1 /* 2 * ***************************************************************************** 3 * 4 * SPDX-License-Identifier: BSD-2-Clause 5 * 6 * Copyright (c) 2018-2024 Gavin D. Howard and contributors. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions are met: 10 * 11 * * Redistributions of source code must retain the above copyright notice, this 12 * list of conditions and the following disclaimer. 13 * 14 * * Redistributions in binary form must reproduce the above copyright notice, 15 * this list of conditions and the following disclaimer in the documentation 16 * and/or other materials provided with the distribution. 17 * 18 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 19 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 21 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE 22 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 23 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 24 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 25 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 26 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 27 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 28 * POSSIBILITY OF SUCH DAMAGE. 29 * 30 * ***************************************************************************** 31 * 32 * Code for the number type. 33 * 34 */ 35 36 #include <assert.h> 37 #include <ctype.h> 38 #include <stdbool.h> 39 #include <stdlib.h> 40 #include <string.h> 41 #include <setjmp.h> 42 #include <limits.h> 43 44 #include <num.h> 45 #include <rand.h> 46 #include <vm.h> 47 #if BC_ENABLE_LIBRARY 48 #include <library.h> 49 #endif // BC_ENABLE_LIBRARY 50 51 // Before you try to understand this code, see the development manual 52 // (manuals/development.md#numbers). 53 54 static void 55 bc_num_m(BcNum* a, BcNum* b, BcNum* restrict c, size_t scale); 56 57 /** 58 * Multiply two numbers and throw a math error if they overflow. 59 * @param a The first operand. 60 * @param b The second operand. 61 * @return The product of the two operands. 62 */ 63 static inline size_t 64 bc_num_mulOverflow(size_t a, size_t b) 65 { 66 size_t res = a * b; 67 if (BC_ERR(BC_VM_MUL_OVERFLOW(a, b, res))) bc_err(BC_ERR_MATH_OVERFLOW); 68 return res; 69 } 70 71 /** 72 * Conditionally negate @a n based on @a neg. Algorithm taken from 73 * https://graphics.stanford.edu/~seander/bithacks.html#ConditionalNegate . 74 * @param n The value to turn into a signed value and negate. 75 * @param neg The condition to negate or not. 76 */ 77 static inline ssize_t 78 bc_num_neg(size_t n, bool neg) 79 { 80 return (((ssize_t) n) ^ -((ssize_t) neg)) + neg; 81 } 82 83 /** 84 * Compare a BcNum against zero. 85 * @param n The number to compare. 86 * @return -1 if the number is less than 0, 1 if greater, and 0 if equal. 87 */ 88 ssize_t 89 bc_num_cmpZero(const BcNum* n) 90 { 91 return bc_num_neg((n)->len != 0, BC_NUM_NEG(n)); 92 } 93 94 /** 95 * Return the number of integer limbs in a BcNum. This is the opposite of rdx. 96 * @param n The number to return the amount of integer limbs for. 97 * @return The amount of integer limbs in @a n. 98 */ 99 static inline size_t 100 bc_num_int(const BcNum* n) 101 { 102 return n->len ? n->len - BC_NUM_RDX_VAL(n) : 0; 103 } 104 105 /** 106 * Expand a number's allocation capacity to at least req limbs. 107 * @param n The number to expand. 108 * @param req The number limbs to expand the allocation capacity to. 109 */ 110 static void 111 bc_num_expand(BcNum* restrict n, size_t req) 112 { 113 assert(n != NULL); 114 115 req = req >= BC_NUM_DEF_SIZE ? req : BC_NUM_DEF_SIZE; 116 117 if (req > n->cap) 118 { 119 BC_SIG_LOCK; 120 121 n->num = bc_vm_realloc(n->num, BC_NUM_SIZE(req)); 122 n->cap = req; 123 124 BC_SIG_UNLOCK; 125 } 126 } 127 128 /** 129 * Set a number to 0 with the specified scale. 130 * @param n The number to set to zero. 131 * @param scale The scale to set the number to. 132 */ 133 static inline void 134 bc_num_setToZero(BcNum* restrict n, size_t scale) 135 { 136 assert(n != NULL); 137 n->scale = scale; 138 n->len = n->rdx = 0; 139 } 140 141 void 142 bc_num_zero(BcNum* restrict n) 143 { 144 bc_num_setToZero(n, 0); 145 } 146 147 void 148 bc_num_one(BcNum* restrict n) 149 { 150 bc_num_zero(n); 151 n->len = 1; 152 n->num[0] = 1; 153 } 154 155 /** 156 * "Cleans" a number, which means reducing the length if the most significant 157 * limbs are zero. 158 * @param n The number to clean. 159 */ 160 static void 161 bc_num_clean(BcNum* restrict n) 162 { 163 // Reduce the length. 164 while (BC_NUM_NONZERO(n) && !n->num[n->len - 1]) 165 { 166 n->len -= 1; 167 } 168 169 // Special cases. 170 if (BC_NUM_ZERO(n)) n->rdx = 0; 171 else 172 { 173 // len must be at least as much as rdx. 174 size_t rdx = BC_NUM_RDX_VAL(n); 175 if (n->len < rdx) n->len = rdx; 176 } 177 } 178 179 /** 180 * Returns the log base 10 of @a i. I could have done this with floating-point 181 * math, and in fact, I originally did. However, that was the only 182 * floating-point code in the entire codebase, and I decided I didn't want any. 183 * This is fast enough. Also, it might handle larger numbers better. 184 * @param i The number to return the log base 10 of. 185 * @return The log base 10 of @a i. 186 */ 187 static size_t 188 bc_num_log10(size_t i) 189 { 190 size_t len; 191 192 for (len = 1; i; i /= BC_BASE, ++len) 193 { 194 continue; 195 } 196 197 assert(len - 1 <= BC_BASE_DIGS + 1); 198 199 return len - 1; 200 } 201 202 /** 203 * Returns the number of decimal digits in a limb that are zero starting at the 204 * most significant digits. This basically returns how much of the limb is used. 205 * @param n The number. 206 * @return The number of decimal digits that are 0 starting at the most 207 * significant digits. 208 */ 209 static inline size_t 210 bc_num_zeroDigits(const BcDig* n) 211 { 212 assert(*n >= 0); 213 assert(((size_t) *n) < BC_BASE_POW); 214 return BC_BASE_DIGS - bc_num_log10((size_t) *n); 215 } 216 217 /** 218 * Returns the power of 10 that the least significant limb should be multiplied 219 * by to put its digits in the right place. For example, if the scale only 220 * reaches 8 places into the limb, this will return 1 (because it should be 221 * multiplied by 10^1) to put the number in the correct place. 222 * @param scale The scale. 223 * @return The power of 10 that the least significant limb should be 224 * multiplied by 225 */ 226 static inline size_t 227 bc_num_leastSigPow(size_t scale) 228 { 229 size_t digs; 230 231 digs = scale % BC_BASE_DIGS; 232 digs = digs != 0 ? BC_BASE_DIGS - digs : 0; 233 234 return bc_num_pow10[digs]; 235 } 236 237 /** 238 * Return the total number of integer digits in a number. This is the opposite 239 * of scale, like bc_num_int() is the opposite of rdx. 240 * @param n The number. 241 * @return The number of integer digits in @a n. 242 */ 243 static size_t 244 bc_num_intDigits(const BcNum* n) 245 { 246 size_t digits = bc_num_int(n) * BC_BASE_DIGS; 247 if (digits > 0) digits -= bc_num_zeroDigits(n->num + n->len - 1); 248 return digits; 249 } 250 251 /** 252 * Returns the number of limbs of a number that are non-zero starting at the 253 * most significant limbs. This expects that there are *no* integer limbs in the 254 * number because it is specifically to figure out how many zero limbs after the 255 * decimal place to ignore. If there are zero limbs after non-zero limbs, they 256 * are counted as non-zero limbs. 257 * @param n The number. 258 * @return The number of non-zero limbs after the decimal point. 259 */ 260 static size_t 261 bc_num_nonZeroLen(const BcNum* restrict n) 262 { 263 size_t i, len = n->len; 264 265 assert(len == BC_NUM_RDX_VAL(n)); 266 267 for (i = len - 1; i < len && !n->num[i]; --i) 268 { 269 continue; 270 } 271 272 assert(i + 1 > 0); 273 274 return i + 1; 275 } 276 277 #if BC_ENABLE_EXTRA_MATH 278 279 /** 280 * Returns the power of 10 that a number with an absolute value less than 1 281 * needs to be multiplied by in order to be greater than 1 or less than -1. 282 * @param n The number. 283 * @return The power of 10 that a number greater than 1 and less than -1 must 284 * be multiplied by to be greater than 1 or less than -1. 285 */ 286 static size_t 287 bc_num_negPow10(const BcNum* restrict n) 288 { 289 // Figure out how many limbs after the decimal point is zero. 290 size_t i, places, idx = bc_num_nonZeroLen(n) - 1; 291 292 places = 1; 293 294 // Figure out how much in the last limb is zero. 295 for (i = BC_BASE_DIGS - 1; i < BC_BASE_DIGS; --i) 296 { 297 if (bc_num_pow10[i] > (BcBigDig) n->num[idx]) places += 1; 298 else break; 299 } 300 301 // Calculate the combination of zero limbs and zero digits in the last 302 // limb. 303 return places + (BC_NUM_RDX_VAL(n) - (idx + 1)) * BC_BASE_DIGS; 304 } 305 306 #endif // BC_ENABLE_EXTRA_MATH 307 308 /** 309 * Performs a one-limb add with a carry. 310 * @param a The first limb. 311 * @param b The second limb. 312 * @param carry An in/out parameter; the carry in from the previous add and the 313 * carry out from this add. 314 * @return The resulting limb sum. 315 */ 316 static BcDig 317 bc_num_addDigits(BcDig a, BcDig b, bool* carry) 318 { 319 assert(((BcBigDig) BC_BASE_POW) * 2 == ((BcDig) BC_BASE_POW) * 2); 320 assert(a < BC_BASE_POW && a >= 0); 321 assert(b < BC_BASE_POW && b >= 0); 322 323 a += b + *carry; 324 *carry = (a >= BC_BASE_POW); 325 if (*carry) a -= BC_BASE_POW; 326 327 assert(a >= 0); 328 assert(a < BC_BASE_POW); 329 330 return a; 331 } 332 333 /** 334 * Performs a one-limb subtract with a carry. 