| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-mrvv-vector-bits=scalable -march=rv32gcv -mabi=ilp32 -fno-schedule-insns -fno-schedule-insns2" } */ |
| 3 |
|
| 4 |
#include "riscv_vector.h" |
| 5 |
|
| 6 |
void f (void * restrict in, void * restrict out, void * restrict mask_in, int n, int cond) |
| 7 |
{ |
| 8 |
vbool64_t mask = *(vbool64_t*)mask_in; |
| 9 |
vfloat32mf2_t vf32mf2 = *(vfloat32mf2_t*)in; |
| 10 |
vint16mf4_t vf16mf4 = *(vint16mf4_t*)(in + 5); |
| 11 |
asm volatile ("":::"memory"); |
| 12 |
vfloat32mf2_t v = __riscv_vle32_v_f32mf2 (in + 10000, 5); |
| 13 |
__riscv_vse32_v_f32mf2 (out + 10000, v, 5); |
| 14 |
|
| 15 |
if (cond) |
| 16 |
{ |
| 17 |
vfloat32mf2_t vt = __riscv_vle32_v_f32mf2_tumu (mask, vf32mf2, in + 20000, 5); |
| 18 |
__riscv_vse32_v_f32mf2 (out + 20000, vt, 5); |
| 19 |
} |
| 20 |
else |
| 21 |
{ |
| 22 |
vint16mf4_t vt = __riscv_vle16_v_i16mf4_tumu (mask, vf16mf4, in + 20000, 5); |
| 23 |
__riscv_vse16_v_i16mf4 (out + 20000, vt, 5); |
| 24 |
} |
| 25 |
|
| 26 |
for (int i = 0; i < n; i++) |
| 27 |
{ |
| 28 |
vfloat32mf2_t v0 = __riscv_vle32_v_f32mf2_tu (v, in + i + 100, 5); |
| 29 |
vint16mf2_t v1 = __riscv_vle16_v_i16mf2 (in + i + 200, 6); |
| 30 |
vint8mf2_t v2 = __riscv_vle8_v_i8mf2 (in + i + 300, 7); |
| 31 |
vint8mf4_t v3 = __riscv_vle8_v_i8mf4 (in + i + 400, 8); |
| 32 |
vint8mf8_t v4 = __riscv_vle8_v_i8mf8 (in + i + 500, 9); |
| 33 |
vint32mf2_t v5 = __riscv_vle32_v_i32mf2 (in + i + 600, 5); |
| 34 |
|
| 35 |
__riscv_vse32_v_f32mf2 (out + i + 100, v0, 5); |
| 36 |
__riscv_vse16_v_i16mf2 (out + i + 200, v1, 6); |
| 37 |
__riscv_vse8_v_i8mf2 (out + i + 300, v2, 7); |
| 38 |
__riscv_vse8_v_i8mf4 (out + i + 400, v3, 8); |
| 39 |
__riscv_vse8_v_i8mf8 (out + i + 500, v4, 9); |
| 40 |
__riscv_vse32_v_i32mf2 (out + i + 600, v5, 5); |
| 41 |
} |
| 42 |
} |
| 43 |
|
| 44 |
/* { dg-final { scan-assembler-times {vsetivli} 11 { target { no-opts "-O0" no-opts "-O1" no-opts "-Os" no-opts "-Oz" no-opts "-g" no-opts "-funroll-loops" } } } } */ |
| 45 |
|