Implement NIC identification and remapping (#420)
* Add 1H16P GPU model * Implement NIC identification and remapping * Revert "Sort IB devices based on device name (#413)" This reverts commit2d0ed8dff6. * Fix permute and check order * Correction on IB speed reporting * Revert "Allow user to link layer with RCCL_IB_HCA_SKIP_LINK_LAYER (#361)" This reverts commitcaf5c9992a.
Этот коммит содержится в:
@@ -489,7 +489,7 @@ ncclResult_t initTransportsRank_1(struct ncclComm* comm, struct allGather1Data_t
|
||||
if (comm->topo->nodes[GPU].count == comm->topo->nRanks && (comm->topo->type & RCCL_TOPO_CR8G))
|
||||
allGather3Data[rank].nc = 4;
|
||||
if (comm->topo->nodes[GPU].count == comm->topo->nRanks && comm->topo->nodes[GPU].nodes[idx].gpu.gcn == 910)
|
||||
allGather3Data[rank].nc = 6;
|
||||
allGather3Data[rank].nc = 4;
|
||||
allGather3Data[rank].tree.pattern = treeGraph.pattern;
|
||||
allGather3Data[rank].tree.nChannels = treeGraph.nChannels;
|
||||
allGather3Data[rank].tree.sameChannels = treeGraph.sameChannels;
|
||||
|
||||
Ссылка в новой задаче
Block a user