335 * @param a The first limb. 336 * @param b The second limb. 337 * @param carry An in/out parameter; the carry in from the previous subtract 338 * and the carry out from this subtract. 339 * @return The resulting limb difference. 340 */ 341 static BcDig 342 bc_num_subDigits(BcDig a, BcDig b, bool* carry) 343 { 344 assert(a < BC_BASE_POW && a >= 0); 345 assert(b < BC_BASE_POW && b >= 0); 346 347 b += *carry; 348 *carry = (a < b); 349 if (*carry) a += BC_BASE_POW; 350 351 assert(a - b >= 0); 352 assert(a - b < BC_BASE_POW); 353 354 return a - b; 355 } 356 357 /** 358 * Add two BcDig arrays and store the result in the first array. 359 * @param a The first operand and out array. 360 * @param b The second operand. 361 * @param len The length of @a b. 362 */ 363 static void 364 bc_num_addArrays(BcDig* restrict a, const BcDig* restrict b, size_t len) 365 { 366 size_t i; 367 bool carry = false; 368 369 for (i = 0; i < len; ++i) 370 { 371 a[i] = bc_num_addDigits(a[i], b[i], &carry); 372 } 373 374 // Take care of the extra limbs in the bigger array. 375 for (; carry; ++i) 376 { 377 a[i] = bc_num_addDigits(a[i], 0, &carry); 378 } 379 } 380 381 /** 382 * Subtract two BcDig arrays and store the result in the first array. 383 * @param a The first operand and out array. 384 * @param b The second operand. 385 * @param len The length of @a b. 386 */ 387 static void 388 bc_num_subArrays(BcDig* restrict a, const BcDig* restrict b, size_t len) 389 { 390 size_t i; 391 bool carry = false; 392 393 for (i = 0; i < len; ++i) 394 { 395 a[i] = bc_num_subDigits(a[i], b[i], &carry); 396 } 397 398 // Take care of the extra limbs in the bigger array. 399 for (; carry; ++i) 400 { 401 a[i] = bc_num_subDigits(a[i], 0, &carry); 402 } 403 } 404 405 /** 406 * Multiply a BcNum array by a one-limb number. This is a faster version of 407 * multiplication for when we can use it. 408 * @param a The BcNum to multiply by the one-limb number. 409 * @param b The one limb of the one-limb number. 410 * @param c The return parameter. 411 */ 412 static void 413 bc_num_mulArray(const BcNum* restrict a, BcBigDig b, BcNum* restrict c) 414 { 415 size_t i; 416 BcBigDig carry = 0; 417 418 assert(b <= BC_BASE_POW); 419 420 // Make sure the return parameter is big enough. 421 if (a->len + 1 > c->cap) bc_num_expand(c, a->len + 1); 422 423 // We want the entire return parameter to be zero for cleaning later. 424 // NOLINTNEXTLINE 425 memset(c->num, 0, BC_NUM_SIZE(c->cap)); 426 427 // Actual multiplication loop. 428 for (i = 0; i < a->len; ++i) 429 { 430 BcBigDig in = ((BcBigDig) a->num[i]) * b + carry; 431 c->num[i] = in % BC_BASE_POW; 432 carry = in / BC_BASE_POW; 433 } 434 435 assert(carry < BC_BASE_POW); 436 437 // Finishing touches. 438 c->num[i] = (BcDig) carry; 439 assert(c->num[i] >= 0 && c->num[i] < BC_BASE_POW); 440 c->len = a->len; 441 c->len += (carry != 0); 442 443 bc_num_clean(c); 444 445 // Postconditions. 446 assert(!BC_NUM_NEG(c) || BC_NUM_NONZERO(c)); 447 assert(BC_NUM_RDX_VAL(c) <= c->len || !c->len); 448 assert(!c->len || c->num[c->len - 1] || BC_NUM_RDX_VAL(c) == c->len); 449 } 450 451 /** 452 * Divide a BcNum array by a one-limb number. This is a faster version of divide 453 * for when we can use it. 454 * @param a The BcNum to multiply by the one-limb number. 455 * @param b The one limb of the one-limb number. 456 * @param c The return parameter for the quotient. 457 * @param rem The return parameter for the remainder. 458 */ 459 static void 460 bc_num_divArray(const BcNum* restrict a, BcBigDig b, BcNum* restrict c, 461 BcBigDig* rem) 462 { 463 size_t i; 464 BcBigDig carry = 0; 465 466 assert(c->cap >= a->len); 467 468 // Actual division loop. 469 for (i = a->len - 1; i < a->len; --i) 470 { 471 BcBigDig in = ((BcBigDig) a->num[i]) + carry * BC_BASE_POW; 472 assert(in / b < BC_BASE_POW); 473 c->num[i] = (BcDig) (in / b); 474 assert(c->num[i] >= 0 && c->num[i] < BC_BASE_POW); 475 carry = in % b; 476 } 477 478 // Finishing touches. 479 c->len = a->len; 480 bc_num_clean(c); 481 *rem = carry; 482 483 // Postconditions. 484 assert(!BC_NUM_NEG(c) || BC_NUM_NONZERO(c)); 485 assert(BC_NUM_RDX_VAL(c) <= c->len || !c->len); 486 assert(!c->len || c->num[c->len - 1] || BC_NUM_RDX_VAL(c) == c->len); 487 } 488 489 /** 490 * Compare two BcDig arrays and return >0 if @a b is greater, <0 if @a b is 491 * less, and 0 if equal. Both @a a and @a b must have the same length. 492 * @param a The first array. 493 * @param b The second array. 494 * @param len The minimum length of the arrays. 495 */ 496 static ssize_t 497 bc_num_compare(const BcDig* restrict a, const BcDig* restrict b, size_t len) 498 { 499 size_t i; 500 BcDig c = 0; 501 for (i = len - 1; i < len && !(c = a[i] - b[i]); --i) 502 { 503 continue; 504 } 505 return bc_num_neg(i + 1, c < 0); 506 } 507 508 ssize_t 509 bc_num_cmp(const BcNum* a, const BcNum* b) 510 { 511 size_t i, min, a_int, b_int, diff, ardx, brdx; 512 BcDig* max_num; 513 BcDig* min_num; 514 bool a_max, neg = false; 515 ssize_t cmp; 516 517 assert(a != NULL && b != NULL); 518 519 // Same num? Equal. 520 if (a == b) return 0; 521 522 // Easy cases. 523 if (BC_NUM_ZERO(a)) return bc_num_neg(b->len != 0, !BC_NUM_NEG(b)); 524 if (BC_NUM_ZERO(b)) return bc_num_cmpZero(a); 525 if (BC_NUM_NEG(a)) 526 { 527 if (BC_NUM_NEG(b)) neg = true; 528 else return -1; 529 } 530 else if (BC_NUM_NEG(b)) return 1; 531 532 // Get the number of int limbs in each number and get the difference. 533 a_int = bc_num_int(a); 534 b_int = bc_num_int(b); 535 a_int -= b_int; 536 537 // If there's a difference, then just return the comparison. 538 if (a_int) return neg ? -((ssize_t) a_int) : (ssize_t) a_int; 539 540 // Get the rdx's and figure out the max. 541 ardx = BC_NUM_RDX_VAL(a); 542 brdx = BC_NUM_RDX_VAL(b); 543 a_max = (ardx > brdx); 544 545 // Set variables based on the above. 546 if (a_max) 547 { 548 min = brdx; 549 diff = ardx - brdx; 550 max_num = a->num + diff; 551 min_num = b->num; 552 } 553 else 554 { 555 min = ardx; 556 diff = brdx - ardx; 557 max_num = b->num + diff; 558 min_num = a->num; 559 } 560 561 // Do a full limb-by-limb comparison. 562 cmp = bc_num_compare(max_num, min_num, b_int + min); 563 564 // If we found a difference, return it based on state. 565 if (cmp) return bc_num_neg((size_t) cmp, !a_max == !neg); 566 567 // If there was no difference, then the final step is to check which number 568 // has greater or lesser limbs beyond the other's. 569 for (max_num -= diff, i = diff - 1; i < diff; --i) 570 { 571 if (max_num[i]) return bc_num_neg(1, !a_max == !neg); 572 } 573 574 return 0; 575 } 576 577 void 578 bc_num_truncate(BcNum* restrict n, size_t places) 579 { 580 size_t nrdx, places_rdx; 581 582 if (!places) return; 583 584 // Grab these needed values; places_rdx is the rdx equivalent to places like 585 // rdx is to scale. 586 nrdx = BC_NUM_RDX_VAL(n); 587 places_rdx = nrdx ? nrdx - BC_NUM_RDX(n->scale - places) : 0; 588 589 // We cannot truncate more places than we have. 590 assert(places <= n->scale && (BC_NUM_ZERO(n) || places_rdx <= n->len)); 591 592 n->scale -= places; 593 BC_NUM_RDX_SET(n, nrdx - places_rdx); 594 595 // Only when the number is nonzero do we need to do the hard stuff. 596 if (BC_NUM_NONZERO(n)) 597 { 598 size_t pow; 599 600 // This calculates how many decimal digits are in the least significant 601 // limb, then gets the power for that. 602 pow = bc_num_leastSigPow(n->scale); 603 604 n->len -= places_rdx; 605 606 // We have to move limbs to maintain invariants. The limbs must begin at 607 // the beginning of the BcNum array. 608 // NOLINTNEXTLINE 609 memmove(n->num, n->num + places_rdx, BC_NUM_SIZE(n->len)); 610 611 // Clear the lower part of the last digit. 612 if (BC_NUM_NONZERO(n)) n->num[0] -= n->num[0] % (BcDig) pow; 613 614 bc_num_clean(n); 615 } 616 } 617 618 void 619 bc_num_extend(BcNum* restrict n, size_t places) 620 { 621 size_t nrdx, places_rdx; 622 623 if (!places) return; 624 625 // Easy case with zero; set the scale. 626 if (BC_NUM_ZERO(n)) 627 { 628 n->scale += places; 629 return; 630 } 631 632 // Grab these needed values; places_rdx is the rdx equivalent to places like 633 // rdx is to scale. 634 nrdx = BC_NUM_RDX_VAL(n); 635 places_rdx = BC_NUM_RDX(places + n->scale) - nrdx; 636 637 // This is the hard case. We need to expand the number, move the limbs, and 638 // set the limbs that were just cleared. 639 if (places_rdx) 640 { 641 bc_num_expand(n, bc_vm_growSize(n->len, places_rdx)); 642 // NOLINTNEXTLINE 643 memmove(n->num + places_rdx, n->num, BC_NUM_SIZE(n->len)); 644 // NOLINTNEXTLINE 645 memset(n->num, 0, BC_NUM_SIZE(places_rdx)); 646 } 647 648 // Finally, set scale and rdx. 649 BC_NUM_RDX_SET(n, nrdx + places_rdx); 650 n->scale += places; 651 n->len += places_rdx; 652 653 assert(BC_NUM_RDX_VAL(n) == BC_NUM_RDX(n->scale)); 654 } 655 656 /** 657 * Retires (finishes) a multiplication or division operation. 658 */ 659 static void 660 bc_num_retireMul(BcNum* restrict n, size_t scale, bool neg1, bool neg2) 661 { 662 // Make sure scale is correct. 