diff options
Diffstat (limited to 'src/video_core')
19 files changed, 266 insertions, 99 deletions
diff --git a/src/video_core/buffer_cache/buffer_cache.h b/src/video_core/buffer_cache/buffer_cache.h index 9bafd8cc0..45977d578 100644 --- a/src/video_core/buffer_cache/buffer_cache.h +++ b/src/video_core/buffer_cache/buffer_cache.h @@ -719,9 +719,15 @@ void BufferCache<P>::BindHostVertexBuffers() { bool any_valid{false}; auto& flags = maxwell3d->dirty.flags; for (u32 index = 0; index < NUM_VERTEX_BUFFERS; ++index) { + const Binding& binding = channel_state->vertex_buffers[index]; + Buffer& buffer = slot_buffers[binding.buffer_id]; + TouchBuffer(buffer, binding.buffer_id); + SynchronizeBuffer(buffer, binding.cpu_addr, binding.size); if (!flags[Dirty::VertexBuffer0 + index]) { continue; } + flags[Dirty::VertexBuffer0 + index] = false; + host_bindings.min_index = std::min(host_bindings.min_index, index); host_bindings.max_index = std::max(host_bindings.max_index, index); any_valid = true; @@ -735,9 +741,6 @@ void BufferCache<P>::BindHostVertexBuffers() { const Binding& binding = channel_state->vertex_buffers[index]; Buffer& buffer = slot_buffers[binding.buffer_id]; - TouchBuffer(buffer, binding.buffer_id); - SynchronizeBuffer(buffer, binding.cpu_addr, binding.size); - const u32 stride = maxwell3d->regs.vertex_streams[index].stride; const u32 offset = buffer.Offset(binding.cpu_addr); diff --git a/src/video_core/renderer_opengl/gl_compute_pipeline.cpp b/src/video_core/renderer_opengl/gl_compute_pipeline.cpp index 1a0cea9b7..3151c0db8 100644 --- a/src/video_core/renderer_opengl/gl_compute_pipeline.cpp +++ b/src/video_core/renderer_opengl/gl_compute_pipeline.cpp @@ -87,7 +87,8 @@ void ComputePipeline::Configure() { texture_cache.SynchronizeComputeDescriptors(); boost::container::static_vector<VideoCommon::ImageViewInOut, MAX_TEXTURES + MAX_IMAGES> views; - std::array<GLuint, MAX_TEXTURES> samplers; + boost::container::static_vector<VideoCommon::SamplerId, MAX_TEXTURES> samplers; + std::array<GLuint, MAX_TEXTURES> gl_samplers; std::array<GLuint, MAX_TEXTURES> textures; std::array<GLuint, MAX_IMAGES> images; GLsizei sampler_binding{}; @@ -131,7 +132,6 @@ void ComputePipeline::Configure() { for (u32 index = 0; index < desc.count; ++index) { const auto handle{read_handle(desc, index)}; views.push_back({handle.first}); - samplers[sampler_binding++] = 0; } } for (const auto& desc : info.image_buffer_descriptors) { @@ -142,8 +142,8 @@ void ComputePipeline::Configure() { const auto handle{read_handle(desc, index)}; views.push_back({handle.first}); - Sampler* const sampler = texture_cache.GetComputeSampler(handle.second); - samplers[sampler_binding++] = sampler->Handle(); + VideoCommon::SamplerId sampler = texture_cache.GetComputeSamplerId(handle.second); + samplers.push_back(sampler); } } for (const auto& desc : info.image_descriptors) { @@ -186,10 +186,17 @@ void ComputePipeline::Configure() { const VideoCommon::ImageViewInOut* views_it{views.data() + num_texture_buffers + num_image_buffers}; + const VideoCommon::SamplerId* samplers_it{samplers.data()}; texture_binding += num_texture_buffers; image_binding += num_image_buffers; u32 texture_scaling_mask{}; + + for (const auto& desc : info.texture_buffer_descriptors) { + for (u32 index = 0; index < desc.count; ++index) { + gl_samplers[sampler_binding++] = 0; + } + } for (const auto& desc : info.texture_descriptors) { for (u32 index = 0; index < desc.count; ++index) { ImageView& image_view{texture_cache.GetImageView((views_it++)->id)}; @@ -198,6 +205,12 @@ void ComputePipeline::Configure() { texture_scaling_mask |= 1u << texture_binding; } ++texture_binding; + + const Sampler& sampler{texture_cache.GetSampler(*(samplers_it++))}; + const bool use_fallback_sampler{sampler.HasAddedAnisotropy() && + !image_view.SupportsAnisotropy()}; + gl_samplers[sampler_binding++] = + use_fallback_sampler ? sampler.HandleWithDefaultAnisotropy() : sampler.Handle(); } } u32 image_scaling_mask{}; @@ -228,7 +241,7 @@ void ComputePipeline::Configure() { if (texture_binding != 0) { ASSERT(texture_binding == sampler_binding); glBindTextures(0, texture_binding, textures.data()); - glBindSamplers(0, sampler_binding, samplers.data()); + glBindSamplers(0, sampler_binding, gl_samplers.data()); } if (image_binding != 0) { glBindImageTextures(0, image_binding, images.data()); diff --git a/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp b/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp index 89000d6e0..c58f760b8 100644 --- a/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp +++ b/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp @@ -275,9 +275,9 @@ GraphicsPipeline::GraphicsPipeline(const Device& device, TextureCache& texture_c template <typename Spec> void GraphicsPipeline::ConfigureImpl(bool is_indexed) { std::array<VideoCommon::ImageViewInOut, MAX_TEXTURES + MAX_IMAGES> views; - std::array<GLuint, MAX_TEXTURES> samplers; + std::array<VideoCommon::SamplerId, MAX_TEXTURES> samplers; size_t views_index{}; - GLsizei sampler_binding{}; + size_t samplers_index{}; texture_cache.SynchronizeGraphicsDescriptors(); @@ -337,7 +337,6 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) { for (u32 index = 0; index < desc.count; ++index) { const auto handle{read_handle(desc, index)}; views[views_index++] = {handle.first}; - samplers[sampler_binding++] = 0; } } } @@ -351,8 +350,8 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) { const auto handle{read_handle(desc, index)}; views[views_index++] = {handle.first}; - Sampler* const sampler{texture_cache.GetGraphicsSampler(handle.second)}; - samplers[sampler_binding++] = sampler->Handle(); + VideoCommon::SamplerId sampler{texture_cache.GetGraphicsSamplerId(handle.second)}; + samplers[samplers_index++] = sampler; } } if constexpr (Spec::has_images) { @@ -445,10 +444,13 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) { program_manager.BindSourcePrograms(source_programs); } const VideoCommon::ImageViewInOut* views_it{views.data()}; + const VideoCommon::SamplerId* samplers_it{samplers.data()}; GLsizei texture_binding = 0; GLsizei image_binding = 0; + GLsizei sampler_binding{}; std::array<GLuint, MAX_TEXTURES> textures; std::array<GLuint, MAX_IMAGES> images; + std::array<GLuint, MAX_TEXTURES> gl_samplers; const auto prepare_stage{[&](size_t stage) { buffer_cache.runtime.SetImagePointers(&textures[texture_binding], &images[image_binding]); buffer_cache.BindHostStageBuffers(stage); @@ -465,6 +467,13 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) { u32 stage_image_binding{}; const auto& info{stage_infos[stage]}; + if constexpr (Spec::has_texture_buffers) { + for (const auto& desc : info.texture_buffer_descriptors) { + for (u32 index = 0; index < desc.count; ++index) { + gl_samplers[sampler_binding++] = 0; + } + } + } for (const auto& desc : info.texture_descriptors) { for (u32 index = 0; index < desc.count; ++index) { ImageView& image_view{texture_cache.GetImageView((views_it++)->id)}; @@ -474,6 +483,12 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) { } ++texture_binding; ++stage_texture_binding; + + const Sampler& sampler{texture_cache.GetSampler(*(samplers_it++))}; + const bool use_fallback_sampler{sampler.HasAddedAnisotropy() && + !image_view.SupportsAnisotropy()}; + gl_samplers[sampler_binding++] = + use_fallback_sampler ? sampler.HandleWithDefaultAnisotropy() : sampler.Handle(); } } for (const auto& desc : info.image_descriptors) { @@ -534,7 +549,7 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) { if (texture_binding != 0) { ASSERT(texture_binding == sampler_binding); glBindTextures(0, texture_binding, textures.data()); - glBindSamplers(0, sampler_binding, samplers.data()); + glBindSamplers(0, sampler_binding, gl_samplers.data()); } if (image_binding != 0) { glBindImageTextures(0, image_binding, images.data()); diff --git a/src/video_core/renderer_opengl/gl_shader_context.h b/src/video_core/renderer_opengl/gl_shader_context.h index 207a75d42..d12cd06fa 100644 --- a/src/video_core/renderer_opengl/gl_shader_context.h +++ b/src/video_core/renderer_opengl/gl_shader_context.h @@ -16,9 +16,9 @@ struct ShaderPools { inst.ReleaseContents(); } - Shader::ObjectPool<Shader::IR::Inst> inst; - Shader::ObjectPool<Shader::IR::Block> block; - Shader::ObjectPool<Shader::Maxwell::Flow::Block> flow_block; + Shader::ObjectPool<Shader::IR::Inst> inst{8192}; + Shader::ObjectPool<Shader::IR::Block> block{32}; + Shader::ObjectPool<Shader::Maxwell::Flow::Block> flow_block{32}; }; struct Context { diff --git a/src/video_core/renderer_opengl/gl_texture_cache.cpp b/src/video_core/renderer_opengl/gl_texture_cache.cpp index 1c5dbcdd8..3b446be07 100644 --- a/src/video_core/renderer_opengl/gl_texture_cache.cpp +++ b/src/video_core/renderer_opengl/gl_texture_cache.cpp @@ -1268,36 +1268,48 @@ Sampler::Sampler(TextureCacheRuntime& runtime, const TSCEntry& config) { UNIMPLEMENTED_IF(config.cubemap_anisotropy != 1); - sampler.Create(); - const GLuint handle = sampler.handle; - glSamplerParameteri(handle, GL_TEXTURE_WRAP_S, MaxwellToGL::WrapMode(config.wrap_u)); - glSamplerParameteri(handle, GL_TEXTURE_WRAP_T, MaxwellToGL::WrapMode(config.wrap_v)); - glSamplerParameteri(handle, GL_TEXTURE_WRAP_R, MaxwellToGL::WrapMode(config.wrap_p)); - glSamplerParameteri(handle, GL_TEXTURE_COMPARE_MODE, compare_mode); - glSamplerParameteri(handle, GL_TEXTURE_COMPARE_FUNC, compare_func); - glSamplerParameteri(handle, GL_TEXTURE_MAG_FILTER, mag); - glSamplerParameteri(handle, GL_TEXTURE_MIN_FILTER, min); - glSamplerParameterf(handle, GL_TEXTURE_LOD_BIAS, config.LodBias()); - glSamplerParameterf(handle, GL_TEXTURE_MIN_LOD, config.MinLod()); - glSamplerParameterf(handle, GL_TEXTURE_MAX_LOD, config.MaxLod()); - glSamplerParameterfv(handle, GL_TEXTURE_BORDER_COLOR, config.BorderColor().data()); - - if (GLAD_GL_ARB_texture_filter_anisotropic || GLAD_GL_EXT_texture_filter_anisotropic) { - const f32 max_anisotropy = std::clamp(config.MaxAnisotropy(), 1.0f, 16.0f); - glSamplerParameterf(handle, GL_TEXTURE_MAX_ANISOTROPY, max_anisotropy); - } else { - LOG_WARNING(Render_OpenGL, "GL_ARB_texture_filter_anisotropic is required"); - } - if (GLAD_GL_ARB_texture_filter_minmax || GLAD_GL_EXT_texture_filter_minmax) { - glSamplerParameteri(handle, GL_TEXTURE_REDUCTION_MODE_ARB, reduction_filter); - } else if (reduction_filter != GL_WEIGHTED_AVERAGE_ARB) { - LOG_WARNING(Render_OpenGL, "GL_ARB_texture_filter_minmax is required"); - } - if (GLAD_GL_ARB_seamless_cubemap_per_texture || GLAD_GL_AMD_seamless_cubemap_per_texture) { - glSamplerParameteri(handle, GL_TEXTURE_CUBE_MAP_SEAMLESS, seamless); - } else if (seamless == GL_FALSE) { - // We default to false because it's more common - LOG_WARNING(Render_OpenGL, "GL_ARB_seamless_cubemap_per_texture is required"); + const f32 max_anisotropy = std::clamp(config.MaxAnisotropy(), 1.0f, 16.0f); + + const auto create_sampler = [&](const f32 anisotropy) { + OGLSampler new_sampler; + new_sampler.Create(); + const GLuint handle = new_sampler.handle; + glSamplerParameteri(handle, GL_TEXTURE_WRAP_S, MaxwellToGL::WrapMode(config.wrap_u)); + glSamplerParameteri(handle, GL_TEXTURE_WRAP_T, MaxwellToGL::WrapMode(config.wrap_v)); + glSamplerParameteri(handle, GL_TEXTURE_WRAP_R, MaxwellToGL::WrapMode(config.wrap_p)); + glSamplerParameteri(handle, GL_TEXTURE_COMPARE_MODE, compare_mode); + glSamplerParameteri(handle, GL_TEXTURE_COMPARE_FUNC, compare_func); + glSamplerParameteri(handle, GL_TEXTURE_MAG_FILTER, mag); + glSamplerParameteri(handle, GL_TEXTURE_MIN_FILTER, min); + glSamplerParameterf(handle, GL_TEXTURE_LOD_BIAS, config.LodBias()); + glSamplerParameterf(handle, GL_TEXTURE_MIN_LOD, config.MinLod()); + glSamplerParameterf(handle, GL_TEXTURE_MAX_LOD, config.MaxLod()); + glSamplerParameterfv(handle, GL_TEXTURE_BORDER_COLOR, config.BorderColor().data()); + + if (GLAD_GL_ARB_texture_filter_anisotropic || GLAD_GL_EXT_texture_filter_anisotropic) { + glSamplerParameterf(handle, GL_TEXTURE_MAX_ANISOTROPY, anisotropy); + } else { + LOG_WARNING(Render_OpenGL, "GL_ARB_texture_filter_anisotropic is required"); + } + if (GLAD_GL_ARB_texture_filter_minmax || GLAD_GL_EXT_texture_filter_minmax) { + glSamplerParameteri(handle, GL_TEXTURE_REDUCTION_MODE_ARB, reduction_filter); + } else if (reduction_filter != GL_WEIGHTED_AVERAGE_ARB) { + LOG_WARNING(Render_OpenGL, "GL_ARB_texture_filter_minmax is required"); + } + if (GLAD_GL_ARB_seamless_cubemap_per_texture || GLAD_GL_AMD_seamless_cubemap_per_texture) { + glSamplerParameteri(handle, GL_TEXTURE_CUBE_MAP_SEAMLESS, seamless); + } else if (seamless == GL_FALSE) { + // We default to false because it's more common + LOG_WARNING(Render_OpenGL, "GL_ARB_seamless_cubemap_per_texture is required"); + } + return new_sampler; + }; + + sampler = create_sampler(max_anisotropy); + + const f32 max_anisotropy_default = static_cast<f32>(1U << config.max_anisotropy); + if (max_anisotropy > max_anisotropy_default) { + sampler_default_anisotropy = create_sampler(max_anisotropy_default); } } diff --git a/src/video_core/renderer_opengl/gl_texture_cache.h b/src/video_core/renderer_opengl/gl_texture_cache.h index 1148b73d7..3676eaaa9 100644 --- a/src/video_core/renderer_opengl/gl_texture_cache.h +++ b/src/video_core/renderer_opengl/gl_texture_cache.h @@ -309,12 +309,21 @@ class Sampler { public: explicit Sampler(TextureCacheRuntime&, const Tegra::Texture::TSCEntry&); - GLuint Handle() const noexcept { + [[nodiscard]] GLuint Handle() const noexcept { return sampler.handle; } + [[nodiscard]] GLuint HandleWithDefaultAnisotropy() const noexcept { + return sampler_default_anisotropy.handle; + } + + [[nodiscard]] bool HasAddedAnisotropy() const noexcept { + return static_cast<bool>(sampler_default_anisotropy.handle); + } + private: OGLSampler sampler; + OGLSampler sampler_default_anisotropy; }; class Framebuffer { diff --git a/src/video_core/renderer_vulkan/pipeline_helper.h b/src/video_core/renderer_vulkan/pipeline_helper.h index 983e1c2e1..71c783709 100644 --- a/src/video_core/renderer_vulkan/pipeline_helper.h +++ b/src/video_core/renderer_vulkan/pipeline_helper.h @@ -178,7 +178,7 @@ public: inline void PushImageDescriptors(TextureCache& texture_cache, GuestDescriptorQueue& guest_descriptor_queue, const Shader::Info& info, RescalingPushConstant& rescaling, - const VkSampler*& samplers, + const VideoCommon::SamplerId*& samplers, const VideoCommon::ImageViewInOut*& views) { const u32 num_texture_buffers = Shader::NumDescriptors(info.texture_buffer_descriptors); const u32 num_image_buffers = Shader::NumDescriptors(info.image_buffer_descriptors); @@ -187,10 +187,15 @@ inline void PushImageDescriptors(TextureCache& texture_cache, for (const auto& desc : info.texture_descriptors) { for (u32 index = 0; index < desc.count; ++index) { const VideoCommon::ImageViewId image_view_id{(views++)->id}; - const VkSampler sampler{*(samplers++)}; + const VideoCommon::SamplerId sampler_id{*(samplers++)}; ImageView& image_view{texture_cache.GetImageView(image_view_id)}; const VkImageView vk_image_view{image_view.Handle(desc.type)}; - guest_descriptor_queue.AddSampledImage(vk_image_view, sampler); + const Sampler& sampler{texture_cache.GetSampler(sampler_id)}; + const bool use_fallback_sampler{sampler.HasAddedAnisotropy() && + !image_view.SupportsAnisotropy()}; + const VkSampler vk_sampler{use_fallback_sampler ? sampler.HandleWithDefaultAnisotropy() + : sampler.Handle()}; + guest_descriptor_queue.AddSampledImage(vk_image_view, vk_sampler); rescaling.PushTexture(texture_cache.IsRescaling(image_view)); } } diff --git a/src/video_core/renderer_vulkan/vk_buffer_cache.cpp b/src/video_core/renderer_vulkan/vk_buffer_cache.cpp index 8c33722d3..e30fcb1ed 100644 --- a/src/video_core/renderer_vulkan/vk_buffer_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_buffer_cache.cpp @@ -516,15 +516,15 @@ void BufferCacheRuntime::BindVertexBuffers(VideoCommon::HostBindings<Buffer>& bi buffer_handles.push_back(handle); } if (device.IsExtExtendedDynamicStateSupported()) { - scheduler.Record([bindings = bindings, - buffer_handles = buffer_handles](vk::CommandBuffer cmdbuf) { + scheduler.Record([bindings = std::move(bindings), + buffer_handles = std::move(buffer_handles)](vk::CommandBuffer cmdbuf) { cmdbuf.BindVertexBuffers2EXT( bindings.min_index, bindings.max_index - bindings.min_index, buffer_handles.data(), bindings.offsets.data(), bindings.sizes.data(), bindings.strides.data()); }); } else { - scheduler.Record([bindings = bindings, - buffer_handles = buffer_handles](vk::CommandBuffer cmdbuf) { + scheduler.Record([bindings = std::move(bindings), + buffer_handles = std::move(buffer_handles)](vk::CommandBuffer cmdbuf) { cmdbuf.BindVertexBuffers(bindings.min_index, bindings.max_index - bindings.min_index, buffer_handles.data(), bindings.offsets.data()); }); @@ -561,12 +561,12 @@ void BufferCacheRuntime::BindTransformFeedbackBuffers(VideoCommon::HostBindings< for (u32 index = 0; index < bindings.buffers.size(); ++index) { buffer_handles.push_back(bindings.buffers[index]->Handle()); } - scheduler.Record( - [bindings = bindings, buffer_handles = buffer_handles](vk::CommandBuffer cmdbuf) { - cmdbuf.BindTransformFeedbackBuffersEXT(0, static_cast<u32>(buffer_handles.size()), - buffer_handles.data(), bindings.offsets.data(), - bindings.sizes.data()); - }); + scheduler.Record([bindings = std::move(bindings), + buffer_handles = std::move(buffer_handles)](vk::CommandBuffer cmdbuf) { + cmdbuf.BindTransformFeedbackBuffersEXT(0, static_cast<u32>(buffer_handles.size()), + buffer_handles.data(), bindings.offsets.data(), + bindings.sizes.data()); + }); } void BufferCacheRuntime::ReserveNullBuffer() { diff --git a/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp b/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp index 733e70d9d..73e585c2b 100644 --- a/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp +++ b/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp @@ -115,7 +115,7 @@ void ComputePipeline::Configure(Tegra::Engines::KeplerCompute& kepler_compute, static constexpr size_t max_elements = 64; boost::container::static_vector<VideoCommon::ImageViewInOut, max_elements> views; - boost::container::static_vector<VkSampler, max_elements> samplers; + boost::container::static_vector<VideoCommon::SamplerId, max_elements> samplers; const auto& qmd{kepler_compute.launch_description}; const auto& cbufs{qmd.const_buffer_config}; @@ -160,8 +160,8 @@ void ComputePipeline::Configure(Tegra::Engines::KeplerCompute& kepler_compute, const auto handle{read_handle(desc, index)}; views.push_back({handle.first}); - Sampler* const sampler = texture_cache.GetComputeSampler(handle.second); - samplers.push_back(sampler->Handle()); + VideoCommon::SamplerId sampler = texture_cache.GetComputeSamplerId(handle.second); + samplers.push_back(sampler); } } for (const auto& desc : info.image_descriptors) { @@ -192,7 +192,7 @@ void ComputePipeline::Configure(Tegra::Engines::KeplerCompute& kepler_compute, buffer_cache.BindHostComputeBuffers(); RescalingPushConstant rescaling; - const VkSampler* samplers_it{samplers.data()}; + const VideoCommon::SamplerId* samplers_it{samplers.data()}; const VideoCommon::ImageViewInOut* views_it{views.data()}; PushImageDescriptors(texture_cache, guest_descriptor_queue, info, rescaling, samplers_it, views_it); diff --git a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp index 506b78f08..c1595642e 100644 --- a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp +++ b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp @@ -298,7 +298,7 @@ void GraphicsPipeline::AddTransition(GraphicsPipeline* transition) { template <typename Spec> void GraphicsPipeline::ConfigureImpl(bool is_indexed) { std::array<VideoCommon::ImageViewInOut, MAX_IMAGE_ELEMENTS> views; - std::array<VkSampler, MAX_IMAGE_ELEMENTS> samplers; + std::array<VideoCommon::SamplerId, MAX_IMAGE_ELEMENTS> samplers; size_t sampler_index{}; size_t view_index{}; @@ -367,8 +367,8 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) { const auto handle{read_handle(desc, index)}; views[view_index++] = {handle.first}; - Sampler* const sampler{texture_cache.GetGraphicsSampler(handle.second)}; - samplers[sampler_index++] = sampler->Handle(); + VideoCommon::SamplerId