| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-march=rv32gc_zve64d -mabi=ilp32 -mpreferred-stack-boundary=3 -O3 -fno-schedule-insns -fno-schedule-insns2" } */ |
| 3 |
/* { dg-final { check-function-bodies "**" "" } } */ |
| 4 |
|
| 5 |
#include "riscv_vector.h" |
| 6 |
#include "macro.h" |
| 7 |
|
| 8 |
/* |
| 9 |
** spill_2: |
| 10 |
** csrr\tt0,vlenb |
| 11 |
** sub\tsp,sp,t0 |
| 12 |
** vsetvli\ta5,zero,e16,mf4,ta,ma |
| 13 |
** vle16.v\tv[0-9]+,0\(a0\) |
| 14 |
** csrr\t[a-x0-9]+,vlenb |
| 15 |
** srli\t[a-x0-9]+,[a-x0-9]+,2 |
| 16 |
** slli\t[a-x0-9]+,[a-x0-9]+,2 |
| 17 |
** sub\t[a-x0-9]+,[a-x0-9]+,[a-x0-9]+ |
| 18 |
** add\t[a-x0-9]+,[a-x0-9]+,sp |
| 19 |
** vse16.v\tv[0-9]+,0\([a-x0-9]+\) |
| 20 |
** ... |
| 21 |
** csrr\t[a-x0-9]+,vlenb |
| 22 |
** srli\t[a-x0-9]+,[a-x0-9]+,2 |
| 23 |
** slli\t[a-x0-9]+,[a-x0-9]+,2 |
| 24 |
** sub\t[a-x0-9]+,[a-x0-9]+,[a-x0-9]+ |
| 25 |
** add\t[a-x0-9]+,[a-x0-9]+,sp |
| 26 |
** vle16.v\tv[0-9]+,0\([a-x0-9]+\) |
| 27 |
** vse16.v\tv[0-9]+,0\([a-x0-9]+\) |
| 28 |
** csrr\tt0,vlenb |
| 29 |
** add\tsp,sp,t0 |
| 30 |
** ... |
| 31 |
** jr\tra |
| 32 |
*/ |
| 33 |
void |
| 34 |
spill_2 (int16_t *in, int16_t *out) |
| 35 |
{ |
| 36 |
vint16mf4_t v1 = *(vint16mf4_t*)in; |
| 37 |
exhaust_vector_regs (); |
| 38 |
*(vint16mf4_t*)out = v1; |
| 39 |
} |
| 40 |
|
| 41 |
/* |
| 42 |
** spill_3: |
| 43 |
** csrr\tt0,vlenb |
| 44 |
** sub\tsp,sp,t0 |
| 45 |
** vsetvli\ta5,zero,e16,mf2,ta,ma |
| 46 |
** vle16.v\tv[0-9]+,0\(a0\) |
| 47 |
** csrr\t[a-x0-9]+,vlenb |
| 48 |
** srli\t[a-x0-9]+,[a-x0-9]+,1 |
| 49 |
** add\t[a-x0-9]+,[a-x0-9]+,sp |
| 50 |
** vse16.v\tv[0-9]+,0\([a-x0-9]+\) |
| 51 |
** ... |
| 52 |
** csrr\t[a-x0-9]+,vlenb |
| 53 |
** srli\t[a-x0-9]+,[a-x0-9]+,1 |
| 54 |
** add\t[a-x0-9]+,[a-x0-9]+,sp |
| 55 |
** vle16.v\tv[0-9]+,0\([a-x0-9]+\) |
| 56 |
** vse16.v\tv[0-9]+,0\([a-x0-9]+\) |
| 57 |
** csrr\tt0,vlenb |
| 58 |
** add\tsp,sp,t0 |
| 59 |
** ... |
| 60 |
** jr\tra |
| 61 |
*/ |
| 62 |
void |
| 63 |
spill_3 (int16_t *in, int16_t *out) |
| 64 |
{ |
| 65 |
vint16mf2_t v1 = *(vint16mf2_t*)in; |
| 66 |
exhaust_vector_regs (); |
| 67 |
*(vint16mf2_t*)out = v1; |
| 68 |
} |
| 69 |
|
| 70 |
/* |
| 71 |
** spill_4: |
| 72 |
** csrr\tt0,vlenb |
| 73 |
** sub\tsp,sp,t0 |
| 74 |
** ... |
| 75 |
** vs1r.v\tv[0-9]+,0\(sp\) |
| 76 |
** ... |
| 77 |
** vl1re16.v\tv2,0\(sp\) |
| 78 |
