diff options
Diffstat (limited to 'servers/rendering/renderer_rd/effects')
| -rw-r--r-- | servers/rendering/renderer_rd/effects/bokeh_dof.cpp | 94 | ||||
| -rw-r--r-- | servers/rendering/renderer_rd/effects/bokeh_dof.h | 11 | ||||
| -rw-r--r-- | servers/rendering/renderer_rd/effects/copy_effects.cpp | 137 | ||||
| -rw-r--r-- | servers/rendering/renderer_rd/effects/copy_effects.h | 46 | ||||
| -rw-r--r-- | servers/rendering/renderer_rd/effects/fsr.cpp | 128 | ||||
| -rw-r--r-- | servers/rendering/renderer_rd/effects/fsr.h | 72 | ||||
| -rw-r--r-- | servers/rendering/renderer_rd/effects/resolve.h | 2 | ||||
| -rw-r--r-- | servers/rendering/renderer_rd/effects/ss_effects.cpp | 1876 | ||||
| -rw-r--r-- | servers/rendering/renderer_rd/effects/ss_effects.h | 552 | ||||
| -rw-r--r-- | servers/rendering/renderer_rd/effects/taa.cpp | 138 | ||||
| -rw-r--r-- | servers/rendering/renderer_rd/effects/taa.h | 68 | ||||
| -rw-r--r-- | servers/rendering/renderer_rd/effects/tone_mapper.cpp | 4 | ||||
| -rw-r--r-- | servers/rendering/renderer_rd/effects/tone_mapper.h | 6 | ||||
| -rw-r--r-- | servers/rendering/renderer_rd/effects/vrs.cpp | 58 | ||||
| -rw-r--r-- | servers/rendering/renderer_rd/effects/vrs.h | 4 |
15 files changed, 3092 insertions, 104 deletions
diff --git a/servers/rendering/renderer_rd/effects/bokeh_dof.cpp b/servers/rendering/renderer_rd/effects/bokeh_dof.cpp index cc7441776d..27850695b0 100644 --- a/servers/rendering/renderer_rd/effects/bokeh_dof.cpp +++ b/servers/rendering/renderer_rd/effects/bokeh_dof.cpp @@ -33,6 +33,8 @@ #include "servers/rendering/renderer_rd/renderer_compositor_rd.h" #include "servers/rendering/renderer_rd/storage_rd/material_storage.h" #include "servers/rendering/renderer_rd/uniform_set_cache_rd.h" +#include "servers/rendering/rendering_server_default.h" +#include "servers/rendering/storage/camera_attributes_storage.h" using namespace RendererRD; @@ -84,7 +86,7 @@ BokehDOF::~BokehDOF() { } } -void BokehDOF::bokeh_dof_compute(const BokehBuffers &p_buffers, 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, RenderingServer::DOFBokehShape p_bokeh_shape, RS::DOFBlurQuality p_quality, bool p_use_jitter, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal) { +void BokehDOF::bokeh_dof_compute(const BokehBuffers &p_buffers, RID p_camera_attributes, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal) { ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use compute version of bokeh depth of field with the mobile renderer."); UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); @@ -92,22 +94,39 @@ void BokehDOF::bokeh_dof_compute(const BokehBuffers &p_buffers, bool p_dof_far, MaterialStorage *material_storage = MaterialStorage::get_singleton(); ERR_FAIL_NULL(material_storage); + bool dof_far = RSG::camera_attributes->camera_attributes_get_dof_far_enabled(p_camera_attributes); + float dof_far_begin = RSG::camera_attributes->camera_attributes_get_dof_far_distance(p_camera_attributes); + float dof_far_size = RSG::camera_attributes->camera_attributes_get_dof_far_transition(p_camera_attributes); + bool dof_near = RSG::camera_attributes->camera_attributes_get_dof_near_enabled(p_camera_attributes); + float dof_near_begin = RSG::camera_attributes->camera_attributes_get_dof_near_distance(p_camera_attributes); + float dof_near_size = RSG::camera_attributes->camera_attributes_get_dof_near_transition(p_camera_attributes); + float bokeh_size = RSG::camera_attributes->camera_attributes_get_dof_blur_amount(p_camera_attributes) * 64; // Base 64 pixel radius. + + bool use_jitter = RSG::camera_attributes->camera_attributes_get_dof_blur_use_jitter(); + RS::DOFBokehShape bokeh_shape = RSG::camera_attributes->camera_attributes_get_dof_blur_bokeh_shape(); + RS::DOFBlurQuality blur_quality = RSG::camera_attributes->camera_attributes_get_dof_blur_quality(); + // setup our push constant memset(&bokeh.push_constant, 0, sizeof(BokehPushConstant)); - bokeh.push_constant.blur_far_active = p_dof_far; - bokeh.push_constant.blur_far_begin = p_dof_far_begin; - bokeh.push_constant.blur_far_end = p_dof_far_begin + p_dof_far_size; - - bokeh.push_constant.blur_near_active = p_dof_near; - bokeh.push_constant.blur_near_begin = p_dof_near_begin; - bokeh.push_constant.blur_near_end = MAX(0, p_dof_near_begin - p_dof_near_size); - bokeh.push_constant.use_jitter = p_use_jitter; + bokeh.push_constant.blur_far_active = dof_far; + bokeh.push_constant.blur_far_begin = dof_far_begin; + bokeh.push_constant.blur_far_end = dof_far_begin + dof_far_size; // Only used with non-physically-based. + bokeh.push_constant.use_physical_far = dof_far_size < 0.0; + bokeh.push_constant.blur_size_far = bokeh_size; // Only used with physically-based. + + bokeh.push_constant.blur_near_active = dof_near; + bokeh.push_constant.blur_near_begin = dof_near_begin; + bokeh.push_constant.blur_near_end = dof_near_begin - dof_near_size; // Only used with non-physically-based. + bokeh.push_constant.use_physical_near = dof_near_size < 0.0; + bokeh.push_constant.blur_size_near = bokeh_size; // Only used with physically-based. + + bokeh.push_constant.use_jitter = use_jitter; bokeh.push_constant.jitter_seed = Math::randf() * 1000.0; bokeh.push_constant.z_near = p_cam_znear; bokeh.push_constant.z_far = p_cam_zfar; bokeh.push_constant.orthogonal = p_cam_orthogonal; - bokeh.push_constant.blur_size = p_bokeh_size; + bokeh.push_constant.blur_size = (dof_near_size < 0.0 && dof_far_size < 0.0) ? 32 : bokeh_size; // Cap with physically-based to keep performance reasonable. bokeh.push_constant.second_pass = false; bokeh.push_constant.half_size = false; @@ -150,9 +169,9 @@ void BokehDOF::bokeh_dof_compute(const BokehBuffers &p_buffers, bool p_dof_far, RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_buffers.base_texture_size.x, p_buffers.base_texture_size.y, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); - if (p_bokeh_shape == RS::DOF_BOKEH_BOX || p_bokeh_shape == RS::DOF_BOKEH_HEXAGON) { + if (bokeh_shape == RS::DOF_BOKEH_BOX || bokeh_shape == RS::DOF_BOKEH_HEXAGON) { //second pass - BokehMode mode = p_bokeh_shape == RS::DOF_BOKEH_BOX ? BOKEH_GEN_BOKEH_BOX : BOKEH_GEN_BOKEH_HEXAGONAL; + BokehMode mode = bokeh_shape == RS::DOF_BOKEH_BOX ? BOKEH_GEN_BOKEH_BOX : BOKEH_GEN_BOKEH_HEXAGONAL; shader = bokeh.compute_shader.version_get_shader(bokeh.shader_version, mode); ERR_FAIL_COND(shader.is_null()); @@ -160,9 +179,9 @@ void BokehDOF::bokeh_dof_compute(const BokehBuffers &p_buffers, bool p_dof_far, static const int quality_samples[4] = { 6, 12, 12, 24 }; - bokeh.push_constant.steps = quality_samples[p_quality]; + bokeh.push_constant.steps = quality_samples[blur_quality]; - if (p_quality == RS::DOF_BLUR_QUALITY_VERY_LOW || p_quality == RS::DOF_BLUR_QUALITY_LOW) { + if (blur_quality == RS::DOF_BLUR_QUALITY_VERY_LOW || blur_quality == RS::DOF_BLUR_QUALITY_LOW) { //box and hexagon are more or less the same, and they can work in either half (very low and low quality) or full (medium and high quality_ sizes) RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_half_image0), 0); @@ -187,7 +206,7 @@ void BokehDOF::bokeh_dof_compute(const BokehBuffers &p_buffers, bool p_dof_far, //third pass bokeh.push_constant.second_pass = true; - if (p_quality == RS::DOF_BLUR_QUALITY_VERY_LOW || p_quality == RS::DOF_BLUR_QUALITY_LOW) { + if (blur_quality == RS::DOF_BLUR_QUALITY_VERY_LOW || blur_quality == RS::DOF_BLUR_QUALITY_LOW) { RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_half_image1), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_half_texture0), 1); } else { @@ -200,7 +219,7 @@ void BokehDOF::bokeh_dof_compute(const BokehBuffers &p_buffers, bool p_dof_far, 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) { + if (blur_quality == RS::DOF_BLUR_QUALITY_VERY_LOW || blur_quality == RS::DOF_BLUR_QUALITY_LOW) { //forth pass, upscale for low quality shader = bokeh.compute_shader.version_get_shader(bokeh.shader_version, BOKEH_COMPOSITE); @@ -232,7 +251,7 @@ void BokehDOF::bokeh_dof_compute(const BokehBuffers &p_buffers, bool p_dof_far, static const float quality_scale[4] = { 8.0, 4.0, 1.0, 0.5 }; bokeh.push_constant.steps = 0; - bokeh.push_constant.blur_scale = quality_scale[p_quality]; + bokeh.push_constant.blur_scale = quality_scale[blur_quality]; //circle always runs in half size, otherwise too expensive @@ -273,7 +292,7 @@ void BokehDOF::bokeh_dof_compute(const BokehBuffers &p_buffers, bool p_dof_far, RD::get_singleton()->compute_list_end(); } -void BokehDOF::bokeh_dof_raster(const BokehBuffers &p_buffers, 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_dof_blur_amount, RenderingServer::DOFBokehShape p_bokeh_shape, RS::DOFBlurQuality p_quality, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal) { +void BokehDOF::bokeh_dof_raster(const BokehBuffers &p_buffers, RID p_camera_attributes, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal) { ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't blur-based depth of field with the clustered renderer."); UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); @@ -281,6 +300,17 @@ void BokehDOF::bokeh_dof_raster(const BokehBuffers &p_buffers, bool p_dof_far, f MaterialStorage *material_storage = MaterialStorage::get_singleton(); ERR_FAIL_NULL(material_storage); + bool dof_far = RSG::camera_attributes->camera_attributes_get_dof_far_enabled(p_camera_attributes); + float dof_far_begin = RSG::camera_attributes->camera_attributes_get_dof_far_distance(p_camera_attributes); + float dof_far_size = RSG::camera_attributes->camera_attributes_get_dof_far_transition(p_camera_attributes); + bool dof_near = RSG::camera_attributes->camera_attributes_get_dof_near_enabled(p_camera_attributes); + float dof_near_begin = RSG::camera_attributes->camera_attributes_get_dof_near_distance(p_camera_attributes); + float dof_near_size = RSG::camera_attributes->camera_attributes_get_dof_near_transition(p_camera_attributes); + float bokeh_size = RSG::camera_attributes->camera_attributes_get_dof_blur_amount(p_camera_attributes) * 64; // Base 64 pixel radius. + + RS::DOFBokehShape bokeh_shape = RSG::camera_attributes->camera_attributes_get_dof_blur_bokeh_shape(); + RS::DOFBlurQuality blur_quality = RSG::camera_attributes->camera_attributes_get_dof_blur_quality(); + // setup our base push constant memset(&bokeh.push_constant, 0, sizeof(BokehPushConstant)); @@ -292,7 +322,7 @@ void BokehDOF::bokeh_dof_raster(const BokehBuffers &p_buffers, bool p_dof_far, f bokeh.push_constant.second_pass = false; bokeh.push_constant.half_size = false; - bokeh.push_constant.blur_size = p_dof_blur_amount; + bokeh.push_constant.blur_size = bokeh_size; // setup our uniforms RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); @@ -307,17 +337,17 @@ void BokehDOF::bokeh_dof_raster(const BokehBuffers &p_buffers, bool p_dof_far, f RD::Uniform u_weight_texture2(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_buffers.weight_texture[2] })); RD::Uniform u_weight_texture3(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_buffers.weight_texture[3] })); - if (p_dof_far || p_dof_near) { - if (p_dof_far) { + if (dof_far || dof_near) { + if (dof_far) { bokeh.push_constant.blur_far_active = true; - bokeh.push_constant.blur_far_begin = p_dof_far_begin; - bokeh.push_constant.blur_far_end = p_dof_far_begin + p_dof_far_size; + bokeh.push_constant.blur_far_begin = dof_far_begin; + bokeh.push_constant.blur_far_end = dof_far_begin + dof_far_size; } - if (p_dof_near) { + if (dof_near) { bokeh.push_constant.blur_near_active = true; - bokeh.push_constant.blur_near_begin = p_dof_near_begin; - bokeh.push_constant.blur_near_end = p_dof_near_begin - p_dof_near_size; + bokeh.push_constant.blur_near_begin = dof_near_begin; + bokeh.push_constant.blur_near_end = dof_near_begin - dof_near_size; } { @@ -337,14 +367,14 @@ void BokehDOF::bokeh_dof_raster(const BokehBuffers &p_buffers, bool p_dof_far, f RD::get_singleton()->draw_list_end(); } - if (p_bokeh_shape == RS::DOF_BOKEH_BOX || p_bokeh_shape == RS::DOF_BOKEH_HEXAGON) { + if (bokeh_shape == RS::DOF_BOKEH_BOX || bokeh_shape == RS::DOF_BOKEH_HEXAGON) { // double pass approach - BokehMode mode = p_bokeh_shape == RS::DOF_BOKEH_BOX ? BOKEH_GEN_BOKEH_BOX : BOKEH_GEN_BOKEH_HEXAGONAL; + BokehMode mode = bokeh_shape == RS::DOF_BOKEH_BOX ? BOKEH_GEN_BOKEH_BOX : BOKEH_GEN_BOKEH_HEXAGONAL; RID shader = bokeh.raster_shader.version_get_shader(bokeh.shader_version, mode); ERR_FAIL_COND(shader.is_null()); - if (p_quality == RS::DOF_BLUR_QUALITY_VERY_LOW || p_quality == RS::DOF_BLUR_QUALITY_LOW) { + if (blur_quality == RS::DOF_BLUR_QUALITY_VERY_LOW || blur_quality == RS::DOF_BLUR_QUALITY_LOW) { //box and hexagon are more or less the same, and they can work in either half (very low and low quality) or full (medium and high quality_ sizes) bokeh.push_constant.size[0] = p_buffers.base_texture_size.x >> 1; bokeh.push_constant.size[1] = p_buffers.base_texture_size.y >> 1; @@ -354,7 +384,7 @@ void BokehDOF::bokeh_dof_raster(const BokehBuffers &p_buffers, bool p_dof_far, f static const int quality_samples[4] = { 6, 12, 12, 24 }; bokeh.push_constant.blur_scale = 0.5; - bokeh.push_constant.steps = quality_samples[p_quality]; + bokeh.push_constant.steps = quality_samples[blur_quality]; RID framebuffer = bokeh.push_constant.half_size ? p_buffers.half_fb[0] : p_buffers.secondary_fb; @@ -373,7 +403,7 @@ void BokehDOF::bokeh_dof_raster(const BokehBuffers &p_buffers, bool p_dof_far, f // Pass 2 if (!bokeh.push_constant.half_size) { // do not output weight, we're writing back into our base buffer - mode = p_bokeh_shape == RS::DOF_BOKEH_BOX ? BOKEH_GEN_BOKEH_BOX_NOWEIGHT : BOKEH_GEN_BOKEH_HEXAGONAL_NOWEIGHT; + mode = bokeh_shape == RS::DOF_BOKEH_BOX ? BOKEH_GEN_BOKEH_BOX_NOWEIGHT : BOKEH_GEN_BOKEH_HEXAGONAL_NOWEIGHT; shader = bokeh.raster_shader.version_get_shader(bokeh.shader_version, mode); ERR_FAIL_COND(shader.is_null()); @@ -432,7 +462,7 @@ void BokehDOF::bokeh_dof_raster(const BokehBuffers &p_buffers, bool p_dof_far, f } static const float quality_scale[4] = { 8.0, 4.0, 1.0, 0.5 }; - bokeh.push_constant.blur_scale = quality_scale[p_quality]; + bokeh.push_constant.blur_scale = quality_scale[blur_quality]; bokeh.push_constant.steps = 0.0; RID framebuffer = bokeh.push_constant.half_size ? p_buffers.half_fb[0] : p_buffers.secondary_fb; diff --git a/servers/rendering/renderer_rd/effects/bokeh_dof.h b/servers/rendering/renderer_rd/effects/bokeh_dof.h index d7b736119c..33dbdfcdc1 100644 --- a/servers/rendering/renderer_rd/effects/bokeh_dof.h +++ b/servers/rendering/renderer_rd/effects/bokeh_dof.h @@ -66,6 +66,11 @@ private: uint32_t use_jitter; float jitter_seed; + uint32_t use_physical_near; + uint32_t use_physical_far; + + float blur_size_near; + float blur_size_far; uint32_t pad[2]; }; @@ -111,10 +116,10 @@ public: BokehDOF(bool p_prefer_raster_effects); ~BokehDOF(); - void bokeh_dof_compute(const BokehBuffers &p_buffers, 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); - void bokeh_dof_raster(const BokehBuffers &p_buffers, 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_dof_blur_amount, RenderingServer::DOFBokehShape p_bokeh_shape, RS::DOFBlurQuality p_quality, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal); + void bokeh_dof_compute(const BokehBuffers &p_buffers, RID p_camera_attributes, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal); + void bokeh_dof_raster(const BokehBuffers &p_buffers, RID p_camera_attributes, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal); }; } // namespace RendererRD -#endif // !BOKEH_DOF_RD_H +#endif // BOKEH_DOF_RD_H diff --git a/servers/rendering/renderer_rd/effects/copy_effects.cpp b/servers/rendering/renderer_rd/effects/copy_effects.cpp index cbf7046887..53237c1dfb 100644 --- a/servers/rendering/renderer_rd/effects/copy_effects.cpp +++ b/servers/rendering/renderer_rd/effects/copy_effects.cpp @@ -249,6 +249,56 @@ CopyEffects::CopyEffects(bool p_prefer_raster_effects) { roughness.raster_pipeline.clear(); } } + + { + Vector<String> specular_modes; + specular_modes.push_back("\n#define MODE_MERGE\n"); // SPECULAR_MERGE_ADD + specular_modes.push_back("\n#define MODE_MERGE\n#define MODE_SSR\n"); // SPECULAR_MERGE_SSR + specular_modes.push_back("\n"); // SPECULAR_MERGE_ADDITIVE_ADD + specular_modes.push_back("\n#define MODE_SSR\n"); // SPECULAR_MERGE_ADDITIVE_SSR + + specular_modes.push_back("\n#define USE_MULTIVIEW\n#define MODE_MERGE\n"); // SPECULAR_MERGE_ADD_MULTIVIEW + specular_modes.push_back("\n#define USE_MULTIVIEW\n#define MODE_MERGE\n#define MODE_SSR\n"); // SPECULAR_MERGE_SSR_MULTIVIEW + specular_modes.push_back("\n#define USE_MULTIVIEW\n"); // SPECULAR_MERGE_ADDITIVE_ADD_MULTIVIEW + specular_modes.push_back("\n#define USE_MULTIVIEW\n#define MODE_SSR\n"); // SPECULAR_MERGE_ADDITIVE_SSR_MULTIVIEW + + specular_merge.shader.initialize(specular_modes); + + if (!RendererCompositorRD::singleton->is_xr_enabled()) { + specular_merge.shader.set_variant_enabled(SPECULAR_MERGE_ADD_MULTIVIEW, false); + specular_merge.shader.set_variant_enabled(SPECULAR_MERGE_SSR_MULTIVIEW, false); + specular_merge.shader.set_variant_enabled(SPECULAR_MERGE_ADDITIVE_ADD_MULTIVIEW, false); + specular_merge.shader.set_variant_enabled(SPECULAR_MERGE_ADDITIVE_SSR_MULTIVIEW, false); + } + + specular_merge.shader_version = specular_merge.shader.version_create(); + + //use additive + + RD::PipelineColorBlendState::Attachment ba; + ba.enable_blend = true; + ba.src_color_blend_factor = RD::BLEND_FACTOR_ONE; + ba.dst_color_blend_factor = RD::BLEND_FACTOR_ONE; + ba.src_alpha_blend_factor = RD::BLEND_FACTOR_ONE; + ba.dst_alpha_blend_factor = RD::BLEND_FACTOR_ONE; + ba.color_blend_op = RD::BLEND_OP_ADD; + ba.alpha_blend_op = RD::BLEND_OP_ADD; + + RD::PipelineColorBlendState blend_additive; + blend_additive.attachments.push_back(ba); + + for (int i = 0; i < SPECULAR_MERGE_MAX; i++) { + if (specular_merge.shader.is_variant_enabled(i)) { + RD::PipelineColorBlendState blend_state; + if (i == SPECULAR_MERGE_ADDITIVE_ADD || i == SPECULAR_MERGE_ADDITIVE_SSR || i == SPECULAR_MERGE_ADDITIVE_ADD_MULTIVIEW || i == SPECULAR_MERGE_ADDITIVE_SSR_MULTIVIEW) { + blend_state = blend_additive; + } else { + blend_state = RD::PipelineColorBlendState::create_disabled(); + } + specular_merge.pipelines[i].setup(specular_merge.shader.version_get_shader(specular_merge.shader_version, i), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), blend_state, 0); + } + } + } } CopyEffects::~CopyEffects() { @@ -264,6 +314,8 @@ CopyEffects::~CopyEffects() { roughness.compute_shader.version_free(roughness.shader_version); } + specular_merge.shader.version_free(specular_merge.shader_version); + RD::get_singleton()->free(filter.coefficient_buffer); if (RD::get_singleton()->uniform_set_is_valid(filter.image_uniform_set)) { @@ -602,7 +654,7 @@ void CopyEffects::gaussian_blur(RID p_source_rd_texture, RID p_texture, const Re RD::get_singleton()->compute_list_end(); } -void CopyEffects::gaussian_glow(RID p_source_rd_texture, RID p_back_texture, const Size2i &p_size, float p_strength, bool p_high_quality, bool p_first_pass, float p_luminance_cap, float p_exposure, float p_bloom, float p_hdr_bleed_threshold, float p_hdr_bleed_scale, RID p_auto_exposure, float p_auto_exposure_grey) { +void CopyEffects::gaussian_glow(RID p_source_rd_texture, RID p_back_texture, const Size2i &p_size, float p_strength, bool p_high_quality, bool p_first_pass, float p_luminance_cap, float p_exposure, float p_bloom, float p_hdr_bleed_threshold, float p_hdr_bleed_scale, RID p_auto_exposure, float p_auto_exposure_scale) { ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use the compute version of the gaussian glow with the mobile renderer."); UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); @@ -626,7 +678,7 @@ void CopyEffects::gaussian_glow(RID p_source_rd_texture, RID p_back_texture, con copy.push_constant.glow_white = 0; //actually unused copy.push_constant.glow_luminance_cap = p_luminance_cap; - copy.push_constant.glow_auto_exposure_grey = p_auto_exposure_grey; //unused also + copy.push_constant.glow_auto_exposure_scale = p_auto_exposure_scale; //unused also // setup our uniforms RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); @@ -653,7 +705,7 @@ void CopyEffects::gaussian_glow(RID p_source_rd_texture, RID p_back_texture, con RD::get_singleton()->compute_list_end(); } -void CopyEffects::gaussian_glow_raster(RID p_source_rd_texture, float p_luminance_multiplier, RID p_framebuffer_half, RID p_rd_texture_half, RID p_dest_framebuffer, const Size2i &p_size, float p_strength, bool p_high_quality, bool p_first_pass, float p_luminance_cap, float p_exposure, float p_bloom, float p_hdr_bleed_threshold, float p_hdr_bleed_scale, RID p_auto_exposure, float p_auto_exposure_grey) { +void CopyEffects::gaussian_glow_raster(RID p_source_rd_texture, RID p_half_texture, RID p_dest_texture, float p_luminance_multiplier, const Size2i &p_size, float p_strength, bool p_high_quality, bool p_first_pass, float p_luminance_cap, float p_exposure, float p_bloom, float p_hdr_bleed_threshold, float p_hdr_bleed_scale, RID p_auto_exposure, float p_auto_exposure_scale) { ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use the raster version of the gaussian glow with the clustered renderer."); UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); @@ -661,6 +713,9 @@ void CopyEffects::gaussian_glow_raster(RID p_source_rd_texture, float p_luminanc MaterialStorage *material_storage = MaterialStorage::get_singleton(); ERR_FAIL_NULL(material_storage); + RID half_framebuffer = FramebufferCacheRD::get_singleton()->get_cache(p_half_texture); + RID dest_framebuffer = FramebufferCacheRD::get_singleton()->get_cache(p_dest_texture); + memset(&blur_raster.push_constant, 0, sizeof(BlurRasterPushConstant)); BlurRasterMode blur_mode = p_first_pass && p_auto_exposure.is_valid() ? BLUR_MODE_GAUSSIAN_GLOW_AUTO_EXPOSURE : BLUR_MODE_GAUSSIAN_GLOW; @@ -677,7 +732,7 @@ void CopyEffects::gaussian_glow_raster(RID p_source_rd_texture, float p_luminanc blur_raster.push_constant.glow_white = 0; //actually unused blur_raster.push_constant.glow_luminance_cap = p_luminance_cap; - blur_raster.push_constant.glow_auto_exposure_grey = p_auto_exposure_grey; //unused also + blur_raster.push_constant.glow_auto_exposure_scale = p_auto_exposure_scale; //unused also blur_raster.push_constant.luminance_multiplier = p_luminance_multiplier; @@ -685,14 +740,14 @@ void CopyEffects::gaussian_glow_raster(RID p_source_rd_texture, float p_luminanc RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); RD::Uniform u_source_rd_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_source_rd_texture })); - RD::Uniform u_rd_texture_half(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_rd_texture_half })); + RD::Uniform u_half_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_half_texture })); RID shader = blur_raster.shader.version_get_shader(blur_raster.shader_version, blur_mode); ERR_FAIL_COND(shader.is_null()); //HORIZONTAL - RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_framebuffer_half, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); - RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, blur_raster.pipelines[blur_mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_framebuffer_half))); + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(half_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, blur_raster.pipelines[blur_mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(half_framebuffer))); RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 0, u_source_rd_texture), 0); if (p_auto_exposure.is_valid() && p_first_pass) { RD::Uniform u_auto_exposure(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_auto_exposure })); @@ -712,9 +767,9 @@ void CopyEffects::gaussian_glow_raster(RID p_source_rd_texture, float p_luminanc ERR_FAIL_COND(shader.is_null()); //VERTICAL - draw_list = RD::get_singleton()->draw_list_begin(p_dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); - RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, blur_raster.pipelines[blur_mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dest_framebuffer))); - RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 0, u_rd_texture_half), 0); + draw_list = RD::get_singleton()->draw_list_begin(dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, blur_raster.pipelines[blur_mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(dest_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 0, u_half_texture), 0); RD::get_singleton()->draw_list_bind_index_array(draw_list, material_storage->get_quad_index_array()); blur_raster.push_constant.flags = base_flags; @@ -758,9 +813,11 @@ void CopyEffects::make_mipmap(RID p_source_rd_texture, RID p_dest_texture, const RD::get_singleton()->compute_list_end(); } -void CopyEffects::make_mipmap_raster(RID p_source_rd_texture, RID p_dest_framebuffer, const Size2i &p_size) { +void CopyEffects::make_mipmap_raster(RID p_source_rd_texture, RID p_dest_texture, const Size2i &p_size) { ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use the raster version of mipmap with the clustered renderer."); + RID dest_framebuffer = FramebufferCacheRD::get_singleton()->get_cache(p_dest_texture); + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); ERR_FAIL_NULL(uniform_set_cache); MaterialStorage *material_storage = MaterialStorage::get_singleton(); @@ -781,8 +838,8 @@ void CopyEffects::make_mipmap_raster(RID p_source_rd_texture, RID p_dest_framebu RID shader = blur_raster.shader.version_get_shader(blur_raster.shader_version, mode); ERR_FAIL_COND(shader.is_null()); - RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); - RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, blur_raster.pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dest_framebuffer))); + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, blur_raster.pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(dest_framebuffer))); RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 0, u_source_rd_texture), 0); RD::get_singleton()->draw_list_bind_index_array(draw_list, material_storage->get_quad_index_array()); RD::get_singleton()->draw_list_set_push_constant(draw_list, &blur_raster.push_constant, sizeof(BlurRasterPushConstant)); @@ -1083,3 +1140,57 @@ void CopyEffects::cubemap_roughness_raster(RID p_source_rd_texture, RID p_dest_f RD::get_singleton()->draw_list_draw(draw_list, true); RD::get_singleton()->draw_list_end(); } + +void CopyEffects::merge_specular(RID p_dest_framebuffer, RID p_specular, RID p_base, RID p_reflection, uint32_t p_view_count) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::get_singleton()->draw_command_begin_label("Merge specular"); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, Vector<Color>()); + + int mode; + if (p_reflection.is_valid()) { + if (p_base.is_valid()) { + mode = SPECULAR_MERGE_SSR; + } else { + mode = SPECULAR_MERGE_ADDITIVE_SSR; + } + } else { + if (p_base.is_valid()) { + mode = SPECULAR_MERGE_ADD; + } else { + mode = SPECULAR_MERGE_ADDITIVE_ADD; + } + } + + if (p_view_count > 1) { + mode += SPECULAR_MERGE_ADD_MULTIVIEW; + } + + RID shader = specular_merge.shader.version_get_shader(specular_merge.shader_version, mode); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, specular_merge.pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dest_framebuffer))); + + if (p_base.is_valid()) { + RD::Uniform u_base(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_base })); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 2, u_base), 2); + } + + RD::Uniform u_specular(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_specular })); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 0, u_specular), 0); + + if (p_reflection.is_valid()) { + RD::Uniform u_reflection(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_reflection })); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uniform_set_cache->get_cache(shader, 1, u_reflection), 1); + } + + RD::get_singleton()->draw_list_bind_index_array(draw_list, material_storage->get_quad_index_array()); + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + + RD::get_singleton()->draw_command_end_label(); +} diff --git a/servers/rendering/renderer_rd/effects/copy_effects.h b/servers/rendering/renderer_rd/effects/copy_effects.h index 882b446964..0ddb60ebef 100644 --- a/servers/rendering/renderer_rd/effects/copy_effects.h +++ b/servers/rendering/renderer_rd/effects/copy_effects.h @@ -28,8 +28,8 @@ /* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ /*************************************************************************/ -#ifndef COPY_RD_H -#define COPY_RD_H +#ifndef COPY_EFFECTS_RD_H +#define COPY_EFFECTS_RD_H #include "servers/rendering/renderer_rd/pipeline_cache_rd.h" #include "servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl.gen.h" @@ -42,6 +42,7 @@ #include "servers/rendering/renderer_rd/shaders/effects/cubemap_filter_raster.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/effects/cubemap_roughness.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/effects/cubemap_roughness_raster.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/specular_merge.glsl.gen.h" #include "servers/rendering/renderer_scene_render.h" #include "servers/rendering_server.h" @@ -85,7 +86,7 @@ private: float glow_exposure; float glow_white; float glow_luminance_cap; - float glow_auto_exposure_grey; + float glow_auto_exposure_scale; float luminance_multiplier; float res1; @@ -147,7 +148,7 @@ private: float glow_exposure; float glow_white; float glow_luminance_cap; - float glow_auto_exposure_grey; + float glow_auto_exposure_scale; // DOF. float camera_z_far; float camera_z_near; @@ -274,6 +275,33 @@ private: PipelineCacheRD raster_pipeline; } roughness; + // Merge specular + + enum SpecularMergeMode { + SPECULAR_MERGE_ADD, + SPECULAR_MERGE_SSR, + SPECULAR_MERGE_ADDITIVE_ADD, + SPECULAR_MERGE_ADDITIVE_SSR, + + SPECULAR_MERGE_ADD_MULTIVIEW, + SPECULAR_MERGE_SSR_MULTIVIEW, + SPECULAR_MERGE_ADDITIVE_ADD_MULTIVIEW, + SPECULAR_MERGE_ADDITIVE_SSR_MULTIVIEW, + + SPECULAR_MERGE_MAX + }; + + /* Specular merge must be done using raster, rather than compute + * because it must continue the existing color buffer + */ + + struct SpecularMerge { + SpecularMergeShaderRD shader; + RID shader_version; + PipelineCacheRD pipelines[SPECULAR_MERGE_MAX]; + + } specular_merge; + static CopyEffects *singleton; public: @@ -293,11 +321,11 @@ public: void copy_raster(RID p_source_texture, RID p_dest_framebuffer); void gaussian_blur(RID p_source_rd_texture, RID p_texture, const Rect2i &p_region, bool p_8bit_dst = false); - void gaussian_glow(RID p_source_rd_texture, RID p_back_texture, const Size2i &p_size, float p_strength = 1.0, bool p_high_quality = false, bool p_first_pass = false, float p_luminance_cap = 16.0, float p_exposure = 1.0, float p_bloom = 0.0, float p_hdr_bleed_threshold = 1.0, float p_hdr_bleed_scale = 1.0, RID p_auto_exposure = RID(), float p_auto_exposure_grey = 1.0); - void gaussian_glow_raster(RID p_source_rd_texture, float p_luminance_multiplier, RID p_framebuffer_half, RID p_rd_texture_half, RID p_dest_framebuffer, const Size2i &p_size, float p_strength = 1.0, bool p_high_quality = false, bool p_first_pass = false, float p_luminance_cap = 16.0, float p_exposure = 1.0, float p_bloom = 0.0, float p_hdr_bleed_threshold = 1.0, float p_hdr_bleed_scale = 1.0, RID p_auto_exposure = RID(), float p_auto_exposure_grey = 1.0); + void gaussian_glow(RID p_source_rd_texture, RID p_back_texture, const Size2i &p_size, float p_strength = 1.0, bool p_high_quality = false, bool p_first_pass = false, float p_luminance_cap = 16.0, float p_exposure = 1.0, float p_bloom = 0.0, float p_hdr_bleed_threshold = 1.0, float p_hdr_bleed_scale = 1.0, RID p_auto_exposure = RID(), float p_auto_exposure_scale = 1.0); + void gaussian_glow_raster(RID p_source_rd_texture, RID p_half_texture, RID p_dest_texture, float p_luminance_multiplier, const Size2i &p_size, float p_strength = 1.0, bool p_high_quality = false, bool p_first_pass = false, float p_luminance_cap = 16.0, float p_exposure = 1.0, float p_bloom = 0.0, float p_hdr_bleed_threshold = 1.0, float p_hdr_bleed_scale = 1.0, RID p_auto_exposure = RID(), float p_auto_exposure_scale = 1.0); void make_mipmap(RID p_source_rd_texture, RID p_dest_texture, const Size2i &p_size); - void make_mipmap_raster(RID p_source_rd_texture, RID p_dest_framebuffer, const Size2i &p_size); + void make_mipmap_raster(RID p_source_rd_texture, RID p_dest_texture, const Size2i &p_size); void set_color(RID p_dest_texture, const Color &p_color, const Rect2i &p_region, bool p_8bit_dst = false); @@ -309,8 +337,10 @@ public: void cubemap_roughness(RID p_source_rd_texture, RID p_dest_texture, uint32_t p_face_id, uint32_t p_sample_count, float p_roughness, float p_size); void cubemap_roughness_raster(RID p_source_rd_texture, RID p_dest_framebuffer, uint32_t p_face_id, uint32_t p_sample_count, float p_roughness, float p_size); + + void merge_specular(RID p_dest_framebuffer, RID p_specular, RID p_base, RID p_reflection, uint32_t p_view_count); }; } // namespace RendererRD -#endif // !COPY_RD_H +#endif // COPY_EFFECTS_RD_H diff --git a/servers/rendering/renderer_rd/effects/fsr.cpp b/servers/rendering/renderer_rd/effects/fsr.cpp new file mode 100644 index 0000000000..92b34ede0e --- /dev/null +++ b/servers/rendering/renderer_rd/effects/fsr.cpp @@ -0,0 +1,128 @@ +/*************************************************************************/ +/* fsr.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "fsr.h" +#include "../storage_rd/material_storage.h" +#include "../uniform_set_cache_rd.h" + +using namespace RendererRD; + +FSR::FSR() { + Vector<String> FSR_upscale_modes; + +#if defined(MACOS_ENABLED) || defined(IOS_ENABLED) + // MoltenVK does not support some of the operations used by the normal mode of FSR. Fallback works just fine though. + FSR_upscale_modes.push_back("\n#define MODE_FSR_UPSCALE_FALLBACK\n"); +#else + // Everyone else can use normal mode when available. + if (RD::get_singleton()->has_feature(RD::SUPPORTS_FSR_HALF_FLOAT)) { + FSR_upscale_modes.push_back("\n#define MODE_FSR_UPSCALE_NORMAL\n"); + } else { + FSR_upscale_modes.push_back("\n#define MODE_FSR_UPSCALE_FALLBACK\n"); + } +#endif + + fsr_shader.initialize(FSR_upscale_modes); + + shader_version = fsr_shader.version_create(); + pipeline = RD::get_singleton()->compute_pipeline_create(fsr_shader.version_get_shader(shader_version, 0)); +} + +FSR::~FSR() { + fsr_shader.version_free(shader_version); +} + +void FSR::fsr_upscale(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_source_rd_texture, RID p_destination_texture) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + Size2i internal_size = p_render_buffers->get_internal_size(); + Size2i target_size = p_render_buffers->get_target_size(); + float fsr_upscale_sharpness = p_render_buffers->get_fsr_sharpness(); + + if (!p_render_buffers->has_texture(SNAME("FSR"), SNAME("upscale_texture"))) { + RD::DataFormat format = p_render_buffers->get_base_data_format(); + uint32_t usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + uint32_t layers = 1; // we only need one layer, in multiview we're processing one layer at a time. + + p_render_buffers->create_texture(SNAME("FSR"), SNAME("upscale_texture"), format, usage_bits, RD::TEXTURE_SAMPLES_1, target_size, layers); + } + + RID upscale_texture = p_render_buffers->get_texture(SNAME("FSR"), SNAME("upscale_texture")); + + FSRUpscalePushConstant push_constant; + memset(&push_constant, 0, sizeof(FSRUpscalePushConstant)); + + int dispatch_x = (target_size.x + 15) / 16; + int dispatch_y = (target_size.y + 15) / 16; + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, pipeline); + + push_constant.resolution_width = internal_size.width; + push_constant.resolution_height = internal_size.height; + push_constant.upscaled_width = target_size.width; + push_constant.upscaled_height = target_size.height; + push_constant.sharpness = fsr_upscale_sharpness; + + RID shader = fsr_shader.version_get_shader(shader_version, 0); + ERR_FAIL_COND(shader.is_null()); + + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + //FSR Easc + RD::Uniform u_source_rd_texture(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, { default_sampler, p_source_rd_texture }); + RD::Uniform u_upscale_texture(RD::UNIFORM_TYPE_IMAGE, 0, { upscale_texture }); + + push_constant.pass = FSR_UPSCALE_PASS_EASU; + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_source_rd_texture), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_upscale_texture), 1); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(FSRUpscalePushConstant)); + + RD::get_singleton()->compute_list_dispatch(compute_list, dispatch_x, dispatch_y, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + //FSR Rcas + RD::Uniform u_upscale_texture_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, { default_sampler, upscale_texture }); + RD::Uniform u_destination_texture(RD::UNIFORM_TYPE_IMAGE, 0, { p_destination_texture }); + + push_constant.pass = FSR_UPSCALE_PASS_RCAS; + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_upscale_texture_with_sampler), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_destination_texture), 1); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(FSRUpscalePushConstant)); + + RD::get_singleton()->compute_list_dispatch(compute_list, dispatch_x, dispatch_y, 1); + + RD::get_singleton()->compute_list_end(compute_list); +} diff --git a/servers/rendering/renderer_rd/effects/fsr.h b/servers/rendering/renderer_rd/effects/fsr.h new file mode 100644 index 0000000000..69088e526a --- /dev/null +++ b/servers/rendering/renderer_rd/effects/fsr.h @@ -0,0 +1,72 @@ +/*************************************************************************/ +/* fsr.