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

[ROCm/rccl commit: 84081064a0]

projects/rccl/ext-tuner/example/nccl/tuner.h

@@ -39,13 +39,17 @@ typedef struct {
   const char* name;
 
   // Initializes tuner states.
-  // nRanks: number of ranks in current communicator. Each communicator initialize its own tuner.
-  // nNodes: number of nodes in current communicator.
-  // logFunction: a logFunction can be useful to integrate logging together with NCCL core.
-  ncclResult_t (*init)(size_t nRanks, size_t nNodes, ncclDebugLogger_t logFunction);
+  // Inputs:
+  //   - nRanks: number of ranks in current communicator. Each communicator initialize its own tuner.
+  //   - nNodes: number of nodes in current communicator.
+  //   - logFunction: a logFunction can be useful to integrate logging together with NCCL core.
+  // Outputs:
+  //   - context: tuner context object
+  ncclResult_t (*init)(size_t nRanks, size_t nNodes, ncclDebugLogger_t logFunction, void **context);
 
   // Gets info (algo, protocol, number of ctas and threads) for a given collective.
   // Inputs:
+  //   - context: tuner context object
   //   - collType: collective type , e.g., allreduce, allgather…
   //   - nBytes: collective size in bytes
   //   - collNetSupport: whether collnet supports this type
@@ -62,16 +66,17 @@ typedef struct {
   // Also, the plugin is allowed to not set any output, or set only the
   // algorithm and protocol, but not only the algorithm or only the protocol.
   // Unset fields will be set automatically by NCCL.
-  ncclResult_t (*getCollInfo)(ncclFunc_t collType, size_t nBytes,
+  ncclResult_t (*getCollInfo)(void* context, ncclFunc_t collType, size_t nBytes,
                               int collNetSupport, int nvlsSupport, int numPipeOps,
                               int *algorithm, int *protocol, int* nChannels);
 
   // Terminates the plugin and cleans up any resources that the plugin allocated.
-  ncclResult_t (*destroy)();
-} ncclTuner_v1_t;
+  // context: tuner context object
+  ncclResult_t (*destroy)(void* context);
+} ncclTuner_v2_t;
 
-typedef ncclTuner_v1_t ncclTuner_t;
+typedef ncclTuner_v2_t ncclTuner_t;
 
-#define NCCL_TUNER_PLUGIN_SYMBOL "ncclTunerPlugin_v1"
+#define NCCL_TUNER_PLUGIN_SYMBOL "ncclTunerPlugin_v2"
 
 #endif

projects/rccl/ext-tuner/example/plugin.c

@@ -86,7 +86,7 @@ ncclResult_t ncclTopoGetAlgoTime_Tuner(ncclFunc_t collType, int algorithm, int p
   return ncclSuccess;
 }
 
-__hidden ncclResult_t pluginInit(size_t nRanks, size_t nNodes, ncclDebugLogger_t logFunction) { 
+__hidden ncclResult_t pluginInit(size_t nRanks, size_t nNodes, ncclDebugLogger_t logFunction, void **context) { 
   if (nRanks <= 1) return ncclSuccess;
   int compCapIndex = HOPPER_COMPCAP_IDX;
   int index2 = nNodes <= 2 ? nNodes-1 : 2;
@@ -195,7 +195,7 @@ __hidden ncclResult_t pluginInit(size_t nRanks, size_t nNodes, ncclDebugLogger_t
   return ncclSuccess;
 }
 
-__hidden ncclResult_t pluginGetCollInfo(ncclFunc_t collType, size_t nBytes,
+__hidden ncclResult_t pluginGetCollInfo(void* context, ncclFunc_t collType, size_t nBytes,
                               int collNetSupport, int nvlsSupport, int numPipeOps,
                               int *algorithm, int *protocol, int* nChannels) {
                                 
@@ -222,11 +222,11 @@ __hidden ncclResult_t pluginGetCollInfo(ncclFunc_t collType, size_t nBytes,
   return ncclSuccess;
 }
 
-__hidden ncclResult_t pluginDestroy() { return ncclSuccess; }
+__hidden ncclResult_t pluginDestroy(void* context) { return ncclSuccess; }
 
 #define PLUGIN_NAME "Example"
 
-const ncclTuner_v1_t ncclTunerPlugin_v1 = {
+const ncclTuner_v2_t ncclTunerPlugin_v2 = {
   .name = PLUGIN_NAME,
   .init = pluginInit,
   .getCollInfo = pluginGetCollInfo,