| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-march=rv32gcv -mabi=ilp32 -O3" } */ |
| 3 |
/* { dg-final { check-function-bodies "**" "" } } */ |
| 4 |
#include "riscv_vector.h" |
| 5 |
|
| 6 |
/* |
| 7 |
** f0: |
| 8 |
** ... |
| 9 |
** vma[c-d][c-d]\.vx\tv[0-9]+,\s*[a-x0-9]+,\s*v[0-9]+ |
| 10 |
** vma[c-d][c-d]\.vx\tv[0-9]+,\s*[a-x0-9]+,\s*v[0-9]+ |
| 11 |
** ... |
| 12 |
** ret |
| 13 |
*/ |
| 14 |
void f0 (void * in, void *out, int64_t x, int n) |
| 15 |
{ |
| 16 |
vint64m1_t v = __riscv_vle64_v_i64m1 (in + 1, 4); |
| 17 |
vint64m1_t v2 = __riscv_vle64_v_i64m1_tu (v, in + 2, 4); |
| 18 |
vint64m1_t v3 = __riscv_vmacc_vx_i64m1_tu (v2, -16, v2, 4); |
| 19 |
vint64m1_t v4 = __riscv_vmacc_vx_i64m1_tu (v3, -16, v3, 4); |
| 20 |
__riscv_vse64_v_i64m1 (out + 2, v4, 4); |
| 21 |
} |
| 22 |
|
| 23 |
/* |
| 24 |
** f1: |
| 25 |
** ... |
| 26 |
** vma[c-d][c-d]\.vx\tv[0-9]+,\s*[a-x0-9]+,\s*v[0-9]+ |
| 27 |
** vma[c-d][c-d]\.vx\tv[0-9]+,\s*[a-x0-9]+,\s*v[0-9]+ |
| 28 |
** ... |
| 29 |
** ret |
| 30 |
*/ |
| 31 |
void f1 (void * in, void *out, int64_t x, int n) |
| 32 |
{ |
| 33 |
vint64m1_t v = __riscv_vle64_v_i64m1 (in + 1, 4); |
| 34 |
vint64m1_t v2 = __riscv_vle64_v_i64m1_tu (v, in + 2, 4); |
| 35 |
vint64m1_t v3 = __riscv_vmacc_vx_i64m1_tu (v2, 15, v2, 4); |
| 36 |
vint64m1_t v4 = __riscv_vmacc_vx_i64m1_tu (v3, 15, v3, 4); |
| 37 |
__riscv_vse64_v_i64m1 (out + 2, v4, 4); |
| 38 |
} |
| 39 |
|
| 40 |
/* |
| 41 |
** f2: |
| 42 |
** ... |
| 43 |
** vma[c-d][c-d]\.vx\tv[0-9]+,\s*[a-x0-9]+,\s*v[0-9]+ |
| 44 |
** vma[c-d][c-d]\.vx\tv[0-9]+,\s*[a-x0-9]+,\s*v[0-9]+ |
| 45 |
** ... |
| 46 |
** ret |
| 47 |
*/ |
| 48 |
void f2 (void * in, void *out, int64_t x, int n) |
| 49 |
{ |
| 50 |
vint64m1_t v = __riscv_vle64_v_i64m1 (in + 1, 4); |
| 51 |
vint64m1_t v2 = __riscv_vle64_v_i64m1_tu (v, in + 2, 4); |
| 52 |
vint64m1_t v3 = __riscv_vmacc_vx_i64m1_tu (v2, 16, v2, 4); |
| 53 |
vint64m1_t v4 = __riscv_vmacc_vx_i64m1_tu (v3, 16, v3, 4); |
| 54 |
__riscv_vse64_v_i64m1 (out + 2, v4, 4); |
| 55 |
} |
| 56 |
|
| 57 |
/* |
| 58 |
** f3: |
| 59 |
** ... |
| 60 |
** vma[c-d][c-d]\.vx\tv[0-9]+,\s*[a-x0-9]+,\s*v[0-9]+ |
| 61 |
** vma[c-d][c-d]\.vx\tv[0-9]+,\s*[a-x0-9]+,\s*v[0-9]+ |
| 62 |
** ... |
| 63 |
** ret |
| 64 |
*/ |
| 65 |
void f3 (void * in, void *out, int64_t x, int n) |
| 66 |
{ |
| 67 |
