| 1 |
// { dg-options "-O -fomit-frame-pointer -fno-optimize-sibling-calls" } |
| 2 |
// { dg-final { check-function-bodies "**" "" } } |
| 3 |
|
| 4 |
#include <arm_neon.h> |
| 5 |
#include <arm_sve.h> |
| 6 |
|
| 7 |
double produce_d0 (); |
| 8 |
void consume_d0 (double); |
| 9 |
|
| 10 |
/* |
| 11 |
** test_d0: |
| 12 |
** ... |
| 13 |
** smstop sm |
| 14 |
** bl produce_d0 |
| 15 |
** fmov x10, d0 |
| 16 |
** smstart sm |
| 17 |
** fmov d0, x10 |
| 18 |
** fmov x10, d0 |
| 19 |
** smstop sm |
| 20 |
** fmov d0, x10 |
| 21 |
** bl consume_d0 |
| 22 |
** ... |
| 23 |
*/ |
| 24 |
void |
| 25 |
test_d0 () [[arm::streaming]] |
| 26 |
{ |
| 27 |
double res = produce_d0 (); |
| 28 |
asm volatile (""); |
| 29 |
consume_d0 (res); |
| 30 |
} |
| 31 |
|
| 32 |
int8x8_t produce_d0_vec (); |
| 33 |
void consume_d0_vec (int8x8_t); |
| 34 |
|
| 35 |
/* |
| 36 |
** test_d0_vec: |
| 37 |
** ... |
| 38 |
** smstop sm |
| 39 |
** bl produce_d0_vec |
| 40 |
** ( |
| 41 |
** fmov x10, d0 |
| 42 |
** | |
| 43 |
** umov x10, v0.d\[0\] |
| 44 |
** ) |
| 45 |
** smstart sm |
| 46 |
** fmov d0, x10 |
| 47 |
** ( |
| 48 |
** fmov x10, d0 |
| 49 |
** | |
| 50 |
** umov x10, v0.d\[0\] |
| 51 |
** ) |
| 52 |
** smstop sm |
| 53 |
** fmov d0, x10 |
| 54 |
** bl consume_d0_vec |
| 55 |
** ... |
| 56 |
*/ |
| 57 |
void |
| 58 |
test_d0_vec () [[arm::streaming]] |
| 59 |
{ |
| 60 |
int8x8_t res = produce_d0_vec (); |
| 61 |
asm volatile (""); |
| 62 |
consume_d0_vec (res); |
| 63 |
} |
| 64 |
|
| 65 |
int8x16_t produce_q0 (); |
| 66 |
void consume_q0 (int8x16_t); |
| 67 |
|
| 68 |
/* |
| 69 |
** test_q0: |
| 70 |
** ... |
| 71 |
** smstop sm |
| 72 |
** bl produce_q0 |
| 73 |
** str q0, \[sp, #?-16\]! |
| 74 |
** smstart sm |
| 75 |
** ldr q0, \[sp\], #?16 |
| 76 |
** str q0, \[sp, #?-16\]! |
| 77 |
** smstop sm |
| 78 |
** ldr q0, \[sp\], #?16 |
| 79 |
** bl consume_q0 |
| 80 |
** ... |
| 81 |
*/ |
| 82 |
void |
| 83 |
test_q0 () [[arm::streaming]] |
| 84 |
{ |
| 85 |
int8x16_t res = produce_q0 (); |
| 86 |
asm volatile (""); |
| 87 |
consume_q0 (res); |
| 88 |
} |
| 89 |
|
| 90 |
int8x16x2_t produce_q1 (); |
| 91 |
void consume_q1 (int8x16x2_t); |
| 92 |
|
| 93 |
/* |
| 94 |
** test_q1: |
| 95 |
** ... |
| 96 |
** smstop sm |
| 97 |
** bl produce_q1 |
| 98 |
** stp q0, q1, \[sp, #?-32\]! |
| 99 |
** smstart sm |
| 100 |
** ldp q0, q1, \[sp\], #?32 |
| 101 |
