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

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

Browse Source
This commit is contained in:
Wenkai Du
2023-04-25 15:38:04 -07:00
parent 36e453c61e 9b7d5edbfc
commit 53a1f91857
79 changed files with 4609 additions and 2880 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/graph/paths.cc
+2 -2
View File
@@ -486,7 +486,7 @@ ncclResult_t ncclTopoGetIntermediateRank(struct ncclTopoSystem* system, int rank
type = node->type;
}
if (type != GPU) {
WARN("Could not find intermediate GPU between GPU rank %d and NIC %d\n", rank, netDev);
WARN("Could not find intermediate GPU between GPU rank %d and NIC %d", rank, netDev);
return ncclInternalError;
}
*intermediateRank = node->gpu.rank[0];
@@ -802,6 +802,7 @@ static int nextPow2(int v) {
}
ncclResult_t ncclTopoComputeP2pChannels(struct ncclComm* comm) {
/* here we already honor comm->max/minCTAs for p2pnChannels. */
comm->p2pnChannels = std::min(comm->nChannels, (int)ncclParamMaxP2pNChannels());
comm->p2pnChannels = std::max(comm->p2pnChannels, (int)ncclParamMinP2pNChannels());
int minChannels = comm->p2pnChannels;
@@ -842,7 +843,6 @@ ncclResult_t ncclTopoComputeP2pChannels(struct ncclComm* comm) {
for (int b=1, mb=(comm->p2pnChannels>>1); b<comm->p2pnChannels; b<<=1, mb>>=1) if (c & b) mirror |= mb;
comm->p2pChannels[c] = mirror;
}
INFO(NCCL_INIT, "%d coll channels, %d p2p channels, %d p2p channels per peer", comm->nChannels, comm->p2pnChannels, comm->p2pnChannelsPerPeer);
return ncclSuccess;
}