AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

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

Browse Source 
* 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
This commit is contained in:
Mustafa Abduljabbar
2025-04-23 15:44:56 -04:00
committed by GitHub
parent 45e1c3f3e2
commit 82afb2bcfe
8 changed files with 110 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/enqueue.cc
+3 -3
View File
@@ -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*/
) {