2.12.7-1

Add network communication through another GPU connected with NVLink
(PXN).
Add aggregation of messages coming from different local GPUs through
PXN and going to the same destination.
Add new v5 plugin API with grouped receives and tags.
Add compat for v4 plugins.
Add naming of NCCL threads to help debugging.
Fix NVLink detection and avoid data corruption when some NVLinks are
down.
Add support for Relaxed Ordering for IB.
Add profiling and timing infrastructure.


[ROCm/rccl commit: 3c223c105a]

projects/rccl/src/include/graph.h

@@ -1,5 +1,5 @@
 /*************************************************************************
- * Copyright (c) 2016-2021, NVIDIA CORPORATION. All rights reserved.
+ * Copyright (c) 2016-2022, NVIDIA CORPORATION. All rights reserved.
  *
  * See LICENSE.txt for license information
  ************************************************************************/
@@ -30,9 +30,12 @@ ncclResult_t ncclTopoComputeP2pChannels(struct ncclComm* comm);
 ncclResult_t ncclTopoGetNvbGpus(struct ncclTopoSystem* system, int rank, int* nranks, int** ranks);
 
 // Query topology
-ncclResult_t ncclTopoGetNetDev(struct ncclTopoSystem* system, int rank, struct ncclTopoGraph* graph, int channelId, int rr, int* net);
+ncclResult_t ncclTopoGetNetDev(struct ncclComm* comm, int rank, struct ncclTopoGraph* graph, int channelId, int peerRank, int* net, int* proxyRank);
 ncclResult_t ncclTopoCheckP2p(struct ncclTopoSystem* system, int64_t id1, int64_t id2, int* p2p, int *read, int* intermediateRank);
 ncclResult_t ncclTopoCheckGdr(struct ncclTopoSystem* topo, int64_t busId, int netDev, int read, int* useGdr);
+int ncclPxnDisable();
+ncclResult_t ncclTopoGetPxnRanks(struct ncclComm* comm, int** intermediateRanks, int* nranks);
+ncclResult_t ncclTopoGetLocalRank(struct ncclTopoSystem* system, int rank, int* localRank);
 
 // Find CPU affinity
 ncclResult_t ncclTopoGetCpuAffinity(struct ncclTopoSystem* system, int rank, cpu_set_t* affinity);
@@ -48,6 +51,7 @@ ncclResult_t ncclTopoGetCpuAffinity(struct ncclTopoSystem* system, int rank, cpu
 #define NCCL_TOPO_CPU_TYPE_YONGFENG 1
 ncclResult_t ncclTopoCpuType(struct ncclTopoSystem* system, int* arch, int* vendor, int* model);
 ncclResult_t ncclTopoGetNetCount(struct ncclTopoSystem* system, int* count);
+ncclResult_t ncclTopoGetLocalNet(struct ncclTopoSystem* system, int rank, int* id);
 
 #define NCCL_TOPO_MAX_NODES 256
 
@@ -70,6 +74,7 @@ struct ncclTopoGraph {
   int nChannels;
   float speedIntra;
   float speedInter;
+  float latencyInter;
   int typeIntra;
   int typeInter;
   int sameChannels;