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, target_ulong xtn, 1903 target_ulong xbn, uint32_t index) 1904 { 1905 ppc_vsr_t xt, xb; 1906 size_t es = sizeof(uint32_t); 1907 uint32_t ext_index; 1908 int i; 1909 1910 getVSR(xbn, &xb, env); 1911 memset(&xt, 0, sizeof(xt)); 1912 1913 ext_index = index; 1914 for (i = 0; i < es; i++, ext_index++) { 1915 xt.VsrB(8 - es + i) = xb.VsrB(ext_index % 16); 1916 } 1917 1918 putVSR(xtn, &xt, env); 1919 } 1920 1921 void helper_xxinsertw(CPUPPCState *env, target_ulong xtn, 1922 target_ulong xbn, uint32_t index) 1923 { 1924 ppc_vsr_t xt, xb; 1925 size_t es = sizeof(uint32_t); 1926 int ins_index, i = 0; 1927 1928 getVSR(xbn, &xb, env); 1929 getVSR(xtn, &xt, env); 1930 1931 ins_index = index; 1932 for (i = 0; i < es && ins_index < 16; i++, ins_index++) { 1933 xt.VsrB(ins_index) = xb.VsrB(8 - es + i); 1934 } 1935 1936 putVSR(xtn, &xt, env); 1937 } 1938 1939 #define VEXT_SIGNED(name, element, cast) \ 1940 void helper_##name(ppc_avr_t *r, ppc_avr_t *b) \ 1941 { \ 1942 int i; \ 1943 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ 1944 r->element[i] = (cast)b->element[i]; \ 1945 } \ 1946 } 1947 VEXT_SIGNED(vextsb2w, s32, int8_t) 1948 VEXT_SIGNED(vextsb2d, s64, int8_t) 1949 VEXT_SIGNED(vextsh2w, s32, int16_t) 1950 VEXT_SIGNED(vextsh2d, s64, int16_t) 1951 VEXT_SIGNED(vextsw2d, s64, int32_t) 1952 #undef VEXT_SIGNED 1953 1954 #define VNEG(name, element) \ 1955 void helper_##name(ppc_avr_t *r, ppc_avr_t *b) \ 1956 { \ 1957 int i; \ 1958 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ 1959 r->element[i] = -b->element[i]; \ 1960 } \ 1961 } 1962 VNEG(vnegw, s32) 1963 VNEG(vnegd, s64) 1964 #undef VNEG 1965 1966 void helper_vsro(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 1967 { 1968 int sh = (b->VsrB(0xf) >> 3) & 0xf; 1969 1970 #if defined(HOST_WORDS_BIGENDIAN) 1971 memmove(&r->u8[sh], &a->u8[0], 16 - sh); 1972 memset(&r->u8[0], 0, sh); 1973 #else 1974 memmove(&r->u8[0], &a->u8[sh], 16 - sh); 1975 memset(&r->u8[16 - sh], 0, sh); 1976 #endif 1977 } 1978 1979 void helper_vsubcuw(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 1980 { 1981 int i; 1982 1983 for (i = 0; i < ARRAY_SIZE(r->u32); i++) { 1984 r->u32[i] = a->u32[i] >= b->u32[i]; 1985 } 1986 } 1987 1988 void helper_vsumsws(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 1989 { 1990 int64_t t; 1991 int i, upper; 1992 ppc_avr_t result; 1993 int sat = 0; 1994 1995 upper = ARRAY_SIZE(r->s32) - 1; 1996 t = (int64_t)b->VsrSW(upper); 1997 for (i = 0; i < ARRAY_SIZE(r->s32); i++) { 1998 t += a->VsrSW(i); 1999 result.VsrSW(i) = 0; 2000 } 2001 result.VsrSW(upper) = cvtsdsw(t, &sat); 2002 *r = result; 2003 2004 if (sat) { 2005 set_vscr_sat(env); 2006 } 2007 } 2008 2009 void helper_vsum2sws(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 2010 { 2011 int i, j, upper; 2012 ppc_avr_t result; 2013 int sat = 0; 2014 2015 upper = 1; 2016 for (i = 0; i < ARRAY_SIZE(r->u64); i++) { 2017 int64_t t = (int64_t)b->VsrSW(upper + i * 2); 2018 2019 result.VsrD(i) = 0; 2020 for (j = 0; j < ARRAY_SIZE(r->u64); j++) { 2021 t += a->VsrSW(2 * i + j); 2022 } 2023 result.VsrSW(upper + i * 2) = cvtsdsw(t, &sat); 2024 } 2025 2026 *r = result; 2027 if (sat) { 2028 set_vscr_sat(env); 2029 } 2030 } 2031 2032 void helper_vsum4sbs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 2033 { 2034 int i, j; 2035 int sat = 0; 2036 2037 for (i = 0; i < ARRAY_SIZE(r->s32); i++) { 2038 int64_t t = (int64_t)b->s32[i]; 2039 2040 for (j = 0; j < ARRAY_SIZE(r->s32); j++) { 2041 t += a->s8[4 * i + j]; 2042 } 2043 r->s32[i] = cvtsdsw(t, &sat); 2044 } 2045 2046 if (sat) { 2047 set_vscr_sat(env); 2048 } 2049 } 2050 2051 void helper_vsum4shs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 2052 { 2053 int sat = 0; 2054 int i; 2055 2056 for (i = 0; i < ARRAY_SIZE(r->s32); i++) { 2057 int64_t t = (int64_t)b->s32[i]; 2058 2059 t += a->s16[2 * i] + a->s16[2 * i + 1]; 2060 r->s32[i] = cvtsdsw(t, &sat); 2061 } 2062 2063 if (sat) { 2064 set_vscr_sat(env); 2065 } 2066 } 2067 2068 void helper_vsum4ubs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 2069 { 2070 int i, j; 2071 int sat = 0; 2072 2073 for (i = 0; i < ARRAY_SIZE(r->u32); i++) { 2074 uint64_t t = (uint64_t)b->u32[i]; 2075 2076 for (j = 0; j < ARRAY_SIZE(r->u32); j++) { 2077 t += a->u8[4 * i + j]; 2078 } 2079 r->u32[i] = cvtuduw(t, &sat); 2080 } 2081 2082 if (sat) { 2083 set_vscr_sat(env); 2084 } 2085 } 2086 2087 #if defined(HOST_WORDS_BIGENDIAN) 2088 #define UPKHI 1 2089 #define UPKLO 0 2090 #else 2091 #define UPKHI 0 2092 #define UPKLO 1 2093 #endif 2094 #define VUPKPX(suffix, hi) \ 2095 void helper_vupk##suffix(ppc_avr_t *r, ppc_avr_t *b) \ 2096 { \ 2097 int i; \ 2098 ppc_avr_t result; \ 2099 \ 2100 for (i = 0; i < ARRAY_SIZE(r->u32); i++) { \ 2101 uint16_t e = b->u16[hi ? i : i + 4]; \ 2102 uint8_t a = (e >> 15) ? 