_roughness_ RD::TextureFormat tf; tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; - tf.width = rb->width; - tf.height = rb->height; + tf.width = rb->internal_width; + tf.height = rb->internal_height; if (half_resolution) { tf.width >>= 1; tf.height >>= 1; @@ -3146,13 +3146,13 @@ void RendererSceneGIRD::process_gi(RID p_render_buffers, RID p_normal_roughness_ PushConstant push_constant; - push_constant.screen_size[0] = rb->width; - push_constant.screen_size[1] = rb->height; + push_constant.screen_size[0] = rb->internal_width; + push_constant.screen_size[1] = rb->internal_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[0] = -2.0f / (rb->internal_width * p_projection.matrix[0][0]); + push_constant.proj_info[1] = -2.0f / (rb->internal_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_voxel_gi_instances = MIN((uint64_t)MAX_VOXEL_GI_INSTANCES, p_voxel_gi_instances.size()); @@ -3344,9 +3344,9 @@ void RendererSceneGIRD::process_gi(RID p_render_buffers, RID p_normal_roughness_ RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(PushConstant)); if (rb->gi.using_half_size_gi) { - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->width >> 1, rb->height >> 1, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->internal_width >> 1, rb->internal_height >> 1, 1); } else { - RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->width, rb->height, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->internal_width, rb->internal_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 diff --git a/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp b/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp index b8e9f40bc4..ae8d91a73b 100644 --- a/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp @@ -1503,8 +1503,8 @@ void RendererSceneRenderRD::_allocate_blur_textures(RenderBuffers *rb) { RD::TextureFormat tf; tf.format = _render_buffers_get_color_format(); // RD::DATA_FORMAT_R16G16B16A16_SFLOAT; - tf.width = rb->width; - tf.height = rb->height; + tf.width = rb->internal_width; + tf.height = rb->internal_height; tf.texture_type = rb->view_count > 1 ? RD::TEXTURE_TYPE_2D_ARRAY : RD::TEXTURE_TYPE_2D; tf.array_layers = rb->view_count; tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; @@ -1515,6 +1515,10 @@ void RendererSceneRenderRD::_allocate_blur_textures(RenderBuffers *rb) { } tf.mipmaps = mipmaps_required; + rb->sss_texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); + + tf.width = rb->internal_width; + tf.height = rb->internal_height; rb->blur[0].texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); //the second one is smaller (only used for separatable part of blur) tf.width >>= 1; @@ -1522,8 +1526,8 @@ void RendererSceneRenderRD::_allocate_blur_textures(RenderBuffers *rb) { tf.mipmaps--; rb->blur[1].texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); - int base_width = rb->width; - int base_height = rb->height; + int base_width = rb->internal_width; + int base_height = rb->internal_height; for (uint32_t i = 0; i < mipmaps_required; i++) { RenderBuffers::Blur::Mipmap mm; @@ -1577,8 +1581,8 @@ void RendererSceneRenderRD::_allocate_blur_textures(RenderBuffers *rb) { // create 4 weight textures, 2 full size, 2 half size tf.format = RD::DATA_FORMAT_R16_SFLOAT; // We could probably use DATA_FORMAT_R8_SNORM if we don't pre-multiply by blur_size but that depends on whether we can remove DEPTH_GAP - tf.width = rb->width; - tf.height = rb->height; + tf.width = rb->internal_width; + tf.height = rb->internal_height; tf.texture_type = rb->view_count > 1 ? RD::TEXTURE_TYPE_2D_ARRAY : RD::TEXTURE_TYPE_2D; tf.array_layers = rb->view_count; tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; @@ -1656,8 +1660,8 @@ void RendererSceneRenderRD::_allocate_depth_backbuffer_textures(RenderBuffers *r void RendererSceneRenderRD::_allocate_luminance_textures(RenderBuffers *rb) { ERR_FAIL_COND(!rb->luminance.current.is_null()); - int w = rb->width; - int h = rb->height; + int w = rb->internal_width; + int h = rb->internal_height; while (true) { w = MAX(w / 8, 1); @@ -1709,9 +1713,26 @@ void RendererSceneRenderRD::_free_render_buffer_data(RenderBuffers *rb) { rb->texture_fb = RID(); } - if (rb->texture.is_valid()) { - RD::get_singleton()->free(rb->texture); + if (rb->internal_texture == rb->texture && rb->internal_texture.is_valid()) { + RD::get_singleton()->free(rb->internal_texture); rb->texture = RID(); + rb->internal_texture = RID(); + rb->upscale_texture = RID(); + } else { + if (rb->texture.is_valid()) { + RD::get_singleton()->free(rb->texture); + rb->texture = RID(); + } + + if (rb->internal_texture.is_valid()) { + RD::get_singleton()->free(rb->internal_texture); + rb->internal_texture = RID(); + } + + if (rb->upscale_texture.is_valid()) { + RD::get_singleton()->free(rb->upscale_texture); + rb->upscale_texture = RID(); + } } if (rb->depth_texture.is_valid()) { @@ -1729,6 +1750,11 @@ void RendererSceneRenderRD::_free_render_buffer_data(RenderBuffers *rb) { rb->depth_back_texture = RID(); } + if (rb->sss_texture.is_valid()) { + RD::get_singleton()->free(rb->sss_texture); + rb->sss_texture = RID(); + } + for (int i = 0; i < 2; i++) { for (int m = 0; m < rb->blur[i].mipmaps.size(); m++) { // do we free the texture slice here? or is it enough to free the main texture? @@ -1818,7 +1844,7 @@ void RendererSceneRenderRD::_process_sss(RID p_render_buffers, const CameraMatri RenderBuffers *rb = render_buffers_owner.get_or_null(p_render_buffers); ERR_FAIL_COND(!rb); - bool can_use_effects = rb->width >= 8 && rb->height >= 8; + bool can_use_effects = rb->internal_width >= 8 && rb->internal_height >= 8; if (!can_use_effects) { //just copy @@ -1829,18 +1855,18 @@ void RendererSceneRenderRD::_process_sss(RID p_render_buffers, const CameraMatri _allocate_blur_textures(rb); } - storage->get_effects()->sub_surface_scattering(rb->texture, rb->blur[0].mipmaps[0].texture, rb->depth_texture, p_camera, Size2i(rb->width, rb->height), sss_scale, sss_depth_scale, sss_quality); + storage->get_effects()->sub_surface_scattering(rb->internal_texture, rb->sss_texture, rb->depth_texture, p_camera, Size2i(rb->internal_width, rb->internal_height), sss_scale, sss_depth_scale, sss_quality); } void RendererSceneRenderRD::_process_ssr(RID p_render_buffers, RID p_dest_framebuffer, RID p_normal_buffer, RID p_specular_buffer, RID p_metallic, const Color &p_metallic_mask, RID