Merge branch 'develop' into 2.13.4

src/include/devcomm.h

@@ -289,6 +289,7 @@ struct ncclDevChannelPeer {
   // instead of the full ncclConnector.
   struct ncclConnInfo send[NCCL_MAX_CONNS];
   struct ncclConnInfo recv[NCCL_MAX_CONNS];
+
 };
 #pragma pack(pop)   /* restore original alignment from stack */
 
@@ -354,6 +355,7 @@ struct alignas(16) ncclDevChannel {
   struct ncclDevChannelPeer *peers;
   struct ncclRing ring;
   struct ncclTree tree;
+  struct ncclTree binTree;
   struct ncclDirect collTree;
   uint32_t* workFifoDone; // Location of done counter, device writes index+1 of last work processed
 };

src/include/graph.h

@@ -115,6 +115,8 @@ ncclResult_t ncclTopoPostset(struct ncclComm* comm, int* firstRanks, int* treePa
 
 ncclResult_t ncclTreeBasePostset(struct ncclComm* comm, struct ncclTopoGraph* treeGraph);
 
+ncclResult_t ncclBinaryTreePostset(struct ncclComm* comm, struct ncclTopoGraph* treeGraph);
+
 ncclResult_t ncclTopoTuneModel(struct ncclComm* comm, int minCompCap, int maxCompCap, struct ncclTopoGraph* treeGraph, struct ncclTopoGraph* ringGraph, struct ncclTopoGraph* collNetGraph);
 #include "info.h"
 ncclResult_t ncclTopoGetAlgoTime(struct ncclInfo* info, int algorithm, int protocol, int numPipeOps, float* time);