0xff : 0; \ 2103 uint8_t r = (e >> 10) & 0x1f; \ 2104 uint8_t g = (e >> 5) & 0x1f; \ 2105 uint8_t b = e & 0x1f; \ 2106 \ 2107 result.u32[i] = (a << 24) | (r << 16) | (g << 8) | b; \ 2108 } \ 2109 *r = result; \ 2110 } 2111 VUPKPX(lpx, UPKLO) 2112 VUPKPX(hpx, UPKHI) 2113 #undef VUPKPX 2114 2115 #define VUPK(suffix, unpacked, packee, hi) \ 2116 void helper_vupk##suffix(ppc_avr_t *r, ppc_avr_t *b) \ 2117 { \ 2118 int i; \ 2119 ppc_avr_t result; \ 2120 \ 2121 if (hi) { \ 2122 for (i = 0; i < ARRAY_SIZE(r->unpacked); i++) { \ 2123 result.unpacked[i] = b->packee[i]; \ 2124 } \ 2125 } else { \ 2126 for (i = ARRAY_SIZE(r->unpacked); i < ARRAY_SIZE(r->packee); \ 2127 i++) { \ 2128 result.unpacked[i - ARRAY_SIZE(r->unpacked)] = b->packee[i]; \ 2129 } \ 2130 } \ 2131 *r = result; \ 2132 } 2133 VUPK(hsb, s16, s8, UPKHI) 2134 VUPK(hsh, s32, s16, UPKHI) 2135 VUPK(hsw, s64, s32, UPKHI) 2136 VUPK(lsb, s16, s8, UPKLO) 2137 VUPK(lsh, s32, s16, UPKLO) 2138 VUPK(lsw, s64, s32, UPKLO) 2139 #undef VUPK 2140 #undef UPKHI 2141 #undef UPKLO 2142 2143 #define VGENERIC_DO(name, element) \ 2144 void helper_v##name(ppc_avr_t *r, ppc_avr_t *b) \ 2145 { \ 2146 int i; \ 2147 \ 2148 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ 2149 r->element[i] = name(b->element[i]); \ 2150 } \ 2151 } 2152 2153 #define clzb(v) ((v) ? clz32((uint32_t)(v) << 24) : 8) 2154 #define clzh(v) ((v) ? clz32((uint32_t)(v) << 16) : 16) 2155 #define clzw(v) clz32((v)) 2156 #define clzd(v) clz64((v)) 2157 2158 VGENERIC_DO(clzb, u8) 2159 VGENERIC_DO(clzh, u16) 2160 VGENERIC_DO(clzw, u32) 2161 VGENERIC_DO(clzd, u64) 2162 2163 #undef clzb 2164 #undef clzh 2165 #undef clzw 2166 #undef clzd 2167 2168 #define ctzb(v) ((v) ? ctz32(v) : 8) 2169 #define ctzh(v) ((v) ? ctz32(v) : 16) 2170 #define ctzw(v) ctz32((v)) 2171 #define ctzd(v) ctz64((v)) 2172 2173 VGENERIC_DO(ctzb, u8) 2174 VGENERIC_DO(ctzh, u16) 2175 VGENERIC_DO(ctzw, u32) 2176 VGENERIC_DO(ctzd, u64) 2177 2178 #undef ctzb 2179 #undef ctzh 2180 #undef ctzw 2181 #undef ctzd 2182 2183 #define popcntb(v) ctpop8(v) 2184 #define popcnth(v) ctpop16(v) 2185 #define popcntw(v) ctpop32(v) 2186 #define popcntd(v) ctpop64(v) 2187 2188 VGENERIC_DO(popcntb, u8) 2189 VGENERIC_DO(popcnth, u16) 2190 VGENERIC_DO(popcntw, u32) 2191 VGENERIC_DO(popcntd, u64) 2192 2193 #undef popcntb 2194 #undef popcnth 2195 #undef popcntw 2196 #undef popcntd 2197 2198 #undef VGENERIC_DO 2199 2200 #if defined(HOST_WORDS_BIGENDIAN) 2201 #define QW_ONE { .u64 = { 0, 1 } } 2202 #else 2203 #define QW_ONE { .u64 = { 1, 0 } } 2204 #endif 2205 2206 #ifndef CONFIG_INT128 2207 2208 static inline void avr_qw_not(ppc_avr_t *t, ppc_avr_t a) 2209 { 2210 t->u64[0] = ~a.u64[0]; 2211 t->u64[1] = ~a.u64[1]; 2212 } 2213 2214 static int avr_qw_cmpu(ppc_avr_t a, ppc_avr_t b) 2215 { 2216 if (a.VsrD(0) < b.VsrD(0)) { 2217 return -1; 2218 } else if (a.VsrD(0) > b.VsrD(0)) { 2219 return 1; 2220 } else if (a.VsrD(1) < b.VsrD(1)) { 2221 return -1; 2222 } else if (a.VsrD(1) > b.VsrD(1)) { 2223 return 1; 2224 } else { 2225 return 0; 2226 } 2227 } 2228 2229 static void avr_qw_add(ppc_avr_t *t, ppc_avr_t a, ppc_avr_t b) 2230 { 2231 t->VsrD(1) = a.VsrD(1) + b.VsrD(1); 2232 t->VsrD(0) = a.VsrD(0) + b.VsrD(0) + 2233 (~a.VsrD(1) < b.VsrD(1)); 2234 } 2235 2236 static int avr_qw_addc(ppc_avr_t *t, ppc_avr_t a, ppc_avr_t b) 2237 { 2238 ppc_avr_t not_a; 2239 t->VsrD(1) = a.VsrD(1) + b.VsrD(1); 2240 t->VsrD(0) = a.VsrD(0) + b.VsrD(0) + 2241 (~a.VsrD(1) < b.VsrD(1)); 2242 avr_qw_not(¬_a, a); 2243 return avr_qw_cmpu(not_a, b) < 0; 2244 } 2245 2246 #endif 2247 2248 void helper_vadduqm(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 2249 { 2250 #ifdef CONFIG_INT128 