p_environment, const CameraMatrix &p_projection, bool p_use_additive) { RenderBuffers *rb = render_buffers_owner.get_or_null(p_render_buffers); ERR_FAIL_COND(!rb); - bool can_use_effects = rb->width >= 8 && rb->height >= 8; + bool can_use_effects = rb->internal_width >= 8 && rb->internal_height >= 8; if (!can_use_effects) { //just copy - storage->get_effects()->merge_specular(p_dest_framebuffer, p_specular_buffer, p_use_additive ? RID() : rb->texture, RID()); + storage->get_effects()->merge_specular(p_dest_framebuffer, p_specular_buffer, p_use_additive ? RID() : rb->internal_texture, RID()); return; } @@ -1852,8 +1878,8 @@ void RendererSceneRenderRD::_process_ssr(RID p_render_buffers, RID p_dest_frameb if (rb->ssr.depth_scaled.is_null()) { RD::TextureFormat tf; tf.format = RD::DATA_FORMAT_R32_SFLOAT; - tf.width = rb->width / 2; - tf.height = rb->height / 2; + tf.width = rb->internal_width / 2; + tf.height = rb->internal_height / 2; tf.texture_type = RD::TEXTURE_TYPE_2D; tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT; @@ -1867,8 +1893,8 @@ void RendererSceneRenderRD::_process_ssr(RID p_render_buffers, RID p_dest_frameb if (ssr_roughness_quality != RS::ENV_SSR_ROUGNESS_QUALITY_DISABLED && !rb->ssr.blur_radius[0].is_valid()) { RD::TextureFormat tf; tf.format = RD::DATA_FORMAT_R8_UNORM; - tf.width = rb->width / 2; - tf.height = rb->height / 2; + tf.width = rb->internal_width / 2; + tf.height = rb->internal_height / 2; tf.texture_type = RD::TEXTURE_TYPE_2D; tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; @@ -1880,8 +1906,8 @@ void RendererSceneRenderRD::_process_ssr(RID p_render_buffers, RID p_dest_frameb _allocate_blur_textures(rb); } - storage->get_effects()->screen_space_reflection(rb->texture, p_normal_buffer, ssr_roughness_quality, rb->ssr.blur_radius[0], rb->ssr.blur_radius[1], p_metallic, p_metallic_mask, rb->depth_texture, rb->ssr.depth_scaled, rb->ssr.normal_scaled, rb->blur[0].mipmaps[1].texture, rb->blur[1].mipmaps[0].texture, Size2i(rb->width / 2, rb->height / 2), env->ssr_max_steps, env->ssr_fade_in, env->ssr_fade_out, env->ssr_depth_tolerance, p_projection); - storage->get_effects()->merge_specular(p_dest_framebuffer, p_specular_buffer, p_use_additive ? RID() : rb->texture, rb->blur[0].mipmaps[1].texture); + storage->get_effects()->screen_space_reflection(rb->internal_texture, p_normal_buffer, ssr_roughness_quality, rb->ssr.blur_radius[0], rb->ssr.blur_radius[1], p_metallic, p_metallic_mask, rb->depth_texture, rb->ssr.depth_scaled, rb->ssr.normal_scaled, rb->blur[0].mipmaps[1].texture, rb->blur[1].mipmaps[0].texture, Size2i(rb->internal_width / 2, rb->internal_height / 2), env->ssr_max_steps, env->ssr_fade_in, env->ssr_fade_out, env->ssr_depth_tolerance, p_projection); + storage->get_effects()->merge_specular(p_dest_framebuffer, p_specular_buffer, p_use_additive ? RID() : rb->internal_texture, rb->blur[0].mipmaps[1].texture); } void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environment, RID p_normal_buffer, const CameraMatrix &p_projection) { @@ -1918,15 +1944,15 @@ void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environmen int half_width; int half_height; if (ssao_half_size) { - buffer_width = (rb->width + 3) / 4; - buffer_height = (rb->height + 3) / 4; - half_width = (rb->width + 7) / 8; - half_height = (rb->height + 7) / 8; + buffer_width = (rb->internal_width + 3) / 4; + buffer_height = (rb->internal_height + 3) / 4; + half_width = (rb->internal_width + 7) / 8; + half_height = (rb->internal_height + 7) / 8; } else { - buffer_width = (rb->width + 1) / 2; - buffer_height = (rb->height + 1) / 2; - half_width = (rb->width + 3) / 4; - half_height = (rb->height + 3) / 4; + buffer_width = (rb->internal_width + 1) / 2; + buffer_height = (rb->internal_height + 1) / 2; + half_width = (rb->internal_width + 3) / 4; + half_height = (rb->internal_height + 3) / 4; } bool uniform_sets_are_invalid = false; if (rb->ssao.depth.is_null()) { @@ -1998,8 +2024,8 @@ void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environmen { RD::TextureFormat tf; tf.format = RD::DATA_FORMAT_R8_UNORM; - tf.width = rb->width; - tf.height = rb->height; + tf.width = rb->internal_width; + tf.height = rb->internal_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"); @@ -2022,7 +2048,7 @@ void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environmen settings.blur_passes = ssao_blur_passes; settings.fadeout_from = ssao_fadeout_from; settings.fadeout_to = ssao_fadeout_to; - settings.full_screen_size = Size2i(rb->width, rb->height); + settings.full_screen_size = Size2i(rb->internal_width, rb->internal_height); settings.half_screen_size = Size2i(buffer_width, buffer_height); settings.quarter_screen_size = Size2i(half_width, half_height); @@ -2102,9 +2128,9 @@ void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(const Rende EffectsRD::BokehBuffers buffers; - // Textures we use. - buffers.base_texture_size = Size2i(rb->width, rb->height); - buffers.base_texture = rb->texture; + // Textures we use + buffers.base_texture_size = Size2i(rb->internal_width, rb->internal_height); + buffers.base_texture = rb->internal_texture; buffers.depth_texture = rb->depth_texture; buffers.secondary_texture = rb->blur[0].mipmaps[0].texture; buffers.half_texture[0] = rb->blur[1].mipmaps[0].texture; @@ -2143,9 +2169,9 @@ void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(const Rende double step = env->auto_exp_speed * time_step; if (can_use_storage) { - storage->get_effects()->luminance_reduction(rb->texture, Size2i(rb->width, rb->height), rb->luminance.reduce, rb->luminance.current, env->min_luminance, env->max_luminance, step, set_immediate); + storage->get_effects()->luminance_reduction(rb->internal_texture, Size2i(rb->internal_width, rb->internal_height), rb->luminance.reduce, rb->luminance.current, env->min_luminance, env->max_luminance, step, set_immediate); } else { - storage->get_effects()->luminance_reduction_raster(rb->texture, Size2i(rb->width, rb->height), rb->luminance.reduce, rb->luminance.fb, rb->luminance.current, env->min_luminance, env->max_luminance, step, set_immediate); + storage->get_effects()->luminance_reduction_raster(rb->internal_texture, Size2i(rb->internal_width, rb->internal_height), rb->luminance.reduce, rb->luminance.fb, rb->luminance.current, env->min_luminance, env->max_luminance, step, set_immediate); } // Swap