diff options
Diffstat (limited to 'src/video_core')
21 files changed, 97 insertions, 26 deletions
diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp index ba9ba082f..789af452d 100644 --- a/src/video_core/gpu.cpp +++ b/src/video_core/gpu.cpp @@ -230,7 +230,7 @@ struct GPU::Impl { void IncrementSyncPoint(u32 syncpoint_id) { auto& syncpoint = syncpoints.at(syncpoint_id); syncpoint++; - std::lock_guard lock{sync_mutex}; + std::scoped_lock lock{sync_mutex}; sync_cv.notify_all(); auto& interrupt = syncpt_interrupts.at(syncpoint_id); if (!interrupt.empty()) { @@ -252,7 +252,7 @@ struct GPU::Impl { } void RegisterSyncptInterrupt(u32 syncpoint_id, u32 value) { - std::lock_guard lock{sync_mutex}; + std::scoped_lock lock{sync_mutex}; auto& interrupt = syncpt_interrupts.at(syncpoint_id); bool contains = std::any_of(interrupt.begin(), interrupt.end(), [value](u32 in_value) { return in_value == value; }); @@ -263,7 +263,7 @@ struct GPU::Impl { } [[nodiscard]] bool CancelSyncptInterrupt(u32 syncpoint_id, u32 value) { - std::lock_guard lock{sync_mutex}; + std::scoped_lock lock{sync_mutex}; auto& interrupt = syncpt_interrupts.at(syncpoint_id); const auto iter = std::find_if(interrupt.begin(), interrupt.end(), diff --git a/src/video_core/gpu_thread.cpp b/src/video_core/gpu_thread.cpp index 9547f277a..4e8999915 100644 --- a/src/video_core/gpu_thread.cpp +++ b/src/video_core/gpu_thread.cpp @@ -56,7 +56,7 @@ static void RunThread(std::stop_token stop_token, Core::System& system, if (next.block) { // We have to lock the write_lock to ensure that the condition_variable wait not get a // race between the check and the lock itself. - std::lock_guard lk(state.write_lock); + std::scoped_lock lk{state.write_lock}; state.cv.notify_all(); } } diff --git a/src/video_core/host_shaders/CMakeLists.txt b/src/video_core/host_shaders/CMakeLists.txt index af05d47d1..190fc6aea 100644 --- a/src/video_core/host_shaders/CMakeLists.txt +++ b/src/video_core/host_shaders/CMakeLists.txt @@ -18,6 +18,7 @@ set(SHADER_FILES full_screen_triangle.vert fxaa.frag fxaa.vert + opengl_convert_s8d24.comp opengl_copy_bc4.comp opengl_present.frag opengl_present.vert diff --git a/src/video_core/host_shaders/opengl_convert_s8d24.comp b/src/video_core/host_shaders/opengl_convert_s8d24.comp new file mode 100644 index 000000000..83e1ab176 --- /dev/null +++ b/src/video_core/host_shaders/opengl_convert_s8d24.comp @@ -0,0 +1,18 @@ +// Copyright 2022 yuzu Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#version 430 core + +layout(local_size_x = 16, local_size_y = 8) in; + +layout(binding = 0, rgba8ui) restrict uniform uimage2D destination; +layout(location = 0) uniform uvec3 size; + +void main() { + if (any(greaterThanEqual(gl_GlobalInvocationID, size))) { + return; + } + uvec4 components = imageLoad(destination, ivec2(gl_GlobalInvocationID.xy)); + imageStore(destination, ivec2(gl_GlobalInvocationID.xy), components.wxyz); +} diff --git a/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp b/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp index 9e6732abd..fd40966d5 100644 --- a/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp +++ b/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp @@ -253,7 +253,7 @@ GraphicsPipeline::GraphicsPipeline( } } if (in_parallel) { - std::lock_guard lock{built_mutex}; + std::scoped_lock lock{built_mutex}; built_fence.Create(); // Flush this context to ensure compilation commands and fence are in the GPU pipe. glFlush(); diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index e6f9ece8b..8ef79753f 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -520,6 +520,8 @@ bool RasterizerOpenGL::AccelerateDisplay(const Tegra::FramebufferConfig& config, // ASSERT_MSG(image_view->size.width == config.width, "Framebuffer width is different"); // ASSERT_MSG(image_view->size.height == config.height, "Framebuffer height is different"); + screen_info.texture.width = image_view->size.width; + screen_info.texture.height = image_view->size.height; screen_info.display_texture = image_view->Handle(Shader::TextureType::Color2D); screen_info.display_srgb = VideoCore::Surface::IsPixelFormatSRGB(image_view->format); return true; @@ -557,12 +559,19 @@ void RasterizerOpenGL::SyncViewport() { const bool dirty_viewport = flags[Dirty::Viewports] || rescale_viewports; const bool dirty_clip_control = flags[Dirty::ClipControl]; - if (dirty_clip_control || flags[Dirty::FrontFace]) { + if (dirty_viewport || dirty_clip_control || flags[Dirty::FrontFace]) { flags[Dirty::FrontFace] = false; GLenum mode = MaxwellToGL::FrontFace(regs.front_face); + bool flip_faces = false; if (regs.screen_y_control.triangle_rast_flip != 0 && regs.viewport_transform[0].scale_y < 0.0f) { + flip_faces = !flip_faces; + } + if (regs.viewport_transform[0].scale_z < 0.0f) { + flip_faces = !flip_faces; + } + if (flip_faces) { switch (mode) { case GL_CW: mode = GL_CCW; diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp index 6423992c3..05c5e702c 100644 --- a/src/video_core/renderer_opengl/gl_shader_cache.cpp +++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp @@ -258,7 +258,7 @@ void ShaderCache::LoadDiskResources(u64 title_id, std::stop_token stop_loading, [this, key, env = std::move(env), &state, &callback](Context* ctx) mutable { ctx->pools.ReleaseContents(); auto pipeline{CreateComputePipeline(ctx->pools, key, env)}; - std::lock_guard lock{state.mutex}; + std::scoped_lock lock{state.mutex}; if (pipeline) { compute_cache.emplace(key, std::move(pipeline)); } @@ -280,7 +280,7 @@ void ShaderCache::LoadDiskResources(u64 title_id, std::stop_token stop_loading, } ctx->pools.ReleaseContents(); auto pipeline{CreateGraphicsPipeline(ctx->pools, key, MakeSpan(env_ptrs), false)}; - std::lock_guard lock{state.mutex}; + std::scoped_lock lock{state.mutex}; if (pipeline) { graphics_cache.emplace(key, std::move(pipeline)); } diff --git a/src/video_core/renderer_opengl/gl_texture_cache.cpp b/src/video_core/renderer_opengl/gl_texture_cache.cpp index 8f9a65beb..d12076358 100644 --- a/src/video_core/renderer_opengl/gl_texture_cache.cpp +++ b/src/video_core/renderer_opengl/gl_texture_cache.cpp @@ -409,8 +409,8 @@ ImageBufferMap::~ImageBufferMap() { TextureCacheRuntime::TextureCacheRuntime(const Device& device_, ProgramManager& program_manager, StateTracker& state_tracker_) - : device{device_}, state_tracker{state_tracker_}, - util_shaders(program_manager), resolution{Settings::values.resolution_info} { + : device{device_}, state_tracker{state_tracker_}, util_shaders(program_manager), + format_conversion_pass{util_shaders}, resolution{Settings::values.resolution_info} { static constexpr std::array TARGETS{GL_TEXTURE_1D_ARRAY, GL_TEXTURE_2D_ARRAY, GL_TEXTURE_3D}; for (size_t i = 0; i < TARGETS.size(); ++i) { const GLenum target = TARGETS[i]; @@ -1325,6 +1325,9 @@ Framebuffer::Framebuffer(TextureCacheRuntime& runtime, std::span<ImageView*, NUM Framebuffer::~Framebuffer() = default; +FormatConversionPass::FormatConversionPass(UtilShaders& util_shaders_) + : util_shaders{util_shaders_} {} + void FormatConversionPass::ConvertImage(Image& dst_image, Image& src_image, std::span<const VideoCommon::ImageCopy> copies) { const GLenum dst_target = ImageTarget(dst_image.info); @@ -1357,6 +1360,12 @@ void FormatConversionPass::ConvertImage(Image& dst_image, Image& src_image, dst_origin.z, region.width, region.height, region.depth, dst_image.GlFormat(), dst_image.GlType(), nullptr); } + + // Swap component order of S8D24 to ABGR8 reinterprets + if (src_image.info.format == PixelFormat::D24_UNORM_S8_UINT && + dst_image.info.format == PixelFormat::A8B8G8R8_UNORM) { + util_shaders.ConvertS8D24(dst_image, copies); + } } } // namespace OpenGL