From d9cb5925a7405499d66fd3978e9134d52cf63001 Mon Sep 17 00:00:00 2001 From: foreman Date: Tue, 7 Mar 2017 14:17:43 -0500 Subject: [PATCH] P4 to Git Change 1382437 by gandryey@gera-w8 on 2017/03/07 13:57:15 SWDEV-115242 - [ROCm CQE][OCLonLC][QR][G] ocltst - OCLImage2DFromBuffer results in Segmentation fault - don't use asImage() virtual method, since coherency layer may require a transfer during object destruction Affected files ... ... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocblit.cpp#15 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocblit.hpp#6 edit [ROCm/clr commit: 6950c861adf4c4d1d2c3ea1a18f939145db6f054] --- .../rocclr/runtime/device/rocm/rocblit.cpp | 119 +++++++++++------- .../rocclr/runtime/device/rocm/rocblit.hpp | 4 + 2 files changed, 78 insertions(+), 45 deletions(-) diff --git a/projects/clr/rocclr/runtime/device/rocm/rocblit.cpp b/projects/clr/rocclr/runtime/device/rocm/rocblit.cpp index db57bc622c..2179ed658a 100644 --- a/projects/clr/rocclr/runtime/device/rocm/rocblit.cpp +++ b/projects/clr/rocclr/runtime/device/rocm/rocblit.cpp @@ -949,7 +949,8 @@ KernelBlitManager::copyBufferToImage( static const bool CopyRect = false; // Flush DMA for ASYNC copy static const bool FlushDMA = true; - size_t imgRowPitch = size[0] * gpuMem(dstMemory).owner()->asImage()->getImageFormat().getElementSize(); + amd::Image* dstImage = static_cast(dstMemory.owner()); + size_t imgRowPitch = size[0] * dstImage->getImageFormat().getElementSize(); size_t imgSlicePitch = imgRowPitch * size[1]; if (setup_.disableCopyBufferToImage_) { @@ -992,7 +993,8 @@ CalcRowSlicePitches( cl_ulong* pitch, const cl_int* copySize, size_t rowPitch, size_t slicePitch, const Memory& mem) { - uint32_t memFmtSize = mem.owner()->asImage()->getImageFormat().getElementSize(); + amd::Image* image = static_cast(mem.owner()); + uint32_t memFmtSize = image->getImageFormat().getElementSize(); bool img1Darray = (mem.owner()->getType() == CL_MEM_OBJECT_IMAGE1D_ARRAY) ? true : false; if (rowPitch == 0) { @@ -1015,7 +1017,7 @@ CalcRowSlicePitches( } } -static void +static inline void setArgument(amd::Kernel* kernel, size_t index, size_t size, const void* value) { kernel->parameters().set(index, size, value); @@ -1036,7 +1038,9 @@ KernelBlitManager::copyBufferToImageKernel( Memory* dstView = &gpuMem(dstMemory); bool releaseView = false; bool result = false; - amd::Image::Format newFormat(gpuMem(dstMemory).owner()->asImage()->getImageFormat()); + amd::Image* dstImage = static_cast(dstMemory.owner()); + amd::Image* srcImage = static_cast(srcMemory.owner()); + amd::Image::Format newFormat(dstImage->getImageFormat()); // Find unsupported formats for (uint i = 0; i < RejectedFormatDataTotal; ++i) { @@ -1059,8 +1063,7 @@ KernelBlitManager::copyBufferToImageKernel( // If the image format was rejected, then attempt to create a view if (rejected && // todo ROC runtime has a problem with a view for this format - (gpuMem(dstMemory).owner()->asImage()-> - getImageFormat().image_channel_data_type != CL_UNORM_INT_101010)) { + (dstImage->getImageFormat().image_channel_data_type != CL_UNORM_INT_101010)) { dstView = createView(gpuMem(dstMemory), newFormat, CL_MEM_WRITE_ONLY); if (dstView != NULL) { rejected = false; @@ -1084,14 +1087,14 @@ KernelBlitManager::copyBufferToImageKernel( // Program the kernels workload depending on the blit dimensions dim = 3; - if (dstMemory.owner()->asImage()->getDims() == 1) { + if (dstImage->getDims() == 1) { globalWorkSize[0] = amd::alignUp(size[0], 256); globalWorkSize[1] = amd::alignUp(size[1], 1); globalWorkSize[2] = amd::alignUp(size[2], 1); localWorkSize[0] = 256; localWorkSize[1] = localWorkSize[2] = 1; } - else if (dstMemory.owner()->asImage()->getDims() == 2) { + else if (dstImage->getDims() == 2) { globalWorkSize[0] = amd::alignUp(size[0], 16); globalWorkSize[1] = amd::alignUp(size[1], 16); globalWorkSize[2] = amd::alignUp(size[2], 1); @@ -1111,8 +1114,8 @@ KernelBlitManager::copyBufferToImageKernel( setArgument(kernels_[blitType], 0, sizeof(cl_mem), &mem); mem = as_cl(dstView->owner()); setArgument(kernels_[blitType], 1, sizeof(cl_mem), &mem); - uint32_t memFmtSize = dstMemory.owner()->asImage()->getImageFormat().getElementSize(); - uint32_t components = dstMemory.owner()->asImage()->getImageFormat().getNumChannels(); + uint32_t memFmtSize = dstImage->getImageFormat().getElementSize(); + uint32_t components = dstImage->getImageFormat().getNumChannels(); // 1 element granularity for writes by default cl_int granularity = 1; @@ -1156,10 +1159,9 @@ KernelBlitManager::copyBufferToImageKernel( globalWorkOffset, globalWorkSize, localWorkSize); // Execute the blit - address parameters = kernels_[blitType]->parameters().capture(dev()); + address parameters = captureArguments(kernels_[blitType]); result = gpu().submitKernelInternal(ndrange, *kernels_[blitType], parameters, NULL); - kernels_[blitType]->parameters().release(const_cast
(parameters), dev()); - + releaseArguments(parameters); if (releaseView) { // todo SRD programming could be changed to avoid a stall gpu().releaseGpuMemoryFence(); @@ -1185,7 +1187,8 @@ KernelBlitManager::copyImageToBuffer( static const bool CopyRect = false; // Flush DMA for ASYNC copy static const bool FlushDMA = true; - size_t imgRowPitch = size[0] * gpuMem(srcMemory).owner()->asImage()->getImageFormat().getElementSize(); + amd::Image* srcImage = static_cast(srcMemory.owner()); + size_t imgRowPitch = size[0] * srcImage->getImageFormat().getElementSize(); size_t imgSlicePitch = imgRowPitch * size[1]; if (setup_.disableCopyImageToBuffer_) { @@ -1240,7 +1243,8 @@ KernelBlitManager::copyImageToBufferKernel( Memory* srcView = &gpuMem(srcMemory); bool releaseView = false; bool result = false; - amd::Image::Format newFormat(gpuMem(srcMemory).owner()->asImage()->getImageFormat()); + amd::Image* srcImage = static_cast(srcMemory.owner()); + amd::Image::Format newFormat(srcImage->getImageFormat()); // Find unsupported formats for (uint i = 0; i < RejectedFormatDataTotal; ++i) { @@ -1263,8 +1267,7 @@ KernelBlitManager::copyImageToBufferKernel( // If the image format was rejected, then attempt to create a view if (rejected && // todo ROC runtime has a problem with a view for this format - (gpuMem(srcMemory).owner()->asImage()-> - getImageFormat().image_channel_data_type != CL_UNORM_INT_101010)) { + (srcImage->getImageFormat().image_channel_data_type != CL_UNORM_INT_101010)) { srcView = createView(gpuMem(srcMemory), newFormat, CL_MEM_READ_ONLY); if (srcView != NULL) { rejected = false; @@ -1288,14 +1291,14 @@ KernelBlitManager::copyImageToBufferKernel( // Program the kernels workload depending on the blit dimensions dim = 3; // Find the current blit type - if (srcMemory.owner()->asImage()->getDims() == 1) { + if (srcImage->getDims() == 1) { globalWorkSize[0] = amd::alignUp(size[0], 256); globalWorkSize[1] = amd::alignUp(size[1], 1); globalWorkSize[2] = amd::alignUp(size[2], 1); localWorkSize[0] = 256; localWorkSize[1] = localWorkSize[2] = 1; } - else if (srcMemory.owner()->asImage()->getDims() == 2) { + else if (srcImage->getDims() == 2) { globalWorkSize[0] = amd::alignUp(size[0], 16); globalWorkSize[1] = amd::alignUp(size[1], 16); globalWorkSize[2] = amd::alignUp(size[2], 1); @@ -1329,8 +1332,8 @@ KernelBlitManager::copyImageToBufferKernel( (cl_int)size[1], (cl_int)size[2], 0 }; setArgument(kernels_[blitType], 4, sizeof(srcOrg), srcOrg); - uint32_t memFmtSize = srcMemory.owner()->asImage()->getImageFormat().getElementSize(); - uint32_t components = srcMemory.owner()->asImage()->getImageFormat().getNumChannels(); + uint32_t memFmtSize = srcImage->getImageFormat().getElementSize(); + uint32_t components = srcImage->getImageFormat().getNumChannels(); // 1 element granularity for writes by default cl_int granularity = 1; @@ -1365,9 +1368,9 @@ KernelBlitManager::copyImageToBufferKernel( globalWorkOffset, globalWorkSize, localWorkSize); // Execute the blit - address parameters = kernels_[blitType]->parameters().capture(dev()); + address parameters = captureArguments(kernels_[blitType]); result = gpu().submitKernelInternal(ndrange, *kernels_[blitType], parameters, NULL); - kernels_[blitType]->parameters().release(const_cast
