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

SWDEV-236178 - Fixing func_attributes that was over written on redesign.

Browse Source 
Change-Id: I157913ae4fa4a3a149817e901f10475b7750188b
This commit is contained in:
kjayapra-amd committed 2020-07-01 11:52:45 -04:00
1 parent f2102adc99
commit 6f15d8ca6b
2 files changed
+12 -32
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

rocclr/hip_global.cpp
+12 -4
View File
@@ -141,11 +141,19 @@ hipError_t Function::getStatFuncAttr(hipFuncAttributes* func_attr, int deviceId)
amd::Kernel* kernel = dFunc_[deviceId]->kernel();
const device::Kernel::WorkGroupInfo* wginfo = kernel->getDeviceKernel(*devices[deviceId])->workGroupInfo();
func_attr->sharedSizeBytes = static_cast<int>(wginfo->localMemSize_);
func_attr->binaryVersion = static_cast<int>(kernel->signature().version());
func_attr->cacheModeCA = 0;
func_attr->constSizeBytes = 0;
func_attr->localSizeBytes = wginfo->privateMemSize_;
func_attr->sharedSizeBytes = wginfo->localMemSize_;
func_attr->maxDynamicSharedSizeBytes = wginfo->availableLDSSize_ - wginfo->localMemSize_;
func_attr->maxThreadsPerBlock = wginfo->size_;
func_attr->numRegs = wginfo->usedVGPRs_;
func_attr->maxDynamicSharedSizeBytes = static_cast<int>(wginfo->availableLDSSize_
- wginfo->localMemSize_);
func_attr->maxThreadsPerBlock = static_cast<int>(wginfo->size_);
func_attr->numRegs = static_cast<int>(wginfo->usedVGPRs_);
func_attr->preferredShmemCarveout = 0;
func_attr->ptxVersion = 30;
return hipSuccess;
}