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

Merge pull request #1022 from PedramAlizadeh/sync_nccl_2.18.6

Browse Source 
Sync to nccl 2.18.6
This commit is contained in:
Pedram Alizadeh authored and GitHub committed 2024-01-09 13:29:29 -05:00
2 parents d9871d171b 0d515f9388
commit aa5c84c997
12 files changed
+107 -130
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

src/init.cc
+1 -1
View File
@@ -672,7 +672,7 @@ static ncclResult_t devCommSetup(ncclComm_t comm) {
NCCLCHECKGOTO(ncclCudaMemcpyAsync(devCommAndChans, &tmpCommAndChans, 1, comm->sharedRes->deviceStream.cudaStream), ret, fail);
exit:
CUDACHECK(cudaStreamSynchronize(comm->sharedRes->deviceStream.cudaStream));
NCCLCHECK(ncclStrongStreamSynchronize(&comm->sharedRes->deviceStream));
NCCLCHECK(ncclStrongStreamRelease(ncclCudaGraphNone(), &comm->sharedRes->deviceStream));
return ret;
fail: