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
2021-05-25 20:49:10 -07:00
parent 256403d4f0 3fec2fa5ee
commit e3abf1c2ec
14 changed files with 133 additions and 43 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/graph/paths.cc
+19
View File
@@ -30,6 +30,8 @@ static ncclResult_t getPath(struct ncclTopoSystem* system, struct ncclTopoNode*
return ncclInternalError;
}
NCCL_PARAM(NvbDisable, "NVB_DISABLE", 0);
static ncclResult_t ncclTopoSetPaths(struct ncclTopoNode* baseNode, struct ncclTopoSystem* system) {
if (baseNode->paths[baseNode->type] == NULL) {
NCCLCHECK(ncclCalloc(baseNode->paths+baseNode->type, system->nodes[baseNode->type].count));
@@ -597,3 +599,20 @@ ncclResult_t ncclTopoComputeP2pChannels(struct ncclComm* comm) {
INFO(NCCL_INIT, "%d coll channels, %d p2p channels, %d p2p channels per peer", comm->nChannels, comm->p2pnChannels, comm->p2pnChannelsPerPeer);
return ncclSuccess;
}
ncclResult_t ncclTopoGetNvbGpus(struct ncclTopoSystem* system, int rank, int* nranks, int** ranks) {
int ngpus = system->nodes[GPU].count;
NCCLCHECK(ncclCalloc(ranks, ngpus));
int nvbGpus = 0;
for (int g=0; g<ngpus; g++) {
struct ncclTopoNode* gpu = system->nodes[GPU].nodes+g;
if (gpu->gpu.rank != rank) continue;
for (int p=0; p<ngpus; p++) {
if (gpu->paths[GPU][p].type == PATH_NVB) {
(*ranks)[nvbGpus++] = system->nodes[GPU].nodes[p].gpu.rank;
}
}
}
*nranks = nvbGpus;
return ncclSuccess;
}