2251 r->u128 = a->u128 + b->u128; 2252 #else 2253 avr_qw_add(r, *a, *b); 2254 #endif 2255 } 2256 2257 void helper_vaddeuqm(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c) 2258 { 2259 #ifdef CONFIG_INT128 2260 r->u128 = a->u128 + b->u128 + (c->u128 & 1); 2261 #else 2262 2263 if (c->VsrD(1) & 1) { 2264 ppc_avr_t tmp; 2265 2266 tmp.VsrD(0) = 0; 2267 tmp.VsrD(1) = c->VsrD(1) & 1; 2268 avr_qw_add(&tmp, *a, tmp); 2269 avr_qw_add(r, tmp, *b); 2270 } else { 2271 avr_qw_add(r, *a, *b); 2272 } 2273 #endif 2274 } 2275 2276 void helper_vaddcuq(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 2277 { 2278 #ifdef CONFIG_INT128 2279 r->u128 = (~a->u128 < b->u128); 2280 #else 2281 ppc_avr_t not_a; 2282 2283 avr_qw_not(¬_a, *a); 2284 2285 r->VsrD(0) = 0; 2286 r->VsrD(1) = (avr_qw_cmpu(not_a, *b) < 0); 2287 #endif 2288 } 2289 2290 void helper_vaddecuq(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c) 2291 { 2292 #ifdef CONFIG_INT128 2293 int carry_out = (~a->u128 < b->u128); 2294 if (!carry_out && (c->u128 & 1)) { 2295 carry_out = ((a->u128 + b->u128 + 1) == 0) && 2296 ((a->u128 != 0) || (b->u128 != 0)); 2297 } 2298 r->u128 = carry_out; 2299 #else 2300 2301 int carry_in = c->VsrD(1) & 1; 2302 int carry_out = 0; 2303 ppc_avr_t tmp; 2304 2305 carry_out = avr_qw_addc(&tmp, *a, *b); 2306 2307 if (!carry_out && carry_in) { 2308 ppc_avr_t one = QW_ONE; 2309 carry_out = avr_qw_addc(&tmp, tmp, one); 2310 } 2311 r->VsrD(0) = 0; 2312 r->VsrD(1) = carry_out; 2313 #endif 2314 } 2315 2316 void helper_vsubuqm(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 2317 { 2318 #ifdef CONFIG_INT128 2319 r->u128 = a->u128 - b->u128; 2320 #else 2321 ppc_avr_t tmp; 2322 ppc_avr_t one = QW_ONE; 2323 2324 avr_qw_not(&tmp, *b); 2325 avr_qw_add(&tmp, *a, tmp); 2326 avr_qw_add(r, tmp, one); 2327 #endif 2328 } 2329 2330 void helper_vsubeuqm(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c) 2331 { 2332 #ifdef CONFIG_INT128 2333 r->u128 = a->u128 + ~b->u128 + (c->u128 & 1); 2334 #else 2335 ppc_avr_t tmp, sum; 2336 2337 avr_qw_not(&tmp, *b); 2338 avr_qw_add(&sum, *a, tmp); 2339 2340 tmp.VsrD(0) = 0; 2341 tmp.VsrD(1) = c->VsrD(1) & 1; 2342 avr_qw_add(r, sum, tmp); 2343 #endif 2344 } 2345 2346 void helper_vsubcuq(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 2347 { 2348 #ifdef CONFIG_INT128 2349 r->u128 = (~a->u128 < ~b->u128) || 2350 (a->u128 + ~b->u128 == (__uint128_t)-1); 2351 #else 2352 int carry = (avr_qw_cmpu(*a, *b) > 0); 2353 if (!carry) { 2354 ppc_avr_t tmp; 2355 avr_qw_not(&tmp, *b); 2356 avr_qw_add(&tmp, *a, tmp); 2357 carry = ((tmp.VsrSD(0) == -1ull) && (tmp.VsrSD(1) == -1ull)); 2358 } 2359 r->VsrD(0) = 0; 2360 r->VsrD(1) = carry; 2361 #endif 2362 } 2363 2364 void helper_vsubecuq(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c) 2365 { 2366 #ifdef CONFIG_INT128 2367 r->u128 = 2368 (~a->u128 < ~b->u128) || 2369 ((c->u128 & 1) && (a->u128 + ~b->u128 == (__uint128_t)-1)); 2370 #else 2371 int carry_in = c->VsrD(1) & 1; 2372 int carry_out = (avr_qw_cmpu(*a, *b) > 0); 2373 if (!carry_out && carry_in) { 2374 ppc_avr_t tmp; 2375 avr_qw_not(&tmp, *b); 2376 avr_qw_add(&tmp, *a, tmp); 2377 carry_out = ((tmp.VsrD(0) == -1ull) && (tmp.VsrD(1) == -1ull)); 2378 } 2379 2380 r->VsrD(0) = 0; 2381 r->VsrD(1) = carry_out; 2382 #endif 2383 } 2384 2385 #define BCD_PLUS_PREF_1 0xC 2386 #define BCD_PLUS_PREF_2 0xF 2387 #define BCD_PLUS_ALT_1 0xA 2388 #define BCD_NEG_PREF 0xD 2389 #define BCD_NEG_ALT 0xB 2390 #define BCD_PLUS_ALT_2 0xE 2391 #define NATIONAL_PLUS 0x2B 2392 #define NATIONAL_NEG 0x2D 2393 2394 #if defined(HOST_WORDS_BIGENDIAN) 2395 #define BCD_DIG_BYTE(n) (15 - ((n) / 2)) 2396 #else 2397 #define BCD_DIG_BYTE(n) ((n) / 2) 2398 #endif 2399 2400 static int bcd_get_sgn(ppc_avr_t *bcd) 2401 { 2402 switch (bcd->u8[BCD_DIG_BYTE(0)] & 0xF) { 2403 case BCD_PLUS_PREF_1: 2404 case BCD_PLUS_PREF_2: 2405 case BCD_PLUS_ALT_1: 2406 case BCD_PLUS_ALT_2: 2407 { 2408 return 1; 2409 } 2410 2411 case BCD_NEG_PREF: 2412 case BCD_NEG_ALT: 2413 { 2414 return -1; 2415 } 2416 2417 default: 2418 { 2419 return 0; 2420 } 2421 } 2422 } 2423 2424 static int bcd_preferred_sgn(int sgn, int ps) 2425 { 2426 if (sgn >= 0) { 2427 return (ps == 0) ? BCD_PLUS_PREF_1 : BCD_PLUS_PREF_2; 2428 } else { 2429 return