Merge remote-tracking branch 'nccl/master' into develop

[ROCm/rccl commit: 3a919c1f49]
This commit is contained in:
Wenkai Du
2021-11-11 14:21:51 -08:00
51 changed files with 1774 additions and 1252 deletions
+7 -1
View File
@@ -52,10 +52,16 @@ ncclResult_t ArgsCheck(struct ncclInfo* info) {
}
if (info->coll == ncclFuncAllGather || info->coll == ncclFuncReduceScatter) info->nBytes *= info->comm->nRanks; // count is per rank
if (info->op < 0 || info->op >= ncclNumOps) {
if (info->op < 0 || ncclMaxRedOp < info->op) {
WARN("%s : invalid reduction operation %d", info->opName, info->op);
return ncclInvalidArgument;
}
int opIx = int(ncclUserRedOpMangle(info->comm, info->op)) - int(ncclNumOps);
if (ncclNumOps <= info->op &&
(info->comm->userRedOpCapacity <= opIx || info->comm->userRedOps[opIx].freeNext != -1)) {
WARN("%s : reduction operation %d unknown to this communicator", info->opName, info->op);
return ncclInvalidArgument;
}
if (info->comm->checkPointers) {
if (info->coll == ncclFuncSendRecv) {