663 if (n->scale < scale) bc_num_extend(n, scale - n->scale); 664 else bc_num_truncate(n, n->scale - scale); 665 666 bc_num_clean(n); 667 668 // We need to ensure rdx is correct. 669 if (BC_NUM_NONZERO(n)) n->rdx = BC_NUM_NEG_VAL(n, !neg1 != !neg2); 670 } 671 672 /** 673 * Splits a number into two BcNum's. This is used in Karatsuba. 674 * @param n The number to split. 675 * @param idx The index to split at. 676 * @param a An out parameter; the low part of @a n. 677 * @param b An out parameter; the high part of @a n. 678 */ 679 static void 680 bc_num_split(const BcNum* restrict n, size_t idx, BcNum* restrict a, 681 BcNum* restrict b) 682 { 683 // We want a and b to be clear. 684 assert(BC_NUM_ZERO(a)); 685 assert(BC_NUM_ZERO(b)); 686 687 // The usual case. 688 if (idx < n->len) 689 { 690 // Set the fields first. 691 b->len = n->len - idx; 692 a->len = idx; 693 a->scale = b->scale = 0; 694 BC_NUM_RDX_SET(a, 0); 695 BC_NUM_RDX_SET(b, 0); 696 697 assert(a->cap >= a->len); 698 assert(b->cap >= b->len); 699 700 // Copy the arrays. This is not necessary for safety, but it is faster, 701 // for some reason. 702 // NOLINTNEXTLINE 703 memcpy(b->num, n->num + idx, BC_NUM_SIZE(b->len)); 704 // NOLINTNEXTLINE 705 memcpy(a->num, n->num, BC_NUM_SIZE(idx)); 706 707 bc_num_clean(b); 708 } 709 // If the index is weird, just skip the split. 710 else bc_num_copy(a, n); 711 712 bc_num_clean(a); 713 } 714 715 /** 716 * Copies a number into another, but shifts the rdx so that the result number 717 * only sees the integer part of the source number. 718 * @param n The number to copy. 719 * @param r The result number with a shifted rdx, len, and num. 720 */ 721 static void 722 bc_num_shiftRdx(const BcNum* restrict n, BcNum* restrict r) 723 { 724 size_t rdx = BC_NUM_RDX_VAL(n); 725 726 r->len = n->len - rdx; 727 r->cap = n->cap - rdx; 728 r->num = n->num + rdx; 729 730 BC_NUM_RDX_SET_NEG(r, 0, BC_NUM_NEG(n)); 731 r->scale = 0; 732 } 733 734 /** 735 * Shifts a number so that all of the least significant limbs of the number are 736 * skipped. This must be undone by bc_num_unshiftZero(). 737 * @param n The number to shift. 738 */ 739 static size_t 740 bc_num_shiftZero(BcNum* restrict n) 741 { 742 // This is volatile to quiet a GCC warning about longjmp() clobbering. 743 volatile size_t i; 744 745 // If we don't have an integer, that is a problem, but it's also a bug 746 // because the caller should have set everything up right. 747 assert(!BC_NUM_RDX_VAL(n) || BC_NUM_ZERO(n)); 748 749 for (i = 0; i < n->len && !n->num[i]; ++i) 750 { 751 continue; 752 } 753 754 n->len -= i; 755 n->num += i; 756 757 return i; 758 } 759 760 /** 761 * Undo the damage done by bc_num_unshiftZero(). This must be called like all 762 * cleanup functions: after a label used by setjmp() and longjmp(). 763 * @param n The number to unshift. 764 * @param places_rdx The amount the number was originally shift. 765 */ 766 static void 767 bc_num_unshiftZero(BcNum* restrict n, size_t places_rdx) 768 { 769 n->len += places_rdx; 770 n->num -= places_rdx; 771 } 772 773 /** 774 * Shifts the digits right within a number by no more than BC_BASE_DIGS - 1. 775 * This is the final step on shifting numbers with the shift operators. It 776 * depends on the caller to shift the limbs properly because all it does is 777 * ensure that the rdx point is realigned to be between limbs. 778 * @param n The number to shift digits in. 779 * @param dig The number of places to shift right. 780 */ 781 static void 782 bc_num_shift(BcNum* restrict n, BcBigDig dig) 783 { 784 size_t i, len = n->len; 785 BcBigDig carry = 0, pow; 786 BcDig* ptr = n->num; 787 788 assert(dig < BC_BASE_DIGS); 789 790 // Figure out the parameters for division. 791 pow = bc_num_pow10[dig]; 792 dig = bc_num_pow10[BC_BASE_DIGS - dig]; 793 794 // Run a series of divisions and mods with carries across the entire number 795 // array. This effectively shifts everything over. 796 for (i = len - 1; i < len; --i) 797 { 798 BcBigDig in, temp; 799 in = ((BcBigDig) ptr[i]); 800 temp = carry * dig; 801 carry = in % pow; 802 ptr[i] = ((BcDig) (in / pow)) + (BcDig) temp; 803 assert(ptr[i] >= 0 && ptr[i] < BC_BASE_POW); 804 } 805 806 assert(!carry); 807 } 808 809 /** 810 * Shift a number left by a certain number of places. This is the workhorse of 811 * the left shift operator. 812 * @param n The number to shift left. 813 * @param places The amount of places to shift @a n left by. 814 */ 815 static void 816 bc_num_shiftLeft(BcNum* restrict n, size_t places) 817 { 818 BcBigDig dig; 819 size_t places_rdx; 820 bool shift; 821 822 if (!places) return; 823 824 // Make sure to grow the number if necessary. 825 if (places > n->scale) 826 { 827 size_t size = bc_vm_growSize(BC_NUM_RDX(places - n->scale), n->len); 828 if (size > SIZE_MAX - 1) bc_err(BC_ERR_MATH_OVERFLOW); 829 } 830 831 // If zero, we can just set the scale and bail. 832 if (BC_NUM_ZERO(n)) 833 { 834 if (n->scale >= places) n->scale -= places; 835 else n->scale = 0; 836 return; 837 } 838 839 // When I changed bc to have multiple digits per limb, this was the hardest 840 // code to change. This and shift right. Make sure you understand this 841 // before attempting anything. 842 843 // This is how many limbs we will shift. 844 dig = (BcBigDig) (places % BC_BASE_DIGS); 845 shift = (dig != 0); 846 847 // Convert places to a rdx value. 848 places_rdx = BC_NUM_RDX(places); 849 850 // If the number is not an integer, we need special care. The reason an 851 // integer doesn't is because left shift would only extend the integer, 852 // whereas a non-integer might have its fractional part eliminated or only 853 // partially eliminated. 854 if (n->scale) 855 { 856 size_t nrdx = BC_NUM_RDX_VAL(n); 857 858 // If the number's rdx is bigger, that's the hard case. 859 if (nrdx >= places_rdx) 860 { 861 size_t mod = n->scale % BC_BASE_DIGS, revdig; 862 863 // We want mod to be in the range [1, BC_BASE_DIGS], not 864 // [0, BC_BASE_DIGS). 865 mod = mod ? mod : BC_BASE_DIGS; 866 867 // We need to reverse dig to get the actual number of digits. 868 revdig = dig ? BC_BASE_DIGS - dig : 0; 869 870 // If the two overflow BC_BASE_DIGS, we need to move an extra place. 871 if (mod + revdig > BC_BASE_DIGS) places_rdx = 1; 872 else places_rdx = 0; 873 } 874 else places_rdx -= nrdx; 875 } 876 877 // If this is non-zero, we need an extra place, so expand, move, and set. 878 if (places_rdx) 879 { 880 bc_num_expand(n, bc_vm_growSize(n->len, places_rdx)); 881 // NOLINTNEXTLINE 882 memmove(n->num + places_rdx, n->num, BC_NUM_SIZE(n->len)); 883 // NOLINTNEXTLINE 884 memset(n->num, 0, BC_NUM_SIZE(places_rdx)); 885 n->len += places_rdx; 886 } 887 888 // Set the scale appropriately. 889 if (places > n->scale) 890 { 891 n->scale = 0; 892 BC_NUM_RDX_SET(n, 0); 893 } 894 else 895 { 896 n->scale -= places; 897 BC_NUM_RDX_SET(n, BC_NUM_RDX(n->scale)); 898 } 899 900 // Finally, shift within limbs. 901 if (shift) bc_num_shift(n, BC_BASE_DIGS - dig); 902 903 bc_num_clean(n); 904 } 905 906 void 907 bc_num_shiftRight(BcNum* restrict n, size_t places) 908 { 909 BcBigDig dig; 910 size_t places_rdx, scale, scale_mod, int_len, expand; 911 bool shift; 912 913 if (!places) return; 914 915 // If zero, we can just set the scale and bail. 916 if (BC_NUM_ZERO(n)) 917 { 918 n->scale += places; 919 bc_num_expand(n, BC_NUM_RDX(n->scale)); 920 return; 921 } 922 923 // Amount within a limb we have to shift by. 924 dig = (BcBigDig) (places % BC_BASE_DIGS); 925 shift = (dig != 0); 926 927 scale = n->scale; 928 929 // Figure out how the scale is affected. 930 scale_mod = scale % BC_BASE_DIGS; 931 scale_mod = scale_mod ? scale_mod : BC_BASE_DIGS; 932 933 // We need to know the int length and rdx for places. 934 int_len = bc_num_int(n); 935 places_rdx = BC_NUM_RDX(places); 936 937 // If we are going to shift past a limb boundary or not, set accordingly. 938 if (scale_mod + dig > BC_BASE_DIGS) 939 { 940 expand = places_rdx - 1; 941 places_rdx = 1; 942 } 943 else 944 { 945 expand = places_rdx; 946 places_rdx = 0; 947 } 948 949 // Clamp expanding. 950 if (expand > int_len) expand -= int_len; 951 else expand = 0; 952 953 // Extend, expand, and zero. 954 bc_num_extend(n, places_rdx * BC_BASE_DIGS); 955 bc_num_expand(n, bc_vm_growSize(expand, n->len)); 956 // NOLINTNEXTLINE 957 memset(n->num + n->len, 0, BC_NUM_SIZE(expand)); 958 959 // Set the fields. 960 n->len += expand; 961 n->scale = 0; 962 BC_NUM_RDX_SET(n, 0); 963 964 // Finally, shift within limbs. 965 if (shift) bc_num_shift(n, dig); 966 967 n->scale = scale + places; 968 BC_NUM_RDX_SET(n, BC_NUM_RDX(n->scale)); 969 970 bc_num_clean(n); 971 972 assert(BC_NUM_RDX_VAL(n) <= n->len && n->len <= n->cap); 973 assert(BC_NUM_RDX_VAL(n) == BC_NUM_RDX(n->scale)); 974 } 975 976 /** 977 * Tests if a number is a integer with scale or not. Returns true if the number 978 * is not an integer. If it is, its integer shifted form is copied into the 979 * result parameter for use where only integers are allowed. 980 * @param n The integer to test and shift. 981 * @param r The number to store the shifted result into. This number should 982 * *not* be allocated. 983 * @return True if the number is a non-integer, false otherwise. 