final reduce with prev luminance. SWAP(rb->luminance.current, rb->luminance.reduce.write[rb->luminance.reduce.size() - 1]); @@ -2188,9 +2214,9 @@ void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(const Rende luminance_texture = rb->luminance.current; } if (can_use_storage) { - storage->get_effects()->gaussian_glow(rb->texture, rb->blur[1].mipmaps[i].texture, Size2i(vp_w, vp_h), env->glow_strength, glow_high_quality, true, env->glow_hdr_luminance_cap, env->exposure, env->glow_bloom, env->glow_hdr_bleed_threshold, env->glow_hdr_bleed_scale, luminance_texture, env->auto_exp_scale); + storage->get_effects()->gaussian_glow(rb->internal_texture, rb->blur[1].mipmaps[i].texture, Size2i(vp_w, vp_h), env->glow_strength, glow_high_quality, true, env->glow_hdr_luminance_cap, env->exposure, env->glow_bloom, env->glow_hdr_bleed_threshold, env->glow_hdr_bleed_scale, luminance_texture, env->auto_exp_scale); } else { - storage->get_effects()->gaussian_glow_raster(rb->texture, rb->blur[1].mipmaps[i].half_fb, rb->blur[1].mipmaps[i].half_texture, rb->blur[1].mipmaps[i].fb, Size2i(vp_w, vp_h), env->glow_strength, glow_high_quality, true, env->glow_hdr_luminance_cap, env->exposure, env->glow_bloom, env->glow_hdr_bleed_threshold, env->glow_hdr_bleed_scale, luminance_texture, env->auto_exp_scale); + storage->get_effects()->gaussian_glow_raster(rb->internal_texture, rb->blur[1].mipmaps[i].half_fb, rb->blur[1].mipmaps[i].half_texture, rb->blur[1].mipmaps[i].fb, Size2i(vp_w, vp_h), env->glow_strength, glow_high_quality, true, env->glow_hdr_luminance_cap, env->exposure, env->glow_bloom, env->glow_hdr_bleed_threshold, env->glow_hdr_bleed_scale, luminance_texture, env->auto_exp_scale); } } else { if (can_use_storage) { @@ -2237,7 +2263,7 @@ void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(const Rende } tonemap.use_debanding = rb->use_debanding; - tonemap.texture_size = Vector2i(rb->width, rb->height); + tonemap.texture_size = Vector2i(rb->internal_width, rb->internal_height); if (env) { tonemap.tonemap_mode = env->tone_mapper; @@ -2268,7 +2294,15 @@ void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(const Rende tonemap.luminance_multiplier = _render_buffers_get_luminance_multiplier(); tonemap.view_count = p_render_data->view_count; - storage->get_effects()->tonemapper(rb->texture, storage->render_target_get_rd_framebuffer(rb->render_target), tonemap); + storage->get_effects()->tonemapper(rb->internal_texture, storage->render_target_get_rd_framebuffer(rb->render_target), tonemap); + + RD::get_singleton()->draw_command_end_label(); + } + + if (can_use_effects && can_use_storage && (rb->internal_width != rb->width || rb->internal_height != rb->height)) { + RD::get_singleton()->draw_command_begin_label("FSR Upscale"); + + storage->get_effects()->fsr_upscale(rb->internal_texture, rb->upscale_texture, rb->texture, Size2i(rb->internal_width, rb->internal_height), Size2i(rb->width, rb->height), rb->fsr_sharpness); RD::get_singleton()->draw_command_end_label(); } @@ -2628,14 +2662,28 @@ bool RendererSceneRenderRD::_render_buffers_can_be_storage() { return true; } -void RendererSceneRenderRD::render_buffers_configure(RID p_render_buffers, RID p_render_target, int p_width, int p_height, RS::ViewportMSAA p_msaa, RenderingServer::ViewportScreenSpaceAA p_screen_space_aa, bool p_use_debanding, uint32_t p_view_count) { +void RendererSceneRenderRD::render_buffers_configure(RID p_render_buffers, RID p_render_target, int p_internal_width, int p_internal_height, int p_width, int p_height, float p_fsr_sharpness, float p_fsr_mipmap_bias, RS::ViewportMSAA p_msaa, RenderingServer::ViewportScreenSpaceAA p_screen_space_aa, bool p_use_debanding, uint32_t p_view_count) { ERR_FAIL_COND_MSG(p_view_count == 0, "Must have at least 1 view"); + if (!_render_buffers_can_be_storage()) { + p_internal_height = p_height; + p_internal_width = p_width; + } + + if (p_width != p_internal_width) { + float fsr_mipmap_bias = -log2f(p_width / p_internal_width) + p_fsr_mipmap_bias; + storage->sampler_rd_configure_custom(fsr_mipmap_bias); + update_uniform_sets(); + } + RenderBuffers *rb = render_buffers_owner.get_or_null(p_render_buffers); // Should we add an overrule per viewport? + rb->internal_width = p_internal_width; + rb->internal_height = p_internal_height; rb->width = p_width; rb->height = p_height; + rb->fsr_sharpness = p_fsr_sharpness; rb->render_target = p_render_target; rb->msaa = p_msaa; rb->screen_space_aa = p_screen_space_aa; @@ -2657,8 +2705,8 @@ void RendererSceneRenderRD::render_buffers_configure(RID p_render_buffers, RID p tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; } tf.format = _render_buffers_get_color_format(); - tf.width = rb->width; - tf.height = rb->height; + tf.width = rb->internal_width; // If set to rb->width, msaa won't crash + tf.height = rb->internal_height; // If set to rb->width, msaa won't crash tf.array_layers = rb->view_count; // create a layer for every view tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | (_render_buffers_can_be_storage() ? RD::TEXTURE_USAGE_STORAGE_BIT : 0) | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; if (rb->msaa != RS::VIEWPORT_MSAA_DISABLED) { @@ -2666,7 +2714,17 @@ void RendererSceneRenderRD::render_buffers_configure(RID p_render_buffers, RID p } tf.usage_bits |= RD::TEXTURE_USAGE_INPUT_ATTACHMENT_BIT; // only needed when using subpasses in the mobile renderer - rb->texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); + rb->internal_texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); + + if ((p_internal_width != p_width || p_internal_height != p_height)) { + tf.width = rb->width; + tf.height = rb->height; + rb->texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); + rb->upscale_texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); + } else { + rb->texture = rb->internal_texture; + rb->upscale_texture = rb->internal_texture; + } } { @@ -2680,8 +2738,8 @@ void RendererSceneRenderRD::render_buffers_configure(RID p_render_buffers, RID p tf.format = RD::DATA_FORMAT_R32_SFLOAT; } - tf.width = rb->width; - tf.height = rb->height; + tf.width = rb->internal_width; + tf.height = rb->internal_height; tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT; tf.array_layers = rb->view_count; // create a layer for every view @@ -2697,16 +2755,16 @@ void RendererSceneRenderRD::render_buffers_configure(RID p_render_buffers, RID p if (!