Fixes for NCCL_MAX_NCHANNELS and topo_expl (#398)
Этот коммит содержится в:
@@ -292,6 +292,7 @@ ncclResult_t ncclTopoPostset(struct ncclComm* comm, int* firstRanks, int* treePa
|
||||
|
||||
// Get number of channels after duplication
|
||||
nc *= comm->nChannels;
|
||||
nc = std::min((int)ncclMaxNchannels(), nc);
|
||||
// Duplication should be complete now
|
||||
nChannels = comm->nChannels = std::min(MAXCHANNELS,nChannels*2);
|
||||
|
||||
|
||||
@@ -404,7 +404,7 @@ ncclResult_t initTransportsRank_1(struct ncclComm* comm, struct allGather1Data_t
|
||||
|
||||
if ((comm->topo->type & RCCL_TOPO_4P2H_ROME) && (comm->topo->type & RCCL_TOPO_GDR_ALL)) {
|
||||
if (rcclParamP2pNetDisable() == 0) {
|
||||
STORE(comm->p2pNet, 1);
|
||||
comm->p2pNet = 1;
|
||||
INFO(NCCL_INIT, "RCCL enabled same node P2P over network");
|
||||
}
|
||||
else
|
||||
|
||||
Ссылка в новой задаче
Block a user