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef FSR_RD_H +#define FSR_RD_H + +#include "../pipeline_cache_rd.h" +#include "../shaders/effects/fsr_upscale.glsl.gen.h" +#include "../storage_rd/render_scene_buffers_rd.h" +#include "servers/rendering/renderer_scene_render.h" +#include "servers/rendering_server.h" + +namespace RendererRD { + +class FSR { +public: + FSR(); + ~FSR(); + + void fsr_upscale(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_source_rd_texture, RID p_destination_texture); + +private: + enum FSRUpscalePass { + FSR_UPSCALE_PASS_EASU = 0, + FSR_UPSCALE_PASS_RCAS = 1 + }; + + struct FSRUpscalePushConstant { + float resolution_width; + float resolution_height; + float upscaled_width; + float upscaled_height; + float sharpness; + int pass; + int _unused0, _unused1; + }; + + FsrUpscaleShaderRD fsr_shader; + RID shader_version; + RID pipeline; +}; + +} // namespace RendererRD + +#endif // FSR_RD_H diff --git a/servers/rendering/renderer_rd/effects/resolve.h b/servers/rendering/renderer_rd/effects/resolve.h index d4b24a610f..2a4cd06827 100644 --- a/servers/rendering/renderer_rd/effects/resolve.h +++ b/servers/rendering/renderer_rd/effects/resolve.h @@ -71,4 +71,4 @@ public: } // namespace RendererRD -#endif // !RESOLVE_RD_H +#endif // RESOLVE_RD_H diff --git a/servers/rendering/renderer_rd/effects/ss_effects.cpp b/servers/rendering/renderer_rd/effects/ss_effects.cpp new file mode 100644 index 0000000000..9653382e96 --- /dev/null +++ b/servers/rendering/renderer_rd/effects/ss_effects.cpp @@ -0,0 +1,1876 @@ +/*************************************************************************/ +/* ss_effects.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "ss_effects.h" + +#include "core/config/project_settings.h" +#include "servers/rendering/renderer_rd/renderer_compositor_rd.h" +#include "servers/rendering/renderer_rd/storage_rd/material_storage.h" +#include "servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.h" +#include "servers/rendering/renderer_rd/uniform_set_cache_rd.h" + +using namespace RendererRD; + +SSEffects *SSEffects::singleton = nullptr; + +static _FORCE_INLINE_ void store_camera(const Projection &p_mtx, float *p_array) { + for (int i = 0; i < 4; i++) { + for (int j = 0; j < 4; j++) { + p_array[i * 4 + j] = p_mtx.columns[i][j]; + } + } +} + +SSEffects::SSEffects() { + singleton = this; + + // Initialize depth buffer for screen space effects + { + Vector<String> downsampler_modes; + downsampler_modes.push_back("\n"); + downsampler_modes.push_back("\n#define USE_HALF_SIZE\n"); + downsampler_modes.push_back("\n#define GENERATE_MIPS\n"); + downsampler_modes.push_back("\n#define GENERATE_MIPS\n#define USE_HALF_SIZE\n"); + downsampler_modes.push_back("\n#define USE_HALF_BUFFERS\n"); + downsampler_modes.push_back("\n#define USE_HALF_BUFFERS\n#define USE_HALF_SIZE\n"); + downsampler_modes.push_back("\n#define GENERATE_MIPS\n#define GENERATE_FULL_MIPS"); + + ss_effects.downsample_shader.initialize(downsampler_modes); + + ss_effects.downsample_shader_version = ss_effects.downsample_shader.version_create(); + + for (int i = 0; i < SS_EFFECTS_MAX; i++) { + ss_effects.pipelines[i] = RD::get_singleton()->compute_pipeline_create(ss_effects.downsample_shader.version_get_shader(ss_effects.downsample_shader_version, i)); + } + + ss_effects.gather_constants_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(SSEffectsGatherConstants)); + SSEffectsGatherConstants gather_constants; + + const int sub_pass_count = 5; + for (int pass = 0; pass < 4; pass++) { + for (int subPass = 0; subPass < sub_pass_count; subPass++) { + int a = pass; + int b = subPass; + + int spmap[5]{ 0, 1, 4, 3, 2 }; + b = spmap[subPass]; + + float ca, sa; + float angle0 = (float(a) + float(b) / float(sub_pass_count)) * Math_PI * 0.5f; + + ca = Math::cos(angle0); + sa = Math::sin(angle0); + + float scale = 1.0f + (a - 1.5f + (b - (sub_pass_count - 1.0f) * 0.5f) / float(sub_pass_count)) * 0.07f; + + gather_constants.rotation_matrices[pass * 20 + subPass * 4 + 0] = scale * ca; + gather_constants.rotation_matrices[pass * 20 + subPass * 4 + 1] = scale * -sa; + gather_constants.rotation_matrices[pass * 20 + subPass * 4 + 2] = -scale * sa; + gather_constants.rotation_matrices[pass * 20 + subPass * 4 + 3] = -scale * ca; + } + } + + RD::get_singleton()->buffer_update(ss_effects.gather_constants_buffer, 0, sizeof(SSEffectsGatherConstants), &gather_constants); + } + + // Initialize Screen Space Indirect Lighting (SSIL) + ssil_set_quality(RS::EnvironmentSSILQuality(int(GLOBAL_GET("rendering/environment/ssil/quality"))), GLOBAL_GET("rendering/environment/ssil/half_size"), GLOBAL_GET("rendering/environment/ssil/adaptive_target"), GLOBAL_GET("rendering/environment/ssil/blur_passes"), GLOBAL_GET("rendering/environment/ssil/fadeout_from"), GLOBAL_GET("rendering/environment/ssil/fadeout_to")); + + { + Vector<String> ssil_modes; + ssil_modes.push_back("\n"); + ssil_modes.push_back("\n#define SSIL_BASE\n"); + ssil_modes.push_back("\n#define ADAPTIVE\n"); + + ssil.gather_shader.initialize(ssil_modes); + + ssil.gather_shader_version = ssil.gather_shader.version_create(); + + for (int i = SSIL_GATHER; i <= SSIL_GATHER_ADAPTIVE; i++) { + ssil.pipelines[i] = RD::get_singleton()->compute_pipeline_create(ssil.gather_shader.version_get_shader(ssil.gather_shader_version, i)); + } + ssil.projection_uniform_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(SSILProjectionUniforms)); + } + + { + Vector<String> ssil_modes; + ssil_modes.push_back("\n#define GENERATE_MAP\n"); + ssil_modes.push_back("\n#define PROCESS_MAPA\n"); + ssil_modes.push_back("\n#define PROCESS_MAPB\n"); + + ssil.importance_map_shader.initialize(ssil_modes); + + ssil.importance_map_shader_version = ssil.importance_map_shader.version_create(); + + for (int i = SSIL_GENERATE_IMPORTANCE_MAP; i <= SSIL_PROCESS_IMPORTANCE_MAPB; i++) { + ssil.pipelines[i] = RD::get_singleton()->compute_pipeline_create(ssil.importance_map_shader.version_get_shader(ssil.importance_map_shader_version, i - SSIL_GENERATE_IMPORTANCE_MAP)); + } + ssil.importance_map_load_counter = RD::get_singleton()->storage_buffer_create(sizeof(uint32_t)); + int zero[1] = { 0 }; + RD::get_singleton()->buffer_update(ssil.importance_map_load_counter, 0, sizeof(uint32_t), &zero); + RD::get_singleton()->set_resource_name(ssil.importance_map_load_counter, "Importance Map Load Counter"); + + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 0; + u.append_id(ssil.importance_map_load_counter); + uniforms.push_back(u); + } + ssil.counter_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssil.importance_map_shader.version_get_shader(ssil.importance_map_shader_version, 2), 2); + RD::get_singleton()->set_resource_name(ssil.counter_uniform_set, "Load Counter Uniform Set"); + } + + { + Vector<String> ssil_modes; + ssil_modes.push_back("\n#define MODE_NON_SMART\n"); + ssil_modes.push_back("\n#define MODE_SMART\n"); + ssil_modes.push_back("\n#define MODE_WIDE\n"); + + ssil.blur_shader.initialize(ssil_modes); + + ssil.blur_shader_version = ssil.blur_shader.version_create(); + for (int i = SSIL_BLUR_PASS; i <= SSIL_BLUR_PASS_WIDE; i++) { + ssil.pipelines[i] = RD::get_singleton()->compute_pipeline_create(ssil.blur_shader.version_get_shader(ssil.blur_shader_version, i - SSIL_BLUR_PASS)); + } + } + + { + Vector<String> ssil_modes; + ssil_modes.push_back("\n#define MODE_NON_SMART\n"); + ssil_modes.push_back("\n#define MODE_SMART\n"); + ssil_modes.push_back("\n#define MODE_HALF\n"); + + ssil.interleave_shader.initialize(ssil_modes); + + ssil.interleave_shader_version = ssil.interleave_shader.version_create(); + for (int i = SSIL_INTERLEAVE; i <= SSIL_INTERLEAVE_HALF; i++) { + ssil.pipelines[i] = RD::get_singleton()->compute_pipeline_create(ssil.interleave_shader.version_get_shader(ssil.interleave_shader_version, i - SSIL_INTERLEAVE)); + } + } + + // Initialize Screen Space Ambient Occlusion (SSAO) + ssao_set_quality(RS::EnvironmentSSAOQuality(int(GLOBAL_GET("rendering/environment/ssao/quality"))), GLOBAL_GET("rendering/environment/ssao/half_size"), GLOBAL_GET("rendering/environment/ssao/adaptive_target"), GLOBAL_GET("rendering/environment/ssao/blur_passes"), GLOBAL_GET("rendering/environment/ssao/fadeout_from"), GLOBAL_GET("rendering/environment/ssao/fadeout_to")); + + { + RD::SamplerState sampler; + sampler.mag_filter = RD::SAMPLER_FILTER_NEAREST; + sampler.min_filter = RD::SAMPLER_FILTER_NEAREST; + sampler.mip_filter = RD::SAMPLER_FILTER_NEAREST; + sampler.repeat_u = RD::SAMPLER_REPEAT_MODE_MIRRORED_REPEAT; + sampler.repeat_v = RD::SAMPLER_REPEAT_MODE_MIRRORED_REPEAT; + sampler.repeat_w = RD::SAMPLER_REPEAT_MODE_MIRRORED_REPEAT; + sampler.max_lod = 4; + + uint32_t pipeline = 0; + { + Vector<String> ssao_modes; + + ssao_modes.push_back("\n"); + ssao_modes.push_back("\n#define SSAO_BASE\n"); + ssao_modes.push_back("\n#define ADAPTIVE\n"); + + ssao.gather_shader.initialize(ssao_modes); + + ssao.gather_shader_version = ssao.gather_shader.version_create(); + + for (int i = 0; i <= SSAO_GATHER_ADAPTIVE; i++) { + ssao.pipelines[pipeline] = RD::get_singleton()->compute_pipeline_create(ssao.gather_shader.version_get_shader(ssao.gather_shader_version, i)); + pipeline++; + } + } + + { + Vector<String> ssao_modes; + ssao_modes.push_back("\n#define GENERATE_MAP\n"); + ssao_modes.push_back("\n#define PROCESS_MAPA\n"); + ssao_modes.push_back("\n#define PROCESS_MAPB\n"); + + ssao.importance_map_shader.initialize(ssao_modes); + + ssao.importance_map_shader_version = ssao.importance_map_shader.version_create(); + + for (int i = SSAO_GENERATE_IMPORTANCE_MAP; i <= SSAO_PROCESS_IMPORTANCE_MAPB; i++) { + ssao.pipelines[pipeline] = RD::get_singleton()->compute_pipeline_create(ssao.importance_map_shader.version_get_shader(ssao.importance_map_shader_version, i - SSAO_GENERATE_IMPORTANCE_MAP)); + + pipeline++; + } + + ssao.importance_map_load_counter = RD::get_singleton()->storage_buffer_create(sizeof(uint32_t)); + int zero[1] = { 0 }; + RD::get_singleton()->buffer_update(ssao.importance_map_load_counter, 0, sizeof(uint32_t), &zero); + RD::get_singleton()->set_resource_name(ssao.importance_map_load_counter, "Importance Map Load Counter"); + + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 0; + u.append_id(ssao.importance_map_load_counter); + uniforms.push_back(u); + } + ssao.counter_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssao.importance_map_shader.version_get_shader(ssao.importance_map_shader_version, 2), 2); + RD::get_singleton()->set_resource_name(ssao.counter_uniform_set, "Load Counter Uniform Set"); + } + + { + Vector<String> ssao_modes; + ssao_modes.push_back("\n#define MODE_NON_SMART\n"); + ssao_modes.push_back("\n#define MODE_SMART\n"); + ssao_modes.push_back("\n#define MODE_WIDE\n"); + + ssao.blur_shader.initialize(ssao_modes); + + ssao.blur_shader_version = ssao.blur_shader.version_create(); + + for (int i = SSAO_BLUR_PASS; i <= SSAO_BLUR_PASS_WIDE; i++) { + ssao.pipelines[pipeline] = RD::get_singleton()->compute_pipeline_create(ssao.blur_shader.version_get_shader(ssao.blur_shader_version, i - SSAO_BLUR_PASS)); + + pipeline++; + } + } + + { + Vector<String> ssao_modes; + ssao_modes.push_back("\n#define MODE_NON_SMART\n"); + ssao_modes.push_back("\n#define MODE_SMART\n"); + ssao_modes.push_back("\n#define MODE_HALF\n"); + + ssao.interleave_shader.initialize(ssao_modes); + + ssao.interleave_shader_version = ssao.interleave_shader.version_create(); + for (int i = SSAO_INTERLEAVE; i <= SSAO_INTERLEAVE_HALF; i++) { + ssao.pipelines[pipeline] = RD::get_singleton()->compute_pipeline_create(ssao.interleave_shader.version_get_shader(ssao.interleave_shader_version, i - SSAO_INTERLEAVE)); + RD::get_singleton()->set_resource_name(ssao.pipelines[pipeline], "Interleave Pipeline " + itos(i)); + pipeline++; + } + } + + ERR_FAIL_COND(pipeline != SSAO_MAX); + + ss_effects.mirror_sampler = RD::get_singleton()->sampler_create(sampler); + } + + // Screen Space Reflections + ssr_roughness_quality = RS::EnvironmentSSRRoughnessQuality(int(GLOBAL_GET("rendering/environment/screen_space_reflection/roughness_quality"))); + + { + Vector<RD::PipelineSpecializationConstant> specialization_constants; + + { + RD::PipelineSpecializationConstant sc; + sc.type = RD::PIPELINE_SPECIALIZATION_CONSTANT_TYPE_BOOL; + sc.constant_id = 0; // SSR_USE_FULL_PROJECTION_MATRIX + sc.bool_value = false; + specialization_constants.push_back(sc); + } + + { + Vector<String> ssr_scale_modes; + ssr_scale_modes.push_back("\n"); + + ssr_scale.shader.initialize(ssr_scale_modes); + ssr_scale.shader_version = ssr_scale.shader.version_create(); + + for (int v = 0; v < SSR_VARIATIONS; v++) { + specialization_constants.ptrw()[0].bool_value = (v & SSR_MULTIVIEW) ? true : false; + ssr_scale.pipelines[v] = RD::get_singleton()->compute_pipeline_create(ssr_scale.shader.version_get_shader(ssr_scale.shader_version, 0), specialization_constants); + } + } + + { + Vector<String> ssr_modes; + ssr_modes.push_back("\n"); // SCREEN_SPACE_REFLECTION_NORMAL + ssr_modes.push_back("\n#define MODE_ROUGH\n"); // SCREEN_SPACE_REFLECTION_ROUGH + + ssr.shader.initialize(ssr_modes); + ssr.shader_version = ssr.shader.version_create(); + + for (int v = 0; v < SSR_VARIATIONS; v++) { + specialization_constants.ptrw()[0].bool_value = (v & SSR_MULTIVIEW) ? true : false; + for (int i = 0; i < SCREEN_SPACE_REFLECTION_MAX; i++) { + ssr.pipelines[v][i] = RD::get_singleton()->compute_pipeline_create(ssr.shader.version_get_shader(ssr.shader_version, i), specialization_constants); + } + } + } + + { + Vector<String> ssr_filter_modes; + ssr_filter_modes.push_back("\n"); // SCREEN_SPACE_REFLECTION_FILTER_HORIZONTAL + ssr_filter_modes.push_back("\n#define VERTICAL_PASS\n"); // SCREEN_SPACE_REFLECTION_FILTER_VERTICAL + + ssr_filter.shader.initialize(ssr_filter_modes); + ssr_filter.shader_version = ssr_filter.shader.version_create(); + + for (int v = 0; v < SSR_VARIATIONS; v++) { + specialization_constants.ptrw()[0].bool_value = (v & SSR_MULTIVIEW) ? true : false; + for (int i = 0; i < SCREEN_SPACE_REFLECTION_FILTER_MAX; i++) { + ssr_filter.pipelines[v][i] = RD::get_singleton()->compute_pipeline_create(ssr_filter.shader.version_get_shader(ssr_filter.shader_version, i), specialization_constants); + } + } + } + } + + // Subsurface scattering + sss_quality = RS::SubSurfaceScatteringQuality(int(GLOBAL_GET("rendering/environment/subsurface_scattering/subsurface_scattering_quality"))); + sss_scale = GLOBAL_GET("rendering/environment/subsurface_scattering/subsurface_scattering_scale"); + sss_depth_scale = GLOBAL_GET("rendering/environment/subsurface_scattering/subsurface_scattering_depth_scale"); + + { + Vector<String> sss_modes; + sss_modes.push_back("\n#define USE_11_SAMPLES\n"); + sss_modes.push_back("\n#define USE_17_SAMPLES\n"); + sss_modes.push_back("\n#define USE_25_SAMPLES\n"); + + sss.shader.initialize(sss_modes); + + sss.shader_version = sss.shader.version_create(); + + for (int i = 0; i < sss_modes.size(); i++) { + sss.pipelines[i] = RD::get_singleton()->compute_pipeline_create(sss.shader.version_get_shader(sss.shader_version, i)); + } + } +} + +SSEffects::~SSEffects() { + { + // Cleanup SS Reflections + ssr.shader.version_free(ssr.shader_version); + ssr_filter.shader.version_free(ssr_filter.shader_version); + ssr_scale.shader.version_free(ssr_scale.shader_version); + + if (ssr.ubo.is_valid()) { + RD::get_singleton()->free(ssr.ubo); + } + } + + { + // Cleanup SS downsampler + ss_effects.downsample_shader.version_free(ss_effects.downsample_shader_version); + + RD::get_singleton()->free(ss_effects.mirror_sampler); + RD::get_singleton()->free(ss_effects.gather_constants_buffer); + } + + { + // Cleanup SSIL + ssil.blur_shader.version_free(ssil.blur_shader_version); + ssil.gather_shader.version_free(ssil.gather_shader_version); + ssil.interleave_shader.version_free(ssil.interleave_shader_version); + ssil.importance_map_shader.version_free(ssil.importance_map_shader_version); + + RD::get_singleton()->free(ssil.importance_map_load_counter); + RD::get_singleton()->free(ssil.projection_uniform_buffer); + } + + { + // Cleanup SSAO + ssao.blur_shader.version_free(ssao.blur_shader_version); + ssao.gather_shader.version_free(ssao.gather_shader_version); + ssao.interleave_shader.version_free(ssao.interleave_shader_version); + ssao.importance_map_shader.version_free(ssao.importance_map_shader_version); + + RD::get_singleton()->free(ssao.importance_map_load_counter); + } + + { + // Cleanup Subsurface scattering + sss.shader.version_free(sss.shader_version); + } + + singleton = nullptr; +} + +/* SS Downsampler */ + +void SSEffects::downsample_depth(RID p_depth_buffer, const Vector<RID> &p_depth_mipmaps, bool p_invalidate_uniform_set, Size2i p_full_screen_size, const Projection &p_projection) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + // Downsample and deinterleave the depth buffer for SSAO and SSIL + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + + int downsample_mode = SS_EFFECTS_DOWNSAMPLE; + bool use_mips = ssao_quality > RS::ENV_SSAO_QUALITY_MEDIUM || ssil_quality > RS::ENV_SSIL_QUALITY_MEDIUM; + + if (ssao_quality == RS::ENV_SSAO_QUALITY_VERY_LOW && ssil_quality == RS::ENV_SSIL_QUALITY_VERY_LOW) { + downsample_mode = SS_EFFECTS_DOWNSAMPLE_HALF; + } else if (use_mips) { + downsample_mode = SS_EFFECTS_DOWNSAMPLE_MIPMAP; + } + + bool use_half_size = false; + bool use_full_mips = false; + + if (ssao_half_size && ssil_half_size) { + downsample_mode++; + use_half_size = true; + } else if (ssao_half_size != ssil_half_size) { + if (use_mips) { + downsample_mode = SS_EFFECTS_DOWNSAMPLE_FULL_MIPS; + use_full_mips = true; + } else { + // Only need the first two mipmaps, but the cost to generate the next two is trivial + // TODO investigate the benefit of a shader version to generate only 2 mips + downsample_mode = SS_EFFECTS_DOWNSAMPLE_MIPMAP; + use_mips = true; + } + } + + int depth_index = use_half_size ? 