1 /* 2 * PowerPC integer and vector emulation helpers for QEMU. 3 * 4 * Copyright (c) 2003-2007 Jocelyn Mayer 5 * 6 * This library is free software; you can redistribute it and/or 7 * modify it under the terms of the GNU Lesser General Public 8 * License as published by the Free Software Foundation; either 9 * version 2 of the License, or (at your option) any later version. 10 * 11 * This library is distributed in the hope that it will be useful, 12 * but WITHOUT ANY WARRANTY; without even the implied warranty of 13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 14 * Lesser General Public License for more details. 15 * 16 * You should have received a copy of the GNU Lesser General Public 17 * License along with this library; if not, see <http://www.gnu.org/licenses/>. 18 */ 19 #include "qemu/osdep.h" 20 #include "cpu.h" 21 #include "internal.h" 22 #include "qemu/host-utils.h" 23 #include "exec/helper-proto.h" 24 #include "crypto/aes.h" 25 #include "fpu/softfloat.h" 26 #include "qapi/error.h" 27 #include "qemu/guest-random.h" 28 29 #include "helper_regs.h" 30 /*****************************************************************************/ 31 /* Fixed point operations helpers */ 32 33 static inline void helper_update_ov_legacy(CPUPPCState *env, int ov) 34 { 35 if (unlikely(ov)) { 36 env->so = env->ov = 1; 37 } else { 38 env->ov = 0; 39 } 40 } 41 42 target_ulong helper_divweu(CPUPPCState *env, target_ulong ra, target_ulong rb, 43 uint32_t oe) 44 { 45 uint64_t rt = 0; 46 int overflow = 0; 47 48 uint64_t dividend = (uint64_t)ra << 32; 49 uint64_t divisor = (uint32_t)rb; 50 51 if (unlikely(divisor == 0)) { 52 overflow = 1; 53 } else { 54 rt = dividend / divisor; 55 overflow = rt > UINT32_MAX; 56 } 57 58 if (unlikely(overflow)) { 59 rt = 0; /* Undefined */ 60 } 61 62 if (oe) { 63 helper_update_ov_legacy(env, overflow); 64 } 65 66 return (target_ulong)rt; 67 } 68 69 target_ulong helper_divwe(CPUPPCState *env, target_ulong ra, target_ulong rb, 70 uint32_t oe) 71 { 72 int64_t rt = 0; 73 int overflow = 0; 74 75 int64_t dividend = (int64_t)ra << 32; 76 int64_t divisor = (int64_t)((int32_t)rb); 77 78 if (unlikely((divisor == 0) || 79 ((divisor == -1ull) && (dividend == INT64_MIN)))) { 80 overflow = 1; 81 } else { 82 rt = dividend / divisor; 83 overflow = rt != (int32_t)rt; 84 } 85 86 if (unlikely(overflow)) { 87 rt = 0; /* Undefined */ 88 } 89 90 if (oe) { 91 helper_update_ov_legacy(env, overflow); 92 } 93 94 return (target_ulong)rt; 95 } 96 97 #if defined(TARGET_PPC64) 98 99 uint64_t helper_divdeu(CPUPPCState *env, uint64_t ra, uint64_t rb, uint32_t oe) 100 { 101 uint64_t rt = 0; 102 int overflow = 0; 103 104 overflow = divu128(&rt, &ra, rb); 105 106 if (unlikely(overflow)) { 107 rt = 0; /* Undefined */ 108 } 109 110 if (oe) { 111 helper_update_ov_legacy(env, overflow); 112 } 113 114 return rt; 115 } 116 117 uint64_t helper_divde(CPUPPCState *env, uint64_t rau, uint64_t rbu, uint32_t oe) 118 { 119 int64_t rt = 0; 120 int64_t ra = (int64_t)rau; 121 int64_t rb = (int64_t)rbu; 122 int overflow = divs128(&rt, &ra, rb); 123 124 if (unlikely(overflow)) { 125 rt = 0; /* Undefined */ 126 } 127 128 if (oe) { 129 helper_update_ov_legacy(env, overflow); 130 } 131 132 return rt; 133 } 134 135 #endif 136 137 138 #if defined(TARGET_PPC64) 139 /* if x = 0xab, returns 0xababababababababa */ 140 #define pattern(x) (((x) & 0xff) * (~(target_ulong)0 / 0xff)) 141 142 /* 143 * subtract 1 from each byte, and with inverse, check if MSB is set at each 144 * byte. 145 * i.e. ((0x00 - 0x01) & ~(0x00)) & 0x80 146 * (0xFF & 0xFF) & 0x80 = 0x80 (zero found) 147 */ 148 #define haszero(v) (((v) - pattern(0x01)) & ~(v) & pattern(0x80)) 149 150 /* When you XOR the pattern and there is a match, that byte will be zero */ 151 #define hasvalue(x, n) (haszero((x) ^ pattern(n))) 152 153 uint32_t helper_cmpeqb(target_ulong ra, target_ulong rb) 154 { 155 return hasvalue(rb, ra) ? CRF_GT : 0; 156 } 157 158 #undef pattern 159 #undef haszero 160 #undef hasvalue 161 162 /* 163 * Return a random number. 164 */ 165 uint64_t helper_darn32(void) 166 { 167 Error *err = NULL; 168 uint32_t ret; 169 170 if (qemu_guest_getrandom(&ret, sizeof(ret), &err) < 0) { 171 qemu_log_mask(LOG_UNIMP, "darn: Crypto failure: %s", 172 error_get_pretty(err)); 173 error_free(err); 174 return -1; 175 } 176 177 return ret; 178 } 179 180 uint64_t helper_darn64(void) 181 { 182 Error *err = NULL; 183 uint64_t ret; 184 185 if (qemu_guest_getrandom(&ret, sizeof(ret), &err) < 0) { 186 qemu_log_mask(LOG_UNIMP, "darn: Crypto failure: %s", 187 error_get_pretty(err)); 188 error_free(err); 189 return -1; 190 } 191 192 return ret; 193 } 194 195 uint64_t helper_bpermd(uint64_t rs, uint64_t rb) 196 { 197 int i; 198 uint64_t ra = 0; 199 200 for (i = 0; i < 8; i++) { 201 int index = (rs >> (i * 8)) & 0xFF; 202 if (index < 64) { 203 if (rb & PPC_BIT(index)) { 204 ra |= 1 << i; 205 } 206 } 207 } 208 return ra; 209 } 210 211 #endif 212 213 target_ulong helper_cmpb(target_ulong rs, target_ulong rb) 214 { 215 target_ulong mask = 0xff; 216 target_ulong ra = 0; 217 int i; 218 219 for (i = 0; i < sizeof(target_ulong); i++) { 220 if ((rs & mask) == (rb & mask)) { 221 ra |= mask; 222 } 223 mask <<= 8; 224 } 225 return ra; 226 } 227 228 /* shift right arithmetic helper */ 229 target_ulong helper_sraw(CPUPPCState *env, target_ulong value, 230 target_ulong shift) 231 { 232 int32_t ret; 233 234 if (likely(!(shift & 0x20))) { 235 if (likely((uint32_t)shift != 0)) { 236 shift &= 0x1f; 237 ret = (int32_t)value >> shift; 238 if (likely(ret >= 0 || (value & ((1 << shift) - 1)) == 0)) { 239 env->ca32 = env->ca = 0; 240 } else { 241 env->ca32 = env->ca = 1; 242 } 243 } else { 244 ret = (int32_t)value; 245 env->ca32 = env->ca = 0; 246 } 247 } else { 248 ret = (int32_t)value >> 31; 249 env->ca32 = env->ca = (ret != 0); 250 } 251 return (target_long)ret; 252 } 253 254 #if defined(TARGET_PPC64) 255 target_ulong helper_srad(CPUPPCState *env, target_ulong value, 256 target_ulong shift) 257 { 258 int64_t ret; 259 260 if (likely(!(shift & 0x40))) { 261 if (likely((uint64_t)shift != 0)) { 262 shift &= 0x3f; 263 ret = (int64_t)value >> shift; 264 if (likely(ret >= 0 || (value & ((1ULL << shift) - 1)) == 0)) { 265 env->ca32 = env->ca = 0; 266 } else { 267 env->ca32 = env->ca = 1; 268 } 269 } else { 270 ret = (int64_t)value; 271 env->ca32 = env->ca = 0; 272 } 273 } else { 274 ret = (int64_t)value >> 63; 275 env->ca32 = env->ca = (ret != 0); 276 } 277 return ret; 278 } 279 #endif 280 281 #if defined(TARGET_PPC64) 282 target_ulong helper_popcntb(target_ulong val) 283 { 284 /* Note that we don't fold past bytes */ 285 val = (val & 0x5555555555555555ULL) + ((val >> 1) & 286 0x5555555555555555ULL); 287 val = (val & 0x3333333333333333ULL) + ((val >> 2) & 288 0x3333333333333333ULL); 289 val = (val & 0x0f0f0f0f0f0f0f0fULL) + ((val >> 4) & 290 0x0f0f0f0f0f0f0f0fULL); 291 return val; 292 } 293 294 target_ulong helper_popcntw(target_ulong val) 295 { 296 /* Note that we don't fold past words. */ 297 val = (val & 0x5555555555555555ULL) + ((val >> 1) & 298 0x5555555555555555ULL); 299 val = (val & 0x3333333333333333ULL) + ((val >> 2) & 300 0x3333333333333333ULL); 301 val = (val & 0x0f0f0f0f0f0f0f0fULL) + ((val >> 4) & 302 0x0f0f0f0f0f0f0f0fULL); 303 val = (val & 0x00ff00ff00ff00ffULL) + ((val >> 8) & 304 0x00ff00ff00ff00ffULL); 305 val = (val & 0x0000ffff0000ffffULL) + ((val >> 16) & 306 0x0000ffff0000ffffULL); 307 return val; 308 } 309 #else 310 target_ulong helper_popcntb(target_ulong val) 311 { 312 /* Note that we don't fold past bytes */ 313 val = (val & 0x55555555) + ((val >> 1) & 0x55555555); 314 val = (val & 0x33333333) + ((val >> 2) & 0x33333333); 315 val = (val & 0x0f0f0f0f) + ((val >> 4) & 0x0f0f0f0f); 316 return val; 317 } 318 #endif 319 320 /*****************************************************************************/ 321 /* PowerPC 601 specific instructions (POWER bridge) */ 322 target_ulong helper_div(CPUPPCState *env, target_ulong arg1, target_ulong arg2) 323 { 324 uint64_t tmp = (uint64_t)arg1 << 32 | env->spr[SPR_MQ]; 325 326 if (((int32_t)tmp == INT32_MIN && (int32_t)arg2 == (int32_t)-1) || 327 (int32_t)arg2 == 0) { 328 env->spr[SPR_MQ] = 0; 329 return INT32_MIN; 330 } else { 331 env->spr[SPR_MQ] = tmp % arg2; 332 return tmp / (int32_t)arg2; 333 } 334 } 335 336 target_ulong helper_divo(CPUPPCState *env, target_ulong arg1, 337 target_ulong arg2) 338 { 339 uint64_t tmp = (uint64_t)arg1 << 32 | env->spr[SPR_MQ]; 340 341 if (((int32_t)tmp == INT32_MIN && (int32_t)arg2 == (int32_t)-1) || 342 (int32_t)arg2 == 0) { 343 env->so = env->ov = 1; 344 env->spr[SPR_MQ] = 0; 345 return INT32_MIN; 346 } else { 347 env->spr[SPR_MQ] = tmp % arg2; 348 tmp /= (int32_t)arg2; 349 if ((int32_t)tmp != tmp) { 350 env->so = env->ov = 1; 351 } else { 352 env->ov = 0; 353 } 354 return tmp; 355 } 356 } 357 358 target_ulong helper_divs(CPUPPCState *env, target_ulong arg1, 359 target_ulong arg2) 360 { 361 if (((int32_t)arg1 == INT32_MIN && (int32_t)arg2 == (int32_t)-1) || 362 (int32_t)arg2 == 0) { 363 env->spr[SPR_MQ] = 0; 364 return INT32_MIN; 365 } else { 366 env->spr[SPR_MQ] = (int32_t)arg1 % (int32_t)arg2; 367 return (int32_t)arg1 / (int32_t)arg2; 368 } 369 } 370 371 target_ulong helper_divso(CPUPPCState *env, target_ulong arg1, 372 target_ulong arg2) 373 { 374 if (((int32_t)arg1 == INT32_MIN && (int32_t)arg2 == (int32_t)-1) || 375 (int32_t)arg2 == 0) { 376 env->so = env->ov = 1; 377 env->spr[SPR_MQ] = 0; 378 return INT32_MIN; 379 } else { 380 env->ov = 0; 381 env->spr[SPR_MQ] = (int32_t)arg1 % (int32_t)arg2; 382 return (int32_t)arg1 / (int32_t)arg2; 383 } 384 } 385 386 /*****************************************************************************/ 387 /* 602 specific instructions */ 388 /* mfrom is the most crazy instruction ever seen, imho ! */ 389 /* Real implementation uses a ROM table. Do the same */ 390 /* 391 * Extremely decomposed: 392 * -arg / 256 393 * return 256 * log10(10 + 1.0) + 0.5 394 */ 395 #if !defined(CONFIG_USER_ONLY) 396 target_ulong helper_602_mfrom(target_ulong arg) 397 { 398 if (likely(arg < 602)) { 399 #include "mfrom_table.inc.c" 400 return mfrom_ROM_table[arg]; 401 } else { 402 return 0; 403 } 404 } 405 #endif 406 407 /*****************************************************************************/ 408 /* Altivec extension helpers */ 409 #if defined(HOST_WORDS_BIGENDIAN) 410 #define VECTOR_FOR_INORDER_I(index, element) \ 411 for (index = 0; index < ARRAY_SIZE(r->element); index++) 412 #else 413 #define VECTOR_FOR_INORDER_I(index, element) \ 414 for (index = ARRAY_SIZE(r->element) - 1; index >= 0; index--) 415 #endif 416 417 /* Saturating arithmetic helpers. */ 418 #define SATCVT(from, to, from_type, to_type, min, max) \ 419 static inline to_type cvt##from##to(from_type x, int *sat) \ 420 { \ 421 to_type r; \ 422 \ 423 if (x < (from_type)min) { \ 424 r = min; \ 425 *sat = 1; \ 426 } else if (x > (from_type)max) { \ 427 r = max; \ 428 *sat = 1; \ 429 } else { \ 430 r = x; \ 431 } \ 432 return r; \ 433 } 434 #define SATCVTU(from, to, from_type, to_type, min, max) \ 435 static inline to_type cvt##from##to(from_type x, int *sat) \ 436 { \ 437 to_type r; \ 438 \ 439 if (x > (from_type)max) { \ 440 r = max; \ 441 *sat = 1; \ 442 } else { \ 443 r = x; \ 444 } \ 445 return r; \ 446 } 447 SATCVT(sh, sb, int16_t, int8_t, INT8_MIN, INT8_MAX) 448 SATCVT(sw, sh, int32_t, int16_t, INT16_MIN, INT16_MAX) 449 SATCVT(sd, sw, int64_t, int32_t, INT32_MIN, INT32_MAX) 450 451 SATCVTU(uh, ub, uint16_t, uint8_t, 0, UINT8_MAX) 452 SATCVTU(uw, uh, uint32_t, uint16_t, 0, UINT16_MAX) 453 SATCVTU(ud, uw, uint64_t, uint32_t, 0, UINT32_MAX) 454 SATCVT(sh, ub, int16_t, uint8_t, 0, UINT8_MAX) 455 SATCVT(sw, uh, int32_t, uint16_t, 0, UINT16_MAX) 456 SATCVT(sd, uw, int64_t, uint32_t, 0, UINT32_MAX) 457 #undef SATCVT 458 #undef SATCVTU 459 460 void helper_lvsl(ppc_avr_t *r, target_ulong sh) 461 { 462 int i, j = (sh & 0xf); 463 464 for (i = 0; i < ARRAY_SIZE(r->u8); i++) { 465 r->VsrB(i) = j++; 466 } 467 } 468 469 void helper_lvsr(ppc_avr_t *r, target_ulong sh) 470 { 471 int i, j = 0x10 - (sh & 0xf); 472 473 for (i = 0; i < ARRAY_SIZE(r->u8); i++) { 474 r->VsrB(i) = j++; 475 } 476 } 477 478 void helper_mtvscr(CPUPPCState *env, uint32_t vscr) 479 { 480 env->vscr = vscr & ~(1u << VSCR_SAT); 481 /* Which bit we set is completely arbitrary, but clear the rest. */ 482 env->vscr_sat.u64[0] = vscr & (1u << VSCR_SAT); 483 env->vscr_sat.u64[1] = 0; 484 set_flush_to_zero((vscr >> VSCR_NJ) & 1, &env->vec_status); 485 } 486 487 uint32_t helper_mfvscr(CPUPPCState *env) 488 { 489 uint32_t sat = (env->vscr_sat.u64[0] | env->vscr_sat.u64[1]) != 0; 490 return env->vscr | (sat << VSCR_SAT); 491 } 492 493 static inline void set_vscr_sat(CPUPPCState *env) 494 { 495 /* The choice of non-zero value is arbitrary. */ 496 env->vscr_sat.u32[0] = 1; 497 } 498 499 void helper_vaddcuw(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 500 { 501 int i; 502 503 for (i = 0; i < ARRAY_SIZE(r->u32); i++) { 504 r->u32[i] = ~a->u32[i] < b->u32[i]; 505 } 506 } 507 508 /* vprtybw */ 509 void helper_vprtybw(ppc_avr_t *r, ppc_avr_t *b) 510 { 511 int i; 512 for (i = 0; i < ARRAY_SIZE(r->u32); i++) { 513 uint64_t res = b->u32[i] ^ (b->u32[i] >> 16); 514 res ^= res >> 8; 515 r->u32[i] = res & 1; 516 } 517 } 518 519 /* vprtybd */ 520 void helper_vprtybd(ppc_avr_t *r, ppc_avr_t *b) 521 { 522 int i; 523 for (i = 0; i < ARRAY_SIZE(r->u64); i++) { 524 uint64_t res = b->u64[i] ^ (b->u64[i] >> 32); 525 res ^= res >> 16; 526 res ^= res >> 8; 527 r->u64[i] = res & 1; 528 } 529 } 530 531 /* vprtybq */ 532 void helper_vprtybq(ppc_avr_t *r, ppc_avr_t *b) 533 { 534 uint64_t res = b->u64[0] ^ b->u64[1]; 535 res ^= res >> 32; 536 res ^= res >> 16; 537 res ^= res >> 8; 538 r->VsrD(1) = res & 1; 539 r->VsrD(0) = 0; 540 } 541 542 #define VARITH_DO(name, op, element) \ 543 void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ 544 { \ 545 int i; \ 546 \ 547 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ 548 r->element[i] = a->element[i] op b->element[i]; \ 549 } \ 550 } 551 VARITH_DO(muluwm, *, u32) 552 #undef VARITH_DO 553 #undef VARITH 554 555 #define VARITHFP(suffix, func) \ 556 void helper_v##suffix(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, \ 557 ppc_avr_t *b) \ 558 { \ 559 int i; \ 560 \ 561 for (i = 0; i < ARRAY_SIZE(r->f32); i++) { \ 562 r->f32[i] = func(a->f32[i], b->f32[i], &env->vec_status); \ 563 } \ 564 } 565 VARITHFP(addfp, float32_add) 566 VARITHFP(subfp, float32_sub) 567 VARITHFP(minfp, float32_min) 568 VARITHFP(maxfp, float32_max) 569 #undef VARITHFP 570 571 #define VARITHFPFMA(suffix, type) \ 572 void helper_v##suffix(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, \ 573 ppc_avr_t *b, ppc_avr_t *c) \ 574 { \ 575 int i; \ 576 for (i = 0; i < ARRAY_SIZE(r->f32); i++) { \ 577 r->f32[i] = float32_muladd(a->f32[i], c->f32[i], b->f32[i], \ 578 type, &env->vec_status); \ 579 } \ 580 } 581 VARITHFPFMA(maddfp, 0); 582 VARITHFPFMA(nmsubfp, float_muladd_negate_result | float_muladd_negate_c); 583 #undef VARITHFPFMA 584 585 #define VARITHSAT_CASE(type, op, cvt, element) \ 586 { \ 587 type result = (type)a->element[i] op (type)b->element[i]; \ 588 r->element[i] = cvt(result, &sat); \ 589 } 590 591 #define VARITHSAT_DO(name, op, optype, cvt, element) \ 592 void helper_v##name(ppc_avr_t *r, ppc_avr_t *vscr_sat, \ 593 ppc_avr_t *a, ppc_avr_t *b, uint32_t desc) \ 594 { \ 595 int sat = 0; \ 596 int i; \ 597 \ 598 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ 599 VARITHSAT_CASE(optype, op, cvt, element); \ 600 } \ 601 if (sat) { \ 602 vscr_sat->u32[0] = 1; \ 603 } \ 604 } 605 #define VARITHSAT_SIGNED(suffix, element, optype, cvt) \ 606 VARITHSAT_DO(adds##suffix##s, +, optype, cvt, element) \ 607 VARITHSAT_DO(subs##suffix##s, -, optype, cvt, element) 608 #define VARITHSAT_UNSIGNED(suffix, element, optype, cvt) \ 609 VARITHSAT_DO(addu##suffix##s, +, optype, cvt, element) \ 610 VARITHSAT_DO(subu##suffix##s, -, optype, cvt, element) 611 VARITHSAT_SIGNED(b, s8, int16_t, cvtshsb) 612 VARITHSAT_SIGNED(h, s16, int32_t, cvtswsh) 613 VARITHSAT_SIGNED(w, s32, int64_t, cvtsdsw) 614 VARITHSAT_UNSIGNED(b, u8, uint16_t, cvtshub) 615 VARITHSAT_UNSIGNED(h, u16, uint32_t, cvtswuh) 616 VARITHSAT_UNSIGNED(w, u32, uint64_t, cvtsduw) 617 #undef VARITHSAT_CASE 618 #undef VARITHSAT_DO 619 #undef VARITHSAT_SIGNED 620 #undef VARITHSAT_UNSIGNED 621 622 #define VAVG_DO(name, element, etype) \ 623 void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ 624 { \ 625 int i; \ 626 \ 627 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ 628 etype x = (etype)a->element[i] + (etype)b->element[i] + 1; \ 629 r->element[i] = x >> 1; \ 630 } \ 631 } 632 633 #define VAVG(type, signed_element, signed_type, unsigned_element, \ 634 unsigned_type) \ 635 VAVG_DO(avgs##type, signed_element, signed_type) \ 636 VAVG_DO(avgu##type, unsigned_element, unsigned_type) 637 VAVG(b, s8, int16_t, u8, uint16_t) 638 VAVG(h, s16, int32_t, u16, uint32_t) 639 VAVG(w, s32, int64_t, u32, uint64_t) 640 #undef VAVG_DO 641 #undef VAVG 642 643 #define VABSDU_DO(name, element) \ 644 void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ 645 { \ 646 int i; \ 647 \ 648 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ 649 r->element[i] = (a->element[i] > b->element[i]) ? \ 650 (a->element[i] - b->element[i]) : \ 651 (b->element[i] - a->element[i]); \ 652 } \ 653 } 654 655 /* 656 * VABSDU - Vector absolute difference unsigned 657 * name - instruction mnemonic suffix (b: byte, h: halfword, w: word) 658 * element - element type to access from vector 659 */ 660 #define VABSDU(type, element) \ 661 VABSDU_DO(absdu##type, element) 662 VABSDU(b, u8) 663 VABSDU(h, u16) 664 VABSDU(w, u32) 665 #undef VABSDU_DO 666 #undef VABSDU 667 668 #define VCF(suffix, cvt, element) \ 669 void helper_vcf##suffix(CPUPPCState *env, ppc_avr_t *r, \ 670 ppc_avr_t *b, uint32_t uim) \ 671 { \ 672 int i; \ 673 \ 674 for (i = 0; i < ARRAY_SIZE(r->f32); i++) { \ 675 float32 t = cvt(b->element[i], &env->vec_status); \ 676 r->f32[i] = float32_scalbn(t, -uim, &env->vec_status); \ 677 } \ 678 } 679 VCF(ux, uint32_to_float32, u32) 680 VCF(sx, int32_to_float32, s32) 681 #undef VCF 682 683 #define VCMP_DO(suffix, compare, element, record) \ 684 void helper_vcmp##suffix(CPUPPCState *env, ppc_avr_t *r, \ 685 ppc_avr_t *a, ppc_avr_t *b) \ 686 { \ 687 uint64_t ones = (uint64_t)-1; \ 688 uint64_t all = ones; \ 689 uint64_t none = 0; \ 690 int i; \ 691 \ 692 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ 693 uint64_t result = (a->element[i] compare b->element[i] ? \ 694 ones : 0x0); \ 695 switch (sizeof(a->element[0])) { \ 696 case 8: \ 697 r->u64[i] = result; \ 698 break; \ 699 case 4: \ 700 r->u32[i] = result; \ 701 break; \ 702 case 2: \ 703 r->u16[i] = result; \ 704 break; \ 705 case 1: \ 706 r->u8[i] = result; \ 707 break; \ 708 } \ 709 all &= result; \ 710 none |= result; \ 711 } \ 712 if (record) { \ 713 env->crf[6] = ((all != 0) << 3) | ((none == 0) << 1); \ 714 } \ 715 } 716 #define VCMP(suffix, compare, element) \ 717 VCMP_DO(suffix, compare, element, 0) \ 718 VCMP_DO(suffix##_dot, compare, element, 1) 719 VCMP(equb, ==, u8) 720 VCMP(equh, ==, u16) 721 VCMP(equw, ==, u32) 722 VCMP(equd, ==, u64) 723 VCMP(gtub, >, u8) 724 VCMP(gtuh, >, u16) 725 VCMP(gtuw, >, u32) 726 VCMP(gtud, >, u64) 727 VCMP(gtsb, >, s8) 728 VCMP(gtsh, >, s16) 729 VCMP(gtsw, >, s32) 730 VCMP(gtsd, >, s64) 731 #undef VCMP_DO 732 #undef VCMP 733 734 #define VCMPNE_DO(suffix, element, etype, cmpzero, record) \ 735 void helper_vcmpne##suffix(CPUPPCState *env, ppc_avr_t *r, \ 736 ppc_avr_t *a, ppc_avr_t *b) \ 737 { \ 738 etype ones = (etype)-1; \ 739 etype all = ones; \ 740 etype result, none = 0; \ 741 int i; \ 742 \ 743 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ 744 if (cmpzero) { \ 745 result = ((a->element[i] == 0) \ 746 || (b->element[i] == 0) \ 747 || (a->element[i] != b->element[i]) ? \ 748 ones : 0x0); \ 749 } else { \ 750 result = (a->element[i] != b->element[i]) ? ones : 0x0; \ 751 } \ 752 r->element[i] = result; \ 753 all &= result; \ 754 none |= result; \ 755 } \ 756 if (record) { \ 757 env->crf[6] = ((all != 0) << 3) | ((none == 0) << 1); \ 758 } \ 759 } 760 761 /* 762 * VCMPNEZ - Vector compare not equal to zero 763 * suffix - instruction mnemonic suffix (b: byte, h: halfword, w: word) 764 * element - element type to access from vector 765 */ 766 #define VCMPNE(suffix, element, etype, cmpzero) \ 767 VCMPNE_DO(suffix, element, etype, cmpzero, 0) \ 768 VCMPNE_DO(suffix##_dot, element, etype, cmpzero, 1) 769 VCMPNE(zb, u8, uint8_t, 1) 770 VCMPNE(zh, u16, uint16_t, 1) 771 VCMPNE(zw, u32, uint32_t, 1) 772 VCMPNE(b, u8, uint8_t, 0) 773 VCMPNE(h, u16, uint16_t, 0) 774 VCMPNE(w, u32, uint32_t, 0) 775 #undef VCMPNE_DO 776 #undef VCMPNE 777 778 #define VCMPFP_DO(suffix, compare, order, record) \ 779 void helper_vcmp##suffix(CPUPPCState *env, ppc_avr_t *r, \ 780 ppc_avr_t *a, ppc_avr_t *b) \ 781 { \ 782 uint32_t ones = (uint32_t)-1; \ 783 uint32_t all = ones; \ 784 uint32_t none = 0; \ 785 int i; \ 786 \ 787 for (i = 0; i < ARRAY_SIZE(r->f32); i++) { \ 788 uint32_t result; \ 789 int rel = float32_compare_quiet(a->f32[i], b->f32[i], \ 790 &env->vec_status); \ 791 if (rel == float_relation_unordered) { \ 792 result = 0; \ 793 } else if (rel compare order) { \ 794 result = ones; \ 795 } else { \ 796 result = 0; \ 797 } \ 798 r->u32[i] = result; \ 799 all &= result; \ 800 none |= result; \ 801 } \ 802 if (record) { \ 803 env->crf[6] = ((all != 0) << 3) | ((none == 0) << 1); \ 804 } \ 805 } 806 #define VCMPFP(suffix, compare, order) \ 807 VCMPFP_DO(suffix, compare, order, 0) \ 808 VCMPFP_DO(suffix##_dot, compare, order, 1) 809 VCMPFP(eqfp, ==, float_relation_equal) 810 VCMPFP(gefp, !