mirrors
/
dxvk
mirror of https://github.com/doitsujin/dxvk
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

[d3d9] Remove direct upload path 

And remove some tracking that will no longer be necessary.
This commit is contained in:
Robin Kertels 2022-06-11 19:38:05 +02:00 committed by Joshie
parent 45c1d7911e
commit ba4d95c5fc
3 changed files with 26 additions and 91 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
src/d3d9/d3d9_common_buffer.h
View File

@ -155,11 +155,6 @@ namespace dxvk {
*/
inline D3D9Range& DirtyRange() { return m_dirtyRange; }
/**
* \brief The range of the buffer that might currently be read by the GPU
*/
inline D3D9Range& GPUReadingRange() { return m_gpuReadingRange; }
/**
* \brief Whether or not the buffer was written to by the GPU (in IDirect3DDevice9::ProcessVertices)
*/
@ -184,15 +179,6 @@ namespace dxvk {
*/
inline bool NeedsUpload() { return m_desc.Pool != D3DPOOL_DEFAULT && !m_dirtyRange.IsDegenerate(); }
inline bool DoesStagingBufferUploads() const { return m_uploadUsingStaging; }
inline void EnableStagingBufferUploads() {
if (GetMapMode() != D3D9_COMMON_BUFFER_MAP_MODE_BUFFER)
return;
m_uploadUsingStaging = true;
}
void PreLoad();
bool HasSequenceNumber() const {
@ -252,7 +238,6 @@ namespace dxvk {
DxvkBufferSliceHandle m_sliceHandle;
D3D9Range m_dirtyRange;
D3D9Range m_gpuReadingRange;
uint32_t m_lockCount = 0;

5
src/d3d9/d3d9_common_texture.h
View File

@ -378,11 +378,6 @@ namespace dxvk {
bool NeedsUpload(UINT Subresource) const { return m_needsUpload.get(Subresource); }
bool NeedsAnyUpload() { return m_needsUpload.any(); }
void ClearNeedsUpload() { return m_needsUpload.clearAll(); }
bool DoesStagingBufferUploads(UINT Subresource) const { return m_uploadUsingStaging.get(Subresource); }
void EnableStagingBufferUploads(UINT Subresource) {
m_uploadUsingStaging.set(Subresource, true);
}
void SetNeedsMipGen(bool value) { m_needsMipGen = value; }
bool NeedsMipGen() const { return m_needsMipGen; }

