Enable CPX mode for MI300X (#1259)
* graph: enable cpx mode for MI300X * graph: tune limits for cpx and cleanup
このコミットが含まれているのは:
+1
-1
@@ -115,7 +115,7 @@ static ncclResult_t ncclTopoSetPaths(struct ncclTopoNode* baseNode, struct ncclT
|
||||
}
|
||||
|
||||
static void printNodePaths(struct ncclTopoSystem* system, struct ncclTopoNode* node) {
|
||||
char line[1024];
|
||||
char line[2048];
|
||||
#ifdef ENABLE_TRACE
|
||||
INFO(NCCL_GRAPH, "Paths from %s/%lX :", topoNodeTypeStr[node->type], node->id);
|
||||
#else
|
||||
|
||||
+1
-1
@@ -295,7 +295,7 @@ static ncclResult_t ncclTopoPrintRec(struct ncclTopoNode* node, struct ncclTopoN
|
||||
|
||||
ncclResult_t ncclTopoPrint(struct ncclTopoSystem* s) {
|
||||
INFO(NCCL_GRAPH, "=== System : maxBw %2.1f totalBw %2.1f ===", s->maxBw, s->totalBw);
|
||||
char line[1024];
|
||||
char line[2048];
|
||||
for (int n=0; n<s->nodes[CPU].count; n++) NCCLCHECK(ncclTopoPrintRec(s->nodes[CPU].nodes+n, NULL, line, 0));
|
||||
INFO(NCCL_GRAPH, "==========================================");
|
||||
NCCLCHECK(ncclTopoPrintPaths(s));
|
||||
|
||||
+2
-1
@@ -94,7 +94,8 @@ struct ncclTopoLink {
|
||||
float bw;
|
||||
struct ncclTopoNode* remNode;
|
||||
};
|
||||
#define NCCL_TOPO_MAX_LINKS 32
|
||||
#define NCCL_TOPO_MAX_LINKS 64 //Changed the value from 32 to 64 for CPX mode
|
||||
|
||||
#define NCCL_TOPO_MAX_HOPS (NCCL_TOPO_MAX_NODES*NCCL_TOPO_NODE_TYPES)
|
||||
|
||||
struct ncclTopoLinkList {
|
||||
|
||||
+1
-1
@@ -17,7 +17,7 @@
|
||||
// A few constraints to make the implementation easy
|
||||
#define MAX_STR_LEN 255
|
||||
#define MAX_ATTR_COUNT 16
|
||||
#define MAX_SUBS 32
|
||||
#define MAX_SUBS 64 //Changed the value from 32 to 64 for CPX mode
|
||||
#define MAX_NODES 1024
|
||||
|
||||
#define NODE_TYPE_NONE 0
|
||||
|
||||
新しいイシューから参照
ユーザーをブロックする