=, float_relation_less) 811 VCMPFP(gtfp, ==, float_relation_greater) 812 #undef VCMPFP_DO 813 #undef VCMPFP 814 815 static inline void vcmpbfp_internal(CPUPPCState *env, ppc_avr_t *r, 816 ppc_avr_t *a, ppc_avr_t *b, int record) 817 { 818 int i; 819 int all_in = 0; 820 821 for (i = 0; i < ARRAY_SIZE(r->f32); i++) { 822 int le_rel = float32_compare_quiet(a->f32[i], b->f32[i], 823 &env->vec_status); 824 if (le_rel == float_relation_unordered) { 825 r->u32[i] = 0xc0000000; 826 all_in = 1; 827 } else { 828 float32 bneg = float32_chs(b->f32[i]); 829 int ge_rel = float32_compare_quiet(a->f32[i], bneg, 830 &env->vec_status); 831 int le = le_rel != float_relation_greater; 832 int ge = ge_rel != float_relation_less; 833 834 r->u32[i] = ((!le) << 31) | ((!ge) << 30); 835 all_in |= (!le | !ge); 836 } 837 } 838 if (record) { 839 env->crf[6] = (all_in == 0) << 1; 840 } 841 } 842 843 void helper_vcmpbfp(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 844 { 845 vcmpbfp_internal(env, r, a, b, 0); 846 } 847 848 void helper_vcmpbfp_dot(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, 849 ppc_avr_t *b) 850 { 851 vcmpbfp_internal(env, r, a, b, 1); 852 } 853 854 #define VCT(suffix, satcvt, element) \ 855 void helper_vct##suffix(CPUPPCState *env, ppc_avr_t *r, \ 856 ppc_avr_t *b, uint32_t uim) \ 857 { \ 858 int i; \ 859 int sat = 0; \ 860 float_status s = env->vec_status; \ 861 \ 862 set_float_rounding_mode(float_round_to_zero, &s); \ 863 for (i = 0; i < ARRAY_SIZE(r->f32); i++) { \ 864 if (float32_is_any_nan(b->f32[i])) { \ 865 r->element[i] = 0; \ 866 } else { \ 867 float64 t = float32_to_float64(b->f32[i], &s); \ 868 int64_t j; \ 869 \ 870 t = float64_scalbn(t, uim, &s); \ 871 j = float64_to_int64(t, &s); \ 872 r->element[i] = satcvt(j, &sat); \ 873 } \ 874 } \ 875 if (sat) { \ 876 set_vscr_sat(env); \ 877 } \ 878 } 879 VCT(uxs, cvtsduw, u32) 880 VCT(sxs, cvtsdsw, s32) 881 #undef VCT 882 883 target_ulong helper_vclzlsbb(ppc_avr_t *r) 884 { 885 target_ulong count = 0; 886 int i; 887 for (i = 0; i < ARRAY_SIZE(r->u8); i++) { 888 if (r->VsrB(i) & 0x01) { 889 break; 890 } 891 count++; 892 } 893 return count; 894 } 895 896 target_ulong helper_vctzlsbb(ppc_avr_t *r) 897 { 898 target_ulong count = 0; 899 int i; 900 for (i = ARRAY_SIZE(r->u8) - 1; i >= 0; i--) { 901 if (r->VsrB(i) & 0x01) { 902 break; 903 } 904 count++; 905 } 906 return count; 907 } 908 909 void helper_vmhaddshs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, 910 ppc_avr_t *b, ppc_avr_t *c) 911 { 912 int sat = 0; 913 int i; 914 915 for (i = 0; i < ARRAY_SIZE(r->s16); i++) { 916 int32_t prod = a->s16[i] * b->s16[i]; 917 int32_t t = (int32_t)c->s16[i] + (prod >> 15); 918 919 r->s16[i] = cvtswsh(t, &sat); 920 } 921 922 if (sat) { 923 set_vscr_sat(env); 924 } 925 } 926 927 void helper_vmhraddshs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, 928 ppc_avr_t *b, ppc_avr_t *c) 929 { 930 int sat = 0; 931 int i; 932 933 for (i = 0; i < ARRAY_SIZE(r->s16); i++) { 934 int32_t prod = a->s16[i] * b->s16[i] + 0x00004000; 935 int32_t t = (int32_t)c->s16[i] + (prod >> 15); 936 r->s16[i] = cvtswsh(t, &sat); 937 } 938 939 if (sat) { 940 set_vscr_sat(env); 941 } 942 } 943 944 void helper_vmladduhm(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c) 945 { 946 int i; 947 948 for (i = 0; i < ARRAY_SIZE(r->s16); i++) { 949 int32_t prod = a->s16[i] * b->s16[i]; 950 r->s16[i] = (int16_t) (prod + c->s16[i]); 951 } 952 } 953 954 #define VMRG_DO(name, element, access, ofs) \ 955 void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ 956 { \ 957 ppc_avr_t result; \ 958 int i, half = ARRAY_SIZE(r->element) / 2; \ 959 \ 960 for (i = 0; i < half; i++) { \ 961 result.access(i * 2 + 0) = a->access(i + ofs); \ 962 result.access(i * 2 + 1) = b->access(i + ofs); \ 963 } \ 964 *r = result; \ 965 } 966 967 #define VMRG(suffix, element, access) \ 968 VMRG_DO(mrgl##suffix, element, access, half) \ 969 VMRG_DO(mrgh##suffix, element, access, 0) 970 VMRG(b, u8, VsrB) 971 VMRG(h, u16, VsrH) 972 VMRG(w, u32, VsrW) 973 #undef VMRG_DO 974 #undef VMRG 975 976 void helper_vmsummbm(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, 977 ppc_avr_t *b, ppc_avr_t *c) 978 { 979 int32_t prod[16]; 980 int i; 981 982 for (i = 0; i < ARRAY_SIZE(r->s8); i++) { 983 prod[i] = (int32_t)a->s8[i] * b->u8[i]; 984 } 985 986 VECTOR_FOR_INORDER_I(i, s32) { 987 r->s32[i] = c->s32[i] + prod[4 * i] + prod[4 * i + 1] + 988 prod[4 * i + 2] + prod[4 * i + 3]; 989 } 990 } 991 992 void helper_vmsumshm(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, 993 ppc_avr_t *b, ppc_avr_t *c) 994 { 995 int32_t prod[8]; 996 int i; 997 998 for (i = 0; i < ARRAY_SIZE(r->s16); i++) { 999 prod[i] = a->s16[i] * b->s16[i]; 1000 } 1001 1002 VECTOR_FOR_INORDER_I(i, s32) { 1003 r->s32[i] = c->s32[i] + prod[2 * i] + prod[2 * i + 1]; 1004 } 1005 } 1006 1007 void helper_vmsumshs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, 1008 ppc_avr_t *b, ppc_avr_t *c) 1009 { 1010 int32_t prod[8]; 1011 int i; 1012 int sat = 0; 1013 1014 for (i = 0; i < ARRAY_SIZE(r->s16); i++) { 1015 prod[i] = (int32_t)a->s16[i] * b->s16[i]; 1016 } 1017 1018 VECTOR_FOR_INORDER_I(i, s32) { 1019 int64_t t = (int64_t)c->s32[i] + prod[2 * i] + prod[2 * i + 1]; 1020 1021 r->u32[i] = cvtsdsw(t, &sat); 1022 } 1023 1024 if (sat) { 1025 set_vscr_sat(env); 1026 } 1027 } 1028 1029 void helper_vmsumubm(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, 1030 ppc_avr_t *b, ppc_avr_t *c) 1031 { 1032 uint16_t prod[16]; 1033 int i; 1034 1035 for (i = 0; i < ARRAY_SIZE(r->u8); i++) { 1036 prod[i] = a->u8[i] * b->u8[i]; 1037 } 1038 1039 VECTOR_FOR_INORDER_I(i, u32) { 1040 r->u32[i] = c->u32[i] + prod[4 * i] + prod[4 * i + 1] + 1041 prod[4 * i + 2] + prod[4 * i + 3]; 1042 } 1043 } 1044 1045 void helper_vmsumuhm(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, 1046 ppc_avr_t *b, ppc_avr_t *c) 1047 { 1048 uint32_t prod[8]; 1049 int i; 1050 1051 for (i = 0; i < ARRAY_SIZE(r->u16); i++) { 1052 prod[i] = a->u16[i] * b->u16[i]; 1053 } 1054 1055 VECTOR_FOR_INORDER_I(i, u32) { 1056 r->u32[i] = c->u32[i] + prod[2 * i] + prod[2 * i + 1]; 1057 } 1058 } 1059 1060 void helper_vmsumuhs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, 1061 ppc_avr_t *b, ppc_avr_t *c) 1062 { 1063 uint32_t prod[8]; 1064 int i; 1065 int sat = 0; 1066 1067 for (i = 0; i < ARRAY_SIZE(r->u16); i++) { 1068 prod[i] = a->u16[i] * b->u16[i]; 1069 } 1070 1071 VECTOR_FOR_INORDER_I(i, s32) { 1072 uint64_t t = (uint64_t)c->u32[i] + prod[2 * i] + prod[2 * i + 1]; 1073 1074 r->u32[i] = cvtuduw(t, &sat); 1075 } 1076 1077 if (sat) { 1078 set_vscr_sat(env); 1079 } 1080 } 1081 1082 #define VMUL_DO_EVN(name, mul_element, mul_access, prod_access, cast) \ 1083 void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ 1084 { \ 1085 int i; \ 1086 \ 1087 for (i = 0; i < ARRAY_SIZE(r->mul_element); i += 2) { \ 1088 r->prod_access(i >> 1) = (cast)a->mul_access(i) * \ 1089 (cast)b->mul_access(i); \ 1090 } \ 1091 } 1092 1093 #define VMUL_DO_ODD(name, mul_element, mul_access, prod_access, cast) \ 1094 void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ 1095 { \ 1096 int i; \ 1097 \ 1098 for (i = 0; i < ARRAY_SIZE(r->mul_element); i += 2) { \ 1099 r->prod_access(i >> 1) = (cast)a->mul_access(i + 1) * \ 1100 (cast)b->mul_access(i + 1); \ 1101 } \ 1102 } 1103 1104 #define VMUL(suffix, mul_element, mul_access, prod_access, cast) \ 1105 VMUL_DO_EVN(mule##suffix, mul_element, mul_access, prod_access, cast) \ 1106 VMUL_DO_ODD(mulo##suffix, mul_element, mul_access, prod_access, cast) 1107 VMUL(sb, s8, VsrSB, VsrSH, int16_t) 1108 VMUL(sh, s16, VsrSH, VsrSW, int32_t) 1109 VMUL(sw, s32, VsrSW, VsrSD, int64_t) 1110 VMUL(ub, u8, VsrB, VsrH, uint16_t) 1111 VMUL(uh, u16, VsrH, VsrW, uint32_t) 1112 VMUL(uw, u32, VsrW, VsrD, uint64_t) 1113 #undef VMUL_DO_EVN 1114 #undef VMUL_DO_ODD 1115 #undef VMUL 1116 1117 void helper_vperm(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, 1118 ppc_avr_t *c) 1119 { 1120 ppc_avr_t result; 1121 int i; 1122 1123 for (i = 0; i < ARRAY_SIZE(r->u8); i++) { 1124 int s = c->VsrB(i) & 0x1f; 1125 int index = s & 0xf; 1126 1127 if (s & 0x10) { 1128 result.VsrB(i) = b->VsrB(index); 1129 } else { 1130 result.VsrB(i) = a->VsrB(index); 1131 } 1132 } 1133 *r = result; 1134 } 1135 1136 void helper_vpermr(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, 1137 ppc_avr_t *c) 1138 { 1139 ppc_avr_t result; 1140 int i; 1141 1142 for (i = 0; i < ARRAY_SIZE(r->u8); i++) { 1143 int s = c->VsrB(i) & 0x1f; 1144 int index = 15 - (s & 0xf); 1145 1146 if (s & 0x10) { 1147 result.VsrB(i) = a->VsrB(index); 1148 } else { 1149 result.VsrB(i) = b->VsrB(index); 1150 } 1151 } 1152 *r = result; 1153 } 1154 1155 #if defined(HOST_WORDS_BIGENDIAN) 1156 #define VBPERMQ_INDEX(avr, i) ((avr)->u8[(i)]) 1157 #define VBPERMD_INDEX(i) (i) 1158 #define VBPERMQ_DW(index) (((index) & 0x40) != 0) 1159 #define EXTRACT_BIT(avr, i, index) (extract64((avr)->u64[i], index, 1)) 1160 #else 1161 #define VBPERMQ_INDEX(avr, i) ((avr)->u8[15 - (i)]) 1162 #define VBPERMD_INDEX(i) (1 - i) 1163 #define VBPERMQ_DW(index) (((index) & 0x40) == 0) 1164 #define EXTRACT_BIT(avr, i, index) \ 1165 (extract64((avr)->u64[1 - i], 63 - index, 1)) 1166 #endif 1167 1168 void helper_vbpermd(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 1169 { 1170 int i, j; 1171 ppc_avr_t result = { .u64 = { 0, 0 } }; 1172 VECTOR_FOR_INORDER_I(i, u64) { 1173 for (j = 0; j < 8; j++) { 1174 int index = VBPERMQ_INDEX(b, (i * 8) + j); 1175 if (index < 64 && EXTRACT_BIT(a, i, index)) { 1176 result.u64[VBPERMD_INDEX(i)] |= (0x80 >> j); 1177 } 1178 } 1179 } 1180 *r = result; 1181 } 1182 1183 void helper_vbpermq(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 1184 { 1185 int i; 1186 uint64_t perm = 0; 1187 1188 VECTOR_FOR_INORDER_I(i, u8) { 1189 int index = VBPERMQ_INDEX(b, i); 1190 1191 if (index < 128) { 1192 uint64_t mask = (1ull << (63 - (index & 0x3F))); 1193 if (a->u64[VBPERMQ_DW(index)] & mask) { 1194 perm |= (0x8000 >> i); 1195 } 1196 } 1197 } 1198 1199 r->VsrD(0) = perm; 1200 r->VsrD(1) = 0; 1201 } 1202 1203 #undef VBPERMQ_INDEX 1204 #undef VBPERMQ_DW 1205 1206 static const uint64_t VGBBD_MASKS[256] = { 1207 0x0000000000000000ull, /* 00 */ 1208 0x0000000000000080ull, /* 01 */ 1209 0x0000000000008000ull, /* 02 */ 1210 0x0000000000008080ull, /* 03 */ 1211 0x0000000000800000ull, /* 04 */ 1212 0x0000000000800080ull, /* 05 */ 1213 0x0000000000808000ull, /* 06 */ 1214 0x0000000000808080ull, /* 07 */ 1215 0x0000000080000000ull, /* 08 */ 1216 0x0000000080000080ull, /* 09 */ 1217 0x0000000080008000ull, /* 0A */ 1218 0x0000000080008080ull, /* 0B */ 1219 0x0000000080800000ull, /* 0C */ 1220 0x0000000080800080ull, /* 0D */ 1221 0x0000000080808000ull, /* 0E */ 1222 0x0000000080808080ull, /* 0F */ 1223 0x0000008000000000ull, /* 10 */ 1224 0x0000008000000080ull, /* 11 */ 1225 0x0000008000008000ull, /* 12 */ 1226 0x0000008000008080ull, /* 13 */ 1227 0x0000008000800000ull, /* 14 */ 1228 0x0000008000800080ull, /* 15 */ 1229 0x0000008000808000ull, /* 16 */ 1230 0x0000008000808080ull, /* 17 */ 1231 0x0000008080000000ull, /* 18 */ 1232 0x0000008080000080ull, /* 19 */ 1233 