Skip to content

Commit edad89e

Browse files
RKSimonRKSimon
authored
[Headers][X86] Update MMX arithmetic intrinsics to be used in constexpr (#152296)
Update the easy add/sub/mul/logic/cmp/scalar_to_vector intrinsics to be constexpr compatible. I'm not expecting anyone to be very interested in using MMX intrinsics, but they're smaller than the other types and are useful to test the constexpr handling and test methods before we start applying them to SSE/AVX2/AVX512 intrinsics.
1 parent a485e0e commit edad89e

File tree

3 files changed

+54
-27
lines changed

3 files changed

+54
-27
lines changed

‎clang/lib/Headers/emmintrin.h

Lines changed: 6 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -2127,8 +2127,9 @@ _mm_add_epi32(__m128i __a, __m128i __b) {
21272127
/// \param __b
21282128
/// A 64-bit integer.
21292129
/// \returns A 64-bit integer containing the sum of both parameters.
2130-
static __inline__ __m64 __DEFAULT_FN_ATTRS _mm_add_si64(__m64 __a, __m64 __b) {
2131-
return (__m64)(((unsigned long long)__a) + ((unsigned long long)__b));
2130+
static __inline__ __m64 __DEFAULT_FN_ATTRS_CONSTEXPR _mm_add_si64(__m64 __a,
2131+
__m64 __b) {
2132+
return (__m64)(((__v1du)__a)[0] + ((__v1du)__b)[0]);
21322133
}
21332134

21342135
/// Adds the corresponding elements of two 128-bit vectors of [2 x i64],
@@ -2557,8 +2558,9 @@ _mm_sub_epi32(__m128i __a, __m128i __b) {
25572558
/// A 64-bit integer vector containing the subtrahend.
25582559
/// \returns A 64-bit integer vector containing the difference of the values in
25592560
/// the operands.
2560-
static __inline__ __m64 __DEFAULT_FN_ATTRS _mm_sub_si64(__m64 __a, __m64 __b) {
2561-
return (__m64)((unsigned long long)__a - (unsigned long long)__b);
2561+
static __inline__ __m64 __DEFAULT_FN_ATTRS_CONSTEXPR _mm_sub_si64(__m64 __a,
2562+
__m64 __b) {
2563+
return (__m64)(((__v1du)__a)[0] - ((__v1du)__b)[0]);
25622564
}
25632565

25642566
/// Subtracts the corresponding elements of two [2 x i64] vectors.

‎clang/lib/Headers/mmintrin.h

Lines changed: 23 additions & 23 deletions
Original file line numberDiff line numberDiff line change
@@ -85,7 +85,7 @@ _mm_empty(void) {
8585
/// A 32-bit integer value.
8686
/// \returns A 64-bit integer vector. The lower 32 bits contain the value of the
8787
/// parameter. The upper 32 bits are set to 0.
88-
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
88+
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
8989
_mm_cvtsi32_si64(int __i)
9090
{
9191
return __extension__ (__m64)(__v2si){__i, 0};
@@ -102,7 +102,7 @@ _mm_cvtsi32_si64(int __i)
102102
/// A 64-bit integer vector.
103103
/// \returns A 32-bit signed integer value containing the lower 32 bits of the
104104
/// parameter.
105-
static __inline__ int __DEFAULT_FN_ATTRS_SSE2
105+
static __inline__ int __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
106106
_mm_cvtsi64_si32(__m64 __m)
107107
{
108108
return ((__v2si)__m)[0];
@@ -118,10 +118,10 @@ _mm_cvtsi64_si32(__m64 __m)
118118
/// A 64-bit signed integer.
119119
/// \returns A 64-bit integer vector containing the same bitwise pattern as the
120120
/// parameter.
121-
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
121+
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
122122
_mm_cvtsi64_m64(long long __i)
123123
{
124-
return (__m64)__i;
124+
return __extension__ (__m64)(__v1di){__i};
125125
}
126126

127127
/// Casts a 64-bit integer vector into a 64-bit signed integer value.
@@ -134,10 +134,10 @@ _mm_cvtsi64_m64(long long __i)
134134
/// A 64-bit integer vector.
135135
/// \returns A 64-bit signed integer containing the same bitwise pattern as the
136136
/// parameter.
137-
static __inline__ long long __DEFAULT_FN_ATTRS_SSE2
137+
static __inline__ long long __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
138138
_mm_cvtm64_si64(__m64 __m)
139139
{
140-
return (long long)__m;
140+
return ((__v1di)__m)[0];
141141
}
142142