984 */ 985 static bool 986 bc_num_nonInt(const BcNum* restrict n, BcNum* restrict r) 987 { 988 bool zero; 989 size_t i, rdx = BC_NUM_RDX_VAL(n); 990 991 if (!rdx) 992 { 993 // NOLINTNEXTLINE 994 memcpy(r, n, sizeof(BcNum)); 995 return false; 996 } 997 998 zero = true; 999 1000 for (i = 0; zero && i < rdx; ++i) 1001 { 1002 zero = (n->num[i] == 0); 1003 } 1004 1005 if (BC_ERR(!zero)) return true; 1006 1007 bc_num_shiftRdx(n, r); 1008 1009 return false; 1010 } 1011 1012 #if BC_ENABLE_EXTRA_MATH 1013 1014 /** 1015 * Execute common code for an operater that needs an integer for the second 1016 * operand and return the integer operand as a BcBigDig. 1017 * @param a The first operand. 1018 * @param b The second operand. 1019 * @param c The result operand. 1020 * @return The second operand as a hardware integer. 1021 */ 1022 static BcBigDig 1023 bc_num_intop(const BcNum* a, const BcNum* b, BcNum* restrict c) 1024 { 1025 BcNum temp; 1026 1027 #if BC_GCC 1028 temp.len = 0; 1029 temp.rdx = 0; 1030 temp.num = NULL; 1031 #endif // BC_GCC 1032 1033 if (BC_ERR(bc_num_nonInt(b, &temp))) bc_err(BC_ERR_MATH_NON_INTEGER); 1034 1035 bc_num_copy(c, a); 1036 1037 return bc_num_bigdig(&temp); 1038 } 1039 #endif // BC_ENABLE_EXTRA_MATH 1040 1041 /** 1042 * This is the actual implementation of add *and* subtract. Since this function 1043 * doesn't need to use scale (per the bc spec), I am hijacking it to say whether 1044 * it's doing an add or a subtract. And then I convert substraction to addition 1045 * of negative second operand. This is a BcNumBinOp function. 1046 * @param a The first operand. 1047 * @param b The second operand. 1048 * @param c The return parameter. 1049 * @param sub Non-zero for a subtract, zero for an add. 1050 */ 1051 static void 1052 bc_num_as(BcNum* a, BcNum* b, BcNum* restrict c, size_t sub) 1053 { 1054 BcDig* ptr_c; 1055 BcDig* ptr_l; 1056 BcDig* ptr_r; 1057 size_t i, min_rdx, max_rdx, diff, a_int, b_int, min_len, max_len, max_int; 1058 size_t len_l, len_r, ardx, brdx; 1059 bool b_neg, do_sub, do_rev_sub, carry, c_neg; 1060 1061 if (BC_NUM_ZERO(b)) 1062 { 1063 bc_num_copy(c, a); 1064 return; 1065 } 1066 1067 if (BC_NUM_ZERO(a)) 1068 { 1069 bc_num_copy(c, b); 1070 c->rdx = BC_NUM_NEG_VAL(c, BC_NUM_NEG(b) != sub); 1071 return; 1072 } 1073 1074 // Invert sign of b if it is to be subtracted. This operation must 1075 // precede the tests for any of the operands being zero. 1076 b_neg = (BC_NUM_NEG(b) != sub); 1077 1078 // Figure out if we will actually add the numbers if their signs are equal 1079 // or subtract. 1080 do_sub = (BC_NUM_NEG(a) != b_neg); 1081 1082 a_int = bc_num_int(a); 1083 b_int = bc_num_int(b); 1084 max_int = BC_MAX(a_int, b_int); 1085 1086 // Figure out which number will have its last limbs copied (for addition) or 1087 // subtracted (for subtraction). 1088 ardx = BC_NUM_RDX_VAL(a); 1089 brdx = BC_NUM_RDX_VAL(b); 1090 min_rdx = BC_MIN(ardx, brdx); 1091 max_rdx = BC_MAX(ardx, brdx); 1092 diff = max_rdx - min_rdx; 1093 1094 max_len = max_int + max_rdx; 1095 1096 if (do_sub) 1097 { 1098 // Check whether b has to be subtracted from a or a from b. 1099 if (a_int != b_int) do_rev_sub = (a_int < b_int); 1100 else if (ardx > brdx) 1101 { 1102 do_rev_sub = (bc_num_compare(a->num + diff, b->num, b->len) < 0); 1103 } 1104 else do_rev_sub = (bc_num_compare(a->num, b->num + diff, a->len) <= 0); 1105 } 1106 else 1107 { 1108 // The result array of the addition might come out one element 1109 // longer than the bigger of the operand arrays. 1110 max_len += 1; 1111 do_rev_sub = (a_int < b_int); 1112 } 1113 1114 assert(max_len <= c->cap); 1115 1116 // Cache values for simple code later. 1117 if (do_rev_sub) 1118 { 1119 ptr_l = b->num; 1120 ptr_r = a->num; 1121 len_l = b->len; 1122 len_r = a->len; 1123 } 1124 else 1125 { 1126 ptr_l = a->num; 1127 ptr_r = b->num; 1128 len_l = a->len; 1129 len_r = b->len; 1130 } 1131 1132 ptr_c = c->num; 1133 carry = false; 1134 1135 // This is true if the numbers have a different number of limbs after the 1136 // decimal point. 1137 if (diff) 1138 { 1139 // If the rdx values of the operands do not match, the result will 1140 // have low end elements that are the positive or negative trailing 1141 // elements of the operand with higher rdx value. 1142 if ((ardx > brdx) != do_rev_sub) 1143 { 1144 // !do_rev_sub && ardx > brdx || do_rev_sub && brdx > ardx 1145 // The left operand has BcDig values that need to be copied, 1146 // either from a or from b (in case of a reversed subtraction). 1147 // NOLINTNEXTLINE 1148 memcpy(ptr_c, ptr_l, BC_NUM_SIZE(diff)); 1149 ptr_l += diff; 1150 len_l -= diff; 1151 } 1152 else 1153 { 1154 // The right operand has BcDig values that need to be copied 1155 // or subtracted from zero (in case of a subtraction). 1156 if (do_sub) 1157 { 1158 // do_sub (do_rev_sub && ardx > brdx || 1159 // !do_rev_sub && brdx > ardx) 1160 for (i = 0; i < diff; i++) 1161 { 1162 ptr_c[i] = bc_num_subDigits(0, ptr_r[i], &carry); 1163 } 1164 } 1165 else 1166 { 1167 // !do_sub && brdx > ardx 1168 // NOLINTNEXTLINE 1169 memcpy(ptr_c, ptr_r, BC_NUM_SIZE(diff)); 1170 } 1171 1172 // Future code needs to ignore the limbs we just did. 1173 ptr_r += diff; 1174 len_r -= diff; 1175 } 1176 1177 // The return value pointer needs to ignore what we just did. 1178 ptr_c += diff; 1179 } 1180 1181 // This is the length that can be directly added/subtracted. 1182 min_len = BC_MIN(len_l, len_r); 1183 1184 // After dealing with possible low array elements that depend on only one 1185 // operand above, the actual add or subtract can be performed as if the rdx 1186 // of both operands was the same. 1187 // 1188 // Inlining takes care of eliminating constant zero arguments to 1189 // addDigit/subDigit (checked in disassembly of resulting bc binary 1190 // compiled with gcc and clang). 1191 if (do_sub) 1192 { 1193 // Actual subtraction. 1194 for (i = 0; i < min_len; ++i) 1195 { 1196 ptr_c[i] = bc_num_subDigits(ptr_l[i], ptr_r[i], &carry); 1197 } 1198 1199 // Finishing the limbs beyond the direct subtraction. 1200 for (; i < len_l; ++i) 1201 { 1202 ptr_c[i] = bc_num_subDigits(ptr_l[i], 0, &carry); 1203 } 1204 } 1205 else 1206 { 1207 // Actual addition. 1208 for (i = 0; i < min_len; ++i) 1209 { 1210 ptr_c[i] = bc_num_addDigits(ptr_l[i], ptr_r[i], &carry); 1211 } 1212 1213 // Finishing the limbs beyond the direct addition. 1214 for (; i < len_l; ++i) 1215 { 1216 ptr_c[i] = bc_num_addDigits(ptr_l[i], 0, &carry); 1217 } 1218 1219 // Addition can create an extra limb. We take care of that here. 1220 ptr_c[i] = bc_num_addDigits(0, 0, &carry); 1221 } 1222 1223 assert(carry == false); 1224 1225 // The result has the same sign as a, unless the operation was a 1226 // reverse subtraction (b - a). 1227 c_neg = BC_NUM_NEG(a) != (do_sub && do_rev_sub); 1228 BC_NUM_RDX_SET_NEG(c, max_rdx, c_neg); 1229 c->len = max_len; 1230 c->scale = BC_MAX(a->scale, b->scale); 1231 1232 bc_num_clean(c); 1233 } 1234 1235 /** 1236 * The simple multiplication that karatsuba dishes out to when the length of the 1237 * numbers gets low enough. This doesn't use scale because it treats the 1238 * operands as though they are integers. 1239 * @param a The first operand. 1240 * @param b The second operand. 1241 * @param c The return parameter. 1242 */ 1243 static void 1244 bc_num_m_simp(const BcNum* a, const BcNum* b, BcNum* restrict c) 1245 { 1246 size_t i, alen = a->len, blen = b->len, clen; 1247 BcDig* ptr_a = a->num; 1248 BcDig* ptr_b = b->num; 1249 BcDig* ptr_c; 1250 BcBigDig sum = 0, carry = 0; 1251 1252 assert(sizeof(sum) >= sizeof(BcDig) * 2); 1253 assert(!BC_NUM_RDX_VAL(a) && !BC_NUM_RDX_VAL(b)); 1254 1255 // Make sure c is big enough. 1256 clen = bc_vm_growSize(alen, blen); 1257 bc_num_expand(c, bc_vm_growSize(clen, 1)); 1258 1259 // If we don't memset, then we might have uninitialized data use later. 1260 ptr_c = c->num; 1261 // NOLINTNEXTLINE 1262 memset(ptr_c, 0, BC_NUM_SIZE(c->cap)); 1263 1264 // This is the actual multiplication loop. It uses the lattice form of long 1265 // multiplication (see the explanation on the web page at 1266 // https://knilt.arcc.albany.edu/What_is_Lattice_Multiplication or the 1267 // explanation at Wikipedia). 1268 for (i = 0; i < clen; ++i) 1269 { 1270 ssize_t sidx = (ssize_t) (i - blen + 1); 1271 size_t j, k; 1272 1273 // These are the start indices. 1274 j = (size_t) BC_MAX(0, sidx); 1275 k = BC_MIN(i, blen - 1); 1276 1277 // On every iteration of this loop, a multiplication happens, then the 1278 // sum is automatically calculated. 1279 for (; j < alen && k < blen; ++j, --k) 1280 { 1281 sum += ((BcBigDig) ptr_a[j]) * ((BcBigDig) ptr_b[k]); 1282 1283 if (sum >= ((BcBigDig) BC_BASE_POW) * BC_BASE_POW) 1284 { 1285 carry += sum / BC_BASE_POW; 1286 sum %= BC_BASE_POW; 1287 } 1288 } 1289 1290 // Calculate the carry. 1291 if (sum >= BC_BASE_POW) 1292 { 1293 carry += sum / BC_BASE_POW; 1294 sum %= BC_BASE_POW; 1295 } 1296 1297 // Store and set up for next iteration. 1298 ptr_c[i] = (BcDig) sum; 1299 assert(ptr_c[i] < BC_BASE_POW); 1300 sum = carry; 1301 carry = 0; 1302 } 1303 1304 // This should always be true because there should be no carry on the last 1305 // digit; multiplication never goes above the sum of both lengths. 1306 assert(!sum); 1307 1308 c->len = clen; 1309 } 1310 1311 /** 1312 * Does a shifted add or subtract for Karatsuba below. This calls either 1313 * bc_num_addArrays() or bc_num_subArrays(). 