| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-mavx512fp16 -mavx512vl -mavx512bw -O2 -mavx512dq -fno-trapping-math" } */ |
| 3 |
/* { dg-final { scan-assembler-times "vcvtdq2pd" 4 { target { ia32 } } } } */ |
| 4 |
/* { dg-final { scan-assembler-times "vcvtdq2pd" 6 { target { ! ia32 } } } } */ |
| 5 |
/* { dg-final { scan-assembler-times "vcvtdq2ps" 6 { target { ia32 } } } } */ |
| 6 |
/* { dg-final { scan-assembler-times "vcvtdq2ps" 8 { target { ! ia32 } } } } */ |
| 7 |
/* { dg-final { scan-assembler-times "vcvtw2ph" 8 { target { ia32 } } } } */ |
| 8 |
/* { dg-final { scan-assembler-times "vcvtw2ph" 10 { target { ! ia32 } } } } */ |
| 9 |
/* { dg-final { scan-assembler-times "vpmovsxbd" 5 { target { ia32 } } } } */ |
| 10 |
/* { dg-final { scan-assembler-times "vpmovsxbd" 7 { target { ! ia32 } } } } */ |
| 11 |
/* { dg-final { scan-assembler-times "vpmovzxbd" 5 { target { ia32 } } } } */ |
| 12 |
/* { dg-final { scan-assembler-times "vpmovzxbd" 7 { target { ! ia32 } } } } */ |
| 13 |
/* { dg-final { scan-assembler-times "vpmovsxbd" 5 { target { ia32 } } } } */ |
| 14 |
/* { dg-final { scan-assembler-times "vpmovsxbd" 7 { target { ! ia32 } } } } */ |
| 15 |
/* { dg-final { scan-assembler-times "vpmovzxbd" 5 { target { ia32 } } } } */ |
| 16 |
/* { dg-final { scan-assembler-times "vpmovzxbd" 7 { target { ! ia32 } } } } */ |
| 17 |
|
| 18 |
#include <x86intrin.h> |
| 19 |
|
| 20 |
typedef char __v2qi __attribute__ ((__vector_size__ (2))); |
| 21 |
typedef char __v4qi __attribute__ ((__vector_size__ (4))); |
| 22 |
typedef char __v8qi __attribute__ ((__vector_size__ (8))); |
| 23 |
typedef char __v16qi __attribute__ ((__vector_size__ (16))); |
| 24 |
typedef unsigned char __v2qu __attribute__ ((vector_size (2))); |
| 25 |
typedef unsigned char __v4qu __attribute__ ((vector_size (4))); |
| 26 |
typedef unsigned char __v8qu __attribute__ ((vector_size (8))); |
| 27 |
typedef unsigned char __v16qu __attribute__ ((vector_size (16))); |
| 28 |
typedef _Float16 __v2hf __attribute__ ((__vector_size__ (4))); |
| 29 |
typedef _Float16 __v4hf __attribute__ ((__vector_size__ (8))); |
| 30 |
typedef _Float16 __v8hf __attribute__ ((__vector_size__ (16))); |
| 31 |
|
| 32 |
__v2df mm_cvtepi8_pd_builtin_convertvector(__v2qi a) |
| 33 |
{ |
| 34 |
return __builtin_convertvector((__v2qi)a, __v2df); |
| 35 |
} |
| 36 |
|
| 37 |
__v4df mm256_cvtepi8_pd_builtin_convertvector(__v4qi a) |
| 38 |
{ |
| 39 |
return __builtin_convertvector((__v4qi)a, __v4df); |
| 40 |
} |
| 41 |
|
| 42 |
__v8df mm512_cvtepi8_pd_builtin_convertvector(__v8qi a) |
| 43 |
{ |
| 44 |
return __builtin_convertvector((__v8qi)a, __v8df); |
| 45 |
} |
| 46 |
|
| 47 |
__v2df mm_cvtepu8_pd_builtin_convertvector(__v2qu a) |
| 48 |
{ |
| 49 |
return __builtin_convertvector((__v2qu)a, __v2df); |
| 50 |
} |
| 51 |
|
| 52 |
__v4df mm256_cvtepu8_pd_builtin_convertvector(__v4qu a) |
| 53 |
{ |
| 54 |
return __builtin_convertvector((__v4qu)a, __v4df); |
| 55 |
} |
| 56 |
|
| 57 |
