| 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 |
** f0: |
| 9 |
** ... |
| 10 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 11 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 12 |
** ... |
| 13 |
** ret |
| 14 |
*/ |
| 15 |
void f0 (void * in, void *out, int64_t x, int n) |
| 16 |
{ |
| 17 |
vint64m1_t v = __riscv_vle64_v_i64m1 (in + 1, 4); |
| 18 |
vint64m1_t v2 = __riscv_vle64_v_i64m1_tu (v, in + 2, 4); |
| 19 |
vint64m1_t v3 = __riscv_vslide1up_vx_i64m1 (v2, -16, 4); |
| 20 |
vint64m1_t v4 = __riscv_vslide1up_vx_i64m1 (v3, -16, 4); |
| 21 |
__riscv_vse64_v_i64m1 (out + 2, v4, 4); |
| 22 |
} |
| 23 |
|
| 24 |
/* |
| 25 |
** f1: |
| 26 |
** ... |
| 27 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 28 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 29 |
** ... |
| 30 |
** ret |
| 31 |
*/ |
| 32 |
void f1 (void * in, void *out, int64_t x, int n) |
| 33 |
{ |
| 34 |
vint64m1_t v = __riscv_vle64_v_i64m1 (in + 1, 4); |
| 35 |
vint64m1_t v2 = __riscv_vle64_v_i64m1_tu (v, in + 2, 4); |
| 36 |
vint64m1_t v3 = __riscv_vslide1up_vx_i64m1 (v2, 15, 4); |
| 37 |
vint64m1_t v4 = __riscv_vslide1up_vx_i64m1 (v3, 15, 4); |
| 38 |
__riscv_vse64_v_i64m1 (out + 2, v4, 4); |
| 39 |
} |
| 40 |
|
| 41 |
/* |
| 42 |
** f2: |
| 43 |
** ... |
| 44 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 45 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 46 |
** ... |
| 47 |
** ret |
| 48 |
*/ |
| 49 |
void f2 (void * in, void *out, int64_t x, int n) |
| 50 |
{ |
| 51 |
vint64m1_t v = __riscv_vle64_v_i64m1 (in + 1, 4); |
| 52 |
vint64m1_t v2 = __riscv_vle64_v_i64m1_tu (v, in + 2, 4); |
| 53 |
vint64m1_t v3 = __riscv_vslide1up_vx_i64m1 (v2, 16, 4); |
| 54 |
vint64m1_t v4 = __riscv_vslide1up_vx_i64m1 (v3, 16, 4); |
| 55 |
__riscv_vse64_v_i64m1 (out + 2, v4, 4); |
| 56 |
} |
| 57 |
|
| 58 |
/* |
| 59 |
** f3: |
| 60 |
** ... |
| 61 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*zero |
| 62 |
** ... |
| 63 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 64 |
** ... |
| 65 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*zero |
| 66 |
** ... |
| 67 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 68 |
** ... |
| 69 |
** ret |
| 70 |
*/ |
| 71 |
void f3 (void * in, void *out, int64_t x, int n) |
| 72 |
{ |
| 73 |
vint64m1_t v = __riscv_vle64_v_i64m1 (in + 1, 4); |
| 74 |
vint64m1_t v2 = __riscv_vle64_v_i64m1_tu (v, in + 2, 4); |
| 75 |
vint64m1_t v3 = __riscv_vslide1up_vx_i64m1 (v2, 0xAAAAAAAA, 4); |
| 76 |
vint64m1_t v4 = __riscv_vslide1up_vx_i64m1 (v3, 0xAAAAAAAA, 4); |
| 77 |
__riscv_vse64_v_i64m1 (out + 2, v4, 4); |
| 78 |
} |
| 79 |
|
| 80 |
/* |
| 81 |
** f4: |
| 82 |
** ... |
| 83 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 84 |
** ... |
| 85 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 86 |
** ... |
| 87 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 88 |
