Skip to content

[SYCL][CUDA][HIP][libclc] Rename libclc NonUniform binops to KHR extensions #6279

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 3 commits into from
Jun 9, 2022
Merged

[SYCL][CUDA][HIP][libclc] Rename libclc NonUniform binops to KHR extensions #6279

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
261fde4
[SYCL][CUDA][HIP][libclc] Rename libclc NonUniform binops to KHR exte…
steffenlarsen Jun 9, 2022
893b592
Remove old comment
steffenlarsen Jun 9, 2022
429ee0a
Rename CLC subgroup collectives
steffenlarsen Jun 9, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
41 changes: 20 additions & 21 deletions libclc/amdgcn-amdhsa/libspirv/group/collectives.cl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -116,16 +116,16 @@ __CLC_SUBGROUP_COLLECTIVE(IAdd, __CLC_ADD, ulong, m, 0)
__CLC_SUBGROUP_COLLECTIVE(FAdd, __CLC_ADD, float, f, 0)
__CLC_SUBGROUP_COLLECTIVE(FAdd, __CLC_ADD, double, d, 0)

__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, char, a, 1)
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, uchar, h, 1)
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, short, s, 1)
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, ushort, t, 1)
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, int, i, 1)
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, uint, j, 1)
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, long, l, 1)
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, ulong, m, 1)
__CLC_SUBGROUP_COLLECTIVE(FMul, __CLC_MUL, float, f, 1)
__CLC_SUBGROUP_COLLECTIVE(FMul, __CLC_MUL, double, d, 1)
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, char, a, 1)
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, uchar, h, 1)
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, short, s, 1)
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, ushort, t, 1)
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, int, i, 1)
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, uint, j, 1)
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, long, l, 1)
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, ulong, m, 1)
__CLC_SUBGROUP_COLLECTIVE(FMulKHR, __CLC_MUL, float, f, 1)
__CLC_SUBGROUP_COLLECTIVE(FMulKHR, __CLC_MUL, double, d, 1)

__CLC_SUBGROUP_COLLECTIVE(SMin, __CLC_MIN, char, a, CHAR_MAX)
__CLC_SUBGROUP_COLLECTIVE(UMin, __CLC_MIN, uchar, h, UCHAR_MAX)
Expand Down Expand Up @@ -234,17 +234,16 @@ __CLC_GROUP_COLLECTIVE(IAdd, __CLC_ADD, ulong, 0)
__CLC_GROUP_COLLECTIVE(FAdd, __CLC_ADD, float, 0)
__CLC_GROUP_COLLECTIVE(FAdd, __CLC_ADD, double, 0)

// There is no Mul group op in SPIR-V, use non-uniform variant instead.
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, char, 1)
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, uchar, 1)
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, short, 1)
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, ushort, 1)
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, int, 1)
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, uint, 1)
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, long, 1)
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, ulong, 1)
__CLC_GROUP_COLLECTIVE(NonUniformFMul, FMul, __CLC_MUL, float, 1)
__CLC_GROUP_COLLECTIVE(NonUniformFMul, FMul, __CLC_MUL, double, 1)
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, char, 1)
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, uchar, 1)
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, short, 1)
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, ushort, 1)
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, int, 1)
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, uint, 1)
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, long, 1)
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, ulong, 1)
__CLC_GROUP_COLLECTIVE(FMulKHR, __CLC_MUL, float, 1)
__CLC_GROUP_COLLECTIVE(FMulKHR, __CLC_MUL, double, 1)

__CLC_GROUP_COLLECTIVE(SMin, __CLC_MIN, char, CHAR_MAX)
__CLC_GROUP_COLLECTIVE(UMin, __CLC_MIN, uchar, UCHAR_MAX)
Expand Down
69 changes: 34 additions & 35 deletions libclc/ptx-nvidiacl/libspirv/group/collectives.cl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -213,17 +213,17 @@ __CLC_SUBGROUP_COLLECTIVE(FAdd, __CLC_ADD, half, 0)
__CLC_SUBGROUP_COLLECTIVE(FAdd, __CLC_ADD, float, 0)
__CLC_SUBGROUP_COLLECTIVE(FAdd, __CLC_ADD, double, 0)

__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, char, 1)
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, uchar, 1)
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, short, 1)
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, ushort, 1)
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, int, 1)
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, uint, 1)
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, long, 1)
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, ulong, 1)
__CLC_SUBGROUP_COLLECTIVE(FMul, __CLC_MUL, half, 1)
__CLC_SUBGROUP_COLLECTIVE(FMul, __CLC_MUL, float, 1)
__CLC_SUBGROUP_COLLECTIVE(FMul, __CLC_MUL, double, 1)
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, char, 1)
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, uchar, 1)
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, short, 1)
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, ushort, 1)
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, int, 1)
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, uint, 1)
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, long, 1)
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, ulong, 1)
__CLC_SUBGROUP_COLLECTIVE(FMulKHR, __CLC_MUL, half, 1)
__CLC_SUBGROUP_COLLECTIVE(FMulKHR, __CLC_MUL, float, 1)
__CLC_SUBGROUP_COLLECTIVE(FMulKHR, __CLC_MUL, double, 1)

__CLC_SUBGROUP_COLLECTIVE(SMin, __CLC_MIN, char, CHAR_MAX)
__CLC_SUBGROUP_COLLECTIVE(UMin, __CLC_MIN, uchar, UCHAR_MAX)
Expand All @@ -249,12 +249,12 @@ __CLC_SUBGROUP_COLLECTIVE(FMax, __CLC_MAX, half, -HALF_MAX)
__CLC_SUBGROUP_COLLECTIVE(FMax, __CLC_MAX, float, -FLT_MAX)
__CLC_SUBGROUP_COLLECTIVE(FMax, __CLC_MAX, double, -DBL_MAX)

__CLC_SUBGROUP_COLLECTIVE_REDUX(NonUniformBitwiseAnd, __CLC_AND, and, uint, ~0)
__CLC_SUBGROUP_COLLECTIVE_REDUX(NonUniformBitwiseOr, __CLC_OR, or, uint, 0)
__CLC_SUBGROUP_COLLECTIVE_REDUX(NonUniformBitwiseXor, __CLC_XOR, xor, uint, 0)
__CLC_SUBGROUP_COLLECTIVE_REDUX(NonUniformBitwiseAnd, __CLC_AND, and, int, ~0)
__CLC_SUBGROUP_COLLECTIVE_REDUX(NonUniformBitwiseOr, __CLC_OR, or, int, 0)
__CLC_SUBGROUP_COLLECTIVE_REDUX(NonUniformBitwiseXor, __CLC_XOR, xor, int, 0)
__CLC_SUBGROUP_COLLECTIVE_REDUX(BitwiseAndKHR, __CLC_AND, and, uint, ~0)
__CLC_SUBGROUP_COLLECTIVE_REDUX(BitwiseOrKHR, __CLC_OR, or, uint, 0)
__CLC_SUBGROUP_COLLECTIVE_REDUX(BitwiseXorKHR, __CLC_XOR, xor, uint, 0)
__CLC_SUBGROUP_COLLECTIVE_REDUX(BitwiseAndKHR, __CLC_AND, and, int, ~0)
__CLC_SUBGROUP_COLLECTIVE_REDUX(BitwiseOrKHR, __CLC_OR, or, int, 0)
__CLC_SUBGROUP_COLLECTIVE_REDUX(BitwiseXorKHR, __CLC_XOR, xor, int, 0)

#undef __CLC_SUBGROUP_COLLECTIVE_BODY
#undef __CLC_SUBGROUP_COLLECTIVE
Expand Down Expand Up @@ -340,18 +340,17 @@ __CLC_GROUP_COLLECTIVE(FAdd, __CLC_ADD, half, 0)
__CLC_GROUP_COLLECTIVE(FAdd, __CLC_ADD, float, 0)
__CLC_GROUP_COLLECTIVE(FAdd, __CLC_ADD, double, 0)

// There is no Mul group op in SPIR-V, use non-uniform variant instead.
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, char, 1)
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, uchar, 1)
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, short, 1)
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, ushort, 1)
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, int, 1)
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, uint, 1)
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, long, 1)
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, ulong, 1)
__CLC_GROUP_COLLECTIVE(NonUniformFMul, FMul, __CLC_MUL, half, 1)
__CLC_GROUP_COLLECTIVE(NonUniformFMul, FMul, __CLC_MUL, float, 1)
__CLC_GROUP_COLLECTIVE(NonUniformFMul, FMul, __CLC_MUL, double, 1)
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, char, 1)
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, uchar, 1)
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, short, 1)
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, ushort, 1)
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, int, 1)
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, uint, 1)
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, long, 1)
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, ulong, 1)
__CLC_GROUP_COLLECTIVE(FMulKHR, __CLC_MUL, half, 1)
__CLC_GROUP_COLLECTIVE(FMulKHR, __CLC_MUL, float, 1)
__CLC_GROUP_COLLECTIVE(FMulKHR, __CLC_MUL, double, 1)

__CLC_GROUP_COLLECTIVE(SMin, __CLC_MIN, char, CHAR_MAX)
__CLC_GROUP_COLLECTIVE(UMin, __CLC_MIN, uchar, UCHAR_MAX)
Expand All @@ -377,12 +376,12 @@ __CLC_GROUP_COLLECTIVE(FMax, __CLC_MAX, half, -HALF_MAX)
__CLC_GROUP_COLLECTIVE(FMax, __CLC_MAX, float, -FLT_MAX)
__CLC_GROUP_COLLECTIVE(FMax, __CLC_MAX, double, -DBL_MAX)

__CLC_GROUP_COLLECTIVE(NonUniformBitwiseAnd, __CLC_AND, uint, ~0)
__CLC_GROUP_COLLECTIVE(NonUniformBitwiseOr, __CLC_OR, uint, 0)
__CLC_GROUP_COLLECTIVE(NonUniformBitwiseXor, __CLC_XOR, uint, 0)
__CLC_GROUP_COLLECTIVE(NonUniformBitwiseAnd, __CLC_AND, int, ~0)
__CLC_GROUP_COLLECTIVE(NonUniformBitwiseOr, __CLC_OR, int, 0)
__CLC_GROUP_COLLECTIVE(NonUniformBitwiseXor, __CLC_XOR, int, 0)
__CLC_GROUP_COLLECTIVE(BitwiseAndKHR, __CLC_AND, uint, ~0)
__CLC_GROUP_COLLECTIVE(BitwiseOrKHR, __CLC_OR, uint, 0)
__CLC_GROUP_COLLECTIVE(BitwiseXorKHR, __CLC_XOR, uint, 0)
__CLC_GROUP_COLLECTIVE(BitwiseAndKHR, __CLC_AND, int, ~0)
__CLC_GROUP_COLLECTIVE(BitwiseOrKHR, __CLC_OR, int, 0)
__CLC_GROUP_COLLECTIVE(BitwiseXorKHR, __CLC_XOR, int, 0)

// half requires additional mangled entry points
_CLC_DEF _CLC_CONVERGENT half _Z17__spirv_GroupFAddjjDF16_(uint scope, uint op,
Expand Down