** stp q0, q1, \[sp, #?-32\]! |
| 102 |
** smstop sm |
| 103 |
** ldp q0, q1, \[sp\], #?32 |
| 104 |
** bl consume_q1 |
| 105 |
** ... |
| 106 |
*/ |
| 107 |
void |
| 108 |
test_q1 () [[arm::streaming]] |
| 109 |
{ |
| 110 |
int8x16x2_t res = produce_q1 (); |
| 111 |
asm volatile (""); |
| 112 |
consume_q1 (res); |
| 113 |
} |
| 114 |
|
| 115 |
int8x16x3_t produce_q2 (); |
| 116 |
void consume_q2 (int8x16x3_t); |
| 117 |
|
| 118 |
/* |
| 119 |
** test_q2: |
| 120 |
** ... |
| 121 |
** smstop sm |
| 122 |
** bl produce_q2 |
| 123 |
** stp q0, q1, \[sp, #?-48\]! |
| 124 |
** str q2, \[sp, #?32\] |
| 125 |
** smstart sm |
| 126 |
** ldr q2, \[sp, #?32\] |
| 127 |
** ldp q0, q1, \[sp\], #?48 |
| 128 |
** stp q0, q1, \[sp, #?-48\]! |
| 129 |
** str q2, \[sp, #?32\] |
| 130 |
** smstop sm |
| 131 |
** ldr q2, \[sp, #?32\] |
| 132 |
** ldp q0, q1, \[sp\], #?48 |
| 133 |
** bl consume_q2 |
| 134 |
** ... |
| 135 |
*/ |
| 136 |
void |
| 137 |
test_q2 () [[arm::streaming]] |
| 138 |
{ |
| 139 |
int8x16x3_t res = produce_q2 (); |
| 140 |
asm volatile (""); |
| 141 |
consume_q2 (res); |
| 142 |
} |
| 143 |
|
| 144 |
int8x16x4_t produce_q3 (); |
| 145 |
void consume_q3 (int8x16x4_t); |
| 146 |
|
| 147 |
/* |
| 148 |
** test_q3: |
| 149 |
** ... |
| 150 |
** smstop sm |
| 151 |
** bl produce_q3 |
| 152 |
** stp q0, q1, \[sp, #?-64\]! |
| 153 |
** stp q2, q3, \[sp, #?32\] |
| 154 |
** smstart sm |
| 155 |
** ldp q2, q3, \[sp, #?32\] |
| 156 |
** ldp q0, q1, \[sp\], #?64 |
| 157 |
** stp q0, q1, \[sp, #?-64\]! |
| 158 |
** stp q2, q3, \[sp, #?32\] |
| 159 |
** smstop sm |
| 160 |
** ldp q2, q3, \[sp, #?32\] |
| 161 |
** ldp q0, q1, \[sp\], #?64 |
| 162 |
** bl consume_q3 |
| 163 |
** ... |
| 164 |
*/ |
| 165 |
void |
| 166 |
test_q3 () [[arm::streaming]] |
| 167 |
{ |
| 168 |
int8x16x4_t res = produce_q3 (); |
| 169 |
asm volatile (""); |
| 170 |
consume_q3 (res); |
| 171 |
} |
| 172 |
|
| 173 |
svint8_t produce_z0 (); |
| 174 |
void consume_z0 (svint8_t); |
| 175 |
|
| 176 |
/* |
| 177 |
** test_z0: |
| 178 |
** ... |
| 179 |
** smstop sm |
| 180 |
** bl produce_z0 |
| 181 |
** addvl sp, sp, #-1 |
| 182 |
** str z0, \[sp\] |
| 183 |
** smstart sm |
| 184 |
** ldr z0, \[sp\] |
| 185 |
** addvl sp, sp, #1 |
| 186 |
** addvl sp, sp, #-1 |
| 187 |
** str z0, \[sp\] |
| 188 |
** smstop sm |
| 189 |
** ldr z0, \[sp\] |
| 190 |
