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

Replace GPU agent with CPU agent properly for memory async copy API

Browse Source 
ihipStream_t::copySync use GPU agent in memory async copy API, even
if the src/dst memory does not belong to GPU, which cause the hsa
runtime to choose a slower copy engine.

SWDEV-95191

Change-Id: If3cab3d493c0c96ed63721cdcf28247a1193887c


[ROCm/clr commit: dc0a787984]
This commit is contained in:
Fan Cao
2016-06-28 15:14:55 -04:00
committed by Maneesh Gupta
parent cf3186cd63
commit eb0db4e549
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/clr/hipamd/src/hip_hcc.cpp
+2 -2
View File
@@ -1403,7 +1403,7 @@ void ihipStream_t::copySync(LockedAccessor_StreamCrit_t &crit, void* dst, const
void *devPtrSrc = srcPtrInfo._devicePointer; void *devPtrSrc = srcPtrInfo._devicePointer;
tprintf(DB_COPY1, "HSA Async_copy dst=%p src=%p sz=%zu\n", dst, src, sizeBytes); tprintf(DB_COPY1, "HSA Async_copy dst=%p src=%p sz=%zu\n", dst, src, sizeBytes);
hsa_status_t hsa_status = hsa_amd_memory_async_copy(dst, dstAgent, devPtrSrc, srcAgent, sizeBytes, depSignalCnt, depSignalCnt ? &depSignal:0x0, copyCompleteSignal); hsa_status_t hsa_status = hsa_amd_memory_async_copy(dst, dstAgent, devPtrSrc, g_cpu_agent, sizeBytes, depSignalCnt, depSignalCnt ? &depSignal:0x0, copyCompleteSignal);
// This is sync copy, so let's wait for copy right here: // This is sync copy, so let's wait for copy right here:
if (hsa_status == HSA_STATUS_SUCCESS) { if (hsa_status == HSA_STATUS_SUCCESS) {
@@ -1443,7 +1443,7 @@ void ihipStream_t::copySync(LockedAccessor_StreamCrit_t &crit, void* dst, const
void *devPtrDst = dstPtrInfo._devicePointer; void *devPtrDst = dstPtrInfo._devicePointer;
tprintf(DB_COPY1, "HSA Async_copy dst=%p src=%p sz=%zu\n", dst, src, sizeBytes); tprintf(DB_COPY1, "HSA Async_copy dst=%p src=%p sz=%zu\n", dst, src, sizeBytes);
hsa_status_t hsa_status = hsa_amd_memory_async_copy(devPtrDst, dstAgent, src, srcAgent, sizeBytes, depSignalCnt, depSignalCnt ? &depSignal:0x0, copyCompleteSignal); hsa_status_t hsa_status = hsa_amd_memory_async_copy(devPtrDst, g_cpu_agent, src, srcAgent, sizeBytes, depSignalCnt, depSignalCnt ? &depSignal:0x0, copyCompleteSignal);
// This is sync copy, so let's wait for copy right here: // This is sync copy, so let's wait for copy right here:
if (hsa_status == HSA_STATUS_SUCCESS) { if (hsa_status == HSA_STATUS_SUCCESS) {