vint64m1_t v = __riscv_vle64_v_i64m1 (in + 1, 4); |
| 68 |
vint64m1_t v2 = __riscv_vle64_v_i64m1_tu (v, in + 2, 4); |
| 69 |
vint64m1_t v3 = __riscv_vmacc_vx_i64m1_tu (v2, 0xAAAAAA, v2, 4); |
| 70 |
vint64m1_t v4 = __riscv_vmacc_vx_i64m1_tu (v3, 0xAAAAAA, v3, 4); |
| 71 |
__riscv_vse64_v_i64m1 (out + 2, v4, 4); |
| 72 |
} |
| 73 |
|
| 74 |
/* |
| 75 |
** f4: |
| 76 |
** ... |
| 77 |
** vma[c-d][c-d]\.vx\tv[0-9]+,\s*[a-x0-9]+,\s*v[0-9]+ |
| 78 |
** vma[c-d][c-d]\.vx\tv[0-9]+,\s*[a-x0-9]+,\s*v[0-9]+ |
| 79 |
** ... |
| 80 |
** ret |
| 81 |
*/ |
| 82 |
void f4 (void * in, void *out, int64_t x, int n) |
| 83 |
{ |
| 84 |
vint64m1_t v = __riscv_vle64_v_i64m1 (in + 1, 4); |
| 85 |
vint64m1_t v2 = __riscv_vle64_v_i64m1_tu (v, in + 2, 4); |
| 86 |
vint64m1_t v3 = __riscv_vmacc_vx_i64m1_tu (v2, 0xAAAAAAA, v2, 4); |
| 87 |
vint64m1_t v4 = __riscv_vmacc_vx_i64m1_tu (v3, 0xAAAAAAA, v3, 4); |
| 88 |
__riscv_vse64_v_i64m1 (out + 2, v4, 4); |
| 89 |
} |
| 90 |
|
| 91 |
/* |
| 92 |
** f5: |
| 93 |
** ... |
| 94 |
** vma[c-d][c-d]\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ |
| 95 |
** vma[c-d][c-d]\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ |
| 96 |
** ... |
| 97 |
*/ |
| 98 |
void f5 (void * in, void *out, int64_t x, int n) |
| 99 |
{ |
| 100 |
vint64m1_t v = __riscv_vle64_v_i64m1 (in + 1, 4); |
| 101 |
vint64m1_t v2 = __riscv_vle64_v_i64m1_tu (v, in + 2, 4); |
| 102 |
vint64m1_t v3 = __riscv_vmacc_vx_i64m1_tu (v2, 0xAAAAAAAAAAAAAAAA, v2, 4); |
| 103 |
vint64m1_t v4 = __riscv_vmacc_vx_i64m1_tu (v3, 0xAAAAAAAAAAAAAAAA, v3, 4); |
| 104 |
__riscv_vse64_v_i64m1 (out + 2, v4, 4); |
| 105 |
} |
| 106 |
|
| 107 |
/* |
| 108 |
** f6: |
| 109 |
** ... |
| 110 |
** vma[c-d][c-d]\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ |
| 111 |
** vma[c-d][c-d]\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ |
| 112 |
** ... |
| 113 |
*/ |
| 114 |
void f6 (void * in, void *out, int64_t x, int n) |
| 115 |
{ |
| 116 |
vint64m1_t v = __riscv_vle64_v_i64m1 (in + 1, 4); |
| 117 |
vint64m1_t v2 = __riscv_vle64_v_i64m1_tu (v, in + 2, 4); |
| 118 |
vint64m1_t v3 = __riscv_vmacc_vx_i64m1_tu (v2, x, v2, 4); |
| 119 |
vint64m1_t v4 = __riscv_vmacc_vx_i64m1_tu (v3, x, v3, 4); |
| 120 |
__riscv_vse64_v_i64m1 (out + 2, v4, 4); |
| 121 |
} |
| 122 |
|
| 123 |
/* { dg-final { scan-assembler-not {vmv} } } */ |