| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-O2 -msse2 -mno-avx" } */ |
| 3 |
|
| 4 |
void foo (int * __restrict a, int *b, int s) |
| 5 |
{ |
| 6 |
for (int i = 0; i < 1024; ++i) |
| 7 |
{ |
| 8 |
a[8*i+0] = b[s*i+0]; |
| 9 |
a[8*i+1] = b[s*i+1]; |
| 10 |
a[8*i+2] = b[s*i+2]; |
| 11 |
a[8*i+3] = b[s*i+3]; |
| 12 |
a[8*i+4] = b[s*i+4]; |
| 13 |
a[8*i+5] = b[s*i+5]; |
| 14 |
a[8*i+6] = b[s*i+4]; |
| 15 |
a[8*i+7] = b[s*i+5]; |
| 16 |
} |
| 17 |
} |
| 18 |
|
| 19 |
/* Three two-element loads, two four-element stores. On ia32 we elide |
| 20 |
a permute and perform a redundant load. */ |
| 21 |
/* { dg-final { scan-assembler-times "movq" 2 } } */ |
| 22 |
/* { dg-final { scan-assembler-times "movhps" 2 { target ia32 } } } */ |
| 23 |
/* { dg-final { scan-assembler-times "movhps" 1 { target { ! ia32 } } } } */ |
| 24 |
/* { dg-final { scan-assembler-times "movups" 2 } } */ |