| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-march=rv32gcv -mabi=ilp32d -fno-schedule-insns -fno-schedule-insns2 -O3" } */ |
| 3 |
/* { dg-final { check-function-bodies "**" "" } } */ |
| 4 |
|
| 5 |
#include "riscv_vector.h" |
| 6 |
|
| 7 |
/* |
| 8 |
** foo: |
| 9 |
** addi\t[a-x0-9]+,\s*[a-x0-9]+,100 |
| 10 |
** vsetvli\tzero,a2,e64,m2,t[au],m[au] |
| 11 |
** vlse64.v\tv[0-9]+,0\([a-x0-9]+\),zero |
| 12 |
** vs2r.v\tv[0-9]+,0\([a-x0-9]+\) |
| 13 |
** ret |
| 14 |
*/ |
| 15 |
void foo (void *base, void *out, size_t vl) |
| 16 |
{ |
| 17 |
int64_t scalar = *(int64_t*)(base + 100); |
| 18 |
vint64m2_t v = __riscv_vmv_s_x_i64m2 (scalar, vl); |
| 19 |
*(vint64m2_t*)out = v; |
| 20 |
} |
| 21 |
|
| 22 |
/* |
| 23 |
** foo2: |
| 24 |
** fld\tfa[0-9]+,\s*100\(a0\) |
| 25 |
** vsetvli\tzero,a2,e64,m2,t[au],m[au] |
| 26 |
** vfmv\.v\.f\tv[0-9]+,\s*fa[0-9]+ |
| 27 |
** vs2r.v\tv[0-9]+,0\([a-x0-9]+\) |
| 28 |
** ret |
| 29 |
*/ |
| 30 |
void foo2 (void *base, void *out, size_t vl) |
| 31 |
{ |
| 32 |
double scalar = *(double*)(base + 100); |
| 33 |
vfloat64m2_t v = __riscv_vfmv_s_f_f64m2 (scalar, vl); |
| 34 |
*(vfloat64m2_t*)out = v; |
| 35 |
} |
| 36 |
|
| 37 |
/* |
| 38 |
** foo3: |
| 39 |
** ... |
| 40 |
** vlse64.v\tv[0-9]+,0\([a-x0-9]+\),zero |
| 41 |
** ... |
| 42 |
** ret |
| 43 |
*/ |
| 44 |
void foo3 (void *base, void *out, size_t vl) |
| 45 |
{ |
| 46 |
int64_t scalar = *(int64_t*)(base + 100); |
| 47 |
vint64m2_t merge = *(vint64m2_t*) (base + 200); |
| 48 |
vint64m2_t v = __riscv_vmv_s_x_i64m2_tu (merge, scalar, vl); |
| 49 |
*(vint64m2_t*)out = v; |
| 50 |
} |
| 51 |
|
| 52 |
/* |
| 53 |
** foo4: |
| 54 |
** ... |
| 55 |
** vfmv\.v\.f\tv[0-9]+,\s*fa[0-9]+ |
| 56 |
** ... |
| 57 |
** ret |
| 58 |
*/ |
| 59 |
void foo4 (void *base, void *out, size_t vl) |
| 60 |
{ |
| 61 |
double scalar = *(double*)(base + 100); |
| 62 |
vfloat64m2_t merge = *(vfloat64m2_t*) (base + 200); |
| 63 |
vfloat64m2_t v = __riscv_vfmv_s_f_f64m2_tu (merge, scalar, vl); |
| 64 |
*(vfloat64m2_t*)out = v; |
| 65 |
} |
| 66 |
|
| 67 |
/* |
| 68 |
** foo5: |
| 69 |
** ... |
| 70 |
** vlse64.v\tv[0-9]+,0\([a-x0-9]+\),zero |
| 71 |
** ... |
| 72 |
*/ |
| 73 |
void foo5 (void *base, void *out, size_t vl, int64_t x) |
| 74 |
{ |
| 75 |
vint64m2_t v = __riscv_vmv_s_x_i64m2 (x, vl); |
| 76 |
*(vint64m2_t*)out = v; |
| 77 |
} |
| 78 |
|
| 79 |
/* |
| 80 |
** foo6: |
| 81 |
** ... |
| 82 |
** vfmv.s.f\tv[0-9]+,\s*[a-x0-9]+ |
| 83 |
** ... |
| 84 |
** ret |
| 85 |
*/ |
| 86 |
void foo6 (void *base, void *out, size_t vl, double x) |
| 87 |
{ |
| 88 |
vfloat64m2_t v = __riscv_vfmv_s_f_f64m2 (x, vl); |
| 89 |
*(vfloat64m2_t*)out = v; |
| 90 |
} |
| 91 |
|
| 92 |
/* |
| 93 |
** foo7: |
| 94 |
** ... |
| 95 |
** vlse64.v\tv[0-9]+,0\([a-x0-9]+\),zero |
| 96 |
** ... |
| 97 |
*/ |
| 98 |
void foo7 (void *base, void *out, size_t vl, int64_t x) |
| 99 |
{ |
| 100 |
vint64m2_t merge = *(vint64m2_t*) (base + 200); |
| 101 |
vint64m2_t v = __riscv_vmv_s_x_i64m2_tu (merge, x, vl); |
| 102 |
*(vint64m2_t*)out = v; |
| 103 |
} |
| 104 |
|
| 105 |
/* |
| 106 |
** foo8: |
| 107 |
** ... |
| 108 |