1 : 0; + + RD::get_singleton()->draw_command_begin_label("Downsample Depth"); + if (p_invalidate_uniform_set || use_full_mips != ss_effects.used_full_mips_last_frame || use_half_size != ss_effects.used_half_size_last_frame || use_mips != ss_effects.used_mips_last_frame) { + if (ss_effects.downsample_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(ss_effects.downsample_uniform_set)) { + RD::get_singleton()->free(ss_effects.downsample_uniform_set); + ss_effects.downsample_uniform_set = RID(); + } + + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 0; + u.append_id(p_depth_mipmaps[depth_index + 1]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 1; + u.append_id(p_depth_mipmaps[depth_index + 2]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 2; + u.append_id(p_depth_mipmaps[depth_index + 3]); + uniforms.push_back(u); + } + if (use_full_mips) { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 3; + u.append_id(p_depth_mipmaps[4]); + uniforms.push_back(u); + } + ss_effects.downsample_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ss_effects.downsample_shader.version_get_shader(ss_effects.downsample_shader_version, use_full_mips ? 6 : 2), 2); + } + + float depth_linearize_mul = -p_projection.columns[3][2]; + float depth_linearize_add = p_projection.columns[2][2]; + if (depth_linearize_mul * depth_linearize_add < 0) { + depth_linearize_add = -depth_linearize_add; + } + + ss_effects.downsample_push_constant.orthogonal = p_projection.is_orthogonal(); + ss_effects.downsample_push_constant.z_near = depth_linearize_mul; + ss_effects.downsample_push_constant.z_far = depth_linearize_add; + if (ss_effects.downsample_push_constant.orthogonal) { + ss_effects.downsample_push_constant.z_near = p_projection.get_z_near(); + ss_effects.downsample_push_constant.z_far = p_projection.get_z_far(); + } + ss_effects.downsample_push_constant.pixel_size[0] = 1.0 / p_full_screen_size.x; + ss_effects.downsample_push_constant.pixel_size[1] = 1.0 / p_full_screen_size.y; + ss_effects.downsample_push_constant.radius_sq = 1.0; + + RID shader = ss_effects.downsample_shader.version_get_shader(ss_effects.downsample_shader_version, downsample_mode); + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::Uniform u_depth_buffer(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_depth_buffer })); + RD::Uniform u_depth_mipmaps(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_depth_mipmaps[depth_index + 0] })); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ss_effects.pipelines[downsample_mode]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_depth_buffer), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_depth_mipmaps), 1); + if (use_mips) { + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, ss_effects.downsample_uniform_set, 2); + } + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ss_effects.downsample_push_constant, sizeof(SSEffectsDownsamplePushConstant)); + + Size2i size(MAX(1, p_full_screen_size.x >> (use_half_size ? 2 : 1)), MAX(1, p_full_screen_size.y >> (use_half_size ? 2 : 1))); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, size.x, size.y, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + RD::get_singleton()->draw_command_end_label(); + + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_COMPUTE); + + ss_effects.used_full_mips_last_frame = use_full_mips; + ss_effects.used_half_size_last_frame = use_half_size; + ss_effects.used_mips_last_frame = use_mips; +} + +/* SSIL */ + +void SSEffects::ssil_set_quality(RS::EnvironmentSSILQuality p_quality, bool p_half_size, float p_adaptive_target, int p_blur_passes, float p_fadeout_from, float p_fadeout_to) { + ssil_quality = p_quality; + ssil_half_size = p_half_size; + ssil_adaptive_target = p_adaptive_target; + ssil_blur_passes = p_blur_passes; + ssil_fadeout_from = p_fadeout_from; + ssil_fadeout_to = p_fadeout_to; +} + +void SSEffects::gather_ssil(RD::ComputeListID p_compute_list, const Vector<RID> p_ssil_slices, const Vector<RID> p_edges_slices, const SSILSettings &p_settings, bool p_adaptive_base_pass, RID p_gather_uniform_set, RID p_importance_map_uniform_set, RID p_projection_uniform_set) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + + RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, p_gather_uniform_set, 0); + if ((ssil_quality == RS::ENV_SSIL_QUALITY_ULTRA) && !p_adaptive_base_pass) { + RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, p_importance_map_uniform_set, 1); + } + RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, p_projection_uniform_set, 3); + + RID shader = ssil.gather_shader.version_get_shader(ssil.gather_shader_version, 0); + + for (int i = 0; i < 4; i++) { + if ((ssil_quality == RS::ENV_SSIL_QUALITY_VERY_LOW) && ((i == 1) || (i == 2))) { + continue; + } + + RD::Uniform u_ssil_slice(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssil_slices[i] })); + RD::Uniform u_edges_slice(RD::UNIFORM_TYPE_IMAGE, 1, Vector<RID>({ p_edges_slices[i] })); + + ssil.gather_push_constant.pass_coord_offset[0] = i % 2; + ssil.gather_push_constant.pass_coord_offset[1] = i / 2; + ssil.gather_push_constant.pass_uv_offset[0] = ((i % 2) - 0.0) / p_settings.full_screen_size.x; + ssil.gather_push_constant.pass_uv_offset[1] = ((i / 2) - 0.0) / p_settings.full_screen_size.y; + ssil.gather_push_constant.pass = i; + RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, uniform_set_cache->get_cache(shader, 2, u_ssil_slice, u_edges_slice), 2); + RD::get_singleton()->compute_list_set_push_constant(p_compute_list, &ssil.gather_push_constant, sizeof(SSILGatherPushConstant)); + + Size2i size = Size2i(p_settings.full_screen_size.x >> (ssil_half_size ? 2 : 1), p_settings.full_screen_size.y >> (ssil_half_size ? 2 : 1)); + + RD::get_singleton()->compute_list_dispatch_threads(p_compute_list, size.x, size.y, 1); + } + RD::get_singleton()->compute_list_add_barrier(p_compute_list); +} + +void SSEffects::ssil_allocate_buffers(SSILRenderBuffers &p_ssil_buffers, const SSILSettings &p_settings, RID p_linear_depth) { + if (p_ssil_buffers.half_size != ssil_half_size) { + ssil_free(p_ssil_buffers); + } + + if (ssil_half_size) { + p_ssil_buffers.buffer_width = (p_settings.full_screen_size.x + 3) / 4; + p_ssil_buffers.buffer_height = (p_settings.full_screen_size.y + 3) / 4; + p_ssil_buffers.half_buffer_width = (p_settings.full_screen_size.x + 7) / 8; + p_ssil_buffers.half_buffer_height = (p_settings.full_screen_size.y + 7) / 8; + } else { + p_ssil_buffers.buffer_width = (p_settings.full_screen_size.x + 1) / 2; + p_ssil_buffers.buffer_height = (p_settings.full_screen_size.y + 1) / 2; + p_ssil_buffers.half_buffer_width = (p_settings.full_screen_size.x + 3) / 4; + p_ssil_buffers.half_buffer_height = (p_settings.full_screen_size.y + 3) / 4; + } + + if (p_ssil_buffers.ssil_final.is_null()) { + { + p_ssil_buffers.depth_texture_view = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_linear_depth, 0, ssil_half_size ? 1 : 0, 4, RD::TEXTURE_SLICE_2D_ARRAY); + } + { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.width = p_settings.full_screen_size.x; + tf.height = p_settings.full_screen_size.y; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; + p_ssil_buffers.ssil_final = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssil_buffers.ssil_final, "SSIL texture"); + RD::get_singleton()->texture_clear(p_ssil_buffers.ssil_final, Color(0, 0, 0, 0), 0, 1, 0, 1); + if (p_ssil_buffers.last_frame.is_null()) { + tf.mipmaps = 6; + p_ssil_buffers.last_frame = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssil_buffers.last_frame, "Last Frame Radiance"); + RD::get_singleton()->texture_clear(p_ssil_buffers.last_frame, Color(0, 0, 0, 0), 0, tf.mipmaps, 0, 1); + for (uint32_t i = 0; i < 6; i++) { + RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_ssil_buffers.last_frame, 0, i); + p_ssil_buffers.last_frame_slices.push_back(slice); + RD::get_singleton()->set_resource_name(slice, "Last Frame Radiance Mip " + itos(i) + " "); + } + } + } + { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; + tf.width = p_ssil_buffers.buffer_width; + tf.height = p_ssil_buffers.buffer_height; + tf.array_layers = 4; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + p_ssil_buffers.deinterleaved = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssil_buffers.deinterleaved, "SSIL deinterleaved buffer"); + for (uint32_t i = 0; i < 4; i++) { + RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_ssil_buffers.deinterleaved, i, 0); + p_ssil_buffers.deinterleaved_slices.push_back(slice); + RD::get_singleton()->set_resource_name(slice, "SSIL deinterleaved buffer array " + itos(i) + " "); + } + } + + { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; + tf.width = p_ssil_buffers.buffer_width; + tf.height = p_ssil_buffers.buffer_height; + tf.array_layers = 4; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + p_ssil_buffers.pong = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssil_buffers.pong, "SSIL deinterleaved pong buffer"); + for (uint32_t i = 0; i < 4; i++) { + RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_ssil_buffers.pong, i, 0); + p_ssil_buffers.pong_slices.push_back(slice); + RD::get_singleton()->set_resource_name(slice, "SSIL deinterleaved buffer pong array " + itos(i) + " "); + } + } + + { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R8_UNORM; + tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; + tf.width = p_ssil_buffers.buffer_width; + tf.height = p_ssil_buffers.buffer_height; + tf.array_layers = 4; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + p_ssil_buffers.edges = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssil_buffers.edges, "SSIL edges buffer"); + for (uint32_t i = 0; i < 4; i++) { + RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_ssil_buffers.edges, i, 0); + p_ssil_buffers.edges_slices.push_back(slice); + RD::get_singleton()->set_resource_name(slice, "SSIL edges buffer slice " + itos(i) + " "); + } + } + + { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R8_UNORM; + tf.width = p_ssil_buffers.half_buffer_width; + tf.height = p_ssil_buffers.half_buffer_height; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + p_ssil_buffers.importance_map[0] = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssil_buffers.importance_map[0], "SSIL Importance Map"); + p_ssil_buffers.importance_map[1] = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssil_buffers.importance_map[1], "SSIL Importance Map Pong"); + } + p_ssil_buffers.half_size = ssil_half_size; + } +} + +void SSEffects::screen_space_indirect_lighting(SSILRenderBuffers &p_ssil_buffers, RID p_normal_buffer, const Projection &p_projection, const Projection &p_last_projection, const SSILSettings &p_settings) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + RD::get_singleton()->draw_command_begin_label("Process Screen Space Indirect Lighting"); + //Store projection info before starting the compute list + SSILProjectionUniforms projection_uniforms; + store_camera(p_last_projection, projection_uniforms.inv_last_frame_projection_matrix); + + RD::get_singleton()->buffer_update(ssil.projection_uniform_buffer, 0, sizeof(SSILProjectionUniforms), &projection_uniforms); + + memset(&ssil.gather_push_constant, 0, sizeof(SSILGatherPushConstant)); + + RID shader = ssil.gather_shader.version_get_shader(ssil.gather_shader_version, 0); + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + RID default_mipmap_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + { + RD::get_singleton()->draw_command_begin_label("Gather Samples"); + ssil.gather_push_constant.screen_size[0] = p_settings.full_screen_size.x; + ssil.gather_push_constant.screen_size[1] = p_settings.full_screen_size.y; + + ssil.gather_push_constant.half_screen_pixel_size[0] = 1.0 / p_ssil_buffers.buffer_width; + ssil.gather_push_constant.half_screen_pixel_size[1] = 1.0 / p_ssil_buffers.buffer_height; + float tan_half_fov_x = 1.0 / p_projection.columns[0][0]; + float tan_half_fov_y = 1.0 / p_projection.columns[1][1]; + ssil.gather_push_constant.NDC_to_view_mul[0] = tan_half_fov_x * 2.0; + ssil.gather_push_constant.NDC_to_view_mul[1] = tan_half_fov_y * -2.0; + ssil.gather_push_constant.NDC_to_view_add[0] = tan_half_fov_x * -1.0; + ssil.gather_push_constant.NDC_to_view_add[1] = tan_half_fov_y; + ssil.gather_push_constant.z_near = p_projection.get_z_near(); + ssil.gather_push_constant.z_far = p_projection.get_z_far(); + ssil.gather_push_constant.is_orthogonal = p_projection.is_orthogonal(); + + ssil.gather_push_constant.half_screen_pixel_size_x025[0] = ssil.gather_push_constant.half_screen_pixel_size[0] * 0.25; + ssil.gather_push_constant.half_screen_pixel_size_x025[1] = ssil.gather_push_constant.half_screen_pixel_size[1] * 0.25; + + ssil.gather_push_constant.radius = p_settings.radius; + float radius_near_limit = (p_settings.radius * 1.2f); + if (ssil_quality <= RS::ENV_SSIL_QUALITY_LOW) { + radius_near_limit *= 1.50f; + + if (ssil_quality == RS::ENV_SSIL_QUALITY_VERY_LOW) { + ssil.gather_push_constant.radius *= 0.8f; + } + } + radius_near_limit /= tan_half_fov_y; + ssil.gather_push_constant.intensity = p_settings.intensity * Math_PI; + ssil.gather_push_constant.fade_out_mul = -1.0 / (ssil_fadeout_to - ssil_fadeout_from); + ssil.gather_push_constant.fade_out_add = ssil_fadeout_from / (ssil_fadeout_to - ssil_fadeout_from) + 1.0; + ssil.gather_push_constant.inv_radius_near_limit = 1.0f / radius_near_limit; + ssil.gather_push_constant.neg_inv_radius = -1.0 / ssil.gather_push_constant.radius; + ssil.gather_push_constant.normal_rejection_amount = p_settings.normal_rejection; + + ssil.gather_push_constant.load_counter_avg_div = 9.0 / float((p_ssil_buffers.half_buffer_width) * (p_ssil_buffers.half_buffer_height) * 255); + ssil.gather_push_constant.adaptive_sample_limit = ssil_adaptive_target; + + ssil.gather_push_constant.quality = MAX(0, ssil_quality - 1); + ssil.gather_push_constant.size_multiplier = ssil_half_size ? 2 : 1; + + if (p_ssil_buffers.projection_uniform_set.is_null()) { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; + u.binding = 0; + u.append_id(default_mipmap_sampler); + u.append_id(p_ssil_buffers.last_frame); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 1; + u.append_id(ssil.projection_uniform_buffer); + uniforms.push_back(u); + } + p_ssil_buffers.projection_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssil.gather_shader.version_get_shader(ssil.gather_shader_version, 0), 3); + } + + if (p_ssil_buffers.gather_uniform_set.is_null()) { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; + u.binding = 0; + u.append_id(default_sampler); + u.append_id(p_ssil_buffers.depth_texture_view); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 1; + u.append_id(p_normal_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 2; + u.append_id(ss_effects.gather_constants_buffer); + uniforms.push_back(u); + } + p_ssil_buffers.gather_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssil.gather_shader.version_get_shader(ssil.gather_shader_version, 0), 0); + } + + if (p_ssil_buffers.importance_map_uniform_set.is_null()) { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 0; + u.append_id(p_ssil_buffers.pong); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; + u.binding = 1; + u.append_id(default_sampler); + u.append_id(p_ssil_buffers.importance_map[0]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 2; + u.append_id(ssil.importance_map_load_counter); + uniforms.push_back(u); + } + p_ssil_buffers.importance_map_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssil.gather_shader.version_get_shader(ssil.gather_shader_version, 2), 1); + } + + if (ssil_quality == RS::ENV_SSIL_QUALITY_ULTRA) { + RD::get_singleton()->draw_command_begin_label("Generate Importance Map"); + ssil.importance_map_push_constant.half_screen_pixel_size[0] = 1.0 / p_ssil_buffers.buffer_width; + ssil.importance_map_push_constant.half_screen_pixel_size[1] = 1.0 / p_ssil_buffers.buffer_height; + ssil.importance_map_push_constant.intensity = p_settings.intensity * Math_PI; + //base pass + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssil.pipelines[SSIL_GATHER_BASE]); + gather_ssil(compute_list, p_ssil_buffers.pong_slices, p_ssil_buffers.edges_slices, p_settings, true, p_ssil_buffers.gather_uniform_set, p_ssil_buffers.importance_map_uniform_set, p_ssil_buffers.projection_uniform_set); + + //generate importance map + RD::Uniform u_ssil_pong_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssil_buffers.pong })); + RD::Uniform u_importance_map(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssil_buffers.importance_map[0] })); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssil.pipelines[SSIL_GENERATE_IMPORTANCE_MAP]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_ssil_pong_with_sampler), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_importance_map), 1); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssil.importance_map_push_constant, sizeof(SSILImportanceMapPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_ssil_buffers.half_buffer_width, p_ssil_buffers.half_buffer_height, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + // process Importance Map A + RD::Uniform u_importance_map_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssil_buffers.importance_map[0] })); + RD::Uniform u_importance_map_pong(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssil_buffers.importance_map[1] })); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssil.pipelines[SSIL_PROCESS_IMPORTANCE_MAPA]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_importance_map_with_sampler), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_importance_map_pong), 1); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssil.importance_map_push_constant, sizeof(SSILImportanceMapPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_ssil_buffers.half_buffer_width, p_ssil_buffers.half_buffer_height, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + // process Importance Map B + RD::Uniform u_importance_map_pong_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssil_buffers.importance_map[1] })); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssil.pipelines[SSIL_PROCESS_IMPORTANCE_MAPB]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_importance_map_pong_with_sampler), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_importance_map), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, ssil.counter_uniform_set, 2); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssil.importance_map_push_constant, sizeof(SSILImportanceMapPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_ssil_buffers.half_buffer_width, p_ssil_buffers.half_buffer_height, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + RD::get_singleton()->draw_command_end_label(); // Importance Map + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssil.pipelines[SSIL_GATHER_ADAPTIVE]); + } else { + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssil.pipelines[SSIL_GATHER]); + } + + gather_ssil(compute_list, p_ssil_buffers.deinterleaved_slices, p_ssil_buffers.edges_slices, p_settings, false, p_ssil_buffers.gather_uniform_set, p_ssil_buffers.importance_map_uniform_set, p_ssil_buffers.projection_uniform_set); + RD::get_singleton()->draw_command_end_label(); //Gather + } + + { + RD::get_singleton()->draw_command_begin_label("Edge Aware Blur"); + ssil.blur_push_constant.edge_sharpness = 1.0 - p_settings.sharpness; + ssil.blur_push_constant.half_screen_pixel_size[0] = 1.0 / p_ssil_buffers.buffer_width; + ssil.blur_push_constant.half_screen_pixel_size[1] = 1.0 / p_ssil_buffers.buffer_height; + + int blur_passes = ssil_quality > RS::ENV_SSIL_QUALITY_VERY_LOW ? ssil_blur_passes : 1; + + shader = ssil.blur_shader.version_get_shader(ssil.blur_shader_version, 0); + + for (int pass = 0; pass < blur_passes; pass++) { + int blur_pipeline = SSIL_BLUR_PASS; + if (ssil_quality > RS::ENV_SSIL_QUALITY_VERY_LOW) { + blur_pipeline = SSIL_BLUR_PASS_SMART; + if (pass < blur_passes - 2) { + blur_pipeline = SSIL_BLUR_PASS_WIDE; + } + } + + for (int i = 0; i < 4; i++) { + if ((ssil_quality == RS::ENV_SSIL_QUALITY_VERY_LOW) && ((i == 1) || (i == 2))) { + continue; + } + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssil.pipelines[blur_pipeline]); + if (pass % 2 == 0) { + if (ssil_quality == RS::ENV_SSIL_QUALITY_VERY_LOW) { + RD::Uniform u_ssil_slice(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssil_buffers.deinterleaved_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_ssil_slice), 0); + } else { + RD::Uniform u_ssil_slice(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ ss_effects.mirror_sampler, p_ssil_buffers.deinterleaved_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_ssil_slice), 0); + } + + RD::Uniform u_ssil_pong_slice(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssil_buffers.pong_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_ssil_pong_slice), 1); + } else { + if (ssil_quality == RS::ENV_SSIL_QUALITY_VERY_LOW) { + RD::Uniform u_ssil_pong_slice(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssil_buffers.pong_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_ssil_pong_slice), 0); + } else { + RD::Uniform u_ssil_pong_slice(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ ss_effects.mirror_sampler, p_ssil_buffers.pong_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_ssil_pong_slice), 0); + } + + RD::Uniform u_ssil_slice(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssil_buffers.deinterleaved_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_ssil_slice), 1); + } + + RD::Uniform u_edges_slice(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssil_buffers.edges_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 2, u_edges_slice), 2); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssil.blur_push_constant, sizeof(SSILBlurPushConstant)); + + int x_groups = (p_settings.full_screen_size.x >> (ssil_half_size ? 2 : 1)); + int y_groups = (p_settings.full_screen_size.y >> (ssil_half_size ? 2 : 1)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, x_groups, y_groups, 1); + if (ssil_quality > RS::ENV_SSIL_QUALITY_VERY_LOW) { + RD::get_singleton()->compute_list_add_barrier(compute_list); + } + } + } + + RD::get_singleton()->draw_command_end_label(); // Blur + } + + { + RD::get_singleton()->draw_command_begin_label("Interleave Buffers"); + ssil.interleave_push_constant.inv_sharpness = 1.0 - p_settings.sharpness; + ssil.interleave_push_constant.pixel_size[0] = 1.0 / p_settings.full_screen_size.x; + ssil.interleave_push_constant.pixel_size[1] = 1.0 / p_settings.full_screen_size.y; + ssil.interleave_push_constant.size_modifier = uint32_t(ssil_half_size ? 4 : 2); + + int interleave_pipeline = SSIL_INTERLEAVE_HALF; + if (ssil_quality == RS::ENV_SSIL_QUALITY_LOW) { + interleave_pipeline = SSIL_INTERLEAVE; + } else if (ssil_quality >= RS::ENV_SSIL_QUALITY_MEDIUM) { + interleave_pipeline = SSIL_INTERLEAVE_SMART; + } + + shader = ssil.interleave_shader.version_get_shader(ssil.interleave_shader_version, 0); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssil.pipelines[interleave_pipeline]); + + RD::Uniform u_destination(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssil_buffers.ssil_final })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_destination), 0); + + if (ssil_quality > RS::ENV_SSIL_QUALITY_VERY_LOW && ssil_blur_passes % 2 == 0) { + RD::Uniform u_ssil(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssil_buffers.deinterleaved })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_ssil), 1); + } else { + RD::Uniform u_ssil_pong(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssil_buffers.pong })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_ssil_pong), 1); + } + + RD::Uniform u_edges(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssil_buffers.edges })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 2, u_edges), 2); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssil.interleave_push_constant, sizeof(SSILInterleavePushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_settings.full_screen_size.x, p_settings.full_screen_size.y, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + RD::get_singleton()->draw_command_end_label(); // Interleave + } + + RD::get_singleton()->draw_command_end_label(); // SSIL + + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_NO_BARRIER); + + int zero[1] = { 0 }; + RD::get_singleton()->buffer_update(ssil.importance_map_load_counter, 0, sizeof(uint32_t), &zero, 0); //no barrier +} + +void SSEffects::ssil_free(SSILRenderBuffers &p_ssil_buffers) { + if (p_ssil_buffers.ssil_final.is_valid()) { + RD::get_singleton()->free(p_ssil_buffers.ssil_final); + RD::get_singleton()->free(p_ssil_buffers.deinterleaved); + RD::get_singleton()->free(p_ssil_buffers.pong); + RD::get_singleton()->free(p_ssil_buffers.edges); + RD::get_singleton()->free(p_ssil_buffers.importance_map[0]); + RD::get_singleton()->free(p_ssil_buffers.importance_map[1]); + RD::get_singleton()->free(p_ssil_buffers.last_frame); + + p_ssil_buffers.ssil_final = RID(); + p_ssil_buffers.deinterleaved = RID(); + p_ssil_buffers.pong = RID(); + p_ssil_buffers.edges = RID(); + p_ssil_buffers.deinterleaved_slices.clear(); + p_ssil_buffers.pong_slices.clear(); + p_ssil_buffers.edges_slices.clear(); + p_ssil_buffers.importance_map[0] = RID(); + p_ssil_buffers.importance_map[1] = RID(); + p_ssil_buffers.last_frame = RID(); + p_ssil_buffers.last_frame_slices.clear(); + + p_ssil_buffers.gather_uniform_set = RID(); + p_ssil_buffers.importance_map_uniform_set = RID(); + p_ssil_buffers.projection_uniform_set = RID(); + } +} + +/* SSAO */ + +void SSEffects::ssao_set_quality(RS::EnvironmentSSAOQuality p_quality, bool p_half_size, float p_adaptive_target, int p_blur_passes, float p_fadeout_from, float p_fadeout_to) { + ssao_quality = p_quality; + ssao_half_size = p_half_size; + ssao_adaptive_target = p_adaptive_target; + ssao_blur_passes = p_blur_passes; + ssao_fadeout_from = p_fadeout_from; + ssao_fadeout_to = p_fadeout_to; +} + +void SSEffects::gather_ssao(RD::ComputeListID p_compute_list, const Vector<RID> p_ao_slices, const SSAOSettings &p_settings, bool p_adaptive_base_pass, RID p_gather_uniform_set, RID p_importance_map_uniform_set) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + + RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, p_gather_uniform_set, 0); + if ((ssao_quality == RS::ENV_SSAO_QUALITY_ULTRA) && !p_adaptive_base_pass) { + RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, p_importance_map_uniform_set, 0); + } + + RID shader = ssao.gather_shader.version_get_shader(ssao.gather_shader_version, 1); // + + for (int i = 0; i < 4; i++) { + if ((ssao_quality == RS::ENV_SSAO_QUALITY_VERY_LOW) && ((i == 1) || (i == 2))) { + continue; + } + + RD::Uniform u_ao_slice(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ao_slices[i] })); + + ssao.gather_push_constant.pass_coord_offset[0] = i % 2; + ssao.gather_push_constant.pass_coord_offset[1] = i / 2; + ssao.gather_push_constant.pass_uv_offset[0] = ((i % 2) - 0.0) / p_settings.full_screen_size.x; + ssao.gather_push_constant.pass_uv_offset[1] = ((i / 2) - 0.0) / p_settings.full_screen_size.y; + ssao.gather_push_constant.pass = i; + RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, uniform_set_cache->get_cache(shader, 2, u_ao_slice), 2); + RD::get_singleton()->compute_list_set_push_constant(p_compute_list, &ssao.gather_push_constant, sizeof(SSAOGatherPushConstant)); + + Size2i size = Size2i(p_settings.full_screen_size.x >> (ssao_half_size ? 2 : 1), p_settings.full_screen_size.y >> (ssao_half_size ? 2 : 1)); + + RD::get_singleton()->compute_list_dispatch_threads(p_compute_list, size.x, size.y, 1); + } + RD::get_singleton()->compute_list_add_barrier(p_compute_list); +} + +void SSEffects::ssao_allocate_buffers(SSAORenderBuffers &p_ssao_buffers, const SSAOSettings &p_settings, RID p_linear_depth) { + if (p_ssao_buffers.half_size != ssao_half_size) { + ssao_free(p_ssao_buffers); + } + + if (ssao_half_size) { + p_ssao_buffers.buffer_width = (p_settings.full_screen_size.x + 3) / 4; + p_ssao_buffers.buffer_height = (p_settings.full_screen_size.y + 3) / 4; + p_ssao_buffers.half_buffer_width = (p_settings.full_screen_size.x + 7) / 8; + p_ssao_buffers.half_buffer_height = (p_settings.full_screen_size.y + 7) / 8; + } else { + p_ssao_buffers.buffer_width = (p_settings.full_screen_size.x + 1) / 2; + p_ssao_buffers.buffer_height = (p_settings.full_screen_size.y + 1) / 2; + p_ssao_buffers.half_buffer_width = (p_settings.full_screen_size.x + 3) / 4; + p_ssao_buffers.half_buffer_height = (p_settings.full_screen_size.y + 3) / 4; + } + + if (p_ssao_buffers.ao_deinterleaved.is_null()) { + { + p_ssao_buffers.depth_texture_view = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_linear_depth, 0, ssao_half_size ? 1 : 0, 4, RD::TEXTURE_SLICE_2D_ARRAY); + } + { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R8G8_UNORM; + tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; + tf.width = p_ssao_buffers.buffer_width; + tf.height = p_ssao_buffers.buffer_height; + tf.array_layers = 4; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + p_ssao_buffers.ao_deinterleaved = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssao_buffers.ao_deinterleaved, "SSAO De-interleaved Array"); + for (uint32_t i = 0; i < 4; i++) { + RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_ssao_buffers.ao_deinterleaved, i, 0); + p_ssao_buffers.ao_deinterleaved_slices.push_back(slice); + RD::get_singleton()->set_resource_name(slice, "SSAO De-interleaved Array Layer " + itos(i) + " "); + } + } + + { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R8G8_UNORM; + tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; + tf.width = p_ssao_buffers.buffer_width; + tf.height = p_ssao_buffers.buffer_height; + tf.array_layers = 4; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + p_ssao_buffers.ao_pong = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssao_buffers.ao_pong, "SSAO De-interleaved Array Pong"); + for (uint32_t i = 0; i < 4; i++) { + RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_ssao_buffers.ao_pong, i, 0); + p_ssao_buffers.ao_pong_slices.push_back(slice); + RD::get_singleton()->set_resource_name(slice, "SSAO De-interleaved Array Layer " + itos(i) + " Pong"); + } + } + + { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R8_UNORM; + tf.width = p_ssao_buffers.buffer_width; + tf.height = p_ssao_buffers.buffer_height; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + p_ssao_buffers.importance_map[0] = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssao_buffers.importance_map[0], "SSAO Importance Map"); + p_ssao_buffers.importance_map[1] = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssao_buffers.importance_map[1], "SSAO Importance Map Pong"); + } + { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R8_UNORM; + tf.width = p_settings.full_screen_size.x; + tf.height = p_settings.full_screen_size.y; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + p_ssao_buffers.ao_final = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssao_buffers.ao_final, "SSAO Final"); + } + p_ssao_buffers.half_size = ssao_half_size; + } +} + +void SSEffects::generate_ssao(SSAORenderBuffers &p_ssao_buffers, RID p_normal_buffer, const Projection &p_projection, const SSAOSettings &p_settings) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + memset(&ssao.gather_push_constant, 0, sizeof(SSAOGatherPushConstant)); + /* FIRST PASS */ + + RID shader = ssao.gather_shader.version_get_shader(ssao.gather_shader_version, 0); + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + RD::get_singleton()->draw_command_begin_label("Process Screen Space Ambient Occlusion"); + /* SECOND PASS */ + // Sample SSAO + { + RD::get_singleton()->draw_command_begin_label("Gather Samples"); + ssao.gather_push_constant.screen_size[0] = p_settings.full_screen_size.x; + ssao.gather_push_constant.screen_size[1] = p_settings.full_screen_size.y; + + ssao.gather_push_constant.half_screen_pixel_size[0] = 1.0 / p_ssao_buffers.buffer_width; + ssao.gather_push_constant.half_screen_pixel_size[1] = 1.0 / p_ssao_buffers.buffer_height; + float tan_half_fov_x = 1.0 / p_projection.columns[0][0]; + float tan_half_fov_y = 1.0 / p_projection.columns[1][1]; + ssao.gather_push_constant.NDC_to_view_mul[0] = tan_half_fov_x * 2.0; + ssao.gather_push_constant.NDC_to_view_mul[1] = tan_half_fov_y * -2.0; + ssao.gather_push_constant.NDC_to_view_add[0] = tan_half_fov_x * -1.0; + ssao.gather_push_constant.NDC_to_view_add[1] = tan_half_fov_y; + ssao.gather_push_constant.is_orthogonal = p_projection.is_orthogonal(); + + ssao.gather_push_constant.half_screen_pixel_size_x025[0] = ssao.gather_push_constant.half_screen_pixel_size[0] * 0.25; + ssao.gather_push_constant.half_screen_pixel_size_x025[1] = ssao.gather_push_constant.half_screen_pixel_size[1] * 0.25; + + ssao.gather_push_constant.radius = p_settings.radius; + float radius_near_limit = (p_settings.radius * 1.2f); + if (ssao_quality <= RS::ENV_SSAO_QUALITY_LOW) { + radius_near_limit *= 1.50f; + + if (ssao_quality == RS::ENV_SSAO_QUALITY_VERY_LOW) { + ssao.gather_push_constant.radius *= 0.8f; + } + } + radius_near_limit /= tan_half_fov_y; + ssao.gather_push_constant.intensity = p_settings.intensity; + ssao.gather_push_constant.shadow_power = p_settings.power; + ssao.gather_push_constant.shadow_clamp = 0.98; + ssao.gather_push_constant.fade_out_mul = -1.0 / (ssao_fadeout_to - ssao_fadeout_from); + ssao.gather_push_constant.fade_out_add = ssao_fadeout_from / (ssao_fadeout_to - ssao_fadeout_from) + 1.0; + ssao.gather_push_constant.horizon_angle_threshold = p_settings.horizon; + ssao.gather_push_constant.inv_radius_near_limit = 1.0f / radius_near_limit; + ssao.gather_push_constant.neg_inv_radius = -1.0 / ssao.gather_push_constant.radius; + + ssao.gather_push_constant.load_counter_avg_div = 9.0 / float((p_ssao_buffers.half_buffer_width) * (p_ssao_buffers.half_buffer_height) * 255); + ssao.gather_push_constant.adaptive_sample_limit = ssao_adaptive_target; + + ssao.gather_push_constant.detail_intensity = p_settings.detail; + ssao.gather_push_constant.quality = MAX(0, ssao_quality - 1); + ssao.gather_push_constant.size_multiplier = ssao_half_size ? 