** addvl sp, sp, #1 |
| 191 |
** bl consume_z0 |
| 192 |
** ... |
| 193 |
*/ |
| 194 |
void |
| 195 |
test_z0 () [[arm::streaming]] |
| 196 |
{ |
| 197 |
svint8_t res = produce_z0 (); |
| 198 |
asm volatile (""); |
| 199 |
consume_z0 (res); |
| 200 |
} |
| 201 |
|
| 202 |
svint8x4_t produce_z3 (); |
| 203 |
void consume_z3 (svint8x4_t); |
| 204 |
|
| 205 |
/* |
| 206 |
** test_z3: |
| 207 |
** ... |
| 208 |
** smstop sm |
| 209 |
** bl produce_z3 |
| 210 |
** addvl sp, sp, #-4 |
| 211 |
** str z0, \[sp\] |
| 212 |
** str z1, \[sp, #1, mul vl\] |
| 213 |
** str z2, \[sp, #2, mul vl\] |
| 214 |
** str z3, \[sp, #3, mul vl\] |
| 215 |
** smstart sm |
| 216 |
** ldr z0, \[sp\] |
| 217 |
** ldr z1, \[sp, #1, mul vl\] |
| 218 |
** ldr z2, \[sp, #2, mul vl\] |
| 219 |
** ldr z3, \[sp, #3, mul vl\] |
| 220 |
** addvl sp, sp, #4 |
| 221 |
** addvl sp, sp, #-4 |
| 222 |
** str z0, \[sp\] |
| 223 |
** str z1, \[sp, #1, mul vl\] |
| 224 |
** str z2, \[sp, #2, mul vl\] |
| 225 |
** str z3, \[sp, #3, mul vl\] |
| 226 |
** smstop sm |
| 227 |
** ldr z0, \[sp\] |
| 228 |
** ldr z1, \[sp, #1, mul vl\] |
| 229 |
** ldr z2, \[sp, #2, mul vl\] |
| 230 |
** ldr z3, \[sp, #3, mul vl\] |
| 231 |
** addvl sp, sp, #4 |
| 232 |
** bl consume_z3 |
| 233 |
** ... |
| 234 |
*/ |
| 235 |
void |
| 236 |
test_z3 () [[arm::streaming]] |
| 237 |
{ |
| 238 |
svint8x4_t res = produce_z3 (); |
| 239 |
asm volatile (""); |
| 240 |
consume_z3 (res); |
| 241 |
} |
| 242 |
|
| 243 |
svbool_t produce_p0 (); |
| 244 |
void consume_p0 (svbool_t); |
| 245 |
|
| 246 |
/* |
| 247 |
** test_p0: |
| 248 |
** ... |
| 249 |
** smstop sm |
| 250 |
** bl produce_p0 |
| 251 |
** addvl sp, sp, #-1 |
| 252 |
** str p0, \[sp\] |
| 253 |
** smstart sm |
| 254 |
** ldr p0, \[sp\] |
| 255 |
** addvl sp, sp, #1 |
| 256 |
** addvl sp, sp, #-1 |
| 257 |
** str p0, \[sp\] |
| 258 |
** smstop sm |
| 259 |
** ldr p0, \[sp\] |
| 260 |
** addvl sp, sp, #1 |
| 261 |
** bl consume_p0 |
| 262 |
** ... |
| 263 |
*/ |
| 264 |
void |
| 265 |
test_p0 () [[arm::streaming]] |
| 266 |
{ |
| 267 |
svbool_t res = produce_p0 (); |
| 268 |
asm volatile (""); |
| 269 |
consume_p0 (res); |
| 270 |
} |
| 271 |
|
| 272 |
void consume_d7 (double, double, double, double, double, double, double, |
| 273 |
double); |
| 274 |
|
| 275 |
/* |
| 276 |
** test_d7: |
| 277 |
** ... |
| 278 |
