| 1 |
/* { dg-do compile { target { ! riscv_abi_e } } } */ |
| 2 |
/* { dg-additional-options "-O1 -fno-schedule-insns -fno-schedule-insns2 -mrvv-max-lmul=m8" } */ |
| 3 |
/* { dg-add-options riscv_v } */ |
| 4 |
/* { dg-final { check-function-bodies "**" "" } } */ |
| 5 |
|
| 6 |
#define MIN_VECTOR_BYTES (__riscv_v_min_vlen / 8) |
| 7 |
|
| 8 |
/* Check that vector memcpy with predicated store uses smaller LMUL where |
| 9 |
possible. |
| 10 |
|
| 11 |
/* m1 |
| 12 |
** f1: |
| 13 |
** ( |
| 14 |
** vsetivli\s+zero,\d+,e8,m1,ta,ma |
| 15 |
** | |
| 16 |
** li\s+[ta][0-7],\d+ |
| 17 |
** vsetvli\s+zero,[ta][0-7],e8,m1,ta,ma |
| 18 |
** ) |
| 19 |
** vle8.v\s+v\d+,0\(a1\) |
| 20 |
** vse8.v\s+v\d+,0\(a0\) |
| 21 |
** ret |
| 22 |
*/ |
| 23 |
|
| 24 |
void f1 (char *d, char *s) |
| 25 |
{ |
| 26 |
__builtin_memcpy (d, s, MIN_VECTOR_BYTES - 1); |
| 27 |
} |
| 28 |
|
| 29 |
/* m2 |
| 30 |
** f2: |
| 31 |
** ( |
| 32 |
** vsetivli\s+zero,\d+,e8,m2,ta,ma |
| 33 |
** | |
| 34 |
** li\s+[ta][0-7],\d+ |
| 35 |
** vsetvli\s+zero,[ta][0-7],e8,m2,ta,ma |
| 36 |
** ) |
| 37 |
** vle8.v\s+v\d+,0\(a1\) |
| 38 |
** vse8.v\s+v\d+,0\(a0\) |
| 39 |
** ret |
| 40 |
*/ |
| 41 |
|
| 42 |
void f2 (char *d, char *s) |
| 43 |
{ |
| 44 |
__builtin_memcpy (d, s, 2 * MIN_VECTOR_BYTES - 1); |
| 45 |
} |
| 46 |
|
| 47 |
/* m4 |
| 48 |
** f3: |
| 49 |
** ( |
| 50 |
** vsetivli\s+zero,\d+,e8,m4,ta,ma |
| 51 |
** | |
| 52 |
** li\s+[ta][0-7],\d+ |
| 53 |
** vsetvli\s+zero,[ta][0-7],e8,m4,ta,ma |
| 54 |
** ) |
| 55 |
** vle8.v\s+v\d+,0\(a1\) |
| 56 |
** vse8.v\s+v\d+,0\(a0\) |
| 57 |
** ret |
| 58 |
*/ |
| 59 |
|
| 60 |
void f3 (char *d, char *s) |
| 61 |
{ |
| 62 |
__builtin_memcpy (d, s, 4 * MIN_VECTOR_BYTES - 1); |
| 63 |
} |
| 64 |
|
| 65 |
/* m8 |
| 66 |
** f4: |
| 67 |
** ( |
| 68 |
** vsetivli\s+zero,\d+,e8,m8,ta,ma |
| 69 |
** | |
| 70 |
** li\s+[ta][0-7],\d+ |
| 71 |
** vsetvli\s+zero,[ta][0-7],e8,m8,ta,ma |
| 72 |
** | |
| 73 |
** li\s+[ta][0-7],\d+ |
| 74 |
** addi\s+[ta][0-7],[ta][0-7],-?\d+ |
| 75 |
** vsetvli\s+zero,[ta][0-7],e8,m8,ta,ma |
| 76 |
** ) |
| 77 |
** vle8.v\s+v\d+,0\(a1\) |
| 78 |
** vse8.v\s+v\d+,0\(a0\) |
| 79 |
** ret |
| 80 |
*/ |
| 81 |
|
| 82 |
void f4 (char *d, char *s) |
| 83 |
{ |
| 84 |
__builtin_memcpy (d, s, 8 * MIN_VECTOR_BYTES - 1); |
| 85 |
} |