_render_buffers_can_be_storage()) { // ONLY USED ON MOBILE RENDERER, ONLY USED FOR POST EFFECTS! Vector<RID> fb; - fb.push_back(rb->texture); + fb.push_back(rb->internal_texture); rb->texture_fb = RD::get_singleton()->framebuffer_create(fb, RenderingDevice::INVALID_ID, rb->view_count); } RID target_texture = storage->render_target_get_rd_texture(rb->render_target); - rb->data->configure(rb->texture, rb->depth_texture, target_texture, rb->width, rb->height, p_msaa, p_view_count); + rb->data->configure(rb->internal_texture, rb->depth_texture, target_texture, p_internal_width, p_internal_height, p_msaa, p_view_count); if (is_clustered_enabled()) { - rb->cluster_builder->setup(Size2i(rb->width, rb->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); + rb->cluster_builder->setup(Size2i(p_internal_width, p_internal_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->internal_texture); } } @@ -3796,6 +3854,18 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e rb->volumetric_fog->fog_map = RD::get_singleton()->texture_create(tf, RD::TextureView()); RD::get_singleton()->set_resource_name(rb->volumetric_fog->fog_map, "Fog map"); +#if defined(OSX_ENABLED) || defined(IPHONE_ENABLED) + Vector<uint8_t> dm; + dm.resize(target_width * target_height * volumetric_fog_depth * 4); + dm.fill(0); + + rb->volumetric_fog->density_map = RD::get_singleton()->storage_buffer_create(dm.size(), dm); + RD::get_singleton()->set_resource_name(rb->volumetric_fog->density_map, "Fog density map"); + rb->volumetric_fog->light_map = RD::get_singleton()->storage_buffer_create(dm.size(), dm); + RD::get_singleton()->set_resource_name(rb->volumetric_fog->light_map, "Fog light map"); + rb->volumetric_fog->emissive_map = RD::get_singleton()->storage_buffer_create(dm.size(), dm); + RD::get_singleton()->set_resource_name(rb->volumetric_fog->emissive_map, "Fog emissive map"); +#else tf.format = RD::DATA_FORMAT_R32_UINT; tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; rb->volumetric_fog->density_map = RD::get_singleton()->texture_create(tf, RD::TextureView()); @@ -3807,6 +3877,7 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e rb->volumetric_fog->emissive_map = RD::get_singleton()->texture_create(tf, RD::TextureView()); RD::get_singleton()->set_resource_name(rb->volumetric_fog->emissive_map, "Fog emissive map"); RD::get_singleton()->texture_clear(rb->volumetric_fog->emissive_map, Color(0, 0, 0, 0), 0, 1, 0, 1); +#endif Vector<RD::Uniform> uniforms; { @@ -3872,7 +3943,11 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e { RD::Uniform u; +#if defined(OSX_ENABLED) || defined(IPHONE_ENABLED) + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; +#else u.uniform_type = RD::UNIFORM_TYPE_IMAGE; +#endif u.binding = 1; u.ids.push_back(rb->volumetric_fog->emissive_map); uniforms.push_back(u); @@ -3888,7 +3963,11 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e { RD::Uniform u; +#if defined(OSX_ENABLED) || defined(IPHONE_ENABLED) + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; +#else u.uniform_type = RD::UNIFORM_TYPE_IMAGE; +#endif u.binding = 3; u.ids.push_back(rb->volumetric_fog->density_map); uniforms.push_back(u); @@ -3896,7 +3975,11 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e { RD::Uniform u; +#if defined(OSX_ENABLED) || defined(IPHONE_ENABLED) + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; +#else u.uniform_type = RD::UNIFORM_TYPE_IMAGE; +#endif u.binding = 4; u.ids.push_back(rb->volumetric_fog->light_map); uniforms.push_back(u); @@ -4163,14 +4246,22 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e } { RD::Uniform u; +#if defined(OSX_ENABLED) || defined(IPHONE_ENABLED) + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; +#else u.uniform_type = RD::UNIFORM_TYPE_IMAGE; +#endif u.binding = 16; u.ids.push_back(rb->volumetric_fog->density_map); uniforms.push_back(u); } { RD::Uniform u; +#if defined(OSX_ENABLED) || defined(IPHONE_ENABLED) + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; +#else u.uniform_type = RD::UNIFORM_TYPE_IMAGE; +#endif u.binding = 17; u.ids.push_back(rb->volumetric_fog->light_map); uniforms.push_back(u); @@ -4178,7 +4269,11 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e { RD::Uniform u; +#if defined(OSX_ENABLED) || defined(IPHONE_ENABLED) + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; +#else u.uniform_type = RD::UNIFORM_TYPE_IMAGE; +#endif u.binding = 18; u.ids.push_back(rb->volumetric_fog->emissive_map); uniforms.push_back(u); @@ -4704,9 +4799,7 @@ void RendererSceneRenderRD::render_scene(RID p_render_buffers, const CameraData if (p_render_buffers.is_valid()) { /* _debug_draw_cluster(p_render_buffers); - RENDER_TIMESTAMP("Tonemap"); - _render_buffers_post_process_and_tonemap(&render_data); */ diff --git a/servers/rendering/renderer_rd/renderer_scene_render_rd.h b/servers/rendering/renderer_rd/renderer_scene_render_rd.h index 740e0e75ce..98ab1a2c3c 100644 --- a/servers/rendering/renderer_rd/renderer_scene_render_rd.h +++ b/servers/rendering/renderer_rd/renderer_scene_render_rd.h @@ -456,7 +456,11 @@ private: struct RenderBuffers { RenderBufferData *data = nullptr; - int width = 0, height = 0; + int internal_width = 0; + int internal_height = 0; + int width = 0; + int height = 0; + float fsr_sharpness = 0.2f; RS::ViewportMSAA msaa = RS::VIEWPORT_MSAA_DISABLED; RS::ViewportScreenSpaceAA screen_space_aa = RS::VIEWPORT_SCREEN_SPACE_AA_DISABLED; bool use_debanding = false; @@ -466,9 +470,12 @@ private: uint64_t auto_exposure_version = 1; - RID texture; //main texture for rendering to, must be filled after done rendering + RID sss_texture; //texture for sss. This needs to be a different resolution than blur[0] + RID internal_texture; //main texture for rendering to, must be filled after done rendering + RID texture; //upscaled version of main texture (This uses the same resource as internal_texture if there is no upscaling) RID depth_texture; //main depth texture RID texture_fb; // framebuffer for the main texture, ONLY USED FOR MOBILE RENDERER POST EFFECTS, DO NOT USE FOR RENDERING 3D!!! + RID upscale_texture; //used when upscaling internal_texture (This uses the same resource as internal_texture if there is no upscaling) RendererSceneGIRD::SDFGI *sdfgi = nullptr; VolumetricFog *volumetric_fog = nullptr; @@ -1332,7 +1339,7 @@ public: virtual RD::DataFormat _render_buffers_get_color_format(); virtual bool _render_buffers_can_be_storage(); virtual RID render_buffers_create() override; - virtual 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, uint32_t p_view_count) override; + virtual void render_buffers_configure(RID p_render_buffers, RID p_render_target, int p_internal_width, int p_internal_height, int p_width, int p_height, float p_fsr_sharpness, float p_fsr_mipmap_bias, RS::ViewportMSAA p_msaa, RS::ViewportScreenSpaceAA p_screen_space_aa, bool p_use_debanding, uint32_t p_view_count) override; virtual void gi_set_use_half_resolution(bool p_enable) override; RID render_buffers_get_depth_texture(RID p_render_buffers); @@ -1363,6 +1370,8 @@ public: float render_buffers_get_volumetric_fog_end(RID p_render_buffers); float render_buffers_get_volumetric_fog_detail_spread(RID p_render_buffers); + virtual void update_uniform_sets(){}; + virtual void render_scene(RID p_render_buffers, const CameraData *p_camera_data, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_voxel_gi_instances, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, const PagedArray<RID> &p_fog_volumes, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_occluder_debug_tex, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_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, RendererScene::RenderInfo *r_render_info = nullptr) override; virtual void render_material(const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) override; diff --git a/servers/rendering/renderer_rd/renderer_storage_rd.cpp b/servers/rendering/renderer_rd/renderer_storage_rd.cpp index 04753d7a9b..29e4a63cbb 100644 --- a/servers/rendering/renderer_rd/renderer_storage_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_storage_rd.cpp @@ -1227,6 +1227,100 @@ RendererStorageRD::CanvasTexture::~CanvasTexture() { clear_sets(); } +void RendererStorageRD::sampler_rd_configure_custom(float p_mipmap_bias) { + for (int i = 1; i < RS::CANVAS_ITEM_TEXTURE_FILTER_MAX; i++) { + for (int j = 1; j < RS::CANVAS_ITEM_TEXTURE_REPEAT_MAX; j++) { + RD::SamplerState sampler_state; + switch (i) { + case RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST: { + sampler_state.mag_filter = RD::SAMPLER_FILTER_NEAREST; + sampler_state.min_filter = RD::SAMPLER_FILTER_NEAREST; + sampler_state.max_lod = 0; + } break; + case RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR: { + sampler_state.mag_filter = RD::SAMPLER_FILTER_LINEAR; + sampler_state.min_filter = RD::SAMPLER_FILTER_LINEAR; + sampler_state.max_lod = 0; + } break; + case RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS: { + sampler_state.mag_filter = RD::SAMPLER_FILTER_NEAREST; + sampler_state.min_filter = RD::SAMPLER_FILTER_NEAREST; + if (GLOBAL_GET("rendering/textures/default_filters/use_nearest_mipmap_filter")) { + sampler_state.mip_filter = RD::SAMPLER_FILTER_NEAREST; + } else { + sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + } + sampler_state.lod_bias = p_mipmap_bias; + } break; + case RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS: { + sampler_state.mag_filter = RD::SAMPLER_FILTER_LINEAR; + sampler_state.min_filter = RD::SAMPLER_FILTER_LINEAR; + if (GLOBAL_GET("rendering/textures/default_filters/use_nearest_mipmap_filter")) { + sampler_state.mip_filter = RD::SAMPLER_FILTER_NEAREST; + } else { + sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + } + sampler_state.lod_bias = p_mipmap_bias; + + } break; + case RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC: { + sampler_state.mag_filter = RD::SAMPLER_FILTER_NEAREST; + sampler_state.min_filter = RD::SAMPLER_FILTER_NEAREST; + if (GLOBAL_GET("rendering/textures/default_filters/use_nearest_mipmap_filter")) { + sampler_state.mip_filter = RD::SAMPLER_FILTER_NEAREST; + } else { + sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + } + sampler_state.lod_bias = p_mipmap_bias; + sampler_state.use_anisotropy = true; + sampler_state.anisotropy_max = 1 << int(GLOBAL_GET("rendering/textures/default_filters/anisotropic_filtering_level")); + } break; + case RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC: { + sampler_state.mag_filter = RD::SAMPLER_FILTER_LINEAR; + sampler_state.min_filter = RD::SAMPLER_FILTER_LINEAR; + if (GLOBAL_GET("rendering/textures/default_filters/use_nearest_mipmap_filter")) { + sampler_state.mip_filter = RD::SAMPLER_FILTER_NEAREST; + } else { + sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + } + sampler_state.lod_bias = p_mipmap_bias; + sampler_state.use_anisotropy = true; + sampler_state.anisotropy_max = 1 << int(GLOBAL_GET("rendering/textures/default_filters/anisotropic_filtering_level")); + + } break; + default: { + } + } + switch (j) { + case RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED: { + sampler_state.repeat_u = RD::SAMPLER_REPEAT_MODE_CLAMP_TO_EDGE; + sampler_state.repeat_v = RD::SAMPLER_REPEAT_MODE_CLAMP_TO_EDGE; + sampler_state.repeat_w = RD::SAMPLER_REPEAT_MODE_CLAMP_TO_EDGE; + + } break; + case RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED: { + sampler_state.repeat_u = RD::SAMPLER_REPEAT_MODE_REPEAT; + sampler_state.repeat_v = RD::SAMPLER_REPEAT_MODE_REPEAT; + sampler_state.repeat_w = RD::SAMPLER_REPEAT_MODE_REPEAT; + } break; + case RS::CANVAS_ITEM_TEXTURE_REPEAT_MIRROR: { + sampler_state.repeat_u = RD::SAMPLER_REPEAT_MODE_MIRRORED_REPEAT; + sampler_state.repeat_v = RD::SAMPLER_REPEAT_MODE_MIRRORED_REPEAT; + sampler_state.repeat_w = RD::SAMPLER_REPEAT_MODE_MIRRORED_REPEAT; + } break; + default: { + } + } + + if (custom_rd_samplers[i][j].is_valid()) { + RD::get_singleton()->free(custom_rd_samplers[i][j]); + } + + custom_rd_samplers[i][j] = RD::get_singleton()->sampler_create(sampler_state); + } + } +} + RID RendererStorageRD::canvas_texture_allocate() { return canvas_texture_owner.allocate_rid(); } @@ -2558,7 +2652,17 @@ void RendererStorageRD::MaterialData::update_uniform_buffer(const Map<StringName //regular uniform uint32_t offset = p_uniform_offsets[E.value.order]; #ifdef DEBUG_ENABLED - uint32_t size = ShaderLanguage::get_type_size(E.value.type); + uint32_t size = 0U; + // The following code enforces a 16-byte alignment of uniform arrays. + if (E.value.array_size > 0) { + size = ShaderLanguage::get_type_size(E.value.type) * E.value.array_size; + int m = (16 * E.value.array_size); + if ((size % m) != 0U) { + size += m - (size % m); + } + } else { + size = ShaderLanguage::get_type_size(E.value.type); + } ERR_CONTINUE(offset + size > p_buffer_size); #endif uint8_t *data = &p_buffer[offset]; @@ -9774,6 +9878,9 @@ RendererStorageRD::RendererStorageRD() { } } + //custom sampler + sampler_rd_configure_custom(0.0f); + //default rd buffers { Vector<uint8_t> buffer; @@ -9926,7 +10033,7 @@ RendererStorageRD::RendererStorageRD() { actions.renames["RESTART"] = "restart"; actions.renames["CUSTOM"] = "PARTICLE.custom"; actions.renames["TRANSFORM"] = "PARTICLE.xform"; - actions.renames["TIME"] = "FRAME.time"; + actions.renames["TIME"] = "frame_history.data[0].time"; actions.renames["PI"] = _MKSTR(Math_PI); actions.renames["TAU"] = _MKSTR(Math_TAU); actions.renames["E"] = _MKSTR(Math_E); @@ -10104,6 +10211,15 @@ RendererStorageRD::~RendererStorageRD() { } } + //custom samplers + for (int i = 1; i < RS::CANVAS_ITEM_TEXTURE_FILTER_MAX; i++) { + for (int j = 0; j < RS::CANVAS_ITEM_TEXTURE_REPEAT_MAX; j++) { + if (custom_rd_samplers[i][j].is_valid()) { + RD::get_singleton()->free(custom_rd_samplers[i][j]); + } + } + } + //def buffers for (int i = 0; i < DEFAULT_RD_BUFFER_MAX; i++) { RD::get_singleton()->free(mesh_default_rd_buffers[i]); diff --git a/servers/rendering/renderer_rd/renderer_storage_rd.h b/servers/rendering/renderer_rd/renderer_storage_rd.h index 2cd3a01c66..9a64480c3e 100644 --- a/servers/rendering/renderer_rd/renderer_storage_rd.h +++ b/servers/rendering/renderer_rd/renderer_storage_rd.h @@ -320,6 +320,7 @@ private: RID default_rd_textures[DEFAULT_RD_TEXTURE_MAX]; RID default_rd_samplers[RS::CANVAS_ITEM_TEXTURE_FILTER_MAX][RS::CANVAS_ITEM_TEXTURE_REPEAT_MAX]; + RID custom_rd_samplers[RS::CANVAS_ITEM_TEXTURE_FILTER_MAX][RS::CANVAS_ITEM_TEXTURE_REPEAT_MAX]; RID default_rd_storage_buffer; /* DECAL ATLAS */ @@ -1391,6 +1392,13 @@ public: _FORCE_INLINE_ RID sampler_rd_get_default(RS::CanvasItemTextureFilter p_filter, RS::CanvasItemTextureRepeat p_repeat) { return default_rd_samplers[p_filter][p_repeat]; } + _FORCE_INLINE_ RID sampler_rd_get_custom(RS::CanvasItemTextureFilter p_filter, RS::CanvasItemTextureRepeat p_repeat) { + return custom_rd_samplers[p_filter][p_repeat]; + } + + void sampler_rd_configure_custom(float mipmap_bias); + + void sampler_rd_set_default(float p_mipmap_bias); /* CANVAS TEXTURE API */ diff --git a/servers/rendering/renderer_rd/shader_compiler_rd.cpp b/servers/rendering/renderer_rd/shader_compiler_rd.cpp index 215959bb6a..77d3a2e766 100644 --- a/servers/rendering/renderer_rd/shader_compiler_rd.cpp +++ b/servers/rendering/renderer_rd/shader_compiler_rd.cpp @@ -227,6 +227,13 @@ static String _prestr(SL::DataPrecision p_pres, bool p_force_highp = false) { return ""; } +static String _constr(bool p_is_const) { + if (p_is_const) { + return "const "; + } + return ""; +} + static String _qualstr(SL::ArgumentQualifier p_qual) { switch (p_qual) { case SL::ARGUMENT_QUALIFIER_IN: @@ -417,9 +424,7 @@ void ShaderCompilerRD::_dump_function_deps(const SL::ShaderNode *p_node, const S if (i > 0) { header += ", "; } - if (fnode->arguments[i].is_const) { - header += "const "; - } + header += _constr(fnode->arguments[i].is_const); if (fnode->arguments[i].type == SL::TYPE_STRUCT) { header += _qualstr(fnode->arguments[i].qualifier) + _mkid(fnode->arguments[i].type_str) + " " + _mkid(fnode->arguments[i].name); } else { @@ -656,6 +661,7 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge uniform_sizes.write[uniform.order] = _get_datatype_size(ShaderLanguage::TYPE_UINT); uniform_alignments.write[uniform.order] = _get_datatype_alignment(ShaderLanguage::TYPE_UINT); } else { + // The following code enforces a 16-byte alignment of uniform arrays. if (uniform.array_size > 0) { int size = _get_datatype_size(uniform.type) * uniform.array_size; int m = (16 * uniform.array_size); @@ -663,10 +669,11 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge size += m - (size % m); } uniform_sizes.write[uniform.order] = size; + uniform_alignments.write[uniform.order] = 16; } else { uniform_sizes.write[uniform.order] = _get_datatype_size(uniform.type); + uniform_alignments.write[uniform.order] = _get_datatype_alignment(uniform.type); } - uniform_alignments.write[uniform.order] = _get_datatype_alignment(uniform.type); } } @@ -791,7 +798,7 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge for (int i = 0; i < pnode->vconstants.size(); i++) { const SL::ShaderNode::Constant &cnode = pnode->vconstants[i]; String gcode; - gcode += "const "; + gcode += _constr(true); gcode += _prestr(cnode.precision, ShaderLanguage::is_float_type(cnode.type)); if (cnode.type == SL::TYPE_STRUCT) { gcode += _mkid(cnode.type_str); @@ -875,9 +882,7 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge SL::VariableDeclarationNode *vdnode = (SL::VariableDeclarationNode *)p_node; String declaration; - if (vdnode->is_const) { - declaration += "const "; - } + declaration += _constr(vdnode->is_const); if (vdnode->datatype == SL::TYPE_STRUCT) { declaration += _mkid(vdnode->struct_name); } else { @@ -997,9 +1002,7 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge case SL::Node::TYPE_ARRAY_DECLARATION: { SL::ArrayDeclarationNode *adnode = (SL::ArrayDeclarationNode *)p_node; String declaration; - if (adnode->is_const) { - declaration += "const "; - } + declaration += _constr(adnode->is_const); if (adnode->datatype == SL::TYPE_STRUCT) { declaration += _mkid(adnode->struct_name); } else { diff --git a/servers/rendering/renderer_rd/shader_rd.cpp b/servers/rendering/renderer_rd/shader_rd.cpp index b9a8947fa2..2568090918 100644 --- a/servers/rendering/renderer_rd/shader_rd.cpp +++ b/servers/rendering/renderer_rd/shader_rd.cpp @@ -177,6 +177,9 @@ void ShaderRD::_build_variant_code(StringBuilder &builder, uint32_t p_variant, c for (const KeyValue<StringName, CharString> &E : p_version->code_sections) { builder.append(String("#define ") + String(E.key) + "_CODE_USED\n"); } +#if defined(OSX_ENABLED) || defined(IPHONE_ENABLED) + builder.append("#define MOLTENVK_USED\n"); +#endif } break; case StageTemplate::Chunk::TYPE_MATERIAL_UNIFORMS: { builder.append(p_version->uniforms.get_data()); //uniforms (same for vertex and fragment) diff --git a/servers/rendering/renderer_rd/shaders/fsr_upscale.glsl b/servers/rendering/renderer_rd/shaders/fsr_upscale.glsl new file mode 100644 index 0000000000..4e2ba84033 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/fsr_upscale.glsl @@ -0,0 +1,173 @@ +/*************************************************************************/ +/* fsr_upscale.glsl */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2021 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2021 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#[compute] + +#version 450 + +#VERSION_DEFINES + +#define A_GPU +#define A_GLSL + +#ifdef MODE_FSR_UPSCALE_NORMAL + +#define A_HALF + +#endif + +#include "thirdparty/amd-fsr/ffx_a.h" + +layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in; + +layout(rgba16f, set = 1, binding = 0) uniform restrict writeonly image2D fsr_image; +layout(set = 0, binding = 0) uniform sampler2D source_image; + +#define FSR_UPSCALE_PASS_TYPE_EASU 0 +#define FSR_UPSCALE_PASS_TYPE_RCAS 1 + +layout(push_constant, binding = 1, std430) uniform Params { + float resolution_width; + float resolution_height; + float upscaled_width; + float upscaled_height; + float sharpness; + int pass; +} +params; + +AU4 Const0, Const1, Const2, Const3; + +#ifdef MODE_FSR_UPSCALE_FALLBACK + +#define FSR_EASU_F +AF4 FsrEasuRF(AF2 p) { + AF4 res = textureGather(source_image, p, 0); + return res; +} +AF4 FsrEasuGF(AF2 p) { + AF4 res = textureGather(source_image, p, 1); + return res; +} +AF4 FsrEasuBF(AF2 p) { + AF4 res = textureGather(source_image, p, 2); + return res; +} + +#define FSR_RCAS_F +AF4 FsrRcasLoadF(ASU2 p) { + return AF4(texelFetch(source_image, ASU2(p), 0)); +} +void FsrRcasInputF(inout AF1 r, inout AF1 g, inout AF1 b) {} + +#else + +#define FSR_EASU_H +AH4 FsrEasuRH(AF2 p) { + AH4 res = AH4(textureGather(source_image, p, 0)); + return res; +} +AH4 FsrEasuGH(AF2 p) { + AH4 res = AH4(textureGather(source_image, p, 1)); + return res; +} +AH4 FsrEasuBH(AF2 p) { + AH4 res = AH4(textureGather(source_image, p, 2)); + return res; +} + +#define FSR_RCAS_H +AH4 FsrRcasLoadH(ASW2 p) { + return AH4(texelFetch(source_image, ASU2(p), 0)); +} +void FsrRcasInputH(inout AH1 r, inout AH1 g, inout AH1 b) {} + +#endif + +#include "thirdparty/amd-fsr/ffx_fsr1.h" + +void fsr_easu_pass(AU2 pos) { +#ifdef MODE_FSR_UPSCALE_NORMAL + + AH3 Gamma2Color = AH3(0, 0, 0); + FsrEasuH(Gamma2Color, pos, Const0, Const1, Const2, Const3); + imageStore(fsr_image, ASU2(pos), AH4(Gamma2Color, 1)); + +#else + + AF3 Gamma2Color = AF3(0, 0, 0); + FsrEasuF(Gamma2Color, pos, Const0, Const1, Const2, Const3); + imageStore(fsr_image, ASU2(pos), AF4(Gamma2Color, 1)); + +#endif +} + +void fsr_rcas_pass(AU2 pos) { +#ifdef MODE_FSR_UPSCALE_NORMAL + + AH3 Gamma2Color = AH3(0, 0, 0); + FsrRcasH(Gamma2Color.r, Gamma2Color.g, Gamma2Color.b, pos, Const0); + imageStore(fsr_image, ASU2(pos), AH4(Gamma2Color, 1)); + +#else + + AF3 Gamma2Color = AF3(0, 0, 0); + FsrRcasF(Gamma2Color.r, Gamma2Color.g, Gamma2Color.b, pos, Const0); + imageStore(fsr_image, ASU2(pos), AF4(Gamma2Color, 1)); + +#endif +} + +void fsr_pass(AU2 pos) { + if (params.pass == FSR_UPSCALE_PASS_TYPE_EASU) { + fsr_easu_pass(pos); + } else if (params.pass == FSR_UPSCALE_PASS_TYPE_RCAS) { + fsr_rcas_pass(pos); + } +} + +void main() { + // Clang does not like unused functions. If ffx_a.h is included in the binary, clang will throw a fit and not compile so we must configure FSR in this shader + if (params.pass == FSR_UPSCALE_PASS_TYPE_EASU) { + FsrEasuCon(Const0, Const1, Const2, Const3, params.resolution_width, params.resolution_height, params.resolution_width, params.resolution_height, params.upscaled_width, params.upscaled_height); + } else if (params.pass == FSR_UPSCALE_PASS_TYPE_RCAS) { + FsrRcasCon(Const0, params.sharpness); + } + + AU2 gxy = ARmp8x8(gl_LocalInvocationID.x) + AU2(gl_WorkGroupID.x << 4u, gl_WorkGroupID.y << 4u); + + fsr_pass(gxy); + gxy.x += 8u; + fsr_pass(gxy); + gxy.y += 8u; + fsr_pass(gxy); + gxy.x -= 8u; + fsr_pass(gxy); +} diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl index 20982a466c..e4628b2d5a 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl @@ -695,7 +695,7 @@ void main() { #endif // ALPHA_ANTIALIASING_EDGE_USED #ifdef USE_OPAQUE_PREPASS - if (alpha < opaque_prepass_threshold) { + if (alpha < scene_data.opaque_prepass_threshold) { discard; } #endif // USE_OPAQUE_PREPASS @@ -1711,7 +1711,7 @@ void main() { #ifdef USE_OPAQUE_PREPASS - if (alpha < opaque_prepass_threshold) { + if (alpha < scene_data.opaque_prepass_threshold) { discard; } @@ -1757,7 +1757,11 @@ void main() { } } +#ifdef MOLTENVK_USED + imageStore(geom_facing_grid, grid_pos, uvec4(imageLoad(geom_facing_grid, grid_pos).r | facing_bits)); //store facing bits +#else imageAtomicOr(geom_facing_grid, grid_pos, facing_bits); //store facing bits +#endif if (length(emission) > 0.001) { float lumas[6]; diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl index c8489f2137..be29cf4f58 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl @@ -208,7 +208,8 @@ layout(set = 1, binding = 0, std140) uniform SceneData { float roughness_limiter_amount; float roughness_limiter_limit; - uvec2 roughness_limiter_pad; + float opaque_prepass_threshold; + uint roughness_limiter_pad; mat4 sdf_to_bounds; diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl index b6fd89d912..e92fbecfd0 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl @@ -705,7 +705,7 @@ void main() { #endif // ALPHA_ANTIALIASING_EDGE_USED #ifdef USE_OPAQUE_PREPASS - if (alpha < opaque_prepass_threshold) { + if (alpha < scene_data.opaque_prepass_threshold) { discard; } #endif // USE_OPAQUE_PREPASS @@ -1476,7 +1476,7 @@ void main() { #ifdef USE_OPAQUE_PREPASS - if (alpha < opaque_prepass_threshold) { + if (alpha < scene_data.opaque_prepass_threshold) { discard; } diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl index eb8fb49598..a9a4fce82a 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl @@ -168,7 +168,8 @@ layout(set = 1, binding = 0, std140) uniform SceneData { mediump float roughness_limiter_amount; mediump float roughness_limiter_limit; - uvec2 roughness_limiter_pad; + mediump float opaque_prepass_threshold; + uint roughness_limiter_pad; bool fog_enabled; highp float fog_density; diff --git a/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl b/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl index afc5d68776..181d3b272f 100644 --- a/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl +++ b/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl @@ -47,7 +47,13 @@ layout(push_constant, binding = 