** ... |
| 89 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 90 |
** ... |
| 91 |
** ret |
| 92 |
*/ |
| 93 |
void f4 (void * in, void *out, int64_t x, int n) |
| 94 |
{ |
| 95 |
vint64m1_t v = __riscv_vle64_v_i64m1 (in + 1, 4); |
| 96 |
vint64m1_t v2 = __riscv_vle64_v_i64m1_tu (v, in + 2, 4); |
| 97 |
vint64m1_t v3 = __riscv_vslide1up_vx_i64m1 (v2, 0xAAAAAAAAAAAAAAAA, 4); |
| 98 |
vint64m1_t v4 = __riscv_vslide1up_vx_i64m1 (v3, 0xAAAAAAAAAAAAAAAA, 4); |
| 99 |
__riscv_vse64_v_i64m1 (out + 2, v4, 4); |
| 100 |
} |
| 101 |
|
| 102 |
/* |
| 103 |
** f5: |
| 104 |
** ... |
| 105 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 106 |
** ... |
| 107 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 108 |
** ... |
| 109 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 110 |
** ... |
| 111 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 112 |
** ... |
| 113 |
** ret |
| 114 |
*/ |
| 115 |
void f5 (void * in, void *out, int64_t x, int n) |
| 116 |
{ |
| 117 |
vint64m1_t v = __riscv_vle64_v_i64m1 (in + 1, 4); |
| 118 |
vint64m1_t v2 = __riscv_vle64_v_i64m1_tu (v, in + 2, 4); |
| 119 |
vint64m1_t v3 = __riscv_vslide1up_vx_i64m1 (v2, 0xAAAAAAAAAAAAAAAA, 4); |
| 120 |
vint64m1_t v4 = __riscv_vslide1up_vx_i64m1 (v3, 0xAAAAAAAAAAAAAAAA, 4); |
| 121 |
__riscv_vse64_v_i64m1 (out + 2, v4, 4); |
| 122 |
} |
| 123 |
|
| 124 |
/* |
| 125 |
** f6: |
| 126 |
** ... |
| 127 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 128 |
** ... |
| 129 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 130 |
** ... |
| 131 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 132 |
** ... |
| 133 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ |
| 134 |
** ... |
| 135 |
** ret |
| 136 |
*/ |
| 137 |
void f6 (void * in, void *out, int64_t x, int n) |
| 138 |
{ |
| 139 |
vint64m1_t v = __riscv_vle64_v_i64m1 (in + 1, 4); |
| 140 |
vint64m1_t v2 = __riscv_vle64_v_i64m1_tu (v, in + 2, 4); |
| 141 |
vint64m1_t v3 = __riscv_vslide1up_vx_i64m1 (v2, x, 4); |
| 142 |
vint64m1_t v4 = __riscv_vslide1up_vx_i64m1 (v3, x, 4); |
| 143 |
__riscv_vse64_v_i64m1 (out + 2, v4, 4); |
| 144 |
} |
| 145 |
|
| 146 |
/* |
| 147 |
** f7: |
| 148 |
** ... |
| 149 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*zero |
| 150 |
** vslide1up\.vx\tv[0-9]+,\s*v[0-9]+,\s*zero |
| 151 |
** ... |
| 152 |
** ret |
| 153 |
*/ |
| 154 |
void f7 (void * in, void *out, int64_t x, int n) |
| 155 |
{ |
| 156 |
vint64m1_t v = __riscv_vle64_v_i64m1 (in + 1, 4); |
| 157 |
vint64m1_t v2 = __riscv_vle64_v_i64m1_tu (v, in + 2, 4); |
| 158 |
vint64m1_t v3 = __riscv_vslide1up_vx_i64m1 (v2, 0, 4); |
| 159 |
vint64m1_t v4 = __riscv_vslide1up_vx_i64m1 (v3, 0, 4); |
| 160 |
__riscv_vse64_v_i64m1 (out + 2, v4, 4); |
| 161 |
} |
| 162 |
|
| 163 |
/* { dg-final { scan-assembler-not {vmv} } } */ |