** vs1r.v\tv2,0\([a-x0-9]+\) |
| 79 |
** ... |
| 80 |
** jr\tra |
| 81 |
*/ |
| 82 |
void |
| 83 |
spill_4 (int16_t *in, int16_t *out) |
| 84 |
{ |
| 85 |
register vint16m1_t v1 asm("v1") = *(vint16m1_t*)in; |
| 86 |
asm volatile ("# %0"::"vr"(v1)); |
| 87 |
exhaust_vector_regs (); |
| 88 |
register vint16m1_t v2 asm("v2") = v1; |
| 89 |
*(vint16m1_t*)out = v2; |
| 90 |
asm volatile ("# %0"::"vr"(v2)); |
| 91 |
} |
| 92 |
|
| 93 |
/* |
| 94 |
** spill_5: |
| 95 |
** csrr\tt0,vlenb |
| 96 |
** slli\tt1,t0,1 |
| 97 |
** sub\tsp,sp,t1 |
| 98 |
** ... |
| 99 |
** vs2r.v\tv[0-9]+,0\(sp\) |
| 100 |
** ... |
| 101 |
** vl2re16.v\tv4,0\(sp\) |
| 102 |
** vs2r.v\tv4,0\([a-x0-9]+\) |
| 103 |
** ... |
| 104 |
** jr\tra |
| 105 |
*/ |
| 106 |
void |
| 107 |
spill_5 (int16_t *in, int16_t *out) |
| 108 |
{ |
| 109 |
register vint16m2_t v2 asm("v2") = *(vint16m2_t*)in; |
| 110 |
asm volatile ("# %0"::"vr"(v2)); |
| 111 |
exhaust_vector_regs (); |
| 112 |
register vint16m2_t v4 asm("v4") = v2; |
| 113 |
*(vint16m2_t*)out = v4; |
| 114 |
asm volatile ("# %0"::"vr"(v4)); |
| 115 |
} |
| 116 |
|
| 117 |
/* |
| 118 |
** spill_6: |
| 119 |
** csrr\tt0,vlenb |
| 120 |
** slli\tt1,t0,2 |
| 121 |
** sub\tsp,sp,t1 |
| 122 |
** ... |
| 123 |
** vs4r.v\tv[0-9]+,0\(sp\) |
| 124 |
** ... |
| 125 |
** vl4re16.v\tv8,0\(sp\) |
| 126 |
** vs4r.v\tv8,0\([a-x0-9]+\) |
| 127 |
** ... |
| 128 |
** jr\tra |
| 129 |
*/ |
| 130 |
void |
| 131 |
spill_6 (int16_t *in, int16_t *out) |
| 132 |
{ |
| 133 |
register vint16m4_t v4 asm("v4") = *(vint16m4_t*)in; |
| 134 |
asm volatile ("# %0"::"vr"(v4)); |
| 135 |
exhaust_vector_regs (); |
| 136 |
register vint16m4_t v8 asm("v8") = v4; |
| 137 |
*(vint16m4_t*)out = v8; |
| 138 |
asm volatile ("# %0"::"vr"(v8)); |
| 139 |
} |
| 140 |
|
| 141 |
/* |
| 142 |
** spill_7: |
| 143 |
** csrr\tt0,vlenb |
| 144 |
** slli\tt1,t0,3 |
| 145 |
** sub\tsp,sp,t1 |
| 146 |
** ... |
| 147 |
** vs8r.v\tv[0-9]+,0\(sp\) |
| 148 |
** ... |
| 149 |
** vl8re16.v\tv16,0\(sp\) |
| 150 |
** vs8r.v\tv16,0\([a-x0-9]+\) |
| 151 |
** ... |
| 152 |
** jr\tra |
| 153 |
*/ |
| 154 |
void |
| 155 |
spill_7 (int16_t *in, int16_t *out) |
| 156 |
{ |
| 157 |
register vint16m8_t v8 asm("v8") = *(vint16m8_t*)in; |
| 158 |
asm volatile ("# %0"::"vr"(v8)); |
| 159 |
exhaust_vector_regs (); |
| 160 |
register vint16m8_t v16 asm("v16") = v8; |
| 161 |
