Improved LL tuning for multi-node jobs.
Improved bootstrap for large job scaling.
Fixed a hang during bootstrap due to socket reuse.
Added operation name to the COLL INFO logging.
Esse commit está contido em:
David Addison
2018-10-24 14:44:59 -07:00
commit b56650c7f5
13 arquivos alterados com 164 adições e 157 exclusões
-1
Ver Arquivo
@@ -13,6 +13,5 @@ ncclResult_t bootstrapCreateRoot(ncclUniqueId* commId, bool idFromEnv);
ncclResult_t bootstrapGetUniqueId(ncclUniqueId* out);
ncclResult_t bootstrapInit(ncclUniqueId* id, int rank, int nranks, void** commState);
ncclResult_t bootstrapAllGather(void* commState, void* allData, int size);
ncclResult_t bootstrapRingExchange(void* commState, void* prevNextData, int prev, int next, int size);
ncclResult_t bootstrapClose(void* commState);
#endif