recreated pr 914 to work with current develop branch (#979)

This commit is contained in:
akolliasAMD
2023-11-28 16:33:47 -07:00
committed by GitHub
parent c71bae1608
commit 56ce9ef05f
5 changed files with 42 additions and 7 deletions
+11 -1
View File
@@ -28,6 +28,14 @@ __global__ void MSCCL_KERNEL_ENTRY_NAME(devredop, type, proto, fullOps)(struct n
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP_TYPE(devredop, double, fullOps) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP_TYPE(devredop, rccl_bfloat16, fullOps)
#define MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP_NOFLOAT(devredop, fullOps) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP_TYPE(devredop, int8_t, fullOps) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP_TYPE(devredop, uint8_t, fullOps) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP_TYPE(devredop, int32_t, fullOps) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP_TYPE(devredop, uint32_t, fullOps) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP_TYPE(devredop, int64_t, fullOps) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP_TYPE(devredop, uint64_t, fullOps)
#define MSCCL_DECL_KERNEL_ENTRY_FUNC() \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP(Sum, false) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP(Prod, false) \
@@ -36,7 +44,9 @@ __global__ void MSCCL_KERNEL_ENTRY_NAME(devredop, type, proto, fullOps)(struct n
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP(Sum, true) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP(Prod, true) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP(Max, true) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP(Min, true)
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP(Min, true) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP(PreMulSum, true) \
MSCCL_DECL_KERNEL_ENTRY_FUNC_DEVREDOP_NOFLOAT(SumPostDiv, true)
MSCCL_DECL_KERNEL_ENTRY_FUNC()