| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-march=x86-64 -mavx512fp16 -mavx512vl -O3" } */ |
| 3 |
/* { dg-final { scan-assembler-times "vcvtps2pd" 2 { target { ia32 } } } } */ |
| 4 |
/* { dg-final { scan-assembler-times "vcvtps2pd" 3 { target { ! ia32 } } } } */ |
| 5 |
/* { dg-final { scan-assembler-times "vcvtph2pd" 3 } } */ |
| 6 |
/* { dg-final { scan-assembler-times "vcvtph2ps" 3 } } */ |
| 7 |
|
| 8 |
#include <x86intrin.h> |
| 9 |
|
| 10 |
typedef _Float16 __v2hf __attribute__ ((__vector_size__ (4))); |
| 11 |
typedef _Float16 __v4hf __attribute__ ((__vector_size__ (8))); |
| 12 |
|
| 13 |
__v2df mm_cvtps_pd_builtin_convertvector(__v2sf a) |
| 14 |
{ |
| 15 |
return __builtin_convertvector(a, __v2df); |
| 16 |
} |
| 17 |
|
| 18 |
__v4df mm256_cvtps_pd_builtin_convertvector(__v4sf a) |
| 19 |
{ |
| 20 |
return __builtin_convertvector(a, __v4df); |
| 21 |
} |
| 22 |
|
| 23 |
__v8df mm512_cvtps_pd_builtin_convertvector(__v8sf a) |
| 24 |
{ |
| 25 |
return __builtin_convertvector(a, __v8df); |
| 26 |
} |
| 27 |
|
| 28 |
__v2df mm_cvtph_pd_builtin_convertvector(__v2hf a) |
| 29 |
{ |
| 30 |
return __builtin_convertvector(a, __v2df); |
| 31 |
} |
| 32 |
|
| 33 |
__v4df mm256_cvtph_pd_builtin_convertvector(__v4hf a) |
| 34 |
{ |
| 35 |
return __builtin_convertvector(a, __v4df); |
| 36 |
} |
| 37 |
|
| 38 |
__v8df mm512_cvtph_pd_builtin_convertvector(__v8hf a) |
| 39 |
{ |
| 40 |
return __builtin_convertvector(a, __v8df); |
| 41 |
} |
| 42 |
|
| 43 |
__v4sf mm_cvtph_ps_builtin_convertvector(__v4hf a) |
| 44 |
{ |
| 45 |
return __builtin_convertvector(a, __v4sf); |
| 46 |
} |
| 47 |
|
| 48 |
__v8sf mm256_cvtph_ps_builtin_convertvector(__v8hf a) |
| 49 |
{ |
| 50 |
return __builtin_convertvector(a, __v8sf); |
| 51 |
} |
| 52 |
|
| 53 |
__v16sf mm512_cvtph_ps_builtin_convertvector(__v16hf a) |
| 54 |
{ |
| 55 |
return __builtin_convertvector(a, __v16sf); |
| 56 |
} |