diff options
Diffstat (limited to 'src/video_core')
36 files changed, 329 insertions, 379 deletions
diff --git a/src/video_core/control/channel_state_cache.h b/src/video_core/control/channel_state_cache.h index cdaf4f8d5..46bc9e322 100644 --- a/src/video_core/control/channel_state_cache.h +++ b/src/video_core/control/channel_state_cache.h @@ -44,7 +44,7 @@ public: template <class P> class ChannelSetupCaches { public: - /// Operations for seting the channel of execution. + /// Operations for setting the channel of execution. virtual ~ChannelSetupCaches(); /// Create channel state. diff --git a/src/video_core/engines/draw_manager.cpp b/src/video_core/engines/draw_manager.cpp index 1d22d25f1..0e94c521a 100644 --- a/src/video_core/engines/draw_manager.cpp +++ b/src/video_core/engines/draw_manager.cpp @@ -164,6 +164,7 @@ void DrawManager::DrawEnd(u32 instance_count, bool force_draw) { draw_state.index_buffer.count = static_cast<u32>(draw_state.inline_index_draw_indexes.size() / 4); draw_state.index_buffer.format = Maxwell3D::Regs::IndexFormat::UnsignedInt; + maxwell3d->dirty.flags[VideoCommon::Dirty::IndexBuffer] = true; ProcessDraw(true, instance_count); draw_state.inline_index_draw_indexes.clear(); break; diff --git a/src/video_core/engines/sw_blitter/blitter.cpp b/src/video_core/engines/sw_blitter/blitter.cpp index 2f1ea4626..3c9f38559 100644 --- a/src/video_core/engines/sw_blitter/blitter.cpp +++ b/src/video_core/engines/sw_blitter/blitter.cpp @@ -193,7 +193,7 @@ bool SoftwareBlitEngine::Blit(Fermi2D::Surface& src, Fermi2D::Surface& dst, output_converter->ConvertFrom(impl->intermediate_dst, impl->dst_buffer); }; - // Do actuall Blit + // Do actual Blit impl->dst_buffer.resize(dst_copy_size); if (src.linear == Fermi2D::MemoryLayout::BlockLinear) { diff --git a/src/video_core/gpu_thread.cpp b/src/video_core/gpu_thread.cpp index 9c103c0d4..050b11874 100644 --- a/src/video_core/gpu_thread.cpp +++ b/src/video_core/gpu_thread.cpp @@ -25,7 +25,7 @@ static void RunThread(std::stop_token stop_token, Core::System& system, SCOPE_EXIT({ MicroProfileOnThreadExit(); }); Common::SetCurrentThreadName(name.c_str()); - Common::SetCurrentThreadPriority(Common::ThreadPriority::High); + Common::SetCurrentThreadPriority(Common::ThreadPriority::Critical); system.RegisterHostThread(); auto current_context = context.Acquire(); diff --git a/src/video_core/host_shaders/astc_decoder.comp b/src/video_core/host_shaders/astc_decoder.comp index d608678a3..bf2693559 100644 --- a/src/video_core/host_shaders/astc_decoder.comp +++ b/src/video_core/host_shaders/astc_decoder.comp @@ -125,7 +125,7 @@ uvec4 local_buff; uvec4 color_endpoint_data; int color_bitsread = 0; -// Four values, two endpoints, four maximum paritions +// Four values, two endpoints, four maximum partitions uint color_values[32]; int colvals_index = 0; diff --git a/src/video_core/host_shaders/opengl_smaa.glsl b/src/video_core/host_shaders/opengl_smaa.glsl index 3cbe87bbf..419f89bca 100644 --- a/src/video_core/host_shaders/opengl_smaa.glsl +++ b/src/video_core/host_shaders/opengl_smaa.glsl @@ -97,7 +97,7 @@ * half-rate linear filtering on GCN. * * If SMAA is applied to 64-bit color buffers, switching to point filtering - * when accesing them will increase the performance. Search for + * when accessing them will increase the performance. Search for * 'SMAASamplePoint' to see which textures may benefit from point * filtering, and where (which is basically the color input in the edge * detection and resolve passes). diff --git a/src/video_core/memory_manager.h b/src/video_core/memory_manager.h index cf56392ef..51ae2de68 100644 --- a/src/video_core/memory_manager.h +++ b/src/video_core/memory_manager.h @@ -103,8 +103,8 @@ public: /** * Returns a vector with all the subranges of cpu addresses mapped beneath. - * if the region is continous, a single pair will be returned. If it's unmapped, an empty vector - * will be returned; + * if the region is continuous, a single pair will be returned. If it's unmapped, an empty + * vector will be returned; */ std::vector<std::pair<GPUVAddr, std::size_t>> GetSubmappedRange(GPUVAddr gpu_addr, std::size_t size) const; diff --git a/src/video_core/query_cache.h b/src/video_core/query_cache.h index 00ce53e3e..8906ba6d8 100644 --- a/src/video_core/query_cache.h +++ b/src/video_core/query_cache.h @@ -341,7 +341,7 @@ public: /// Flushes the query to guest memory. virtual void Flush() { - // When counter is nullptr it means that it's just been reseted. We are supposed to write a + // When counter is nullptr it means that it's just been reset. We are supposed to write a // zero in these cases. const u64 value = counter ? counter->Query() : 0; std::memcpy(host_ptr, &value, sizeof(u64)); diff --git a/src/video_core/renderer_opengl/blit_image.cpp b/src/video_core/renderer_opengl/blit_image.cpp index 9a560a73b..3b03e8d5a 100644 --- a/src/video_core/renderer_opengl/blit_image.cpp +++ b/src/video_core/renderer_opengl/blit_image.cpp @@ -22,7 +22,7 @@ BlitImageHelper::~BlitImageHelper() = default; void BlitImageHelper::BlitColor(GLuint dst_framebuffer, GLuint src_image_view, GLuint src_sampler, const Region2D& dst_region, const Region2D& src_region, const Extent3D& src_size) { - glEnable(GL_CULL_FACE); + glDisable(GL_CULL_FACE); glDisable(GL_COLOR_LOGIC_OP); glDisable(GL_DEPTH_TEST); glDisable(GL_STENCIL_TEST); @@ -31,7 +31,6 @@ void BlitImageHelper::BlitColor(GLuint dst_framebuffer, GLuint src_image_view, G glDisable(GL_ALPHA_TEST); glDisablei(GL_BLEND, 0); glPolygonMode(GL_FRONT_AND_BACK, GL_FILL); - glCullFace(GL_BACK); glFrontFace(GL_CW); glColorMaski(0, GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE); glDepthRangeIndexed(0, 0.0, 0.0); diff --git a/src/video_core/renderer_opengl/gl_fence_manager.cpp b/src/video_core/renderer_opengl/gl_fence_manager.cpp index 91463f854..5326172af 100644 --- a/src/video_core/renderer_opengl/gl_fence_manager.cpp +++ b/src/video_core/renderer_opengl/gl_fence_manager.cpp @@ -27,9 +27,7 @@ bool GLInnerFence::IsSignaled() const { return true; } ASSERT(sync_object.handle != 0); - GLint sync_status; - glGetSynciv(sync_object.handle, GL_SYNC_STATUS, 1, nullptr, &sync_status); - return sync_status == GL_SIGNALED; + return sync_object.IsSignaled(); } void GLInnerFence::Wait() { diff --git a/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp b/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp index 29491e762..89000d6e0 100644 --- a/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp +++ b/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp @@ -621,10 +621,7 @@ bool GraphicsPipeline::IsBuilt() noexcept { if (built_fence.handle == 0) { return false; } - // Timeout of zero means this is non-blocking - const auto sync_status = glClientWaitSync(built_fence.handle, 0, 0); - ASSERT(sync_status != GL_WAIT_FAILED); - is_built = sync_status != GL_TIMEOUT_EXPIRED; + is_built = built_fence.IsSignaled(); return is_built; } diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index 7bced675c..90e35e307 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -63,7 +63,7 @@ RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& emu_window_, Tegra