| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-mavx512f -mavx512vl -O2" } */ |
| 3 |
/* { dg-final { scan-assembler-times "vpexpandd\[ \\t\]+\[^\{\n\]*\\(\[^\n\]*%ymm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */ |
| 4 |
/* { dg-final { scan-assembler-times "vpexpandd\[ \\t\]+\[^\{\n\(]*%ymm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */ |
| 5 |
/* { dg-final { scan-assembler-times "vpexpandq\[ \\t\]+\[^\{\n\]*\\(\[^\n\]*%ymm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */ |
| 6 |
/* { dg-final { scan-assembler-times "vpexpandq\[ \\t\]+\[^\{\n\(]*%ymm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */ |
| 7 |
/* { dg-final { scan-assembler-times "vexpandpd\[ \\t\]+\[^\{\n\]*\\(\[^\n\]*%ymm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */ |
| 8 |
/* { dg-final { scan-assembler-times "vexpandpd\[ \\t\]+\[^\{\n\(]*%ymm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */ |
| 9 |
/* { dg-final { scan-assembler-times "vexpandps\[ \\t\]+\[^\{\n\]*\\(\[^\n\]*%ymm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */ |
| 10 |
/* { dg-final { scan-assembler-times "vexpandps\[ \\t\]+\[^\{\n\(]*%ymm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */ |
| 11 |
|
| 12 |
/* { dg-final { scan-assembler-times "vmov\[a-z0-9\]*\[ \\t\]+\[^\{\n\]*\\(\[^\n\]*%ymm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 4 } } */ |
| 13 |
/* { dg-final { scan-assembler-times "(?:vmov\[a-z0-9\]*\[ \\t\]+\[^\{\n\(]*%ymm\[0-9\]+\{%k\[1-7\]\}|blend\[a-z]+\[ \\t\]+\[^\{\n\(]*%ymm\[0-9\]+)(?:\n|\[ \\t\]+#)" 4 } } */ |
| 14 |
/* { dg-final { scan-assembler-times "vpexpandd\[ \\t\]+\[^\{\n\]*\\(\[^\n\]*%xmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */ |
| 15 |
/* { dg-final { scan-assembler-times "vpexpandd\[ \\t\]+\[^\{\n\(]*%xmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */ |
| 16 |
/* { dg-final { scan-assembler-times "vpexpandq\[ \\t\]+\[^\{\n\]*\\(\[^\n\]*%xmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */ |
| 17 |
/* { dg-final { scan-assembler-times "vpexpandq\[ \\t\]+\[^\{\n\(]*%xmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */ |
| 18 |
/* { dg-final { scan-assembler-times "vexpandpd\[ \\t\]+\[^\{\n\]*\\(\[^\n\]*%xmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */ |
| 19 |
/* { dg-final { scan-assembler-times "vexpandpd\[ \\t\]+\[^\{\n\(]*%xmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */ |
| 20 |
/* { dg-final { scan-assembler-times "vexpandps\[ \\t\]+\[^\{\n\]*\\(\[^\n\]*%xmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */ |
| 21 |
