Author: Simon Pilgrim Date: 2024-10-29T14:34:19Z New Revision: e281d96a81bca896692da4a07ca1423ee6dc1f53
URL: https://github.com/llvm/llvm-project/commit/e281d96a81bca896692da4a07ca1423ee6dc1f53 DIFF: https://github.com/llvm/llvm-project/commit/e281d96a81bca896692da4a07ca1423ee6dc1f53.diff LOG: [clang][x86] Add constexpr support for _mm_add_epi32/64 and _mm_sub_epi32/64 Added: Modified: clang/lib/Headers/emmintrin.h clang/test/CodeGen/X86/builtin_test_helpers.h clang/test/CodeGen/X86/sse2-builtins.c Removed: ################################################################################ diff --git a/clang/lib/Headers/emmintrin.h b/clang/lib/Headers/emmintrin.h index d6494762169b25..778cdf99a12964 100644 --- a/clang/lib/Headers/emmintrin.h +++ b/clang/lib/Headers/emmintrin.h @@ -2110,8 +2110,8 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_add_epi16(__m128i __a, /// A 128-bit vector of [4 x i32]. /// \returns A 128-bit vector of [4 x i32] containing the sums of both /// parameters. -static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_add_epi32(__m128i __a, - __m128i __b) { +static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR +_mm_add_epi32(__m128i __a, __m128i __b) { return (__m128i)((__v4su)__a + (__v4su)__b); } @@ -2147,8 +2147,8 @@ static __inline__ __m64 __DEFAULT_FN_ATTRS _mm_add_si64(__m64 __a, __m64 __b) { /// A 128-bit vector of [2 x i64]. /// \returns A 128-bit vector of [2 x i64] containing the sums of both /// parameters. -static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_add_epi64(__m128i __a, - __m128i __b) { +static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR +_mm_add_epi64(__m128i __a, __m128i __b) { return (__m128i)((__v2du)__a + (__v2du)__b); } @@ -2539,8 +2539,8 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_sub_epi16(__m128i __a, /// A 128-bit integer vector containing the subtrahends. /// \returns A 128-bit integer vector containing the diff erences of the values /// in the operands. -static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_sub_epi32(__m128i __a, - __m128i __b) { +static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR +_mm_sub_epi32(__m128i __a, __m128i __b) { return (__m128i)((__v4su)__a - (__v4su)__b); } @@ -2573,8 +2573,8 @@ static __inline__ __m64 __DEFAULT_FN_ATTRS _mm_sub_si64(__m64 __a, __m64 __b) { /// A 128-bit integer vector containing the subtrahends. /// \returns A 128-bit integer vector containing the diff erences of the values /// in the operands. -static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_sub_epi64(__m128i __a, - __m128i __b) { +static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR +_mm_sub_epi64(__m128i __a, __m128i __b) { return (__m128i)((__v2du)__a - (__v2du)__b); } diff --git a/clang/test/CodeGen/X86/builtin_test_helpers.h b/clang/test/CodeGen/X86/builtin_test_helpers.h index 5e77ff3a7ca45e..f6547d4cb29ca3 100644 --- a/clang/test/CodeGen/X86/builtin_test_helpers.h +++ b/clang/test/CodeGen/X86/builtin_test_helpers.h @@ -16,6 +16,15 @@ constexpr bool match_m128i(__m128i v, unsigned long long a, unsigned long long b return v[0] == a && v[1] == b; } +constexpr bool match_v2di(__m128i v, long long a, long long b) { + return v[0] == a && v[1] == b; +} + +constexpr bool match_v4si(__m128i _v, int a, int b, int c, int d) { + __v4si v = (__v4si)_v; + return v[0] == a && v[1] == b && v[2] == c && v[3] == d; +} + constexpr bool match_m256(__m256 v, float a, float b, float c, float d, float e, float f, float g, float h) { return v[0] == a && v[1] == b && v[2] == c && v[3] == d && v[4] == e && v[5] == f && v[6] == g && v[7] == h; } diff --git a/clang/test/CodeGen/X86/sse2-builtins.c b/clang/test/CodeGen/X86/sse2-builtins.c index 4287d3d4b5ec4e..82aa7a2d2b49d1 100644 --- a/clang/test/CodeGen/X86/sse2-builtins.c +++ b/clang/test/CodeGen/X86/sse2-builtins.c @@ -32,12 +32,14 @@ __m128i test_mm_add_epi32(__m128i A, __m128i B) { // CHECK: add <4 x i32> return _mm_add_epi32(A, B); } +TEST_CONSTEXPR(match_v4si(_mm_add_epi32((__m128i)(__v4si){+1, -2, +3, -4}, (__m128i)(__v4si){-10, +8, +6, -4}), -9, +6, +9, -8)); __m128i test_mm_add_epi64(__m128i A, __m128i B) { // CHECK-LABEL: test_mm_add_epi64 // CHECK: add <2 x i64> return _mm_add_epi64(A, B); } +TEST_CONSTEXPR(match_v2di(_mm_add_epi64((__m128i)(__v2di){+5, -3}, (__m128i)(__v2di){-9, +8}), -4, +5)); __m128d test_mm_add_pd(__m128d A, __m128d B) { // CHECK-LABEL: test_mm_add_pd @@ -1634,12 +1636,14 @@ __m128i test_mm_sub_epi32(__m128i A, __m128i B) { // CHECK: sub <4 x i32> return _mm_sub_epi32(A, B); } +TEST_CONSTEXPR(match_v4si(_mm_sub_epi32((__m128i)(__v4si){+1, -2, +3, -4}, (__m128i)(__v4si){-10, +8, +6, -4}), +11, -10, -3, 0)); __m128i test_mm_sub_epi64(__m128i A, __m128i B) { // CHECK-LABEL: test_mm_sub_epi64 // CHECK: sub <2 x i64> return _mm_sub_epi64(A, B); } +TEST_CONSTEXPR(match_v2di(_mm_sub_epi64((__m128i)(__v2di){+5, -3}, (__m128i)(__v2di){-9, +8}), +14, -11)); __m128d test_mm_sub_pd(__m128d A, __m128d B) { // CHECK-LABEL: test_mm_sub_pd _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
