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

P4 to Git Change 1293023 by wchau@wchau_WIN_OCL_HSA on 2016/07/19 10:24:01

Browse Source 
SWDEV-89502 - [OCL] Support AMD DVR Core functionalities - add support for PFPA_DOPP_SUBMISSION & GL_RGB10_A2 format (changes in OCL RT)

Affected files ...

... //depot/stg/opencl/drivers/opencl/api/opencl/amdocl/cl_gl.cpp#51 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/amdocl/cl_svm.cpp#18 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl1.2/CL/cl_ext.h#12 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl2.0/CL/cl_ext.h#26 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl2.1/CL/cl_ext.h#5 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpukernel.cpp#317 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpukernel.hpp#125 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuvirtual.cpp#404 edit
... //depot/stg/opencl/drivers/opencl/runtime/platform/kernel.hpp#17 edit
This commit is contained in:
foreman
2016-07-19 10:45:55 -04:00
parent 84b2757ab7
commit 55eab0bebd
4 changed files with 25 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files
rocclr/runtime/device/gpu/gpukernel.cpp
+2 -2
View File
@@ -1567,7 +1567,7 @@ Kernel::loadParameters(
}
bool
Kernel::run(VirtualGPU& gpu, GpuEvent* calEvent, bool lastRun, bool lastDoppCmd) const
Kernel::run(VirtualGPU& gpu, GpuEvent* calEvent, bool lastRun, bool lastDoppCmd, bool pfpaDoppCmd) const
{
const VirtualGPU::CalVirtualDesc* dispatch = gpu.cal();
@@ -1576,7 +1576,7 @@ Kernel::run(VirtualGPU& gpu, GpuEvent* calEvent, bool lastRun, bool lastDoppCmd)
gpu.eventBegin(MainEngine);
gpu.rs()->Dispatch(gpu.cs(), &dispatch->gridBlock, &dispatch->partialGridBlock,
&dispatch->gridSize, dispatch->localSize, gpu.vmMems(), dispatch->memCount_, lastDoppCmd);
&dispatch->gridSize, dispatch->localSize, gpu.vmMems(), dispatch->memCount_, lastDoppCmd, pfpaDoppCmd);
gpu.eventEnd(MainEngine, *calEvent);
// Unbind all resources
rocclr/runtime/device/gpu/gpukernel.hpp
+2 -1
View File
@@ -621,7 +621,8 @@ public:
VirtualGPU& gpu, //!< virtual GPU device object
GpuEvent* gpuEvent, //!< Pointer to the GPU event
bool lastRun, //!< Last run in the split execution
bool lastDoppCmd //!< info for kernel dispatch
bool lastDoppCmd, //!< for last dopp submission kernel dispatch
bool pfpaDoppCmd //!< for PFPA dopp submission kernel dispatch
) const;
//! Help function to debug the kernel output
rocclr/runtime/device/gpu/gpuvirtual.cpp
+3 -1
View File
@@ -1851,6 +1851,7 @@ VirtualGPU::submitKernelInternalHSA(
dispatchInfo.kernelInfo = pKernelInfo;
dispatchInfo.wavesPerSH = hsaKernel.getWavesPerSH(this);
dispatchInfo.lastDoppSubmission = kernel.parameters().getExecNewVcop();
dispatchInfo.pfpaDoppSubmission = kernel.parameters().getExecPfpaVcop();
GpuEvent gpuEvent;
// Run AQL dispatch in HW
@@ -2155,7 +2156,8 @@ VirtualGPU::submitKernelInternal(
}
// Execute the kernel
if (gpuKernel.run(*this, &gpuEvent, lastRun, kernel.parameters().getExecNewVcop())) {
if (gpuKernel.run(*this, &gpuEvent, lastRun, kernel.parameters().getExecNewVcop(),
kernel.parameters().getExecPfpaVcop())) {
//! @todo A flush is necessary to make sure
// that 2 consecutive runs won't access to the same
// private/local memory. CAL has to generate cache flush