diff options
| author | dim <dim@FreeBSD.org> | 2012-05-03 16:53:59 +0000 |
|---|---|---|
| committer | dim <dim@FreeBSD.org> | 2012-05-03 16:53:59 +0000 |
| commit | 822bde9df508e0b9afac5e581b0d6ab403417a28 (patch) | |
| tree | 2e51705e103e92c7be1b21e8bd8ffd5b5d0e4d52 /lib/Headers | |
| parent | 50b73317314e889cf39c7b1d6cbf419fa7502f22 (diff) | |
| download | FreeBSD-src-822bde9df508e0b9afac5e581b0d6ab403417a28.zip FreeBSD-src-822bde9df508e0b9afac5e581b0d6ab403417a28.tar.gz | |
Vendor import of clang release_31 branch r155985:
http://llvm.org/svn/llvm-project/cfe/branches/release_31@155985
Diffstat (limited to 'lib/Headers')
| -rw-r--r-- | lib/Headers/avx2intrin.h | 14 | ||||
| -rw-r--r-- | lib/Headers/avxintrin.h | 26 |
2 files changed, 10 insertions, 30 deletions
diff --git a/lib/Headers/avx2intrin.h b/lib/Headers/avx2intrin.h index d165f1f..884c46d 100644 --- a/lib/Headers/avx2intrin.h +++ b/lib/Headers/avx2intrin.h @@ -822,7 +822,9 @@ _mm256_permutevar8x32_epi32(__m256i a, __m256i b) #define _mm256_permute4x64_pd(V, M) __extension__ ({ \ __m256d __V = (V); \ - (__m256d)__builtin_ia32_permdf256((__v4df)__V, (M)); }) + (__m256d)__builtin_shufflevector((__v4df)__V, (__v4df) _mm256_setzero_pd(), \ + (M) & 0x3, ((M) & 0xc) >> 2, \ + ((M) & 0x30) >> 4, ((M) & 0xc0) >> 6); }) static __inline__ __m256 __attribute__((__always_inline__, __nodebug__)) _mm256_permutevar8x32_ps(__m256 a, __m256 b) @@ -832,16 +834,14 @@ _mm256_permutevar8x32_ps(__m256 a, __m256 b) #define _mm256_permute4x64_epi64(V, M) __extension__ ({ \ __m256i __V = (V); \ - (__m256i)__builtin_ia32_permdi256(__V, (M)); }) + (__m256i)__builtin_shufflevector((__v4di)__V, (__v4di) _mm256_setzero_si256(), \ + (M) & 0x3, ((M) & 0xc) >> 2, \ + ((M) & 0x30) >> 4, ((M) & 0xc0) >> 6); }) #define _mm256_permute2x128_si256(V1, V2, M) __extension__ ({ \ __m256i __V1 = (V1); \ __m256i __V2 = (V2); \ - __builtin_shufflevector(__V1, __V2, \ - ((M) & 0x3) * 2, \ - ((M) & 0x3) * 2 + 1, \ - (((M) & 0x30) >> 4) * 2, \ - (((M) & 0x30) >> 4) * 2 + 1); }) + (__m256i)__builtin_ia32_permti256(__V1, __V2, (M)); }) #define _mm256_extracti128_si256(A, O) __extension__ ({ \ __m256i __A = (A); \ diff --git a/lib/Headers/avxintrin.h b/lib/Headers/avxintrin.h index 7a0ec3f..ee7f835 100644 --- a/lib/Headers/avxintrin.h +++ b/lib/Headers/avxintrin.h @@ -289,37 +289,17 @@ _mm256_permutevar_ps(__m256 a, __m256i c) #define _mm256_permute2f128_pd(V1, V2, M) __extension__ ({ \ __m256d __V1 = (V1); \ __m256d __V2 = (V2); \ - (__m256d)__builtin_shufflevector((__v4df)__V1, (__v4df)__V2, \ - ((M) & 0x3) * 2, \ - ((M) & 0x3) * 2 + 1, \ - (((M) & 0x30) >> 4) * 2, \ - (((M) & 0x30) >> 4) * 2 + 1); }) + (__m256d)__builtin_ia32_vperm2f128_pd256((__v4df)__V1, (__v4df)__V2, (M)); }) #define _mm256_permute2f128_ps(V1, V2, M) __extension__ ({ \ __m256 __V1 = (V1); \ __m256 __V2 = (V2); \ - (__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); }) + (__m256)__builtin_ia32_vperm2f128_ps256((__v8sf)__V1, (__v8sf)__V2, (M)); }) #define _mm256_permute2f128_si256(V1, V2, M) __extension__ ({ \ __m256i __V1 = (V1); \ __m256i __V2 = (V2); \ - (__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); }) + (__m256i)__builtin_ia32_vperm2f128_si256((__v8si)__V1, (__v8si)__V2, (M)); }) /* Vector Blend */ #define _mm256_blend_pd(V1, V2, M) __extension__ ({ \ |
