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

P4 to Git Change 1056280 by gandryey@gera-dev-w7 on 2014/07/17 11:55:30

Browse Source 
ECR #304775 - Device enqueuing
	- Add extra sync point for the scheduler exit.

Affected files ...

... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpusched.hpp#9 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuschedcl.cpp#21 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuvirtual.cpp#323 edit
This commit is contained in:
foreman
2014-07-17 12:20:49 -04:00
parent 4798473034
commit 6627cf5cf7
3 changed files with 24 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files
rocclr/runtime/device/gpu/gpuvirtual.cpp
+2 -1
View File
@@ -1792,7 +1792,7 @@ VirtualGPU::submitKernelInternalHSA(
address argum = gpuDefQueue->virtualQueue_->data() + offsArg;
print << "Kernel: " << child->name() << "\n";
static const char* Names[HSAILKernel::ExtraArguments] = {
"Offset0: ", "Offset1: ","Offset2: ","PrintfBuf: ", "VqueuePtr: ", "AqlWarap: "};
"Offset0: ", "Offset1: ","Offset2: ","PrintfBuf: ", "VqueuePtr: ", "AqlWrap: "};
for (j = 0; j < HSAILKernel::ExtraArguments; ++j) {
print << "\t" << Names[j] << *(size_t*)argum;
print << "\n";
@@ -1832,6 +1832,7 @@ VirtualGPU::submitKernelInternalHSA(
param->hw_queue = patchStart + sizeof(uint32_t)/* Rewind packet*/;
param->hsa_queue = gpuDefQueue->hsaQueueMem()->vmAddress();
param->launch = 0;
param->releaseHostCP = 0;
// Fill the scratch buffer information
if (hsaKernel.prog().maxScratchRegs() > 0) {
gpu::Memory* scratchBuf = dev().scratch(gpuDefQueue->hwRing())->memObjs_[0];