143143
/// Converts, with saturation, 16-bit signed integers from both 64-bit integer
@@ -379,7 +379,7 @@ _mm_unpacklo_pi32(__m64 __m1, __m64 __m2)
379379
/// A 64-bit integer vector of [8 x i8].
380380
/// \returns A 64-bit integer vector of [8 x i8] containing the sums of both
381381
/// parameters.
382-
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
382+
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
383383
_mm_add_pi8(__m64 __m1, __m64 __m2)
384384
{
385385
return (__m64)(((__v8qu)__m1) + ((__v8qu)__m2));
@@ -400,7 +400,7 @@ _mm_add_pi8(__m64 __m1, __m64 __m2)
400400
/// A 64-bit integer vector of [4 x i16].
401401
/// \returns A 64-bit integer vector of [4 x i16] containing the sums of both
402402
/// parameters.
403-
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
403+
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
404404
_mm_add_pi16(__m64 __m1, __m64 __m2)
405405
{
406406
return (__m64)(((__v4hu)__m1) + ((__v4hu)__m2));
@@ -421,7 +421,7 @@ _mm_add_pi16(__m64 __m1, __m64 __m2)
421421
/// A 64-bit integer vector of [2 x i32].
422422
/// \returns A 64-bit integer vector of [2 x i32] containing the sums of both
423423
/// parameters.
424-
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
424+
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
425425
_mm_add_pi32(__m64 __m1, __m64 __m2)
426426
{
427427
return (__m64)(((__v2su)__m1) + ((__v2su)__m2));
@@ -536,7 +536,7 @@ _mm_adds_pu16(__m64 __m1, __m64 __m2)
536536
/// A 64-bit integer vector of [8 x i8] containing the subtrahends.
537537
/// \returns A 64-bit integer vector of [8 x i8] containing the differences of
538538
/// both parameters.
539-
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
539+
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
540540
_mm_sub_pi8(__m64 __m1, __m64 __m2)
541541
{
542542
return (__m64)(((__v8qu)__m1) - ((__v8qu)__m2));
@@ -557,7 +557,7 @@ _mm_sub_pi8(__m64 __m1, __m64 __m2)
557557
/// A 64-bit integer vector of [4 x i16] containing the subtrahends.
558558
/// \returns A 64-bit integer vector of [4 x i16] containing the differences of
559559
/// both parameters.
560-
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
560+
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
561561
_mm_sub_pi16(__m64 __m1, __m64 __m2)
562562
{
563563
return (__m64)(((__v4hu)__m1) - ((__v4hu)__m2));
@@ -578,7 +578,7 @@ _mm_sub_pi16(__m64 __m1, __m64 __m2)
578578
/// A 64-bit integer vector of [2 x i32] containing the subtrahends.
579579
/// \returns A 64-bit integer vector of [2 x i32] containing the differences of
580580
/// both parameters.
581-
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
581+
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
582582
_mm_sub_pi32(__m64 __m1, __m64 __m2)
583583
{
584584
return (__m64)(((__v2su)__m1) - ((__v2su)__m2));
@@ -745,7 +745,7 @@ _mm_mulhi_pi16(__m64 __m1, __m64 __m2)
745745
/// A 64-bit integer vector of [4 x i16].
746746
/// \returns A 64-bit integer vector of [4 x i16] containing the lower 16 bits
747747
/// of the products of both parameters.
748-
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
748+
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
749749
_mm_mullo_pi16(__m64 __m1, __m64 __m2)
750750
{
751751
return (__m64)(((__v4hu)__m1) * ((__v4hu)__m2));
@@ -1134,7 +1134,7 @@ _mm_srli_si64(__m64 __m, int __count)
11341134
/// A 64-bit integer vector.
11351135
/// \returns A 64-bit integer vector containing the bitwise AND of both
11361136
/// parameters.
1137-
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
1137+
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
11381138
_mm_and_si64(__m64 __m1, __m64 __m2)
11391139
{
11401140
return (__m64)(((__v1du)__m1) & ((__v1du)__m2));
@@ -1155,7 +1155,7 @@ _mm_and_si64(__m64 __m1, __m64 __m2)
11551155
/// A 64-bit integer vector.
11561156
/// \returns A 64-bit integer vector containing the bitwise AND of the second
11571157