/* { dg-final { scan-assembler-times "vexpandps\[ \\t\]+\[^\{\n\(]*%xmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */ |
| 22 |
|
| 23 |
/* { dg-final { scan-assembler-times "vmov\[a-z0-9\]*\[ \\t\]+\[^\{\n\]*\\(\[^\n\]*%xmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 4 } } */ |
| 24 |
/* { dg-final { scan-assembler-times "(?:vmov\[a-z0-9\]*\[ \\t\]+\[^\{\n\(]*%xmm\[0-9\]+\{%k\[1-7\]\}|(?:blend\[a-z]+|movsd)\[ \\t\]+\[^\{\n\(]*%xmm\[0-9\]+)(?:\n|\[ \\t\]+#)" 4 } } */ |
| 25 |
|
| 26 |
#include <immintrin.h> |
| 27 |
|
| 28 |
int *pi32; |
| 29 |
long long *pi64; |
| 30 |
double *pd; |
| 31 |
float *pf; |
| 32 |
volatile __m256i xi32, xi64; |
| 33 |
volatile __m256d xd; |
| 34 |
volatile __m256 xf; |
| 35 |
|
| 36 |
volatile __m128i xi32_128, xi64_128; |
| 37 |
volatile __m128d xd_128; |
| 38 |
volatile __m128 xf_128; |
| 39 |
|
| 40 |
void extern |
| 41 |
avx512vl_test (void) |
| 42 |
{ |
| 43 |
xi32 = _mm256_mask_expand_epi32 (xi32, 0, xi32); |
| 44 |
xi32 = _mm256_mask_expand_epi32 (xi32, -1, xi32); |
| 45 |
xi32 = _mm256_mask_expand_epi32 (xi32, (1 << 4) - 1, xi32); |
| 46 |
xi32 = _mm256_mask_expand_epi32 (xi32, (1 << 4) + 1, xi32); |
| 47 |
|
| 48 |
xi32 = _mm256_mask_expandloadu_epi32 (xi32, 0, pi32); |
| 49 |
xi32 = _mm256_mask_expandloadu_epi32 (xi32, (1 << 8) - 1, pi32); |
| 50 |
xi32 = _mm256_mask_expandloadu_epi32 (xi32, (1 << 6) - 1, pi32); |
| 51 |
xi32 = _mm256_mask_expandloadu_epi32 (xi32, (1 << 6) + 3, pi32); |
| 52 |
|
| 53 |
xi64 = _mm256_mask_expand_epi64 (xi64, 0, xi64); |
| 54 |
xi64 = _mm256_mask_expand_epi64 (xi64, -1, xi64); |
| 55 |
xi64 = _mm256_mask_expand_epi64 (xi64, (1 << 3) - 1, xi64); |
| 56 |
xi64 = _mm256_mask_expand_epi64 (xi64, (1 << 3) + 2, xi64); |
| 57 |
|
| 58 |
xi64 = _mm256_mask_expandloadu_epi64 (xi64, 0, pi64); |
| 59 |
xi64 = _mm256_mask_expandloadu_epi64 (xi64, (1 << 4) - 1, pi64); |
| 60 |
xi64 = _mm256_mask_expandloadu_epi64 (xi64, (1 << 2) - 1, pi64); |
| 61 |
xi64 = _mm256_mask_expandloadu_epi64 (xi64, (1 << 2), pi64); |
| 62 |
|
| 63 |
xf = _mm256_mask_expand_ps (xf, 0, xf); |
| 64 |
xf = _mm256_mask_expand_ps (xf, (1 << 8) - 1, xf); |
| 65 |
xf = _mm256_mask_expand_ps (xf, (1 << 6) - 1, xf); |
| 66 |
xf = _mm256_mask_expand_ps (xf, (1 << 6) + 3, xf); |
| 67 |
|
| 68 |
xf = _mm256_mask_expandloadu_ps (xf, 0, pf); |
| 69 |
xf = _mm256_mask_expandloadu_ps (xf, -1, pf); |
| 70 |
xf = _mm256_mask_expandloadu_ps (xf, (1 << 7) - 1, pf); |
| 71 |
xf = _mm256_mask_expandloadu_ps (xf, (1 << 7) + 5, pf); |
| 72 |
|
| 73 |
xd = _mm256_mask_expand_pd (xd, 0, xd); |
| 74 |
xd = _mm256_mask_expand_pd (xd, (1 << 4) - 1, xd); |
| 75 |