diff --git a/src/video_core/renderer_opengl/gl_texture_cache.h b/src/video_core/renderer_opengl/gl_texture_cache.h index 53088b66e..672fa8dde 100644 --- a/src/video_core/renderer_opengl/gl_texture_cache.h +++ b/src/video_core/renderer_opengl/gl_texture_cache.h @@ -55,13 +55,14 @@ struct FormatProperties { class FormatConversionPass { public: - FormatConversionPass() = default; + explicit FormatConversionPass(UtilShaders& util_shaders); ~FormatConversionPass() = default; void ConvertImage(Image& dst_image, Image& src_image, std::span<const VideoCommon::ImageCopy> copies); private: + UtilShaders& util_shaders; OGLBuffer intermediate_pbo; size_t pbo_size{}; }; diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp index f8f29013a..3a3c213bb 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.cpp +++ b/src/video_core/renderer_opengl/renderer_opengl.cpp @@ -208,6 +208,8 @@ void RendererOpenGL::LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuf // Framebuffer orientation handling framebuffer_transform_flags = framebuffer.transform_flags; framebuffer_crop_rect = framebuffer.crop_rect; + framebuffer_width = framebuffer.width; + framebuffer_height = framebuffer.height; const VAddr framebuffer_addr{framebuffer.address + framebuffer.offset}; screen_info.was_accelerated = @@ -480,9 +482,12 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) { ASSERT_MSG(framebuffer_crop_rect.top == 0, "Unimplemented"); ASSERT_MSG(framebuffer_crop_rect.left == 0, "Unimplemented"); + f32 scale_u = static_cast<f32>(framebuffer_width) / static_cast<f32>(screen_info.texture.width); + f32 scale_v = + static_cast<f32>(framebuffer_height) / static_cast<f32>(screen_info.texture.height); + // Scale the output by the crop width/height. This is commonly used with 1280x720 rendering // (e.g. handheld mode) on a 1920x1080 framebuffer. - f32 scale_u = 1.f, scale_v = 1.f; if (framebuffer_crop_rect.GetWidth() > 0) { scale_u = static_cast<f32>(framebuffer_crop_rect.GetWidth()) / static_cast<f32>(screen_info.texture.width); diff --git a/src/video_core/renderer_opengl/renderer_opengl.h b/src/video_core/renderer_opengl/renderer_opengl.h index aa206878b..ae9558a33 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.h +++ b/src/video_core/renderer_opengl/renderer_opengl.h @@ -137,6 +137,8 @@ private: /// Used for transforming the framebuffer orientation Service::android::BufferTransformFlags framebuffer_transform_flags{}; Common::Rectangle<int> framebuffer_crop_rect; + u32 framebuffer_width; + u32 framebuffer_height; }; } // namespace OpenGL diff --git a/src/video_core/renderer_opengl/util_shaders.cpp b/src/video_core/renderer_opengl/util_shaders.cpp index 897c380b3..04c482a09 100644 --- a/src/video_core/renderer_opengl/util_shaders.cpp +++ b/src/video_core/renderer_opengl/util_shaders.cpp @@ -13,6 +13,7 @@ #include "video_core/host_shaders/astc_decoder_comp.h" #include "video_core/host_shaders/block_linear_unswizzle_2d_comp.h" #include "video_core/host_shaders/block_linear_unswizzle_3d_comp.h" +#include "video_core/host_shaders/opengl_convert_s8d24_comp.h" #include "video_core/host_shaders/opengl_copy_bc4_comp.h" #include "video_core/host_shaders/pitch_unswizzle_comp.h" #include "video_core/renderer_opengl/gl_shader_manager.h" @@ -50,7 +51,8 @@ UtilShaders::UtilShaders(ProgramManager& program_manager_) block_linear_unswizzle_2d_program(MakeProgram(BLOCK_LINEAR_UNSWIZZLE_2D_COMP)), block_linear_unswizzle_3d_program(MakeProgram(BLOCK_LINEAR_UNSWIZZLE_3D_COMP)), pitch_unswizzle_program(MakeProgram(PITCH_UNSWIZZLE_COMP)), - copy_bc4_program(MakeProgram(OPENGL_COPY_BC4_COMP)) { + copy_bc4_program(MakeProgram(OPENGL_COPY_BC4_COMP)), + convert_s8d24_program(MakeProgram(OPENGL_CONVERT_S8D24_COMP)) { const auto swizzle_table = Tegra::Texture::MakeSwizzleTable(); swizzle_table_buffer.Create(); glNamedBufferStorage(swizzle_table_buffer.handle, sizeof(swizzle_table), &swizzle_table, 