97
src/d3d9/d3d9_device.cpp
View File

@ -4142,10 +4142,6 @@ namespace dxvk {
VkExtent3D levelExtent = pResource->GetExtentMip(MipLevel);
VkExtent3D blockCount = util::computeBlockCount(levelExtent, formatInfo->blockSize);
const bool systemmem = desc.Pool == D3DPOOL_SYSTEMMEM;
const bool managed = IsPoolManaged(desc.Pool);
const bool scratch = desc.Pool == D3DPOOL_SCRATCH;
bool fullResource = pBox == nullptr;
if (unlikely(!fullResource)) {
VkOffset3D lockOffset;
@ -4187,7 +4183,7 @@ namespace dxvk {
DxvkBufferSliceHandle physSlice;
if (Flags & D3DLOCK_DISCARD) {
if ((Flags & D3DLOCK_DISCARD) && needsReadback) {
// We do not have to preserve the contents of the
// buffer if the entire image gets discarded.
physSlice = pResource->DiscardMapSlice(Subresource);
@ -4201,18 +4197,8 @@ namespace dxvk {
} else {
physSlice = pResource->GetMappedSlice(Subresource);
// We do not need to wait for the resource in the event the
// calling app promises not to overwrite data that is in use
// or is reading. Remember! This will only trigger for MANAGED resources
// that cannot get affected by GPU, therefore readonly is A-OK for NOT waiting.
const bool usesStagingBuffer = pResource->DoesStagingBufferUploads(Subresource);
const bool skipWait = (scratch || managed || systemmem) && !needsReadback
&& (usesStagingBuffer || readOnly);
if (alloced && !needsReadback) {
std::memset(physSlice.mapPtr, 0, physSlice.length);
}
else if (!skipWait) {
if (needsReadback) {
const Rc<DxvkBuffer> mappedBuffer = pResource->GetBuffer(Subresource);
if (unlikely(needsReadback) && pResource->GetImage() != nullptr) {
Rc<DxvkImage> resourceImage = pResource->GetImage();
@ -4283,12 +4269,12 @@ namespace dxvk {
}
});
TrackTextureMappingBufferSequenceNumber(pResource, Subresource);
} else if (!(Flags & D3DLOCK_DONOTWAIT) && !WaitForResource(mappedBuffer, pResource->GetMappingBufferSequenceNumber(Subresource), D3DLOCK_DONOTWAIT)) {
pResource->EnableStagingBufferUploads(Subresource);
}
if (!WaitForResource(mappedBuffer, pResource->GetMappingBufferSequenceNumber(Subresource), Flags))
return D3DERR_WASSTILLDRAWING;
} else if (alloced) {
std::memset(physSlice.mapPtr, 0, physSlice.length);
}
}
@ -4324,7 +4310,7 @@ namespace dxvk {
}
}
if (managed && !readOnly) {
if (IsPoolManaged(desc.Pool) && !readOnly) {
pResource->SetNeedsUpload(Subresource, true);
for (uint32_t i : bit::BitMask(m_activeTextures)) {
@ -4475,38 +4461,25 @@ namespace dxvk {
+ srcOffsetBlockCount.y * pitch
+ srcOffsetBlockCount.x * formatInfo->elementSize;
VkDeviceSize sliceAlignment = 1;
VkDeviceSize rowAlignment = 1;
DxvkBufferSlice copySrcSlice;
if (pSrcTexture->DoesStagingBufferUploads(SrcSubresource)) {
VkDeviceSize dirtySize = extentBlockCount.width * extentBlockCount.height * extentBlockCount.depth * formatInfo->elementSize;
D3D9BufferSlice slice = AllocStagingBuffer(dirtySize);
copySrcSlice = slice.slice;
void* srcData = reinterpret_cast<uint8_t*>(srcSlice.mapPtr) + copySrcOffset;
util::packImageData(
slice.mapPtr, srcData, extentBlockCount, formatInfo->elementSize,
pitch, pitch * srcTexLevelExtentBlockCount.height);
} else {
copySrcSlice = DxvkBufferSlice(pSrcTexture->GetBuffer(SrcSubresource), copySrcOffset, srcSlice.length);
// row/slice alignment can act as the pitch parameter
rowAlignment = pitch;
sliceAlignment = srcTexLevelExtentBlockCount.height * pitch;
}
VkDeviceSize dirtySize = extentBlockCount.width * extentBlockCount.height * extentBlockCount.depth * formatInfo->elementSize;
D3D9BufferSlice slice = AllocStagingBuffer(dirtySize);
const void* srcData = reinterpret_cast<const uint8_t*>(srcSlice.mapPtr) + copySrcOffset;
util::packImageData(
slice.mapPtr, srcData, extentBlockCount, formatInfo->elementSize,
pitch, pitch * srcTexLevelExtentBlockCount.height);
EmitCs([
cSrcSlice = std::move(copySrcSlice),
cSrcSlice = slice.slice,
cDstImage = image,
cDstLayers = dstLayers,
cDstLevelExtent = alignedExtent,
cOffset = alignedDestOffset,
cRowAlignment = rowAlignment,
cSliceAlignment = sliceAlignment
cOffset = alignedDestOffset
