[rocprofiler-sdk] Add support for new RCCL API (#771)

* [rocprofiler-sdk] Add support for new RCCL API

Add support for `ncclAllReduceWithBias`

* Move func to be in sync with rccl header

projects/rocprofiler-sdk/source/include/rocprofiler-sdk/cxx/enum_string.hpp

@@ -1106,7 +1106,20 @@ ROCPROFILER_ENUM_LABEL(ROCPROFILER_RCCL_API_ID_mscclRunAlgo);
 ROCPROFILER_ENUM_LABEL(ROCPROFILER_RCCL_API_ID_mscclUnloadAlgo);
 ROCPROFILER_ENUM_LABEL(ROCPROFILER_RCCL_API_ID_ncclCommRegister);
 ROCPROFILER_ENUM_LABEL(ROCPROFILER_RCCL_API_ID_ncclCommDeregister);
+#if RCCL_API_TRACE_VERSION_PATCH >= 1
+ROCPROFILER_ENUM_LABEL(ROCPROFILER_RCCL_API_ID_ncclAllReduceWithBias);
+#endif
+
+#if RCCL_API_TRACE_VERSION_PATCH == 0
 static_assert(ROCPROFILER_RCCL_API_ID_LAST == 37);
+#elif RCCL_API_TRACE_VERSION_PATCH == 1
+static_assert(ROCPROFILER_RCCL_API_ID_LAST == 38);
+#else
+#    if !defined(ROCPROFILER_UNSAFE_NO_VERSION_CHECK) &&                                           \
+        (defined(ROCPROFILER_CI) && ROCPROFILER_CI > 0)
+static_assert(false, "Support for new RCCL_API_TRACE_VERSION_PATCH enumerations is required");
+#    endif
+#endif
 
 // rocprofiler_rocdecode_api_id_t
 ROCPROFILER_ENUM_INFO(rocprofiler_rocdecode_api_id_t, 0, ROCPROFILER_ROCDECODE_API_ID_LAST, false)

projects/rocprofiler-sdk/source/include/rocprofiler-sdk/rccl/api_args.h

@@ -317,7 +317,19 @@ typedef union rocprofiler_rccl_api_args_t
         ncclComm_t comm;
         void*      handle;
     } ncclCommDeregister;
-
+#if RCCL_API_TRACE_VERSION_PATCH >= 1
+    struct
+    {
+        const void*      sendbuff;
+        void*            recvbuff;
+        size_t           count;
+        ncclDataType_t   datatype;
+        ncclRedOp_t      op;
+        struct ncclComm* comm;
+        hipStream_t      stream;
+        const void*      acc;
+    } ncclAllReduceWithBias;
+#endif
 } rocprofiler_rccl_api_args_t;
 
 ROCPROFILER_EXTERN_C_FINI

projects/rocprofiler-sdk/source/include/rocprofiler-sdk/rccl/api_id.h

@@ -68,6 +68,8 @@ typedef enum rocprofiler_rccl_api_id_t  // NOLINT(performance-enum-size)
     ROCPROFILER_RCCL_API_ID_mscclUnloadAlgo,
     ROCPROFILER_RCCL_API_ID_ncclCommRegister,
     ROCPROFILER_RCCL_API_ID_ncclCommDeregister,
-
+#if RCCL_API_TRACE_VERSION_PATCH >= 1
+    ROCPROFILER_RCCL_API_ID_ncclAllReduceWithBias,
+#endif
     ROCPROFILER_RCCL_API_ID_LAST,
 } rocprofiler_rccl_api_id_t;

projects/rocprofiler-sdk/source/include/rocprofiler-sdk/rccl/details/api_trace.h

@@ -47,7 +47,7 @@
 #define RCCL_API_TRACE_VERSION_MAJOR 0
 
 // should be increased every time new members are added to existing dispatch tables
-#define RCCL_API_TRACE_VERSION_PATCH 0
+#define RCCL_API_TRACE_VERSION_PATCH 1
 
 #if !defined(RCCL_EXTERN_C_INIT)
 #    ifdef __cplusplus
@@ -81,6 +81,14 @@ typedef ncclResult_t (*ncclAllReduce_fn_t)(const void*      sendbuff,
                                            ncclRedOp_t      op,
                                            struct ncclComm* comm,
                                            hipStream_t      stream);
+typedef ncclResult_t (*ncclAllReduceWithBias_fn_t)(const void*      sendbuff,
+                                                   void*            recvbuff,
+                                                   size_t           count,
+                                                   ncclDataType_t   datatype,
+                                                   ncclRedOp_t      op,
+                                                   struct ncclComm* comm,
+                                                   hipStream_t      stream,
+                                                   const void*      acc);
 typedef ncclResult_t (*ncclAllToAll_fn_t)(const void*    sendbuff,
                                           void*          recvbuff,
                                           size_t         count,
@@ -264,7 +272,7 @@ typedef struct rcclApiFuncTable
     mscclUnloadAlgo_fn_t          mscclUnloadAlgo_fn;
     ncclCommRegister_fn_t         ncclCommRegister_fn;
     ncclCommDeregister_fn_t       ncclCommDeregister_fn;
-
+    ncclAllReduceWithBias_fn_t    ncclAllReduceWithBias_fn;
 } rcclApiFuncTable;
 
 RCCL_EXTERN_C_FINI