0x0000008080008000ull, /* 1A */ 1234 0x0000008080008080ull, /* 1B */ 1235 0x0000008080800000ull, /* 1C */ 1236 0x0000008080800080ull, /* 1D */ 1237 0x0000008080808000ull, /* 1E */ 1238 0x0000008080808080ull, /* 1F */ 1239 0x0000800000000000ull, /* 20 */ 1240 0x0000800000000080ull, /* 21 */ 1241 0x0000800000008000ull, /* 22 */ 1242 0x0000800000008080ull, /* 23 */ 1243 0x0000800000800000ull, /* 24 */ 1244 0x0000800000800080ull, /* 25 */ 1245 0x0000800000808000ull, /* 26 */ 1246 0x0000800000808080ull, /* 27 */ 1247 0x0000800080000000ull, /* 28 */ 1248 0x0000800080000080ull, /* 29 */ 1249 0x0000800080008000ull, /* 2A */ 1250 0x0000800080008080ull, /* 2B */ 1251 0x0000800080800000ull, /* 2C */ 1252 0x0000800080800080ull, /* 2D */ 1253 0x0000800080808000ull, /* 2E */ 1254 0x0000800080808080ull, /* 2F */ 1255 0x0000808000000000ull, /* 30 */ 1256 0x0000808000000080ull, /* 31 */ 1257 0x0000808000008000ull, /* 32 */ 1258 0x0000808000008080ull, /* 33 */ 1259 0x0000808000800000ull, /* 34 */ 1260 0x0000808000800080ull, /* 35 */ 1261 0x0000808000808000ull, /* 36 */ 1262 0x0000808000808080ull, /* 37 */ 1263 0x0000808080000000ull, /* 38 */ 1264 0x0000808080000080ull, /* 39 */ 1265 0x0000808080008000ull, /* 3A */ 1266 0x0000808080008080ull, /* 3B */ 1267 0x0000808080800000ull, /* 3C */ 1268 0x0000808080800080ull, /* 3D */ 1269 0x0000808080808000ull, /* 3E */ 1270 0x0000808080808080ull, /* 3F */ 1271 0x0080000000000000ull, /* 40 */ 1272 0x0080000000000080ull, /* 41 */ 1273 0x0080000000008000ull, /* 42 */ 1274 0x0080000000008080ull, /* 43 */ 1275 0x0080000000800000ull, /* 44 */ 1276 0x0080000000800080ull, /* 45 */ 1277 0x0080000000808000ull, /* 46 */ 1278 0x0080000000808080ull, /* 47 */ 1279 0x0080000080000000ull, /* 48 */ 1280 0x0080000080000080ull, /* 49 */ 1281 0x0080000080008000ull, /* 4A */ 1282 0x0080000080008080ull, /* 4B */ 1283 0x0080000080800000ull, /* 4C */ 1284 0x0080000080800080ull, /* 4D */ 1285 0x0080000080808000ull, /* 4E */ 1286 0x0080000080808080ull, /* 4F */ 1287 0x0080008000000000ull, /* 50 */ 1288 0x0080008000000080ull, /* 51 */ 1289 0x0080008000008000ull, /* 52 */ 1290 0x0080008000008080ull, /* 53 */ 1291 0x0080008000800000ull, /* 54 */ 1292 0x0080008000800080ull, /* 55 */ 1293 0x0080008000808000ull, /* 56 */ 1294 0x0080008000808080ull, /* 57 */ 1295 0x0080008080000000ull, /* 58 */ 1296 0x0080008080000080ull, /* 59 */ 1297 0x0080008080008000ull, /* 5A */ 1298 0x0080008080008080ull, /* 5B */ 1299 0x0080008080800000ull, /* 5C */ 1300 0x0080008080800080ull, /* 5D */ 1301 0x0080008080808000ull, /* 5E */ 1302 0x0080008080808080ull, /* 5F */ 1303 0x0080800000000000ull, /* 60 */ 1304 0x0080800000000080ull, /* 61 */ 1305 0x0080800000008000ull, /* 62 */ 1306 0x0080800000008080ull, /* 63 */ 1307 0x0080800000800000ull, /* 64 */ 1308 0x0080800000800080ull, /* 65 */ 1309 0x0080800000808000ull, /* 66 */ 1310 0x0080800000808080ull, /* 67 */ 1311 0x0080800080000000ull, /* 68 */ 1312 0x0080800080000080ull, /* 69 */ 1313 0x0080800080008000ull, /* 6A */ 1314 0x0080800080008080ull, /* 6B */ 1315 0x0080800080800000ull, /* 6C */ 1316 0x0080800080800080ull, /* 6D */ 1317 0x0080800080808000ull, /* 6E */ 1318 0x0080800080808080ull, /* 6F */ 1319 0x0080808000000000ull, /* 70 */ 1320 0x0080808000000080ull, /* 71 */ 1321 0x0080808000008000ull, /* 72 */ 1322 0x0080808000008080ull, /* 73 */ 1323 0x0080808000800000ull, /* 74 */ 1324 0x0080808000800080ull, /* 75 */ 1325 0x0080808000808000ull, /* 76 */ 1326 0x0080808000808080ull, /* 77 */ 1327 0x0080808080000000ull, /* 78 */ 1328 0x0080808080000080ull, /* 79 */ 1329 0x0080808080008000ull, /* 7A */ 1330 0x0080808080008080ull, /* 7B */ 1331 0x0080808080800000ull, /* 7C */ 1332 0x0080808080800080ull, /* 7D */ 1333 0x0080808080808000ull, /* 7E */ 1334 0x0080808080808080ull, /* 7F */ 1335 0x8000000000000000ull, /* 80 */ 1336 0x8000000000000080ull, /* 81 */ 1337 0x8000000000008000ull, /* 82 */ 1338 0x8000000000008080ull, /* 83 */ 1339 0x8000000000800000ull, /* 84 */ 1340 0x8000000000800080ull, /* 85 */ 1341 0x8000000000808000ull, /* 86 */ 1342 0x8000000000808080ull, /* 87 */ 1343 0x8000000080000000ull, /* 88 */ 1344 0x8000000080000080ull, /* 89 */ 1345 0x8000000080008000ull, /* 8A */ 1346 0x8000000080008080ull, /* 8B */ 1347 0x8000000080800000ull, /* 8C */ 1348 0x8000000080800080ull, /* 8D */ 1349 0x8000000080808000ull, /* 8E */ 1350 0x8000000080808080ull, /* 8F */ 1351 0x8000008000000000ull, /* 90 */ 1352 0x8000008000000080ull, /* 91 */ 1353 0x8000008000008000ull, /* 92 */ 1354 0x8000008000008080ull, /* 93 */ 1355 0x8000008000800000ull, /* 94 */ 1356 0x8000008000800080ull, /* 95 */ 1357 0x8000008000808000ull, /* 96 */ 1358 0x8000008000808080ull, /* 97 */ 1359 0x8000008080000000ull, /* 98 */ 1360 0x8000008080000080ull, /* 99 */ 1361 0x8000008080008000ull, /* 9A */ 1362 0x8000008080008080ull, /* 9B */ 1363 0x8000008080800000ull, /* 9C */ 1364 0x8000008080800080ull, /* 9D */ 1365 0x8000008080808000ull, /* 9E */ 1366 0x8000008080808080ull, /* 9F */ 1367 0x8000800000000000ull, /* A0 */ 1368 0x8000800000000080ull, /* A1 */ 1369 0x8000800000008000ull, /* A2 */ 1370 0x8000800000008080ull, /* A3 */ 1371 0x8000800000800000ull, /* A4 */ 1372 0x8000800000800080ull, /* A5 */ 1373 0x8000800000808000ull, /* A6 */ 1374 0x8000800000808080ull, /* A7 */ 1375 0x8000800080000000ull, /* A8 */ 1376 0x8000800080000080ull, /* A9 */ 1377 0x8000800080008000ull, /* AA */ 1378 0x8000800080008080ull, /* AB */ 1379 0x8000800080800000ull, /* AC */ 1380 0x8000800080800080ull, /* AD */ 1381 0x8000800080808000ull, /* AE */ 1382 0x8000800080808080ull, /* AF */ 1383 0x8000808000000000ull, /* B0 */ 1384 0x8000808000000080ull, /* B1 */ 1385 0x8000808000008000ull, /* B2 */ 1386 0x8000808000008080ull, /* B3 */ 1387 0x8000808000800000ull, /* B4 */ 1388 0x8000808000800080ull, /* B5 */ 1389 0x8000808000808000ull, /* B6 */ 1390 0x8000808000808080ull, /* B7 */ 1391 0x8000808080000000ull, /* B8 */ 1392 0x8000808080000080ull, /* B9 */ 1393 0x8000808080008000ull, /* BA */ 1394 0x8000808080008080ull, /* BB */ 1395 0x8000808080800000ull, /* BC */ 1396 0x8000808080800080ull, /* BD */ 1397 0x8000808080808000ull, /* BE */ 1398 0x8000808080808080ull, /* BF */ 1399 0x8080000000000000ull, /* C0 */ 1400 0x8080000000000080ull, /* C1 */ 1401 0x8080000000008000ull, /* C2 */ 1402 0x8080000000008080ull, /* C3 */ 1403 0x8080000000800000ull, /* C4 */ 1404 0x8080000000800080ull, /* C5 */ 1405 0x8080000000808000ull, /* C6 */ 1406 0x8080000000808080ull, /* C7 */ 1407 0x8080000080000000ull, /* C8 */ 1408 0x8080000080000080ull, /* C9 */ 1409 0x8080000080008000ull, /* CA */ 1410 0x8080000080008080ull, /* CB */ 1411 0x8080000080800000ull, /* CC */ 1412 0x8080000080800080ull, /* CD */ 1413 0x8080000080808000ull, /* CE */ 1414 0x8080000080808080ull, /* CF */ 1415 0x8080008000000000ull, /* D0 */ 1416 0x8080008000000080ull, /* D1 */ 1417 0x8080008000008000ull, /* D2 */ 1418 0x8080008000008080ull, /* D3 */ 1419 0x8080008000800000ull, /* D4 */ 1420 0x8080008000800080ull, /* D5 */ 1421 0x8080008000808000ull, /* D6 */ 1422 0x8080008000808080ull, /* D7 */ 1423 0x8080008080000000ull, /* D8 */ 1424 0x8080008080000080ull, /* D9 */ 1425 0x8080008080008000ull, /* DA */ 1426 0x8080008080008080ull, /* DB */ 1427 0x8080008080800000ull, /* DC */ 1428 0x8080008080800080ull, /* DD */ 1429 0x8080008080808000ull, /* DE */ 1430 0x8080008080808080ull, /* DF */ 1431 0x8080800000000000ull, /* E0 */ 1432 0x8080800000000080ull, /* E1 */ 1433 0x8080800000008000ull, /* E2 */ 1434 0x8080800000008080ull, /* E3 */ 1435 0x8080800000800000ull, /* E4 */ 1436 0x8080800000800080ull, /* E5 */ 1437 0x8080800000808000ull, /* E6 */ 1438 0x8080800000808080ull, /* E7 */ 1439 0x8080800080000000ull, /* E8 */ 1440 0x8080800080000080ull, /* E9 */ 1441 0x8080800080008000ull, /* EA */ 1442 0x8080800080008080ull, /* EB */ 1443 0x8080800080800000ull, /* EC */ 1444 0x8080800080800080ull, /* ED */ 1445 0x8080800080808000ull, /* EE */ 1446 0x8080800080808080ull, /* EF */ 1447 0x8080808000000000ull, /* F0 */ 1448 0x8080808000000080ull, /* F1 */ 1449 0x8080808000008000ull, /* F2 */ 1450 0x8080808000008080ull, /* F3 */ 1451 0x8080808000800000ull, /* F4 */ 1452 0x8080808000800080ull, /* F5 */ 1453 0x8080808000808000ull, /* F6 */ 1454 0x8080808000808080ull, /* F7 */ 1455 0x8080808080000000ull, /* F8 */ 1456 0x8080808080000080ull, /* F9 */ 1457 0x8080808080008000ull, /* FA */ 1458 0x8080808080008080ull, /* FB */ 1459 0x8080808080800000ull, /* FC */ 1460 0x8080808080800080ull, /* FD */ 1461 0x8080808080808000ull, /* FE */ 1462 0x8080808080808080ull, /* FF */ 1463 }; 1464 1465 void helper_vgbbd(ppc_avr_t *r, ppc_avr_t *b) 1466 { 1467 int i; 1468 uint64_t t[2] = { 0, 0 }; 1469 1470 VECTOR_FOR_INORDER_I(i, u8) { 1471 #if defined(HOST_WORDS_BIGENDIAN) 1472 t[i >> 3] |= VGBBD_MASKS[b->u8[i]] >> (i & 7); 1473 #else 1474 t[i >> 3] |= VGBBD_MASKS[b->u8[i]] >> (7 - (i & 7)); 1475 #endif 1476 } 1477 1478 r->u64[0] = t[0]; 1479 r->u64[1] = t[1]; 1480 } 1481 1482 #define PMSUM(name, srcfld, trgfld, trgtyp) \ 1483 void helper_##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ 1484 { \ 1485 int i, j; \ 1486 trgtyp prod[sizeof(ppc_avr_t) / sizeof(a->srcfld[0])]; \ 1487 \ 1488 VECTOR_FOR_INORDER_I(i, srcfld) { \ 1489 prod[i] = 0; \ 1490 for (j = 0; j < sizeof(a->srcfld[0]) * 8; j++) { \ 1491 if (a->srcfld[i] & (1ull << j)) { \ 1492 prod[i] ^= ((trgtyp)b->srcfld[i] << j); \ 1493 } \ 1494 } \ 1495 } \ 1496 \ 1497 VECTOR_FOR_INORDER_I(i, trgfld) { \ 1498 r->trgfld[i] = prod[2 * i] ^ prod[2 * i + 1]; \ 1499 } \ 1500 } 1501 1502 PMSUM(vpmsumb, u8, u16, uint16_t) 1503 PMSUM(vpmsumh, u16, u32, uint32_t) 1504 PMSUM(vpmsumw, u32, u64, uint64_t) 1505 1506 void helper_vpmsumd(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 1507 { 1508 1509 #ifdef CONFIG_INT128 1510 int i, j; 1511 __uint128_t prod[2]; 1512 1513 VECTOR_FOR_INORDER_I(i, u64) { 1514 prod[i] = 0; 1515 for (j = 0; j < 64; j++) { 1516 if (a->u64[i] & (1ull << j)) { 1517 prod[i] ^= (((__uint128_t)b->u64[i]) << j); 1518 } 1519 } 1520 } 1521 1522 r->u128 = prod[0] ^ prod[1]; 1523 1524 #else 1525 int i, j; 1526 ppc_avr_t prod[2]; 1527 1528 VECTOR_FOR_INORDER_I(i, u64) { 1529 prod[i].VsrD(1) = prod[i].VsrD(0) = 0; 1530 for (j = 0; j < 64; j++) { 1531 if (a->u64[i] & (1ull << j)) { 1532 ppc_avr_t bshift; 1533 if (j == 0) { 1534 bshift.VsrD(0) = 0; 1535 bshift.VsrD(1) = b->u64[i]; 1536 } else { 1537 bshift.VsrD(0) = b->u64[i] >> (64 - j); 1538 bshift.VsrD(1) = b->u64[i] << j; 1539 } 1540 prod[i].VsrD(1) ^= bshift.VsrD(1); 1541 prod[i].VsrD(0) ^= bshift.VsrD(0); 1542 } 1543 } 1544 } 1545 1546 r->VsrD(1) = prod[0].VsrD(1) ^ prod[1].VsrD(1); 1547 r->VsrD(0) = prod[0].VsrD(0) ^ prod[1].VsrD(0); 1548 #endif 1549 } 1550 1551 1552 #if defined(HOST_WORDS_BIGENDIAN) 1553 #define PKBIG 1 1554 #else 1555 #define PKBIG 0 1556 #endif 1557 void helper_vpkpx(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 1558 { 1559 int i, j; 1560 ppc_avr_t result; 1561 #if defined(HOST_WORDS_BIGENDIAN) 1562 const ppc_avr_t *x[2] = { a, b }; 1563 #else 1564 const ppc_avr_t *x[2] = { b, a }; 1565 #endif 1566 1567 VECTOR_FOR_INORDER_I(i, u64) { 1568 VECTOR_FOR_INORDER_I(j, u32) { 1569 uint32_t e = x[i]->u32[j]; 1570 1571 result.u16[4 * i + j] = (((e >> 9) & 0xfc00) | 1572 ((e >> 6) & 0x3e0) | 1573 ((e >> 3) & 0x1f)); 1574 } 1575 } 1576 *r = result; 1577 } 1578 1579 #define VPK(suffix, from, to, cvt, dosat) \ 1580 void helper_vpk##suffix(CPUPPCState *env, ppc_avr_t *r, \ 1581 ppc_avr_t *a, ppc_avr_t *b) \ 1582 { \ 1583 int i; \ 1584 int sat = 0; \ 1585 ppc_avr_t result; \ 1586 ppc_avr_t *a0 = PKBIG ? a : b; \ 1587 ppc_avr_t *a1 = PKBIG ? b : a; \ 1588 \ 1589 VECTOR_FOR_INORDER_I(i, from) { \ 1590 result.to[i] = cvt(a0->from[i], &sat); \ 1591 result.to[i + ARRAY_SIZE(r->from)] = cvt(a1->from[i], &sat);\ 1592 } \ 1593 *r = result; \ 1594 if (dosat && sat) { \ 1595 set_vscr_sat(env); \ 1596 } \ 1597 } 1598 #define I(x, y) (x) 1599 VPK(shss, s16, s8, cvtshsb, 1) 1600 VPK(shus, s16, u8, cvtshub, 1) 1601 VPK(swss, s32, s16, cvtswsh, 1) 1602 VPK(swus, s32, u16, cvtswuh, 1) 1603 VPK(sdss, s64, s32, cvtsdsw, 1) 1604 VPK(sdus, s64, u32, cvtsduw, 1) 1605 VPK(uhus, u16, u8, cvtuhub, 1) 1606 VPK(uwus, u32, u16, cvtuwuh, 1) 1607 VPK(udus, u64, u32, cvtuduw, 1) 1608 VPK(uhum, u16, u8, I, 0) 1609 VPK(uwum, u32, u16, I, 0) 1610 VPK(udum, u64, u32, I, 0) 1611 #undef I 1612 #undef VPK 1613 #undef PKBIG 1614 1615 void helper_vrefp(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *b) 1616 { 1617 int i; 1618 1619 for (i = 0; i < ARRAY_SIZE(r->f32); i++) { 1620 r->f32[i] = float32_div(float32_one, b->f32[i], &env->vec_status); 1621 } 1622 } 1623 1624 #define VRFI(suffix, rounding) \ 1625 void helper_vrfi##suffix(CPUPPCState *env, ppc_avr_t *r, \ 1626 ppc_avr_t *b) \ 1627 { \ 1628 int i; \ 1629 float_status s = env->vec_status; \ 1630 \ 1631 set_float_rounding_mode(rounding, &s); \ 1632 for (i = 0; i < ARRAY_SIZE(r->f32); i++) { \ 1633 r->f32[i] = float32_round_to_int (b->f32[i], &s); \ 1634 } \ 1635 } 1636 VRFI(n, float_round_nearest_even) 1637 VRFI(m, float_round_down) 1638 VRFI(p, float_round_up) 1639 VRFI(z, float_round_to_zero) 1640 #undef VRFI 1641 1642 #define VROTATE(suffix, element, mask) \ 1643 void helper_vrl##suffix(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ 1644 { \ 1645 int i; \ 1646 \ 1647 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ 1648 unsigned int shift = b->element[i] & mask; \ 1649 r->element[i] = (a->element[i] << shift) | \ 1650 (a->element[i] >> (sizeof(a->element[0]) * 8 - shift)); \ 1651 } \ 1652 } 1653 VROTATE(b, u8, 0x7) 1654 VROTATE(h, u16, 0xF) 1655 VROTATE(w, u32, 0x1F) 1656 VROTATE(d, u64, 0x3F) 1657 #undef VROTATE 1658 1659 void helper_vrsqrtefp(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *b) 1660 { 1661 int i; 1662 1663 for (i = 0; i < ARRAY_SIZE(r->f32); i++) { 1664 float32 t = float32_sqrt(b->f32[i], &env->vec_status); 1665 1666 r->f32[i] = float32_div(float32_one, t, &env->vec_status); 1667 } 1668 } 1669 1670 #define VRLMI(name, size, element, insert) \ 1671 void helper_##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ 1672 { \ 1673 int i; \ 1674 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ 1675 uint##size##_t src1 = a->element[i]; \ 1676 uint##size##_t src2 = b->element[i]; \ 1677 uint##size##_t src3 = r->element[i]; \ 1678 uint##size##_t begin, end, shift, mask, rot_val; \ 1679 \ 1680 shift = extract##size(src2, 0, 6); \ 1681 end = extract##size(src2, 8, 6); \ 1682 begin = extract##size(src2, 16, 6); \ 1683 rot_val = rol##size(src1, shift); \ 1684 mask = mask_u##size(begin, end); \ 1685 if (insert) { \ 1686 r->element[i] = (rot_val & mask) | (src3 & ~mask); \ 1687 } else { \ 1688 r->element[i] = (rot_val & mask); \ 1689 } \ 1690 } \ 1691 } 1692 1693 VRLMI(vrldmi, 64, u64, 1); 1694 VRLMI(vrlwmi, 32, u32, 1); 1695 VRLMI(vrldnm, 64, u64, 0); 1696 VRLMI(vrlwnm, 32, u32, 0); 1697 1698 void helper_vsel(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, 1699 ppc_avr_t *c) 1700 { 1701 r->u64[0] = (a->u64[0] & ~c->u64[0]) | (b->u64[0] & c->u64[0]); 1702 r->u64[1] = (a->u64[1] & ~c->u64[1]) | (b->u64[1] & c->u64[1]); 1703 } 1704 1705 void helper_vexptefp(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *b) 1706 { 1707 int i; 1708 1709 for (i = 0; i < ARRAY_SIZE(r->f32); i++) { 1710 r->f32[i] = float32_exp2(b->f32[i], &env->vec_status); 1711 } 1712 } 1713 1714 void helper_vlogefp(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *b) 1715 { 1716 int i; 1717 1718 for (i = 0; i < ARRAY_SIZE(r->f32); i++) { 1719 r->f32[i] = float32_log2(b->f32[i], &env->vec_status); 1720 } 1721 } 1722 1723 #if defined(HOST_WORDS_BIGENDIAN) 1724 #define VEXTU_X_DO(name, size, left) \ 1725 target_ulong glue(helper_, name)(target_ulong a, ppc_avr_t *b) \ 1726 { \ 1727 int index; \ 1728 if (left) { \ 1729 index = (a & 0xf) * 8; \ 1730 } else { \ 1731 index = ((15 - (a & 0xf) + 1) * 8) - size; \ 1732 } \ 1733 return int128_getlo(int128_rshift(b->s128, index)) & \ 1734 MAKE_64BIT_MASK(0, size); \ 1735 } 1736 #else 1737 #define VEXTU_X_DO(name, size, left) \ 1738 target_ulong glue(helper_, name)(target_ulong a, ppc_avr_t *b) \ 1739 { \ 1740 int index; \ 1741 if (left) { \ 1742 index = ((15 - (a & 0xf) + 1) * 8) - size; \ 1743 } else { \ 1744 index = (a & 0xf) * 8; \ 1745 } \ 1746 return int128_getlo(int128_rshift(b->s128, index)) & \ 1747 MAKE_64BIT_MASK(0, size); \ 1748 } 1749 #endif 1750 1751 VEXTU_X_DO(vextublx, 8, 1) 1752 VEXTU_X_DO(vextuhlx, 16, 1) 1753 VEXTU_X_DO(vextuwlx, 32, 1) 1754 VEXTU_X_DO(vextubrx, 8, 0) 1755 VEXTU_X_DO(vextuhrx, 16, 0) 1756 VEXTU_X_DO(vextuwrx, 32, 0) 1757 #undef VEXTU_X_DO 1758 1759 /* 1760 * The specification says that the results are undefined if all of the 1761 * shift counts are not identical. We check to make sure that they 1762 * are to conform to what real hardware appears to do. 1763 */ 1764 #define VSHIFT(suffix, leftp) \ 1765 void helper_vs##suffix(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ 1766 { \ 1767 int shift = b->VsrB(15) & 0x7; \ 1768 int doit = 1; \ 1769 int i; \ 1770 \ 1771 for (i = 0; i < ARRAY_SIZE(r->u8); i++) { \ 1772 doit = doit && ((b->u8[i] & 0x7) == shift); \ 1773 } \ 1774 if (doit) { \ 1775 if (shift == 0) { \ 1776 *r = *a; \ 1777 } else if (leftp) { \ 1778 uint64_t carry = a->VsrD(1) >> (64 - shift); \ 1779 \ 1780 r->VsrD(0) = (a->VsrD(0) << shift) | carry; \ 1781 r->VsrD(1) = a->VsrD(1) << shift; \ 1782 } else { \ 1783 uint64_t carry = a->VsrD(0) << (64 - shift); \ 1784 \ 1785 r->VsrD(1) = (a->VsrD(1) >> shift) | carry; \ 1786 r->VsrD(0) = a->VsrD(0) >> shift; \ 1787 } \ 1788 } \ 1789 } 1790 VSHIFT(l, 1) 1791 VSHIFT(r, 0) 1792 #undef VSHIFT 1793 1794 void helper_vslv(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 1795 { 1796 int i; 1797 unsigned int shift, bytes, size; 1798 1799 size = ARRAY_SIZE(r->u8); 1800 for (i = 0; i < size; i++) { 1801 shift = b->VsrB(i) & 0x7; /* extract shift value */ 1802 bytes = (a->VsrB(i) << 8) + /* extract adjacent bytes */ 1803 (((i + 1) < size) ? a->VsrB(i + 1) : 0); 1804 r->VsrB(i) = (bytes << shift) >> 8; /* shift and store result */ 1805 } 1806 } 1807 1808 void helper_vsrv(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 1809 { 1810 int i; 1811 unsigned int shift, bytes; 1812 1813 /* 1814 * Use reverse order, as destination and source register can be 1815 * same. Its being modified in place saving temporary, reverse 1816 * order will guarantee that computed result is not fed back. 1817 */ 1818 for (i = ARRAY_SIZE(r->u8) - 1; i >= 0; i--) { 1819 shift = b->VsrB(i) & 0x7; /* extract shift value */ 1820 bytes = ((i ? a->VsrB(i - 1) : 0) << 8) + a->VsrB(i); 1821 /* extract adjacent bytes */ 1822 r->VsrB(i) = (bytes >> shift) & 0xFF; /* shift and store result */ 1823 } 1824 } 1825 1826 void helper_vsldoi(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t shift) 1827 { 1828 int sh = shift & 0xf; 1829 int i; 1830 ppc_avr_t result; 1831 1832 for (i = 0; i < ARRAY_SIZE(r->u8); i++) { 1833 int index = sh + i; 1834 if (index > 0xf) { 1835 result.VsrB(i) = b->VsrB(index - 0x10); 1836 } else { 1837 result.VsrB(i) = a->VsrB(index); 1838 } 1839 } 1840 *r = result; 1841 } 1842 1843 void helper_vslo(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 1844 { 1845 int sh = (b->VsrB(0xf) >> 3) & 0xf; 1846 1847 #if defined(HOST_WORDS_BIGENDIAN) 1848 memmove(&r->u8[0], &a->u8[sh], 16 - sh); 1849 memset(&r->u8[16 - sh], 0, sh); 1850 #else 1851 memmove(&r->u8[sh], &a->u8[0], 16 - sh); 1852 memset(&r->u8[0], 0, sh); 1853 #endif 1854 } 1855 1856 #if defined(HOST_WORDS_BIGENDIAN) 1857 #define VINSERT(suffix, element) \ 1858 void helper_vinsert##suffix(ppc_avr_t *r, ppc_avr_t *b, uint32_t index) \ 1859 { \ 1860 memmove(&r->u8[index], &b->u8[8 - sizeof(r->element[0])], \ 1861 sizeof(r->element[0])); \ 1862 } 1863 #else 1864 #define VINSERT(suffix, element) \ 1865 void helper_vinsert##suffix(ppc_avr_t *r, ppc_avr_t *b, uint32_t index) \ 1866 { \ 1867 uint32_t d = (16 - index) - sizeof(r->element[0]); \ 1868 memmove(&r->u8[d], &b->u8[8], sizeof(r->element[0])); \ 1869 } 1870 #endif 1871 VINSERT(b, u8) 1872 VINSERT(h, u16) 1873 VINSERT(w, u32) 1874 VINSERT(d, u64) 1875 #undef VINSERT 1876 #if defined(HOST_WORDS_BIGENDIAN) 1877 #define VEXTRACT(suffix, element) \ 1878 void helper_vextract##suffix(ppc_avr_t *r, ppc_avr_t *b, uint32_t index) \ 1879 { \ 1880 uint32_t es = sizeof(r->element[0]); \ 1881 memmove(&r->u8[8 - es], &b->u8[index], es); \ 1882 memset(&r->u8[8], 0, 8); \ 1883 memset(&r->u8[0], 0, 8 - es); \ 1884 } 1885 #else 1886 #define VEXTRACT(suffix, element) \ 1887 void helper_vextract##suffix(ppc_avr_t *r, ppc_avr_t *b, uint32_t index) \ 1888 { \ 1889 uint32_t es = sizeof(r->element[0]); \ 1890 uint32_t s = (16 - index) - es; \ 1891 memmove(&r->u8[8], &b->u8[s], es); \ 1892 memset(&r->u8[0], 0, 8); \ 1893 memset(&r->u8[8 + es], 0, 8 - es); \ 1894 } 1895 #endif 1896 VEXTRACT(ub, u8) 1897 VEXTRACT(uh, u16) 1898 VEXTRACT(uw, u32) 1899 VEXTRACT(d, u64) 1900 #undef VEXTRACT 1901 1902 void helper_xxextractuw(CPUPPCState *env, ppc_vsr_t *xt, 1903 ppc_vsr_t *xb, uint32_t index) 1904 { 1905 ppc_vsr_t t = { }; 1906 size_t es = sizeof(uint32_t); 1907 uint32_t ext_index; 1908 int i; 1909 1910 ext_index = index; 1911 for (i = 0; i < es; i++, ext_index++) { 1912 t.VsrB(8 - es + i) = xb->VsrB(ext_index % 16); 1913 } 1914 1915 *xt = t; 1916 } 1917 1918 void helper_xxinsertw(CPUPPCState *env, ppc_vsr_t *xt, 1919 ppc_vsr_t *xb, uint32_t index) 1920 { 1921 ppc_vsr_t t = *xt; 1922 size_t es = sizeof(uint32_t); 1923 int ins_index, i = 0; 1924 1925 ins_index = index; 1926 for (i = 0; i < es && ins_index < 16; i++, ins_index++) { 1927 t.VsrB(ins_index) = xb->VsrB(8 - es + i); 1928 } 1929 1930 *xt = t; 1931 } 1932 1933 #define VEXT_SIGNED(name, element, cast) \ 1934 void helper_##name(ppc_avr_t *r, ppc_avr_t *b) \ 1935 { \ 1936 int i; \ 1937 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ 1938 r->element[i] = (cast)b->element[i]; \ 1939 } \ 1940 } 1941 VEXT_SIGNED(vextsb2w, s32, int8_t) 1942 VEXT_SIGNED(vextsb2d, s64, int8_t) 1943 VEXT_SIGNED(vextsh2w, s32, int16_t) 1944 VEXT_SIGNED(vextsh2d, s64, int16_t) 1945 VEXT_SIGNED(vextsw2d, s64, int32_t) 1946 #undef VEXT_SIGNED 1947 1948 #define VNEG(name, element) \ 1949 void helper_##name(ppc_avr_t *r, ppc_avr_t *b) \ 1950 { \ 1951 int i; \ 1952 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ 1953 r->element[i] = -b->element[i]; \ 1954 } \ 1955 } 1956 VNEG(vnegw, s32) 1957 VNEG(vnegd, s64) 1958 #undef VNEG 1959 1960 void helper_vsro(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 1961 { 1962 int sh = (b->VsrB(0xf) >> 3) & 0xf; 1963 1964 #if defined(HOST_WORDS_BIGENDIAN) 1965 memmove(&r->u8[sh], &a->u8[0], 16 - sh); 1966 memset(&r->u8[0], 0, sh); 1967 #else 1968 memmove(&r->u8[0], &a->u8[sh], 16 - sh); 1969 memset(&r->u8[16 - sh], 0, sh); 1970 #endif 1971 } 1972 1973 void helper_vsubcuw(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 1974 { 1975 int i; 1976 1977 for (i = 0; i < ARRAY_SIZE(r->u32); i++) { 1978 r->u32[i] = a->u32[i] >= b->u32[i]; 1979 } 1980 } 1981 1982 void helper_vsumsws(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 1983 { 1984 int64_t t; 1985 int i, upper; 1986 ppc_avr_t result; 1987 int sat = 0; 1988 1989 upper = ARRAY_SIZE(r->s32) - 1; 1990 t = (int64_t)b->VsrSW(upper); 1991 for (i = 0; i < ARRAY_SIZE(r->s32); i++) { 1992 t += a->VsrSW(i); 1993 result.VsrSW(i) = 0; 1994 } 1995 result.VsrSW(upper) = cvtsdsw(t, &sat); 1996 *r = result; 1997 1998 if (sat) { 1999 set_vscr_sat(env); 2000 } 2001 } 2002 2003 void helper_vsum2sws(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 2004 { 2005 int i, j, upper; 2006 ppc_avr_t result; 2007 int sat = 0; 2008 2009 upper = 1; 2010 for (i = 0; i < ARRAY_SIZE(r->u64); i++) { 2011 int64_t t = (int64_t)b->VsrSW(upper + i * 2); 2012 2013 result.VsrD(i) = 0; 2014 for (j = 0; j < ARRAY_SIZE(r->u64); j++) { 2015 t += a->VsrSW(2 * i + j); 2016 } 2017 result.VsrSW(upper + i * 2) = cvtsdsw(t, &sat); 2018 } 2019 2020 *r = result; 2021 if (sat) { 2022 set_vscr_sat(env); 2023 } 2024 } 2025 2026 void helper_vsum4sbs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 2027 { 2028 int i, j; 2029 int sat = 0; 2030 2031 for (i = 0; i < ARRAY_SIZE(r->s32); i++) { 2032 int64_t t = (int64_t)b->s32[i]; 2033 2034 for (j = 0; j < ARRAY_SIZE(r->s32); j++) { 2035 t += a->s8[4 * i + j]; 2036 } 2037 r->s32[i] = cvtsdsw(t, &sat); 2038 } 2039 2040 if (sat) { 2041 set_vscr_sat(env); 2042 } 2043 } 2044 2045 void helper_vsum4shs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 2046 { 2047 int sat = 0; 2048 int i; 2049 2050 for (i = 0; i < ARRAY_SIZE(r->s32); i++) { 2051 int64_t t = (int64_t)b->s32[i]; 2052 2053 t += a->s16[2 * i] + a->s16[2 * i + 1]; 2054 r->s32[i] = cvtsdsw(t, &sat); 2055 } 2056 2057 if (sat) { 2058 set_vscr_sat(env); 2059 } 2060 } 2061 2062 void helper_vsum4ubs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 2063 { 2064 int i, j; 2065 int sat = 0; 2066 2067 for (i = 0; i < ARRAY_SIZE(r->u32); i++) { 2068 uint64_t t = (uint64_t)b->u32[i]; 2069 2070 for (j = 0; j < ARRAY_SIZE(r->u32); j++) { 2071 t += a->u8[4 * i + j]; 2072 } 2073 r->u32[i] = cvtuduw(t, &sat); 2074 } 2075 2076 if (sat) { 2077 set_vscr_sat(env); 2078 } 2079 } 2080 2081 #if defined(HOST_WORDS_BIGENDIAN) 2082 #define UPKHI 1 2083 #define UPKLO 0 2084 #else 2085 #define UPKHI 0 2086 #define UPKLO 1 2087 #endif 2088 #define VUPKPX(suffix, hi) \ 2089 void helper_vupk##suffix(ppc_avr_t *r, ppc_avr_t *b) \ 2090 { \ 2091 int i; \ 2092 ppc_avr_t result; \ 2093 \ 2094 for (i = 0; i < ARRAY_SIZE(r->u32); i++) { \ 2095 uint16_t e = b->u16[hi ? i : i + 4]; \ 2096 uint8_t a = (e >> 15) ? 0xff : 0; \ 2097 uint8_t r = (e >> 10) & 0x1f; \ 2098 uint8_t g = (e >> 5) & 0x1f; \ 2099 uint8_t b = e & 0x1f; \ 2100 \ 2101 result.u32[i] = (a << 24) | (r << 16) | (g << 8) | b; \ 2102 } \ 2103 *r = result; \ 2104 } 2105 VUPKPX(lpx, UPKLO) 2106 VUPKPX(hpx, UPKHI) 2107 #undef VUPKPX 2108 2109 #define VUPK(suffix, unpacked, packee, hi) \ 2110 void helper_vupk##suffix(ppc_avr_t *r, ppc_avr_t *b) \ 2111 { \ 2112 int i; \ 2113 ppc_avr_t result; \ 2114 \ 2115 if (hi) { \ 2116 for (i = 0; i < ARRAY_SIZE(r->unpacked); i++) { \ 2117 result.unpacked[i] = b->packee[i]; \ 2118 } \ 2119 } else { \ 2120 for (i = ARRAY_SIZE(r->unpacked); i < ARRAY_SIZE(r->packee); \ 2121 i++) { \ 2122 result.unpacked[i - ARRAY_SIZE(r->unpacked)] = b->packee[i]; \ 2123 } \ 2124 } \ 2125 *r = result; \ 2126 } 2127 VUPK(hsb, s16, s8, UPKHI) 2128 VUPK(hsh, s32, s16, UPKHI) 2129 VUPK(hsw, s64, s32, UPKHI) 2130 VUPK(lsb, s16, s8, UPKLO) 2131 VUPK(lsh, s32, s16, UPKLO) 2132 VUPK(lsw, s64, s32, UPKLO) 2133 #undef VUPK 2134 #undef UPKHI 2135 #undef UPKLO 2136 2137 #define VGENERIC_DO(name, element) \ 2138 void helper_v##name(ppc_avr_t *r, ppc_avr_t *b) \ 2139 { \ 2140 int i; \ 2141 \ 2142 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ 2143 r->element[i] = name(b->element[i]); \ 2144 } \ 2145 } 2146 2147 #define clzb(v) ((v) ? clz32((uint32_t)(v) << 24) : 8) 2148 #define clzh(v) ((v) ? clz32((uint32_t)(v) << 16) : 16) 2149 #define clzw(v) clz32((v)) 2150 #define clzd(v) clz64((v)) 2151 2152 VGENERIC_DO(clzb, u8) 2153 VGENERIC_DO(clzh, u16) 2154 VGENERIC_DO(clzw, u32) 2155 VGENERIC_DO(clzd, u64) 2156 2157 #undef clzb 2158 #undef clzh 2159 #undef clzw 2160 #undef clzd 2161 2162 #define ctzb(v) ((v) ? ctz32(v) : 8) 2163 #define ctzh(v) ((v) ? ctz32(v) : 16) 2164 #define ctzw(v) ctz32((v)) 2165 #define ctzd(v) ctz64((v)) 2166 2167 VGENERIC_DO(ctzb, u8) 2168 VGENERIC_DO(ctzh, u16) 2169 VGENERIC_DO(ctzw, u32) 2170 VGENERIC_DO(ctzd, u64) 2171 2172 #undef ctzb 2173 #undef ctzh 2174 #undef ctzw 2175 #undef ctzd 2176 2177 #define popcntb(v) ctpop8(v) 2178 #define popcnth(v) ctpop16(v) 2179 #define popcntw(v) ctpop32(v) 2180 #define popcntd(v) ctpop64(v) 2181 2182 VGENERIC_DO(popcntb, u8) 2183 VGENERIC_DO(popcnth, u16) 2184 VGENERIC_DO(popcntw, u32) 2185 VGENERIC_DO(popcntd, u64) 2186 2187 #undef popcntb 2188 #undef popcnth 2189 #undef popcntw 2190 #undef popcntd 2191 2192 #undef VGENERIC_DO 2193 2194 #if defined(HOST_WORDS_BIGENDIAN) 2195 #define QW_ONE { .u64 = { 0, 1 } } 2196 #else 2197 #define QW_ONE { .u64 = { 1, 0 } } 2198 #endif 2199 2200 #ifndef CONFIG_INT128 2201 2202 static inline void avr_qw_not(ppc_avr_t *t, ppc_avr_t a) 2203 { 2204 t->u64[0] = ~a.u64[0]; 2205 t->u64[1] = ~a.u64[1]; 2206 } 2207 2208 static int avr_qw_cmpu(ppc_avr_t a, ppc_avr_t b) 2209 { 2210 if (a.VsrD(0) < b.VsrD(0)) { 2211 return -1; 2212 } else if (a.VsrD(0) > b.VsrD(0)) { 2213 return 1; 2214 } else if (a.VsrD(1) < b.VsrD(1)) { 2215 return -1; 2216 } else if (a.VsrD(1) > b.VsrD(1)) { 2217 return 1; 2218 } else { 2219 return 0; 2220 } 2221 } 2222 2223 static void avr_qw_add(ppc_avr_t *t, ppc_avr_t a, ppc_avr_t b) 2224 { 2225 t->VsrD(1) = a.VsrD(1) + b.VsrD(1); 2226 t->VsrD(0) = a.VsrD(0) + b.VsrD(0) + 2227 (~a.VsrD(1) < b.VsrD(1)); 2228 } 2229 2230 static int avr_qw_addc(ppc_avr_t *t, ppc_avr_t a, ppc_avr_t b) 2231 { 2232 ppc_avr_t not_a; 2233 t->VsrD(1) = a.VsrD(1) + b.VsrD(1); 2234 t->VsrD(0) = a.VsrD(0) + b.VsrD(0) + 2235 (~a.VsrD(1) < b.VsrD(1)); 2236 avr_qw_not(¬_a, a); 2237 return avr_qw_cmpu(not_a, b) < 0; 2238 } 2239 2240 #endif 2241 2242 void helper_vadduqm(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 2243 { 2244 #ifdef CONFIG_INT128 2245 r->u128 = a->u128 + b->u128; 2246 #else 2247 avr_qw_add(r, *a, *b); 2248 #endif 2249 } 2250 2251 void helper_vaddeuqm(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c) 2252 { 2253 #ifdef CONFIG_INT128 2254 r->u128 = a->u128 + b->u128 + (c->u128 & 1); 2255 #else 2256 2257 if (c->VsrD(1) & 1) { 2258 ppc_avr_t tmp; 2259 2260 tmp.VsrD(0) = 0; 2261 tmp.VsrD(1) = c->VsrD(1) & 1; 2262 avr_qw_add(&tmp, *a, tmp); 2263 avr_qw_add(r, tmp, *b); 2264 } else { 2265 avr_qw_add(r, *a, *b); 2266 } 2267 #endif 2268 } 2269 2270 void helper_vaddcuq(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 2271 { 2272 #ifdef CONFIG_INT128 2273 r->u128 = (~a->u128 < b->u128); 2274 #else 2275 ppc_avr_t not_a; 2276 2277 avr_qw_not(¬_a, *a); 2278 2279 r->VsrD(0) = 0; 2280 r->VsrD(1) = (avr_qw_cmpu(not_a, *b) < 0); 2281 #endif 2282 } 2283 2284 void helper_vaddecuq(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c) 2285 { 2286 #ifdef CONFIG_INT128 2287 int carry_out = (~a->u128 < b->u128); 2288 if (!carry_out && (c->u128 & 1)) { 2289 carry_out = ((a->u128 + b->u128 + 1) == 0) && 2290 ((a->u128 != 0) || (b->u128 != 0)); 2291 } 2292 r->u128 = carry_out; 2293 #else 2294 2295 int carry_in = c->VsrD(1) & 1; 2296 int carry_out = 0; 2297 ppc_avr_t tmp; 2298 2299 carry_out = avr_qw_addc(&tmp, *a, *b); 2300 2301 if (!carry_out && carry_in) { 2302 ppc_avr_t one = QW_ONE; 2303 carry_out = avr_qw_addc(&tmp, tmp, one); 2304 } 2305 r->VsrD(0) = 0; 2306 r->VsrD(1) = carry_out; 2307 #endif 2308 } 2309 2310 void helper_vsubuqm(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 2311 { 2312 #ifdef CONFIG_INT128 2313 r->u128 = a->u128 - b->u128; 2314 #else 2315 ppc_avr_t tmp; 2316 ppc_avr_t one = QW_ONE; 2317 2318 avr_qw_not(&tmp, *b); 2319 avr_qw_add(&tmp, *a, tmp); 2320 avr_qw_add(r, tmp, one); 2321 #endif 2322 } 2323 2324 void helper_vsubeuqm(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c) 2325 { 2326 #ifdef CONFIG_INT128 2327 r->u128 = a->u128 + ~b->u128 + (c->u128 & 1); 2328 #else 2329 ppc_avr_t tmp, sum; 2330 2331 avr_qw_not(&tmp, *b); 2332 avr_qw_add(&sum, *a, tmp); 2333 2334 tmp.VsrD(0) = 0; 2335 tmp.VsrD(1) = c->VsrD(1) & 1; 2336 avr_qw_add(r, sum, tmp); 2337 #endif 2338 } 2339 2340 void helper_vsubcuq(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 2341 { 2342 #ifdef CONFIG_INT128 2343 r->u128 = (~a->u128 < ~b->u128) || 2344 (a->u128 + ~b->u128 == (__uint128_t)-1); 2345 #else 2346 int carry = (avr_qw_cmpu(*a, *b) > 0); 2347 if (!carry) { 2348 ppc_avr_t tmp; 2349 avr_qw_not(&tmp, *b); 2350 avr_qw_add(&tmp, *a, tmp); 2351 carry = ((tmp.VsrSD(0) == -1ull) && (tmp.VsrSD(1) == -1ull)); 2352 } 2353 r->VsrD(0) = 0; 2354 r->VsrD(1) = carry; 2355 #endif 2356 } 2357 2358 void helper_vsubecuq(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c) 2359 { 2360 #ifdef CONFIG_INT128 2361 r->u128 = 2362 (~a->u128 < ~b->u128) || 2363 ((c->u128 & 1) && (a->u128 + ~b->u128 == (__uint128_t)-1)); 2364 #else 2365 int carry_in = c->VsrD(1) & 1; 2366 int carry_out = (avr_qw_cmpu(*a, *b) > 0); 2367 if (!carry_out && carry_in) { 2368 ppc_avr_t tmp; 2369 avr_qw_not(&tmp, *b); 2370 avr_qw_add(&tmp, *a, tmp); 2371 carry_out = ((tmp.VsrD(0) == -1ull) && (tmp.VsrD(1) == -1ull)); 2372 } 2373 2374 r->VsrD(0) = 0; 2375 r->VsrD(1) = carry_out; 2376 #endif 2377 } 2378 2379 #define BCD_PLUS_PREF_1 0xC 2380 #define BCD_PLUS_PREF_2 0xF 2381 #define BCD_PLUS_ALT_1 0xA 2382 #define BCD_NEG_PREF 0xD 2383 #define BCD_NEG_ALT 0xB 2384 #define BCD_PLUS_ALT_2 0xE 2385 #define NATIONAL_PLUS 0x2B 2386 #define NATIONAL_NEG 0x2D 2387 2388 #if defined(HOST_WORDS_BIGENDIAN) 2389 #define