__v8df mm512_cvtepu8_pd_builtin_convertvector(__v8qu a) |
| 58 |
{ |
| 59 |
return __builtin_convertvector((__v8qu)a, __v8df); |
| 60 |
} |
| 61 |
|
| 62 |
__v2sf mm64_cvtepi8_ps_builtin_convertvector(__v2qi a) |
| 63 |
{ |
| 64 |
return __builtin_convertvector((__v2qi)a, __v2sf); |
| 65 |
} |
| 66 |
|
| 67 |
__v4sf mm128_cvtepi8_ps_builtin_convertvector(__v4qi a) |
| 68 |
{ |
| 69 |
return __builtin_convertvector((__v4qi)a, __v4sf); |
| 70 |
} |
| 71 |
|
| 72 |
__v8sf mm256_cvtepi8_ps_builtin_convertvector(__v8qi a) |
| 73 |
{ |
| 74 |
return __builtin_convertvector((__v8qi)a, __v8sf); |
| 75 |
} |
| 76 |
|
| 77 |
__v16sf mm512_cvtepi8_ps_builtin_convertvector(__v16qi a) |
| 78 |
{ |
| 79 |
return __builtin_convertvector((__v16qi)a, __v16sf); |
| 80 |
} |
| 81 |
|
| 82 |
__v2sf mm64_cvtepu8_ps_builtin_convertvector(__v2qu a) |
| 83 |
{ |
| 84 |
return __builtin_convertvector((__v2qu)a, __v2sf); |
| 85 |
} |
| 86 |
|
| 87 |
__v4sf mm128_cvtepu8_ps_builtin_convertvector(__v4qu a) |
| 88 |
{ |
| 89 |
return __builtin_convertvector((__v4qu)a, __v4sf); |
| 90 |
} |
| 91 |
|
| 92 |
__v8sf mm256_cvtepu8_ps_builtin_convertvector(__v8qu a) |
| 93 |
{ |
| 94 |
return __builtin_convertvector((__v8qu)a, __v8sf); |
| 95 |
} |
| 96 |
|
| 97 |
__v16sf mm512_cvtepu8_ps_builtin_convertvector(__v16qu a) |
| 98 |
{ |
| 99 |
return __builtin_convertvector((__v16qu)a, __v16sf); |
| 100 |
} |
| 101 |
|
| 102 |
__v2hf mm32_cvtepi8_ph_builtin_convertvector(__v2qi a) |
| 103 |
{ |
| 104 |
return __builtin_convertvector((__v2qi)a, __v2hf); |
| 105 |
} |
| 106 |
|
| 107 |
__v4hf mm64_cvtepi8_ph_builtin_convertvector(__v4qi a) |
| 108 |
{ |
| 109 |
return __builtin_convertvector((__v4qi)a, __v4hf); |
| 110 |
} |
| 111 |
|
| 112 |
__v8hf mm128_cvtepi8_ph_builtin_convertvector(__v8qi a) |
| 113 |
{ |
| 114 |
return __builtin_convertvector((__v8qi)a, __v8hf); |
| 115 |
} |
| 116 |
|
| 117 |
__v16hf mm256_cvtepi8_ph_builtin_convertvector(__v16qi a) |
| 118 |
{ |
| 119 |
return __builtin_convertvector((__v16qi)a, __v16hf); |
| 120 |
} |
| 121 |
|
| 122 |
__v32hf mm512_cvtepi8_ph_builtin_convertvector(__v32qi a) |
| 123 |
{ |
| 124 |
return __builtin_convertvector((__v32qi)a, __v32hf); |
| 125 |
} |
| 126 |
|
| 127 |
__v2hf mm32_cvtepu8_ph_builtin_convertvector(__v2qu a) |
| 128 |
{ |
| 129 |
return __builtin_convertvector((__v2qu)a, __v2hf); |
| 130 |
} |
| 131 |
|
| 132 |
__v4hf mm64_cvtepu8_ph_builtin_convertvector(__v4qu a) |
| 133 |
{ |
| 134 |
return __builtin_convertvector((__v4qu)a, __v4hf); |
| 135 |
} |
| 136 |
|
| 137 |
__v8hf mm128_cvtepu8_ph_builtin_convertvector(__v8qu a) |
| 138 |
{ |
| 139 |
return __builtin_convertvector((__v8qu)a, __v8hf); |
| 140 |
} |
| 141 |
|
| 142 |
__v16hf mm256_cvtepu8_ph_builtin_convertvector(__v16qu a) |
| 143 |
{ |
| 144 |
return __builtin_convertvector((__v16qu)a, __v16hf); |
| 145 |
} |
| 146 |
|
| 147 |
__v32hf mm512_cvtepu8_ph_builtin_convertvector(__v32qu a) |
| 148 |
{ |
| 149 |
return __builtin_convertvector((__v32qu)a, __v32hf); |
| 150 |
} |