Moved mscclpp_ncclGetUniqueId call into ncclCommInitRankFunc (#1332)

* Moved call to `mscclpp_ncclGetUniqueId` into `ncclCommInitRankFunc` to avoid setting up transport early in environments where MSCCL++ isn't valid.

* Checking `mscclEnabled` for the process and the topology to gate MSCCL++.

* Allowed `mscclForceEnable` to enable MSCCL++.
This commit is contained in:
corey-derochie-amd
2024-09-16 16:41:40 -06:00
committed by GitHub
parent ad94c651ad
commit 853a0586b4
3 changed files with 38 additions and 33 deletions
+6
View File
@@ -84,6 +84,12 @@ static bool mscclCommCompatible(ncclComm_t comm) {
return true;
}
#ifdef ENABLE_MSCCLPP
bool mscclppCommCompatible(ncclComm_t comm) {
return mscclCommCompatible(comm);
}
#endif
const char *mscclFuncNames[] = {
"mscclFuncReduce",
"mscclFuncBroadcast",