| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-march=rv64gcv -mabi=lp64d -O3" } */ |
| 3 |
|
| 4 |
#include "riscv_vector.h" |
| 5 |
|
| 6 |
void |
| 7 |
foo (uint32_t *outputMat, uint32_t *inputMat) |
| 8 |
{ |
| 9 |
vuint32m1_t matRegIn0 = __riscv_vle32_v_u32m1 (inputMat, 4); |
| 10 |
vuint32m1_t matRegIn1 = __riscv_vle32_v_u32m1 (inputMat + 4, 4); |
| 11 |
vuint32m1_t matRegIn2 = __riscv_vle32_v_u32m1 (inputMat + 8, 4); |
| 12 |
vuint32m1_t matRegIn3 = __riscv_vle32_v_u32m1 (inputMat + 12, 4); |
| 13 |
|
| 14 |
vbool32_t oddMask |
| 15 |
= __riscv_vreinterpret_v_u32m1_b32 (__riscv_vmv_v_x_u32m1 (0xaaaa, 1)); |
| 16 |
|
| 17 |
vuint32m1_t smallTransposeMat0 |
| 18 |
= __riscv_vslideup_vx_u32m1_tumu (oddMask, matRegIn0, matRegIn1, 1, 4); |
| 19 |
vuint32m1_t smallTransposeMat2 |
| 20 |
= __riscv_vslideup_vx_u32m1_tumu (oddMask, matRegIn2, matRegIn3, 1, 4); |
| 21 |
|
| 22 |
vuint32m1_t outMat0 = __riscv_vslideup_vx_u32m1_tu (smallTransposeMat0, |
| 23 |
smallTransposeMat2, 2, 4); |
| 24 |
|
| 25 |
__riscv_vse32_v_u32m1 (outputMat, outMat0, 4); |
| 26 |
} |
| 27 |
|
| 28 |
void |
| 29 |
foo2 (void *outputMat, void *inputMat) |
| 30 |
{ |
| 31 |
vfloat32m1_t v = __riscv_vfmv_v_f_f32m1 (0xaaaa, 1); |
| 32 |
__riscv_vse32_v_f32m1 (outputMat, v, 4); |
| 33 |
} |
| 34 |
|
| 35 |
/* { dg-final { scan-assembler-times {vsetivli\s+zero,\s*4,\s*e32,\s*m1,\s*t[au],\s*m[au]} 2 } } */ |
| 36 |
/* { dg-final { scan-assembler-times {vsetivli} 2 } } */ |
| 37 |
/* { dg-final { scan-assembler-not {vsetvli} } } */ |