AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

collnet: support multiple NICs (#335)

Browse Source 
[ROCm/rccl commit: d87dc7c2e8]
This commit is contained in:
Wenkai Du
2021-03-25 20:59:32 -07:00
committed by GitHub
parent e85a70a967
commit 065bde98d8
10 changed files with 209 additions and 36 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/rccl/tools/topo_expl/topo_expl.cpp
+6 -8
View File
@@ -120,6 +120,7 @@ NodeModelDesc model_descs[] = {
{1, "topo_8p_rome_4nics.xml", "single node 8 gfx908 Rome 4 NICs"},
{4, "topo_8p_rome_4nics.xml", "4 nodes node 8 gfx908 Rome 4 NICs"},
{4, "topo_collnet_n1.xml", "4 nodes collnet 1 NICs"},
{4, "topo_collnet_n4.xml", "4 nodes collnet 4 NICs"},
};
int main(int argc,char* argv[])
@@ -148,6 +149,8 @@ int main(int argc,char* argv[])
NetworkModel network;
NodeModel* node;
initCollNet();
NodeModelDesc *desc = &model_descs[model_id];
for (int i=0; i<desc->num_nodes; i++) {
node = new NodeModel(desc->filename);
@@ -190,14 +193,9 @@ int main(int argc,char* argv[])
}
struct ncclTopoGraph *treeGraph, *ringGraph, *collNetGraph;
treeGraph = (struct ncclTopoGraph *)malloc(sizeof(struct ncclTopoGraph)*nranks);
ringGraph = (struct ncclTopoGraph *)malloc(sizeof(struct ncclTopoGraph)*nranks);
collNetGraph = (struct ncclTopoGraph *)malloc(sizeof(struct ncclTopoGraph)*nranks);
if (!treeGraph || !ringGraph || !collNetGraph) {
printf("Failed to allocate memory for graphs\n");
return -1;
}
NCCLCHECK(ncclCalloc(&treeGraph, nranks));
NCCLCHECK(ncclCalloc(&ringGraph, nranks));
NCCLCHECK(ncclCalloc(&collNetGraph, nranks));
for (int i = 0; i < nranks; i++) {
node_model = network.GetNode(i);
assert(node_model!=0);