0); @@ -248,6 +250,26 @@ void UtilShaders::CopyBC4(Image& dst_image, Image& src_image, std::span<const Im program_manager.RestoreGuestCompute(); } +void UtilShaders::ConvertS8D24(Image& dst_image, std::span<const ImageCopy> copies) { + static constexpr GLuint BINDING_DESTINATION = 0; + static constexpr GLuint LOC_SIZE = 0; + + program_manager.BindComputeProgram(convert_s8d24_program.handle); + for (const ImageCopy& copy : copies) { + ASSERT(copy.src_subresource.base_layer == 0); + ASSERT(copy.src_subresource.num_layers == 1); + ASSERT(copy.dst_subresource.base_layer == 0); + ASSERT(copy.dst_subresource.num_layers == 1); + + glUniform3ui(LOC_SIZE, copy.extent.width, copy.extent.height, copy.extent.depth); + glBindImageTexture(BINDING_DESTINATION, dst_image.StorageHandle(), + copy.dst_subresource.base_level, GL_TRUE, 0, GL_READ_WRITE, GL_RGBA8UI); + glDispatchCompute(Common::DivCeil(copy.extent.width, 16u), + Common::DivCeil(copy.extent.height, 8u), copy.extent.depth); + } + program_manager.RestoreGuestCompute(); +} + GLenum StoreFormat(u32 bytes_per_block) { switch (bytes_per_block) { case 1: diff --git a/src/video_core/renderer_opengl/util_shaders.h b/src/video_core/renderer_opengl/util_shaders.h index 5de95ea7a..5c132e67f 100644 --- a/src/video_core/renderer_opengl/util_shaders.h +++ b/src/video_core/renderer_opengl/util_shaders.h @@ -39,6 +39,8 @@ public: void CopyBC4(Image& dst_image, Image& src_image, std::span<const VideoCommon::ImageCopy> copies); + void ConvertS8D24(Image& dst_image, std::span<const VideoCommon::ImageCopy> copies); + private: ProgramManager& program_manager; @@ -49,6 +51,7 @@ private: OGLProgram block_linear_unswizzle_3d_program; OGLProgram pitch_unswizzle_program; OGLProgram copy_bc4_program; + OGLProgram convert_s8d24_program; }; GLenum StoreFormat(u32 bytes_per_block); diff --git a/src/video_core/renderer_vulkan/pipeline_statistics.cpp b/src/video_core/renderer_vulkan/pipeline_statistics.cpp index bfec931a6..7ccadf084 100644 --- a/src/video_core/renderer_vulkan/pipeline_statistics.cpp +++ b/src/video_core/renderer_vulkan/pipeline_statistics.cpp @@ -57,7 +57,7 @@ void PipelineStatistics::Collect(VkPipeline pipeline) { stage_stats.basic_block_count = GetUint64(statistic); } } - std::lock_guard lock{mutex}; + std::scoped_lock lock{mutex}; collected_stats.push_back(stage_stats); } } @@ -66,7 +66,7 @@ void PipelineStatistics::Report() const { double num{}; Stats total; { - std::lock_guard lock{mutex}; + std::scoped_lock lock{mutex}; for (const Stats& stats : collected_stats) { total.code_size += stats.code_size; total.register_count += stats.register_count; diff --git a/src/video_core/renderer_vulkan/vk_blit_screen.cpp b/src/video_core/renderer_vulkan/vk_blit_screen.cpp index d893c1952..b866e9103 100644 --- a/src/video_core/renderer_vulkan/vk_blit_screen.cpp +++ b/src/video_core/renderer_vulkan/vk_blit_screen.cpp @@ -1406,8 +1406,9 @@ void VKBlitScreen::SetVertexData(BufferData& data, const Tegra::FramebufferConfi UNIMPLEMENTED_IF(framebuffer_crop_rect.top != 0); UNIMPLEMENTED_IF(framebuffer_crop_rect.left != 0); - f32 scale_u = 1.0f; - f32 scale_v = 1.0f; + f32 scale_u = static_cast<f32>(framebuffer.width) / static_cast<f32>(screen_info.width); + f32 scale_v = static_cast<f32>(framebuffer.height) / static_cast<f32>(screen_info.height); + // Scale the output by the crop width/height. This is commonly used with 1280x720 rendering // (e.g. handheld mode) on a 1920x1080 framebuffer. if (!fsr) { diff --git a/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp b/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp index de36bcdb7..97b3594c2 100644 --- a/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp +++ b/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp @@ -77,7 +77,7 @@ ComputePipeline::ComputePipeline(const Device& device_, DescriptorPool& descript if (pipeline_statistics) { pipeline_statistics->Collect(*pipeline); } - std::lock_guard lock{build_mutex}; + std::scoped_lock lock{build_mutex}; is_built = true; build_condvar.notify_one(); if (shader_notify) { diff --git a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp index d514b71d0..8959d6059 100644 --- a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp +++ b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp @@ -258,7 +258,7 @@ GraphicsPipeline::GraphicsPipeline( pipeline_statistics->Collect(*pipeline); } - std::lock_guard lock{build_mutex}; + std::scoped_lock lock{build_mutex}; is_built = true; build_condvar.notify_one(); if (shader_notify) { diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp index 27e59df73..336d1e9dc 100644 --- a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp @@ -404,7 +404,7 @@ void PipelineCache::LoadDiskResources(u64 title_id, std::stop_token stop_loading workers.QueueWork([this, key, env = std::move(env), &state, &callback]() mutable { ShaderPools pools; auto pipeline{CreateComputePipeline(pools, key, env, state.statistics.get(), false)}; - std::lock_guard lock{state.mutex}; + std::scoped_lock lock{state.mutex}; if (pipeline) { compute_cache.emplace(key, std::move(pipeline)); } @@ -434,7 +434,7 @@ void PipelineCache::LoadDiskResources(u64 title_id, std::stop_token stop_loading auto pipeline{CreateGraphicsPipeline(pools, key, MakeSpan(env_ptrs), state.statistics.get(), false)}; - std::lock_guard lock{state.mutex}; + std::scoped_lock lock{state.mutex}; graphics_cache.emplace(key, std::move(pipeline)); ++state.built; if (state.has_loaded) { diff --git a/src/video_core/renderer_vulkan/vk_render_pass_cache.cpp b/src/video_core/renderer_vulkan/vk_render_pass_cache.cpp index 451ffe019..d22bb6694 100644 --- a/src/video_core/renderer_vulkan/vk_render_pass_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_render_pass_cache.cpp @@ -36,7 +36,7 @@ VkAttachmentDescription AttachmentDescription(const Device& device, PixelFormat RenderPassCache::RenderPassCache(const Device& device_) : device{&device_} {} VkRenderPass RenderPassCache::Get(const RenderPassKey& key) { - std::lock_guard lock{mutex}; + std::scoped_lock lock{mutex}; const auto [pair, is_new] = cache.try_emplace(key); if (!is_new) { return *pair->second; diff --git a/src/video_core/renderer_vulkan/vk_scheduler.cpp b/src/video_core/renderer_vulkan/vk_scheduler.cpp index ad320991b..6a9416457 100644 --- a/src/video_core/renderer_vulkan/vk_scheduler.cpp +++ b/src/video_core/renderer_vulkan/vk_scheduler.cpp @@ -73,7 +73,7 @@ void VKScheduler::DispatchWork() { return; } { - std::lock_guard lock{work_mutex}; + std::scoped_lock lock{work_mutex}; work_queue.push(std::move(chunk)); } work_cv.notify_one(); @@ -157,7 +157,7 @@ void VKScheduler::WorkerThread(std::stop_token stop_token) { if (has_submit) { AllocateWorkerCommandBuffer(); } - std::lock_guard reserve_lock{reserve_mutex}; + std::scoped_lock reserve_lock{reserve_mutex}; chunk_reserve.push_back(std::move(work)); } while (!stop_token.stop_requested()); } @@ -282,7 +282,7 @@ void VKScheduler::EndRenderPass() { } void VKScheduler::AcquireNewChunk() { - std::lock_guard lock{reserve_mutex}; + std::scoped_lock lock{reserve_mutex}; if (chunk_reserve.empty()) { chunk = std::make_unique<CommandChunk>(); return; diff --git a/src/video_core/shader_cache.cpp b/src/video_core/shader_cache.cpp index 87636857d..75031767a 100644 --- a/src/video_core/shader_cache.cpp +++ b/src/video_core/shader_cache.cpp @@ -25,7 +25,7 @@ void ShaderCache::InvalidateRegion(VAddr addr, size_t size) { } void ShaderCache::OnCPUWrite(VAddr addr, size_t size) { - std::lock_guard lock{invalidation_mutex}; + std::scoped_lock lock{invalidation_mutex}; InvalidatePagesInRegion(addr, size); } |