/// parameter and the one's complement of the first parameter.
1158-
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
1158+
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
11591159
_mm_andnot_si64(__m64 __m1, __m64 __m2)
11601160
{
11611161
return (__m64)(~((__v1du)__m1) & ((__v1du)__m2));
@@ -1173,7 +1173,7 @@ _mm_andnot_si64(__m64 __m1, __m64 __m2)
11731173
/// A 64-bit integer vector.
11741174
/// \returns A 64-bit integer vector containing the bitwise OR of both
11751175
/// parameters.
1176-
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
1176+
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
11771177
_mm_or_si64(__m64 __m1, __m64 __m2)
11781178
{
11791179
return (__m64)(((__v1du)__m1) | ((__v1du)__m2));
@@ -1191,7 +1191,7 @@ _mm_or_si64(__m64 __m1, __m64 __m2)
11911191
/// A 64-bit integer vector.
11921192
/// \returns A 64-bit integer vector containing the bitwise exclusive OR of both
11931193
/// parameters.
1194-
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
1194+
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
11951195
_mm_xor_si64(__m64 __m1, __m64 __m2)
11961196
{
11971197
return (__m64)(((__v1du)__m1) ^ ((__v1du)__m2));
@@ -1213,7 +1213,7 @@ _mm_xor_si64(__m64 __m1, __m64 __m2)
12131213
/// A 64-bit integer vector of [8 x i8].
12141214
/// \returns A 64-bit integer vector of [8 x i8] containing the comparison
12151215
/// results.
1216-
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
1216+
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
12171217
_mm_cmpeq_pi8(__m64 __m1, __m64 __m2)
12181218
{
12191219
return (__m64)(((__v8qi)__m1) == ((__v8qi)__m2));
@@ -1235,7 +1235,7 @@ _mm_cmpeq_pi8(__m64 __m1, __m64 __m2)
12351235
/// A 64-bit integer vector of [4 x i16].
12361236
/// \returns A 64-bit integer vector of [4 x i16] containing the comparison
12371237
/// results.
1238-
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
1238+
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
12391239
_mm_cmpeq_pi16(__m64 __m1, __m64 __m2)
12401240
{
12411241
return (__m64)(((__v4hi)__m1) == ((__v4hi)__m2));
@@ -1257,7 +1257,7 @@ _mm_cmpeq_pi16(__m64 __m1, __m64 __m2)
12571257
/// A 64-bit integer vector of [2 x i32].
12581258
/// \returns A 64-bit integer vector of [2 x i32] containing the comparison
12591259
/// results.
1260-
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
1260+
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
12611261
_mm_cmpeq_pi32(__m64 __m1, __m64 __m2)
12621262
{
12631263
return (__m64)(((__v2si)__m1) == ((__v2si)__m2));
@@ -1279,7 +1279,7 @@ _mm_cmpeq_pi32(__m64 __m1, __m64 __m2)
12791279
/// A 64-bit integer vector of [8 x i8].
12801280
/// \returns A 64-bit integer vector of [8 x i8] containing the comparison
12811281
/// results.
1282-
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
1282+
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
12831283
_mm_cmpgt_pi8(__m64 __m1, __m64 __m2)
12841284
{
12851285
/* This function always performs a signed comparison, but __v8qi is a char
@@ -1303,7 +1303,7 @@ _mm_cmpgt_pi8(__m64 __m1, __m64 __m2)
13031303
/// A 64-bit integer vector of [4 x i16].
13041304
/// \returns A 64-bit integer vector of [4 x i16] containing the comparison
13051305
/// results.
1306-
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
1306+
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
13071307
_mm_cmpgt_pi16(__m64 __m1, __m64 __m2)
13081308
{
13091309
return (__m64)((__v4hi)__m1 > (__v4hi)__m2);
@@ -1325,7 +1325,7 @@ _mm_cmpgt_pi16(__m64 __m1, __m64 __m2)
13251325
/// A 64-bit integer vector of [2 x i32].
13261326
/// \returns A 64-bit integer vector of [2 x i32] containing the comparison
13271327
/// results.
1328-
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
1328+
static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
13291329
_mm_cmpgt_pi32(__m64 __m1, __m64 __m2)
13301330
{
13311331
return (__m64)((__v2si)__m1 > (__v2si)__m2);

0 commit comments

Comments
 (0)