Add local user buffer registration for NVLink SHARP.
Add tuning plugin support.
Increase net API to v7 to allow for device-side packet reordering;
remove support for v4 plugins.
Add support for RoCE ECE.
Add support for C2C links.
Better detect SHM allocation failures to avoid crash with Bus Error.
Fix missing thread unlocks in bootstrap (Fixes #936).
Disable network flush by default on H100.
Move device code from src/collectives/device to src/device.


[ROCm/rccl commit: f9c3dc251e]
This commit is contained in:
Sylvain Jeaugey
2023-09-26 05:47:28 -07:00
والد 26b91b9dbb
کامیت 506d6c332c
108فایلهای تغییر یافته به همراه4749 افزوده شده و 2059 حذف شده
@@ -8,7 +8,7 @@
#define NCCL_GRAPH_H_
#include "nccl.h"
#include "devcomm.h"
#include "device.h"
#include <limits.h>
#include <stdlib.h>
#include <ctype.h>
@@ -112,6 +112,6 @@ ncclResult_t ncclTopoPostset(struct ncclComm* comm, int* firstRanks, int* treePa
ncclResult_t ncclTopoTuneModel(struct ncclComm* comm, int minCompCap, int maxCompCap, struct ncclTopoGraph** graphs);
#include "info.h"
ncclResult_t ncclTopoGetAlgoTime(struct ncclInfo* info, int algorithm, int protocol, int numPipeOps, float* time);
ncclResult_t ncclTopoGetAlgoTime(struct ncclInfo* info, int algorithm, int protocol, int numPipeOps, float* time, bool* backup = NULL);
#endif