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

SWDEV-359287 - uniform_work_group_size is interger type is 1 for true and 0 for false.

Browse Source 
Change-Id: Iac5584c3cdd3bf78e3f1e89033b9ce7e07afefca
This commit is contained in:
Jaydeep Patel
2024-01-24 04:02:51 +00:00
committed by Maneesh Gupta
parent 8f9a20738e
commit 4177b4b675
3 changed files with 7 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
rocclr/device/devkernel.cpp
+1 -1
View File
@@ -575,7 +575,7 @@ static amd_comgr_status_t populateKernelMetaV3(const amd_comgr_metadata_node_t k
kernel->SetWGPMode(buf.compare("true") == 0); kernel->SetWGPMode(buf.compare("true") == 0);
break; break;
case KernelField::UniformWrokGroupSize: case KernelField::UniformWrokGroupSize:
kernel->setUniformWorkGroupSize(buf.compare("true") == 0); kernel->setUniformWorkGroupSize(buf.compare("1") == 0);
break; break;
default: default:
return AMD_COMGR_STATUS_ERROR; return AMD_COMGR_STATUS_ERROR;
rocclr/device/pal/palprogram.cpp
+3 -1
View File
@@ -762,7 +762,9 @@ bool LightningProgram::createKernels(void* binary, size_t binSize, bool useUnifo
} }
kernels()[kernelName] = kernel; kernels()[kernelName] = kernel;
kernel->setUniformWorkGroupSize(useUniformWorkGroupSize); if (codeObjectVer() < 5) {
kernel->setUniformWorkGroupSize(useUniformWorkGroupSize);
}
} }
} }
executable_ = loader_->CreateExecutable(HSA_PROFILE_FULL, nullptr); executable_ = loader_->CreateExecutable(HSA_PROFILE_FULL, nullptr);
rocclr/device/rocm/rocprogram.cpp
+3 -1
View File
@@ -274,7 +274,9 @@ bool LightningProgram::createKernels(void* binary, size_t binSize, bool useUnifo
if (!aKernel->init()) { if (!aKernel->init()) {
return false; return false;
} }
aKernel->setUniformWorkGroupSize(useUniformWorkGroupSize); if (codeObjectVer() < 5) {
aKernel->setUniformWorkGroupSize(useUniformWorkGroupSize);
}
aKernel->setInternalKernelFlag(internalKernel); aKernel->setInternalKernelFlag(internalKernel);
kernels()[kernelName] = aKernel; kernels()[kernelName] = aKernel;
} }