diff options
Diffstat (limited to 'lib')
| -rw-r--r-- | lib/Headers/avx2intrin.h | 6 | ||||
| -rw-r--r-- | lib/Headers/avxintrin.h | 26 |
2 files changed, 28 insertions, 4 deletions
diff --git a/lib/Headers/avx2intrin.h b/lib/Headers/avx2intrin.h index cc3ffe69b3..359c0aad3c 100644 --- a/lib/Headers/avx2intrin.h +++ b/lib/Headers/avx2intrin.h @@ -837,7 +837,11 @@ _mm256_permutevar8x32_ps(__m256 a, __m256 b) #define _mm256_permute2x128_si256(V1, V2, M) __extension__ ({ \ __m256i __V1 = (V1); \ __m256i __V2 = (V2); \ - (__m256i)__builtin_ia32_permti256(__V1, __V2, (M)); }) + __builtin_shufflevector(__V1, __V2, \ + ((M) & 0x3) * 2, \ + ((M) & 0x3) * 2 + 1, \ + (((M) & 0x30) >> 4) * 2, \ + (((M) & 0x30) >> 4) * 2 + 1); }) #define _mm256_extracti128_si256(A, O) __extension__ ({ \ __m256i __A = (A); \ diff --git a/lib/Headers/avxintrin.h b/lib/Headers/avxintrin.h index 0758a653bf..d8ecd1d413 100644 --- a/lib/Headers/avxintrin.h +++ b/lib/Headers/avxintrin.h @@ -289,17 +289,37 @@ _mm256_permutevar_ps(__m256 a, __m256i c) #define _mm256_permute2f128_pd(V1, V2, M) __extension__ ({ \ __m256d __V1 = (V1); \ __m256d __V2 = (V2); \ - (__m256d)__builtin_ia32_vperm2f128_pd256((__v4df)__V1, (__v4df)__V2, (M)); }) + (__m256d)__builtin_shufflevector((__v4df)__V1, (__v4df)__V2, \ + ((M) & 0x3) * 2, \ + ((M) & 0x3) * 2 + 1, \ + (((M) & 0x30) >> 4) * 2, \ + (((M) & 0x30) >> 4) * 2 + 1); }) #define _mm256_permute2f128_ps(V1, V2, M) __extension__ ({ \ __m256 __V1 = (V1); \ __m256 __V2 = (V2); \ - (__m256)__builtin_ia32_vperm2f128_ps256((__v8sf)__V1, (__v8sf)__V2, (M)); }) + (__m256)__builtin_shufflevector((__v8sf)__V1, (__v8sf)__V2, \ + ((M) & 0x3) * 4, \ + ((M) & 0x3) * 4 + 1, \ + ((M) & 0x3) * 4 + 2, \ + ((M) & 0x3) * 4 + 3, \ + (((M) & 0x30) >> 4) * 4, \ + (((M) & 0x30) >> 4) * 4 + 1, \ + (((M) & 0x30) >> 4) * 4 + 2, \ + (((M) & 0x30) >> 4) * 4 + 3); }) #define _mm256_permute2f128_si256(V1, V2, M) __extension__ ({ \ __m256i __V1 = (V1); \ __m256i __V2 = (V2); \ - (__m256i)__builtin_ia32_vperm2f128_si256((__v8si)__V1, (__v8si)__V2, (M)); }) + (__m256i)__builtin_shufflevector((__v8si)__V1, (__v8si)__V2, \ + ((M) & 0x3) * 4, \ + ((M) & 0x3) * 4 + 1, \ + ((M) & 0x3) * 4 + 2, \ + ((M) & 0x3) * 4 + 3, \ + (((M) & 0x30) >> 4) * 4, \ + (((M) & 0x30) >> 4) * 4 + 1, \ + (((M) & 0x30) >> 4) * 4 + 2, \ + (((M) & 0x30) >> 4) * 4 + 3); }) /* Vector Blend */ #define _mm256_blend_pd(V1, V2, M) __extension__ ({ \ |