** vfmv.s.f\tv[0-9]+,\s*[a-x0-9]+ |
| 109 |
** ... |
| 110 |
** ret |
| 111 |
*/ |
| 112 |
void foo8 (void *base, void *out, size_t vl, double x) |
| 113 |
{ |
| 114 |
vfloat64m2_t merge = *(vfloat64m2_t*) (base + 200); |
| 115 |
vfloat64m2_t v = __riscv_vfmv_s_f_f64m2_tu (merge, x, vl); |
| 116 |
*(vfloat64m2_t*)out = v; |
| 117 |
} |
| 118 |
|
| 119 |
/* |
| 120 |
** foo9: |
| 121 |
** ... |
| 122 |
** vmv.s.x\tv[0-9]+,\s*[a-x0-9]+ |
| 123 |
** ... |
| 124 |
** ret |
| 125 |
*/ |
| 126 |
void foo9 (void *base, void *out, size_t vl) |
| 127 |
{ |
| 128 |
int64_t scalar = *(int64_t*)(base + 100); |
| 129 |
vint64m2_t v = __riscv_vmv_s_x_i64m2 (-15, vl); |
| 130 |
*(vint64m2_t*)out = v; |
| 131 |
} |
| 132 |
|
| 133 |
/* |
| 134 |
** foo10: |
| 135 |
** ... |
| 136 |
** vmv.s.x\tv[0-9]+,\s*[a-x0-9]+ |
| 137 |
** ... |
| 138 |
*/ |
| 139 |
void foo10 (void *base, void *out, size_t vl) |
| 140 |
{ |
| 141 |
int64_t scalar = *(int64_t*)(base + 100); |
| 142 |
vint64m2_t merge = *(vint64m2_t*) (base + 200); |
| 143 |
vint64m2_t v = __riscv_vmv_s_x_i64m2_tu (merge, -15, vl); |
| 144 |
*(vint64m2_t*)out = v; |
| 145 |
} |
| 146 |
|
| 147 |
/* |
| 148 |
** foo11: |
| 149 |
** ... |
| 150 |
** vmv.s.x\tv[0-9]+,\s*zero |
| 151 |
** ... |
| 152 |
** ret |
| 153 |
*/ |
| 154 |
void foo11 (void *base, void *out, size_t vl) |
| 155 |
{ |
| 156 |
double scalar = *(double*)(base + 100); |
| 157 |
vfloat64m2_t v = __riscv_vfmv_s_f_f64m2 (0, vl); |
| 158 |
*(vfloat64m2_t*)out = v; |
| 159 |
} |
| 160 |
|
| 161 |
/* |
| 162 |
** foo12: |
| 163 |
** ... |
| 164 |
** vmv.s.x\tv[0-9]+,\s*zero |
| 165 |
** ... |
| 166 |
** ret |
| 167 |
*/ |
| 168 |
void foo12 (void *base, void *out, size_t vl) |
| 169 |
{ |
| 170 |
vfloat64m2_t merge = *(vfloat64m2_t*) (base + 200); |
| 171 |
vfloat64m2_t v = __riscv_vfmv_s_f_f64m2_tu (merge, 0, vl); |
| 172 |
*(vfloat64m2_t*)out = v; |
| 173 |
} |
| 174 |
|
| 175 |
/* |
| 176 |
** foo12_1: |
| 177 |
** ... |
| 178 |
** vfmv.s.f\tv[0-9]+,\s*[a-x0-9]+ |
| 179 |
** ... |
| 180 |
** ret |
| 181 |
*/ |
| 182 |
void foo12_1 (void *base, void *out, size_t vl) |
| 183 |
{ |
| 184 |
vfloat64m2_t merge = *(vfloat64m2_t*) (base + 200); |
| 185 |
vfloat64m2_t v = __riscv_vfmv_s_f_f64m2_tu (merge, 0.2, vl); |
| 186 |
*(vfloat64m2_t*)out = v; |
| 187 |
} |
| 188 |
|
| 189 |
/* |
| 190 |
** foo13: |
| 191 |
** ... |
| 192 |
** vmv.s.x\tv[0-9]+,\s*[a-x0-9]+ |
| 193 |
** ... |
| 194 |
** ret |
| 195 |
*/ |
| 196 |
void foo13 (void *base, void *out, size_t vl) |
| 197 |
{ |
| 198 |
int64_t scalar = *(int64_t*)(base + 100); |
| 199 |
vint64m2_t v = __riscv_vmv_s_x_i64m2 (0xAAAAA, vl); |
| 200 |
*(vint64m2_t*)out = v; |
| 201 |
} |
| 202 |
|
| 203 |
/* |
| 204 |
** foo14: |
| 205 |
** ... |
| 206 |
** vmv.s.x\tv[0-9]+,\s*[a-x0-9]+ |
| 207 |
** ... |
| 208 |
*/ |
| 209 |
void foo14 (void *base, void *out, size_t vl) |
| 210 |
{ |
| 211 |
int64_t scalar = *(int64_t*)(base + 100); |
| 212 |
vint64m2_t merge = *(vint64m2_t*) (base + 200); |
| 213 |
vint64m2_t v = __riscv_vmv_s_x_i64m2_tu (merge, 0xAAAAA, vl); |
| 214 |
*(vint64m2_t*)out = v; |
| 215 |
} |