Add tree algorithms for allreduce to improve performance at scale.
Add ncclCommAbort() and ncclCommGetAsyncError() to properly handle
network errors and be permit recover.
Detect initial CPU affinity and no longer escape it.


[ROCm/rccl commit: 1450d42675]
Bu işleme şunda yer alıyor:
Sylvain Jeaugey
2018-12-13 15:56:12 -08:00
ebeveyn 3a7974d4c4
işleme ffc55560a1
66 değiştirilmiş dosya ile 3746 ekleme ve 3251 silme
+13
Dosyayı Görüntüle
@@ -0,0 +1,13 @@
/*************************************************************************
* Copyright (c) 2015-2018, NVIDIA CORPORATION. All rights reserved.
*
* See LICENSE.txt for license information
************************************************************************/
#ifndef NCCL_TREES_H_
#define NCCL_TREES_H_
ncclResult_t ncclGetBtree(int nranks, int rank, int* u0, int* d1, int* d0);
ncclResult_t ncclGetDtree(int nranks, int rank, int* u0, int* d0_0, int* d0_1, int* u1, int* d1_0, int* d1_1);
#endif