graph: restrict maxChannels to 64 for multi-node and RCCL_ENABLE_INTRANET=1

[ROCm/rccl commit: 526cce9bf4]
Этот коммит содержится в:
Nusrat Islam
2024-06-06 07:56:41 -05:00
родитель bf35178250
Коммит 3cbf67715f
+4
Просмотреть файл
@@ -627,6 +627,10 @@ ncclResult_t ncclTopoPostset(struct ncclComm* comm, int* firstRanks, int* treePa
// Only use full MAXCHANNELS for gfx94x
int maxChannels = IsArchMatch(comm->topo->nodes[GPU].nodes[0].gpu.gcn, "gfx94") ? MAXCHANNELS : 2*CHANNEL_LIMIT;
if (graphs[NCCL_ALGO_RING]->nIntraChannels > 0 || comm->nNodes > 1) {
maxChannels = std::min(64, maxChannels);
}
// Duplicate ringPrev/ringNext for ncclBuildRing
if (nChannels <= maxChannels/2) memcpy(ringPrev+nChannels*nranks, ringPrev, nChannels*nranks*sizeof(int));
if (nChannels <= maxChannels/2) memcpy(ringNext+nChannels*nranks, ringNext, nChannels*nranks*sizeof(int));