*(vint16m8_t*)out = v16; |
| 162 |
asm volatile ("# %0"::"vr"(v16)); |
| 163 |
} |
| 164 |
|
| 165 |
/* |
| 166 |
** spill_9: |
| 167 |
** csrr\tt0,vlenb |
| 168 |
** sub\tsp,sp,t0 |
| 169 |
** vsetvli\ta5,zero,e16,mf4,ta,ma |
| 170 |
** vle16.v\tv[0-9]+,0\(a0\) |
| 171 |
** csrr\t[a-x0-9]+,vlenb |
| 172 |
** srli\t[a-x0-9]+,[a-x0-9]+,2 |
| 173 |
** slli\t[a-x0-9]+,[a-x0-9]+,2 |
| 174 |
** sub\t[a-x0-9]+,[a-x0-9]+,[a-x0-9]+ |
| 175 |
** add\t[a-x0-9]+,[a-x0-9]+,sp |
| 176 |
** vse16.v\tv[0-9]+,0\([a-x0-9]+\) |
| 177 |
** ... |
| 178 |
** csrr\t[a-x0-9]+,vlenb |
| 179 |
** srli\t[a-x0-9]+,[a-x0-9]+,2 |
| 180 |
** slli\t[a-x0-9]+,[a-x0-9]+,2 |
| 181 |
** sub\t[a-x0-9]+,[a-x0-9]+,[a-x0-9]+ |
| 182 |
** add\t[a-x0-9]+,[a-x0-9]+,sp |
| 183 |
** vle16.v\tv[0-9]+,0\([a-x0-9]+\) |
| 184 |
** vse16.v\tv[0-9]+,0\([a-x0-9]+\) |
| 185 |
** csrr\tt0,vlenb |
| 186 |
** add\tsp,sp,t0 |
| 187 |
** ... |
| 188 |
** jr\tra |
| 189 |
*/ |
| 190 |
void |
| 191 |
spill_9 (uint16_t *in, uint16_t *out) |
| 192 |
{ |
| 193 |
vuint16mf4_t v1 = *(vuint16mf4_t*)in; |
| 194 |
exhaust_vector_regs (); |
| 195 |
*(vuint16mf4_t*)out = v1; |
| 196 |
} |
| 197 |
|
| 198 |
/* |
| 199 |
** spill_10: |
| 200 |
** csrr\tt0,vlenb |
| 201 |
** sub\tsp,sp,t0 |
| 202 |
** vsetvli\ta5,zero,e16,mf2,ta,ma |
| 203 |
** vle16.v\tv[0-9]+,0\(a0\) |
| 204 |
** csrr\t[a-x0-9]+,vlenb |
| 205 |
** srli\t[a-x0-9]+,[a-x0-9]+,1 |
| 206 |
** add\t[a-x0-9]+,[a-x0-9]+,sp |
| 207 |
** vse16.v\tv[0-9]+,0\([a-x0-9]+\) |
| 208 |
** ... |
| 209 |
** csrr\t[a-x0-9]+,vlenb |
| 210 |
** srli\t[a-x0-9]+,[a-x0-9]+,1 |
| 211 |
** add\t[a-x0-9]+,[a-x0-9]+,sp |
| 212 |
** vle16.v\tv[0-9]+,0\([a-x0-9]+\) |
| 213 |
** vse16.v\tv[0-9]+,0\([a-x0-9]+\) |
| 214 |
** csrr\tt0,vlenb |
| 215 |
** add\tsp,sp,t0 |
| 216 |
** ... |
| 217 |
** jr\tra |
| 218 |
*/ |
| 219 |
void |
| 220 |
spill_10 (uint16_t *in, uint16_t *out) |
| 221 |
{ |
| 222 |
vuint16mf2_t v1 = *(vuint16mf2_t*)in; |
| 223 |
exhaust_vector_regs (); |
| 224 |
*(vuint16mf2_t*)out = v1; |
| 225 |
} |
| 226 |
|
| 227 |
/* |
| 228 |
** spill_11: |
| 229 |
** csrr\tt0,vlenb |
| 230 |
** sub\tsp,sp,t0 |
| 231 |
** ... |
| 232 |
** vs1r.v\tv[0-9]+,0\(sp\) |
| 233 |
** ... |
| 234 |
** vl1re16.v\tv2,0\(sp\) |
| 235 |
** vs1r.v\tv2,0\([a-x0-9]+\) |
| 236 |
** ... |
| 237 |
** jr\tra |
| 238 |
*/ |
| 239 |
void |
| 240 |
spill_11 (uint16_t *in, uint16_t *out) |