** fmov x10, d0 |
| 279 |
** fmov x11, d1 |
| 280 |
** fmov x12, d2 |
| 281 |
** fmov x13, d3 |
| 282 |
** fmov x14, d4 |
| 283 |
** fmov x15, d5 |
| 284 |
** fmov x16, d6 |
| 285 |
** fmov x17, d7 |
| 286 |
** smstop sm |
| 287 |
** fmov d0, x10 |
| 288 |
** fmov d1, x11 |
| 289 |
** fmov d2, x12 |
| 290 |
** fmov d3, x13 |
| 291 |
** fmov d4, x14 |
| 292 |
** fmov d5, x15 |
| 293 |
** fmov d6, x16 |
| 294 |
** fmov d7, x17 |
| 295 |
** bl consume_d7 |
| 296 |
** ... |
| 297 |
*/ |
| 298 |
void |
| 299 |
test_d7 () [[arm::streaming]] |
| 300 |
{ |
| 301 |
consume_d7 (1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0); |
| 302 |
} |
| 303 |
|
| 304 |
void consume_d7_vec (int8x8_t, int8x8_t, int8x8_t, int8x8_t, int8x8_t, |
| 305 |
int8x8_t, int8x8_t, int8x8_t); |
| 306 |
|
| 307 |
/* |
| 308 |
** test_d7_vec: |
| 309 |
** ... |
| 310 |
** ( |
| 311 |
** fmov x10, d0 |
| 312 |
** fmov x11, d1 |
| 313 |
** fmov x12, d2 |
| 314 |
** fmov x13, d3 |
| 315 |
** fmov x14, d4 |
| 316 |
** fmov x15, d5 |
| 317 |
** fmov x16, d6 |
| 318 |
** fmov x17, d7 |
| 319 |
** | |
| 320 |
** umov x10, v0.d\[0\] |
| 321 |
** umov x11, v1.d\[0\] |
| 322 |
** umov x12, v2.d\[0\] |
| 323 |
** umov x13, v3.d\[0\] |
| 324 |
** umov x14, v4.d\[0\] |
| 325 |
** umov x15, v5.d\[0\] |
| 326 |
** umov x16, v6.d\[0\] |
| 327 |
** umov x17, v7.d\[0\] |
| 328 |
** ) |
| 329 |
** smstop sm |
| 330 |
** fmov d0, x10 |
| 331 |
** fmov d1, x11 |
| 332 |
** fmov d2, x12 |
| 333 |
** fmov d3, x13 |
| 334 |
** fmov d4, x14 |
| 335 |
** fmov d5, x15 |
| 336 |
** fmov d6, x16 |
| 337 |
** fmov d7, x17 |
| 338 |
** bl consume_d7_vec |
| 339 |
** ... |
| 340 |
*/ |
| 341 |
void |
| 342 |
test_d7_vec (int8x8_t *ptr) [[arm::streaming]] |
| 343 |
{ |
| 344 |
consume_d7_vec (*ptr, *ptr, *ptr, *ptr, *ptr, *ptr, *ptr, *ptr); |
| 345 |
} |
| 346 |
|
| 347 |
void consume_q7 (int8x16_t, int8x16_t, int8x16_t, int8x16_t, int8x16_t, |
| 348 |
int8x16_t, int8x16_t, int8x16_t); |
| 349 |
|
| 350 |
/* |
| 351 |
** test_q7: |
| 352 |
** ... |
| 353 |
** stp q0, q1, \[sp, #?-128\]! |
| 354 |
** stp q2, q3, \[sp, #?32\] |
| 355 |
** stp q4, q5, \[sp, #?64\] |
| 356 |
** stp q6, q7, \[sp, #?96\] |
| 357 |
** smstop sm |
| 358 |
** ldp q2, q3, \[sp, #?32\] |
| 359 |
** ldp q4, q5, \[sp, #?64\] |
| 360 |
** ldp q6, q7, \[sp, #?96\] |
| 361 |
** ldp q0, q1, \[sp\], #?128 |
| 362 |
