flodavid/suyu
1
0
Fork 0
forked from suyu/suyu
Code Pull requests Activity

Merge pull request #11278 from Kelebek1/dma_sync

Browse source 
Mark accelerated DMA destination buffers and images as GPU-modified
This commit is contained in:
liamwhite 2023-08-18 09:12:27 -04:00 committed by GitHub
commit ae1421265a
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
3 changed files with 15 additions and 5 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

7
src/video_core/renderer_opengl/gl_rasterizer.cpp
View file

@ -1335,7 +1335,8 @@ bool AccelerateDMA::DmaBufferImageCopy(const Tegra::DMA::ImageCopy& copy_info,
} }
const u32 buffer_size = static_cast<u32>(buffer_operand.pitch * buffer_operand.height); const u32 buffer_size = static_cast<u32>(buffer_operand.pitch * buffer_operand.height);
static constexpr auto sync_info = VideoCommon::ObtainBufferSynchronize::FullSynchronize; static constexpr auto sync_info = VideoCommon::ObtainBufferSynchronize::FullSynchronize;
const auto post_op = VideoCommon::ObtainBufferOperation::DoNothing; const auto post_op = IS_IMAGE_UPLOAD ? VideoCommon::ObtainBufferOperation::DoNothing
: VideoCommon::ObtainBufferOperation::MarkAsWritten;
const auto [buffer, offset] = const auto [buffer, offset] =
buffer_cache.ObtainBuffer(buffer_operand.address, buffer_size, sync_info, post_op); buffer_cache.ObtainBuffer(buffer_operand.address, buffer_size, sync_info, post_op);
@ -1344,8 +1345,12 @@ bool AccelerateDMA::DmaBufferImageCopy(const Tegra::DMA::ImageCopy& copy_info,
const std::span copy_span{&copy, 1}; const std::span copy_span{&copy, 1};
if constexpr (IS_IMAGE_UPLOAD) { if constexpr (IS_IMAGE_UPLOAD) {
texture_cache.PrepareImage(image_id, true, false);
image->UploadMemory(buffer->Handle(), offset, copy_span); image->UploadMemory(buffer->Handle(), offset, copy_span);
} else { } else {
if (offset % BytesPerBlock(image->info.format)) {
return false;
}
texture_cache.DownloadImageIntoBuffer(image, buffer->Handle(), offset, copy_span, texture_cache.DownloadImageIntoBuffer(image, buffer->Handle(), offset, copy_span,
buffer_operand.address, buffer_size); buffer_operand.address, buffer_size);
} }

7
src/video_core/renderer_vulkan/vk_rasterizer.cpp
View file

@ -830,7 +830,8 @@ bool AccelerateDMA::DmaBufferImageCopy(const Tegra::DMA::ImageCopy& copy_info,
} }
const u32 buffer_size = static_cast<u32>(buffer_operand.pitch * buffer_operand.height); const u32 buffer_size = static_cast<u32>(buffer_operand.pitch * buffer_operand.height);
static constexpr auto sync_info = VideoCommon::ObtainBufferSynchronize::FullSynchronize; static constexpr auto sync_info = VideoCommon::ObtainBufferSynchronize::FullSynchronize;
const auto post_op = VideoCommon::ObtainBufferOperation::DoNothing; const auto post_op = IS_IMAGE_UPLOAD ? VideoCommon::ObtainBufferOperation::DoNothing
: VideoCommon::ObtainBufferOperation::MarkAsWritten;
const auto [buffer, offset] = const auto [buffer, offset] =
buffer_cache.ObtainBuffer(buffer_operand.address, buffer_size, sync_info, post_op); buffer_cache.ObtainBuffer(buffer_operand.address, buffer_size, sync_info, post_op);
@ -839,8 +840,12 @@ bool AccelerateDMA::DmaBufferImageCopy(const Tegra::DMA::ImageCopy& copy_info,
const std::span copy_span{&copy, 1}; const std::span copy_span{&copy, 1};
if constexpr (IS_IMAGE_UPLOAD) { if constexpr (IS_IMAGE_UPLOAD) {
texture_cache.PrepareImage(image_id, true, false);
image->UploadMemory(buffer->Handle(), offset, copy_span); image->UploadMemory(buffer->Handle(), offset, copy_span);
} else { } else {
if (offset % BytesPerBlock(image->info.format)) {
return false;
}
texture_cache.DownloadImageIntoBuffer(image, buffer->Handle(), offset, copy_span, texture_cache.DownloadImageIntoBuffer(image, buffer->Handle(), offset, copy_span,
buffer_operand.address, buffer_size); buffer_operand.address, buffer_size);
} }

6
src/video_core/texture_cache/texture_cache_base.h
View file

@ -243,6 +243,9 @@ public:
/// Create channel state. /// Create channel state.
void CreateChannel(Tegra::Control::ChannelState& channel) final override; void CreateChannel(Tegra::Control::ChannelState& channel) final override;
/// Prepare an image to be used
void PrepareImage(ImageId image_id, bool is_modification, bool invalidate);
std::recursive_mutex mutex; std::recursive_mutex mutex;
private: private:
@ -387,9 +390,6 @@ private:
/// Synchronize image aliases, copying data if needed /// Synchronize image aliases, copying data if needed
void SynchronizeAliases(ImageId image_id); void SynchronizeAliases(ImageId image_id);
/// Prepare an image to be used
void PrepareImage(ImageId image_id, bool is_modification, bool invalidate);
/// Prepare an image view to be used /// Prepare an image view to be used
void PrepareImageView(ImageViewId image_view_id, bool is_modification, bool invalidate); void PrepareImageView(ImageViewId image_view_id, bool is_modification, bool invalidate);