1314 * @param n An in/out parameter; the first operand and return parameter. 1315 * @param a The second operand. 1316 * @param shift The amount to shift @a n by when adding/subtracting. 1317 * @param op The function to call, either bc_num_addArrays() or 1318 * bc_num_subArrays(). 1319 */ 1320 static void 1321 bc_num_shiftAddSub(BcNum* restrict n, const BcNum* restrict a, size_t shift, 1322 BcNumShiftAddOp op) 1323 { 1324 assert(n->len >= shift + a->len); 1325 assert(!BC_NUM_RDX_VAL(n) && !BC_NUM_RDX_VAL(a)); 1326 op(n->num + shift, a->num, a->len); 1327 } 1328 1329 /** 1330 * Implements the Karatsuba algorithm. 1331 */ 1332 static void 1333 bc_num_k(const BcNum* a, const BcNum* b, BcNum* restrict c) 1334 { 1335 size_t max, max2, total; 1336 BcNum l1, h1, l2, h2, m2, m1, z0, z1, z2, temp; 1337 BcDig* digs; 1338 BcDig* dig_ptr; 1339 BcNumShiftAddOp op; 1340 bool aone = BC_NUM_ONE(a); 1341 #if BC_ENABLE_LIBRARY 1342 BcVm* vm = bcl_getspecific(); 1343 #endif // BC_ENABLE_LIBRARY 1344 1345 assert(BC_NUM_ZERO(c)); 1346 1347 if (BC_NUM_ZERO(a) || BC_NUM_ZERO(b)) return; 1348 1349 if (aone || BC_NUM_ONE(b)) 1350 { 1351 bc_num_copy(c, aone ? b : a); 1352 if ((aone && BC_NUM_NEG(a)) || BC_NUM_NEG(b)) BC_NUM_NEG_TGL(c); 1353 return; 1354 } 1355 1356 // Shell out to the simple algorithm with certain conditions. 1357 if (a->len < BC_NUM_KARATSUBA_LEN || b->len < BC_NUM_KARATSUBA_LEN) 1358 { 1359 bc_num_m_simp(a, b, c); 1360 return; 1361 } 1362 1363 // We need to calculate the max size of the numbers that can result from the 1364 // operations. 1365 max = BC_MAX(a->len, b->len); 1366 max = BC_MAX(max, BC_NUM_DEF_SIZE); 1367 max2 = (max + 1) / 2; 1368 1369 // Calculate the space needed for all of the temporary allocations. We do 1370 // this to just allocate once. 1371 total = bc_vm_arraySize(BC_NUM_KARATSUBA_ALLOCS, max); 1372 1373 BC_SIG_LOCK; 1374 1375 // Allocate space for all of the temporaries. 1376 digs = dig_ptr = bc_vm_malloc(BC_NUM_SIZE(total)); 1377 1378 // Set up the temporaries. 1379 bc_num_setup(&l1, dig_ptr, max); 1380 dig_ptr += max; 1381 bc_num_setup(&h1, dig_ptr, max); 1382 dig_ptr += max; 1383 bc_num_setup(&l2, dig_ptr, max); 1384 dig_ptr += max; 1385 bc_num_setup(&h2, dig_ptr, max); 1386 dig_ptr += max; 1387 bc_num_setup(&m1, dig_ptr, max); 1388 dig_ptr += max; 1389 bc_num_setup(&m2, dig_ptr, max); 1390 1391 // Some temporaries need the ability to grow, so we allocate them 1392 // separately. 1393 max = bc_vm_growSize(max, 1); 1394 bc_num_init(&z0, max); 1395 bc_num_init(&z1, max); 1396 bc_num_init(&z2, max); 1397 max = bc_vm_growSize(max, max) + 1; 1398 bc_num_init(&temp, max); 1399 1400 BC_SETJMP_LOCKED(vm, err); 1401 1402 BC_SIG_UNLOCK; 1403 1404 // First, set up c. 1405 bc_num_expand(c, max); 1406 c->len = max; 1407 // NOLINTNEXTLINE 1408 memset(c->num, 0, BC_NUM_SIZE(c->len)); 1409 1410 // Split the parameters. 1411 bc_num_split(a, max2, &l1, &h1); 1412 bc_num_split(b, max2, &l2, &h2); 1413 1414 // Do the subtraction. 1415 bc_num_sub(&h1, &l1, &m1, 0); 1416 bc_num_sub(&l2, &h2, &m2, 0); 1417 1418 // The if statements below are there for efficiency reasons. The best way to 1419 // understand them is to understand the Karatsuba algorithm because now that 1420 // the ollocations and splits are done, the algorithm is pretty 1421 // straightforward. 1422 1423 if (BC_NUM_NONZERO(&h1) && BC_NUM_NONZERO(&h2)) 1424 { 1425 assert(BC_NUM_RDX_VALID_NP(h1)); 1426 assert(BC_NUM_RDX_VALID_NP(h2)); 1427 1428 bc_num_m(&h1, &h2, &z2, 0); 1429 bc_num_clean(&z2); 1430 1431 bc_num_shiftAddSub(c, &z2, max2 * 2, bc_num_addArrays); 1432 bc_num_shiftAddSub(c, &z2, max2, bc_num_addArrays); 1433 } 1434 1435 if (BC_NUM_NONZERO(&l1) && BC_NUM_NONZERO(&l2)) 1436 { 1437 assert(BC_NUM_RDX_VALID_NP(l1)); 1438 assert(BC_NUM_RDX_VALID_NP(l2)); 1439 1440 bc_num_m(&l1, &l2, &z0, 0); 1441 bc_num_clean(&z0); 1442 1443 bc_num_shiftAddSub(c, &z0, max2, bc_num_addArrays); 1444 bc_num_shiftAddSub(c, &z0, 0, bc_num_addArrays); 1445 } 1446 1447 if (BC_NUM_NONZERO(&m1) && BC_NUM_NONZERO(&m2)) 1448 { 1449 assert(BC_NUM_RDX_VALID_NP(m1)); 1450 assert(BC_NUM_RDX_VALID_NP(m1)); 1451 1452 bc_num_m(&m1, &m2, &z1, 0); 1453 bc_num_clean(&z1); 1454 1455 op = (BC_NUM_NEG_NP(m1) != BC_NUM_NEG_NP(m2)) ? 1456 bc_num_subArrays : 1457 bc_num_addArrays; 1458 bc_num_shiftAddSub(c, &z1, max2, op); 1459 } 1460 1461 err: 1462 BC_SIG_MAYLOCK; 1463 free(digs); 1464 bc_num_free(&temp); 1465 bc_num_free(&z2); 1466 bc_num_free(&z1); 1467 bc_num_free(&z0); 1468 BC_LONGJMP_CONT(vm); 1469 } 1470 1471 /** 1472 * Does checks for Karatsuba. It also changes things to ensure that the 1473 * Karatsuba and simple multiplication can treat the numbers as integers. This 1474 * is a BcNumBinOp function. 1475 * @param a The first operand. 1476 * @param b The second operand. 1477 * @param c The return parameter. 1478 * @param scale The current scale. 1479 */ 1480 static void 1481 bc_num_m(BcNum* a, BcNum* b, BcNum* restrict c, size_t scale) 1482 { 1483 BcNum cpa, cpb; 1484 size_t ascale, bscale, ardx, brdx, zero, len, rscale; 1485 // These are meant to quiet warnings on GCC about longjmp() clobbering. 1486 // The problem is real here. 1487 size_t scale1, scale2, realscale; 1488 // These are meant to quiet the GCC longjmp() clobbering, even though it 1489 // does not apply here. 1490 volatile size_t azero; 1491 volatile size_t bzero; 1492 #if BC_ENABLE_LIBRARY 1493 BcVm* vm = bcl_getspecific(); 1494 #endif // BC_ENABLE_LIBRARY 1495 1496 assert(BC_NUM_RDX_VALID(a)); 1497 assert(BC_NUM_RDX_VALID(b)); 1498 1499 bc_num_zero(c); 1500 1501 ascale = a->scale; 1502 bscale = b->scale; 1503 1504 // This sets the final scale according to the bc spec. 1505 scale1 = BC_MAX(scale, ascale); 1506 scale2 = BC_MAX(scale1, bscale); 1507 rscale = ascale + bscale; 1508 realscale = BC_MIN(rscale, scale2); 1509 1510 // If this condition is true, we can use bc_num_mulArray(), which would be 1511 // much faster. 1512 if ((a->len == 1 || b->len == 1) && !a->rdx && !b->rdx) 1513 { 1514 BcNum* operand; 1515 BcBigDig dig; 1516 1517 // Set the correct operands. 1518 if (a->len == 1) 1519 { 1520 dig = (BcBigDig) a->num[0]; 1521 operand = b; 1522 } 1523 else 1524 { 1525 dig = (BcBigDig) b->num[0]; 1526 operand = a; 1527 } 1528 1529 bc_num_mulArray(operand, dig, c); 1530 1531 // Need to make sure the sign is correct. 1532 if (BC_NUM_NONZERO(c)) 1533 { 1534 c->rdx = BC_NUM_NEG_VAL(c, BC_NUM_NEG(a) != BC_NUM_NEG(b)); 1535 } 1536 1537 return; 1538 } 1539 1540 assert(BC_NUM_RDX_VALID(a)); 1541 assert(BC_NUM_RDX_VALID(b)); 1542 1543 BC_SIG_LOCK; 1544 1545 // We need copies because of all of the mutation needed to make Karatsuba 1546 // think the numbers are integers. 1547 bc_num_init(&cpa, a->len + BC_NUM_RDX_VAL(a)); 1548 bc_num_init(&cpb, b->len + BC_NUM_RDX_VAL(b)); 1549 1550 BC_SETJMP_LOCKED(vm, init_err); 1551 1552 BC_SIG_UNLOCK; 1553 1554 bc_num_copy(&cpa, a); 1555 bc_num_copy(&cpb, b); 1556 1557 assert(BC_NUM_RDX_VALID_NP(cpa)); 1558 assert(BC_NUM_RDX_VALID_NP(cpb)); 1559 1560 BC_NUM_NEG_CLR_NP(cpa); 1561 BC_NUM_NEG_CLR_NP(cpb); 1562 1563 assert(BC_NUM_RDX_VALID_NP(cpa)); 1564 assert(BC_NUM_RDX_VALID_NP(cpb)); 1565 1566 // These are what makes them appear like integers. 1567 ardx = BC_NUM_RDX_VAL_NP(cpa) * BC_BASE_DIGS; 1568 bc_num_shiftLeft(&cpa, ardx); 1569 1570 brdx = BC_NUM_RDX_VAL_NP(cpb) * BC_BASE_DIGS; 1571 bc_num_shiftLeft(&cpb, brdx); 1572 1573 // We need to reset the jump here because azero and bzero are used in the 1574 // cleanup, and local variables are not guaranteed to be the same after a 1575 // jump. 1576 BC_SIG_LOCK; 1577 1578 BC_UNSETJMP(vm); 1579 1580 // We want to ignore zero limbs. 1581 azero = bc_num_shiftZero(&cpa); 1582 bzero = bc_num_shiftZero(&cpb); 1583 1584 BC_SETJMP_LOCKED(vm, err); 1585 1586 BC_SIG_UNLOCK; 1587 1588 bc_num_clean(&cpa); 1589 bc_num_clean(&cpb); 1590 1591 bc_num_k(&cpa, &cpb, c); 1592 1593 // The return parameter needs to have its scale set. This is the start. It 1594 // also needs to be shifted by the same amount as a and b have limbs after 1595 // the decimal point. 1596 zero = bc_vm_growSize(azero, bzero); 1597 len = bc_vm_growSize(c->len, zero); 1598 1599 bc_num_expand(c, len); 1600 1601 // Shift c based on the limbs after the decimal point in a and b. 1602 bc_num_shiftLeft(c, (len - c->len) * BC_BASE_DIGS); 1603 bc_num_shiftRight(c, ardx + brdx); 1604 1605 bc_num_retireMul(c, realscale, BC_NUM_NEG(a), BC_NUM_NEG(b)); 1606 1607 err: 1608 BC_SIG_MAYLOCK; 1609 bc_num_unshiftZero(&cpb, bzero); 1610 bc_num_unshiftZero(&cpa, azero); 1611 init_err: 1612 BC_SIG_MAYLOCK; 1613 bc_num_free(&cpb); 1614 bc_num_free(&cpa); 1615 BC_LONGJMP_CONT(vm); 1616 } 1617 1618 /** 1619 * Returns true if the BcDig array has non-zero limbs, false otherwise. 1620 * @param a The array to test. 1621 * @param len The length of the array. 1622 * @return True if @a has any non-zero limbs, false otherwise. 1623 */ 1624 static bool 1625 bc_num_nonZeroDig(BcDig* restrict a, size_t len) 1626 { 1627 size_t i; 1628 1629 for (i = len - 1; i < len; --i) 1630 { 1631 if (a[i] != 0) return true; 1632 } 1633 1634 return false; 1635 } 1636 1637 /** 1638 * Compares a BcDig array against a BcNum. This is especially suited for 1639 * division. Returns >0 if @a a is greater than @a b, <0 if it is less, and =0 1640 * if they are equal. 