Fix memory leaks.
Fix crash in bootstrap error case.
Fix Collnet clean-up issue.
Make PCI switch vendor/device optional for XML injection.
Add support for nvidia-peermem module.
This commit is contained in:
Sylvain Jeaugey
2021-04-26 14:24:50 -07:00
parent a46ea10583
commit ca8485b0d0
9 changed files with 35 additions and 20 deletions
+1
View File
@@ -760,6 +760,7 @@ static ncclResult_t initTransportsRank(struct ncclComm* comm, ncclUniqueId* comm
collNetSetupFail = 1;
}
}
free(heads);
// Verify CollNet setup across ranks
NCCLCHECK(ncclTransportCollNetCheck(comm, collNetSetupFail));
if (comm->collNetSupport) {