BCD_NEG_PREF; 2430 } 2431 } 2432 2433 static uint8_t bcd_get_digit(ppc_avr_t *bcd, int n, int *invalid) 2434 { 2435 uint8_t result; 2436 if (n & 1) { 2437 result = bcd->u8[BCD_DIG_BYTE(n)] >> 4; 2438 } else { 2439 result = bcd->u8[BCD_DIG_BYTE(n)] & 0xF; 2440 } 2441 2442 if (unlikely(result > 9)) { 2443 *invalid = true; 2444 } 2445 return result; 2446 } 2447 2448 static void bcd_put_digit(ppc_avr_t *bcd, uint8_t digit, int n) 2449 { 2450 if (n & 1) { 2451 bcd->u8[BCD_DIG_BYTE(n)] &= 0x0F; 2452 bcd->u8[BCD_DIG_BYTE(n)] |= (digit << 4); 2453 } else { 2454 bcd->u8[BCD_DIG_BYTE(n)] &= 0xF0; 2455 bcd->u8[BCD_DIG_BYTE(n)] |= digit; 2456 } 2457 } 2458 2459 static bool bcd_is_valid(ppc_avr_t *bcd) 2460 { 2461 int i; 2462 int invalid = 0; 2463 2464 if (bcd_get_sgn(bcd) == 0) { 2465 return false; 2466 } 2467 2468 for (i = 1; i < 32; i++) { 2469 bcd_get_digit(bcd, i, &invalid); 2470 if (unlikely(invalid)) { 2471 return false; 2472 } 2473 } 2474 return true; 2475 } 2476 2477 static int bcd_cmp_zero(ppc_avr_t *bcd) 2478 { 2479 if (bcd->VsrD(0) == 0 && (bcd->VsrD(1) >> 4) == 0) { 2480 return CRF_EQ; 2481 } else { 2482 return (bcd_get_sgn(bcd) == 1) ? CRF_GT : CRF_LT; 2483 } 2484 } 2485 2486 static uint16_t get_national_digit(ppc_avr_t *reg, int n) 2487 { 2488 return reg->VsrH(7 - n); 2489 } 2490 2491 static void set_national_digit(ppc_avr_t *reg, uint8_t val, int n) 2492 { 2493 reg->VsrH(7 - n) = val; 2494 } 2495 2496 static int bcd_cmp_mag(ppc_avr_t *a, ppc_avr_t *b) 2497 { 2498 int i; 2499 int invalid = 0; 2500 for (i = 31; i > 0; i--) { 2501 uint8_t dig_a = bcd_get_digit(a, i, &invalid); 2502 uint8_t dig_b = bcd_get_digit(b, i, &invalid); 2503 if (unlikely(invalid)) { 2504 return 0; /* doesn't matter */ 2505 } else if (dig_a > dig_b) { 2506 return 1; 2507 } else if (dig_a < dig_b) { 2508 return -1; 2509 } 2510 } 2511 2512 return 0; 2513 } 2514 2515 static void bcd_add_mag(ppc_avr_t *t, ppc_avr_t *a, ppc_avr_t *b, int *invalid, 2516 int *overflow) 2517 { 2518 int carry = 0; 2519 int i; 2520 for (i = 1; i <= 31; i++) { 2521 uint8_t digit = bcd_get_digit(a, i, invalid) + 2522 bcd_get_digit(b, i, invalid) + carry; 2523 if (digit > 9) { 2524 carry = 1; 2525 digit -= 10; 2526 } else { 2527 carry = 0; 2528 } 2529 2530 bcd_put_digit(t, digit, i); 2531 } 2532 2533 *overflow = carry; 2534 } 2535 2536 static void bcd_sub_mag(ppc_avr_t *t, ppc_avr_t *a, ppc_avr_t *b, int *invalid, 2537 int *overflow) 2538 { 2539 int carry = 0; 2540 int i; 2541 2542 for (i = 1; i <= 31; i++) { 2543 uint8_t digit = bcd_get_digit(a, i, invalid) - 2544 bcd_get_digit(b, i, invalid) + carry; 2545 if (digit & 0x80) { 2546 carry = -1; 2547 digit += 10; 2548 } else { 2549 carry = 0; 2550 } 2551 2552 bcd_put_digit(t, digit, i); 2553 } 2554 2555 *overflow = carry; 2556 } 2557 2558 uint32_t helper_bcdadd(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t ps) 2559 { 2560 2561 int sgna = bcd_get_sgn(a); 2562 int sgnb = bcd_get_sgn(b); 2563 int invalid = (sgna == 0) || (sgnb == 0); 2564 int overflow = 0; 2565 uint32_t cr = 0; 2566 ppc_avr_t result = { .u64 = { 0, 0 } }; 2567 2568 if (!invalid) { 2569 if (sgna == sgnb) { 2570 result.u8[BCD_DIG_BYTE(0)] = bcd_preferred_sgn(sgna, ps); 2571 bcd_add_mag(&result, a, b, &invalid, &overflow); 2572 cr = bcd_cmp_zero(&result); 2573 } else { 2574 int magnitude = bcd_cmp_mag(a, b); 2575 if (magnitude > 0) { 2576 result.u8[BCD_DIG_BYTE(0)] = bcd_preferred_sgn(sgna, ps); 2577 bcd_sub_mag(&result, a, b, &invalid, &overflow); 2578 cr = (sgna > 0) ? CRF_GT : CRF_LT; 2579 } else if (magnitude < 0) { 2580 result.u8[BCD_DIG_BYTE(0)] = bcd_preferred_sgn(sgnb, ps); 2581 bcd_sub_mag(&result, b, a, &invalid, &overflow); 2582 cr = (sgnb > 0) ? CRF_GT : CRF_LT; 2583 } else { 2584 result.u8[BCD_DIG_BYTE(0)] = bcd_preferred_sgn(0, ps); 2585 cr = CRF_EQ; 2586 } 2587 } 2588 } 2589 2590 if (unlikely(invalid)) { 2591 result.VsrD(0) = result.VsrD(1) = -1; 2592 cr = CRF_SO; 2593 } else if (overflow) { 2594 cr |= CRF_SO; 2595 } 2596 2597 *r = result; 2598 2599 return cr; 2600 } 2601 2602 uint32_t helper_bcdsub(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t ps) 2603 { 2604 ppc_avr_t bcopy = *b; 2605 int sgnb = bcd_get_sgn(b); 2606 if (sgnb < 0) { 2607 bcd_put_digit(&bcopy, BCD_PLUS_PREF_1, 0); 2608 } else if (sgnb > 0) { 2609 bcd_put_digit(&bcopy, BCD_NEG_PREF, 0); 2610 } 2611 /* else invalid ... defer to bcdadd code for proper handling */ 2612 2613 return helper_bcdadd(r, a, &bcopy, ps); 2614 } 2615 2616 uint32_t helper_bcdcfn(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps) 2617 { 2618 int i; 2619 int cr = 0; 2620 uint16_t national = 0; 2621 uint16_t sgnb = get_national_digit(b, 0); 2622 ppc_avr_t ret = { .u64 = { 0, 0 } }; 2623 int invalid = (sgnb != NATIONAL_PLUS && sgnb != NATIONAL_NEG); 2624 2625 for (i = 1; i < 8; i++) { 2626 national = get_national_digit(b, i); 2627 if (unlikely(national < 0x30 || national > 0x39)) { 2628 invalid = 1; 2629 break; 2630 } 2631 2632 bcd_put_digit(&ret, national & 0xf, i); 2633 } 2634 2635 if (sgnb == NATIONAL_PLUS) { 2636 bcd_put_digit(&ret, (ps == 0) ? BCD_PLUS_PREF_1 : BCD_PLUS_PREF_2, 0); 2637 } else { 2638 bcd_put_digit(&ret, BCD_NEG_PREF, 0); 2639 } 2640 2641 cr = bcd_cmp_zero(&ret); 2642 2643 if (unlikely(invalid)) { 2644 cr = CRF_SO; 2645 } 2646 2647 *r = ret; 2648 2649 return cr; 2650 } 2651 2652 uint32_t helper_bcdctn(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps) 2653 { 2654 int i; 2655 int cr = 0; 2656 int sgnb = bcd_get_sgn(b); 2657 int invalid = (sgnb == 0); 2658 ppc_avr_t ret = { .u64 = { 0, 0 } }; 2659 2660 int ox_flag = (b->VsrD(0) != 0) || ((b->VsrD(1) >> 32) != 0); 2661 2662 for (i = 1; i < 8; i++) { 2663 set_national_digit(&ret, 0x30 + bcd_get_digit(b, i, &invalid), i); 2664 2665 if (unlikely(invalid)) { 2666 break; 2667 } 2668 } 2669 set_national_digit(&ret, (sgnb == -1) ? NATIONAL_NEG : NATIONAL_PLUS, 0); 2670 2671 cr = bcd_cmp_zero(b); 2672 2673 if (ox_flag) { 2674 cr |= CRF_SO; 2675 } 2676 2677 if (unlikely(invalid)) { 2678 cr = CRF_SO; 2679 } 2680 2681 *r = ret; 2682 2683 return cr; 2684 } 2685 2686 uint32_t helper_bcdcfz(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps) 2687 { 2688 int i; 2689 int cr = 0; 2690 int invalid = 0; 2691 int zone_digit = 0; 2692 int zone_lead = ps ? 0xF : 0x3; 2693 int digit = 0; 2694 ppc_avr_t ret = { .u64 = { 0, 0 } }; 2695 int sgnb = b->u8[BCD_DIG_BYTE(0)] >> 4; 2696 2697 if (unlikely((sgnb < 0xA) && ps)) { 2698 invalid = 1; 2699 } 2700 2701 for (i = 0; i < 16; i++) { 2702 zone_digit = i ? b->u8[BCD_DIG_BYTE(i * 2)] >> 4 : zone_lead; 2703 digit = b->u8[BCD_DIG_BYTE(i * 2)] & 0xF; 2704 if (unlikely(zone_digit != zone_lead || digit > 0x9)) { 2705 invalid = 1; 2706 break; 2707 } 2708 2709 bcd_put_digit(&ret, digit, i + 1); 2710 } 2711 2712 if ((ps && (sgnb == 0xB || sgnb == 0xD)) || 2713 (!ps && (sgnb & 0x4))) { 2714 bcd_put_digit(&ret, BCD_NEG_PREF, 0); 2715 } else { 2716 bcd_put_digit(&ret, BCD_PLUS_PREF_1, 0); 2717 } 2718 2719 cr = bcd_cmp_zero(&ret); 2720 2721 if (unlikely(invalid)) { 2722 cr = CRF_SO; 2723 } 2724 2725 *r = ret; 2726 2727 return cr; 2728 } 2729 2730 uint32_t helper_bcdctz(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps) 2731 { 2732 int i; 2733 int cr = 0; 2734 uint8_t digit = 0; 2735 int sgnb = bcd_get_sgn(b); 2736 int zone_lead = (ps) ? 0xF0 : 0x30; 2737 int invalid = (sgnb == 0); 2738 ppc_avr_t ret = { .u64 = { 0, 0 } }; 2739 2740 int ox_flag = ((b->VsrD(0) >> 4) != 0); 2741 2742 for (i = 0; i < 16; i++) { 2743 digit = bcd_get_digit(b, i + 1, &invalid); 2744 2745 if (unlikely(invalid)) { 2746 break; 2747 } 2748 2749 ret.u8[BCD_DIG_BYTE(i * 2)] = zone_lead + digit; 2750 } 2751 2752 if (ps) { 2753 bcd_put_digit(&ret, (sgnb == 1) ? 0xC : 0xD, 1); 2754 } else { 2755 bcd_put_digit(&ret, (sgnb == 1) ? 0x3 : 0x7, 1); 2756 } 2757 2758 cr = bcd_cmp_zero(b); 2759 2760 if (ox_flag) { 2761 cr |= CRF_SO; 2762 } 2763 2764 if (unlikely(invalid)) { 2765 cr = CRF_SO; 2766 } 2767 2768 *r = ret; 2769 2770 return cr; 2771 } 2772 2773 uint32_t helper_bcdcfsq(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps) 2774 { 2775 int i; 2776 int cr = 0; 2777 uint64_t lo_value; 2778 uint64_t hi_value; 2779 ppc_avr_t ret = { .u64 = { 0, 0 } }; 2780 2781 if (b->VsrSD(0) < 0) { 2782 lo_value = -b->VsrSD(1); 2783 hi_value = ~b->VsrD(0) + !lo_value; 2784 bcd_put_digit(&ret, 0xD, 0); 2785 } else { 2786 lo_value = b->VsrD(1); 2787 hi_value = b->VsrD(0); 2788 bcd_put_digit(&ret, bcd_preferred_sgn(0, ps), 0); 2789 } 2790 2791 if (divu128(&lo_value, &hi_value, 