| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-Ofast -march=sapphirerapids" } */ |
| 3 |
/* { dg-final { scan-assembler-times "vmovaps" 1 } } */ |
| 4 |
/* { dg-final { scan-assembler-times "vblendps\t\\\$15" 1 } } */ |
| 5 |
/* { dg-final { scan-assembler-times "vblendps\t\\\$240" 5 } } */ |
| 6 |
|
| 7 |
#include<x86intrin.h> |
| 8 |
|
| 9 |
/* Vpermi128/Vpermf128 */ |
| 10 |
__m256i |
| 11 |
perm0 (__m256i a, __m256i b) |
| 12 |
{ |
| 13 |
return _mm256_permute2x128_si256 (a, b, 50); |
| 14 |
} |
| 15 |
|
| 16 |
__m256i |
| 17 |
perm1 (__m256i a, __m256i b) |
| 18 |
{ |
| 19 |
return _mm256_permute2x128_si256 (a, b, 18); |
| 20 |
} |
| 21 |
|
| 22 |
__m256i |
| 23 |
perm2 (__m256i a, __m256i b) |
| 24 |
{ |
| 25 |
return _mm256_permute2x128_si256 (a, b, 48); |
| 26 |
} |
| 27 |
|
| 28 |
/* vshuf{i,f}{32x4,64x2} ymm .*/ |
| 29 |
__m256i |
| 30 |
shuff0 (__m256i a, __m256i b) |
| 31 |
{ |
| 32 |
return _mm256_shuffle_i32x4(a, b, 2); |
| 33 |
} |
| 34 |
|
| 35 |
__m256 |
| 36 |
shuff1 (__m256 a, __m256 b) |
| 37 |
{ |
| 38 |
return _mm256_shuffle_f32x4(a, b, 2); |
| 39 |
} |
| 40 |
|
| 41 |
__m256i |
| 42 |
shuff2 (__m256i a, __m256i b) |
| 43 |
{ |
| 44 |
return _mm256_shuffle_i64x2(a, b, 2); |
| 45 |
} |
| 46 |
|
| 47 |
__m256d |
| 48 |
shuff3 (__m256d a, __m256d b) |
| 49 |
{ |
| 50 |
return _mm256_shuffle_f64x2(a, b, 2); |
| 51 |
} |