0, std430) uniform Params { } params; +#ifdef MOLTENVK_USED +layout(set = 1, binding = 1) volatile buffer emissive_only_map_buffer { + uint emissive_only_map[]; +}; +#else layout(r32ui, set = 1, binding = 1) uniform volatile uimage3D emissive_only_map; +#endif layout(set = 1, binding = 2, std140) uniform SceneParams { vec2 fog_frustum_size_begin; @@ -71,8 +77,17 @@ layout(set = 1, binding = 2, std140) uniform SceneParams { } scene_params; +#ifdef MOLTENVK_USED +layout(set = 1, binding = 3) volatile buffer density_only_map_buffer { + uint density_only_map[]; +}; +layout(set = 1, binding = 4) volatile buffer light_only_map_buffer { + uint light_only_map[]; +}; +#else layout(r32ui, set = 1, binding = 3) uniform volatile uimage3D density_only_map; layout(r32ui, set = 1, binding = 4) uniform volatile uimage3D light_only_map; +#endif #ifdef MATERIAL_UNIFORMS_USED layout(set = 2, binding = 0, std140) uniform MaterialUniforms{ @@ -115,6 +130,9 @@ void main() { if (any(greaterThanEqual(pos, scene_params.fog_volume_size))) { return; //do not compute } +#ifdef MOLTENVK_USED + uint lpos = pos.z * scene_params.fog_volume_size.x * scene_params.fog_volume_size.y + pos.y * scene_params.fog_volume_size.x + pos.x; +#endif vec3 posf = vec3(pos); @@ -197,7 +215,11 @@ void main() { density *= cull_mask; if (abs(density) > 0.001) { int final_density = int(density * DENSITY_SCALE); +#ifdef MOLTENVK_USED + atomicAdd(density_only_map[lpos], uint(final_density)); +#else imageAtomicAdd(density_only_map, pos, uint(final_density)); +#endif #ifdef EMISSION_USED { @@ -207,7 +229,11 @@ void main() { uvec3 emission_u = uvec3(emission.r * 511.0, emission.g * 511.0, emission.b * 255.0); // R and G have 11 bits each and B has 10. Then pack them into a 32 bit uint uint final_emission = emission_u.r << 21 | emission_u.g << 10 | emission_u.b; +#ifdef MOLTENVK_USED + uint prev_emission = atomicAdd(emissive_only_map[lpos], final_emission); +#else uint prev_emission = imageAtomicAdd(emissive_only_map, pos, final_emission); +#endif // Adding can lead to colors overflowing, so validate uvec3 prev_emission_u = uvec3(prev_emission >> 21, (prev_emission << 11) >> 21, prev_emission % 1024); @@ -219,7 +245,11 @@ void main() { if (any(overflowing)) { uvec3 overflow_factor = mix(uvec3(0), uvec3(2047 << 21, 2047 << 10, 1023), overflowing); uint force_max = overflow_factor.r | overflow_factor.g | overflow_factor.b; +#ifdef MOLTENVK_USED + atomicOr(emissive_only_map[lpos], force_max); +#else imageAtomicOr(emissive_only_map, pos, force_max); +#endif } } #endif @@ -230,7 +260,11 @@ void main() { uvec3 scattering_u = uvec3(scattering.r * 2047.0, scattering.g * 2047.0, scattering.b * 1023.0); // R and G have 11 bits each and B has 10. Then pack them into a 32 bit uint uint final_scattering = scattering_u.r << 21 | scattering_u.g << 10 | scattering_u.b; +#ifdef MOLTENVK_USED + uint prev_scattering = atomicAdd(light_only_map[lpos], final_scattering); +#else uint prev_scattering = imageAtomicAdd(light_only_map, pos, final_scattering); +#endif // Adding can lead to colors overflowing, so validate uvec3 prev_scattering_u = uvec3(prev_scattering >> 21, (prev_scattering << 11) >> 21, prev_scattering % 1024); @@ -242,7 +276,11 @@ void main() { if (any(overflowing)) { uvec3 overflow_factor = mix(uvec3(0), uvec3(2047 << 21, 2047 << 10, 1023), overflowing); uint force_max = overflow_factor.r | overflow_factor.g | overflow_factor.b; +#ifdef MOLTENVK_USED + atomicOr(light_only_map[lpos], force_max); +#else imageAtomicOr(light_only_map, pos, force_max); +#endif } } #endif // ALBEDO_USED diff --git a/servers/rendering/renderer_rd/shaders/volumetric_fog_process.glsl b/servers/rendering/renderer_rd/shaders/volumetric_fog_process.glsl index 3d6fbb5653..747f88960c 100644 --- a/servers/rendering/renderer_rd/shaders/volumetric_fog_process.glsl +++ b/servers/rendering/renderer_rd/shaders/volumetric_fog_process.glsl @@ -190,9 +190,22 @@ params; #ifndef MODE_COPY layout(set = 0, binding = 15) uniform texture3D prev_density_texture; +#ifdef MOLTENVK_USED +layout(set = 0, binding = 16) buffer density_only_map_buffer { + uint density_only_map[]; +}; +layout(set = 0, binding = 17) buffer light_only_map_buffer { + uint light_only_map[]; +}; +layout(set = 0, binding = 18) buffer emissive_only_map_buffer { + uint emissive_only_map[]; +}; +#else layout(r32ui, set = 0, binding = 16) uniform uimage3D density_only_map; layout(r32ui, set = 0, binding = 17) uniform uimage3D light_only_map; layout(r32ui, set = 0, binding = 18) uniform uimage3D emissive_only_map; +#endif + #ifdef USE_RADIANCE_CUBEMAP_ARRAY layout(set = 0, binding = 19) uniform textureCubeArray sky_texture; #else @@ -272,6 +285,9 @@ void main() { if (any(greaterThanEqual(pos, params.fog_volume_size))) { return; //do not compute } +#ifdef MOLTENVK_USED + uint lpos = pos.z * params.fog_volume_size.x * params.fog_volume_size.y + pos.y * params.fog_volume_size.x + pos.x; +#endif vec3 posf = vec3(pos); @@ -335,15 +351,28 @@ void main() { vec3 total_light = vec3(0.0); float total_density = params.base_density; +#ifdef MOLTENVK_USED + uint local_density = density_only_map[lpos]; +#else uint local_density = imageLoad(density_only_map, pos).x; +#endif + total_density += float(int(local_density)) / DENSITY_SCALE; total_density = max(0.0, total_density); +#ifdef MOLTENVK_USED + uint scattering_u = light_only_map[lpos]; +#else uint scattering_u = imageLoad(light_only_map, pos).x; +#endif vec3 scattering = vec3(scattering_u >> 21, (scattering_u << 11) >> 21, scattering_u % 1024) / vec3(2047.0, 2047.0, 1023.0); scattering += params.base_scattering * params.base_density; +#ifdef MOLTENVK_USED + uint emission_u = emissive_only_map[lpos]; +#else uint emission_u = imageLoad(emissive_only_map, pos).x; +#endif vec3 emission = vec3(emission_u >> 21, (emission_u << 11) >> 21, emission_u % 1024) / vec3(511.0, 511.0, 255.0); emission += params.base_emission * params.base_density; @@ -673,10 +702,16 @@ void main() { final_density = mix(final_density, reprojected_density, reproject_amount); imageStore(density_map, pos, final_density); +#ifdef MOLTENVK_USED + density_only_map[lpos] = 0; + light_only_map[lpos] = 0; + emissive_only_map[lpos] = 0; +#else imageStore(density_only_map, pos, uvec4(0)); imageStore(light_only_map, pos, uvec4(0)); imageStore(emissive_only_map, pos, uvec4(0)); #endif +#endif #ifdef MODE_FOG |