1641 * @param a The array. 1642 * @param b The number. 1643 * @param len The length to assume the arrays are. This is always less than the 1644 * actual length because of how this is implemented. 1645 */ 1646 static ssize_t 1647 bc_num_divCmp(const BcDig* a, const BcNum* b, size_t len) 1648 { 1649 ssize_t cmp; 1650 1651 if (b->len > len && a[len]) cmp = bc_num_compare(a, b->num, len + 1); 1652 else if (b->len <= len) 1653 { 1654 if (a[len]) cmp = 1; 1655 else cmp = bc_num_compare(a, b->num, len); 1656 } 1657 else cmp = -1; 1658 1659 return cmp; 1660 } 1661 1662 /** 1663 * Extends the two operands of a division by BC_BASE_DIGS minus the number of 1664 * digits in the divisor estimate. In other words, it is shifting the numbers in 1665 * order to force the divisor estimate to fill the limb. 1666 * @param a The first operand. 1667 * @param b The second operand. 1668 * @param divisor The divisor estimate. 1669 */ 1670 static void 1671 bc_num_divExtend(BcNum* restrict a, BcNum* restrict b, BcBigDig divisor) 1672 { 1673 size_t pow; 1674 1675 assert(divisor < BC_BASE_POW); 1676 1677 pow = BC_BASE_DIGS - bc_num_log10((size_t) divisor); 1678 1679 bc_num_shiftLeft(a, pow); 1680 bc_num_shiftLeft(b, pow); 1681 } 1682 1683 /** 1684 * Actually does division. This is a rewrite of my original code by Stefan Esser 1685 * from FreeBSD. 1686 * @param a The first operand. 1687 * @param b The second operand. 1688 * @param c The return parameter. 1689 * @param scale The current scale. 1690 */ 1691 static void 1692 bc_num_d_long(BcNum* restrict a, BcNum* restrict b, BcNum* restrict c, 1693 size_t scale) 1694 { 1695 BcBigDig divisor; 1696 size_t i, rdx; 1697 // This is volatile and len 2 and reallen exist to quiet the GCC warning 1698 // about clobbering on longjmp(). This one is possible, I think. 1699 volatile size_t len; 1700 size_t len2, reallen; 1701 // This is volatile and realend exists to quiet the GCC warning about 1702 // clobbering on longjmp(). This one is possible, I think. 1703 volatile size_t end; 1704 size_t realend; 1705 BcNum cpb; 1706 // This is volatile and realnonzero exists to quiet the GCC warning about 1707 // clobbering on longjmp(). This one is possible, I think. 1708 volatile bool nonzero; 1709 bool realnonzero; 1710 #if BC_ENABLE_LIBRARY 1711 BcVm* vm = bcl_getspecific(); 1712 #endif // BC_ENABLE_LIBRARY 1713 1714 assert(b->len < a->len); 1715 1716 len = b->len; 1717 end = a->len - len; 1718 1719 assert(len >= 1); 1720 1721 // This is a final time to make sure c is big enough and that its array is 1722 // properly zeroed. 1723 bc_num_expand(c, a->len); 1724 // NOLINTNEXTLINE 1725 memset(c->num, 0, c->cap * sizeof(BcDig)); 1726 1727 // Setup. 1728 BC_NUM_RDX_SET(c, BC_NUM_RDX_VAL(a)); 1729 c->scale = a->scale; 1730 c->len = a->len; 1731 1732 // This is pulling the most significant limb of b in order to establish a 1733 // good "estimate" for the actual divisor. 1734 divisor = (BcBigDig) b->num[len - 1]; 1735 1736 // The entire bit of code in this if statement is to tighten the estimate of 1737 // the divisor. The condition asks if b has any other non-zero limbs. 1738 if (len > 1 && bc_num_nonZeroDig(b->num, len - 1)) 1739 { 1740 // This takes a little bit of understanding. The "10*BC_BASE_DIGS/6+1" 1741 // results in either 16 for 64-bit 9-digit limbs or 7 for 32-bit 4-digit 1742 // limbs. Then it shifts a 1 by that many, which in both cases, puts the 1743 // result above *half* of the max value a limb can store. Basically, 1744 // this quickly calculates if the divisor is greater than half the max 1745 // of a limb. 1746 nonzero = (divisor > 1 << ((10 * BC_BASE_DIGS) / 6 + 1)); 1747 1748 // If the divisor is *not* greater than half the limb... 1749 if (!nonzero) 1750 { 1751 // Extend the parameters by the number of missing digits in the 1752 // divisor. 1753 bc_num_divExtend(a, b, divisor); 1754 1755 // Check bc_num_d(). In there, we grow a again and again. We do it 1756 // again here; we *always* want to be sure it is big enough. 1757 len2 = BC_MAX(a->len, b->len); 1758 bc_num_expand(a, len2 + 1); 1759 1760 // Make a have a zero most significant limb to match the len. 1761 if (len2 + 1 > a->len) a->len = len2 + 1; 1762 1763 // Grab the new divisor estimate, new because the shift has made it 1764 // different. 1765 reallen = b->len; 1766 realend = a->len - reallen; 1767 divisor = (BcBigDig) b->num[reallen - 1]; 1768 1769 realnonzero = bc_num_nonZeroDig(b->num, reallen - 1); 1770 } 1771 else 1772 { 1773 realend = end; 1774 realnonzero = nonzero; 1775 } 1776 } 1777 else 1778 { 1779 realend = end; 1780 realnonzero = false; 1781 } 1782 1783 // If b has other nonzero limbs, we want the divisor to be one higher, so 1784 // that it is an upper bound. 1785 divisor += realnonzero; 1786 1787 // Make sure c can fit the new length. 1788 bc_num_expand(c, a->len); 1789 // NOLINTNEXTLINE 1790 memset(c->num, 0, BC_NUM_SIZE(c->cap)); 1791 1792 assert(c->scale >= scale); 1793 rdx = BC_NUM_RDX_VAL(c) - BC_NUM_RDX(scale); 1794 1795 BC_SIG_LOCK; 1796 1797 bc_num_init(&cpb, len + 1); 1798 1799 BC_SETJMP_LOCKED(vm, err); 1800 1801 BC_SIG_UNLOCK; 1802 1803 // This is the actual division loop. 1804 for (i = realend - 1; i < realend && i >= rdx && BC_NUM_NONZERO(a); --i) 1805 { 1806 ssize_t cmp; 1807 BcDig* n; 1808 BcBigDig result; 1809 1810 n = a->num + i; 1811 assert(n >= a->num); 1812 result = 0; 1813 1814 cmp = bc_num_divCmp(n, b, len); 1815 1816 // This is true if n is greater than b, which means that division can 1817 // proceed, so this inner loop is the part that implements one instance 1818 // of the division. 1819 while (cmp >= 0) 1820 { 1821 BcBigDig n1, dividend, quotient; 1822 1823 // These should be named obviously enough. Just imagine that it's a 1824 // division of one limb. Because that's what it is. 1825 n1 = (BcBigDig) n[len]; 1826 dividend = n1 * BC_BASE_POW + (BcBigDig) n[len - 1]; 1827 quotient = (dividend / divisor); 1828 1829 // If this is true, then we can just subtract. Remember: setting 1830 // quotient to 1 is not bad because we already know that n is 1831 // greater than b. 1832 if (quotient <= 1) 1833 { 1834 quotient = 1; 1835 bc_num_subArrays(n, b->num, len); 1836 } 1837 else 1838 { 1839 assert(quotient <= BC_BASE_POW); 1840 1841 // We need to multiply and subtract for a quotient above 1. 1842 bc_num_mulArray(b, (BcBigDig) quotient, &cpb); 1843 bc_num_subArrays(n, cpb.num, cpb.len); 1844 } 1845 1846 // The result is the *real* quotient, by the way, but it might take 1847 // multiple trips around this loop to get it. 1848 result += quotient; 1849 assert(result <= BC_BASE_POW); 1850 1851 // And here's why it might take multiple trips: n might *still* be 1852 // greater than b. So we have to loop again. That's what this is 1853 // setting up for: the condition of the while loop. 1854 if (realnonzero) cmp = bc_num_divCmp(n, b, len); 1855 else cmp = -1; 1856 } 1857 1858 assert(result < BC_BASE_POW); 1859 1860 // Store the actual limb quotient. 1861 c->num[i] = (BcDig) result; 1862 } 1863 1864 err: 1865 BC_SIG_MAYLOCK; 1866 bc_num_free(&cpb); 1867 BC_LONGJMP_CONT(vm); 1868 } 1869 1870 /** 1871 * Implements division. This is a BcNumBinOp function. 1872 * @param a The first operand. 1873 * @param b The second operand. 1874 * @param c The return parameter. 1875 * @param scale The current scale. 1876 */ 1877 static void 1878 bc_num_d(BcNum* a, BcNum* b, BcNum* restrict c, size_t scale) 1879 { 1880 size_t len, cpardx; 1881 BcNum cpa, cpb; 1882 #if BC_ENABLE_LIBRARY 1883 BcVm* vm = bcl_getspecific(); 1884 #endif // BC_ENABLE_LIBRARY 1885 1886 if (BC_NUM_ZERO(b)) bc_err(BC_ERR_MATH_DIVIDE_BY_ZERO); 1887 1888 if (BC_NUM_ZERO(a)) 1889 { 1890 bc_num_setToZero(c, scale); 1891 return; 1892 } 1893 1894 if (BC_NUM_ONE(b)) 1895 { 1896 bc_num_copy(c, a); 1897 bc_num_retireMul(c, scale, BC_NUM_NEG(a), BC_NUM_NEG(b)); 1898 return; 1899 } 1900 1901 // If this is true, we can use bc_num_divArray(), which would be faster. 1902 if (!BC_NUM_RDX_VAL(a) && !BC_NUM_RDX_VAL(b) && b->len == 1 && !scale) 1903 { 1904 BcBigDig rem; 1905 bc_num_divArray(a, (BcBigDig) b->num[0], c, &rem); 1906 bc_num_retireMul(c, scale, BC_NUM_NEG(a), BC_NUM_NEG(b)); 1907 return; 1908 } 1909 1910 len = bc_num_divReq(a, b, scale); 1911 1912 BC_SIG_LOCK; 1913 1914 // Initialize copies of the parameters. We want the length of the first 1915 // operand copy to be as big as the result because of the way the division 1916 // is implemented. 1917 bc_num_init(&cpa, len); 1918 bc_num_copy(&cpa, a); 1919 bc_num_createCopy(&cpb, b); 1920 1921 BC_SETJMP_LOCKED(vm, err); 1922 1923 BC_SIG_UNLOCK; 1924 1925 len = b->len; 1926 1927 // Like the above comment, we want the copy of the first parameter to be 1928 // larger than the second parameter. 1929 if (len > cpa.len) 1930 { 1931 bc_num_expand(&cpa, bc_vm_growSize(len, 2)); 1932 bc_num_extend(&cpa, (len - cpa.len) * BC_BASE_DIGS); 1933 } 1934 1935 cpardx = BC_NUM_RDX_VAL_NP(cpa); 1936 cpa.scale = cpardx * BC_BASE_DIGS; 1937 1938 // This is just setting up the scale in preparation for the division. 1939 bc_num_extend(&cpa, b->scale); 1940 cpardx = BC_NUM_RDX_VAL_NP(cpa) - BC_NUM_RDX(b->scale); 1941 BC_NUM_RDX_SET_NP(cpa, cpardx); 1942 cpa.scale = cpardx * BC_BASE_DIGS; 1943 1944 // Once again, just setting things up, this time to match scale. 