Merge remote-tracking branch 'nccl/master' into 2.10.3

[ROCm/rccl commit: bf2339f93e]
2021-07-28 13:27:06 -07:00
parent d4bdf8fab7 58a9f19dc3
commit 4b89e98675
53 changed files with 3806 additions and 3023 deletions
@@ -82,6 +82,7 @@ struct ncclComm {
  int nRanks;  // number of GPUs in communicator
  int cudaDev; // my cuda device index
  int64_t busId;   // my PCI bus ID in int format
+  cpu_set_t cpuAffinity; // CPU affinity of the GPU

  int node;
  int nNodes;
@@ -162,11 +163,13 @@ struct ncclComm {
  struct ncclInfo* asyncOps;
  int asyncOpCount;
  size_t asyncTotalSize;
+  ssize_t channelSize;
  int lastChannel;
+  enum { ROUND_ROBIN, SHORTEST_QUEUE } asyncAllocMode;

  //list of async p2p operation queued in a group semantics
-  struct ncclP2Plist* p2pSends;
-  struct ncclP2Plist* p2pRecvs;
+  ncclP2Plist** p2pSends;
+  ncclP2Plist** p2pRecvs;
  int p2pSendCount;
  int p2pRecvCount;