xd = _mm256_mask_expand_pd (xd, (1 << 2) - 1, xd); |
| 76 |
xd = _mm256_mask_expand_pd (xd, (1 << 2), xd); |
| 77 |
|
| 78 |
xd = _mm256_mask_expandloadu_pd (xd, 0, pd); |
| 79 |
xd = _mm256_mask_expandloadu_pd (xd, -1, pd); |
| 80 |
xd = _mm256_mask_expandloadu_pd (xd, (1 << 2) - 1, pd); |
| 81 |
xd = _mm256_mask_expandloadu_pd (xd, (1 << 2), pd); |
| 82 |
|
| 83 |
xi32_128 = _mm_mask_expand_epi32 (xi32_128, 0, xi32_128); |
| 84 |
xi32_128 = _mm_mask_expand_epi32 (xi32_128, -1, xi32_128); |
| 85 |
xi32_128 = _mm_mask_expand_epi32 (xi32_128, (1 << 3) - 1, xi32_128); |
| 86 |
xi32_128 = _mm_mask_expand_epi32 (xi32_128, (1 << 3) + 1, xi32_128); |
| 87 |
|
| 88 |
xi32_128 = _mm_mask_expandloadu_epi32 (xi32_128, 0, pi32); |
| 89 |
xi32_128 = _mm_mask_expandloadu_epi32 (xi32_128, (1 << 4) - 1, pi32); |
| 90 |
xi32_128 = _mm_mask_expandloadu_epi32 (xi32_128, (1 << 2) - 1, pi32); |
| 91 |
xi32_128 = _mm_mask_expandloadu_epi32 (xi32_128, (1 << 1) + 3, pi32); |
| 92 |
|
| 93 |
xi64_128 = _mm_mask_expand_epi64 (xi64_128, 0, xi64_128); |
| 94 |
xi64_128 = _mm_mask_expand_epi64 (xi64_128, -1, xi64_128); |
| 95 |
xi64_128 = _mm_mask_expand_epi64 (xi64_128, (1 << 1) - 1, xi64_128); |
| 96 |
xi64_128 = _mm_mask_expand_epi64 (xi64_128, 2, xi64_128); |
| 97 |
|
| 98 |
xi64_128 = _mm_mask_expandloadu_epi64 (xi64_128, 0, pi64); |
| 99 |
xi64_128 = _mm_mask_expandloadu_epi64 (xi64_128, 3, pi64); |
| 100 |
xi64_128 = _mm_mask_expandloadu_epi64 (xi64_128, 1, pi64); |
| 101 |
xi64_128 = _mm_mask_expandloadu_epi64 (xi64_128, 2, pi64); |
| 102 |
|
| 103 |
xf_128 = _mm_mask_expand_ps (xf_128, 0, xf_128); |
| 104 |
xf_128 = _mm_mask_expand_ps (xf_128, (1 << 4) - 1, xf_128); |
| 105 |
xf_128 = _mm_mask_expand_ps (xf_128, (1 << 3) - 1, xf_128); |
| 106 |
xf_128 = _mm_mask_expand_ps (xf_128, (1 << 2), xf_128); |
| 107 |
|
| 108 |
xf_128 = _mm_mask_expandloadu_ps (xf_128, 0, pf); |
| 109 |
xf_128 = _mm_mask_expandloadu_ps (xf_128, -1, pf); |
| 110 |
xf_128 = _mm_mask_expandloadu_ps (xf_128, (1 << 3) - 1, pf); |
| 111 |
xf_128 = _mm_mask_expandloadu_ps (xf_128, (1 << 1), pf); |
| 112 |
|
| 113 |
xd_128 = _mm_mask_expand_pd (xd_128, 0, xd_128); |
| 114 |
xd_128 = _mm_mask_expand_pd (xd_128, (1 << 2) - 1, xd_128); |
| 115 |
xd_128 = _mm_mask_expand_pd (xd_128, 1, xd_128); |
| 116 |
xd_128 = _mm_mask_expand_pd (xd_128, 2, xd_128); |
| 117 |
|
| 118 |
xd_128 = _mm_mask_expandloadu_pd (xd_128, 0, pd); |
| 119 |
xd_128 = _mm_mask_expandloadu_pd (xd_128, -1, pd); |
| 120 |
xd_128 = _mm_mask_expandloadu_pd (xd_128, 1, pd); |
| 121 |
xd_128 = _mm_mask_expandloadu_pd (xd_128, 2, pd); |
| 122 |
} |