(parameters), dev()); + releaseArguments(parameters); if (releaseView) { // todo SRD programming could be changed to avoid a stall gpu().releaseGpuMemoryFence(); @@ -1392,7 +1395,9 @@ KernelBlitManager::copyImage( Memory* dstView = &gpuMem(dstMemory); bool releaseView = false; bool result = false; - amd::Image::Format newFormat(gpuMem(srcMemory).owner()->asImage()->getImageFormat()); + amd::Image* srcImage = static_cast(srcMemory.owner()); + amd::Image* dstImage = static_cast(dstMemory.owner()); + amd::Image::Format newFormat(srcImage->getImageFormat()); // Find unsupported formats for (uint i = 0; i < RejectedFormatDataTotal; ++i) { @@ -1448,16 +1453,16 @@ KernelBlitManager::copyImage( // Program the kernels workload depending on the blit dimensions dim = 3; // Find the current blit type - if ((srcMemory.owner()->asImage()->getDims() == 1) || - (dstMemory.owner()->asImage()->getDims() == 1)) { + if ((srcImage->getDims() == 1) || + (dstImage->getDims() == 1)) { globalWorkSize[0] = amd::alignUp(size[0], 256); globalWorkSize[1] = amd::alignUp(size[1], 1); globalWorkSize[2] = amd::alignUp(size[2], 1); localWorkSize[0] = 256; localWorkSize[1] = localWorkSize[2] = 1; } - else if ((srcMemory.owner()->asImage()->getDims() == 2) || - (dstMemory.owner()->asImage()->getDims() == 2)) { + else if ((srcImage->getDims() == 2) || + (dstImage->getDims() == 2)) { globalWorkSize[0] = amd::alignUp(size[0], 16); globalWorkSize[1] = amd::alignUp(size[1], 16); globalWorkSize[2] = amd::alignUp(size[2], 1); @@ -1507,9 +1512,9 @@ KernelBlitManager::copyImage( globalWorkOffset, globalWorkSize, localWorkSize); // Execute the blit - address parameters = kernels_[blitType]->parameters().capture(dev()); + address parameters = captureArguments(kernels_[blitType]); result = gpu().submitKernelInternal(ndrange, *kernels_[blitType], parameters, NULL); - kernels_[blitType]->parameters().release(const_cast
(parameters), dev()); + releaseArguments(parameters); if (releaseView) { // todo SRD programming could be changed to avoid a stall gpu().releaseGpuMemoryFence(); @@ -1527,7 +1532,8 @@ FindPinSize( size_t& pinSize, const amd::Coord3D& size, size_t& rowPitch, size_t& slicePitch, const Memory& mem) { - pinSize = size[0] * mem.owner()->asImage()->getImageFormat().getElementSize(); + amd::Image* image = static_cast(mem.owner()); + pinSize = size[0] * image->getImageFormat().getElementSize(); if ((rowPitch == 0) || (rowPitch == pinSize)) { rowPitch = 0; } @@ -1536,7 +1542,7 @@ FindPinSize( } // Calculate the pin size, which should be equal to the copy size - for (uint i = 1; i < mem.owner()->asImage()->getDims(); ++i) { + for (uint i = 1; i < image->getDims(); ++i) { pinSize *= size[i]; if (i == 1) { if ((slicePitch == 0) || (slicePitch == pinSize)) { @@ -1782,10 +1788,9 @@ KernelBlitManager::copyBufferRect( globalWorkOffset, globalWorkSize, localWorkSize); // Execute the blit - address parameters = kernels_[blitType]->parameters().capture(dev()); + address parameters = captureArguments(kernels_[blitType]); result = gpu().submitKernelInternal(ndrange, *kernels_[blitType], parameters, NULL); - kernels_[blitType]->parameters().release(const_cast
(parameters), dev()); - + releaseArguments(parameters); synchronize(); return result; @@ -2085,9 +2090,9 @@ KernelBlitManager::fillBuffer( globalWorkOffset, &globalWorkSize, &localWorkSize); // Execute the blit - address parameters = kernels_[fillType]->parameters().capture(dev()); + address parameters = captureArguments(kernels_[fillType]); result = gpu().submitKernelInternal(ndrange, *kernels_[fillType], parameters, NULL); - kernels_[fillType]->parameters().release(const_cast
(parameters), dev()); + releaseArguments(parameters); } synchronize(); @@ -2181,9 +2186,9 @@ KernelBlitManager::copyBuffer( globalWorkOffset, &globalWorkSize, &localWorkSize); // Execute the blit - address parameters = kernels_[blitType]->parameters().capture(dev()); + address parameters = captureArguments(kernels_[blitType]); result = gpu().submitKernelInternal(ndrange, *kernels_[blitType], parameters, NULL); - kernels_[blitType]->parameters().release(const_cast
