AI/rocm-systems
2
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix min, max and avg (#745)

Browse Source
This commit is contained in:
Ziyue Yang authored and GitHub committed 2023-05-18 11:02:59 -07:00
1 parent edb2ee1a44
commit 11676267b5
4 files changed
+10 -38
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

src/include/msccl/msccl_kernel.h
+1 -11
View File
@@ -28,21 +28,11 @@ __global__ void MSCCL_KERNEL_ENTRY_NAME(devredop, type, proto)(struct ncclDevCom
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP_TYPE(devredop, double) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP_TYPE(devredop, rccl_bfloat16)
#define MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP_NOFLOAT(devredop) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP_TYPE(devredop, int8_t) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP_TYPE(devredop, uint8_t) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP_TYPE(devredop, int32_t) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP_TYPE(devredop, uint32_t) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP_TYPE(devredop, int64_t) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP_TYPE(devredop, uint64_t)
#define MSCCL_DECL_KERNEL_ENTRY_FUNC() \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP(Sum) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP(Prod) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP(Min) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP(Max) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP(PreMulSum) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP_NOFLOAT(SumPostDiv)
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP(Min)
MSCCL_DECL_KERNEL_ENTRY_FUNC()