diff options
Diffstat (limited to 'src/collectives/device/common_kernel.h')
| -rw-r--r-- | src/collectives/device/common_kernel.h | 6 |
1 files changed, 2 insertions, 4 deletions
diff --git a/src/collectives/device/common_kernel.h b/src/collectives/device/common_kernel.h index e1fb096..aa1e936 100644 --- a/src/collectives/device/common_kernel.h +++ b/src/collectives/device/common_kernel.h @@ -1,5 +1,5 @@ /************************************************************************* - * Copyright (c) 2015-2018, NVIDIA CORPORATION. All rights reserved. + * Copyright (c) 2015-2019, NVIDIA CORPORATION. All rights reserved. * * See LICENSE.txt for license information ************************************************************************/ @@ -7,7 +7,7 @@ #ifndef NCCL_COMMON_KERNEL_H_ #define NCCL_COMMON_KERNEL_H_ -#include "core.h" +#include "devcomm.h" #include <cstdio> #include <cstdint> @@ -263,8 +263,6 @@ __device__ __forceinline__ void ReduceCopyMulti(const int tid, const int nthread } } -#define WARP_SIZE 32 - template<class FUNC, typename T, int UNROLL, int MINSRCS, int MAXSRCS, int MINDSTS, int MAXDSTS> __device__ __forceinline__ void ReduceCopy128bMulti( const int w, const int nw, const int t, int nsrcs, const T* s[MAXSRCS], int ndsts, T* d[MAXDSTS], |