Merge remote-tracking branch 'remotes/nccl/master' into rccl_2.5.6

[ROCm/rccl commit: 6648c81dc6]

projects/rccl/src/include/core.h

@@ -1,5 +1,6 @@
 /*************************************************************************
  * Copyright (c) 2015-2019, NVIDIA CORPORATION. All rights reserved.
+ * Modifications Copyright (c) 2019 Advanced Micro Devices, Inc. All rights reserved.
  *
  * See LICENSE.txt for license information
  ************************************************************************/
@@ -8,19 +9,11 @@
 #define NCCL_CORE_H_
 
 #include <pthread.h>
-#include <algorithm>
-#include "nccl.h"
-#include "debug.h"
-#include "checks.h"
-#include "alloc.h"
-#include "transport.h"
-#include "devcomm.h"
-#include "comm.h"
-#include "info.h"
-#include "argcheck.h"
-#include <cstdio>
 #include <unistd.h>
 #include <stdlib.h>
+#include <stdint.h>
+#include <algorithm> // For std::min/std::max
+#include "nccl.h"
 
 #ifdef PROFAPI
 #define NCCL_API(ret, func, args...)        \
@@ -38,10 +31,6 @@
     ret func(args)
 #endif // end PROFAPI
 
-int ncclCudaCompCap();
-ncclResult_t ncclNvlinkGpu(int* nvlink);
-int64_t ncclTreeThreshold();
-
 static __inline__ int ncclTypeSize(ncclDataType_t type) {
   switch (type) {
     case ncclInt8:
@@ -63,4 +52,22 @@ static __inline__ int ncclTypeSize(ncclDataType_t type) {
   }
 }
 
+#define NCCL_NUM_FUNCTIONS 5
+typedef enum { ncclCollBroadcast, ncclCollReduce, ncclCollAllGather, ncclCollReduceScatter, ncclCollAllReduce } ncclFunc_t;
+
+#define NCCL_NUM_ALGORITHMS 2 // Tree/Ring
+#define NCCL_ALGO_TREE 0
+#define NCCL_ALGO_RING 1
+
+#define NCCL_NUM_PROTOCOLS 3 // Simple/LL/LL128
+#define NCCL_PROTO_LL 0
+#define NCCL_PROTO_LL128 1
+#define NCCL_PROTO_SIMPLE 2
+
+#include "debug.h"
+#include "checks.h"
+#include "alloc.h"
+#include "utils.h"
+#include "param.h"
+
 #endif // end include guard