| 241 |
{ |
| 242 |
register vuint16m1_t v1 asm("v1") = *(vuint16m1_t*)in; |
| 243 |
asm volatile ("# %0"::"vr"(v1)); |
| 244 |
exhaust_vector_regs (); |
| 245 |
register vuint16m1_t v2 asm("v2") = v1; |
| 246 |
*(vuint16m1_t*)out = v2; |
| 247 |
asm volatile ("# %0"::"vr"(v2)); |
| 248 |
} |
| 249 |
|
| 250 |
/* |
| 251 |
** spill_12: |
| 252 |
** csrr\tt0,vlenb |
| 253 |
** slli\tt1,t0,1 |
| 254 |
** sub\tsp,sp,t1 |
| 255 |
** ... |
| 256 |
** vs2r.v\tv[0-9]+,0\(sp\) |
| 257 |
** ... |
| 258 |
** vl2re16.v\tv4,0\(sp\) |
| 259 |
** vs2r.v\tv4,0\([a-x0-9]+\) |
| 260 |
** ... |
| 261 |
** jr\tra |
| 262 |
*/ |
| 263 |
void |
| 264 |
spill_12 (uint16_t *in, uint16_t *out) |
| 265 |
{ |
| 266 |
register vuint16m2_t v2 asm("v2") = *(vuint16m2_t*)in; |
| 267 |
asm volatile ("# %0"::"vr"(v2)); |
| 268 |
exhaust_vector_regs (); |
| 269 |
register vuint16m2_t v4 asm("v4") = v2; |
| 270 |
*(vuint16m2_t*)out = v4; |
| 271 |
asm volatile ("# %0"::"vr"(v4)); |
| 272 |
} |
| 273 |
|
| 274 |
/* |
| 275 |
** spill_13: |
| 276 |
** csrr\tt0,vlenb |
| 277 |
** slli\tt1,t0,2 |
| 278 |
** sub\tsp,sp,t1 |
| 279 |
** ... |
| 280 |
** vs4r.v\tv[0-9]+,0\(sp\) |
| 281 |
** ... |
| 282 |
** vl4re16.v\tv8,0\(sp\) |
| 283 |
** vs4r.v\tv8,0\([a-x0-9]+\) |
| 284 |
** ... |
| 285 |
** jr\tra |
| 286 |
*/ |
| 287 |
void |
| 288 |
spill_13 (uint16_t *in, uint16_t *out) |
| 289 |
{ |
| 290 |
register vuint16m4_t v4 asm("v4") = *(vuint16m4_t*)in; |
| 291 |
asm volatile ("# %0"::"vr"(v4)); |
| 292 |
exhaust_vector_regs (); |
| 293 |
register vuint16m4_t v8 asm("v8") = v4; |
| 294 |
*(vuint16m4_t*)out = v8; |
| 295 |
asm volatile ("# %0"::"vr"(v8)); |
| 296 |
} |
| 297 |
|
| 298 |
/* |
| 299 |
** spill_14: |
| 300 |
** csrr\tt0,vlenb |
| 301 |
** slli\tt1,t0,3 |
| 302 |
** sub\tsp,sp,t1 |
| 303 |
** ... |
| 304 |
** vs8r.v\tv[0-9]+,0\(sp\) |
| 305 |
** ... |
| 306 |
** vl8re16.v\tv16,0\(sp\) |
| 307 |
** vs8r.v\tv16,0\([a-x0-9]+\) |
| 308 |
** ... |
| 309 |
** jr\tra |
| 310 |
*/ |
| 311 |
void |
| 312 |
spill_14 (uint16_t *in, uint16_t *out) |
| 313 |
{ |
| 314 |
register vuint16m8_t v8 asm("v8") = *(vuint16m8_t*)in; |
| 315 |
asm volatile ("# %0"::"vr"(v8)); |
| 316 |
exhaust_vector_regs (); |
| 317 |
register vuint16m8_t v16 asm("v16") = v8; |
| 318 |
*(vuint16m8_t*)out = v16; |
| 319 |
asm volatile ("# %0"::"vr"(v16)); |
| 320 |
} |