BCD_DIG_BYTE(n) (15 - ((n) / 2)) 2390 #else 2391 #define BCD_DIG_BYTE(n) ((n) / 2) 2392 #endif 2393 2394 static int bcd_get_sgn(ppc_avr_t *bcd) 2395 { 2396 switch (bcd->u8[BCD_DIG_BYTE(0)] & 0xF) { 2397 case BCD_PLUS_PREF_1: 2398 case BCD_PLUS_PREF_2: 2399 case BCD_PLUS_ALT_1: 2400 case BCD_PLUS_ALT_2: 2401 { 2402 return 1; 2403 } 2404 2405 case BCD_NEG_PREF: 2406 case BCD_NEG_ALT: 2407 { 2408 return -1; 2409 } 2410 2411 default: 2412 { 2413 return 0; 2414 } 2415 } 2416 } 2417 2418 static int bcd_preferred_sgn(int sgn, int ps) 2419 { 2420 if (sgn >= 0) { 2421 return (ps == 0) ? BCD_PLUS_PREF_1 : BCD_PLUS_PREF_2; 2422 } else { 2423 return BCD_NEG_PREF; 2424 } 2425 } 2426 2427 static uint8_t bcd_get_digit(ppc_avr_t *bcd, int n, int *invalid) 2428 { 2429 uint8_t result; 2430 if (n & 1) { 2431 result = bcd->u8[BCD_DIG_BYTE(n)] >> 4; 2432 } else { 2433 result = bcd->u8[BCD_DIG_BYTE(n)] & 0xF; 2434 } 2435 2436 if (unlikely(result > 9)) { 2437 *invalid = true; 2438 } 2439 return result; 2440 } 2441 2442 static void bcd_put_digit(ppc_avr_t *bcd, uint8_t digit, int n) 2443 { 2444 if (n & 1) { 2445 bcd->u8[BCD_DIG_BYTE(n)] &= 0x0F; 2446 bcd->u8[BCD_DIG_BYTE(n)] |= (digit << 4); 2447 } else { 2448 bcd->u8[BCD_DIG_BYTE(n)] &= 0xF0; 2449 bcd->u8[BCD_DIG_BYTE(n)] |= digit; 2450 } 2451 } 2452 2453 static bool bcd_is_valid(ppc_avr_t *bcd) 2454 { 2455 int i; 2456 int invalid = 0; 2457 2458 if (bcd_get_sgn(bcd) == 0) { 2459 return false; 2460 } 2461 2462 for (i = 1; i < 32; i++) { 2463 bcd_get_digit(bcd, i, &invalid); 2464 if (unlikely(invalid)) { 2465 return false; 2466 } 2467 } 2468 return true; 2469 } 2470 2471 static int bcd_cmp_zero(ppc_avr_t *bcd) 2472 { 2473 if (bcd->VsrD(0) == 0 && (bcd->VsrD(1) >> 4) == 0) { 2474 return CRF_EQ; 2475 } else { 2476 return (bcd_get_sgn(bcd) == 1) ? CRF_GT : CRF_LT; 2477 } 2478 } 2479 2480 static uint16_t get_national_digit(ppc_avr_t *reg, int n) 2481 { 2482 return reg->VsrH(7 - n); 2483 } 2484 2485 static void set_national_digit(ppc_avr_t *reg, uint8_t val, int n) 2486 { 2487 reg->VsrH(7 - n) = val; 2488 } 2489 2490 static int bcd_cmp_mag(ppc_avr_t *a, ppc_avr_t *b) 2491 { 2492 int i; 2493 int invalid = 0; 2494 for (i = 31; i > 0; i--) { 2495 uint8_t dig_a = bcd_get_digit(a, i, &invalid); 2496 uint8_t dig_b = bcd_get_digit(b, i, &invalid); 2497 if (unlikely(invalid)) { 2498 return 0; /* doesn't matter */ 2499 } else if (dig_a > dig_b) { 2500 return 1; 2501 } else if (dig_a < dig_b) { 2502 return -1; 2503 } 2504 } 2505 2506 return 0; 2507 } 2508 2509 static void bcd_add_mag(ppc_avr_t *t, ppc_avr_t *a, ppc_avr_t *b, int *invalid, 2510 int *overflow) 2511 { 2512 int carry = 0; 2513 int i; 2514 for (i = 1; i <= 31; i++) { 2515 uint8_t digit = bcd_get_digit(a, i, invalid) + 2516 bcd_get_digit(b, i, invalid) + carry; 2517 if (digit > 9) { 2518 carry = 1; 2519 digit -= 10; 2520 } else { 2521 carry = 0; 2522 } 2523 2524 bcd_put_digit(t, digit, i); 2525 } 2526 2527 *overflow = carry; 2528 } 2529 2530 static void bcd_sub_mag(ppc_avr_t *t, ppc_avr_t *a, ppc_avr_t *b, int *invalid, 2531 int *overflow) 2532 { 2533 int carry = 0; 2534 int i; 2535 2536 for (i = 1; i <= 31; i++) { 2537 uint8_t digit = bcd_get_digit(a, i, invalid) - 2538 bcd_get_digit(b, i, invalid) + carry; 2539 if (digit & 0x80) { 2540 carry = -1; 2541 digit += 10; 2542 } else { 2543 carry = 0; 2544 } 2545 2546 bcd_put_digit(t, digit, i); 2547 } 2548 2549 *overflow = carry; 2550 } 2551 2552 uint32_t helper_bcdadd(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t ps) 2553 { 2554 2555 int sgna = bcd_get_sgn(a); 2556 int sgnb = bcd_get_sgn(b); 2557 int invalid = (sgna == 0) || (sgnb == 0); 2558 int overflow = 0; 2559 uint32_t cr = 0; 2560 ppc_avr_t result = { .u64 = { 0, 0 } }; 2561 2562 if (!invalid) { 2563 if (sgna == sgnb) { 2564 result.u8[BCD_DIG_BYTE(0)] = bcd_preferred_sgn(sgna, ps); 2565 bcd_add_mag(&result, a, b, &invalid, &overflow); 2566 cr = bcd_cmp_zero(&result); 2567 } else { 2568 int magnitude = bcd_cmp_mag(a, b); 2569 if (magnitude > 0) { 2570 result.u8[BCD_DIG_BYTE(0)] = bcd_preferred_sgn(sgna, ps); 2571 bcd_sub_mag(&result, a, b, &invalid, &overflow); 2572 cr = (sgna > 0) ? CRF_GT : CRF_LT; 2573 } else if (magnitude < 0) { 2574 result.u8[BCD_DIG_BYTE(0)] = bcd_preferred_sgn(sgnb, ps); 2575 bcd_sub_mag(&result, b, a, &invalid, &overflow); 2576 cr = (sgnb > 0) ? CRF_GT : CRF_LT; 2577 } else { 2578 result.u8[BCD_DIG_BYTE(0)] = bcd_preferred_sgn(0, ps); 2579 cr = CRF_EQ; 2580 } 2581 } 2582 } 2583 2584 if (unlikely(invalid)) { 2585 result.VsrD(0) = result.VsrD(1) = -1; 2586 cr = CRF_SO; 2587 } else if (overflow) { 2588 cr |= CRF_SO; 2589 } 2590 2591 *r = result; 2592 2593 return cr; 2594 } 2595 2596 uint32_t helper_bcdsub(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t ps) 2597 { 2598 ppc_avr_t bcopy = *b; 2599 int sgnb = bcd_get_sgn(b); 2600 if (sgnb < 0) { 2601 bcd_put_digit(&bcopy, BCD_PLUS_PREF_1, 0); 2602 } else if (sgnb > 0) { 2603 bcd_put_digit(&bcopy, BCD_NEG_PREF, 0); 2604 } 2605 /* else invalid ... defer to bcdadd code for proper handling */ 2606 2607 return helper_bcdadd(r, a, &bcopy, ps); 2608 } 2609 2610 uint32_t helper_bcdcfn(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps) 2611 { 2612 int i; 2613 int cr = 0; 2614 uint16_t national = 0; 2615 uint16_t sgnb = get_national_digit(b, 0); 2616 ppc_avr_t ret = { .u64 = { 0, 0 } }; 2617 int invalid = (sgnb != NATIONAL_PLUS && sgnb != NATIONAL_NEG); 2618 2619 for (i = 1; i < 8; i++) { 2620 national = get_national_digit(b, i); 2621 if (unlikely(national < 0x30 || national > 0x39)) { 2622 invalid = 1; 2623 break; 2624 } 2625 2626 bcd_put_digit(&ret, national & 0xf, i); 2627 } 2628 2629 if (sgnb == NATIONAL_PLUS) { 2630 bcd_put_digit(&ret, (ps == 0) ? BCD_PLUS_PREF_1 : BCD_PLUS_PREF_2, 0); 2631 } else { 2632 bcd_put_digit(&ret, BCD_NEG_PREF, 0); 2633 } 2634 2635 cr = bcd_cmp_zero(&ret); 2636 2637 if (unlikely(invalid)) { 2638 cr = CRF_SO; 2639 } 2640 2641 *r = ret; 2642 2643 return cr; 2644 } 2645 2646 uint32_t helper_bcdctn(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps) 2647 { 2648 int i; 2649 int cr = 0; 2650 int sgnb = bcd_get_sgn(b); 2651 int invalid = (sgnb == 0); 2652 ppc_avr_t ret = { .u64 = { 0, 0 } }; 2653 2654 int ox_flag = (b->VsrD(0) != 0) || ((b->VsrD(1) >> 32) != 0); 2655 2656 for (i = 1; i < 8; i++) { 2657 set_national_digit(&ret, 0x30 + bcd_get_digit(b, i, &invalid), i); 2658 2659 if (unlikely(invalid)) { 2660 break; 2661 } 2662 } 2663 set_national_digit(&ret, (sgnb == -1) ? NATIONAL_NEG : NATIONAL_PLUS, 0); 2664 2665 cr = bcd_cmp_zero(b); 2666 2667 if (ox_flag) { 2668 cr |= CRF_SO; 2669 } 2670 2671 if (unlikely(invalid)) { 2672 cr = CRF_SO; 2673 } 2674 2675 *r = ret; 2676 2677 return cr; 2678 } 2679 2680 uint32_t helper_bcdcfz(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps) 2681 { 2682 int i; 2683 int cr = 0; 2684 int invalid = 0; 2685 int zone_digit = 0; 2686 int zone_lead = ps ? 0xF : 0x3; 2687 int digit = 0; 2688 ppc_avr_t ret = { .u64 = { 0, 0 } }; 2689 int sgnb = b->u8[BCD_DIG_BYTE(0)] >> 4; 2690 2691 if (unlikely((sgnb < 0xA) && ps)) { 2692 invalid = 1; 2693 } 2694 2695 for (i = 0; i < 16; i++) { 2696 zone_digit = i ? b->u8[BCD_DIG_BYTE(i * 2)] >> 4 : zone_lead; 2697 digit = b->u8[BCD_DIG_BYTE(i * 2)] & 0xF; 2698 if (unlikely(zone_digit != zone_lead || digit > 0x9)) { 2699 invalid = 1; 2700 break; 2701 } 2702 2703 bcd_put_digit(&ret, digit, i + 1); 2704 } 2705 2706 if ((ps && (sgnb == 0xB || sgnb == 0xD)) || 2707 (!ps && (sgnb & 0x4))) { 2708 bcd_put_digit(&ret, BCD_NEG_PREF, 0); 2709 } else { 2710 bcd_put_digit(&ret, BCD_PLUS_PREF_1, 0); 2711 } 2712 2713 cr = bcd_cmp_zero(&ret); 2714 2715 if (unlikely(invalid)) { 2716 cr = CRF_SO; 2717 } 2718 2719 *r = ret; 2720 2721 return cr; 2722 } 2723 2724 uint32_t helper_bcdctz(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps) 2725 { 2726 int i; 2727 int cr = 0; 2728 uint8_t digit = 0; 2729 int sgnb = bcd_get_sgn(b); 2730 int zone_lead = (ps) ? 0xF0 : 0x30; 2731 int invalid = (sgnb == 0); 2732 ppc_avr_t ret = { .u64 = { 0, 0 } }; 2733 2734 int ox_flag = ((b->VsrD(0) >> 4) != 0); 2735 2736 for (i = 0; i < 16; i++) { 2737 digit = bcd_get_digit(b, i + 1, &invalid); 2738 2739 if (unlikely(invalid)) { 2740 break; 2741 } 2742 2743 ret.u8[BCD_DIG_BYTE(i * 2)] = zone_lead + digit; 2744 } 2745 2746 if (ps) { 2747 bcd_put_digit(&ret, (sgnb == 1) ? 0xC : 0xD, 1); 2748 } else { 2749 bcd_put_digit(&ret, (sgnb == 1) ? 0x3 : 0x7, 1); 2750 } 2751 2752 cr = bcd_cmp_zero(b); 2753 2754 if (ox_flag) { 2755 cr |= CRF_SO; 2756 } 2757 2758 if (unlikely(invalid)) { 2759 cr = CRF_SO; 2760 } 2761 2762 *r = ret; 2763 2764 return cr; 2765 } 2766 2767 uint32_t helper_bcdcfsq(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps) 2768 { 2769 int i; 2770 int cr = 0; 2771 uint64_t lo_value; 2772 uint64_t hi_value; 2773 ppc_avr_t ret = { .u64 = { 0, 0 } }; 2774 2775 if (b->VsrSD(0) < 0) { 2776 lo_value = -b->VsrSD(1); 2777 hi_value = ~b->VsrD(0) + !lo_value; 2778 bcd_put_digit(&ret, 0xD, 0); 2779 } else { 2780 lo_value = b->VsrD(1); 2781 hi_value = b->VsrD(0); 2782 bcd_put_digit(&ret, bcd_preferred_sgn(0, ps), 0); 2783 } 2784 2785 if (divu128(&lo_value, &hi_value, 1000000000000000ULL) || 2786 lo_value > 9999999999999999ULL) { 2787 cr = CRF_SO; 2788 } 2789 2790 for (i = 1; i < 16; hi_value /= 10, i++) { 2791 bcd_put_digit(&ret, hi_value % 10, i); 2792 } 2793 2794 for (; i < 32; lo_value /= 10, i++) { 2795 bcd_put_digit(&ret, lo_value % 10, i); 2796 } 2797 2798 cr |= bcd_cmp_zero(&ret); 2799 2800 *r = ret; 2801 2802 return cr; 2803 } 2804 2805 uint32_t helper_bcdctsq(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps) 2806 { 2807 uint8_t i; 2808 int cr; 2809 uint64_t carry; 2810 uint64_t unused; 2811 uint64_t lo_value; 2812 uint64_t hi_value = 0; 2813 int sgnb = bcd_get_sgn(b); 2814 int invalid = (sgnb == 0); 2815 2816 lo_value = bcd_get_digit(b, 31, &invalid); 2817 for (i = 30; i > 0; i--) { 2818 mulu64(&lo_value, &carry, lo_value, 10ULL); 2819 mulu64(&hi_value, &unused, hi_value, 10ULL); 2820 lo_value += bcd_get_digit(b, i, &invalid); 2821 hi_value += carry; 2822 2823 if (unlikely(invalid)) { 2824 break; 2825 } 2826 } 2827 2828 if (sgnb == -1) { 2829 r->VsrSD(1) = -lo_value; 2830 r->VsrSD(0) = ~hi_value + !r->VsrSD(1); 2831 } else { 2832 r->VsrSD(1) = lo_value; 2833 r->VsrSD(0) = hi_value; 2834 } 2835 2836 cr = bcd_cmp_zero(b); 2837 2838 if (unlikely(invalid)) { 2839 cr = CRF_SO; 2840 } 2841 2842 return cr; 2843 } 2844 2845 uint32_t helper_bcdcpsgn(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t ps) 2846 { 2847 int i; 2848 int invalid = 0; 2849 2850 if (bcd_get_sgn(a) == 0 || bcd_get_sgn(b) == 0) { 2851 return CRF_SO; 2852 } 2853 2854 *r = *a; 2855 bcd_put_digit(r, b->u8[BCD_DIG_BYTE(0)] & 0xF, 0); 2856 2857 for (i = 1; i < 32; i++) { 2858 bcd_get_digit(a, i, &invalid); 2859 bcd_get_digit(b, i, &invalid); 2860 if (unlikely(invalid)) { 2861 return CRF_SO; 2862 } 2863 } 2864 2865 return bcd_cmp_zero(r); 2866 } 2867 2868 uint32_t helper_bcdsetsgn(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps) 2869 { 2870 int sgnb = bcd_get_sgn(b); 2871 2872 *r = *b; 2873 bcd_put_digit(r, bcd_preferred_sgn(sgnb, ps), 0); 2874 2875 if (bcd_is_valid(b) == false) { 2876 return CRF_SO; 2877 } 2878 2879 return bcd_cmp_zero(r); 2880 } 2881 2882 uint32_t helper_bcds(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t ps) 2883 { 2884 int cr; 2885 #if defined(HOST_WORDS_BIGENDIAN) 2886 int i = a->s8[7]; 2887 #else 2888 int i = a->s8[8]; 2889 #endif 2890 bool ox_flag = false; 2891 int sgnb = bcd_get_sgn(b); 2892 ppc_avr_t ret = *b; 2893 ret.VsrD(1) &= ~0xf; 2894 2895 if (bcd_is_valid(b) == false) { 2896 return CRF_SO; 2897 } 2898 2899 if (unlikely(i > 31)) { 2900 i = 31; 2901 } else if (unlikely(i < -31)) { 2902 i = -31; 2903 } 2904 2905 if (i > 0) { 2906 ulshift(&ret.VsrD(1), &ret.VsrD(0), i * 4, &ox_flag); 2907 } else { 2908 urshift(&ret.VsrD(1), &ret.VsrD(0), -i * 4); 2909 } 2910 bcd_put_digit(&ret, bcd_preferred_sgn(sgnb, ps), 0); 2911 2912 *r = ret; 2913 2914 cr = bcd_cmp_zero(r); 2915 if (ox_flag) { 2916 cr |= CRF_SO; 2917 } 2918 2919 return cr; 2920 } 2921 2922 uint32_t helper_bcdus(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t ps) 2923 { 2924 int cr; 2925 int i; 2926 int invalid = 0; 2927 bool ox_flag = false; 2928 ppc_avr_t ret = *b; 2929 2930 for (i = 0; i < 32; i++) { 2931 bcd_get_digit(b, i, &invalid); 2932 2933 if (unlikely(invalid)) { 2934 return CRF_SO; 2935 } 2936 } 2937 2938 #if defined(HOST_WORDS_BIGENDIAN) 2939 i = a->s8[7]; 2940 #else 2941 i = a->s8[8]; 2942 #endif 2943 if (i >= 32) { 2944 ox_flag = true; 2945 ret.VsrD(1) = ret.VsrD(0) = 0; 2946 } else if (i <= -32) { 2947 ret.VsrD(1) = ret.VsrD(0) = 0; 2948 } else if (i > 0) { 2949 ulshift(&ret.VsrD(1), &ret.VsrD(0), i * 4, &ox_flag); 2950 } else { 2951 urshift(&ret.VsrD(1), &ret.VsrD(0), -i * 4); 2952 } 2953 *r = ret; 2954 2955 cr = bcd_cmp_zero(r); 2956 if (ox_flag) { 2957 cr |= CRF_SO; 2958 } 2959 2960 return cr; 2961 } 2962 2963 uint32_t helper_bcdsr(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t ps) 2964 { 2965 int cr; 2966 int unused = 0; 2967 int invalid = 0; 2968 bool ox_flag = false; 2969 int sgnb = bcd_get_sgn(b); 2970 ppc_avr_t ret = *b; 2971 ret.VsrD(1) &= ~0xf; 2972 2973 #if defined(HOST_WORDS_BIGENDIAN) 2974 int i = a->s8[7]; 2975 ppc_avr_t bcd_one = { .u64 = { 0, 0x10 } }; 2976 #else 2977 int i = a->s8[8]; 2978 ppc_avr_t bcd_one = { .u64 = { 0x10, 0 } }; 2979 #endif 2980 2981 if (bcd_is_valid(b) == false) { 2982 return CRF_SO; 2983 } 2984 2985 if (unlikely(i > 31)) { 2986 i = 31; 2987 } else if (unlikely(i < -31)) { 2988 i = -31; 2989 } 2990 2991 if (i > 0) { 2992 ulshift(&ret.VsrD(1), &ret.VsrD(0), i * 4, &ox_flag); 2993 } else { 2994 urshift(&ret.VsrD(1), &ret.VsrD(0), -i * 4); 2995 2996 if (bcd_get_digit(&ret, 0, &invalid) >= 5) { 2997 bcd_add_mag(&ret, &ret, &bcd_one, &invalid, &unused); 2998 } 2999 } 3000 bcd_put_digit(&ret, bcd_preferred_sgn(sgnb, ps), 0); 3001 3002 cr = bcd_cmp_zero(&ret); 3003 if (ox_flag) { 3004 cr |= CRF_SO; 3005 } 3006 *r = ret; 3007 3008 return cr; 3009 } 3010 3011 uint32_t helper_bcdtrunc(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t ps) 3012 { 3013 uint64_t mask; 3014 uint32_t ox_flag = 0; 3015 #if defined(HOST_WORDS_BIGENDIAN) 3016 int i = a->s16[3] + 1; 3017 #else 3018 int i = a->s16[4] + 1; 3019 #endif 3020 ppc_avr_t ret = *b; 3021 3022 if (bcd_is_valid(b) == false) { 3023 return CRF_SO; 3024 } 3025 3026 if (i > 16 && i < 32) { 3027 mask = (uint64_t)-1 >> (128 - i * 4); 3028 if (ret.VsrD(0) & ~mask) { 3029 ox_flag = CRF_SO; 3030 } 3031 3032 ret.VsrD(0) &= mask; 3033 } else if (i >= 0 && i <= 16) { 3034 mask = (uint64_t)-1 >> (64 - i * 4); 3035 if (ret.VsrD(0) || (ret.VsrD(1) & ~mask)) { 3036 ox_flag = CRF_SO; 3037 } 3038 3039 ret.VsrD(1) &= mask; 3040 ret.VsrD(0) = 0; 3041 } 3042 bcd_put_digit(&ret, bcd_preferred_sgn(bcd_get_sgn(b), ps), 0); 3043 *r = ret; 3044 3045 return bcd_cmp_zero(&ret) | ox_flag; 3046 } 3047 3048 uint32_t helper_bcdutrunc(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t ps) 3049 { 3050 int i; 3051 uint64_t mask; 3052 uint32_t ox_flag = 0; 3053 int invalid = 0; 3054 ppc_avr_t ret = *b; 3055 3056 for (i = 0; i < 32; i++) { 3057 bcd_get_digit(b, i, &invalid); 3058 3059 if (unlikely(invalid)) { 3060 return CRF_SO; 3061 } 3062 } 3063 3064 #if defined(HOST_WORDS_BIGENDIAN) 3065 i = a->s16[3]; 3066 #else 3067 i = a->s16[4]; 3068 #endif 3069 if (i > 16 && i < 33) { 3070 mask = (uint64_t)-1 >> (128 - i * 4); 3071 if (ret.VsrD(0) & ~mask) { 3072 ox_flag = CRF_SO; 3073 } 3074 3075 ret.VsrD(0) &= mask; 3076 } else if (i > 0 && i <= 16) { 3077 mask = (uint64_t)-1 >> (64 - i * 4); 3078 if (ret.VsrD(0) || (ret.VsrD(1) & ~mask)) { 3079 ox_flag = CRF_SO; 3080 } 3081 3082 ret.VsrD(1) &= mask; 3083 ret.VsrD(0) = 0; 3084 } else if (i == 0) { 3085 if (ret.VsrD(0) || ret.VsrD(1)) { 3086 ox_flag = CRF_SO; 3087 } 3088 ret.VsrD(0) = ret.VsrD(1) = 0; 3089 } 3090 3091 *r = ret; 3092 if (r->VsrD(0) == 0 && r->VsrD(1) == 0) { 3093 return ox_flag | CRF_EQ; 3094 } 3095 3096 return ox_flag | CRF_GT; 3097 } 3098 3099 void helper_vsbox(ppc_avr_t *r, ppc_avr_t *a) 3100 { 3101 int i; 3102 VECTOR_FOR_INORDER_I(i, u8) { 3103 r->u8[i] = AES_sbox[a->u8[i]]; 3104 } 3105 } 3106 3107 void helper_vcipher(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 3108 { 3109 ppc_avr_t result; 3110 int i; 3111 3112 VECTOR_FOR_INORDER_I(i, u32) { 3113 result.VsrW(i) = b->VsrW(i) ^ 3114 (AES_Te0[a->VsrB(AES_shifts[4 * i + 0])] ^ 3115 AES_Te1[a->VsrB(AES_shifts[4 * i + 1])] ^ 3116 AES_Te2[a->VsrB(AES_shifts[4 * i + 2])] ^ 3117 AES_Te3[a->VsrB(AES_shifts[4 * i + 3])]); 3118 } 3119 *r = result; 3120 } 3121 3122 void helper_vcipherlast(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 3123 { 3124 ppc_avr_t result; 3125 int i; 3126 3127 VECTOR_FOR_INORDER_I(i, u8) { 3128 result.VsrB(i) = b->VsrB(i) ^ (AES_sbox[a->VsrB(AES_shifts[i])]); 3129 } 3130 *r = result; 3131 } 3132 3133 void helper_vncipher(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 3134 { 3135 /* This differs from what is written in ISA V2.07. The RTL is */ 3136 /* incorrect and will be fixed in V2.07B. */ 3137 int i; 3138 ppc_avr_t tmp; 3139 3140 VECTOR_FOR_INORDER_I(i, u8) { 3141 tmp.VsrB(i) = b->VsrB(i) ^ AES_isbox[a->VsrB(AES_ishifts[i])]; 3142 } 3143 3144 VECTOR_FOR_INORDER_I(i, u32) { 3145 r->VsrW(i) = 3146 AES_imc[tmp.VsrB(4 * i + 0)][0] ^ 3147 AES_imc[tmp.VsrB(4 * i + 1)][1] ^ 3148 AES_imc[tmp.VsrB(4 * i + 2)][2] ^ 3149 AES_imc[tmp.VsrB(4 * i + 3)][3]; 3150 } 3151 } 3152 3153 void helper_vncipherlast(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 3154 { 3155 ppc_avr_t result; 3156 int i; 3157 3158 VECTOR_FOR_INORDER_I(i, u8) { 3159 result.VsrB(i) = b->VsrB(i) ^ (AES_isbox[a->VsrB(AES_ishifts[i])]); 3160 } 3161 *r = result; 3162 } 3163 3164 void helper_vshasigmaw(ppc_avr_t *r, ppc_avr_t *a, uint32_t st_six) 3165 { 3166 int st = (st_six & 0x10) != 0; 3167 int six = st_six & 0xF; 3168 int i; 3169 3170 for (i = 0; i < ARRAY_SIZE(r->u32); i++) { 3171 if (st == 0) { 3172 if ((six & (0x8 >> i)) == 0) { 3173 r->VsrW(i) = ror32(a->VsrW(i), 7) ^ 3174 ror32(a->VsrW(i), 18) ^ 3175 (a->VsrW(i) >> 3); 3176 } else { /* six.bit[i] == 1 */ 3177 r->VsrW(i) = ror32(a->VsrW(i), 17) ^ 3178 ror32(a->VsrW(i), 19) ^ 3179 (a->VsrW(i) >> 10); 3180 } 3181 } else { /* st == 1 */ 3182 if ((six & (0x8 >> i)) == 0) { 3183 r->VsrW(i) = ror32(a->VsrW(i), 2) ^ 3184 ror32(a->VsrW(i), 13) ^ 3185 ror32(a->VsrW(i), 22); 3186 } else { /* six.bit[i] == 1 */ 3187 r->VsrW(i) = ror32(a->VsrW(i), 6) ^ 3188 ror32(a->VsrW(i), 11) ^ 3189 ror32(a->VsrW(i), 25); 3190 } 3191 } 3192 } 3193 } 3194 3195 void helper_vshasigmad(ppc_avr_t *r, ppc_avr_t *a, uint32_t st_six) 3196 { 3197 int st = (st_six & 0x10) != 0; 3198 int six = st_six & 0xF; 3199 int i; 3200 3201 for (i = 0; i < ARRAY_SIZE(r->u64); i++) { 3202 if (st == 0) { 3203 if ((six & (0x8 >> (2 * i))) == 0) { 3204 r->VsrD(i) = ror64(a->VsrD(i), 1) ^ 3205 ror64(a->VsrD(i), 8) ^ 3206 (a->VsrD(i) >> 7); 3207 } else { /* six.bit[2*i] == 1 */ 3208 r->VsrD(i) = ror64(a->VsrD(i), 19) ^ 3209 ror64(a->VsrD(i), 61) ^ 3210 (a->VsrD(i) >> 6); 3211 } 3212 } else { /* st == 1 */ 3213 if ((six & (0x8 >> (2 * i))) == 0) { 3214 r->VsrD(i) = ror64(a->VsrD(i), 28) ^ 3215 ror64(a->VsrD(i), 34) ^ 3216 ror64(a->VsrD(i), 39); 3217 } else { /* six.bit[2*i] == 1 */ 3218 r->VsrD(i) = ror64(a->VsrD(i), 14) ^ 3219 ror64(a->VsrD(i), 18) ^ 3220 ror64(a->VsrD(i), 41); 3221 } 3222 } 3223 } 3224 } 3225 3226 void helper_vpermxor(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c) 3227 { 3228 ppc_avr_t result; 3229 int i; 3230 3231 for (i = 0; i < ARRAY_SIZE(r->u8); i++) { 3232 int indexA = c->VsrB(i) >> 4; 3233 int indexB = c->VsrB(i) & 0xF; 3234 3235 result.VsrB(i) = a->VsrB(indexA) ^ b->VsrB(indexB); 3236 } 3237 *r = result; 3238 } 3239 3240 #undef VECTOR_FOR_INORDER_I 3241 3242 /*****************************************************************************/ 3243 /* SPE extension helpers */ 3244 /* Use a table to make this quicker */ 3245 static const uint8_t hbrev[16] = { 3246 0x0, 0x8, 0x4, 0xC, 0x2, 0xA, 0x6, 0xE, 3247 0x1, 0x9, 0x5, 0xD, 0x3, 0xB, 0x7, 0xF, 3248 }; 3249 3250 static inline uint8_t byte_reverse(uint8_t val) 3251 { 3252 return hbrev[val >> 4] | (hbrev[val & 0xF] << 4); 3253 } 3254 3255 static inline uint32_t word_reverse(uint32_t val) 3256 { 3257 return byte_reverse(val >> 24) | (byte_reverse(val >> 16) << 8) | 3258 (byte_reverse(val >> 8) << 16) | (byte_reverse(val) << 24); 3259 } 3260 3261 #define MASKBITS 16 /* Random value - to be fixed (implementation dependent) */ 3262 target_ulong helper_brinc(target_ulong arg1, target_ulong arg2) 3263 { 3264 uint32_t a, b, d, mask; 3265 3266 mask = UINT32_MAX >> (32 - MASKBITS); 3267 a = arg1 & mask; 3268 b = arg2 & mask; 3269 d = word_reverse(1 + word_reverse(a | ~b)); 3270 return (arg1 & ~mask) | (d & b); 3271 } 3272 3273 uint32_t helper_cntlsw32(uint32_t val) 3274 { 3275 if (val & 0x80000000) { 3276 return clz32(~val); 3277 } else { 3278 return clz32(val); 3279 } 3280 } 3281 3282 uint32_t helper_cntlzw32(uint32_t val) 3283 { 3284 return clz32(val); 3285 } 3286 3287 /* 440 specific */ 3288 target_ulong helper_dlmzb(CPUPPCState *env, target_ulong high, 3289 target_ulong low, uint32_t update_Rc) 3290 { 3291 target_ulong mask; 3292 int i; 3293 3294 i = 1; 3295 for (mask = 0xFF000000; mask != 0; mask = mask >> 8) { 3296 if ((high & mask) == 0) { 3297 if (update_Rc) { 3298 env->crf[0] = 0x4; 3299 } 3300 goto done; 3301 } 3302 i++; 3303 } 3304 for (mask = 0xFF000000; mask != 0; mask = mask >> 8) { 3305 if ((low & mask) == 0) { 3306 if (update_Rc) { 3307 env->crf[0] = 0x8; 3308 } 3309 goto done; 3310 } 3311 i++; 3312 } 3313 i = 8; 3314 if (update_Rc) { 3315 env->crf[0] = 0x2; 3316 } 3317 done: 3318 env->xer = (env->xer & ~0x7F) | i; 3319 if (update_Rc) { 3320 env->crf[0] |= xer_so; 3321 } 3322 return i; 3323 } 3324