1000000000000000ULL) || 2792 lo_value > 9999999999999999ULL) { 2793 cr = CRF_SO; 2794 } 2795 2796 for (i = 1; i < 16; hi_value /= 10, i++) { 2797 bcd_put_digit(&ret, hi_value % 10, i); 2798 } 2799 2800 for (; i < 32; lo_value /= 10, i++) { 2801 bcd_put_digit(&ret, lo_value % 10, i); 2802 } 2803 2804 cr |= bcd_cmp_zero(&ret); 2805 2806 *r = ret; 2807 2808 return cr; 2809 } 2810 2811 uint32_t helper_bcdctsq(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps) 2812 { 2813 uint8_t i; 2814 int cr; 2815 uint64_t carry; 2816 uint64_t unused; 2817 uint64_t lo_value; 2818 uint64_t hi_value = 0; 2819 int sgnb = bcd_get_sgn(b); 2820 int invalid = (sgnb == 0); 2821 2822 lo_value = bcd_get_digit(b, 31, &invalid); 2823 for (i = 30; i > 0; i--) { 2824 mulu64(&lo_value, &carry, lo_value, 10ULL); 2825 mulu64(&hi_value, &unused, hi_value, 10ULL); 2826 lo_value += bcd_get_digit(b, i, &invalid); 2827 hi_value += carry; 2828 2829 if (unlikely(invalid)) { 2830 break; 2831 } 2832 } 2833 2834 if (sgnb == -1) { 2835 r->VsrSD(1) = -lo_value; 2836 r->VsrSD(0) = ~hi_value + !r->VsrSD(1); 2837 } else { 2838 r->VsrSD(1) = lo_value; 2839 r->VsrSD(0) = hi_value; 2840 } 2841 2842 cr = bcd_cmp_zero(b); 2843 2844 if (unlikely(invalid)) { 2845 cr = CRF_SO; 2846 } 2847 2848 return cr; 2849 } 2850 2851 uint32_t helper_bcdcpsgn(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t ps) 2852 { 2853 int i; 2854 int invalid = 0; 2855 2856 if (bcd_get_sgn(a) == 0 || bcd_get_sgn(b) == 0) { 2857 return CRF_SO; 2858 } 2859 2860 *r = *a; 2861 bcd_put_digit(r, b->u8[BCD_DIG_BYTE(0)] & 0xF, 0); 2862 2863 for (i = 1; i < 32; i++) { 2864 bcd_get_digit(a, i, &invalid); 2865 bcd_get_digit(b, i, &invalid); 2866 if (unlikely(invalid)) { 2867 return CRF_SO; 2868 } 2869 } 2870 2871 return bcd_cmp_zero(r); 2872 } 2873 2874 uint32_t helper_bcdsetsgn(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps) 2875 { 2876 int sgnb = bcd_get_sgn(b); 2877 2878 *r = *b; 2879 bcd_put_digit(r, bcd_preferred_sgn(sgnb, ps), 0); 2880 2881 if (bcd_is_valid(b) == false) { 2882 return CRF_SO; 2883 } 2884 2885 return bcd_cmp_zero(r); 2886 } 2887 2888 uint32_t helper_bcds(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t ps) 2889 { 2890 int cr; 2891 #if defined(HOST_WORDS_BIGENDIAN) 2892 int i = a->s8[7]; 2893 #else 2894 int i = a->s8[8]; 2895 #endif 2896 bool ox_flag = false; 2897 int sgnb = bcd_get_sgn(b); 2898 ppc_avr_t ret = *b; 2899 ret.VsrD(1) &= ~0xf; 2900 2901 if (bcd_is_valid(b) == false) { 2902 return CRF_SO; 2903 } 2904 2905 if (unlikely(i > 31)) { 2906 i = 31; 2907 } else if (unlikely(i < -31)) { 2908 i = -31; 2909 } 2910 2911 if (i > 0) { 2912 ulshift(&ret.VsrD(1), &ret.VsrD(0), i * 4, &ox_flag); 2913 } else { 2914 urshift(&ret.VsrD(1), &ret.VsrD(0), -i * 4); 2915 } 2916 bcd_put_digit(&ret, bcd_preferred_sgn(sgnb, ps), 0); 2917 2918 *r = ret; 2919 2920 cr = bcd_cmp_zero(r); 2921 if (ox_flag) { 2922 cr |= CRF_SO; 2923 } 2924 2925 return cr; 2926 } 2927 2928 uint32_t helper_bcdus(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t ps) 2929 { 2930 int cr; 2931 int i; 2932 int invalid = 0; 2933 bool ox_flag = false; 2934 ppc_avr_t ret = *b; 2935 2936 for (i = 0; i < 32; i++) { 2937 bcd_get_digit(b, i, &invalid); 2938 2939 if (unlikely(invalid)) { 2940 return CRF_SO; 2941 } 2942 } 2943 2944 #if defined(HOST_WORDS_BIGENDIAN) 2945 i = a->s8[7]; 2946 #else 2947 i = a->s8[8]; 2948 #endif 2949 if (i >= 32) { 2950 ox_flag = true; 2951 ret.VsrD(1) = ret.VsrD(0) = 0; 2952 } else if (i <= -32) { 2953 ret.VsrD(1) = ret.VsrD(0) = 0; 2954 } else if (i > 0) { 2955 ulshift(&ret.VsrD(1), &ret.VsrD(0), i * 4, &ox_flag); 2956 } else { 2957 urshift(&ret.VsrD(1), &ret.VsrD(0), -i * 4); 2958 } 2959 *r = ret; 2960 2961 cr = bcd_cmp_zero(r); 2962 if (ox_flag) { 2963 cr |= CRF_SO; 2964 } 2965 2966 return cr; 2967 } 2968 2969 uint32_t helper_bcdsr(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t ps) 2970 { 2971 int cr; 2972 int unused = 0; 2973 int invalid = 0; 2974 bool ox_flag = false; 2975 int sgnb = bcd_get_sgn(b); 2976 ppc_avr_t ret = *b; 2977 ret.VsrD(1) &= ~0xf; 2978 2979 #if defined(HOST_WORDS_BIGENDIAN) 2980 int i = a->s8[7]; 2981 ppc_avr_t bcd_one = { .u64 = { 0, 0x10 } }; 2982 #else 2983 int i = a->s8[8]; 2984 ppc_avr_t