Merge pull request #194 from wenkaidu/search
Fix incorrect next device ID in PCI ordered search
[ROCm/rccl commit: cf5070f6c0]
This commit is contained in:
@@ -345,7 +345,7 @@ ncclResult_t ncclTopoSearchRecGpu(struct ncclTopoSystem* system, struct ncclTopo
|
||||
int next[NCCL_TOPO_MAX_NODES];
|
||||
int count;
|
||||
if (forcedOrder == FORCED_ORDER_PCI) { // Try the PCI order
|
||||
next[0] = (busIdToCudaDev(gpu->id)+1)%system->nodes[GPU].count;
|
||||
next[0] = step+1;
|
||||
count = 1;
|
||||
} else if (forcedOrder == FORCED_ORDER_REPLAY) { // Try last channel order
|
||||
NCCLCHECK(ncclTopoReplayGetGpu(system, graph, step, next));
|
||||
|
||||
Fai riferimento in un nuovo problema
Block a user