** bl consume_q7 |
| 363 |
** ... |
| 364 |
*/ |
| 365 |
void |
| 366 |
test_q7 (int8x16_t *ptr) [[arm::streaming]] |
| 367 |
{ |
| 368 |
consume_q7 (*ptr, *ptr, *ptr, *ptr, *ptr, *ptr, *ptr, *ptr); |
| 369 |
} |
| 370 |
|
| 371 |
void consume_z7 (svint8_t, svint8_t, svint8_t, svint8_t, svint8_t, |
| 372 |
svint8_t, svint8_t, svint8_t); |
| 373 |
|
| 374 |
/* |
| 375 |
** test_z7: |
| 376 |
** ... |
| 377 |
** addvl sp, sp, #-8 |
| 378 |
** str z0, \[sp\] |
| 379 |
** str z1, \[sp, #1, mul vl\] |
| 380 |
** str z2, \[sp, #2, mul vl\] |
| 381 |
** str z3, \[sp, #3, mul vl\] |
| 382 |
** str z4, \[sp, #4, mul vl\] |
| 383 |
** str z5, \[sp, #5, mul vl\] |
| 384 |
** str z6, \[sp, #6, mul vl\] |
| 385 |
** str z7, \[sp, #7, mul vl\] |
| 386 |
** smstop sm |
| 387 |
** ldr z0, \[sp\] |
| 388 |
** ldr z1, \[sp, #1, mul vl\] |
| 389 |
** ldr z2, \[sp, #2, mul vl\] |
| 390 |
** ldr z3, \[sp, #3, mul vl\] |
| 391 |
** ldr z4, \[sp, #4, mul vl\] |
| 392 |
** ldr z5, \[sp, #5, mul vl\] |
| 393 |
** ldr z6, \[sp, #6, mul vl\] |
| 394 |
** ldr z7, \[sp, #7, mul vl\] |
| 395 |
** addvl sp, sp, #8 |
| 396 |
** bl consume_z7 |
| 397 |
** ... |
| 398 |
*/ |
| 399 |
void |
| 400 |
test_z7 (svint8_t *ptr) [[arm::streaming]] |
| 401 |
{ |
| 402 |
consume_z7 (*ptr, *ptr, *ptr, *ptr, *ptr, *ptr, *ptr, *ptr); |
| 403 |
} |
| 404 |
|
| 405 |
void consume_p3 (svbool_t, svbool_t, svbool_t, svbool_t); |
| 406 |
|
| 407 |
/* |
| 408 |
** test_p3: |
| 409 |
** ... |
| 410 |
** addvl sp, sp, #-1 |
| 411 |
** str p0, \[sp\] |
| 412 |
** str p1, \[sp, #1, mul vl\] |
| 413 |
** str p2, \[sp, #2, mul vl\] |
| 414 |
** str p3, \[sp, #3, mul vl\] |
| 415 |
** smstop sm |
| 416 |
** ldr p0, \[sp\] |
| 417 |
** ldr p1, \[sp, #1, mul vl\] |
| 418 |
** ldr p2, \[sp, #2, mul vl\] |
| 419 |
** ldr p3, \[sp, #3, mul vl\] |
| 420 |
** addvl sp, sp, #1 |
| 421 |
** bl consume_p3 |
| 422 |
** ... |
| 423 |
*/ |
| 424 |
void |
| 425 |
test_p3 (svbool_t *ptr) [[arm::streaming]] |
| 426 |
{ |
| 427 |
consume_p3 (*ptr, *ptr, *ptr, *ptr); |
| 428 |
} |
| 429 |
|
| 430 |
void consume_mixed (float, double, float32x4_t, svfloat32_t, |
| 431 |
float, double, float64x2_t, svfloat64_t, |
| 432 |
svbool_t, svbool_t, svbool_t, svbool_t); |
| 433 |
|
| 434 |
/* |
| 435 |
** test_mixed: |
| 436 |
** ... |
| 437 |
** addvl sp, sp, #-3 |
| 438 |
** str p0, \[sp\] |
| 439 |