bcd_one = { .u64 = { 0x10, 0 } }; 2985 #endif 2986 2987 if (bcd_is_valid(b) == false) { 2988 return CRF_SO; 2989 } 2990 2991 if (unlikely(i > 31)) { 2992 i = 31; 2993 } else if (unlikely(i < -31)) { 2994 i = -31; 2995 } 2996 2997 if (i > 0) { 2998 ulshift(&ret.VsrD(1), &ret.VsrD(0), i * 4, &ox_flag); 2999 } else { 3000 urshift(&ret.VsrD(1), &ret.VsrD(0), -i * 4); 3001 3002 if (bcd_get_digit(&ret, 0, &invalid) >= 5) { 3003 bcd_add_mag(&ret, &ret, &bcd_one, &invalid, &unused); 3004 } 3005 } 3006 bcd_put_digit(&ret, bcd_preferred_sgn(sgnb, ps), 0); 3007 3008 cr = bcd_cmp_zero(&ret); 3009 if (ox_flag) { 3010 cr |= CRF_SO; 3011 } 3012 *r = ret; 3013 3014 return cr; 3015 } 3016 3017 uint32_t helper_bcdtrunc(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t ps) 3018 { 3019 uint64_t mask; 3020 uint32_t ox_flag = 0; 3021 #if defined(HOST_WORDS_BIGENDIAN) 3022 int i = a->s16[3] + 1; 3023 #else 3024 int i = a->s16[4] + 1; 3025 #endif 3026 ppc_avr_t ret = *b; 3027 3028 if (bcd_is_valid(b) == false) { 3029 return CRF_SO; 3030 } 3031 3032 if (i > 16 && i < 32) { 3033 mask = (uint64_t)-1 >> (128 - i * 4); 3034 if (ret.VsrD(0) & ~mask) { 3035 ox_flag = CRF_SO; 3036 } 3037 3038 ret.VsrD(0) &= mask; 3039 } else if (i >= 0 && i <= 16) { 3040 mask = (uint64_t)-1 >> (64 - i * 4); 3041 if (ret.VsrD(0) || (ret.VsrD(1) & ~mask)) { 3042 ox_flag = CRF_SO; 3043 } 3044 3045 ret.VsrD(1) &= mask; 3046 ret.VsrD(0) = 0; 3047 } 3048 bcd_put_digit(&ret, bcd_preferred_sgn(bcd_get_sgn(b), ps), 0); 3049 *r = ret; 3050 3051 return bcd_cmp_zero(&ret) | ox_flag; 3052 } 3053 3054 uint32_t helper_bcdutrunc(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t ps) 3055 { 3056 int i; 3057 uint64_t mask; 3058 uint32_t ox_flag = 0; 3059 int invalid = 0; 3060 ppc_avr_t ret = *b; 3061 3062 for (i = 0; i < 32; i++) { 3063 bcd_get_digit(b, i, &invalid); 3064 3065 if (unlikely(invalid)) { 3066 return CRF_SO; 3067 } 3068 } 3069 3070 #if defined(HOST_WORDS_BIGENDIAN) 3071 i = a->s16[3]; 3072 #else 3073 i = a->s16[4]; 3074 #endif 3075 if (i > 16 && i < 33) { 3076 mask = (uint64_t)-1 >> (128 - i * 4); 3077 if (ret.VsrD(0) & ~mask) { 3078 ox_flag = CRF_SO; 3079 } 3080 3081 ret.VsrD(0) &= mask; 3082 } else if (i > 0 && i <= 16) { 3083 mask = (uint64_t)-1 >> (64 - i * 4); 3084 if (ret.VsrD(0) || (ret.VsrD(1) & ~mask)) { 3085 ox_flag = CRF_SO; 3086 } 3087 3088 ret.VsrD(1) &= mask; 3089 ret.VsrD(0) = 0; 3090 } else if (i == 0) { 3091 if (ret.VsrD(0) || ret.VsrD(1)) { 3092 ox_flag = CRF_SO; 3093 } 3094 ret.VsrD(0) = ret.VsrD(1) = 0; 3095 } 3096 3097 *r = ret; 3098 if (r->VsrD(0) == 0 && r->VsrD(1) == 0) { 3099 return ox_flag | CRF_EQ; 3100 } 3101 3102 return ox_flag | CRF_GT; 3103 } 3104 3105 void helper_vsbox(ppc_avr_t *r, ppc_avr_t *a) 3106 { 3107 int i; 3108 VECTOR_FOR_INORDER_I(i, u8) { 3109 r->u8[i] = AES_sbox[a->u8[i]]; 3110 } 3111 } 3112 3113 void helper_vcipher(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 3114 { 3115 ppc_avr_t result; 3116 int i; 3117 3118 VECTOR_FOR_INORDER_I(i, u32) { 3119 result.VsrW(i) = b->VsrW(i) ^ 3120 (AES_Te0[a->VsrB(AES_shifts[4 * i + 0])] ^ 3121 AES_Te1[a->VsrB(AES_shifts[4 * i + 1])] ^ 3122 AES_Te2[a->VsrB(AES_shifts[4 * i + 2])] ^ 3123 AES_Te3[a->VsrB(AES_shifts[4 * i + 3])]); 3124 } 3125 *r = result; 3126 } 3127 3128 void helper_vcipherlast(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 3129 { 3130 ppc_avr_t result; 3131 int i; 3132 3133 VECTOR_FOR_INORDER_I(i, u8) { 3134 result.VsrB(i) = b->VsrB(i) ^ (AES_sbox[a->VsrB(AES_shifts[i])]); 3135 } 3136 *r = result; 3137 } 3138 3139 void helper_vncipher(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 3140 { 3141 /* This differs from what is written in ISA V2.07. The RTL is */ 3142 /* incorrect and will be fixed in V2.07B. */ 3143 int i; 3144 ppc_avr_t tmp; 3145 3146 VECTOR_FOR_INORDER_I(i, u8) { 3147 tmp.VsrB(i) = b->VsrB(i) ^ AES_isbox[a->VsrB(AES_ishifts[i])]; 3148 } 3149 3150 VECTOR_FOR_INORDER_I(i, u32) { 3151 r->VsrW(i) = 3152 AES_imc[tmp.VsrB(4 * i + 0)][0] ^ 3153 AES_imc[tmp.VsrB(4 * i + 1)][1] ^ 3154 AES_imc[tmp.VsrB(4 * i + 2)][2] ^ 3155 AES_imc[tmp.VsrB(4 * i + 3)][3]; 3156 } 3157 } 3158 3159 void helper_vncipherlast(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) 3160 { 3161 ppc_avr_t result; 3162 