Enable NcclGather and NcclBroadcast.

PiperOrigin-RevId: 355859246
Change-Id: I80831dbecb4858073c8bf61906fac9b651868bf9

tensorflow/core/common_runtime/collective_param_resolver_local.cc

@@ -59,13 +59,13 @@ namespace {
 const char* GetCollectiveName(const CollectiveParams* cp, bool nccl) {
   switch (cp->instance.type) {
     case BROADCAST_COLLECTIVE:
-      return "HierarchicalTreeBroadcast";
+      return nccl ? "NcclBroadcast" : "HierarchicalTreeBroadcast";
 
     case REDUCTION_COLLECTIVE:
       return nccl ? "NcclReduce" : "RingReduce";
 
     case GATHER_COLLECTIVE:
-      return "RingGather";
+      return nccl ? "NcclGather" : "RingGather";
 
     case PERMUTE_COLLECTIVE:
       return "Permute";

tensorflow/python/kernel_tests/collective_ops_test.py

@@ -18,6 +18,7 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 
+import os
 import threading
 import time
 
@@ -952,5 +953,6 @@ def _setup_context():
 
 
 if __name__ == '__main__':
+  os.environ['NCCL_DEBUG'] = 'INFO'
   v2_compat.enable_v2_behavior()
   test.main()