** str p1, \[sp, #1, mul vl\] |
| 440 |
** str p2, \[sp, #2, mul vl\] |
| 441 |
** str p3, \[sp, #3, mul vl\] |
| 442 |
** str z3, \[sp, #1, mul vl\] |
| 443 |
** str z7, \[sp, #2, mul vl\] |
| 444 |
** stp q2, q6, \[sp, #?-32\]! |
| 445 |
** fmov w10, s0 |
| 446 |
** fmov x11, d1 |
| 447 |
** fmov w12, s4 |
| 448 |
** fmov x13, d5 |
| 449 |
** smstop sm |
| 450 |
** fmov s0, w10 |
| 451 |
** fmov d1, x11 |
| 452 |
** fmov s4, w12 |
| 453 |
** fmov d5, x13 |
| 454 |
** ldp q2, q6, \[sp\], #?32 |
| 455 |
** ldr p0, \[sp\] |
| 456 |
** ldr p1, \[sp, #1, mul vl\] |
| 457 |
** ldr p2, \[sp, #2, mul vl\] |
| 458 |
** ldr p3, \[sp, #3, mul vl\] |
| 459 |
** ldr z3, \[sp, #1, mul vl\] |
| 460 |
** ldr z7, \[sp, #2, mul vl\] |
| 461 |
** addvl sp, sp, #3 |
| 462 |
** bl consume_mixed |
| 463 |
** ... |
| 464 |
*/ |
| 465 |
void |
| 466 |
test_mixed (float32x4_t *float32x4_ptr, |
| 467 |
svfloat32_t *svfloat32_ptr, |
| 468 |
float64x2_t *float64x2_ptr, |
| 469 |
svfloat64_t *svfloat64_ptr, |
| 470 |
svbool_t *svbool_ptr) [[arm::streaming]] |
| 471 |
{ |
| 472 |
consume_mixed (1.0f, 2.0, *float32x4_ptr, *svfloat32_ptr, |
| 473 |
3.0f, 4.0, *float64x2_ptr, *svfloat64_ptr, |
| 474 |
*svbool_ptr, *svbool_ptr, *svbool_ptr, *svbool_ptr); |
| 475 |
} |
| 476 |
|
| 477 |
void consume_varargs (float, ...); |
| 478 |
|
| 479 |
/* |
| 480 |
** test_varargs: |
| 481 |
** ... |
| 482 |
** stp q3, q7, \[sp, #?-32\]! |
| 483 |
** fmov w10, s0 |
| 484 |
** fmov x11, d1 |
| 485 |
** ( |
| 486 |
** fmov x12, d2 |
| 487 |
** | |
| 488 |
** umov x12, v2.d\[0\] |
| 489 |
** ) |
| 490 |
** fmov x13, d4 |
| 491 |
** fmov x14, d5 |
| 492 |
** ( |
| 493 |
** fmov x15, d6 |
| 494 |
** | |
| 495 |
** umov x15, v6.d\[0\] |
| 496 |
** ) |
| 497 |
** smstop sm |
| 498 |
** fmov s0, w10 |
| 499 |
** fmov d1, x11 |
| 500 |
** fmov d2, x12 |
| 501 |
** fmov d4, x13 |
| 502 |
** fmov d5, x14 |
| 503 |
** fmov d6, x15 |
| 504 |
** ldp q3, q7, \[sp\], #?32 |
| 505 |
** bl consume_varargs |
| 506 |
** ... |
| 507 |
*/ |
| 508 |
void |
| 509 |
test_varargs (float32x2_t *float32x2_ptr, |
| 510 |
float32x4_t *float32x4_ptr, |
| 511 |
float64x1_t *float64x1_ptr, |
| 512 |
float64x2_t *float64x2_ptr) [[arm::streaming]] |
| 513 |
{ |
| 514 |
consume_varargs (1.0f, 2.0, *float32x2_ptr, *float32x4_ptr, |
| 515 |
3.0f, 4.0, *float64x1_ptr, *float64x2_ptr); |
| 516 |
} |