| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-march=x86-64 -mavx512bw -mavx512vl -O3" } */ |
| 3 |
/* { dg-final { scan-assembler-times "vpmovsxdq" 3 } } */ |
| 4 |
/* { dg-final { scan-assembler-times "vpmovsxwq" 3 } } */ |
| 5 |
/* { dg-final { scan-assembler-times "vpmovsxbq" 3 } } */ |
| 6 |
/* { dg-final { scan-assembler-times "vpmovsxwd" 3 } } */ |
| 7 |
/* { dg-final { scan-assembler-times "vpmovsxbd" 3 } } */ |
| 8 |
/* { dg-final { scan-assembler-times "vpmovsxbw" 3 } } */ |
| 9 |
|
| 10 |
#include <x86intrin.h> |
| 11 |
|
| 12 |
typedef short __v2hi __attribute__ ((__vector_size__ (4))); |
| 13 |
typedef char __v2qi __attribute__ ((__vector_size__ (2))); |
| 14 |
typedef char __v4qi __attribute__ ((__vector_size__ (4))); |
| 15 |
typedef char __v8qi __attribute__ ((__vector_size__ (8))); |
| 16 |
|
| 17 |
__m128i mm_cvtepi32_epi64_builtin_convertvector(__v2si a) |
| 18 |
{ |
| 19 |
return __builtin_convertvector(a, __v2di); |
| 20 |
} |
| 21 |
|
| 22 |
__m256i mm256_cvtepi32_epi64_builtin_convertvector(__v4si a) |
| 23 |
{ |
| 24 |
return (__m256i)__builtin_convertvector(a, __v4di); |
| 25 |
} |
| 26 |
|
| 27 |
__m512i mm512_cvtepi32_epi64_builtin_convertvector(__v8si a) |
| 28 |
{ |
| 29 |
return (__m512i)__builtin_convertvector(a, __v8di); |
| 30 |
} |
| 31 |
|
| 32 |
__m128i mm_cvtepi16_epi64_builtin_convertvector(__v2hi a) |
| 33 |
{ |
| 34 |
return __builtin_convertvector(a, __v2di); |
| 35 |
} |
| 36 |
|
| 37 |
__m256i mm256_cvtepi16_epi64_builtin_convertvector(__v4hi a) |
| 38 |
{ |
| 39 |
return (__m256i)__builtin_convertvector(a, __v4di); |
| 40 |
} |
| 41 |
|
| 42 |
__m512i mm512_cvtepi16_epi64_builtin_convertvector(__v8hi a) |
| 43 |
{ |
| 44 |
return (__m512i)__builtin_convertvector(a, __v8di); |
| 45 |
} |
| 46 |
|
| 47 |
__m128i mm_cvtepi8_epi64_builtin_convertvector(__v2qi a) |
| 48 |
{ |
| 49 |
return __builtin_convertvector(a, __v2di); |
| 50 |
} |
| 51 |
|
| 52 |
__m256i mm256_cvtepi8_epi64_builtin_convertvector(__v4qi a) |
| 53 |
{ |
| 54 |
return (__m256i)__builtin_convertvector(a, __v4di); |
| 55 |
} |
| 56 |
|
| 57 |
__m512i mm512_cvtepi8_epi64_builtin_convertvector(__v8qi a) |
| 58 |
{ |
| 59 |
return (__m512i)__builtin_convertvector(a, __v8di); |
| 60 |
} |
| 61 |
|
| 62 |
__m128i mm_cvtepi16_epi32_builtin_convertvector(__v4hi a) |
| 63 |
{ |
| 64 |
return (__m128i)__builtin_convertvector(a, __v4si); |
| 65 |
} |
| 66 |
|
| 67 |
__m256i mm256_cvtepi16_epi32_builtin_convertvector(__v8hi a) |
| 68 |
{ |
| 69 |
return (__m256i)__builtin_convertvector(a, __v8si); |
| 70 |
} |
| 71 |
|
| 72 |
__m512i mm512_cvtepi16_epi32_builtin_convertvector(__v16hi a) |
| 73 |
{ |
| 74 |
return (__m512i)__builtin_convertvector(a, __v16si); |
| 75 |
} |
| 76 |
|
| 77 |
__m128i mm_cvtepi8_epi32_builtin_convertvector(__v4qi a) |
| 78 |
{ |
| 79 |
return (__m128i)__builtin_convertvector(a, __v4si); |
| 80 |
} |
| 81 |
|
| 82 |
__m256i mm256_cvtepi8_epi32_builtin_convertvector(__v8qi a) |
| 83 |
{ |
| 84 |
return (__m256i)__builtin_convertvector(a, __v8si); |
| 85 |
} |
| 86 |
|
| 87 |
__m512i mm512_cvtepi8_epi32_builtin_convertvector(__v16qi a) |
| 88 |
{ |
| 89 |
return (__m512i)__builtin_convertvector(a, __v16si); |
| 90 |
} |
| 91 |
|
| 92 |
__m128i mm_cvtepi8_epi16_builtin_convertvector(__v8qi a) |
| 93 |
{ |
| 94 |
return (__m128i)__builtin_convertvector(a, __v8hi); |
| 95 |
} |
| 96 |
|
| 97 |
__m256i mm256_cvtepi8_epi16_builtin_convertvector(__v16qi a) |
| 98 |
{ |
| 99 |
return (__m256i)__builtin_convertvector(a, __v16hi); |
| 100 |
} |
| 101 |
|
| 102 |
__v32hi mm512_cvtepi8_epi16_builtin_convertvector(__v32qi a) |
| 103 |
{ |
| 104 |
return __builtin_convertvector(a, __v32hi); |
| 105 |
} |