| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-mavx512f -O2" } */ |
| 3 |
/* { dg-final { scan-assembler-times "vpexpandd\[ \\t\]+\[^\{\n\]*\\(\[^\n\]*%zmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */ |
| 4 |
/* { dg-final { scan-assembler-times "vpexpandd\[ \\t\]+\[^\{\n\(]*%zmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */ |
| 5 |
/* { dg-final { scan-assembler-times "vpexpandq\[ \\t\]+\[^\{\n\]*\\(\[^\n\]*%zmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */ |
| 6 |
/* { dg-final { scan-assembler-times "vpexpandq\[ \\t\]+\[^\{\n\(]*%zmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */ |
| 7 |
/* { dg-final { scan-assembler-times "vexpandpd\[ \\t\]+\[^\{\n\]*\\(\[^\n\]*%zmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */ |
| 8 |
/* { dg-final { scan-assembler-times "vexpandpd\[ \\t\]+\[^\{\n\(]*%zmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */ |
| 9 |
/* { dg-final { scan-assembler-times "vexpandps\[ \\t\]+\[^\{\n\]*\\(\[^\n\]*%zmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */ |
| 10 |
/* { dg-final { scan-assembler-times "vexpandps\[ \\t\]+\[^\{\n\(]*%zmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */ |
| 11 |
|
| 12 |
/* { dg-final { scan-assembler-times "vmov\[a-z0-9\]*\[ \\t\]+\[^\{\n\]*\\(\[^\n\]*%zmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 4 } } */ |
| 13 |
/* { dg-final { scan-assembler-times "vmov\[a-z0-9\]*\[ \\t\]+\[^\{\n\(]*%zmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 4 } } */ |
| 14 |
#include <immintrin.h> |
| 15 |
|
| 16 |
int *pi32; |
| 17 |
long long *pi64; |
| 18 |
double *pd; |
| 19 |
float *pf; |
| 20 |
volatile __m512i xi32, xi64; |
| 21 |
volatile __m512d xd; |
| 22 |
volatile __m512 xf; |
| 23 |
|
| 24 |
void extern |
| 25 |
avx512f_test (void) |
| 26 |
{ |
| 27 |
xi32 = _mm512_mask_expand_epi32 (xi32, 0, xi32); |
| 28 |
xi32 = _mm512_mask_expand_epi32 (xi32, -1, xi32); |
| 29 |
xi32 = _mm512_mask_expand_epi32 (xi32, (1 << 8) - 1, xi32); |
| 30 |
xi32 = _mm512_mask_expand_epi32 (xi32, (1 << 8) + 1, xi32); |
| 31 |
|
| 32 |
xi32 = _mm512_mask_expandloadu_epi32 (xi32, 0, pi32); |
| 33 |
xi32 = _mm512_mask_expandloadu_epi32 (xi32, (1 << 16) - 1, pi32); |
| 34 |
xi32 = _mm512_mask_expandloadu_epi32 (xi32, (1 << 6) - 1, pi32); |
| 35 |
xi32 = _mm512_mask_expandloadu_epi32 (xi32, (1 << 6) + 3, pi32); |
| 36 |
|
| 37 |
xi64 = _mm512_mask_expand_epi64 (xi64, 0, xi64); |
| 38 |
xi64 = _mm512_mask_expand_epi64 (xi64, -1, xi64); |
| 39 |
xi64 = _mm512_mask_expand_epi64 (xi64, (1 << 3) - 1, xi64); |
| 40 |
xi64 = _mm512_mask_expand_epi64 (xi64, (1 << 3) + 2, xi64); |
| 41 |
|
| 42 |
xi64 = _mm512_mask_expandloadu_epi64 (xi64, 0, pi64); |
| 43 |
xi64 = _mm512_mask_expandloadu_epi64 (xi64, (1 << 8) - 1, pi64); |
| 44 |
xi64 = _mm512_mask_expandloadu_epi64 (xi64, (1 << 7) - 1, pi64); |
| 45 |
xi64 = _mm512_mask_expandloadu_epi64 (xi64, (1 << 7) + 7, pi64); |
| 46 |
|
| 47 |
xf = _mm512_mask_expand_ps (xf, 0, xf); |
| 48 |
xf = _mm512_mask_expand_ps (xf, (1 << 16) - 1, xf); |
| 49 |
xf = _mm512_mask_expand_ps (xf, (1 << 15) - 1, xf); |
| 50 |
xf = _mm512_mask_expand_ps (xf, (1 << 14) + 3, xf); |
| 51 |
|
| 52 |
xf = _mm512_mask_expandloadu_ps (xf, 0, pf); |
| 53 |
xf = _mm512_mask_expandloadu_ps (xf, -1, pf); |
| 54 |
xf = _mm512_mask_expandloadu_ps (xf, (1 << 13) - 1, pf); |
| 55 |
xf = _mm512_mask_expandloadu_ps (xf, (1 << 13) + 5, pf); |
| 56 |
|
| 57 |
xd = _mm512_mask_expand_pd (xd, 0, xd); |
| 58 |
xd = _mm512_mask_expand_pd (xd, (1 << 8) - 1, xd); |
| 59 |
xd = _mm512_mask_expand_pd (xd, (1 << 4) - 1, xd); |
| 60 |
xd = _mm512_mask_expand_pd (xd, (1 << 4) + 1, xd); |
| 61 |
|
| 62 |
xd = _mm512_mask_expandloadu_pd (xd, 0, pd); |
| 63 |
xd = _mm512_mask_expandloadu_pd (xd, -1, pd); |
| 64 |
xd = _mm512_mask_expandloadu_pd (xd, (1 << 5) - 1, pd); |
| 65 |
xd = _mm512_mask_expandloadu_pd (xd, (1 << 5), pd); |
| 66 |
} |