buffer_cache(*this, cpu_memory_, buffer_cache_runtime), shader_cache(*this, emu_window_, device, texture_cache, buffer_cache, program_manager, state_tracker, gpu.ShaderNotify()), - query_cache(*this), accelerate_dma(buffer_cache), + query_cache(*this), accelerate_dma(buffer_cache, texture_cache), fence_manager(*this, gpu, texture_cache, buffer_cache, query_cache), blit_image(program_manager_) {} @@ -357,6 +357,7 @@ void RasterizerOpenGL::DrawTexture() { .y = static_cast<s32>(draw_texture_state.src_y1)}}; blit_image.BlitColor(texture_cache.GetFramebuffer()->Handle(), texture.DefaultHandle(), sampler->Handle(), dst_region, src_region, texture.size); + state_tracker.InvalidateState(); } ++num_queued_commands; @@ -576,7 +577,7 @@ bool RasterizerOpenGL::AccelerateConditionalRendering() { // Reimplement Host conditional rendering. return false; } - // Medium / Low Hack: stub any checks on queries writen into the buffer cache. + // Medium / Low Hack: stub any checks on queries written into the buffer cache. const GPUVAddr condition_address{maxwell3d->regs.render_enable.Address()}; Maxwell::ReportSemaphore::Compare cmp; if (gpu_memory->IsMemoryDirty(condition_address, sizeof(cmp), @@ -1262,7 +1263,8 @@ void RasterizerOpenGL::ReleaseChannel(s32 channel_id) { query_cache.EraseChannel(channel_id); } -AccelerateDMA::AccelerateDMA(BufferCache& buffer_cache_) : buffer_cache{buffer_cache_} {} +AccelerateDMA::AccelerateDMA(BufferCache& buffer_cache_, TextureCache& texture_cache_) + : buffer_cache{buffer_cache_}, texture_cache{texture_cache_} {} bool AccelerateDMA::BufferCopy(GPUVAddr src_address, GPUVAddr dest_address, u64 amount) { std::scoped_lock lock{buffer_cache.mutex}; @@ -1274,4 +1276,44 @@ bool AccelerateDMA::BufferClear(GPUVAddr src_address, u64 amount, u32 value) { return buffer_cache.DMAClear(src_address, amount, value); } +template <bool IS_IMAGE_UPLOAD> +bool AccelerateDMA::DmaBufferImageCopy(const Tegra::DMA::ImageCopy& copy_info, + const Tegra::DMA::BufferOperand& buffer_operand, + const Tegra::DMA::ImageOperand& image_operand) { + std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex}; + const auto image_id = texture_cache.DmaImageId(image_operand); + if (image_id == VideoCommon::NULL_IMAGE_ID) { + return false; + } + const u32 buffer_size = static_cast<u32>(buffer_operand.pitch * buffer_operand.height); + static constexpr auto sync_info = VideoCommon::ObtainBufferSynchronize::FullSynchronize; + const auto post_op = IS_IMAGE_UPLOAD ? VideoCommon::ObtainBufferOperation::DoNothing + : VideoCommon::ObtainBufferOperation::MarkAsWritten; + const auto [buffer, offset] = + buffer_cache.ObtainBuffer(buffer_operand.address, buffer_size, sync_info, post_op); + + const auto [image, copy] = texture_cache.DmaBufferImageCopy( + copy_info, buffer_operand, image_operand, image_id, IS_IMAGE_UPLOAD); + const std::span copy_span{©, 1}; + + if constexpr (IS_IMAGE_UPLOAD) { + image->UploadMemory(buffer->Handle(), offset, copy_span); + } else { + image->DownloadMemory(buffer->Handle(), offset, copy_span); + } + return true; +} + +bool AccelerateDMA::ImageToBuffer(const Tegra::DMA::ImageCopy& copy_info, + const Tegra::DMA::ImageOperand& image_operand, + const Tegra::DMA::BufferOperand& buffer_operand) { + return DmaBufferImageCopy<false>(copy_info, buffer_operand, image_operand); +} + +bool AccelerateDMA::BufferToImage(const Tegra::DMA::ImageCopy& copy_info, + const Tegra::DMA::BufferOperand& buffer_operand, + const Tegra::DMA::ImageOperand& image_operand) { + return DmaBufferImageCopy<true>(copy_info, buffer_operand, image_operand); +} + } // namespace OpenGL diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h index 7e21fc43d..ad6978bd0 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.h +++ b/src/video_core/renderer_opengl/gl_rasterizer.h @@ -50,24 +50,26 @@ static_assert(sizeof(BindlessSSBO) * CHAR_BIT == 128); class AccelerateDMA : public Tegra::Engines::AccelerateDMAInterface { public: - explicit AccelerateDMA(BufferCache& buffer_cache); + explicit AccelerateDMA(BufferCache& buffer_cache, TextureCache& texture_cache); bool BufferCopy(GPUVAddr src_address, GPUVAddr dest_address, u64 amount) override; bool BufferClear(GPUVAddr src_address, u64 amount, u32 value) override; bool ImageToBuffer(const Tegra::DMA::ImageCopy& copy_info, const Tegra::DMA::ImageOperand& src, - const Tegra::DMA::BufferOperand& dst) override { - return false; - } + const Tegra::DMA::BufferOperand& dst) override; bool BufferToImage(const Tegra::DMA::ImageCopy& copy_info, const Tegra::DMA::BufferOperand& src, - const Tegra::DMA::ImageOperand& dst) override { - return false; - } + const Tegra::DMA::ImageOperand& dst) override; private: + template <bool IS_IMAGE_UPLOAD> + bool DmaBufferImageCopy(const Tegra::DMA::ImageCopy& copy_info, + const Tegra::DMA::BufferOperand& src, + const Tegra::DMA::ImageOperand& dst); + BufferCache& buffer_cache; + TextureCache& texture_cache; }; class RasterizerOpenGL : public VideoCore::RasterizerAccelerated, @@ -160,7 +162,7 @@ private: /// Syncs the cull mode to match the guest state void SyncCullMode(); - /// Syncs the primitve restart to match the guest state + /// Syncs the primitive restart to match the guest state void SyncPrimitiveRestart(); /// Syncs the depth test state to match the guest state @@ -244,7 +246,7 @@ private: std::array<GLuint, MAX_TEXTURES> texture_handles{}; std::array<GLuint, MAX_IMAGES> image_handles{}; - /// Number of commands queued to the OpenGL driver. Resetted on flush. + /// Number of commands queued to the OpenGL driver. Reset on flush. size_t num_queued_commands = 0; bool has_written_global_memory = false; diff --git a/src/video_core/renderer_opengl/gl_resource_manager.cpp b/src/video_core/renderer_opengl/gl_resource_manager.cpp index 3a664fdec..eae8fd110 100644 --- a/src/video_core/renderer_opengl/gl_resource_manager.cpp +++ b/src/video_core/renderer_opengl/gl_resource_manager.cpp @@ -3,6 +3,7 @@ #include <string_view> #include <glad/glad.h> +#include "common/assert.h" #include "common/microprofile.h" #include "video_core/renderer_opengl/gl_resource_manager.h" #include "video_core/renderer_opengl/gl_shader_util.h" @@ -158,6 +159,15 @@ void OGLSync::Release() { handle = 0; } +bool OGLSync::IsSignaled() const noexcept { + // At least on Nvidia, glClientWaitSync with a timeout of 0 + // is faster than glGetSynciv of GL_SYNC_STATUS. + // Timeout of 0 means this check is non-blocking. + const auto sync_status = glClientWaitSync(handle, 0, 0); + ASSERT(sync_status != GL_WAIT_FAILED); + return sync_status != GL_TIMEOUT_EXPIRED; +} + void OGLFramebuffer::Create() { if (handle != 0) return; diff --git a/src/video_core/renderer_opengl/gl_resource_manager.h b/src/video_core/renderer_opengl/gl_resource_manager.h index bc05ba4bd..77362acd2 100644 --- a/src/video_core/renderer_opengl/gl_resource_manager.h +++ b/src/video_core/renderer_opengl/gl_resource_manager.h @@ -263,6 +263,9 @@ public: /// Deletes the internal OpenGL resource void Release(); + /// Checks if the sync has been signaled + bool IsSignaled() const noexcept; + GLsync handle = 0; }; diff --git a/src/video_core/renderer_opengl/gl_texture_cache.cpp b/src/video_core/renderer_opengl/gl_texture_cache.cpp index d3eabd686..0b9c4a904 100644 --- a/src/video_core/renderer_opengl/gl_texture_cache.cpp +++ b/src/video_core/renderer_opengl/gl_texture_cache.cpp @@ -717,9 +717,7 @@ std::optional<size_t> TextureCacheRuntime::StagingBuffers::FindBuffer(size_t req continue; } if (syncs[index].handle != 0) { - GLint status; - glGetSynciv(syncs[index].handle, GL_SYNC_STATUS, 1, nullptr, &status); - if (status != GL_SIGNALED) { + if (!syncs[index].IsSignaled()) { continue; } syncs[index].Release(); @@ -765,14 +763,14 @@ Image::Image(const VideoCommon::NullImageParams& params) : VideoCommon::ImageBas Image::~Image() = default; -void Image::UploadMemory(const ImageBufferMap& map, +void Image::UploadMemory(GLuint buffer_handle, size_t buffer_offset, std::span<const VideoCommon::BufferImageCopy> copies) { const bool is_rescaled = True(flags & ImageFlagBits::Rescaled); if (is_rescaled) { ScaleDown(true); } - glBindBuffer(GL_PIXEL_UNPACK_BUFFER, map.buffer); - glFlushMappedBufferRange(GL_PIXEL_UNPACK_BUFFER, map.offset, unswizzled_size_bytes); + glBindBuffer(GL_PIXEL_UNPACK_BUFFER, buffer_handle); + glFlushMappedBufferRange(GL_PIXEL_UNPACK_BUFFER, buffer_offset, unswizzled_size_bytes); glPixelStorei(GL_UNPACK_ALIGNMENT, 1); @@ -791,21 +789,26 @@ void Image::UploadMemory(const ImageBufferMap& map, current_image_height = copy.buffer_image_height; glPixelStorei(GL_UNPACK_IMAGE_HEIGHT, current_image_height); } - CopyBufferToImage(copy, map.offset); + CopyBufferToImage(copy, buffer_offset); } if (is_rescaled) { ScaleUp(); } } -void Image::DownloadMemory(ImageBufferMap& map, +void Image::UploadMemory(const ImageBufferMap& map, + std::span<const VideoCommon::BufferImageCopy> copies) { + UploadMemory(map.buffer, map.offset, copies); +} + +void Image::DownloadMemory(GLuint buffer_handle, size_t buffer_offset, std::span<const VideoCommon::BufferImageCopy> copies) { const bool is_rescaled = True(flags & ImageFlagBits::Rescaled); if (is_rescaled) { ScaleDown(); } glMemoryBarrier(GL_PIXEL_BUFFER_BARRIER_BIT); // TODO: Move this to its own API - glBindBuffer(GL_PIXEL_PACK_BUFFER, map.buffer); + glBindBuffer(GL_PIXEL_PACK_BUFFER, buffer_handle); glPixelStorei(GL_PACK_ALIGNMENT, 1); u32 current_row_length = std::numeric_limits<u32>::max(); @@ -823,13 +826,18 @@ void Image::DownloadMemory(ImageBufferMap& map, current_image_height = copy.buffer_image_height; glPixelStorei(GL_PACK_IMAGE_HEIGHT, current_image_height); } - CopyImageToBuffer(copy, map.offset); + CopyImageToBuffer(copy, buffer_offset); } if (is_rescaled) { ScaleUp(true); } } +void Image::DownloadMemory(ImageBufferMap& map, + std::span<const VideoCommon::BufferImageCopy> copies) { + DownloadMemory(map.buffer, map.offset, copies); +} + GLuint Image::StorageHandle() noexcept { switch (info.format) { case PixelFormat::A8B8G8R8_SRGB: diff --git a/src/video_core/renderer_opengl/gl_texture_cache.h b/src/video_core/renderer_opengl/gl_texture_cache.h index e30875496..911e4607a 100644 --- a/src/video_core/renderer_opengl/gl_texture_cache.h +++ b/src/video_core/renderer_opengl/gl_texture_cache.h @@ -206,9 +206,15 @@ public: Image(Image&&) = default; Image& operator=(Image&&) = default; + void UploadMemory(GLuint buffer_handle, size_t buffer_offset, + std::span<const VideoCommon::BufferImageCopy> copies); + void UploadMemory(const ImageBufferMap& map, std::span<const VideoCommon::BufferImageCopy> copies); + void DownloadMemory(GLuint buffer_handle, size_t buffer_offset, + std::span<const VideoCommon::BufferImageCopy> copies); + void DownloadMemory(ImageBufferMap& map, std::span<const VideoCommon::BufferImageCopy> copies); GLuint StorageHandle() noexcept; diff --git a/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp b/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp index f8398b511..e7df32d84 100644 --- a/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp +++ b/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp @@ -271,7 +271,7 @@ bool FixedPipelineState::operator==(const FixedPipelineState& rhs) const noexcep u32 FixedPipelineState::PackComparisonOp(Maxwell::ComparisonOp op) noexcept { // OpenGL enums go from 0x200 to 0x207 and the others from 1 to 8 - // If we substract 0x200 to OpenGL enums and 1 to the others we get a 0-7 range. + // If we subtract 0x200 to OpenGL enums and 1 to the others we get a 0-7 range. // Perfect for a hash. const u32 value = static_cast<u32>(op); return value - (value >= 0x200 ? 0x200 : 1); @@ -322,8 +322,8 @@ Maxwell::StencilOp::Op FixedPipelineState::UnpackStencilOp(u32 packed) noexcept } u32 FixedPipelineState::PackCullFace(Maxwell::CullFace cull) noexcept { - // FrontAndBack is 0x408, by substracting 0x406 in it we get 2. - // Individual cull faces are in 0x404 and 0x405, substracting 0x404 we get 0 and 1. + // FrontAndBack is 0x408, by subtracting 0x406 in it we get 2. + // Individual cull faces are in 0x404 and 0x405, subtracting 0x404 we get 0 and 1. const u32 value = static_cast<u32>(cull); return value - (value == 0x408 ? 0x406 : 0x404); } diff --git a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp index ca52e2389..5dce51be8 100644 --- a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp +++ b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp @@ -166,7 +166,7 @@ struct FormatTuple { {VK_FORMAT_R16G16_UINT, Attachable | Storage}, // R16G16_UINT {VK_FORMAT_R16G16_SINT, Attachable | Storage}, // R16G16_SINT {VK_FORMAT_R16G16_SNORM, Attachable | Storage}, // R16G16_SNORM - {VK_FORMAT_UNDEFINED}, // R32G32B32_FLOAT + {VK_FORMAT_R32G32B32_SFLOAT}, // R32G32B32_FLOAT {VK_FORMAT_A8B8G8R8_SRGB_PACK32, Attachable}, // A8B8G8R8_SRGB {VK_FORMAT_R8G8_UNORM, Attachable | Storage}, // R8G8_UNORM {VK_FORMAT_R8G8_SNORM, Attachable | Storage}, // R8G8_SNORM @@ -234,11 +234,6 @@ FormatInfo SurfaceFormat(const Device& device, FormatType format_type, bool with PixelFormat pixel_format) { ASSERT(static_cast<size_t>(pixel_format) < std::size(tex_format_tuples)); FormatTuple tuple = tex_format_tuples[static_cast<size_t>(pixel_format)]; - if (tuple.format == VK_FORMAT_UNDEFINED) { - UNIMPLEMENTED_MSG("Unimplemented texture format with pixel format={}", pixel_format); - return FormatInfo{VK_FORMAT_A8B8G8R8_UNORM_PACK32, true, true}; - } - // Use A8B8G8R8_UNORM on hardware that doesn't support ASTC natively if (!device.IsOptimalAstcSupported() && VideoCore::Surface::IsPixelFormatASTC(pixel_format)) { const bool is_srgb = with_srgb && VideoCore::Surface::IsPixelFormatSRGB(pixel_format); diff --git a/src/video_core/renderer_vulkan/vk_command_pool.cpp b/src/video_core/renderer_vulkan/vk_command_pool.cpp index 2f09de1c1..d0dbf7ca5 100644 --- a/src/video_core/renderer_vulkan/vk_command_pool.cpp +++ b/src/video_core/renderer_vulkan/vk_command_pool.cpp @@ -22,8 +22,8 @@ CommandPool::CommandPool(MasterSemaphore& master_semaphore_, const Device& devic CommandPool::~CommandPool() = default; void CommandPool::Allocate(size_t begin, size_t end) { - // Command buffers are going to be commited, recorded, executed every single usage cycle. - // They are also going to be reseted when commited. + // Command buffers are going to be committed, recorded, executed every single usage cycle. + // They are also going to be reset when committed. Pool& pool = pools.emplace_back(); pool.handle = device.GetLogical().CreateCommandPool({ .sType = VK_STRUCTURE_TYPE_COMMAND_POOL_CREATE_INFO, diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp index 25965b684..673ab478e 100644 --- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp +++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp @@ -671,7 +671,7 @@ bool RasterizerVulkan::AccelerateConditionalRendering() { // TODO(Blinkhawk): Reimplement Host conditional rendering. return false; } - // Medium / Low Hack: stub any checks on queries writen into the buffer cache. + // Medium / Low Hack: stub any checks on queries written into the buffer cache. const GPUVAddr condition_address{maxwell3d->regs.render_enable.Address()}; Maxwell::ReportSemaphore::Compare cmp; if (gpu_memory->IsMemoryDirty(condition_address, sizeof(cmp), @@ -770,232 +770,44 @@ bool AccelerateDMA::BufferCopy(GPUVAddr src_address, GPUVAddr dest_address, u64 return buffer_cache.DMACopy(src_address, dest_address, amount); } -bool AccelerateDMA::ImageToBuffer(const Tegra::DMA::ImageCopy& copy_info, - const Tegra::DMA::ImageOperand& src, - const Tegra::DMA::BufferOperand& dst) { +template <bool IS_IMAGE_UPLOAD> +bool AccelerateDMA::DmaBufferImageCopy(const Tegra::DMA::ImageCopy& copy_info, + const Tegra::DMA::BufferOperand& buffer_operand, + const Tegra::DMA::ImageOperand& image_operand) { std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex}; - auto query_image = texture_cache.ObtainImage(src, false); - if (!query_image) { + const auto image_id = texture_cache.DmaImageId(image_operand); + if (image_id == VideoCommon::NULL_IMAGE_ID) { return false; } - auto* image = query_image->first; - auto [level, base] = query_image->second; - const u32 buffer_size = static_cast<u32>(dst.pitch * dst.height); - const auto [buffer, offset] = buffer_cache.ObtainBuffer( - dst.address, buffer_size, VideoCommon::ObtainBufferSynchronize::FullSynchronize, - VideoCommon::ObtainBufferOperation::MarkAsWritten); - - const bool is_rescaled = image->IsRescaled(); - if (is_rescaled) { - image->ScaleDown(); - } - VkImageSubresourceLayers subresources{ - .aspectMask = image->AspectMask(), - .mipLevel = level, - .baseArrayLayer = base, - .layerCount = 1, - }; - const u32 bpp = VideoCore::Surface::BytesPerBlock(image->info.format); - const auto convert = [old_bpp = src.bytes_per_pixel, bpp](u32 value) { - return (old_bpp * value) / bpp; - }; - const u32 base_x = convert(src.params.origin.x.Value()); - const u32 base_y = src.params.origin.y.Value(); - const u32 length_x = convert(copy_info.length_x); - const u32 length_y = copy_info.length_y; - VkOffset3D image_offset{ - .x = static_cast<s32>(base_x), - .y = static_cast<s32>(base_y), - .z = 0, - }; - VkExtent3D image_extent{ - .width = length_x, - .height = length_y, - .depth = 1, - }; - auto buff_info(dst); - buff_info.pitch = convert(dst.pitch); - scheduler.RequestOutsideRenderPassOperationContext(); - scheduler.Record([src_image = image->Handle(), dst_buffer = buffer->Handle(), - buffer_offset = offset, subresources, image_offset, image_extent, - buff_info](vk::CommandBuffer cmdbuf) { - const std::array buffer_copy_info{ - VkBufferImageCopy{ - .bufferOffset = buffer_offset, - .bufferRowLength = buff_info.pitch, - .bufferImageHeight = buff_info.height, - .imageSubresource = subresources, - .imageOffset = image_offset, - .imageExtent = image_extent, - }, - }; - const VkImageSubresourceRange range{ - .aspectMask = subresources.aspectMask, - .baseMipLevel = subresources.mipLevel, - .levelCount = 1, - .baseArrayLayer = subresources.baseArrayLayer, - .layerCount = 1, - }; - static constexpr VkMemoryBarrier WRITE_BARRIER{ - .sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER, - .pNext = nullptr, - .srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT, - .dstAccessMask = VK_ACCESS_MEMORY_READ_BIT | VK_ACCESS_MEMORY_WRITE_BIT, - }; - const std::array pre_barriers{ - VkImageMemoryBarrier{ - .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER, - .pNext = nullptr, - .srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT | - VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT | - VK_ACCESS_TRANSFER_WRITE_BIT, - .dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT, - .oldLayout = VK_IMAGE_LAYOUT_GENERAL, - .newLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, - .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED, - .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED, - .image = src_image, - .subresourceRange = range, - }, - }; - const std::array post_barriers{ - VkImageMemoryBarrier{ - .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER, - .pNext = nullptr, - .srcAccessMask = 0, - .dstAccessMask = 0, - .oldLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, - .newLayout = VK_IMAGE_LAYOUT_GENERAL, - .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED, - .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED, - .image = src_image, - .subresourceRange = range, - }, - }; - cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, - 0, {}, {}, pre_barriers); - cmdbuf.CopyImageToBuffer(src_image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, dst_buffer, - buffer_copy_info); - cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, - 0, WRITE_BARRIER, nullptr, post_barriers); - }); - if (is_rescaled) { - image->ScaleUp(true); + const u32 buffer_size = static_cast<u32>(buffer_operand.pitch * buffer_operand.height); + static constexpr auto sync_info = VideoCommon::ObtainBufferSynchronize::FullSynchronize; + const auto post_op = IS_IMAGE_UPLOAD ? VideoCommon::ObtainBufferOperation::DoNothing + : VideoCommon::ObtainBufferOperation::MarkAsWritten; + const auto [buffer, offset] = + buffer_cache.ObtainBuffer(buffer_operand.address, buffer_size, sync_info, post_op); + + const auto [image, copy] = texture_cache.DmaBufferImageCopy( + copy_info, buffer_operand, image_operand, image_id, IS_IMAGE_UPLOAD); + const std::span copy_span{©, 1}; + + if constexpr (IS_IMAGE_UPLOAD) { + image->UploadMemory(buffer->Handle(), offset, copy_span); + } else { + image->DownloadMemory(buffer->Handle(), offset, copy_span); } return true; } +bool AccelerateDMA::ImageToBuffer(const Tegra::DMA::ImageCopy& copy_info, + const Tegra::DMA::ImageOperand& image_operand, + const Tegra::DMA::BufferOperand& buffer_operand) { + return DmaBufferImageCopy<false>(copy_info, buffer_operand, image_operand); +} + bool AccelerateDMA::BufferToImage(const Tegra::DMA::ImageCopy& copy_info, - const Tegra::DMA::BufferOperand& src, - const Tegra::DMA::ImageOperand& dst) { - std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex}; - auto query_image = texture_cache.ObtainImage(dst, true); - if (!query_image) { - return false; - } - auto* image = query_image->first; - auto [level, base] = query_image->second; - const u32 buffer_size = static_cast<u32>(src.pitch * src.height); - const auto [buffer, offset] = buffer_cache.ObtainBuffer( - src.address, buffer_size, VideoCommon::ObtainBufferSynchronize::FullSynchronize, - VideoCommon::ObtainBufferOperation::DoNothing); - const bool is_rescaled = image->IsRescaled(); - if (is_rescaled) { - image->ScaleDown(true); - } - VkImageSubresourceLayers subresources{ - .aspectMask = image->AspectMask(), - .mipLevel = level, - .baseArrayLayer = base, - .layerCount = 1, - }; - const u32 bpp = VideoCore::Surface::BytesPerBlock(image->info.format); - const auto convert = [old_bpp = dst.bytes_per_pixel, bpp](u32 value) { - return (old_bpp * value) / bpp; - }; - const u32 base_x = convert(dst.params.origin.x.Value()); - const u32 base_y = dst.params.origin.y.Value(); - const u32 length_x = convert(copy_info.length_x); - const u32 length_y = copy_info.length_y; - VkOffset3D image_offset{ - .x = static_cast<s32>(base_x), - .y = static_cast<s32>(base_y), - .z = 0, - }; - VkExtent3D image_extent{ - .width = length_x, - .height = length_y, - .depth = 1, - }; - auto buff_info(src); - buff_info.pitch = convert(src.pitch); - scheduler.RequestOutsideRenderPassOperationContext(); - scheduler.Record([dst_image = image->Handle(), src_buffer = buffer->Handle(), - buffer_offset = offset, subresources, image_offset, image_extent, - buff_info](vk::CommandBuffer cmdbuf) { - const std::array buffer_copy_info{ - VkBufferImageCopy{ - .bufferOffset = buffer_offset, - .bufferRowLength = buff_info.pitch, - .bufferImageHeight = buff_info.height, - .imageSubresource = subresources, - .imageOffset = image_offset, - .imageExtent = image_extent, - }, - }; - const VkImageSubresourceRange range{ - .aspectMask = subresources.aspectMask, - .baseMipLevel = subresources.mipLevel, - .levelCount = 1, - .baseArrayLayer = subresources.baseArrayLayer, - .layerCount = 1, - }; - static constexpr VkMemoryBarrier READ_BARRIER{ - .sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER, - .pNext = nullptr, - .srcAccessMask = VK_ACCESS_MEMORY_WRITE_BIT, - .dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT | VK_ACCESS_TRANSFER_WRITE_BIT, - }; - const std::array pre_barriers{ - VkImageMemoryBarrier{ - .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER, - .pNext = nullptr, - .srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT | - VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT | - VK_ACCESS_TRANSFER_WRITE_BIT, - .dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT, - .oldLayout = VK_IMAGE_LAYOUT_GENERAL, - .newLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, - .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED, - .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED, - .image = dst_image, - .subresourceRange = range, - }, - }; - const std::array post_barriers{ - VkImageMemoryBarrier{ - .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER, - .pNext = nullptr, - .srcAccessMask = 0, - .dstAccessMask = 0, - .oldLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, - .newLayout = VK_IMAGE_LAYOUT_GENERAL, - .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED, - .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED, - .image = dst_image, - .subresourceRange = range, - }, - }; - cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, - 0, READ_BARRIER, {}, pre_barriers); - cmdbuf.CopyBufferToImage(src_buffer, dst_image, VK_IMAGE_LAYOUT_GENERAL, buffer_copy_info); - cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, - 0, nullptr, nullptr, post_barriers); - }); - if (is_rescaled) { - image->ScaleUp(); - } - return true; + const Tegra::DMA::BufferOperand& buffer_operand, + const Tegra::DMA::ImageOperand& image_operand) { + return DmaBufferImageCopy<true>(copy_info, buffer_operand, image_operand); } void RasterizerVulkan::UpdateDynamicStates() { diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.h b/src/video_core/renderer_vulkan/vk_rasterizer.h index 7746c5434..1659fbc13 100644 --- a/src/video_core/renderer_vulkan/vk_rasterizer.h +++ b/src/video_core/renderer_vulkan/vk_rasterizer.h @@ -59,6 +59,11 @@ public: const Tegra::DMA::ImageOperand& dst) override; private: + template <bool IS_IMAGE_UPLOAD> + bool DmaBufferImageCopy(const Tegra::DMA::ImageCopy& copy_info, + const Tegra::DMA::BufferOperand& src, + const Tegra::DMA::ImageOperand& dst); + BufferCache& buffer_cache; TextureCache& texture_cache; Scheduler& scheduler; diff --git a/src/video_core/renderer_vulkan/vk_resource_pool.cpp b/src/video_core/renderer_vulkan/vk_resource_pool.cpp index 6c8ac22f4..6572f82ba 100644 --- a/src/video_core/renderer_vulkan/vk_resource_pool.cpp +++ b/src/video_core/renderer_vulkan/vk_resource_pool.cpp @@ -37,7 +37,7 @@ size_t ResourcePool::CommitResource() { found = free_resource; } } - // Free iterator is hinted to the resource after the one that's been commited. + // Free iterator is hinted to the resource after the one that's been committed. hint_iterator = (*found + 1) % ticks.size(); return *found; } @@ -46,7 +46,7 @@ size_t ResourcePool::ManageOverflow() { const size_t old_capacity = ticks.size(); Grow(); - // The last entry is guaranted to be free, since it's the first element of the freshly + // The last entry is guaranteed to be free, since it's the first element of the freshly // allocated resources. return old_capacity; } diff --git a/src/video_core/renderer_vulkan/vk_swapchain.cpp b/src/video_core/renderer_vulkan/vk_swapchain.cpp index b6810eef9..85fdce6e5 100644 --- a/src/video_core/renderer_vulkan/vk_swapchain.cpp +++ b/src/video_core/renderer_vulkan/vk_swapchain.cpp @@ -159,7 +159,7 @@ void Swapchain::CreateSwapchain(const VkSurfaceCapabilitiesKHR& capabilities, bo present_mode = ChooseSwapPresentMode(present_modes); u32 requested_image_count{capabilities.minImageCount + 1}; - // Ensure Tripple buffering if possible. + // Ensure Triple buffering if possible. if (capabilities.maxImageCount > 0) { if (requested_image_count > capabilities.maxImageCount) { requested_image_count = capabilities.maxImageCount; diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.cpp b/src/video_core/renderer_vulkan/vk_texture_cache.cpp index e013d1c60..ae15f6976 100644 --- a/src/video_core/renderer_vulkan/vk_texture_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_texture_cache.cpp @@ -1315,15 +1315,16 @@ Image::Image(const VideoCommon::NullImageParams& params) : VideoCommon::ImageBas Image::~Image() = default; -void Image::UploadMemory(const StagingBufferRef& map, std::span<const BufferImageCopy> copies) { +void Image::UploadMemory(VkBuffer buffer, VkDeviceSize offset, + std::span<const VideoCommon::BufferImageCopy> copies) { // TODO: Move this to another API const bool is_rescaled = True(flags & ImageFlagBits::Rescaled); if (is_rescaled) { ScaleDown(true); } scheduler->RequestOutsideRenderPassOperationContext(); - std::vector vk_copies = TransformBufferImageCopies(copies, map.offset, aspect_mask); - const VkBuffer src_buffer = map.buffer; + std::vector vk_copies = TransformBufferImageCopies(copies, offset, aspect_mask); + const VkBuffer src_buffer = buffer; const VkImage vk_image = *original_image; const VkImageAspectFlags vk_aspect_mask = aspect_mask; const bool is_initialized = std::exchange(initialized, true); @@ -1336,14 +1337,19 @@ void Image::UploadMemory(const StagingBufferRef& map, std::span<const BufferImag } } -void Image::DownloadMemory(const StagingBufferRef& map, std::span<const BufferImageCopy> copies) { +void Image::UploadMemory(const StagingBufferRef& map, std::span<const BufferImageCopy> copies) { + UploadMemory(map.buffer, map.offset, copies); +} + +void Image::DownloadMemory(VkBuffer buffer, VkDeviceSize offset, + std::span<const VideoCommon::BufferImageCopy> copies) { const bool is_rescaled = True(flags & ImageFlagBits::Rescaled); if (is_rescaled) { ScaleDown(); } - std::vector vk_copies = TransformBufferImageCopies(copies, map.offset, aspect_mask); + std::vector vk_copies = TransformBufferImageCopies(copies, offset, aspect_mask); scheduler->RequestOutsideRenderPassOperationContext(); - scheduler->Record([buffer = map.buffer, image = *original_image, aspect_mask = aspect_mask, + scheduler->Record([buffer, image = *original_image, aspect_mask = aspect_mask, vk_copies](vk::CommandBuffer cmdbuf) { const VkImageMemoryBarrier read_barrier{ .