Expose production tuning table in topo_explorer using internal RCCL/NCCL logic (#1628)

* Internal RCCL/NCCL functionality exposed when RCCL_EXPOSE_STATIC is enabled
* Algo/protocol/max channels can be obtained with the new RCCL API
* Introduce rccl_static and rccl_static_inline macros to work around invisible functions in core source files like enqueue.cc
* Add usage example in topo-explorer tool

src/device/common.h

@@ -14,7 +14,6 @@
 #include "reduce_kernel.h"
 #include "device_table.h"
 #include "network/unpack/unpack_defs.h"
-
 #define NCCL_MAX_DEV_ARITY (NCCL_MAX_TREE_ARITY-1)  // Using balanced tree instead of split tree
 
 #define __syncwarp()

src/enqueue.cc

@@ -111,7 +111,7 @@ static inline size_t ncclFuncSendCount(ncclFunc_t func, int nRanks, size_t count
 static inline size_t ncclFuncRecvCount(ncclFunc_t func, int nRanks, size_t count) {
   return func == ncclFuncAllGather ? nRanks*count : count;
 }
-static inline size_t ncclFuncMaxSendRecvCount(ncclFunc_t func, int nRanks, size_t count) {
+rccl_static_inline size_t ncclFuncMaxSendRecvCount(ncclFunc_t func, int nRanks, size_t count) {
   return func == ncclFuncAllGather || func == ncclFuncReduceScatter ? nRanks*count : count;
 }
 
@@ -569,7 +569,7 @@ static ncclResult_t registerP2pBuffer(struct ncclComm* comm, void* userbuff, int
 }
 
 static ncclResult_t getCollNetSupport(struct ncclComm* comm, struct ncclTaskColl* task, int* collNetSupport);
-static ncclResult_t getAlgoInfo(
+rccl_static ncclResult_t getAlgoInfo(
   struct ncclComm* comm, struct ncclTaskColl* task,
   int collNetSupport, int nvlsSupport, int numPipeOps, ncclSimInfo_t* simInfo = NULL
 );
@@ -2022,7 +2022,7 @@ static ncclResult_t topoGetAlgoInfo(
 // Call the plugin first. Let it set algo+proto, and/or nChannels.
 // Then, topoGetAlgoInfo will set algo/proto if not set, then nChannels and nThreads based on algo/proto.
 // Finally, nChannels will be overriden by the plugin setting.
-static ncclResult_t getAlgoInfo(
+rccl_static ncclResult_t getAlgoInfo(
     struct ncclComm* comm, struct ncclTaskColl* info,
     int collNetSupport, int nvlsSupport, int numPipeOps, ncclSimInfo_t* simInfo/* = NULL*/
   ) {

src/include/rccl_common.h

@@ -22,7 +22,7 @@ THE SOFTWARE.
 #ifndef RCCL_COMMON_H_
 #define RCCL_COMMON_H_
 #include "nccl_common.h"
-
+#include "nccl.h"
 typedef enum RcclTunableColls {
   RCCL_UNSUPPORTED_TUNABLE = -1,
   RCCL_RS_TUNABLE = 0,    // reduce_scatter index
@@ -38,6 +38,16 @@ typedef enum RcclTunableColls {
 #define RCCL_PROTOCOL_MAX_IDX 1
 #define RCCL_PROTOCOL_FACTOR_IDX 2
 
+#ifdef RCCL_EXPOSE_STATIC
+#define RCCL_STATIC_EXPOSE_CHECK()
+#else
+#define RCCL_STATIC_EXPOSE_CHECK() \
+  do { \
+    WARN("Attempting to use internal logic while required static functions are not exposed. Rebuild with RCCL_EXPOSE_STATIC enabled"); \
+    return ncclInvalidUsage; \
+  } while (0)
+#endif
+
 inline rcclTunableIndex_t rcclGetTunableIndex(ncclFunc_t const& func) {
   switch (func) {
     case ncclFuncReduceScatter:
@@ -61,4 +71,12 @@ inline size_t rcclGetSizePerRank(ncclFunc_t const& func, size_t const& nBytes, i
   return (func == ncclFuncReduceScatter || func == ncclFuncAllGather) ? nBytes / nRanks : nBytes;
 }
 void rcclUpdateCollectiveProtocol(struct ncclComm* comm, size_t const& nBytes, struct ncclTaskColl* info);
+
+
+ncclResult_t rcclGetAlgoInfo(struct ncclComm* comm, ncclFunc_t coll, uint64_t count, ncclDataType_t dataType,
+                             int collNetSupport, int nvlsSupport, int numPipeOps,
+                             int* algo, int* protocol, int* maxChannels);
+
+ncclResult_t rcclFuncMaxSendRecvCount(ncclFunc_t func, int nRanks, size_t count, size_t& maxCount);
+
 #endif

src/include/rccl_vars.h

@@ -27,4 +27,12 @@ THE SOFTWARE.
 
 RCCL_PARAM_DECLARE(EnableHipGraph);  // Opt-in environment variable for enabling hipGraph
 
+#ifdef RCCL_EXPOSE_STATIC
+#define rccl_static
+#define rccl_static_inline
+#else
+#define rccl_static static
+#define rccl_static_inline static inline
+#endif
+
 #endif

src/rccl_wrap.cc

@@ -71,3 +71,31 @@ void rcclUpdateCollectiveProtocol(struct ncclComm* comm, size_t const& nBytes, s
     }
   }
 }
+
+extern size_t ncclFuncMaxSendRecvCount(ncclFunc_t func, int nRanks, size_t count);
+extern ncclResult_t getAlgoInfo(
+    struct ncclComm* comm, struct ncclTaskColl* task,
+    int collNetSupport, int nvlsSupport, int numPipeOps, ncclSimInfo_t* simInfo = NULL
+);
+
+ncclResult_t rcclGetAlgoInfo(struct ncclComm* comm, ncclFunc_t coll, uint64_t count, ncclDataType_t dataType,
+                             int collNetSupport, int nvlsSupport, int numPipeOps,
+                             int* algo, int* protocol, int* maxChannels) {
+  RCCL_STATIC_EXPOSE_CHECK();
+  struct ncclTaskColl task;
+  task.func = coll;
+  task.count = count;
+  task.datatype = dataType;
+  NCCLCHECK(getAlgoInfo(comm, &task, collNetSupport, nvlsSupport, numPipeOps));
+  *algo = task.algorithm;
+  *protocol = task.protocol;
+  *maxChannels = task.nMaxChannels;
+  return ncclSuccess;
+}
+
+
+ncclResult_t rcclFuncMaxSendRecvCount(ncclFunc_t func, int nRanks, size_t count, size_t& maxCount) {
+  RCCL_STATIC_EXPOSE_CHECK();
+  maxCount = ncclFuncMaxSendRecvCount(func, nRanks, count);
+  return ncclSuccess;
+}