sampler{texture_cache.GetGraphicsSamplerId(handle.second)}; + samplers[sampler_index++] = sampler; } } if constexpr (Spec::has_images) { @@ -453,7 +453,7 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) { RescalingPushConstant rescaling; RenderAreaPushConstant render_area; - const VkSampler* samplers_it{samplers.data()}; + const VideoCommon::SamplerId* samplers_it{samplers.data()}; const VideoCommon::ImageViewInOut* views_it{views.data()}; const auto prepare_stage{[&](size_t stage) LAMBDA_FORCEINLINE { buffer_cache.BindHostStageBuffers(stage); diff --git a/src/video_core/renderer_vulkan/vk_master_semaphore.cpp b/src/video_core/renderer_vulkan/vk_master_semaphore.cpp index b128c4f6e..5eeda08d2 100644 --- a/src/video_core/renderer_vulkan/vk_master_semaphore.cpp +++ b/src/video_core/renderer_vulkan/vk_master_semaphore.cpp @@ -3,6 +3,7 @@ #include <thread> +#include "common/polyfill_ranges.h" #include "common/settings.h" #include "video_core/renderer_vulkan/vk_master_semaphore.h" #include "video_core/vulkan_common/vulkan_device.h" diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.h b/src/video_core/renderer_vulkan/vk_pipeline_cache.h index 15aa7e224..e323ea0fd 100644 --- a/src/video_core/renderer_vulkan/vk_pipeline_cache.h +++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.h @@ -92,9 +92,9 @@ struct ShaderPools { inst.ReleaseContents(); } - Shader::ObjectPool<Shader::IR::Inst> inst; - Shader::ObjectPool<Shader::IR::Block> block; - Shader::ObjectPool<Shader::Maxwell::Flow::Block> flow_block; + Shader::ObjectPool<Shader::IR::Inst> inst{8192}; + Shader::ObjectPool<Shader::IR::Block> block{32}; + Shader::ObjectPool<Shader::Maxwell::Flow::Block> flow_block{32}; }; class PipelineCache : public VideoCommon::ShaderCache { diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.cpp b/src/video_core/renderer_vulkan/vk_texture_cache.cpp index 8711e2a87..f025f618b 100644 --- a/src/video_core/renderer_vulkan/vk_texture_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_texture_cache.cpp @@ -1802,27 +1802,36 @@ Sampler::Sampler(TextureCacheRuntime& runtime, const Tegra::Texture::TSCEntry& t // Some games have samplers with garbage. Sanitize them here. const f32 max_anisotropy = std::clamp(tsc.MaxAnisotropy(), 1.0f, 16.0f); - sampler = device.GetLogical().CreateSampler(VkSamplerCreateInfo{ - .sType = VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO, - .pNext = pnext, - .flags = 0, - .magFilter = MaxwellToVK::Sampler::Filter(tsc.mag_filter), - .minFilter = MaxwellToVK::Sampler::Filter(tsc.min_filter), - .mipmapMode = MaxwellToVK::Sampler::MipmapMode(tsc.mipmap_filter), - .addressModeU = MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_u, tsc.mag_filter), - .addressModeV = MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_v, tsc.mag_filter), - .addressModeW = MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_p, tsc.mag_filter), - .mipLodBias = tsc.LodBias(), - .anisotropyEnable = static_cast<VkBool32>(max_anisotropy > 1.0f ? VK_TRUE : VK_FALSE), - .maxAnisotropy = max_anisotropy, - .compareEnable = tsc.depth_compare_enabled, - .compareOp = MaxwellToVK::Sampler::DepthCompareFunction(tsc.depth_compare_func), - .minLod = tsc.mipmap_filter == TextureMipmapFilter::None ? 0.0f : tsc.MinLod(), - .maxLod = tsc.mipmap_filter == TextureMipmapFilter::None ? 0.25f : tsc.MaxLod(), - .borderColor = - arbitrary_borders ? VK_BORDER_COLOR_FLOAT_CUSTOM_EXT : ConvertBorderColor(color), - .unnormalizedCoordinates = VK_FALSE, - }); + const auto create_sampler = [&](const f32 anisotropy) { + return device.GetLogical().CreateSampler(VkSamplerCreateInfo{ + .sType = VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO, + .pNext = pnext, + .flags = 0, + .magFilter = MaxwellToVK::Sampler::Filter(tsc.mag_filter), + .minFilter = MaxwellToVK::Sampler::Filter(tsc.min_filter), + .mipmapMode = MaxwellToVK::Sampler::MipmapMode(tsc.mipmap_filter), + .addressModeU = MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_u, tsc.mag_filter), + .addressModeV = MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_v, tsc.mag_filter), + .addressModeW = MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_p, tsc.mag_filter), + .mipLodBias = tsc.LodBias(), + .anisotropyEnable = static_cast<VkBool32>(anisotropy > 1.0f ? VK_TRUE : VK_FALSE), + .maxAnisotropy = anisotropy, + .compareEnable = tsc.depth_compare_enabled, + .compareOp = MaxwellToVK::Sampler::DepthCompareFunction(tsc.depth_compare_func), + .minLod = tsc.mipmap_filter == TextureMipmapFilter::None ? 