1945 if (scale > cpa.scale) 1946 { 1947 bc_num_extend(&cpa, scale); 1948 cpardx = BC_NUM_RDX_VAL_NP(cpa); 1949 cpa.scale = cpardx * BC_BASE_DIGS; 1950 } 1951 1952 // Grow if necessary. 1953 if (cpa.cap == cpa.len) bc_num_expand(&cpa, bc_vm_growSize(cpa.len, 1)); 1954 1955 // We want an extra zero in front to make things simpler. 1956 cpa.num[cpa.len++] = 0; 1957 1958 // Still setting things up. Why all of these things are needed is not 1959 // something that can be easily explained, but it has to do with making the 1960 // actual algorithm easier to understand because it can assume a lot of 1961 // things. Thus, you should view all of this setup code as establishing 1962 // assumptions for bc_num_d_long(), where the actual division happens. 1963 // 1964 // But in short, this setup makes it so bc_num_d_long() can pretend the 1965 // numbers are integers. 1966 if (cpardx == cpa.len) cpa.len = bc_num_nonZeroLen(&cpa); 1967 if (BC_NUM_RDX_VAL_NP(cpb) == cpb.len) cpb.len = bc_num_nonZeroLen(&cpb); 1968 cpb.scale = 0; 1969 BC_NUM_RDX_SET_NP(cpb, 0); 1970 1971 bc_num_d_long(&cpa, &cpb, c, scale); 1972 1973 bc_num_retireMul(c, scale, BC_NUM_NEG(a), BC_NUM_NEG(b)); 1974 1975 err: 1976 BC_SIG_MAYLOCK; 1977 bc_num_free(&cpb); 1978 bc_num_free(&cpa); 1979 BC_LONGJMP_CONT(vm); 1980 } 1981 1982 /** 1983 * Implements divmod. This is the actual modulus function; since modulus 1984 * requires a division anyway, this returns the quotient and modulus. Either can 1985 * be thrown out as desired. 1986 * @param a The first operand. 1987 * @param b The second operand. 1988 * @param c The return parameter for the quotient. 1989 * @param d The return parameter for the modulus. 1990 * @param scale The current scale. 1991 * @param ts The scale that the operation should be done to. Yes, it's not 1992 * necessarily the same as scale, per the bc spec. 1993 */ 1994 static void 1995 bc_num_r(BcNum* a, BcNum* b, BcNum* restrict c, BcNum* restrict d, size_t scale, 1996 size_t ts) 1997 { 1998 BcNum temp; 1999 // realscale is meant to quiet a warning on GCC about longjmp() clobbering. 2000 // This one is real. 2001 size_t realscale; 2002 bool neg; 2003 #if BC_ENABLE_LIBRARY 2004 BcVm* vm = bcl_getspecific(); 2005 #endif // BC_ENABLE_LIBRARY 2006 2007 if (BC_NUM_ZERO(b)) bc_err(BC_ERR_MATH_DIVIDE_BY_ZERO); 2008 2009 if (BC_NUM_ZERO(a)) 2010 { 2011 bc_num_setToZero(c, ts); 2012 bc_num_setToZero(d, ts); 2013 return; 2014 } 2015 2016 BC_SIG_LOCK; 2017 2018 bc_num_init(&temp, d->cap); 2019 2020 BC_SETJMP_LOCKED(vm, err); 2021 2022 BC_SIG_UNLOCK; 2023 2024 // Division. 2025 bc_num_d(a, b, c, scale); 2026 2027 // We want an extra digit so we can safely truncate. 2028 if (scale) realscale = ts + 1; 2029 else realscale = scale; 2030 2031 assert(BC_NUM_RDX_VALID(c)); 2032 assert(BC_NUM_RDX_VALID(b)); 2033 2034 // Implement the rest of the (a - (a / b) * b) formula. 2035 bc_num_m(c, b, &temp, realscale); 2036 bc_num_sub(a, &temp, d, realscale); 2037 2038 // Extend if necessary. 2039 if (ts > d->scale && BC_NUM_NONZERO(d)) bc_num_extend(d, ts - d->scale); 2040 2041 neg = BC_NUM_NEG(d); 2042 bc_num_retireMul(d, ts, BC_NUM_NEG(a), BC_NUM_NEG(b)); 2043 d->rdx = BC_NUM_NEG_VAL(d, BC_NUM_NONZERO(d) ? neg : false); 2044 2045 err: 2046 BC_SIG_MAYLOCK; 2047 bc_num_free(&temp); 2048 BC_LONGJMP_CONT(vm); 2049 } 2050 2051 /** 2052 * Implements modulus/remainder. (Yes, I know they are different, but not in the 2053 * context of bc.) This is a BcNumBinOp function. 2054 * @param a The first operand. 2055 * @param b The second operand. 2056 * @param c The return parameter. 2057 * @param scale The current scale. 2058 */ 2059 static void 2060 bc_num_rem(BcNum* a, BcNum* b, BcNum* restrict c, size_t scale) 2061 { 2062 BcNum c1; 2063 size_t ts; 2064 #if BC_ENABLE_LIBRARY 2065 BcVm* vm = bcl_getspecific(); 2066 #endif // BC_ENABLE_LIBRARY 2067 2068 ts = bc_vm_growSize(scale, b->scale); 2069 ts = BC_MAX(ts, a->scale); 2070 2071 BC_SIG_LOCK; 2072 2073 // Need a temp for the quotient. 2074 bc_num_init(&c1, bc_num_mulReq(a, b, ts)); 2075 2076 BC_SETJMP_LOCKED(vm, err); 2077 2078 BC_SIG_UNLOCK; 2079 2080 bc_num_r(a, b, &c1, c, scale, ts); 2081 2082 err: 2083 BC_SIG_MAYLOCK; 2084 bc_num_free(&c1); 2085 BC_LONGJMP_CONT(vm); 2086 } 2087 2088 /** 2089 * Implements power (exponentiation). This is a BcNumBinOp function. 2090 * @param a The first operand. 2091 * @param b The second operand. 2092 * @param c The return parameter. 2093 * @param scale The current scale. 2094 */ 2095 static void 2096 bc_num_p(BcNum* a, BcNum* b, BcNum* restrict c, size_t scale) 2097 { 2098 BcNum copy, btemp; 2099 BcBigDig exp; 2100 // realscale is meant to quiet a warning on GCC about longjmp() clobbering. 2101 // This one is real. 2102 size_t powrdx, resrdx, realscale; 2103 bool neg; 2104 #if BC_ENABLE_LIBRARY 2105 BcVm* vm = bcl_getspecific(); 2106 #endif // BC_ENABLE_LIBRARY 2107 2108 // This is here to silence a warning from GCC. 2109 #if BC_GCC 2110 btemp.len = 0; 2111 btemp.rdx = 0; 2112 btemp.num = NULL; 2113 #endif // BC_GCC 2114 2115 if (BC_ERR(bc_num_nonInt(b, &btemp))) bc_err(BC_ERR_MATH_NON_INTEGER); 2116 2117 assert(btemp.len == 0 || btemp.num != NULL); 2118 2119 if (BC_NUM_ZERO(&btemp)) 2120 { 2121 bc_num_one(c); 2122 return; 2123 } 2124 2125 if (BC_NUM_ZERO(a)) 2126 { 2127 if (BC_NUM_NEG_NP(btemp)) bc_err(BC_ERR_MATH_DIVIDE_BY_ZERO); 2128 bc_num_setToZero(c, scale); 2129 return; 2130 } 2131 2132 if (BC_NUM_ONE(&btemp)) 2133 { 2134 if (!BC_NUM_NEG_NP(btemp)) bc_num_copy(c, a); 2135 else bc_num_inv(a, c, scale); 2136 return; 2137 } 2138 2139 neg = BC_NUM_NEG_NP(btemp); 2140 BC_NUM_NEG_CLR_NP(btemp); 2141 2142 exp = bc_num_bigdig(&btemp); 2143 2144 BC_SIG_LOCK; 2145 2146 bc_num_createCopy(©, a); 2147 2148 BC_SETJMP_LOCKED(vm, err); 2149 2150 BC_SIG_UNLOCK; 2151 2152 // If this is true, then we do not have to do a division, and we need to 2153 // set scale accordingly. 2154 if (!neg) 2155 { 2156 size_t max = BC_MAX(scale, a->scale), scalepow; 2157 scalepow = bc_num_mulOverflow(a->scale, exp); 2158 realscale = BC_MIN(scalepow, max); 2159 } 2160 else realscale = scale; 2161 2162 // This is only implementing the first exponentiation by squaring, until it 2163 // reaches the first time where the square is actually used. 2164 for (powrdx = a->scale; !(exp & 1); exp >>= 1) 2165 { 2166 powrdx <<= 1; 2167 assert(BC_NUM_RDX_VALID_NP(copy)); 2168 bc_num_mul(©, ©, ©, powrdx); 2169 } 2170 2171 // Make c a copy of copy for the purpose of saving the squares that should 2172 // be saved. 2173 bc_num_copy(c, ©); 2174 resrdx = powrdx; 2175 2176 // Now finish the exponentiation by squaring, this time saving the squares 2177 // as necessary. 2178 while (exp >>= 1) 2179 { 2180 powrdx <<= 1; 2181 assert(BC_NUM_RDX_VALID_NP(copy)); 2182 bc_num_mul(©, ©, ©, powrdx); 2183 2184 // If this is true, we want to save that particular square. This does 2185 // that by multiplying c with copy. 2186 if (exp & 1) 2187 { 2188 resrdx += powrdx; 2189 assert(BC_NUM_RDX_VALID(c)); 2190 assert(BC_NUM_RDX_VALID_NP(copy)); 2191 bc_num_mul(c, ©, c, resrdx); 2192 } 2193 } 2194 2195 // Invert if necessary. 2196 if (neg) bc_num_inv(c, c, realscale); 2197 2198 // Truncate if necessary. 2199 if (c->scale > realscale) bc_num_truncate(c, c->scale - realscale); 2200 2201 bc_num_clean(c); 2202 2203 err: 2204 BC_SIG_MAYLOCK; 2205 bc_num_free(©); 2206 BC_LONGJMP_CONT(vm); 2207 } 2208 2209 #if BC_ENABLE_EXTRA_MATH 2210 /** 2211 * Implements the places operator. This is a BcNumBinOp function. 2212 * @param a The first operand. 2213 * @param b The second operand. 2214 * @param c The return parameter. 2215 * @param scale The current scale. 2216 */ 2217 static void 2218 bc_num_place(BcNum* a, BcNum* b, BcNum* restrict c, size_t scale) 2219 { 2220 BcBigDig val; 2221 2222 BC_UNUSED(scale); 2223 2224 val = bc_num_intop(a, b, c); 2225 2226 // Just truncate or extend as appropriate. 2227 if (val < c->scale) bc_num_truncate(c, c->scale - val); 2228 else if (val > c->scale) bc_num_extend(c, val - c->scale); 2229 } 2230 2231 /** 2232 * Implements the left shift operator. This is a BcNumBinOp function. 2233 */ 2234 static void 2235 bc_num_left(BcNum* a, BcNum* b, BcNum* restrict c, size_t scale) 2236 { 2237 BcBigDig val; 2238 2239 BC_UNUSED(scale); 2240 2241 val = bc_num_intop(a, b, c); 2242 2243 bc_num_shiftLeft(c, (size_t) val); 2244 } 2245 2246 /** 2247 * Implements the right shift operator. This is a BcNumBinOp function. 2248 */ 2249 static void 2250 bc_num_right(BcNum* a, BcNum* b, BcNum* restrict c, size_t scale) 2251 { 2252 BcBigDig val; 2253 2254 BC_UNUSED(scale); 2255 2256 val = bc_num_intop(a, b, c); 2257 2258 if (BC_NUM_ZERO(c)) return; 2259 2260 bc_num_shiftRight(c, (size_t) val); 2261 } 2262 #endif // BC_ENABLE_EXTRA_MATH 2263 2264 /** 2265 * Prepares for, and calls, a binary operator function. This is probably the 2266 * most important function in the entire file because it establishes assumptions 2267 * that make the rest of the code so easy. Those assumptions include: 2268 * 2269 * - a is not the same pointer as c. 2270 * - b is not the same pointer as c. 2271 * - there is enough room in c for the result. 2272 * 2273 * Without these, this whole function would basically have to be duplicated for 2274 * *all* binary operators. 