int i; 3163 3164 VECTOR_FOR_INORDER_I(i, u8) { 3165 result.VsrB(i) = b->VsrB(i) ^ (AES_isbox[a->VsrB(AES_ishifts[i])]); 3166 } 3167 *r = result; 3168 } 3169 3170 void helper_vshasigmaw(ppc_avr_t *r, ppc_avr_t *a, uint32_t st_six) 3171 { 3172 int st = (st_six & 0x10) != 0; 3173 int six = st_six & 0xF; 3174 int i; 3175 3176 for (i = 0; i < ARRAY_SIZE(r->u32); i++) { 3177 if (st == 0) { 3178 if ((six & (0x8 >> i)) == 0) { 3179 r->VsrW(i) = ror32(a->VsrW(i), 7) ^ 3180 ror32(a->VsrW(i), 18) ^ 3181 (a->VsrW(i) >> 3); 3182 } else { /* six.bit[i] == 1 */ 3183 r->VsrW(i) = ror32(a->VsrW(i), 17) ^ 3184 ror32(a->VsrW(i), 19) ^ 3185 (a->VsrW(i) >> 10); 3186 } 3187 } else { /* st == 1 */ 3188 if ((six & (0x8 >> i)) == 0) { 3189 r->VsrW(i) = ror32(a->VsrW(i), 2) ^ 3190 ror32(a->VsrW(i), 13) ^ 3191 ror32(a->VsrW(i), 22); 3192 } else { /* six.bit[i] == 1 */ 3193 r->VsrW(i) = ror32(a->VsrW(i), 6) ^ 3194 ror32(a->VsrW(i), 11) ^ 3195 ror32(a->VsrW(i), 25); 3196 } 3197 } 3198 } 3199 } 3200 3201 void helper_vshasigmad(ppc_avr_t *r, ppc_avr_t *a, uint32_t st_six) 3202 { 3203 int st = (st_six & 0x10) != 0; 3204 int six = st_six & 0xF; 3205 int i; 3206 3207 for (i = 0; i < ARRAY_SIZE(r->u64); i++) { 3208 if (st == 0) { 3209 if ((six & (0x8 >> (2 * i))) == 0) { 3210 r->VsrD(i) = ror64(a->VsrD(i), 1) ^ 3211 ror64(a->VsrD(i), 8) ^ 3212 (a->VsrD(i) >> 7); 3213 } else { /* six.bit[2*i] == 1 */ 3214 r->VsrD(i) = ror64(a->VsrD(i), 19) ^ 3215 ror64(a->VsrD(i), 61) ^ 3216 (a->VsrD(i) >> 6); 3217 } 3218 } else { /* st == 1 */ 3219 if ((six & (0x8 >> (2 * i))) == 0) { 3220 r->VsrD(i) = ror64(a->VsrD(i), 28) ^ 3221 ror64(a->VsrD(i), 34) ^ 3222 ror64(a->VsrD(i), 39); 3223 } else { /* six.bit[2*i] == 1 */ 3224 r->VsrD(i) = ror64(a->VsrD(i), 14) ^ 3225 ror64(a->VsrD(i), 18) ^ 3226 ror64(a->VsrD(i), 41); 3227 } 3228 } 3229 } 3230 } 3231 3232 void helper_vpermxor(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c) 3233 { 3234 ppc_avr_t result; 3235 int i; 3236 3237 for (i = 0; i < ARRAY_SIZE(r->u8); i++) { 3238 int indexA = c->VsrB(i) >> 4; 3239 int indexB = c->VsrB(i) & 0xF; 3240 3241 result.VsrB(i) = a->VsrB(indexA) ^ b->VsrB(indexB); 3242 } 3243 *r = result; 3244 } 3245 3246 #undef VECTOR_FOR_INORDER_I 3247 3248 /*****************************************************************************/ 3249 /* SPE extension helpers */ 3250 /* Use a table to make this quicker */ 3251 static const uint8_t hbrev[16] = { 3252 0x0, 0x8, 0x4, 0xC, 0x2, 0xA, 0x6, 0xE, 3253 0x1, 0x9, 0x5, 0xD, 0x3, 0xB, 0x7, 0xF, 3254 }; 3255 3256 static inline uint8_t byte_reverse(uint8_t val) 3257 { 3258 return hbrev[val >> 4] | (hbrev[val & 0xF] << 4); 3259 } 3260 3261 static inline uint32_t word_reverse(uint32_t val) 3262 { 3263 return byte_reverse(val >> 24) | (byte_reverse(val >> 16) << 8) | 3264 (byte_reverse(val >> 8) << 16) | (byte_reverse(val) << 24); 3265 } 3266 3267 #define MASKBITS 16 /* Random value - to be fixed (implementation dependent) */ 3268 target_ulong helper_brinc(target_ulong arg1, target_ulong arg2) 3269 { 3270 uint32_t a, b, d, mask; 3271 3272 mask = UINT32_MAX >> (32 - MASKBITS); 3273 a = arg1 & mask; 3274 b = arg2 & mask; 3275 d = word_reverse(1 + word_reverse(a | ~b)); 3276 return (arg1 & ~mask) | (d & b); 3277 } 3278 3279 uint32_t helper_cntlsw32(uint32_t val) 3280 { 3281 if (val & 0x80000000) { 3282 return clz32(~val); 3283 } else { 3284 return clz32(val); 3285 } 3286 } 3287 3288 uint32_t helper_cntlzw32(uint32_t val) 3289 { 3290 return clz32(val); 3291 } 3292 3293 /* 440 specific */ 3294 target_ulong helper_dlmzb(CPUPPCState *env, target_ulong high, 3295 target_ulong low, uint32_t update_Rc) 3296 { 3297 target_ulong mask; 3298 int i; 3299 3300 i = 1; 3301 for (mask = 0xFF000000; mask != 0; mask = mask >> 8) { 3302 if ((high & mask) == 0) { 3303 if (update_Rc) { 3304 env->crf[0] = 0x4; 3305 } 3306 goto done; 3307 } 3308 i++; 3309 } 3310 for (mask = 0xFF000000; mask != 0; mask = mask >> 8) { 3311 if ((low & mask) == 0) { 3312 if (update_Rc) { 3313 env->crf[0] = 0x8; 3314 } 3315 goto done; 3316 } 3317 i++; 3318 } 3319 i = 8; 3320 if (update_Rc) { 3321 env->crf[0] = 0x2; 3322 } 3323 done: 3324 env->xer = (env->xer & ~0x7F) | i; 3325 if (update_Rc) { 3326 env->crf[0] |= xer_so; 3327 } 3328 return i; 3329 } 3330