modifying the tuning table to improve the performance of allreduce for 8MB and 16MB for single-node MI300X (#1063)

src/graph/tuning.cc

@@ -362,6 +362,7 @@ ncclResult_t ncclTopoTuneModel(struct ncclComm* comm, int minCompCap, int maxCom
       if (coll == ncclFuncAllGather && a != NCCL_ALGO_RING) continue;
 
       for (int p=0; p<NCCL_NUM_PROTOCOLS; p++) {
+        if (a == NCCL_ALGO_TREE && p == NCCL_PROTO_SIMPLE && IsArchMatch(comm->topo->nodes[GPU].nodes[0].gpu.gcn, "gfx94") && comm->topo->nodes[GPU].count == comm->topo->nRanks) continue;
         if ((a == NCCL_ALGO_NVLS || a == NCCL_ALGO_NVLS_TREE) && p != NCCL_PROTO_SIMPLE) continue;
         int collnet = (a == NCCL_ALGO_COLLNET_DIRECT || a == NCCL_ALGO_COLLNET_CHAIN) ? 1 : 0;
         float bw = nNodes <= 2 || collnet ? graphs[a]->bwIntra : graphs[a]->bwInter;