| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-mavx512fp16 -mavx512vl -mavx512bw -O2 -mavx512dq -fno-trapping-math" } */ |
| 3 |
/* { dg-final { scan-assembler-times "vcvttpd2dq" 4 { target { ia32 } } } } */ |
| 4 |
/* { dg-final { scan-assembler-times "vcvttpd2dq" 6 { target { ! ia32 } } } } */ |
| 5 |
/* { dg-final { scan-assembler-times "vcvttps2dq" 6 { target { ia32 } } } } */ |
| 6 |
/* { dg-final { scan-assembler-times "vcvttps2dq" 8 { target { ! ia32 } } } } */ |
| 7 |
/* { dg-final { scan-assembler-times "vcvttph2w" 8 { target { ia32 } } } } */ |
| 8 |
/* { dg-final { scan-assembler-times "vcvttph2w" 10 { target { ! ia32 } } } } */ |
| 9 |
/* { dg-final { scan-assembler-times "vpmovdb" 10 { target { ia32 } } } } */ |
| 10 |
/* { dg-final { scan-assembler-times "vpmovdb" 14 { target { ! ia32 } } } } */ |
| 11 |
/* { dg-final { scan-assembler-times "vpmovwb" 8 { target { ia32 } } } } */ |
| 12 |
/* { dg-final { scan-assembler-times "vpmovwb" 10 { target { ! ia32 } } } } */ |
| 13 |
|
| 14 |
#include <x86intrin.h> |
| 15 |
|
| 16 |
typedef char __v2qi __attribute__ ((__vector_size__ (2))); |
| 17 |
typedef char __v4qi __attribute__ ((__vector_size__ (4))); |
| 18 |
typedef char __v8qi __attribute__ ((__vector_size__ (8))); |
| 19 |
typedef char __v16qi __attribute__ ((__vector_size__ (16))); |
| 20 |
typedef unsigned char __v2qu __attribute__ ((vector_size (2))); |
| 21 |
typedef unsigned char __v4qu __attribute__ ((vector_size (4))); |
| 22 |
typedef unsigned char __v8qu __attribute__ ((vector_size (8))); |
| 23 |
typedef unsigned char __v16qu __attribute__ ((vector_size (16))); |
| 24 |
typedef _Float16 __v2hf __attribute__ ((__vector_size__ (4))); |
| 25 |
typedef _Float16 __v4hf __attribute__ ((__vector_size__ (8))); |
| 26 |
typedef _Float16 __v8hf __attribute__ ((__vector_size__ (16))); |
| 27 |
|
| 28 |
__v2qi mm_cvtpd_epi8_builtin_convertvector(__v2df a) |
| 29 |
{ |
| 30 |
return __builtin_convertvector((__v2df)a, __v2qi); |
| 31 |
} |
| 32 |
|
| 33 |
__v4qi mm256_cvtpd_epi8_builtin_convertvector(__v4df a) |
| 34 |
{ |
| 35 |
return __builtin_convertvector((__v4df)a, __v4qi); |
| 36 |
} |
| 37 |
|
| 38 |
__v8qi mm512_cvtpd_epi8_builtin_convertvector(__v8df a) |
| 39 |
{ |
| 40 |
return __builtin_convertvector((__v8df)a, __v8qi); |
| 41 |
} |
| 42 |
|
| 43 |
__v2qu mm_cvtpd_epu8_builtin_convertvector(__v2df a) |
| 44 |
{ |
| 45 |
return __builtin_convertvector((__v2df)a, __v2qu); |
| 46 |
} |
| 47 |
|
| 48 |
__v4qu mm256_cvtpd_epu8_builtin_convertvector(__v4df a) |
| 49 |
{ |
| 50 |
return __builtin_convertvector((__v4df)a, __v4qu); |
| 51 |
} |
| 52 |
|
| 53 |
__v8qu mm512_cvtpd_epu8_builtin_convertvector(__v8df a) |
| 54 |
{ |
| 55 |
return __builtin_convertvector((__v8df)a, __v8qu); |
| 56 |
} |
| 57 |
|
| 58 |
__v2qi mm64_cvtps_epi8_builtin_convertvector(__v2sf a) |
| 59 |
