| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-march=rv32gcv -mabi=ilp32d -O3 -fno-schedule-insns -fno-schedule-insns2" } */ |
| 3 |
/* { dg-final { check-function-bodies "**" "" } } */ |
| 4 |
|
| 5 |
#include "riscv_vector.h" |
| 6 |
|
| 7 |
/* |
| 8 |
** f1: |
| 9 |
** ... |
| 10 |
** vslide1down\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 11 |
** ... |
| 12 |
** vslide1down\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 13 |
** ... |
| 14 |
** vslide1down\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 15 |
** ... |
| 16 |
** vslide1down\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 17 |
** ... |
| 18 |
** vslide1down\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 19 |
** ... |
| 20 |
** vslide1down\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 21 |
** ... |
| 22 |
** vmerge\.vvm\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+,\s*v[0-9]+ |
| 23 |
** ... |
| 24 |
** ret |
| 25 |
*/ |
| 26 |
void f1 (void * in, void *out, int64_t x, int vl) |
| 27 |
{ |
| 28 |
vbool64_t m = __riscv_vlm_v_b64 (in, vl); |
| 29 |
vint64m1_t v = __riscv_vle64_v_i64m1 (in + 1, vl); |
| 30 |
vint64m1_t v2 = __riscv_vle64_v_i64m1_tu (v, in + 2, vl); |
| 31 |
vint64m1_t v3 = __riscv_vslide1down_vx_i64m1 (v2, x, 0x80000000); |
| 32 |
vint64m1_t v4 = __riscv_vslide1down_vx_i64m1_tu (v3, v3, x, 0x80000000); |
| 33 |
vint64m1_t v5 = __riscv_vslide1down_vx_i64m1_tumu (m, v4, v4, x, 0x80000000); |
| 34 |
__riscv_vse64_v_i64m1 (out + 2, v5, vl); |
| 35 |
} |
| 36 |
|
| 37 |
/* |
| 38 |
** f2: |
| 39 |
** ... |
| 40 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 41 |
** ... |
| 42 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 43 |
** ... |
| 44 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 45 |
** ... |
| 46 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 47 |
** ... |
| 48 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 49 |
** ... |
| 50 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 51 |
** ... |
| 52 |
** vmerge\.vvm\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+,\s*v[0-9]+ |
| 53 |
** ... |
| 54 |
** ret |
| 55 |
*/ |
| 56 |
void f2 (void * in, void *out, int64_t x, int vl) |
| 57 |
{ |
| 58 |
vbool64_t m = __riscv_vlm_v_b64 (in, vl); |
| 59 |
vint64m1_t v = __riscv_vle64_v_i64m1 (in + 1, vl); |
| 60 |
vint64m1_t v2 = __riscv_vle64_v_i64m1_tu (v, in + 2, vl); |
| 61 |
vint64m1_t v3 = __riscv_vslide1up_vx_i64m1 (v2, x, 0x80000000); |
| 62 |
vint64m1_t v4 = __riscv_vslide1up_vx_i64m1_tu (v3, v3, x, 0x80000000); |
| 63 |
vint64m1_t v5 = __riscv_vslide1up_vx_i64m1_tumu (m, v4, v4, x, 0x80000000); |
| 64 |
__riscv_vse64_v_i64m1 (out + 2, v5, vl); |
| 65 |
} |
| 66 |
|
| 67 |
/* { dg-final { scan-assembler-times {vmv} 3 } } */ |