AI/rocm-systems
2
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

P4 to Git Change 1968527 by gandryey@gera-win10 on 2019/07/16 14:52:45

Browse Source 
SWDEV-195023 - [CQE OCL][Navi10][RESOLVE] corruption seen in thumbnail for mxf clip after enabling temporal denoiser in Davinci resolve app
	- Add a workaround for missing custom pitch in gfx10 HW. It can be disabled with GPU_IMAGE_BUFFER_WAR=0. Workaround implements double copy with an image without pitch.

Affected files ...

... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palmemory.cpp#26 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palmemory.hpp#12 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palsettings.cpp#89 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palsettings.hpp#24 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.cpp#138 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.hpp#62 edit
... //depot/stg/opencl/drivers/opencl/runtime/utils/flags.hpp#313 edit
This commit is contained in:
foreman committed 2019-07-16 14:56:08 -04:00
1 parent 901336f846
commit 582fae6820
7 files changed
+99 -4
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

rocclr/runtime/device/pal/palvirtual.cpp
+52
View File
@@ -1105,6 +1105,13 @@ void VirtualGPU::submitReadMemory(amd::ReadMemoryCommand& vcmd) {
}
} break;
case CL_COMMAND_READ_IMAGE:
if (memory->memoryType() == Resource::ImageBuffer) {
Image* imageBuffer = static_cast<Image*>(memory);
// Check if synchronization has to be performed
if (imageBuffer->CopyImageBuffer() != nullptr) {
memory = imageBuffer->CopyImageBuffer();
}
}
if (hostMemory != nullptr) {
// Accelerated image to buffer transfer without pinning
amd::Coord3D dstOrigin(offset);
@@ -2398,6 +2405,25 @@ bool VirtualGPU::submitKernelInternal(const amd::NDRangeContainer& sizes, const
}
}
// Check if image buffer write back is required
if (state_.imageBufferWrtBack_) {
// Avoid recursive write back
state_.imageBufferWrtBack_ = false;
// Make sure the original kernel execution is done
addBarrier(RgpSqqtBarrierReason::MemDependency);
for (const auto imageBuffer : wrtBackImageBuffer_) {
Memory* buffer = dev().getGpuMemory(imageBuffer->owner()->parent());
amd::Image* image = imageBuffer->owner()->asImage();
amd::Coord3D offs(0);
// Copy memory from the the backing store image into original buffer
bool result = blitMgr().copyImageToBuffer(
*imageBuffer->CopyImageBuffer(), *buffer, offs, offs,
image->getRegion(), true,
image->getRowPitch(), image->getSlicePitch());
}
wrtBackImageBuffer_.clear();
}
// Perform post dispatch logic for RGP traces
if (rgpCaptureEna()) {
dev().rgpCaptureMgr()->PostDispatch(this);
@@ -3256,6 +3282,32 @@ bool VirtualGPU::processMemObjectsHSA(const amd::Kernel& kernel, const_address p
mem->signalWrite(&dev());
}
if (info.oclObject_ == amd::KernelParameterDescriptor::ImageObject) {
if (gpuMem->memoryType() == Resource::ImageBuffer) {
Image* imageBuffer = static_cast<Image*>(gpuMem);
// Check if synchronization has to be performed
if (imageBuffer->CopyImageBuffer() != nullptr) {
Memory* buffer = dev().getGpuMemory(mem->parent());
amd::Image* image = mem->asImage();
amd::Coord3D offs(0);
// Copy memory from the original image buffer into the backing store image
bool result = blitMgr().copyBufferToImage(
*buffer, *imageBuffer->CopyImageBuffer(), offs, offs,
image->getRegion(), true, image->getRowPitch(), image->getSlicePitch());
// Make sure the copy operation is done
addBarrier(RgpSqqtBarrierReason::MemDependency);
// Use backing store SRD as the replacment
uint64_t srd = imageBuffer->CopyImageBuffer()->hwSrd();
WriteAqlArgAt(const_cast<address>(params), &srd, sizeof(srd), desc.offset_);
// Add backing store image to the list of memory handles
addVmMemory(imageBuffer->CopyImageBuffer());
// If it's not a read only resource, then runtime has to write back
if (!info.readOnly_) {
wrtBackImageBuffer_.push_back(imageBuffer);
state_.imageBufferWrtBack_ = true;
}
}
}
//! \note Special case for the image views.
//! Copy SRD to CB1, so blit manager will be able to release
//! this view without a wait for SRD resource.