| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-Ofast -march=sapphirerapids" } */ |
| 3 |
/* { dg-final { scan-assembler-times "vmov..." 3 } } */ |
| 4 |
/* { dg-final { scan-assembler-times "vblendps\t\\\$15" 3 } } */ |
| 5 |
/* { dg-final { scan-assembler-times "vblendps\t\\\$240" 3 } } */ |
| 6 |
/* { dg-final { scan-assembler-not "vperm2f128" } } */ |
| 7 |
|
| 8 |
#include<x86intrin.h> |
| 9 |
|
| 10 |
/* Vpermf128 */ |
| 11 |
__m256 |
| 12 |
perm0 (__m256 a, __m256 b) |
| 13 |
{ |
| 14 |
return _mm256_permute2f128_ps (a, b, 50); |
| 15 |
} |
| 16 |
|
| 17 |
__m256 |
| 18 |
perm1 (__m256 a, __m256 b) |
| 19 |
{ |
| 20 |
return _mm256_permute2f128_ps (a, b, 18); |
| 21 |
} |
| 22 |
|
| 23 |
__m256 |
| 24 |
perm2 (__m256 a, __m256 b) |
| 25 |
{ |
| 26 |
return _mm256_permute2f128_ps (a, b, 48); |
| 27 |
} |
| 28 |
|
| 29 |
__m256i |
| 30 |
perm3 (__m256i a, __m256i b) |
| 31 |
{ |
| 32 |
return _mm256_permute2f128_si256 (a, b, 50); |
| 33 |
} |
| 34 |
|
| 35 |
__m256i |
| 36 |
perm4 (__m256i a, __m256i b) |
| 37 |
{ |
| 38 |
return _mm256_permute2f128_si256 (a, b, 18); |
| 39 |
} |
| 40 |
|
| 41 |
__m256i |
| 42 |
perm5 (__m256i a, __m256i b) |
| 43 |
{ |
| 44 |
return _mm256_permute2f128_si256 (a, b, 48); |
| 45 |
} |
| 46 |
|
| 47 |
__m256d |
| 48 |
perm6 (__m256d a, __m256d b) |
| 49 |
{ |
| 50 |
return _mm256_permute2f128_pd (a, b, 50); |
| 51 |
} |
| 52 |
|
| 53 |
__m256d |
| 54 |
perm7 (__m256d a, __m256d b) |
| 55 |
{ |
| 56 |
return _mm256_permute2f128_pd (a, b, 18); |
| 57 |
} |
| 58 |
|
| 59 |
__m256d |
| 60 |
perm8 (__m256d a, __m256d b) |
| 61 |
{ |
| 62 |
return _mm256_permute2f128_pd (a, b, 48); |
| 63 |
} |