P4 to Git Change 1537173 by gandryey@gera-w8 on 2018/04/05 13:34:21

SWDEV-149190 - [CQE OCL][DTB-BLOCKER][WIN][LNX][QR]observed soft hang while running oclruntime sub-test of ocltst due to faulty CL#1529531
	- Better solution for CL#1537141. Use the offset in the bind for the thread trace object

Affected files ...

... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palthreadtrace.cpp#6 edit

rocclr/runtime/device/pal/palthreadtrace.cpp

@@ -58,11 +58,7 @@ bool PalThreadTraceReference::finalize() {
   // Acquire GPU memory for the query from the pool and bind it.
   Pal::GpuMemoryRequirements gpuMemReqs = {};
   iPerf()->GetGpuMemoryRequirements(&gpuMemReqs);
-  // Note: Pal:::PerfExperiments can't apply the runtime offsets for suballocaiotns,
-  // thus disable suballocation.
-  Pal::gpusize bufSize = std::max(gpuMemReqs.size,
-    gpu_.dev().settings().subAllocationMaxSize_ + gpuMemReqs.alignment);
-  memory_ = new Memory(gpu().dev(), amd::alignUp(bufSize, gpuMemReqs.alignment));
+  memory_ = new Memory(gpu().dev(), amd::alignUp(gpuMemReqs.size, gpuMemReqs.alignment));
 
   if (nullptr == memory_) {
     return false;
@@ -74,7 +70,7 @@ bool PalThreadTraceReference::finalize() {
 
   gpu_.queue(gpu_.engineID_).addMemRef(memory_->iMem());
 
-  result = iPerf()->BindGpuMemory(memory_->iMem(), 0);
+  result = iPerf()->BindGpuMemory(memory_->iMem(), memory_->offset());
 
   if (result != Pal::Result::Success) {
     return false;