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER, @@ -1398,6 +1404,10 @@ void Image::DownloadMemory(const StagingBufferRef& map, std::span<const BufferIm } } +void Image::DownloadMemory(const StagingBufferRef& map, std::span<const BufferImageCopy> copies) { + DownloadMemory(map.buffer, map.offset, copies); +} + bool Image::IsRescaled() const noexcept { return True(flags & ImageFlagBits::Rescaled); } diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.h b/src/video_core/renderer_vulkan/vk_texture_cache.h index 0ce39616f..d5ee23f8d 100644 --- a/src/video_core/renderer_vulkan/vk_texture_cache.h +++ b/src/video_core/renderer_vulkan/vk_texture_cache.h @@ -132,9 +132,15 @@ public: Image(Image&&) = default; Image& operator=(Image&&) = default; + void UploadMemory(VkBuffer buffer, VkDeviceSize offset, + std::span<const VideoCommon::BufferImageCopy> copies); + void UploadMemory(const StagingBufferRef& map, std::span<const VideoCommon::BufferImageCopy> copies); + void DownloadMemory(VkBuffer buffer, VkDeviceSize offset, + std::span<const VideoCommon::BufferImageCopy> copies); + void DownloadMemory(const StagingBufferRef& map, std::span<const VideoCommon::BufferImageCopy> copies); diff --git a/src/video_core/renderer_vulkan/vk_update_descriptor.cpp b/src/video_core/renderer_vulkan/vk_update_descriptor.cpp index 4d4a6753b..009dab0b6 100644 --- a/src/video_core/renderer_vulkan/vk_update_descriptor.cpp +++ b/src/video_core/renderer_vulkan/vk_update_descriptor.cpp @@ -25,7 +25,7 @@ void UpdateDescriptorQueue::TickFrame() { void UpdateDescriptorQueue::Acquire() { // Minimum number of entries required. - // This is the maximum number of entries a single draw call migth use. + // This is the maximum number of entries a single draw call might use. static constexpr size_t MIN_ENTRIES = 0x400; if (std::distance(payload.data(), payload_cursor) + MIN_ENTRIES >= payload.max_size()) { diff --git a/src/video_core/texture_cache/format_lookup_table.cpp b/src/video_core/texture_cache/format_lookup_table.cpp index 08aa8ca33..5fc2b2fec 100644 --- a/src/video_core/texture_cache/format_lookup_table.cpp +++ b/src/video_core/texture_cache/format_lookup_table.cpp @@ -42,15 +42,15 @@ PixelFormat PixelFormatFromTextureInfo(TextureFormat format, ComponentType red, ComponentType blue, ComponentType alpha, bool is_srgb) noexcept { switch (Hash(format, red, green, blue, alpha, is_srgb)) { - case Hash(TextureFormat::A8R8G8B8, UNORM): + case Hash(TextureFormat::A8B8G8R8, UNORM): return PixelFormat::A8B8G8R8_UNORM; - case Hash(TextureFormat::A8R8G8B8, SNORM): + case Hash(TextureFormat::A8B8G8R8, SNORM): return PixelFormat::A8B8G8R8_SNORM; - case Hash(TextureFormat::A8R8G8B8, UINT): + case Hash(TextureFormat::A8B8G8R8, UINT): return PixelFormat::A8B8G8R8_UINT; - case Hash(TextureFormat::A8R8G8B8, SINT): + case Hash(TextureFormat::A8B8G8R8, SINT): return PixelFormat::A8B8G8R8_SINT; - case Hash(TextureFormat::A8R8G8B8, UNORM, SRGB): + case Hash(TextureFormat::A8B8G8R8, UNORM, SRGB): return PixelFormat::A8B8G8R8_SRGB; case Hash(TextureFormat::B5G6R5, UNORM): return PixelFormat::B5G6R5_UNORM; @@ -74,13 +74,13 @@ PixelFormat PixelFormatFromTextureInfo(TextureFormat format, ComponentType red, return PixelFormat::R8_UINT; case Hash(TextureFormat::R8, SINT): return PixelFormat::R8_SINT; - case Hash(TextureFormat::R8G8, UNORM): + case Hash(TextureFormat::G8R8, UNORM): return PixelFormat::R8G8_UNORM; - case Hash(TextureFormat::R8G8, SNORM): + case Hash(TextureFormat::G8R8, SNORM): return PixelFormat::R8G8_SNORM; - case Hash(TextureFormat::R8G8, UINT): + case Hash(TextureFormat::G8R8, UINT): return PixelFormat::R8G8_UINT; - case Hash(TextureFormat::R8G8, SINT): + case Hash(TextureFormat::G8R8, SINT): return PixelFormat::R8G8_SINT; case Hash(TextureFormat::R16G16B16A16, FLOAT): return PixelFormat::R16G16B16A16_FLOAT; @@ -136,49 +136,49 @@ PixelFormat PixelFormatFromTextureInfo(TextureFormat format, ComponentType red, return PixelFormat::R32_SINT; case Hash(TextureFormat::E5B9G9R9, FLOAT): return PixelFormat::E5B9G9R9_FLOAT; - case Hash(TextureFormat::D32, FLOAT): + case Hash(TextureFormat::Z32, FLOAT): return PixelFormat::D32_FLOAT; - case Hash(TextureFormat::D16, UNORM): + case Hash(TextureFormat::Z16, UNORM): return PixelFormat::D16_UNORM; - case Hash(TextureFormat::S8D24, UINT, UNORM, UNORM, UNORM, LINEAR): + case Hash(TextureFormat::Z24S8, UINT, UNORM, UNORM, UNORM, LINEAR): return PixelFormat::S8_UINT_D24_UNORM; - case Hash(TextureFormat::S8D24, UINT, UNORM, UINT, UINT, LINEAR): + case Hash(TextureFormat::Z24S8, UINT, UNORM, UINT, UINT, LINEAR): return PixelFormat::S8_UINT_D24_UNORM; - case Hash(TextureFormat::R8G24, UINT, UNORM, UNORM, UNORM, LINEAR): + case Hash(TextureFormat::G24R8, UINT, UNORM, UNORM, UNORM, LINEAR): return PixelFormat::S8_UINT_D24_UNORM; - case Hash(TextureFormat::D24S8, UNORM, UINT, UINT, UINT, LINEAR): + case Hash(TextureFormat::S8Z24, UNORM, UINT, UINT, UINT, LINEAR): return PixelFormat::D24_UNORM_S8_UINT; - case Hash(TextureFormat::D32S8, FLOAT, UINT, UNORM, UNORM, LINEAR): + case Hash(TextureFormat::Z32_X24S8, FLOAT, UINT, UNORM, UNORM, LINEAR): return PixelFormat::D32_FLOAT_S8_UINT; - case Hash(TextureFormat::R32_B24G8, FLOAT, UINT, UNORM, UNORM, LINEAR): + case Hash(TextureFormat::R32B24G8, FLOAT, UINT, UNORM, UNORM, LINEAR): return PixelFormat::D32_FLOAT_S8_UINT; - case Hash(TextureFormat::BC1_RGBA, UNORM, LINEAR): + case Hash(TextureFormat::DXT1, UNORM, LINEAR): return PixelFormat::BC1_RGBA_UNORM; - case Hash(TextureFormat::BC1_RGBA, UNORM, SRGB): + case Hash(TextureFormat::DXT1, UNORM, SRGB): return PixelFormat::BC1_RGBA_SRGB; - case Hash(TextureFormat::BC2, UNORM, LINEAR): + case Hash(TextureFormat::DXT23, UNORM, LINEAR): return PixelFormat::BC2_UNORM; - case Hash(TextureFormat::BC2, UNORM, SRGB): + case Hash(TextureFormat::DXT23, UNORM, SRGB): return PixelFormat::BC2_SRGB; - case Hash(TextureFormat::BC3, UNORM, LINEAR): + case Hash(TextureFormat::DXT45, UNORM, LINEAR): return PixelFormat::BC3_UNORM; - case Hash(TextureFormat::BC3, UNORM, SRGB): + case Hash(TextureFormat::DXT45, UNORM, SRGB): return PixelFormat::BC3_SRGB; - case Hash(TextureFormat::BC4, UNORM): + case Hash(TextureFormat::DXN1, UNORM): return PixelFormat::BC4_UNORM; - case Hash(TextureFormat::BC4, SNORM): + case Hash(TextureFormat::DXN1, SNORM): return PixelFormat::BC4_SNORM; - case Hash(TextureFormat::BC5, UNORM): + case Hash(TextureFormat::DXN2, UNORM): return PixelFormat::BC5_UNORM; - case Hash(TextureFormat::BC5, SNORM): + case