] (DxvkContext* ctx) {
ctx->copyBufferToImage(
cDstImage, cDstLayers,
cOffset, cDstLevelExtent,
cSrcSlice.buffer(), cSrcSlice.offset(),
cRowAlignment, cSliceAlignment);
1, 1);
});
TrackTextureMappingBufferSequenceNumber(pSrcTexture, SrcSubresource);
@ -4607,11 +4580,14 @@ namespace dxvk {
if ((desc.Pool == D3DPOOL_DEFAULT || !(Flags & D3DLOCK_NO_DIRTY_UPDATE)) && !(Flags & D3DLOCK_READONLY))
pResource->DirtyRange().Conjoin(lockRange);
const bool directMapping = pResource->GetMapMode() == D3D9_COMMON_BUFFER_MAP_MODE_DIRECT;
const bool needsReadback = pResource->NeedsReadback();
Rc<DxvkBuffer> mappingBuffer = pResource->GetBuffer<D3D9_COMMON_BUFFER_TYPE_MAPPING>();
DxvkBufferSliceHandle physSlice;
if (Flags & D3DLOCK_DISCARD) {
if ((Flags & D3DLOCK_DISCARD) && (directMapping || needsReadback)) {
// Allocate a new backing slice for the buffer and set
// it as the 'new' mapped slice. This assumes that the
// only way to invalidate a buffer is by mapping it.
@ -4625,7 +4601,6 @@ namespace dxvk {
});
pResource->SetNeedsReadback(false);
pResource->GPUReadingRange().Clear();
}
else {
// Use map pointer from previous map operation. This
@ -4633,31 +4608,18 @@ namespace dxvk {
// if the map mode is D3DLOCK_NOOVERWRITE.
physSlice = pResource->GetMappedSlice();
// NOOVERWRITE promises that they will not write in a currently used area.
// Therefore we can skip waiting for these two cases.
// We can also skip waiting if there is not dirty range overlap, if we are one of those resources.
// If we are respecting the bounds ie. (MANAGED) we can test overlap
// of our bounds, otherwise we just ignore this and go for it all the time.
const bool needsReadback = pResource->NeedsReadback();
const bool readOnly = Flags & D3DLOCK_READONLY;
const bool noOverlap = !pResource->GPUReadingRange().Overlaps(lockRange);
// NOOVERWRITE promises that they will not write in a currently used area.
const bool noOverwrite = Flags & D3DLOCK_NOOVERWRITE;
const bool usesStagingBuffer = pResource->DoesStagingBufferUploads();
const bool directMapping = pResource->GetMapMode() == D3D9_COMMON_BUFFER_MAP_MODE_DIRECT;
const bool skipWait = (!needsReadback && (usesStagingBuffer || readOnly || (noOverlap && !directMapping))) || noOverwrite;
const bool skipWait = (!needsReadback && (readOnly || !directMapping)) || noOverwrite;
if (!skipWait) {
if (unlikely(needsReadback)) {
Logger::warn("Buffer readback is unimplemented.");
// Remember to update the sequence number when implementing buffer readback.
} else if (!(Flags & D3DLOCK_DONOTWAIT) && !WaitForResource(mappingBuffer, pResource->GetMappingBufferSequenceNumber(), D3DLOCK_DONOTWAIT))
pResource->EnableStagingBufferUploads();
const Rc<DxvkBuffer> mappingBuffer = pResource->GetBuffer<D3D9_COMMON_BUFFER_TYPE_MAPPING>();
if (!WaitForResource(mappingBuffer, pResource->GetMappingBufferSequenceNumber(), Flags))
return D3DERR_WASSTILLDRAWING;
pResource->SetNeedsReadback(false);
pResource->GPUReadingRange().Clear();
}
}
@ -4688,19 +4650,13 @@ namespace dxvk {
D3D9Range& range = pResource->DirtyRange();
DxvkBufferSlice copySrcSlice;
if (pResource->DoesStagingBufferUploads()) {
D3D9BufferSlice slice = AllocStagingBuffer(range.max - range.min);
copySrcSlice = slice.slice;
void* srcData = reinterpret_cast<uint8_t*>(srcSlice.mapPtr) + range.min;
memcpy(slice.mapPtr, srcData, range.max - range.min);
} else {
copySrcSlice = DxvkBufferSlice(pResource->GetBuffer<D3D9_COMMON_BUFFER_TYPE_MAPPING>(), range.min, range.max - range.min);
}
D3D9BufferSlice slice = AllocStagingBuffer(range.max - range.min);
void* srcData = reinterpret_cast<uint8_t*>(srcSlice.mapPtr) + range.min;
memcpy(slice.mapPtr, srcData, range.max - range.min);
EmitCs([
cDstSlice = dstBuffer,
cSrcSlice = copySrcSlice,
cSrcSlice = slice.slice,
cDstOffset = range.min,
cLength = range.max - range.min
] (DxvkContext* ctx) {
@ -4712,7 +4668,6 @@ namespace dxvk {
cLength);
});
pResource->GPUReadingRange().Conjoin(pResource->DirtyRange());
pResource->DirtyRange().Clear();
TrackBufferMappingBufferSequenceNumber(pResource);