Commit 85897758 authored by Oliveira, Daniel's avatar Oliveira, Daniel
Browse files

[SWDEV-SWDEV-568366] fix: Nv4x 95% perf drops



Fix the memory coherency model with Navi 4x due to hipDeviceMallocUncached flag when allocating fine grained memory
Signed-off-by: default avatarOliveira, Daniel <daniel.oliveira@amd.com>
parent a7dfea3d
...@@ -819,7 +819,7 @@ namespace { ...@@ -819,7 +819,7 @@ namespace {
#if defined (__NVCC__) #if defined (__NVCC__)
return {ERR_FATAL, "Fine-grained GPU memory not supported on NVIDIA platform"}; return {ERR_FATAL, "Fine-grained GPU memory not supported on NVIDIA platform"};
#else #else
int flag = hipDeviceMallocUncached; int flag = hipDeviceMallocFinegrained;
ERR_CHECK(hipExtMallocWithFlags((void**)memPtr, numBytes, flag)); ERR_CHECK(hipExtMallocWithFlags((void**)memPtr, numBytes, flag));
#endif #endif
} else if (memType == MEM_MANAGED) { } else if (memType == MEM_MANAGED) {
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment