Skip to content

Commit 10c19ae

Browse files
steffenlarsensteffenlarsen
authored
[SYCL][CUDA][HIP][libclc] Rename libclc NonUniform binops to KHR extensions (#6279)
#5705 changed the use of NonUniform group binary operators to the KHR extension operators. These changes rename the implementation of these in libclc for both amdgcn and PTX to the new naming. Signed-off-by: Larsen, Steffen <[email protected]>
1 parent fe2b28e commit 10c19ae

File tree

2 files changed

+54
-56
lines changed

2 files changed

+54
-56
lines changed

‎libclc/amdgcn-amdhsa/libspirv/group/collectives.cl

Lines changed: 20 additions & 21 deletions
Original file line numberDiff line numberDiff line change
@@ -116,16 +116,16 @@ __CLC_SUBGROUP_COLLECTIVE(IAdd, __CLC_ADD, ulong, m, 0)
116116
__CLC_SUBGROUP_COLLECTIVE(FAdd, __CLC_ADD, float, f, 0)
117117
__CLC_SUBGROUP_COLLECTIVE(FAdd, __CLC_ADD, double, d, 0)
118118

119-
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, char, a, 1)
120-
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, uchar, h, 1)
121-
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, short, s, 1)
122-
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, ushort, t, 1)
123-
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, int, i, 1)
124-
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, uint, j, 1)
125-
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, long, l, 1)
126-
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, ulong, m, 1)
127-
__CLC_SUBGROUP_COLLECTIVE(FMul, __CLC_MUL, float, f, 1)
128-
__CLC_SUBGROUP_COLLECTIVE(FMul, __CLC_MUL, double, d, 1)
119+
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, char, a, 1)
120+
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, uchar, h, 1)
121+
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, short, s, 1)
122+
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, ushort, t, 1)
123+
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, int, i, 1)
124+
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, uint, j, 1)
125+
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, long, l, 1)
126+
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, ulong, m, 1)
127+
__CLC_SUBGROUP_COLLECTIVE(FMulKHR, __CLC_MUL, float, f, 1)
128+
__CLC_SUBGROUP_COLLECTIVE(FMulKHR, __CLC_MUL, double, d, 1)
129129

130130
__CLC_SUBGROUP_COLLECTIVE(SMin, __CLC_MIN, char, a, CHAR_MAX)
131131
__CLC_SUBGROUP_COLLECTIVE(UMin, __CLC_MIN, uchar, h, UCHAR_MAX)
@@ -234,17 +234,16 @@ __CLC_GROUP_COLLECTIVE(IAdd, __CLC_ADD, ulong, 0)
234234
__CLC_GROUP_COLLECTIVE(FAdd, __CLC_ADD, float, 0)
235235
__CLC_GROUP_COLLECTIVE(FAdd, __CLC_ADD, double, 0)
236236

237-
// There is no Mul group op in SPIR-V, use non-uniform variant instead.
238-
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, char, 1)
239-
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, uchar, 1)
240-
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, short, 1)
241-
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, ushort, 1)
242-
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, int, 1)
243-
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, uint, 1)
244-
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, long, 1)
245-
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, ulong, 1)
246-
__CLC_GROUP_COLLECTIVE(NonUniformFMul, FMul, __CLC_MUL, float, 1)
247-
__CLC_GROUP_COLLECTIVE(NonUniformFMul, FMul, __CLC_MUL, double, 1)
237+
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, char, 1)
238+
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, uchar, 1)
239+
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, short, 1)
240+
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, ushort, 1)
241+
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, int, 1)
242+
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, uint, 1)
243+
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, long, 1)
244+
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, ulong, 1)
245+
__CLC_GROUP_COLLECTIVE(FMulKHR, __CLC_MUL, float, 1)
246+
__CLC_GROUP_COLLECTIVE(FMulKHR, __CLC_MUL, double, 1)
248247

249248
__CLC_GROUP_COLLECTIVE(SMin, __CLC_MIN, char, CHAR_MAX)
250249
__CLC_GROUP_COLLECTIVE(UMin, __CLC_MIN, uchar, UCHAR_MAX)

‎libclc/ptx-nvidiacl/libspirv/group/collectives.cl

Lines changed: 34 additions & 35 deletions
Original file line numberDiff line numberDiff line change
@@ -213,17 +213,17 @@ __CLC_SUBGROUP_COLLECTIVE(FAdd, __CLC_ADD, half, 0)
213213
__CLC_SUBGROUP_COLLECTIVE(FAdd, __CLC_ADD, float, 0)
214214
__CLC_SUBGROUP_COLLECTIVE(FAdd, __CLC_ADD, double, 0)
215215

216-
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, char, 1)
217-
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, uchar, 1)
218-
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, short, 1)
219-
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, ushort, 1)
220-
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, int, 1)
221-
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, uint, 1)
222-
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, long, 1)
223-
__CLC_SUBGROUP_COLLECTIVE(IMul, __CLC_MUL, ulong, 1)
224-
__CLC_SUBGROUP_COLLECTIVE(FMul, __CLC_MUL, half, 1)
225-
__CLC_SUBGROUP_COLLECTIVE(FMul, __CLC_MUL, float, 1)
226-
__CLC_SUBGROUP_COLLECTIVE(FMul, __CLC_MUL, double, 1)
216+
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, char, 1)
217+
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, uchar, 1)
218+
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, short, 1)
219+
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, ushort, 1)
220+
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, int, 1)
221+
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, uint, 1)
222+
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, long, 1)
223+
__CLC_SUBGROUP_COLLECTIVE(IMulKHR, __CLC_MUL, ulong, 1)
224+
__CLC_SUBGROUP_COLLECTIVE(FMulKHR, __CLC_MUL, half, 1)
225+
__CLC_SUBGROUP_COLLECTIVE(FMulKHR, __CLC_MUL, float, 1)
226+
__CLC_SUBGROUP_COLLECTIVE(FMulKHR, __CLC_MUL, double, 1)
227227

228228
__CLC_SUBGROUP_COLLECTIVE(SMin, __CLC_MIN, char, CHAR_MAX)
229229
__CLC_SUBGROUP_COLLECTIVE(UMin, __CLC_MIN, uchar, UCHAR_MAX)
@@ -249,12 +249,12 @@ __CLC_SUBGROUP_COLLECTIVE(FMax, __CLC_MAX, half, -HALF_MAX)
249249
__CLC_SUBGROUP_COLLECTIVE(FMax, __CLC_MAX, float, -FLT_MAX)
250250
__CLC_SUBGROUP_COLLECTIVE(FMax, __CLC_MAX, double, -DBL_MAX)
251251

252-
__CLC_SUBGROUP_COLLECTIVE_REDUX(NonUniformBitwiseAnd, __CLC_AND, and, uint, ~0)
253-
__CLC_SUBGROUP_COLLECTIVE_REDUX(NonUniformBitwiseOr, __CLC_OR, or, uint, 0)
254-
__CLC_SUBGROUP_COLLECTIVE_REDUX(NonUniformBitwiseXor, __CLC_XOR, xor, uint, 0)
255-
__CLC_SUBGROUP_COLLECTIVE_REDUX(NonUniformBitwiseAnd, __CLC_AND, and, int, ~0)
256-
__CLC_SUBGROUP_COLLECTIVE_REDUX(NonUniformBitwiseOr, __CLC_OR, or, int, 0)
257-
__CLC_SUBGROUP_COLLECTIVE_REDUX(NonUniformBitwiseXor, __CLC_XOR, xor, int, 0)
252+
__CLC_SUBGROUP_COLLECTIVE_REDUX(BitwiseAndKHR, __CLC_AND, and, uint, ~0)
253+
__CLC_SUBGROUP_COLLECTIVE_REDUX(BitwiseOrKHR, __CLC_OR, or, uint, 0)
254+
__CLC_SUBGROUP_COLLECTIVE_REDUX(BitwiseXorKHR, __CLC_XOR, xor, uint, 0)
255+
__CLC_SUBGROUP_COLLECTIVE_REDUX(BitwiseAndKHR, __CLC_AND, and, int, ~0)
256+
__CLC_SUBGROUP_COLLECTIVE_REDUX(BitwiseOrKHR, __CLC_OR, or, int, 0)
257+
__CLC_SUBGROUP_COLLECTIVE_REDUX(BitwiseXorKHR, __CLC_XOR, xor, int, 0)
258258

259259
#undef __CLC_SUBGROUP_COLLECTIVE_BODY
260260
#undef __CLC_SUBGROUP_COLLECTIVE
@@ -340,18 +340,17 @@ __CLC_GROUP_COLLECTIVE(FAdd, __CLC_ADD, half, 0)
340340
__CLC_GROUP_COLLECTIVE(FAdd, __CLC_ADD, float, 0)
341341
__CLC_GROUP_COLLECTIVE(FAdd, __CLC_ADD, double, 0)
342342

343-
// There is no Mul group op in SPIR-V, use non-uniform variant instead.
344-
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, char, 1)
345-
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, uchar, 1)
346-
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, short, 1)
347-
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, ushort, 1)
348-
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, int, 1)
349-
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, uint, 1)
350-
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, long, 1)
351-
__CLC_GROUP_COLLECTIVE(NonUniformIMul, IMul, __CLC_MUL, ulong, 1)
352-
__CLC_GROUP_COLLECTIVE(NonUniformFMul, FMul, __CLC_MUL, half, 1)
353-
__CLC_GROUP_COLLECTIVE(NonUniformFMul, FMul, __CLC_MUL, float, 1)
354-
__CLC_GROUP_COLLECTIVE(NonUniformFMul, FMul, __CLC_MUL, double, 1)
343+
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, char, 1)
344+
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, uchar, 1)
345+
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, short, 1)
346+
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, ushort, 1)
347+
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, int, 1)
348+
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, uint, 1)
349+
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, long, 1)
350+
__CLC_GROUP_COLLECTIVE(IMulKHR, __CLC_MUL, ulong, 1)
351+
__CLC_GROUP_COLLECTIVE(FMulKHR, __CLC_MUL, half, 1)
352+
__CLC_GROUP_COLLECTIVE(FMulKHR, __CLC_MUL, float, 1)
353+
__CLC_GROUP_COLLECTIVE(FMulKHR, __CLC_MUL, double, 1)
355354

356355
__CLC_GROUP_COLLECTIVE(SMin, __CLC_MIN, char, CHAR_MAX)
357356
__CLC_GROUP_COLLECTIVE(UMin, __CLC_MIN, uchar, UCHAR_MAX)
@@ -377,12 +376,12 @@ __CLC_GROUP_COLLECTIVE(FMax, __CLC_MAX, half, -HALF_MAX)
377376
__CLC_GROUP_COLLECTIVE(FMax, __CLC_MAX, float, -FLT_MAX)
378377
__CLC_GROUP_COLLECTIVE(FMax, __CLC_MAX, double, -DBL_MAX)
379378

380-
__CLC_GROUP_COLLECTIVE(NonUniformBitwiseAnd, __CLC_AND, uint, ~0)
381-
__CLC_GROUP_COLLECTIVE(NonUniformBitwiseOr, __CLC_OR, uint, 0)
382-
__CLC_GROUP_COLLECTIVE(NonUniformBitwiseXor, __CLC_XOR, uint, 0)
383-
__CLC_GROUP_COLLECTIVE(NonUniformBitwiseAnd, __CLC_AND, int, ~0)
384-
__CLC_GROUP_COLLECTIVE(NonUniformBitwiseOr, __CLC_OR, int, 0)
385-
__CLC_GROUP_COLLECTIVE(NonUniformBitwiseXor, __CLC_XOR, int, 0)
379+
__CLC_GROUP_COLLECTIVE(BitwiseAndKHR, __CLC_AND, uint, ~0)
380+
__CLC_GROUP_COLLECTIVE(BitwiseOrKHR, __CLC_OR, uint, 0)
381+
__CLC_GROUP_COLLECTIVE(BitwiseXorKHR, __CLC_XOR, uint, 0)
382+
__CLC_GROUP_COLLECTIVE(BitwiseAndKHR, __CLC_AND, int, ~0)
383+
__CLC_GROUP_COLLECTIVE(BitwiseOrKHR, __CLC_OR, int, 0)
384+
__CLC_GROUP_COLLECTIVE(BitwiseXorKHR, __CLC_XOR, int, 0)
386385

387386
// half requires additional mangled entry points
388387
_CLC_DEF _CLC_CONVERGENT half _Z17__spirv_GroupFAddjjDF16_(uint scope, uint op,

0 commit comments

Comments
 (0)