| 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 |
#include "riscv_vector.h" |
| 5 |
|
| 6 |
/* |
| 7 |
** f1: |
| 8 |
** ... |
| 9 |
** vmseq\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 10 |
** vmseq\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+,v0.t |
| 11 |
** ... |
| 12 |
** ret |
| 13 |
*/ |
| 14 |
void f1 (void * in, void * in2, void *out, int32_t x) |
| 15 |
{ |
| 16 |
vint32m1_t v = __riscv_vle32_v_i32m1 (in, 4); |
| 17 |
vint32m1_t v2 = __riscv_vle32_v_i32m1 (in2, 4); |
| 18 |
vbool32_t m3 = __riscv_vmseq_vx_i32m1_b32 (v, 16, 4); |
| 19 |
vbool32_t m4 = __riscv_vmseq_vx_i32m1_b32_mu (m3, m3, v2, 16, 4); |
| 20 |
__riscv_vsm_v_b32 (out, m4, 4); |
| 21 |
} |
| 22 |
|
| 23 |
/* |
| 24 |
** f2: |
| 25 |
** ... |
| 26 |
** vmseq\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 27 |
** vmseq\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t |
| 28 |
** ... |
| 29 |
** ret |
| 30 |
*/ |
| 31 |
void f2 (void * in, void *out, int32_t x) |
| 32 |
{ |
| 33 |
vbool32_t mask = *(vbool32_t*)in; |
| 34 |
asm volatile ("":::"memory"); |
| 35 |
vint32m1_t v = __riscv_vle32_v_i32m1 (in, 4); |
| 36 |
vint32m1_t v2 = __riscv_vle32_v_i32m1_m (mask, in, 4); |
| 37 |
vbool32_t m3 = __riscv_vmseq_vx_i32m1_b32 (v, 16, 4); |
| 38 |
vbool32_t m4 = __riscv_vmseq_vx_i32m1_b32_mu (mask, m3, v2, 16, 4); |
| 39 |
__riscv_vsm_v_b32 (out, m4, 4); |
| 40 |
} |
| 41 |
|
| 42 |
/* |
| 43 |
** f3: |
| 44 |
** ... |
| 45 |
** vmseq\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 46 |
** vmseq\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t |
| 47 |
** ... |
| 48 |
** ret |
| 49 |
*/ |
| 50 |
void f3 (void * in, void *out, int32_t x) |
| 51 |
{ |
| 52 |
vbool32_t mask = *(vbool32_t*)in; |
| 53 |
asm volatile ("":::"memory"); |
| 54 |
vint32m1_t v = __riscv_vle32_v_i32m1 (in, 4); |
| 55 |
vint32m1_t v2 = __riscv_vle32_v_i32m1_m (mask, in, 4); |
| 56 |
vbool32_t m3 = __riscv_vmseq_vx_i32m1_b32 (v, 16, 4); |
| 57 |
vbool32_t m4 = __riscv_vmseq_vx_i32m1_b32_m (m3, v2, 16, 4); |
| 58 |
__riscv_vsm_v_b32 (out, m4, 4); |
| 59 |
} |