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

Update NCCL tests

Browse Source
This commit is contained in:
Sylvain Jeaugey
2022-09-20 02:21:36 -07:00
parent 749573f2d6
commit d313d20a26
15 changed files with 206 additions and 137 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/hypercube.cu
+2 -3
View File
@@ -1,5 +1,5 @@
/*************************************************************************
* Copyright (c) 2016-2021, NVIDIA CORPORATION. All rights reserved.
* Copyright (c) 2016-2022, NVIDIA CORPORATION. All rights reserved.
*
* See LICENSE.txt for license information
************************************************************************/
@@ -24,8 +24,7 @@ testResult_t HyperCubeInitData(struct threadArgs* args, ncclDataType_t type, ncc
int nranks = args->nProcs*args->nThreads*args->nGpus;
for (int i=0; i<args->nGpus; i++) {
int gpuid = args->localRank*args->nThreads*args->nGpus + args->thread*args->nGpus + i;
CUDACHECK(cudaSetDevice(gpuid));
CUDACHECK(cudaSetDevice(args->gpus[i]));
int rank = ((args->proc*args->nThreads + args->thread)*args->nGpus + i);
CUDACHECK(cudaMemset(args->recvbuffs[i], 0, args->expectedBytes));
void* data = in_place ? ((char*)args->recvbuffs[i])+rank*args->sendBytes : args->sendbuffs[i];