| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-O3 -mcpu=cortex-a55+aes" } */ |
| 3 |
#include <arm_neon.h> |
| 4 |
|
| 5 |
#define AESE(r, v, key) (r = vaeseq_u8 ((v), (key))); |
| 6 |
#define AESD(r, v, key) (r = vaesdq_u8 ((v), (key))); |
| 7 |
|
| 8 |
const uint8x16_t zero = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; |
| 9 |
|
| 10 |
uint8x16_t foo0 (uint8x16_t a, uint8x16_t b) |
| 11 |
{ |
| 12 |
uint8x16_t dummy; |
| 13 |
AESE(dummy, a ^ b, zero); |
| 14 |
return dummy; |
| 15 |
} |
| 16 |
|
| 17 |
uint8x16_t foo1 (uint8x16_t a, uint8x16_t b) |
| 18 |
{ |
| 19 |
uint8x16_t dummy; |
| 20 |
AESE(dummy, a ^ b, zero); |
| 21 |
AESE(dummy, dummy ^ a, zero); |
| 22 |
return dummy; |
| 23 |
} |
| 24 |
|
| 25 |
uint8x16_t bar0 (uint8x16_t a, uint8x16_t b) |
| 26 |
{ |
| 27 |
uint8x16_t dummy; |
| 28 |
AESE(dummy, zero, a ^ b); |
| 29 |
return dummy; |
| 30 |
} |
| 31 |
|
| 32 |
uint8x16_t bar1 (uint8x16_t a, uint8x16_t b) |
| 33 |
{ |
| 34 |
uint8x16_t dummy; |
| 35 |
AESE(dummy, zero, a ^ b); |
| 36 |
AESE(dummy, zero, b ^ dummy); |
| 37 |
return dummy; |
| 38 |
} |
| 39 |
|
| 40 |
uint8x16_t foo2 (uint8x16_t a, uint8x16_t b) |
| 41 |
{ |
| 42 |
uint8x16_t dummy; |
| 43 |
AESD(dummy, a ^ b, zero); |
| 44 |
return dummy; |
| 45 |
} |
| 46 |
|
| 47 |
uint8x16_t foo3 (uint8x16_t a, uint8x16_t b) |
| 48 |
{ |
| 49 |
uint8x16_t dummy; |
| 50 |
AESD(dummy, a ^ b, zero); |
| 51 |
AESD(dummy, dummy ^ a, zero); |
| 52 |
return dummy; |
| 53 |
} |
| 54 |
|
| 55 |
uint8x16_t bar2 (uint8x16_t a, uint8x16_t b) |
| 56 |
{ |
| 57 |
uint8x16_t dummy; |
| 58 |
AESD(dummy, zero, a ^ b); |
| 59 |
return dummy; |
| 60 |
} |
| 61 |
|
| 62 |
uint8x16_t bar3 (uint8x16_t a, uint8x16_t b) |
| 63 |
{ |
| 64 |
uint8x16_t dummy; |
| 65 |
AESD(dummy, zero, a ^ b); |
| 66 |
AESD(dummy, zero, b ^ dummy); |
| 67 |
return dummy; |
| 68 |
} |
| 69 |
/* { dg-final { scan-assembler-not "eor" } } */ |
| 70 |
/* { dg-final { scan-assembler-not "mov" } } */ |