Hash(TextureFormat::DXN2, SNORM): return PixelFormat::BC5_SNORM; - case Hash(TextureFormat::BC7, UNORM, LINEAR): + case Hash(TextureFormat::BC7U, UNORM, LINEAR): return PixelFormat::BC7_UNORM; - case Hash(TextureFormat::BC7, UNORM, SRGB): + case Hash(TextureFormat::BC7U, UNORM, SRGB): return PixelFormat::BC7_SRGB; - case Hash(TextureFormat::BC6H_SFLOAT, FLOAT): + case Hash(TextureFormat::BC6H_S16, FLOAT): return PixelFormat::BC6H_SFLOAT; - case Hash(TextureFormat::BC6H_UFLOAT, FLOAT): + case Hash(TextureFormat::BC6H_U16, FLOAT): return PixelFormat::BC6H_UFLOAT; case Hash(TextureFormat::ASTC_2D_4X4, UNORM, LINEAR): return PixelFormat::ASTC_2D_4X4_UNORM; diff --git a/src/video_core/texture_cache/image_base.h b/src/video_core/texture_cache/image_base.h index e8fa592d2..329396bb6 100644 --- a/src/video_core/texture_cache/image_base.h +++ b/src/video_core/texture_cache/image_base.h @@ -25,7 +25,7 @@ enum class ImageFlagBits : u32 { Registered = 1 << 6, ///< True when the image is registered Picked = 1 << 7, ///< Temporary flag to mark the image as picked Remapped = 1 << 8, ///< Image has been remapped. - Sparse = 1 << 9, ///< Image has non continous submemory. + Sparse = 1 << 9, ///< Image has non continuous submemory. // Garbage Collection Flags BadOverlap = 1 << 10, ///< This image overlaps other but doesn't fit, has higher diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h index 335338434..8e8b9a5e6 100644 --- a/src/video_core/texture_cache/texture_cache.h +++ b/src/video_core/texture_cache/texture_cache.h @@ -745,6 +745,25 @@ void TextureCache<P>::PopAsyncFlushes() { } template <class P> +ImageId TextureCache<P>::DmaImageId(const Tegra::DMA::ImageOperand& operand) { + const ImageInfo dst_info(operand); + const ImageId dst_id = FindDMAImage(dst_info, operand.address); + if (!dst_id) { + return NULL_IMAGE_ID; + } + const auto& image = slot_images[dst_id]; + if (False(image.flags & ImageFlagBits::GpuModified)) { + // No need to waste time on an image that's synced with guest + return NULL_IMAGE_ID; + } + const auto base = image.TryFindBase(operand.address); + if (!base) { + return NULL_IMAGE_ID; + } + return dst_id; +} + +template <class P> bool TextureCache<P>::IsRescaling() const noexcept { return is_rescaling; } @@ -772,6 +791,49 @@ bool TextureCache<P>::IsRegionGpuModified(VAddr addr, size_t size) { } template <class P> +std::pair<typename TextureCache<P>::Image*, BufferImageCopy> TextureCache<P>::DmaBufferImageCopy( + const Tegra::DMA::ImageCopy& copy_info, const Tegra::DMA::BufferOperand& buffer_operand, + const Tegra::DMA::ImageOperand& image_operand, ImageId image_id, bool modifies_image) { + const auto [level, base] = PrepareDmaImage(image_id, image_operand.address, modifies_image); + auto* image = &slot_images[image_id]; + const u32 buffer_size = static_cast<u32>(buffer_operand.pitch * buffer_operand.height); + const u32 bpp = VideoCore::Surface::BytesPerBlock(image->info.format); + const auto convert = [old_bpp = image_operand.bytes_per_pixel, bpp](u32 value) { + return (old_bpp * value) / bpp; + }; + const u32 base_x = convert(image_operand.params.origin.x.Value()); + const u32 base_y = image_operand.params.origin.y.Value(); + const u32 length_x = convert(copy_info.length_x); + const u32 length_y = copy_info.length_y; + + const BufferImageCopy copy{ + .buffer_offset = 0, + .buffer_size = buffer_size, + .buffer_row_length = convert(buffer_operand.pitch), + .buffer_image_height = buffer_operand.height, + .image_subresource = + { + .base_level = static_cast<s32>(level), + .base_layer = static_cast<s32>(base), + .num_layers = 1, + }, + .image_offset = + { + .x = static_cast<s32>(base_x), + .y = static_cast<s32>(base_y), + .z = 0, + }, + .image_extent = + { + .width = length_x, + .height = length_y, + .depth = 1, + }, + }; + return {image, copy}; +} + +template <class P> void TextureCache<P>::RefreshContents(Image& image, ImageId image_id) { if (False(image.flags & ImageFlagBits::CpuModified)) { // Only upload modified images @@ -1405,26 +1467,14 @@ ImageId TextureCache<P>::FindDMAImage(const ImageInfo& info, GPUVAddr gpu_addr) } template <class P> -std::optional<std::pair<typename TextureCache<P>::Image*, std::pair<u32, u32>>> -TextureCache<P>::ObtainImage(const Tegra::DMA::ImageOperand& operand, bool mark_as_modified) { - ImageInfo dst_info(operand); - ImageId dst_id = FindDMAImage(dst_info, operand.address); - if (!dst_id) { - return std::nullopt; - } - auto& image = slot_images[dst_id]; - auto base = image.TryFindBase(operand.address); - if (!base) { - return std::nullopt; - } - if (False(image.flags & ImageFlagBits::GpuModified)) { - // No need to waste time on an image that's synced with guest - return std::nullopt; - } +std::pair<u32, u32> TextureCache<P>::PrepareDmaImage(ImageId dst_id, GPUVAddr base_addr, + bool mark_as_modified) { + const auto& image = slot_images[dst_id]; + const auto base = image.TryFindBase(base_addr); PrepareImage(dst_id, mark_as_modified, false); - auto& new_image = slot_images[dst_id]; + const auto& new_image = slot_images[dst_id]; lru_cache.Touch(new_image.lru_index, frame_tick); - return std::make_pair(&new_image, std::make_pair(base->level, base->layer)); + return std::make_pair(base->level, base->layer); } 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 848a5d9ea..5a5b4179c 100644 --- a/src/video_core/texture_cache/texture_cache_base.h +++ b/src/video_core/texture_cache/texture_cache_base.h @@ -209,8 +209,11 @@ public: /// Pop asynchronous downloads void PopAsyncFlushes(); - [[nodiscard]] std::optional<std::pair<Image*, std::pair<u32, u32>>> ObtainImage( - const Tegra::DMA::ImageOperand& operand, bool mark_as_modified); + [[nodiscard]] ImageId DmaImageId(const Tegra::DMA::ImageOperand& operand); + + [[nodiscard]] std::pair<Image*, BufferImageCopy> DmaBufferImageCopy( + const Tegra::DMA::ImageCopy& copy_info, const Tegra::DMA::BufferOperand& buffer_operand, + const Tegra::DMA::ImageOperand& image_operand, ImageId image_id, bool modifies_image); /// Return true when a CPU region is modified from the GPU [[nodiscard]] bool IsRegionGpuModified(VAddr addr, size_t size); @@ -386,6 +389,9 @@ private: /// Returns true if the current clear parameters clear the whole image of a given image view [[nodiscard]] bool IsFullClear(ImageViewId id); + [[nodiscard]] std::pair<u32, u32> PrepareDmaImage(ImageId dst_id, GPUVAddr base_addr, + bool mark_as_modified); + bool ImageCanRescale(ImageBase& image); void InvalidateScale(Image& image); bool ScaleUp(Image& image); diff --git a/src/video_core/textures/astc.cpp b/src/video_core/textures/astc.cpp index 4381eed1d..a68bc0d77 100644 --- a/src/video_core/textures/astc.cpp +++ b/src/video_core/textures/astc.cpp @@ -1571,7 +1571,7 @@ static void DecompressBlock(std::span<const u8, 16> inBuf, const u32 blockWidth, assert(strm.GetBitsRead() + weightParams.GetPackedBitSize() == 128); // Decode both color data and texel weight data - u32 colorValues[32]; // Four values, two endpoints, four maximum paritions + u32 colorValues[32]; // Four values, two endpoints, four maximum partitions DecodeColorValues(colorValues, colorEndpointData, colorEndpointMode, nPartitions, colorDataBits); diff --git a/src/video_core/textures/texture.h b/src/video_core/textures/texture.h index 7c4553a53..7e5837b20 100644 --- a/src/video_core/textures/texture.h +++ b/src/video_core/textures/texture.h @@ -15,26 +15,26 @@ enum class TextureFormat : u32 { R32G32B32 = 0x02, R16G16B16A16 = 0x03, R32G32 = 0x04, - R32_B24G8 = 0x05, + R32B24G8 = 0x05, ETC2_RGB = 0x06, X8B8G8R8 = 0x07, - A8R8G8B8 = 0x08, + A8B8G8R8 = 0x08, A2B10G10R10 = 0x09, ETC2_RGB_PTA = 0x0a, ETC2_RGBA = 0x0b, R16G16 = 0x0c, - R24G8 = 0x0d, - R8G24 = 0x0e, + G8R24 = 0x0d, + G24R8 = 0x0e, R32 = 0x0f, - BC6H_SFLOAT = 0x10, - BC6H_UFLOAT = 0x11, + BC6H_S16 = 0x10, + BC6H_U16 = 0x11, A4B4G4R4 = 0x12, A5B5G5R1 = 0x13, A1B5G5R5 = 0x14, B5G6R5 = 0x15, B6G5R5 = 0x16, - BC7 = 0x17, - R8G8 = 0x18, + BC7U = 0x17, + G8R8 = 0x18, EAC = 0x19, EACX2 = 0x1a, R16 = 0x1b, @@ -46,33 +46,33 @@ enum class TextureFormat : u32 { B10G11R11 = 0x21, G8B8G8R8 = 0x22, B8G8R8G8 = 0x23, - BC1_RGBA = 0x24, - BC2 = 0x25, - BC3 = 0x26, - BC4 = 0x27, - BC5 = 0x28, - S8D24 = 0x29, - X8D24 = 0x2a, - D24S8 = 0x2b, - X4V4D24__COV4R4V = 0x2c, - X4V4D24__COV8R8V = 0x2d, - V8D24__COV4R12V = 0x2e, - D32 = 0x2f, - D32S8 = 0x30, - X8D24_X20V4S8__COV4R4V = 0x31, - X8D24_X20V4S8__COV8R8V = 0x32, - D32_X20V4X8__COV4R4V = 0x33, - D32_X20V4X8__COV8R8V = 0x34, - D32_X20V4S8__COV4R4V = 0x35, - D32_X20V4S8__COV8R8V = 0x36, - X8D24_X16V8S8__COV4R12V = 0x37, - D32_X16V8X8__COV4R12V = 0x38, - D32_X16V8S8__COV4R12V = 0x39, - D16 = 0x3a, - V8D24__COV8R24V = 0x3b, - X8D24_X16V8S8__COV8R24V = 0x3c, - D32_X16V8X8__COV8R24V = 0x3d, - D32_X16V8S8__COV8R24V = 0x3e, + DXT1 = 0x24, + DXT23 = 0x25, + DXT45 = 0x26, + DXN1 = 0x27, + DXN2 = 0x28, + Z24S8 = 0x29, + X8Z24 = 0x2a, + S8Z24 = 0x2b, + X4V4Z24__COV4R4V = 0x2c, + X4V4Z24__COV8R8V = 0x2d, + V8Z24__COV4R12V = 0x2e, + Z32 = 0x2f, + Z32_X24S8 = 0x30, + X8Z24_X20V4S8__COV4R4V = 0x31, + X8Z24_X20V4S8__COV8R8V = 0x32, + Z32_X20V4X8__COV4R4V = 0x33, + Z32_X20V4X8__COV8R8V = 0x34, + Z32_X20V4S8__COV4R4V = 0x35, + Z32_X20V4S8__COV8R8V = 0x36, + X8Z24_X16V8S8__COV4R12V = 0x37, + Z32_X16V8X8__COV4R12V = 0x38, + Z32_X16V8S8__COV4R12V = 0x39, + Z16 = 0x3a, + V8Z24__COV8R24V = 0x3b, + X8Z24_X16V8S8__COV8R24V = 0x3c, + Z32_X16V8X8__COV8R24V = 0x3d, + Z32_X16V8S8__COV8R24V = 0x3e, ASTC_2D_4X4 = 0x40, ASTC_2D_5X5 = 0x41, ASTC_2D_6X6 = 0x42, diff --git a/src/video_core/vulkan_common/vulkan_device.cpp b/src/video_core/vulkan_common/vulkan_device.cpp index 48f1a3d14..df348af55 100644 --- a/src/video_core/vulkan_common/vulkan_device.cpp +++ b/src/video_core/vulkan_common/vulkan_device.cpp @@ -417,7 +417,7 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR sets_per_pool = 64; if (is_amd_driver) { - // AMD drivers need a higher amount of Sets per Pool in certain circunstances like in XC2. + // AMD drivers need a higher amount of Sets per Pool in certain circumstances like in XC2. sets_per_pool = 96; // Disable VK_IMAGE_CREATE_CUBE_COMPATIBLE_BIT on AMD GCN4 and lower as it is broken. if (!features.shader_float16_int8.shaderFloat16) { diff --git a/src/video_core/vulkan_common/vulkan_device.h b/src/video_core/vulkan_common/vulkan_device.h index 0662a2d9f..41b5da18a 100644 --- a/src/video_core/vulkan_common/vulkan_device.h +++ b/src/video_core/vulkan_common/vulkan_device.h @@ -180,7 +180,7 @@ public: ~Device(); /** - * Returns a format supported by the device for the passed requeriments. + * Returns a format supported by the device for the passed requirements. * @param wanted_format The ideal format to be returned. It may not be the returned format. * @param wanted_usage The usage that must be fulfilled even if the format is not supported. * @param format_type Format type usage. @@ -259,12 +259,12 @@ public: bool ShouldBoostClocks() const; - /// Returns uniform buffer alignment requeriment. + /// Returns uniform buffer alignment requirement. VkDeviceSize GetUniformBufferAlignment() const { return properties.properties.limits.minUniformBufferOffsetAlignment; } - /// Returns storage alignment requeriment. + /// Returns storage alignment requirement. VkDeviceSize GetStorageBufferAlignment() const { return properties.properties.limits.minStorageBufferOffsetAlignment; } @@ -656,7 +656,7 @@ private: bool is_integrated{}; ///< Is GPU an iGPU. bool is_virtual{}; ///< Is GPU a virtual GPU. bool is_non_gpu{}; ///< Is SoftwareRasterizer, FPGA, non-GPU device. - bool has_broken_cube_compatibility{}; ///< Has broken cube compatiblity bit + bool has_broken_cube_compatibility{}; ///< Has broken cube compatibility bit bool has_renderdoc{}; ///< Has RenderDoc attached bool has_nsight_graphics{}; ///< Has Nsight Graphics attached bool supports_d24_depth{}; ///< Supports D24 depth buffers. diff --git a/src/video_core/vulkan_common/vulkan_wrapper.h b/src/video_core/vulkan_common/vulkan_wrapper.h index e86f661cb..4ff328a21 100644 --- a/src/video_core/vulkan_common/vulkan_wrapper.h +++ b/src/video_core/vulkan_common/vulkan_wrapper.h @@ -68,7 +68,7 @@ public: constexpr Span(const Range& range) : ptr{std::data(range)}, num{std::size(range)} {} /// Construct a span from a pointer and a size. - /// This is inteded for subranges. + /// This is intended for subranges. constexpr Span(const T* ptr_, std::size_t num_) noexcept : ptr{ptr_}, num{num_} {} /// Returns the data pointer by the span. @@ -390,11 +390,11 @@ public: Handle(const Handle&) = delete; Handle& operator=(const Handle&) = delete; - /// Construct a handle transfering the ownership from another handle. + /// Construct a handle transferring the ownership from another handle. Handle(Handle&& rhs) noexcept : handle{std::exchange(rhs.handle, nullptr)}, owner{rhs.owner}, dld{rhs.dld} {} - /// Assign the current handle transfering the ownership from another handle. + /// Assign the current handle transferring the ownership from another handle. /// Destroys any previously held object. Handle& operator=(Handle&& rhs) noexcept { Release(); @@ -463,10 +463,10 @@ public: Handle(const Handle&) = delete; Handle& operator=(const Handle&) = delete; - /// Construct a handle transfering ownership from another handle. + /// Construct a handle transferring ownership from another handle. Handle(Handle&& rhs) noexcept : handle{std::exchange(rhs.handle, nullptr)}, dld{rhs.dld} {} - /// Assign the current handle transfering the ownership from another handle. + /// Assign the current handle transferring the ownership from another handle. /// Destroys any previously held object. Handle& operator=(Handle&& rhs) noexcept { Release(); @@ -533,12 +533,12 @@ public: PoolAllocations(const PoolAllocations&) = delete; PoolAllocations& operator=(const PoolAllocations&) = delete; - /// Construct an allocation transfering ownership from another allocation. + /// Construct an allocation transferring ownership from another allocation. PoolAllocations(PoolAllocations&& rhs) noexcept : allocations{std::move(rhs.allocations)}, num{rhs.num}, device{rhs.device}, pool{rhs.pool}, dld{rhs.dld} {} - /// Assign an allocation transfering ownership from another allocation. + /// Assign an allocation transferring ownership from another allocation. PoolAllocations& operator=(PoolAllocations&& rhs) noexcept { allocations = std::move(rhs.allocations); num = rhs.num; |