0.0f : tsc.MinLod(), + .maxLod = tsc.mipmap_filter == TextureMipmapFilter::None ? 0.25f : tsc.MaxLod(), + .borderColor = + arbitrary_borders ? VK_BORDER_COLOR_FLOAT_CUSTOM_EXT : ConvertBorderColor(color), + .unnormalizedCoordinates = VK_FALSE, + }); + }; + + sampler = create_sampler(max_anisotropy); + + const f32 max_anisotropy_default = static_cast<f32>(1U << tsc.max_anisotropy); + if (max_anisotropy > max_anisotropy_default) { + sampler_default_anisotropy = create_sampler(max_anisotropy_default); + } } Framebuffer::Framebuffer(TextureCacheRuntime& runtime, std::span<ImageView*, NUM_RT> color_buffers, diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.h b/src/video_core/renderer_vulkan/vk_texture_cache.h index 0f7a5ffd4..f14525dcb 100644 --- a/src/video_core/renderer_vulkan/vk_texture_cache.h +++ b/src/video_core/renderer_vulkan/vk_texture_cache.h @@ -279,8 +279,17 @@ public: return *sampler; } + [[nodiscard]] VkSampler HandleWithDefaultAnisotropy() const noexcept { + return *sampler_default_anisotropy; + } + + [[nodiscard]] bool HasAddedAnisotropy() const noexcept { + return static_cast<bool>(sampler_default_anisotropy); + } + private: vk::Sampler sampler; + vk::Sampler sampler_default_anisotropy; }; class Framebuffer { diff --git a/src/video_core/texture_cache/image_view_base.cpp b/src/video_core/texture_cache/image_view_base.cpp index d134b6738..0c5f4450d 100644 --- a/src/video_core/texture_cache/image_view_base.cpp +++ b/src/video_core/texture_cache/image_view_base.cpp @@ -45,4 +45,56 @@ ImageViewBase::ImageViewBase(const ImageInfo& info, const ImageViewInfo& view_in ImageViewBase::ImageViewBase(const NullImageViewParams&) : image_id{NULL_IMAGE_ID} {} +bool ImageViewBase::SupportsAnisotropy() const noexcept { + const bool has_mips = range.extent.levels > 1; + const bool is_2d = type == ImageViewType::e2D || type == ImageViewType::e2DArray; + if (!has_mips || !is_2d) { + return false; + } + + switch (format) { + case PixelFormat::R8_UNORM: + case PixelFormat::R8_SNORM: + case PixelFormat::R8_SINT: + case PixelFormat::R8_UINT: + case PixelFormat::BC4_UNORM: + case PixelFormat::BC4_SNORM: + case PixelFormat::BC5_UNORM: + case PixelFormat::BC5_SNORM: + case PixelFormat::R32G32_FLOAT: + case PixelFormat::R32G32_SINT: + case PixelFormat::R32_FLOAT: + case PixelFormat::R16_FLOAT: + case PixelFormat::R16_UNORM: + case PixelFormat::R16_SNORM: + case PixelFormat::R16_UINT: + case PixelFormat::R16_SINT: + case PixelFormat::R16G16_UNORM: + case PixelFormat::R16G16_FLOAT: + case PixelFormat::R16G16_UINT: + case PixelFormat::R16G16_SINT: + case PixelFormat::R16G16_SNORM: + case PixelFormat::R8G8_UNORM: + case PixelFormat::R8G8_SNORM: + case PixelFormat::R8G8_SINT: + case PixelFormat::R8G8_UINT: + case PixelFormat::R32G32_UINT: + case PixelFormat::R32_UINT: + case PixelFormat::R32_SINT: + case PixelFormat::G4R4_UNORM: + // Depth formats + case PixelFormat::D32_FLOAT: + case PixelFormat::D16_UNORM: + // Stencil formats + case PixelFormat::S8_UINT: + // DepthStencil formats + case PixelFormat::D24_UNORM_S8_UINT: + case PixelFormat::S8_UINT_D24_UNORM: + case PixelFormat::D32_FLOAT_S8_UINT: + return false; + default: + return true; + } +} + } // namespace VideoCommon diff --git a/src/video_core/texture_cache/image_view_base.h b/src/video_core/texture_cache/image_view_base.h index a25ae1d4a..87549ffff 100644 --- a/src/video_core/texture_cache/image_view_base.h +++ b/src/video_core/texture_cache/image_view_base.h @@ -33,6 +33,8 @@ struct ImageViewBase { return type == ImageViewType::Buffer; } + [[nodiscard]] bool SupportsAnisotropy() const noexcept; + ImageId image_id{}; GPUVAddr gpu_addr = 0; PixelFormat format{}; diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h index c7f7448e9..4027d860b 100644 --- a/src/video_core/texture_cache/texture_cache.h +++ b/src/video_core/texture_cache/texture_cache.h @@ -222,30 +222,50 @@ void TextureCache<P>::CheckFeedbackLoop(std::span<const ImageViewInOut> views) { template <class P> typename P::Sampler* TextureCache<P>::GetGraphicsSampler(u32 index) { + return &slot_samplers[GetGraphicsSamplerId(index)]; +} + +template <class P> +typename P::Sampler* TextureCache<P>::GetComputeSampler(u32 index) { + return &slot_samplers[GetComputeSamplerId(index)]; +} + +template <class P> +SamplerId TextureCache<P>::GetGraphicsSamplerId(u32 index) { if (index > channel_state->graphics_sampler_table.Limit()) { LOG_DEBUG(HW_GPU, "Invalid sampler index={}", index); - return &slot_samplers[NULL_SAMPLER_ID]; + return NULL_SAMPLER_ID; } const auto [descriptor, is_new] = channel_state->graphics_sampler_table.Read(index); SamplerId& id = channel_state->graphics_sampler_ids[index]; if (is_new) { id = FindSampler(descriptor); } - return &slot_samplers[id]; + return id; } template <class P> -typename P::Sampler* TextureCache<P>::GetComputeSampler(u32 index) { +SamplerId TextureCache<P>::GetComputeSamplerId(u32 index) { if (index > channel_state->compute_sampler_table.Limit()) { LOG_DEBUG(HW_GPU, "Invalid sampler index={}", index); - return &slot_samplers[NULL_SAMPLER_ID]; + return NULL_SAMPLER_ID; } const auto [descriptor, is_new] = channel_state->compute_sampler_table.Read(index); SamplerId& id = channel_state->compute_sampler_ids[index]; if (is_new) { id = FindSampler(descriptor); } - return &slot_samplers[id]; + return id; +} + +template <class P> +const typename P::Sampler& TextureCache<P>::GetSampler(SamplerId id) const noexcept { + return slot_samplers[id]; +} + +template <class P> +typename P::Sampler& TextureCache<P>::GetSampler(SamplerId id) noexcept { + return slot_samplers[id]; } template <class P> diff --git a/src/video_core/texture_cache/texture_cache_base.h b/src/video_core/texture_cache/texture_cache_base.h index 3bfa92154..d96ddea9d 100644 --- a/src/video_core/texture_cache/texture_cache_base.h +++ b/src/video_core/texture_cache/texture_cache_base.h @@ -159,6 +159,18 @@ public: /// Get the sampler from the compute descriptor table in the specified index Sampler* GetComputeSampler(u32 index); + /// Get the sampler id from the graphics descriptor table in the specified index + SamplerId GetGraphicsSamplerId(u32 index); + + /// Get the sampler id from the compute descriptor table in the specified index + SamplerId GetComputeSamplerId(u32 index); + + /// Return a constant reference to the given sampler id + [[nodiscard]] const Sampler& GetSampler(SamplerId id) const noexcept; + + /// Return a reference to the given sampler id + [[nodiscard]] Sampler& GetSampler(SamplerId id) noexcept; + /// Refresh the state for graphics image view and sampler descriptors void SynchronizeGraphicsDescriptors(); diff --git a/src/video_core/textures/texture.cpp b/src/video_core/textures/texture.cpp index 4a80a59f9..d8b88d9bc 100644 --- a/src/video_core/textures/texture.cpp +++ b/src/video_core/textures/texture.cpp @@ -62,7 +62,12 @@ std::array<float, 4> TSCEntry::BorderColor() const noexcept { } float TSCEntry::MaxAnisotropy() const noexcept { - if (max_anisotropy == 0 && mipmap_filter != TextureMipmapFilter::Linear) { + const bool is_suitable_mipmap_filter = mipmap_filter != TextureMipmapFilter::None; + const bool has_regular_lods = min_lod_clamp == 0 && max_lod_clamp >= 256; + const bool is_bilinear_filter = min_filter == TextureFilter::Linear && + reduction_filter == SamplerReduction::WeightedAverage; + if (max_anisotropy == 0 && (!is_suitable_mipmap_filter || !has_regular_lods || + !is_bilinear_filter || depth_compare_enabled)) { return 1.0f; } const auto anisotropic_settings = Settings::values.max_anisotropy.GetValue(); |