(parameters), dev()); + releaseArguments(parameters); } else { result = DmaBlitManager::copyBuffer( @@ -2222,7 +2227,8 @@ KernelBlitManager::fillImage( size_t globalWorkSize[3]; size_t localWorkSize[3]; Memory* memView = &gpuMem(memory); - amd::Image::Format newFormat(gpuMem(memory).owner()->asImage()->getImageFormat()); + amd::Image* image = static_cast(memory.owner()); + amd::Image::Format newFormat(image->getImageFormat()); // Program the kernels workload depending on the fill dimensions fillType = FillImage; @@ -2278,14 +2284,14 @@ KernelBlitManager::fillImage( // Perform workload split to allow multiple operations in a single thread globalWorkSize[0] = (size[0] + TransferSplitSize - 1) / TransferSplitSize; // Find the current blit type - if (memView->owner()->asImage()->getDims() == 1) { + if (image->getDims() == 1) { globalWorkSize[0] = amd::alignUp(globalWorkSize[0], 256); globalWorkSize[1] = amd::alignUp(size[1], 1); globalWorkSize[2] = amd::alignUp(size[2], 1); localWorkSize[0] = 256; localWorkSize[1] = localWorkSize[2] = 1; } - else if (memView->owner()->asImage()->getDims()== 2) { + else if (image->getDims()== 2) { globalWorkSize[0] = amd::alignUp(globalWorkSize[0], 16); globalWorkSize[1] = amd::alignUp(size[1], 16); globalWorkSize[2] = amd::alignUp(size[2], 1); @@ -2348,9 +2354,9 @@ KernelBlitManager::fillImage( globalWorkOffset, globalWorkSize, localWorkSize); // Execute the blit - address parameters = kernels_[fillType]->parameters().capture(dev()); + address parameters = captureArguments(kernels_[fillType]); result = gpu().submitKernelInternal(ndrange, *kernels_[fillType], parameters, NULL); - kernels_[fillType]->parameters().release(const_cast
(parameters), dev()); + releaseArguments(parameters); if (releaseView) { // todo SRD programming could be changed to avoid a stall gpu().releaseGpuMemoryFence(); @@ -2423,7 +2429,8 @@ KernelBlitManager::createView( cl_mem_flags flags) const { assert((parent.owner()->asBuffer() == nullptr) && "View supports images only"); - amd::Image *image = parent.owner()->asImage()->createView( + amd::Image* parentImage = static_cast(parent.owner()); + amd::Image* image = parentImage->createView( parent.owner()->getContext(), format, &gpu(), 0, flags); if (image == NULL) { @@ -2450,4 +2457,26 @@ KernelBlitManager::createView( return devImage; } +address +KernelBlitManager::captureArguments(const amd::Kernel* kernel) const +{ + const size_t stackSize = kernel->signature().paramsSize(); + const size_t svmInfoSize = kernel->signature().numParameters() * sizeof(bool); + address args = reinterpret_cast
(amd::AlignedMemory::allocate( + stackSize + svmInfoSize, PARAMETERS_MIN_ALIGNMENT)); + if (args == nullptr) { + LogWarning("Failed to allocate memory for arguments"); + return nullptr; + } + memcpy(args, kernel->parameters().values(), kernel->signature().paramsSize()); + memset(args + stackSize, 0, svmInfoSize); + return args; +} + +void +KernelBlitManager::releaseArguments(address args) const +{ + amd::AlignedMemory::deallocate(args); +} + } // namespace pal diff --git a/projects/clr/rocclr/runtime/device/rocm/rocblit.hpp b/projects/clr/rocclr/runtime/device/rocm/rocblit.hpp index 7a22a42cb9..e425598893 100644 --- a/projects/clr/rocclr/runtime/device/rocm/rocblit.hpp +++ b/projects/clr/rocclr/runtime/device/rocm/rocblit.hpp @@ -407,6 +407,7 @@ public: private: static const size_t MaxXferBuffers = 2; static const uint TransferSplitSize = 1; + static const uint MaxNumIssuedTransfers = 3; //! Copies a buffer object to an image object bool copyBufferToImageKernel( @@ -444,6 +445,9 @@ private: cl_mem_flags flags //!< Memory flags ) const; + address captureArguments(const amd::Kernel* kernel) const; + void releaseArguments(address args) const; + //! Disable copy constructor KernelBlitManager(const KernelBlitManager&);