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

P4 to Git Change 2026859 by axie@axie-hip-rocm on 2019/11/07 15:54:13

Browse Source 
SWDEV-198863 - Options for hip-clang-vdi path to provide the chicken bits, or functional equivalents to HCC_DB (phase 3)
	Use ClPrint to implement other log functions.
	Move some funtion to use new log functions.
	This is the final change of the JIRA.

	Tests:
	1. Linux HIP ROCM platform. VEGA10. Driver is release build.
	1.1 export LOG_LEVEL=3
	      ./hipModule
	      There are many logs.
	1.2 export GPU_LOG_MASK=0
	     ./hipModule
	     There is no log
	2. Windows HIP PAL platform. VEGA10, Driver is release build.
	2.1 set LOG_LEVEL=3
	    run test hipPrintfKernel
	    There are many logs
	2.2 set GPU_LOG_MASK=0
	    run test hipPrintfKernel
	    There is no log

	3. http://ocltc.amd.com:8111/viewModification.html?modId=128490&personal=true&tab=vcsModificationBuilds

	ReviewBoard: http://ocltc.amd.com/reviews/r/18247/

Affected files ...

... //depot/stg/opencl/drivers/opencl/runtime/device/pal/paldevice.cpp#175 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.cpp#155 edit
... //depot/stg/opencl/drivers/opencl/runtime/utils/debug.hpp#12 edit
This commit is contained in:
foreman
2019-11-07 16:06:51 -05:00
parent 4fbf7389b2
commit c10b29f0e0
3 changed files with 23 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files
rocclr/runtime/device/pal/palvirtual.cpp
+6 -6
View File
@@ -2392,7 +2392,7 @@ bool VirtualGPU::submitKernelInternal(const amd::NDRangeContainer& sizes, const
}
size_t ldsSize;
LogPrintfInfo("!\tShaderName : %s\n", hsaKernel.name().c_str());
ClPrint(amd::LOG_INFO, amd::LOG_KERN, "!\tShaderName : %s\n", hsaKernel.name().c_str());
// Check memory dependency and SVM objects
if (!processMemObjectsHSA(kernel, parameters, nativeMem, ldsSize)) {
@@ -3381,11 +3381,11 @@ bool VirtualGPU::processMemObjectsHSA(const amd::Kernel& kernel, const_address p
addVmMemory(gpuMem);
const void* globalAddress = *reinterpret_cast<const void* const*>(params + desc.offset_);
LogPrintfInfo("!\targ%d: %s %s = ptr:%p obj:[%p-%p] threadId : %zx\n", index,
desc.typeName_.c_str(), desc.name_.c_str(), globalAddress,
reinterpret_cast<void*>(gpuMem->vmAddress()),
reinterpret_cast<void*>(gpuMem->vmAddress() + gpuMem->size()),
std::this_thread::get_id());
ClPrint(amd::LOG_INFO, amd::LOG_KERN, "!\targ%d: %s %s = ptr:%p obj:[%p-%p] threadId : %zx\n", index,
desc.typeName_.c_str(), desc.name_.c_str(), globalAddress,
reinterpret_cast<void*>(gpuMem->vmAddress()),
reinterpret_cast<void*>(gpuMem->vmAddress() + gpuMem->size()),
std::this_thread::get_id());
//! Check if compiler expects read/write.
//! Note: SVM with subbuffers has an issue with tracking.