2 : 1; + + if (p_ssao_buffers.gather_uniform_set.is_null()) { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; + u.binding = 0; + u.append_id(default_sampler); + u.append_id(p_ssao_buffers.depth_texture_view); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 1; + u.append_id(p_normal_buffer); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 2; + u.append_id(ss_effects.gather_constants_buffer); + uniforms.push_back(u); + } + p_ssao_buffers.gather_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, shader, 0); + RD::get_singleton()->set_resource_name(p_ssao_buffers.gather_uniform_set, "SSAO Gather Uniform Set"); + } + + if (p_ssao_buffers.importance_map_uniform_set.is_null()) { + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_IMAGE; + u.binding = 0; + u.append_id(p_ssao_buffers.ao_pong); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; + u.binding = 1; + u.append_id(default_sampler); + u.append_id(p_ssao_buffers.importance_map[0]); + uniforms.push_back(u); + } + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 2; + u.append_id(ssao.importance_map_load_counter); + uniforms.push_back(u); + } + p_ssao_buffers.importance_map_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssao.gather_shader.version_get_shader(ssao.gather_shader_version, 2), 1); + RD::get_singleton()->set_resource_name(p_ssao_buffers.importance_map_uniform_set, "SSAO Importance Map Uniform Set"); + } + + if (ssao_quality == RS::ENV_SSAO_QUALITY_ULTRA) { + RD::get_singleton()->draw_command_begin_label("Generate Importance Map"); + ssao.importance_map_push_constant.half_screen_pixel_size[0] = 1.0 / p_ssao_buffers.buffer_width; + ssao.importance_map_push_constant.half_screen_pixel_size[1] = 1.0 / p_ssao_buffers.buffer_height; + ssao.importance_map_push_constant.intensity = p_settings.intensity; + ssao.importance_map_push_constant.power = p_settings.power; + + //base pass + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GATHER_BASE]); + gather_ssao(compute_list, p_ssao_buffers.ao_pong_slices, p_settings, true, p_ssao_buffers.gather_uniform_set, RID()); + + //generate importance map + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GENERATE_IMPORTANCE_MAP]); + + RD::Uniform u_ao_pong_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssao_buffers.ao_pong })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_ao_pong_with_sampler), 0); + + RD::Uniform u_importance_map(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssao_buffers.importance_map[0] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_importance_map), 1); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.importance_map_push_constant, sizeof(SSAOImportanceMapPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_ssao_buffers.half_buffer_width, p_ssao_buffers.half_buffer_height, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + //process importance map A + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_PROCESS_IMPORTANCE_MAPA]); + + RD::Uniform u_importance_map_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssao_buffers.importance_map[0] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_importance_map_with_sampler), 0); + + RD::Uniform u_importance_map_pong(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssao_buffers.importance_map[1] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_importance_map_pong), 1); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.importance_map_push_constant, sizeof(SSAOImportanceMapPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_ssao_buffers.half_buffer_width, p_ssao_buffers.half_buffer_height, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + //process Importance Map B + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_PROCESS_IMPORTANCE_MAPB]); + + RD::Uniform u_importance_map_pong_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssao_buffers.importance_map[1] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_importance_map_pong_with_sampler), 0); + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_importance_map), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, ssao.counter_uniform_set, 2); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.importance_map_push_constant, sizeof(SSAOImportanceMapPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_ssao_buffers.half_buffer_width, p_ssao_buffers.half_buffer_height, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GATHER_ADAPTIVE]); + RD::get_singleton()->draw_command_end_label(); // Importance Map + } else { + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GATHER]); + } + + gather_ssao(compute_list, p_ssao_buffers.ao_deinterleaved_slices, p_settings, false, p_ssao_buffers.gather_uniform_set, p_ssao_buffers.importance_map_uniform_set); + RD::get_singleton()->draw_command_end_label(); // Gather SSAO + } + + // /* THIRD PASS */ + // // Blur + // + { + RD::get_singleton()->draw_command_begin_label("Edge Aware Blur"); + ssao.blur_push_constant.edge_sharpness = 1.0 - p_settings.sharpness; + ssao.blur_push_constant.half_screen_pixel_size[0] = 1.0 / p_ssao_buffers.buffer_width; + ssao.blur_push_constant.half_screen_pixel_size[1] = 1.0 / p_ssao_buffers.buffer_height; + + int blur_passes = ssao_quality > RS::ENV_SSAO_QUALITY_VERY_LOW ? ssao_blur_passes : 1; + + shader = ssao.blur_shader.version_get_shader(ssao.blur_shader_version, 0); + + for (int pass = 0; pass < blur_passes; pass++) { + int blur_pipeline = SSAO_BLUR_PASS; + if (ssao_quality > RS::ENV_SSAO_QUALITY_VERY_LOW) { + blur_pipeline = SSAO_BLUR_PASS_SMART; + if (pass < blur_passes - 2) { + blur_pipeline = SSAO_BLUR_PASS_WIDE; + } else { + blur_pipeline = SSAO_BLUR_PASS_SMART; + } + } + + for (int i = 0; i < 4; i++) { + if ((ssao_quality == RS::ENV_SSAO_QUALITY_VERY_LOW) && ((i == 1) || (i == 2))) { + continue; + } + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[blur_pipeline]); + if (pass % 2 == 0) { + if (ssao_quality == RS::ENV_SSAO_QUALITY_VERY_LOW) { + RD::Uniform u_ao_slices_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssao_buffers.ao_deinterleaved_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_ao_slices_with_sampler), 0); + } else { + RD::Uniform u_ao_slices_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ ss_effects.mirror_sampler, p_ssao_buffers.ao_deinterleaved_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_ao_slices_with_sampler), 0); + } + + RD::Uniform u_ao_pong_slices(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssao_buffers.ao_pong_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_ao_pong_slices), 1); + } else { + if (ssao_quality == RS::ENV_SSAO_QUALITY_VERY_LOW) { + RD::Uniform u_ao_pong_slices_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssao_buffers.ao_pong_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_ao_pong_slices_with_sampler), 0); + } else { + RD::Uniform u_ao_pong_slices_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ ss_effects.mirror_sampler, p_ssao_buffers.ao_pong_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_ao_pong_slices_with_sampler), 0); + } + + RD::Uniform u_ao_slices(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssao_buffers.ao_deinterleaved_slices[i] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_ao_slices), 1); + } + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.blur_push_constant, sizeof(SSAOBlurPushConstant)); + + Size2i size(p_settings.full_screen_size.x >> (ssao_half_size ? 2 : 1), p_settings.full_screen_size.y >> (ssao_half_size ? 2 : 1)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, size.x, size.y, 1); + } + + if (ssao_quality > RS::ENV_SSAO_QUALITY_VERY_LOW) { + RD::get_singleton()->compute_list_add_barrier(compute_list); + } + } + RD::get_singleton()->draw_command_end_label(); // Blur + } + + /* FOURTH PASS */ + // Interleave buffers + // back to full size + { + RD::get_singleton()->draw_command_begin_label("Interleave Buffers"); + ssao.interleave_push_constant.inv_sharpness = 1.0 - p_settings.sharpness; + ssao.interleave_push_constant.pixel_size[0] = 1.0 / p_settings.full_screen_size.x; + ssao.interleave_push_constant.pixel_size[1] = 1.0 / p_settings.full_screen_size.y; + ssao.interleave_push_constant.size_modifier = uint32_t(ssao_half_size ? 4 : 2); + + shader = ssao.interleave_shader.version_get_shader(ssao.interleave_shader_version, 0); + + int interleave_pipeline = SSAO_INTERLEAVE_HALF; + if (ssao_quality == RS::ENV_SSAO_QUALITY_LOW) { + interleave_pipeline = SSAO_INTERLEAVE; + } else if (ssao_quality >= RS::ENV_SSAO_QUALITY_MEDIUM) { + interleave_pipeline = SSAO_INTERLEAVE_SMART; + } + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[interleave_pipeline]); + + RD::Uniform u_upscale_buffer(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssao_buffers.ao_final })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_upscale_buffer), 0); + + if (ssao_quality > RS::ENV_SSAO_QUALITY_VERY_LOW && ssao_blur_passes % 2 == 0) { + RD::Uniform u_ao(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssao_buffers.ao_deinterleaved })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_ao), 1); + } else { + RD::Uniform u_ao(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_ssao_buffers.ao_pong })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_ao), 1); + } + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.interleave_push_constant, sizeof(SSAOInterleavePushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_settings.full_screen_size.x, p_settings.full_screen_size.y, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + RD::get_singleton()->draw_command_end_label(); // Interleave + } + RD::get_singleton()->draw_command_end_label(); //SSAO + RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_NO_BARRIER); //wait for upcoming transfer + + int zero[1] = { 0 }; + RD::get_singleton()->buffer_update(ssao.importance_map_load_counter, 0, sizeof(uint32_t), &zero, 0); //no barrier +} + +void SSEffects::ssao_free(SSAORenderBuffers &p_ssao_buffers) { + if (p_ssao_buffers.ao_final.is_valid()) { + RD::get_singleton()->free(p_ssao_buffers.ao_deinterleaved); + RD::get_singleton()->free(p_ssao_buffers.ao_pong); + RD::get_singleton()->free(p_ssao_buffers.ao_final); + + RD::get_singleton()->free(p_ssao_buffers.importance_map[0]); + RD::get_singleton()->free(p_ssao_buffers.importance_map[1]); + + p_ssao_buffers.ao_deinterleaved = RID(); + p_ssao_buffers.ao_pong = RID(); + p_ssao_buffers.ao_final = RID(); + p_ssao_buffers.importance_map[0] = RID(); + p_ssao_buffers.importance_map[1] = RID(); + p_ssao_buffers.ao_deinterleaved_slices.clear(); + p_ssao_buffers.ao_pong_slices.clear(); + + p_ssao_buffers.gather_uniform_set = RID(); + p_ssao_buffers.importance_map_uniform_set = RID(); + } +} + +/* Screen Space Reflection */ + +void SSEffects::ssr_set_roughness_quality(RS::EnvironmentSSRRoughnessQuality p_quality) { + ssr_roughness_quality = p_quality; +} + +void SSEffects::ssr_allocate_buffers(SSRRenderBuffers &p_ssr_buffers, const RenderingDevice::DataFormat p_color_format, const Size2i &p_screen_size, const uint32_t p_view_count) { + // As we are processing one view at a time, we can reuse buffers, only our output needs to have layers for each view. + if (p_ssr_buffers.size != p_screen_size || p_ssr_buffers.roughness_quality != ssr_roughness_quality) { + ssr_free(p_ssr_buffers); + } + + if (p_ssr_buffers.output.is_valid()) { + // already allocated + return; + } + + p_ssr_buffers.size = p_screen_size; + p_ssr_buffers.roughness_quality = ssr_roughness_quality; + + if (p_ssr_buffers.depth_scaled.is_null()) { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R32_SFLOAT; + tf.width = p_screen_size.x; + tf.height = p_screen_size.y; + tf.texture_type = RD::TEXTURE_TYPE_2D; + tf.array_layers = 1; + tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT; + + p_ssr_buffers.depth_scaled = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssr_buffers.depth_scaled, "SSR Depth Scaled"); + + tf.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + + p_ssr_buffers.normal_scaled = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssr_buffers.normal_scaled, "SSR Normal Scaled"); + } + + if (ssr_roughness_quality != RS::ENV_SSR_ROUGHNESS_QUALITY_DISABLED && !p_ssr_buffers.blur_radius[0].is_valid()) { + RD::TextureFormat tf; + tf.format = RD::DATA_FORMAT_R8_UNORM; + tf.width = p_screen_size.x; + tf.height = p_screen_size.y; + tf.texture_type = RD::TEXTURE_TYPE_2D; + tf.array_layers = 1; + tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; + + p_ssr_buffers.blur_radius[0] = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssr_buffers.blur_radius[0], "SSR Blur Radius 0"); + p_ssr_buffers.blur_radius[1] = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssr_buffers.blur_radius[1], "SSR Blur Radius 1"); + } + + if (p_ssr_buffers.intermediate.is_null()) { + RD::TextureFormat tf; + tf.format = p_color_format; + tf.width = p_screen_size.x; + tf.height = p_screen_size.y; + tf.texture_type = RD::TEXTURE_TYPE_2D; + tf.array_layers = 1; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + + p_ssr_buffers.intermediate = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssr_buffers.intermediate, "SSR Intermediate"); + + if (p_view_count > 1) { + tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; + tf.array_layers = p_view_count; + } else { + tf.texture_type = RD::TEXTURE_TYPE_2D; + tf.array_layers = 1; + } + + p_ssr_buffers.output = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(p_ssr_buffers.output, "SSR Output"); + + for (uint32_t v = 0; v < p_view_count; v++) { + p_ssr_buffers.output_slices[v] = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_ssr_buffers.output, v, 0); + } + } +} + +void SSEffects::screen_space_reflection(SSRRenderBuffers &p_ssr_buffers, const RID *p_diffuse_slices, const RID *p_normal_roughness_slices, const RID *p_metallic_slices, const RID *p_depth_slices, const Size2i &p_screen_size, int p_max_steps, float p_fade_in, float p_fade_out, float p_tolerance, const uint32_t p_view_count, const Projection *p_projections, const Vector3 *p_eye_offsets) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + { + // Store some scene data in a UBO, in the near future we will use a UBO shared with other shaders + ScreenSpaceReflectionSceneData scene_data; + + if (ssr.ubo.is_null()) { + ssr.ubo = RD::get_singleton()->uniform_buffer_create(sizeof(ScreenSpaceReflectionSceneData)); + } + + for (uint32_t v = 0; v < p_view_count; v++) { + store_camera(p_projections[v], scene_data.projection[v]); + store_camera(p_projections[v].inverse(), scene_data.inv_projection[v]); + scene_data.eye_offset[v][0] = p_eye_offsets[v].x; + scene_data.eye_offset[v][1] = p_eye_offsets[v].y; + scene_data.eye_offset[v][2] = p_eye_offsets[v].z; + scene_data.eye_offset[v][3] = 0.0; + } + + RD::get_singleton()->buffer_update(ssr.ubo, 0, sizeof(ScreenSpaceReflectionSceneData), &scene_data, RD::BARRIER_MASK_COMPUTE); + } + + uint32_t pipeline_specialization = 0; + if (p_view_count > 1) { + pipeline_specialization |= SSR_MULTIVIEW; + } + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + + for (uint32_t v = 0; v < p_view_count; v++) { + RD::get_singleton()->draw_command_begin_label(String("SSR View ") + itos(v)); + + { //scale color and depth to half + RD::get_singleton()->draw_command_begin_label("SSR Scale"); + + ScreenSpaceReflectionScalePushConstant push_constant; + push_constant.view_index = v; + push_constant.camera_z_far = p_projections[v].get_z_far(); + push_constant.camera_z_near = p_projections[v].get_z_near(); + push_constant.orthogonal = p_projections[v].is_orthogonal(); + push_constant.filter = false; //enabling causes arctifacts + push_constant.screen_size[0] = p_screen_size.x; + push_constant.screen_size[1] = p_screen_size.y; + + RID shader = ssr_scale.shader.version_get_shader(ssr_scale.shader_version, 0); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssr_scale.pipelines[pipeline_specialization]); + + RD::Uniform u_diffuse(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_diffuse_slices[v] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_diffuse), 0); + + RD::Uniform u_depth(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_depth_slices[v] })); + RD::Uniform u_normal_roughness(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 1, Vector<RID>({ default_sampler, p_normal_roughness_slices[v] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_depth, u_normal_roughness), 1); + + if (ssr_roughness_quality != RS::ENV_SSR_ROUGHNESS_QUALITY_DISABLED) { + RD::Uniform u_output(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssr_buffers.output_slices[v] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 2, u_output), 2); + } else { + RD::Uniform u_intermediate(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssr_buffers.intermediate })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 2, u_intermediate), 2); + } + + RD::Uniform u_scale_depth(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssr_buffers.depth_scaled })); + RD::Uniform u_scale_normal(RD::UNIFORM_TYPE_IMAGE, 1, Vector<RID>({ p_ssr_buffers.normal_scaled })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 3, u_scale_depth, u_scale_normal), 3); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ScreenSpaceReflectionScalePushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); + + RD::get_singleton()->compute_list_add_barrier(compute_list); + + RD::get_singleton()->draw_command_end_label(); + } + + { + RD::get_singleton()->draw_command_begin_label("SSR main"); + + ScreenSpaceReflectionPushConstant push_constant; + push_constant.view_index = v; + push_constant.camera_z_far = p_projections[v].get_z_far(); + push_constant.camera_z_near = p_projections[v].get_z_near(); + push_constant.orthogonal = p_projections[v].is_orthogonal(); + push_constant.screen_size[0] = p_screen_size.x; + push_constant.screen_size[1] = p_screen_size.y; + push_constant.curve_fade_in = p_fade_in; + push_constant.distance_fade = p_fade_out; + push_constant.num_steps = p_max_steps; + push_constant.depth_tolerance = p_tolerance; + push_constant.use_half_res = true; + push_constant.proj_info[0] = -2.0f / (p_screen_size.width * p_projections[v].columns[0][0]); + push_constant.proj_info[1] = -2.0f / (p_screen_size.height * p_projections[v].columns[1][1]); + push_constant.proj_info[2] = (1.0f - p_projections[v].columns[0][2]) / p_projections[v].columns[0][0]; + push_constant.proj_info[3] = (1.0f + p_projections[v].columns[1][2]) / p_projections[v].columns[1][1]; + + ScreenSpaceReflectionMode mode = (ssr_roughness_quality != RS::ENV_SSR_ROUGHNESS_QUALITY_DISABLED) ? SCREEN_SPACE_REFLECTION_ROUGH : SCREEN_SPACE_REFLECTION_NORMAL; + RID shader = ssr.shader.version_get_shader(ssr.shader_version, mode); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssr.pipelines[pipeline_specialization][mode]); + + RD::Uniform u_scene_data(RD::UNIFORM_TYPE_UNIFORM_BUFFER, 0, ssr.ubo); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 4, u_scene_data), 4); + + if (ssr_roughness_quality != RS::ENV_SSR_ROUGHNESS_QUALITY_DISABLED) { + // read from output slices (our scale wrote into these) + RD::Uniform u_output(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssr_buffers.output_slices[v] })); + RD::Uniform u_scale_depth(RD::UNIFORM_TYPE_IMAGE, 1, Vector<RID>({ p_ssr_buffers.depth_scaled })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_output, u_scale_depth), 0); + + // write to intermediate (our roughness pass will output into output slices) + RD::Uniform u_intermediate(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssr_buffers.intermediate })); + RD::Uniform u_blur_radius(RD::UNIFORM_TYPE_IMAGE, 1, Vector<RID>({ p_ssr_buffers.blur_radius[0] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_intermediate, u_blur_radius), 1); + } else { + // read from intermediate (our scale wrote into these) + RD::Uniform u_intermediate(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssr_buffers.intermediate })); + RD::Uniform u_scale_depth(RD::UNIFORM_TYPE_IMAGE, 1, Vector<RID>({ p_ssr_buffers.depth_scaled })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_intermediate, u_scale_depth), 0); + + // We are not performing our blur so go directly to output. + RD::Uniform u_output(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssr_buffers.output_slices[v] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_output), 1); + } + + RD::Uniform u_scale_normal(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssr_buffers.normal_scaled })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 2, u_scale_normal), 2); + + RD::Uniform u_metallic(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_metallic_slices[v] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 3, u_metallic), 3); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ScreenSpaceReflectionPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); + + RD::get_singleton()->draw_command_end_label(); + } + + if (ssr_roughness_quality != RS::ENV_SSR_ROUGHNESS_QUALITY_DISABLED) { + RD::get_singleton()->draw_command_begin_label("SSR filter"); + //blur + + RD::get_singleton()->compute_list_add_barrier(compute_list); + + ScreenSpaceReflectionFilterPushConstant push_constant; + push_constant.view_index = v; + push_constant.orthogonal = p_projections[v].is_orthogonal(); + push_constant.edge_tolerance = Math::sin(Math::deg_to_rad(15.0)); + push_constant.proj_info[0] = -2.0f / (p_screen_size.width * p_projections[v].columns[0][0]); + push_constant.proj_info[1] = -2.0f / (p_screen_size.height * p_projections[v].columns[1][1]); + push_constant.proj_info[2] = (1.0f - p_projections[v].columns[0][2]) / p_projections[v].columns[0][0]; + push_constant.proj_info[3] = (1.0f + p_projections[v].columns[1][2]) / p_projections[v].columns[1][1]; + push_constant.vertical = 0; + if (ssr_roughness_quality == RS::ENV_SSR_ROUGHNESS_QUALITY_LOW) { + push_constant.steps = p_max_steps / 3; + push_constant.increment = 3; + } else if (ssr_roughness_quality == RS::ENV_SSR_ROUGHNESS_QUALITY_MEDIUM) { + push_constant.steps = p_max_steps / 2; + push_constant.increment = 2; + } else { + push_constant.steps = p_max_steps; + push_constant.increment = 1; + } + + push_constant.screen_size[0] = p_screen_size.width; + push_constant.screen_size[1] = p_screen_size.height; + + // Horizontal pass + + SSRReflectionMode mode = SCREEN_SPACE_REFLECTION_FILTER_HORIZONTAL; + + RID shader = ssr_filter.shader.version_get_shader(ssr_filter.shader_version, mode); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssr_filter.pipelines[pipeline_specialization][mode]); + + RD::Uniform u_scene_data(RD::UNIFORM_TYPE_UNIFORM_BUFFER, 0, ssr.ubo); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 4, u_scene_data), 4); + + RD::Uniform u_intermediate(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssr_buffers.intermediate })); + RD::Uniform u_blur_radius(RD::UNIFORM_TYPE_IMAGE, 1, Vector<RID>({ p_ssr_buffers.blur_radius[0] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_intermediate, u_blur_radius), 0); + + RD::Uniform u_scale_normal(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssr_buffers.normal_scaled })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_scale_normal), 1); + + RD::Uniform u_output_blur(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssr_buffers.output_slices[v] })); + RD::Uniform u_blur_radius2(RD::UNIFORM_TYPE_IMAGE, 1, Vector<RID>({ p_ssr_buffers.blur_radius[1] })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 2, u_output_blur, u_blur_radius2), 2); + + RD::Uniform u_scale_depth(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_ssr_buffers.depth_scaled })); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 3, u_scale_depth), 3); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ScreenSpaceReflectionFilterPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); + RD::get_singleton()->compute_list_add_barrier(compute_list); + + // Vertical pass + + mode = SCREEN_SPACE_REFLECTION_FILTER_VERTICAL; + shader = ssr_filter.shader.version_get_shader(ssr_filter.shader_version, mode); + + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssr_filter.pipelines[pipeline_specialization][mode]); + + push_constant.vertical = 1; + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_output_blur, u_blur_radius2), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_scale_normal), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 2, u_intermediate), 2); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 3, u_scale_depth), 3); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 4, u_scene_data), 4); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ScreenSpaceReflectionFilterPushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); + + if (v != p_view_count - 1) { + RD::get_singleton()->compute_list_add_barrier(compute_list); + } + + RD::get_singleton()->draw_command_end_label(); + } + + RD::get_singleton()->draw_command_end_label(); + } + + RD::get_singleton()->compute_list_end(); +} + +void SSEffects::ssr_free(SSRRenderBuffers &p_ssr_buffers) { + for (uint32_t v = 0; v < RendererSceneRender::MAX_RENDER_VIEWS; v++) { + p_ssr_buffers.output_slices[v] = RID(); + } + + if (p_ssr_buffers.output.is_valid()) { + RD::get_singleton()->free(p_ssr_buffers.output); + p_ssr_buffers.output = RID(); + } + + if (p_ssr_buffers.intermediate.is_valid()) { + RD::get_singleton()->free(p_ssr_buffers.intermediate); + p_ssr_buffers.intermediate = RID(); + } + + if (p_ssr_buffers.blur_radius[0].is_valid()) { + RD::get_singleton()->free(p_ssr_buffers.blur_radius[0]); + RD::get_singleton()->free(p_ssr_buffers.blur_radius[1]); + p_ssr_buffers.blur_radius[0] = RID(); + p_ssr_buffers.blur_radius[1] = RID(); + } + + if (p_ssr_buffers.depth_scaled.is_valid()) { + RD::get_singleton()->free(p_ssr_buffers.depth_scaled); + p_ssr_buffers.depth_scaled = RID(); + RD::get_singleton()->free(p_ssr_buffers.normal_scaled); + p_ssr_buffers.normal_scaled = RID(); + } +} + +/* Subsurface scattering */ + +void SSEffects::sss_set_quality(RS::SubSurfaceScatteringQuality p_quality) { + sss_quality = p_quality; +} + +RS::SubSurfaceScatteringQuality SSEffects::sss_get_quality() const { + return sss_quality; +} + +void SSEffects::sss_set_scale(float p_scale, float p_depth_scale) { + sss_scale = p_scale; + sss_depth_scale = p_depth_scale; +} + +void SSEffects::sub_surface_scattering(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_diffuse, RID p_depth, const Projection &p_camera, const Size2i &p_screen_size) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + // Our intermediate buffer is only created if we haven't created it already. + RD::DataFormat format = p_render_buffers->get_base_data_format(); + uint32_t usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + uint32_t layers = 1; // We only need one layer, we're handling one view at a time + uint32_t mipmaps = 1; // Image::get_image_required_mipmaps(p_screen_size.x, p_screen_size.y, Image::FORMAT_RGBAH); + RID intermediate = p_render_buffers->create_texture(SNAME("SSR"), SNAME("intermediate"), format, usage_bits, RD::TEXTURE_SAMPLES_1, p_screen_size, layers, mipmaps); + + Plane p = p_camera.xform4(Plane(1, 0, -1, 1)); + p.normal /= p.d; + float unit_size = p.normal.x; + + { //scale color and depth to half + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + + sss.push_constant.camera_z_far = p_camera.get_z_far(); + sss.push_constant.camera_z_near = p_camera.get_z_near(); + sss.push_constant.orthogonal = p_camera.is_orthogonal(); + sss.push_constant.unit_size = unit_size; + sss.push_constant.screen_size[0] = p_screen_size.x; + sss.push_constant.screen_size[1] = p_screen_size.y; + sss.push_constant.vertical = false; + sss.push_constant.scale = sss_scale; + sss.push_constant.depth_scale = sss_depth_scale; + + RID shader = sss.shader.version_get_shader(sss.shader_version, sss_quality - 1); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sss.pipelines[sss_quality - 1]); + + RD::Uniform u_diffuse_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_diffuse })); + RD::Uniform u_diffuse(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ p_diffuse })); + RD::Uniform u_intermediate_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, intermediate })); + RD::Uniform u_intermediate(RD::UNIFORM_TYPE_IMAGE, 0, Vector<RID>({ intermediate })); + RD::Uniform u_depth_with_sampler(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 0, Vector<RID>({ default_sampler, p_depth })); + + // horizontal + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_diffuse_with_sampler), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_intermediate), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 2, u_depth_with_sampler), 2); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &sss.push_constant, sizeof(SubSurfaceScatteringPushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); + + RD::get_singleton()->compute_list_add_barrier(compute_list); + + // vertical + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_intermediate_with_sampler), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 1, u_diffuse), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 2, u_depth_with_sampler), 2); + + sss.push_constant.vertical = true; + RD::get_singleton()->compute_list_set_push_constant(compute_list, &sss.push_constant, sizeof(SubSurfaceScatteringPushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); + + RD::get_singleton()->compute_list_end(); + } +} diff --git a/servers/rendering/renderer_rd/effects/ss_effects.h b/servers/rendering/renderer_rd/effects/ss_effects.h new file mode 100644 index 0000000000..dfaf3881bb --- /dev/null +++ b/servers/rendering/renderer_rd/effects/ss_effects.h @@ -0,0 +1,552 @@ +/*************************************************************************/ +/* ss_effects.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef SS_EFFECTS_RD_H +#define SS_EFFECTS_RD_H + +#include "servers/rendering/renderer_rd/pipeline_cache_rd.h" +#include "servers/rendering/renderer_rd/shaders/effects/screen_space_reflection.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/screen_space_reflection_filter.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/screen_space_reflection_scale.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/ss_effects_downsample.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/ssao.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/ssao_blur.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/ssao_importance_map.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/ssao_interleave.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/ssil.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/ssil_blur.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/ssil_importance_map.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/ssil_interleave.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/effects/subsurface_scattering.glsl.gen.h" +#include "servers/rendering/renderer_scene_render.h" +#include "servers/rendering_server.h" + +class RenderSceneBuffersRD; + +namespace RendererRD { + +class SSEffects { +private: + static SSEffects *singleton; + +public: + static SSEffects *get_singleton() { return singleton; } + + SSEffects(); + ~SSEffects(); + + /* SS Downsampler */ + + void downsample_depth(RID p_depth_buffer, const Vector<RID> &p_depth_mipmaps, bool p_invalidate_uniform_set, Size2i p_full_screen_size, const Projection &p_projection); + + /* SSIL */ + void ssil_set_quality(RS::EnvironmentSSILQuality p_quality, bool p_half_size, float p_adaptive_target, int p_blur_passes, float p_fadeout_from, float p_fadeout_to); + + struct SSILRenderBuffers { + bool half_size = false; + int buffer_width; + int buffer_height; + int half_buffer_width; + int half_buffer_height; + + RID ssil_final; + RID deinterleaved; + Vector<RID> deinterleaved_slices; + RID pong; + Vector<RID> pong_slices; + RID edges; + Vector<RID> edges_slices; + RID importance_map[2]; + RID depth_texture_view; + + RID last_frame; + Vector<RID> last_frame_slices; + + RID gather_uniform_set; + RID importance_map_uniform_set; + RID projection_uniform_set; + }; + + struct SSILSettings { + float radius = 1.0; + float intensity = 2.0; + float sharpness = 0.98; + float normal_rejection = 1.0; + + Size2i full_screen_size = Size2i(); + }; + + void ssil_allocate_buffers(SSILRenderBuffers &p_ssil_buffers, const SSILSettings &p_settings, RID p_linear_depth); + void screen_space_indirect_lighting(SSILRenderBuffers &p_ssil_buffers, RID p_normal_buffer, const Projection &p_projection, const Projection &p_last_projection, const SSILSettings &p_settings); + void ssil_free(SSILRenderBuffers &p_ssil_buffers); + + /* SSAO */ + void ssao_set_quality(RS::EnvironmentSSAOQuality p_quality, bool p_half_size, float p_adaptive_target, int p_blur_passes, float p_fadeout_from, float p_fadeout_to); + + struct SSAORenderBuffers { + bool half_size = false; + int buffer_width; + int buffer_height; + int half_buffer_width; + int half_buffer_height; + + RID ao_deinterleaved; + Vector<RID> ao_deinterleaved_slices; + RID ao_pong; + Vector<RID> ao_pong_slices; + RID ao_final; + RID importance_map[2]; + RID depth_texture_view; + + RID gather_uniform_set; + RID importance_map_uniform_set; + }; + + struct SSAOSettings { + float radius = 1.0; + float intensity = 2.0; + float power = 1.5; + float detail = 0.5; + float horizon = 0.06; + float sharpness = 0.98; + + Size2i full_screen_size = Size2i(); + }; + + void ssao_allocate_buffers(SSAORenderBuffers &p_ssao_buffers, const SSAOSettings &p_settings, RID p_linear_depth); + void generate_ssao(SSAORenderBuffers &p_ssao_buffers, RID p_normal_buffer, const Projection &p_projection, const SSAOSettings &p_settings); + void ssao_free(SSAORenderBuffers &p_ssao_buffers); + + /* Screen Space Reflection */ + void ssr_set_roughness_quality(RS::EnvironmentSSRRoughnessQuality p_quality); + + struct SSRRenderBuffers { + Size2i size; + RenderingServer::EnvironmentSSRRoughnessQuality roughness_quality = RenderingServer::ENV_SSR_ROUGHNESS_QUALITY_DISABLED; + + RID normal_scaled; + RID depth_scaled; + RID blur_radius[2]; + RID intermediate; + RID output; + RID output_slices[RendererSceneRender::MAX_RENDER_VIEWS]; + }; + + void ssr_allocate_buffers(SSRRenderBuffers &p_ssr_buffers, const RenderingDevice::DataFormat p_color_format, const Size2i &p_screen_size, const uint32_t p_view_count); + void screen_space_reflection(SSRRenderBuffers &p_ssr_buffers, const RID *p_diffuse_slices, const RID *p_normal_roughness_slices, const RID *p_metallic_slices, const RID *p_depth_slices, const Size2i &p_screen_size, int p_max_steps, float p_fade_in, float p_fade_out, float p_tolerance, const uint32_t p_view_count, const Projection *p_projections, const Vector3 *p_eye_offsets); + void ssr_free(SSRRenderBuffers &p_ssr_buffers); + + /* subsurface scattering */ + void sss_set_quality(RS::SubSurfaceScatteringQuality p_quality); + RS::SubSurfaceScatteringQuality sss_get_quality() const; + void sss_set_scale(float p_scale, float p_depth_scale); + + void sub_surface_scattering(Ref<RenderSceneBuffersRD> p_render_buffers, RID p_diffuse, RID p_depth, const Projection &p_camera, const Size2i &p_screen_size); + +private: + /* Settings */ + + RS::EnvironmentSSAOQuality ssao_quality = RS::ENV_SSAO_QUALITY_MEDIUM; + bool ssao_half_size = false; + float ssao_adaptive_target = 0.5; + int ssao_blur_passes = 2; + float ssao_fadeout_from = 50.0; + float ssao_fadeout_to = 300.0; + + RS::EnvironmentSSILQuality ssil_quality = RS::ENV_SSIL_QUALITY_MEDIUM; + bool ssil_half_size = false; + float ssil_adaptive_target = 0.5; + int ssil_blur_passes = 4; + float ssil_fadeout_from = 50.0; + float ssil_fadeout_to = 300.0; + + RS::EnvironmentSSRRoughnessQuality ssr_roughness_quality = RS::ENV_SSR_ROUGHNESS_QUALITY_LOW; + + RS::SubSurfaceScatteringQuality sss_quality = RS::SUB_SURFACE_SCATTERING_QUALITY_MEDIUM; + float sss_scale = 0.05; + float sss_depth_scale = 0.01; + + /* SS Downsampler */ + + struct SSEffectsDownsamplePushConstant { + float pixel_size[2]; + float z_far; + float z_near; + uint32_t orthogonal; + float radius_sq; + uint32_t pad[2]; + }; + + enum SSEffectsMode { + SS_EFFECTS_DOWNSAMPLE, + SS_EFFECTS_DOWNSAMPLE_HALF_RES, + SS_EFFECTS_DOWNSAMPLE_MIPMAP, + SS_EFFECTS_DOWNSAMPLE_MIPMAP_HALF_RES, + SS_EFFECTS_DOWNSAMPLE_HALF, + SS_EFFECTS_DOWNSAMPLE_HALF_RES_HALF, + SS_EFFECTS_DOWNSAMPLE_FULL_MIPS, + SS_EFFECTS_MAX + }; + + struct SSEffectsGatherConstants { + float rotation_matrices[80]; //5 vec4s * 4 + }; + + struct SSEffectsShader { + SSEffectsDownsamplePushConstant downsample_push_constant; + SsEffectsDownsampleShaderRD downsample_shader; + RID downsample_shader_version; + RID downsample_uniform_set; + bool used_half_size_last_frame = false; + bool used_mips_last_frame = false; + bool used_full_mips_last_frame = false; + + RID gather_constants_buffer; + + RID mirror_sampler; + + RID pipelines[SS_EFFECTS_MAX]; + } ss_effects; + + /* SSIL */ + + enum SSILMode { + SSIL_GATHER, + SSIL_GATHER_BASE, + SSIL_GATHER_ADAPTIVE, + SSIL_GENERATE_IMPORTANCE_MAP, + SSIL_PROCESS_IMPORTANCE_MAPA, + SSIL_PROCESS_IMPORTANCE_MAPB, + SSIL_BLUR_PASS, + SSIL_BLUR_PASS_SMART, + SSIL_BLUR_PASS_WIDE, + SSIL_INTERLEAVE, + SSIL_INTERLEAVE_SMART, + SSIL_INTERLEAVE_HALF, + SSIL_MAX + }; + + struct SSILGatherPushConstant { + int32_t screen_size[2]; + int pass; + int quality; + + float half_screen_pixel_size[2]; + float half_screen_pixel_size_x025[2]; + + float NDC_to_view_mul[2]; + float NDC_to_view_add[2]; + + float pad2[2]; + float z_near; + float z_far; + + float radius; + float intensity; + int size_multiplier; + int pad; + + float fade_out_mul; + float fade_out_add; + float normal_rejection_amount; + float inv_radius_near_limit; + + uint32_t is_orthogonal; + float neg_inv_radius; + float load_counter_avg_div; + float adaptive_sample_limit; + + int32_t pass_coord_offset[2]; + float pass_uv_offset[2]; + }; + + struct SSILImportanceMapPushConstant { + float half_screen_pixel_size[2]; + float intensity; + float pad; + }; + + struct SSILBlurPushConstant { + float edge_sharpness; + float pad; + float half_screen_pixel_size[2]; + }; + + struct SSILInterleavePushConstant { + float inv_sharpness; + uint32_t size_modifier; + float pixel_size[2]; + }; + + struct SSILProjectionUniforms { + float inv_last_frame_projection_matrix[16]; + }; + + struct SSIL { + SSILGatherPushConstant gather_push_constant; + SsilShaderRD gather_shader; + RID gather_shader_version; + RID projection_uniform_buffer; + + SSILImportanceMapPushConstant importance_map_push_constant; + SsilImportanceMapShaderRD importance_map_shader; + RID importance_map_shader_version; + RID importance_map_load_counter; + RID counter_uniform_set; + + SSILBlurPushConstant blur_push_constant; + SsilBlurShaderRD blur_shader; + RID blur_shader_version; + + SSILInterleavePushConstant interleave_push_constant; + SsilInterleaveShaderRD interleave_shader; + RID interleave_shader_version; + + RID pipelines[SSIL_MAX]; + } ssil; + + void gather_ssil(RD::ComputeListID p_compute_list, const Vector<RID> p_ssil_slices, const Vector<RID> p_edges_slices, const SSILSettings &p_settings, bool p_adaptive_base_pass, RID p_gather_uniform_set, RID p_importance_map_uniform_set, RID p_projection_uniform_set); + + /* SSAO */ + + enum SSAOMode { + SSAO_GATHER, + SSAO_GATHER_BASE, + SSAO_GATHER_ADAPTIVE, + SSAO_GENERATE_IMPORTANCE_MAP, + SSAO_PROCESS_IMPORTANCE_MAPA, + SSAO_PROCESS_IMPORTANCE_MAPB, + SSAO_BLUR_PASS, + SSAO_BLUR_PASS_SMART, + SSAO_BLUR_PASS_WIDE, + SSAO_INTERLEAVE, + SSAO_INTERLEAVE_SMART, + SSAO_INTERLEAVE_HALF, + SSAO_MAX + }; + + struct SSAOGatherPushConstant { + int32_t screen_size[2]; + int pass; + int quality; + + float half_screen_pixel_size[2]; + int size_multiplier; + float detail_intensity; + + float NDC_to_view_mul[2]; + float NDC_to_view_add[2]; + + float pad[2]; + float half_screen_pixel_size_x025[2]; + + float radius; + float intensity; + float shadow_power; + float shadow_clamp; + + float fade_out_mul; + float fade_out_add; + float horizon_angle_threshold; + float inv_radius_near_limit; + + uint32_t is_orthogonal; + float neg_inv_radius; + float load_counter_avg_div; + float adaptive_sample_limit; + + int32_t pass_coord_offset[2]; + float pass_uv_offset[2]; + }; + + struct SSAOImportanceMapPushConstant { + float half_screen_pixel_size[2]; + float intensity; + float power; + }; + + struct SSAOBlurPushConstant { + float edge_sharpness; + float pad; + float half_screen_pixel_size[2]; + }; + + struct SSAOInterleavePushConstant { + float inv_sharpness; + uint32_t size_modifier; + float pixel_size[2]; + }; + + struct SSAO { + SSAOGatherPushConstant gather_push_constant; + SsaoShaderRD gather_shader; + RID gather_shader_version; + + SSAOImportanceMapPushConstant importance_map_push_constant; + SsaoImportanceMapShaderRD importance_map_shader; + RID importance_map_shader_version; + RID importance_map_load_counter; + RID counter_uniform_set; + + SSAOBlurPushConstant blur_push_constant; + SsaoBlurShaderRD blur_shader; + RID blur_shader_version; + + SSAOInterleavePushConstant interleave_push_constant; + SsaoInterleaveShaderRD interleave_shader; + RID interleave_shader_version; + + RID pipelines[SSAO_MAX]; + } ssao; + + void gather_ssao(RD::ComputeListID p_compute_list, const Vector<RID> p_ao_slices, const SSAOSettings &p_settings, bool p_adaptive_base_pass, RID p_gather_uniform_set, RID p_importance_map_uniform_set); + + /* Screen Space Reflection */ + + enum SSRShaderSpecializations { + SSR_MULTIVIEW = 1 << 0, + SSR_VARIATIONS = 2, + }; + + struct ScreenSpaceReflectionSceneData { + float projection[2][16]; + float inv_projection[2][16]; + float eye_offset[2][4]; + }; + + // SSR Scale + + struct ScreenSpaceReflectionScalePushConstant { + int32_t screen_size[2]; + float camera_z_near; + float camera_z_far; + + uint32_t orthogonal; + uint32_t filter; + uint32_t view_index; + uint32_t pad1; + }; + + struct ScreenSpaceReflectionScale { + ScreenSpaceReflectionScaleShaderRD shader; + RID shader_version; + RID pipelines[SSR_VARIATIONS]; + } ssr_scale; + + // SSR main + + enum ScreenSpaceReflectionMode { + SCREEN_SPACE_REFLECTION_NORMAL, + SCREEN_SPACE_REFLECTION_ROUGH, + SCREEN_SPACE_REFLECTION_MAX, + }; + + struct ScreenSpaceReflectionPushConstant { + float proj_info[4]; // 16 - 16 + + int32_t screen_size[2]; // 8 - 24 + float camera_z_near; // 4 - 28 + float camera_z_far; // 4 - 32 + + int32_t num_steps; // 4 - 36 + float depth_tolerance; // 4 - 40 + float distance_fade; // 4 - 44 + float curve_fade_in; // 4 - 48 + + uint32_t orthogonal; // 4 - 52 + float filter_mipmap_levels; // 4 - 56 + uint32_t use_half_res; // 4 - 60 + uint32_t view_index; // 4 - 64 + + // float projection[16]; // this is in our ScreenSpaceReflectionSceneData now + }; + + struct ScreenSpaceReflection { + ScreenSpaceReflectionShaderRD shader; + RID shader_version; + RID pipelines[SSR_VARIATIONS][SCREEN_SPACE_REFLECTION_MAX]; + + RID ubo; + } ssr; + + // SSR Filter + + struct ScreenSpaceReflectionFilterPushConstant { + float proj_info[4]; // 16 - 16 + + uint32_t orthogonal; // 4 - 20 + float edge_tolerance; // 4 - 24 + int32_t increment; // 4 - 28 + uint32_t view_index; // 4 - 32 + + int32_t screen_size[2]; // 8 - 40 + uint32_t vertical; // 4 - 44 + uint32_t steps; // 4 - 48 + }; + + enum SSRReflectionMode { + SCREEN_SPACE_REFLECTION_FILTER_HORIZONTAL, + SCREEN_SPACE_REFLECTION_FILTER_VERTICAL, + SCREEN_SPACE_REFLECTION_FILTER_MAX, + }; + + struct ScreenSpaceReflectionFilter { + ScreenSpaceReflectionFilterShaderRD shader; + RID shader_version; + RID pipelines[SSR_VARIATIONS][SCREEN_SPACE_REFLECTION_FILTER_MAX]; + } ssr_filter; + + /* Subsurface scattering */ + + struct SubSurfaceScatteringPushConstant { + int32_t screen_size[2]; + float camera_z_far; + float camera_z_near; + + uint32_t vertical; + uint32_t orthogonal; + float unit_size; + float scale; + + float depth_scale; + uint32_t pad[3]; + }; + + struct SubSurfaceScattering { + SubSurfaceScatteringPushConstant push_constant; + SubsurfaceScatteringShaderRD shader; + RID shader_version; + RID pipelines[3]; //3 quality levels + } sss; +}; + +} // namespace RendererRD + +#endif // SS_EFFECTS_RD_H diff --git a/servers/rendering/renderer_rd/effects/taa.cpp b/servers/rendering/renderer_rd/effects/taa.cpp new file mode 100644 index 0000000000..657385a509 --- /dev/null +++ b/servers/rendering/renderer_rd/effects/taa.cpp @@ -0,0 +1,138 @@ +/*************************************************************************/ +/* taa.cpp */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#include "taa.h" +#include "servers/rendering/renderer_rd/effects/copy_effects.h" +#include "servers/rendering/renderer_rd/storage_rd/material_storage.h" +#include "servers/rendering/renderer_rd/uniform_set_cache_rd.h" + +using namespace RendererRD; + +TAA::TAA() { + Vector<String> taa_modes; + taa_modes.push_back("\n#define MODE_TAA_RESOLVE"); + taa_shader.initialize(taa_modes); + shader_version = taa_shader.version_create(); + pipeline = RD::get_singleton()->compute_pipeline_create(taa_shader.version_get_shader(shader_version, 0)); +} + +TAA::~TAA() { + taa_shader.version_free(shader_version); +} + +void TAA::msaa_resolve(Ref<RenderSceneBuffersRD> p_render_buffers) { + if (!p_render_buffers->has_velocity_buffer(true)) { + // nothing to resolve + return; + } + + for (uint32_t v = 0; v < p_render_buffers->get_view_count(); v++) { + RID velocity_buffer_msaa = p_render_buffers->get_velocity_buffer(true, v); + RID velocity_buffer = p_render_buffers->get_velocity_buffer(false, v); + + RD::get_singleton()->texture_resolve_multisample(velocity_buffer_msaa, velocity_buffer); + } +} + +void TAA::resolve(RID p_frame, RID p_temp, RID p_depth, RID p_velocity, RID p_prev_velocity, RID p_history, Size2 p_resolution, float p_z_near, float p_z_far) { + UniformSetCacheRD *uniform_set_cache = UniformSetCacheRD::get_singleton(); + ERR_FAIL_NULL(uniform_set_cache); + MaterialStorage *material_storage = MaterialStorage::get_singleton(); + ERR_FAIL_NULL(material_storage); + + RID shader = taa_shader.version_get_shader(shader_version, 0); + ERR_FAIL_COND(shader.is_null()); + + RID default_sampler = material_storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + + TAAResolvePushConstant push_constant; + memset(&push_constant, 0, sizeof(TAAResolvePushConstant)); + push_constant.resolution_width = p_resolution.width; + push_constant.resolution_height = p_resolution.height; + push_constant.disocclusion_threshold = 0.025f; + push_constant.disocclusion_scale = 10.0f; + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, pipeline); + + RD::Uniform u_frame_source(RD::UNIFORM_TYPE_IMAGE, 0, { p_frame }); + RD::Uniform u_depth(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 1, { default_sampler, p_depth }); + RD::Uniform u_velocity(RD::UNIFORM_TYPE_IMAGE, 2, { p_velocity }); + RD::Uniform u_prev_velocity(RD::UNIFORM_TYPE_IMAGE, 3, { p_prev_velocity }); + RD::Uniform u_history(RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE, 4, { default_sampler, p_history }); + RD::Uniform u_frame_dest(RD::UNIFORM_TYPE_IMAGE, 5, { p_temp }); + + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, uniform_set_cache->get_cache(shader, 0, u_frame_source, u_depth, u_velocity, u_prev_velocity, u_history, u_frame_dest), 0); + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(TAAResolvePushConstant)); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_resolution.width, p_resolution.height, 1); + RD::get_singleton()->compute_list_end(); +} + +void TAA::process(Ref<RenderSceneBuffersRD> p_render_buffers, RD::DataFormat p_format, float p_z_near, float p_z_far) { + CopyEffects *copy_effects = CopyEffects::get_singleton(); + + uint32_t view_count = p_render_buffers->get_view_count(); + Size2i internal_size = p_render_buffers->get_internal_size(); + Size2i target_size = p_render_buffers->get_target_size(); + + bool just_allocated = false; + if (!p_render_buffers->has_texture(SNAME("taa"), SNAME("history"))) { + uint32_t usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + + p_render_buffers->create_texture(SNAME("taa"), SNAME("history"), p_format, usage_bits); + p_render_buffers->create_texture(SNAME("taa"), SNAME("temp"), p_format, usage_bits); + + p_render_buffers->create_texture(SNAME("taa"), SNAME("prev_velocity"), RD::DATA_FORMAT_R16G16_SFLOAT, usage_bits); + + just_allocated = true; + } + + RD::get_singleton()->draw_command_begin_label("TAA"); + + for (uint32_t v = 0; v < view_count; v++) { + // Get our (cached) slices + RID internal_texture = p_render_buffers->get_internal_texture(v); + RID velocity_buffer = p_render_buffers->get_velocity_buffer(false, v); + RID taa_history = p_render_buffers->get_texture_slice(SNAME("taa"), SNAME("history"), v, 0); + RID taa_prev_velocity = p_render_buffers->get_texture_slice(SNAME("taa"), SNAME("prev_velocity"), v, 0); + + if (!just_allocated) { + RID depth_texture = p_render_buffers->get_depth_texture(v); + RID taa_temp = p_render_buffers->get_texture_slice(SNAME("taa"), SNAME("temp"), v, 0); + resolve(internal_texture, taa_temp, depth_texture, velocity_buffer, taa_prev_velocity, taa_history, Size2(internal_size.x, internal_size.y), p_z_near, p_z_far); + copy_effects->copy_to_rect(taa_temp, internal_texture, Rect2(0, 0, internal_size.x, internal_size.y)); + } + + copy_effects->copy_to_rect(internal_texture, taa_history, Rect2(0, 0, internal_size.x, internal_size.y)); + copy_effects->copy_to_rect(velocity_buffer, taa_prev_velocity, Rect2(0, 0, target_size.x, target_size.y)); + } + + RD::get_singleton()->draw_command_end_label(); +} diff --git a/servers/rendering/renderer_rd/effects/taa.h b/servers/rendering/renderer_rd/effects/taa.h new file mode 100644 index 0000000000..ce4af18866 --- /dev/null +++ b/servers/rendering/renderer_rd/effects/taa.h @@ -0,0 +1,68 @@ +/*************************************************************************/ +/* taa.h */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#ifndef TAA_RD_H +#define TAA_RD_H + +#include "servers/rendering/renderer_rd/pipeline_cache_rd.h" +#include "servers/rendering/renderer_rd/shaders/effects/taa_resolve.glsl.gen.h" +#include "servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.h" +#include "servers/rendering/renderer_scene_render.h" + +#include "servers/rendering_server.h" + +namespace RendererRD { + +class TAA { +public: + TAA(); + ~TAA(); + + void msaa_resolve(Ref<RenderSceneBuffersRD> p_render_buffers); + void process(Ref<RenderSceneBuffersRD> p_render_buffers, RD::DataFormat p_format, float p_z_near, float p_z_far); + +private: + struct TAAResolvePushConstant { + float resolution_width; + float resolution_height; + float disocclusion_threshold; + float disocclusion_scale; + }; + + TaaResolveShaderRD taa_shader; + RID shader_version; + RID pipeline; + + void resolve(RID p_frame, RID p_temp, RID p_depth, RID p_velocity, RID p_prev_velocity, RID p_history, Size2 p_resolution, float p_z_near, float p_z_far); +}; + +} // namespace RendererRD + +#endif // TAA_RD_H diff --git a/servers/rendering/renderer_rd/effects/tone_mapper.cpp b/servers/rendering/renderer_rd/effects/tone_mapper.cpp index 38a4a37b8a..3a47b1420b 100644 --- a/servers/rendering/renderer_rd/effects/tone_mapper.cpp +++ b/servers/rendering/renderer_rd/effects/tone_mapper.cpp @@ -117,7 +117,7 @@ void ToneMapper::tonemapper(RID p_source_color, RID p_dst_framebuffer, const Ton tonemap.push_constant.use_auto_exposure = p_settings.use_auto_exposure; tonemap.push_constant.exposure = p_settings.exposure; tonemap.push_constant.white = p_settings.white; - tonemap.push_constant.auto_exposure_grey = p_settings.auto_exposure_grey; + tonemap.push_constant.auto_exposure_scale = p_settings.auto_exposure_scale; tonemap.push_constant.luminance_multiplier = p_settings.luminance_multiplier; tonemap.push_constant.use_color_correction = p_settings.use_color_correction; @@ -203,7 +203,7 @@ void ToneMapper::tonemapper(RD::DrawListID p_subpass_draw_list, RID p_source_col tonemap.push_constant.use_auto_exposure = p_settings.use_auto_exposure; tonemap.push_constant.exposure = p_settings.exposure; tonemap.push_constant.white = p_settings.white; - tonemap.push_constant.auto_exposure_grey = p_settings.auto_exposure_grey; + tonemap.push_constant.auto_exposure_scale = p_settings.auto_exposure_scale; tonemap.push_constant.use_color_correction = p_settings.use_color_correction; diff --git a/servers/rendering/renderer_rd/effects/tone_mapper.h b/servers/rendering/renderer_rd/effects/tone_mapper.h index a90849dbeb..e91118e241 100644 --- a/servers/rendering/renderer_rd/effects/tone_mapper.h +++ b/servers/rendering/renderer_rd/effects/tone_mapper.h @@ -77,7 +77,7 @@ private: float exposure; // 4 - 84 float white; // 4 - 88 - float auto_exposure_grey; // 4 - 92 + float auto_exposure_scale; // 4 - 92 float luminance_multiplier; // 4 - 96 float pixel_size[2]; // 8 - 104 @@ -124,7 +124,7 @@ public: float white = 1.0; bool use_auto_exposure = false; - float auto_exposure_grey = 0.5; + float auto_exposure_scale = 0.5; RID exposure_texture; float luminance_multiplier = 1.0; @@ -149,4 +149,4 @@ public: } // namespace RendererRD -#endif // !TONE_MAPPER_RD_H +#endif // TONE_MAPPER_RD_H diff --git a/servers/rendering/renderer_rd/effects/vrs.cpp b/servers/rendering/renderer_rd/effects/vrs.cpp index 505a35a269..5ff00aa94c 100644 --- a/servers/rendering/renderer_rd/effects/vrs.cpp +++ b/servers/rendering/renderer_rd/effects/vrs.cpp @@ -91,47 +91,22 @@ void VRS::copy_vrs(RID p_source_rd_texture, RID p_dest_framebuffer, bool p_multi RD::get_singleton()->draw_list_end(); } -void VRS::create_vrs_texture(const int p_base_width, const int p_base_height, const uint32_t p_view_count, RID &p_vrs_texture, RID &p_vrs_fb) { - // TODO find a way to skip this if VRS is not supported, but we don't have access to VulkanContext here, even though we're in vulkan.. hmmm - +Size2i VRS::get_vrs_texture_size(const Size2i p_base_size) const { // TODO we should find some way to store this properly, we're assuming 16x16 as this seems to be the standard but in our vrs_capacities we - // obtain a minimum and maximum size, and we should choose something within this range and then make sure that is consistantly set when creating + // obtain a minimum and maximum size, and we should choose something within this range and then make sure that is consistently set when creating // our frame buffer. Also it is important that we make the resulting size we calculate down below available to the end user so they know the size // of the VRS buffer to supply. Size2i texel_size = Size2i(16, 16); - RD::TextureFormat tf; - if (p_view_count > 1) { - tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; - } else { - tf.texture_type = RD::TEXTURE_TYPE_2D; - } - tf.format = RD::DATA_FORMAT_R8_UINT; - tf.width = p_base_width / texel_size.x; - if (p_base_width % texel_size.x != 0) { - tf.width++; + int width = p_base_size.x / texel_size.x; + if (p_base_size.x % texel_size.x != 0) { + width++; } - tf.height = p_base_height / texel_size.y; - if (p_base_height % texel_size.y != 0) { - tf.height++; + int height = p_base_size.y / texel_size.y; + if (p_base_size.y % texel_size.y != 0) { + height++; } - tf.array_layers = p_view_count; // create a layer for every view - tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_VRS_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; - tf.samples = RD::TEXTURE_SAMPLES_1; - - p_vrs_texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); - - // by default VRS is assumed to be our VRS attachment, but if we need to write into it, we need a bit more control - Vector<RID> fb; - fb.push_back(p_vrs_texture); - - RD::FramebufferPass pass; - pass.color_attachments.push_back(0); - - Vector<RD::FramebufferPass> passes; - passes.push_back(pass); - - p_vrs_fb = RD::get_singleton()->framebuffer_create_multipass(fb, passes, RenderingDevice::INVALID_ID, p_view_count); + return Size2i(width, height); } void VRS::update_vrs_texture(RID p_vrs_fb, RID p_render_target) { @@ -146,10 +121,11 @@ void VRS::update_vrs_texture(RID p_vrs_fb, RID p_render_target) { if (vrs_mode == RS::VIEWPORT_VRS_TEXTURE) { RID vrs_texture = texture_storage->render_target_get_vrs_texture(p_render_target); if (vrs_texture.is_valid()) { - Texture *texture = texture_storage->get_texture(vrs_texture); - if (texture) { + RID rd_texture = texture_storage->texture_get_rd_texture(vrs_texture); + int layers = texture_storage->texture_get_layers(vrs_texture); + if (rd_texture.is_valid()) { // Copy into our density buffer - copy_vrs(texture->rd_texture, p_vrs_fb, texture->layers > 1); + copy_vrs(rd_texture, p_vrs_fb, layers > 1); } } } else if (vrs_mode == RS::VIEWPORT_VRS_XR) { @@ -157,10 +133,12 @@ void VRS::update_vrs_texture(RID p_vrs_fb, RID p_render_target) { if (interface.is_valid()) { RID vrs_texture = interface->get_vrs_texture(); if (vrs_texture.is_valid()) { - Texture *texture = texture_storage->get_texture(vrs_texture); - if (texture) { + RID rd_texture = texture_storage->texture_get_rd_texture(vrs_texture); + int layers = texture_storage->texture_get_layers(vrs_texture); + + if (rd_texture.is_valid()) { // Copy into our density buffer - copy_vrs(texture->rd_texture, p_vrs_fb, texture->layers > 1); + copy_vrs(rd_texture, p_vrs_fb, layers > 1); } } } diff --git a/servers/rendering/renderer_rd/effects/vrs.h b/servers/rendering/renderer_rd/effects/vrs.h index 0f2bdd31b6..7125c6455d 100644 --- a/servers/rendering/renderer_rd/effects/vrs.h +++ b/servers/rendering/renderer_rd/effects/vrs.h @@ -66,10 +66,10 @@ public: void copy_vrs(RID p_source_rd_texture, RID p_dest_framebuffer, bool p_multiview = false); - void create_vrs_texture(const int p_base_width, const int p_base_height, const uint32_t p_view_count, RID &p_vrs_texture, RID &p_vrs_fb); + Size2i get_vrs_texture_size(const Size2i p_base_size) const; void update_vrs_texture(RID p_vrs_fb, RID p_render_target); }; } // namespace RendererRD -#endif // !VRS_RD_H +#endif // VRS_RD_H |