{ |
| 60 |
return __builtin_convertvector((__v2sf)a, __v2qi); |
| 61 |
} |
| 62 |
|
| 63 |
__v4qi mm128_cvtps_epi8_builtin_convertvector(__v4sf a) |
| 64 |
{ |
| 65 |
return __builtin_convertvector((__v4sf)a, __v4qi); |
| 66 |
} |
| 67 |
|
| 68 |
__v8qi mm256_cvtps_epi8_builtin_convertvector(__v8sf a) |
| 69 |
{ |
| 70 |
return __builtin_convertvector((__v8sf)a, __v8qi); |
| 71 |
} |
| 72 |
|
| 73 |
__v16qi mm512_cvtps_epi8_builtin_convertvector(__v16sf a) |
| 74 |
{ |
| 75 |
return __builtin_convertvector((__v16sf)a, __v16qi); |
| 76 |
} |
| 77 |
|
| 78 |
__v2qu mm64_cvtps_epu8_builtin_convertvector(__v2sf a) |
| 79 |
{ |
| 80 |
return __builtin_convertvector((__v2sf)a, __v2qu); |
| 81 |
} |
| 82 |
|
| 83 |
__v4qu mm128_cvtps_epu8_builtin_convertvector(__v4sf a) |
| 84 |
{ |
| 85 |
return __builtin_convertvector((__v4sf)a, __v4qu); |
| 86 |
} |
| 87 |
|
| 88 |
__v8qu mm256_cvtps_epu8_builtin_convertvector(__v8sf a) |
| 89 |
{ |
| 90 |
return __builtin_convertvector((__v8sf)a, __v8qu); |
| 91 |
} |
| 92 |
|
| 93 |
__v16qu mm512_cvtps_epu8_builtin_convertvector(__v16sf a) |
| 94 |
{ |
| 95 |
return __builtin_convertvector((__v16sf)a, __v16qu); |
| 96 |
} |
| 97 |
|
| 98 |
__v2qi mm32_cvtph_epi8_builtin_convertvector(__v2hf a) |
| 99 |
{ |
| 100 |
return __builtin_convertvector((__v2hf)a, __v2qi); |
| 101 |
} |
| 102 |
|
| 103 |
__v4qi mm64_cvtph_epi8_builtin_convertvector(__v4hf a) |
| 104 |
{ |
| 105 |
return __builtin_convertvector((__v4hf)a, __v4qi); |
| 106 |
} |
| 107 |
|
| 108 |
__v8qi mm128_cvtph_epi8_builtin_convertvector(__v8hf a) |
| 109 |
{ |
| 110 |
return __builtin_convertvector((__v8hf)a, __v8qi); |
| 111 |
} |
| 112 |
|
| 113 |
__v16qi mm256_cvtph_epi8_builtin_convertvector(__v16hf a) |
| 114 |
{ |
| 115 |
return __builtin_convertvector((__v16hf)a, __v16qi); |
| 116 |
} |
| 117 |
|
| 118 |
__v32qi mm512_cvtph_epi8_builtin_convertvector(__v32hf a) |
| 119 |
{ |
| 120 |
return __builtin_convertvector((__v32hf)a, __v32qi); |
| 121 |
} |
| 122 |
|
| 123 |
__v2qu mm32_cvtph_epu8_builtin_convertvector(__v2hf a) |
| 124 |
{ |
| 125 |
return __builtin_convertvector((__v2hf)a, __v2qu); |
| 126 |
} |
| 127 |
|
| 128 |
__v4qu mm64_cvtph_epu8_builtin_convertvector(__v4hf a) |
| 129 |
{ |
| 130 |
return __builtin_convertvector((__v4hf)a, __v4qu); |
| 131 |
} |
| 132 |
|
| 133 |
__v8qu mm128_cvtph_epu8_builtin_convertvector(__v8hf a) |
| 134 |
{ |
| 135 |
return __builtin_convertvector((__v8hf)a, __v8qu); |
| 136 |
} |
| 137 |
|
| 138 |
__v16qu mm256_cvtph_epu8_builtin_convertvector(__v16hf a) |
| 139 |
{ |
| 140 |
return __builtin_convertvector((__v16hf)a, __v16qu); |
| 141 |
} |
| 142 |
|
| 143 |
__v32qu mm512_cvtph_epu8_builtin_convertvector(__v32hf a) |
| 144 |
{ |
| 145 |
return __builtin_convertvector((__v32hf)a, __v32qu); |
| 146 |
} |