Improve allreduce performance when we have more than one network interface per
GPU and we need to use PXN to close rings.
Add support for PCI Gen5 on 5.4 kernels.
Fix crash when setting NCCL_SET_THREAD_NAME.
Fix random crash in init due to uninitialized struct.
Fix hang on cubemesh topologies.
Add P2P_DIRECT_DISABLE parameter to disable direct access to pointers within a
process.


[ROCm/rccl commit: 7aa1c46fd5]
This commit is contained in:
Sylvain Jeaugey
2022-05-03 01:30:26 -07:00
والد e89ff21d35
کامیت 1c5734046d
12فایلهای تغییر یافته به همراه94 افزوده شده و 50 حذف شده
@@ -165,8 +165,8 @@ ncclResult_t bootstrapCreateRoot(ncclUniqueId* id, bool idFromEnv) {
memcpy(id, &listenSock->addr, sizeof(union ncclSocketAddress));
pthread_t thread;
pthread_create(&thread, NULL, bootstrapRoot, (void*)listenSock);
pthread_detach(thread); // will not be pthread_join()'d
ncclSetThreadName(thread, "NCCL BootstrapR");
pthread_detach(thread); // will not be pthread_join()'d
return ncclSuccess;
}