2275 * 2276 * @param a The first operand. 2277 * @param b The second operand. 2278 * @param c The return parameter. 2279 * @param scale The current scale. 2280 * @param req The number of limbs needed to fit the result. 2281 */ 2282 static void 2283 bc_num_binary(BcNum* a, BcNum* b, BcNum* c, size_t scale, BcNumBinOp op, 2284 size_t req) 2285 { 2286 BcNum* ptr_a; 2287 BcNum* ptr_b; 2288 BcNum num2; 2289 #if BC_ENABLE_LIBRARY 2290 BcVm* vm = NULL; 2291 #endif // BC_ENABLE_LIBRARY 2292 2293 assert(a != NULL && b != NULL && c != NULL && op != NULL); 2294 2295 assert(BC_NUM_RDX_VALID(a)); 2296 assert(BC_NUM_RDX_VALID(b)); 2297 2298 BC_SIG_LOCK; 2299 2300 ptr_a = c == a ? &num2 : a; 2301 ptr_b = c == b ? &num2 : b; 2302 2303 // Actually reallocate. If we don't reallocate, we want to expand at the 2304 // very least. 2305 if (c == a || c == b) 2306 { 2307 #if BC_ENABLE_LIBRARY 2308 vm = bcl_getspecific(); 2309 #endif // BC_ENABLE_LIBRARY 2310 2311 // NOLINTNEXTLINE 2312 memcpy(&num2, c, sizeof(BcNum)); 2313 2314 bc_num_init(c, req); 2315 2316 // Must prepare for cleanup. We want this here so that locals that got 2317 // set stay set since a longjmp() is not guaranteed to preserve locals. 2318 BC_SETJMP_LOCKED(vm, err); 2319 BC_SIG_UNLOCK; 2320 } 2321 else 2322 { 2323 BC_SIG_UNLOCK; 2324 bc_num_expand(c, req); 2325 } 2326 2327 // It is okay for a and b to be the same. If a binary operator function does 2328 // need them to be different, the binary operator function is responsible 2329 // for that. 2330 2331 // Call the actual binary operator function. 2332 op(ptr_a, ptr_b, c, scale); 2333 2334 assert(!BC_NUM_NEG(c) || BC_NUM_NONZERO(c)); 2335 assert(BC_NUM_RDX_VAL(c) <= c->len || !c->len); 2336 assert(BC_NUM_RDX_VALID(c)); 2337 assert(!c->len || c->num[c->len - 1] || BC_NUM_RDX_VAL(c) == c->len); 2338 2339 err: 2340 // Cleanup only needed if we initialized c to a new number. 2341 if (c == a || c == b) 2342 { 2343 BC_SIG_MAYLOCK; 2344 bc_num_free(&num2); 2345 BC_LONGJMP_CONT(vm); 2346 } 2347 } 2348 2349 /** 2350 * Tests a number string for validity. This function has a history; I originally 2351 * wrote it because I did not trust my parser. Over time, however, I came to 2352 * trust it, so I was able to relegate this function to debug builds only, and I 2353 * used it in assert()'s. But then I created the library, and well, I can't 2354 * trust users, so I reused this for yelling at users. 2355 * @param val The string to check to see if it's a valid number string. 2356 * @return True if the string is a valid number string, false otherwise. 2357 */ 2358 bool 2359 bc_num_strValid(const char* restrict val) 2360 { 2361 bool radix = false; 2362 size_t i, len = strlen(val); 2363 2364 // Notice that I don't check if there is a negative sign. That is not part 2365 // of a valid number, except in the library. The library-specific code takes 2366 // care of that part. 2367 2368 // Nothing in the string is okay. 2369 if (!len) return true; 2370 2371 // Loop through the characters. 2372 for (i = 0; i < len; ++i) 2373 { 2374 BcDig c = val[i]; 2375 2376 // If we have found a radix point... 2377 if (c == '.') 2378 { 2379 // We don't allow two radices. 2380 if (radix) return false; 2381 2382 radix = true; 2383 continue; 2384 } 2385 2386 // We only allow digits and uppercase letters. 2387 if (!(isdigit(c) || isupper(c))) return false; 2388 } 2389 2390 return true; 2391 } 2392 2393 /** 2394 * Parses one character and returns the digit that corresponds to that 2395 * character according to the base. 2396 * @param c The character to parse. 2397 * @param base The base. 2398 * @return The character as a digit. 2399 */ 2400 static BcBigDig 2401 bc_num_parseChar(char c, size_t base) 2402 { 2403 assert(isupper(c) || isdigit(c)); 2404 2405 // If a letter... 2406 if (isupper(c)) 2407 { 2408 #if BC_ENABLE_LIBRARY 2409 BcVm* vm = bcl_getspecific(); 2410 #endif // BC_ENABLE_LIBRARY 2411 2412 // This returns the digit that directly corresponds with the letter. 2413 c = BC_NUM_NUM_LETTER(c); 2414 2415 // If the digit is greater than the base, we clamp. 2416 if (BC_DIGIT_CLAMP) 2417 { 2418 c = ((size_t) c) >= base ? (char) base - 1 : c; 2419 } 2420 } 2421 // Straight convert the digit to a number. 2422 else c -= '0'; 2423 2424 return (BcBigDig) (uchar) c; 2425 } 2426 2427 /** 2428 * Parses a string as a decimal number. This is separate because it's going to 2429 * be the most used, and it can be heavily optimized for decimal only. 2430 * @param n The number to parse into and return. Must be preallocated. 2431 * @param val The string to parse. 2432 */ 2433 static void 2434 bc_num_parseDecimal(BcNum* restrict n, const char* restrict val) 2435 { 2436 size_t len, i, temp, mod; 2437 const char* ptr; 2438 bool zero = true, rdx; 2439 #if BC_ENABLE_LIBRARY 2440 BcVm* vm = bcl_getspecific(); 2441 #endif // BC_ENABLE_LIBRARY 2442 2443 // Eat leading zeroes. 2444 for (i = 0; val[i] == '0'; ++i) 2445 { 2446 continue; 2447 } 2448 2449 val += i; 2450 assert(!val[0] || isalnum(val[0]) || val[0] == '.'); 2451 2452 // All 0's. We can just return, since this procedure expects a virgin 2453 // (already 0) BcNum. 2454 if (!val[0]) return; 2455 2456 // The length of the string is the length of the number, except it might be 2457 // one bigger because of a decimal point. 2458 len = strlen(val); 2459 2460 // Find the location of the decimal point. 2461 ptr = strchr(val, '.'); 2462 rdx = (ptr != NULL); 2463 2464 // We eat leading zeroes again. These leading zeroes are different because 2465 // they will come after the decimal point if they exist, and since that's 2466 // the case, they must be preserved. 2467 for (i = 0; i < len && (zero = (val[i] == '0' || val[i] == '.')); ++i) 2468 { 2469 continue; 2470 } 2471 2472 // Set the scale of the number based on the location of the decimal point. 2473 // The casts to uintptr_t is to ensure that bc does not hit undefined 2474 // behavior when doing math on the values. 2475 n->scale = (size_t) (rdx * 2476 (((uintptr_t) (val + len)) - (((uintptr_t) ptr) + 1))); 2477 2478 // Set rdx. 2479 BC_NUM_RDX_SET(n, BC_NUM_RDX(n->scale)); 2480 2481 // Calculate length. First, the length of the integer, then the number of 2482 // digits in the last limb, then the length. 2483 i = len - (ptr == val ? 0 : i) - rdx; 2484 temp = BC_NUM_ROUND_POW(i); 2485 mod = n->scale % BC_BASE_DIGS; 2486 i = mod ? BC_BASE_DIGS - mod : 0; 2487 n->len = ((temp + i) / BC_BASE_DIGS); 2488 2489 // Expand and zero. The plus extra is in case the lack of clamping causes 2490 // the number to overflow the original bounds. 2491 bc_num_expand(n, n->len + !BC_DIGIT_CLAMP); 2492 // NOLINTNEXTLINE 2493 memset(n->num, 0, BC_NUM_SIZE(n->len + !BC_DIGIT_CLAMP)); 2494 2495 if (zero) 2496 { 2497 // I think I can set rdx directly to zero here because n should be a 2498 // new number with sign set to false. 2499 n->len = n->rdx = 0; 2500 } 2501 else 2502 { 2503 // There is actually stuff to parse if we make it here. Yay... 2504 BcBigDig exp, pow; 2505 2506 assert(i <= BC_NUM_BIGDIG_MAX); 2507 2508 // The exponent and power. 2509 exp = (BcBigDig) i; 2510 pow = bc_num_pow10[exp]; 2511 2512 // Parse loop. We parse backwards because numbers are stored little 2513 // endian. 2514 for (i = len - 1; i < len; --i, ++exp) 2515 { 2516 char c = val[i]; 2517 2518 // Skip the decimal point. 2519 if (c == '.') exp -= 1; 2520 else 2521 { 2522 // The index of the limb. 2523 size_t idx = exp / BC_BASE_DIGS; 2524 BcBigDig dig; 2525 2526 if (isupper(c)) 2527 { 2528 // Clamp for the base. 2529 if (!BC_DIGIT_CLAMP) c = BC_NUM_NUM_LETTER(c); 2530 else c = 9; 2531 } 2532 else c -= '0'; 2533 2534 // Add the digit to the limb. This takes care of overflow from 2535 // lack of clamping. 2536 dig = ((BcBigDig) n->num[idx]) + ((BcBigDig) c) * pow; 2537 if (dig >= BC_BASE_POW) 2538 { 2539 // We cannot go over BC_BASE_POW with clamping. 2540 assert(!BC_DIGIT_CLAMP); 2541 2542 n->num[idx + 1] = (BcDig) (dig / BC_BASE_POW); 2543 n->num[idx] = (BcDig) (dig % BC_BASE_POW); 2544 assert(n->num[idx] >= 0 && n->num[idx] < BC_BASE_POW); 2545 assert(n->num[idx + 1] >= 0 && 2546 n->num[idx + 1] < BC_BASE_POW); 2547 } 2548 else 2549 { 2550 n->num[idx] = (BcDig) dig; 2551 assert(n->num[idx] >= 0 && n->num[idx] < BC_BASE_POW); 2552 } 2553 2554 // Adjust the power and exponent. 2555 if ((exp + 1) % BC_BASE_DIGS == 0) pow = 1; 2556 else pow *= BC_BASE; 2557 } 2558 } 2559 } 2560 2561 // Make sure to add one to the length if needed from lack of clamping. 2562 n->len += (!BC_DIGIT_CLAMP && n->num[n->len] != 0); 2563 } 2564 2565 /** 2566 * Parse a number in any base (besides decimal). 2567 * @param n The number to parse into and return. Must be preallocated. 2568 * @param val The string to parse. 2569 * @param base The base to parse as. 2570 */ 2571 static void 2572 bc_num_parseBase(BcNum* restrict n, const char* restrict val, BcBigDig base) 2573 { 2574 BcNum temp, mult1, mult2, result1, result2; 2575 BcNum* m1; 2576 BcNum* m2; 2577 BcNum* ptr; 2578 char c = 0; 2579 bool zero = true; 2580 BcBigDig v; 2581 size_t digs, len = strlen(val); 2582 // This is volatile to quiet a warning on GCC about longjmp() clobbering. 2583 volatile size_t i; 2584 #if BC_ENABLE_LIBRARY 2585 BcVm* vm = bcl_getspecific(); 2586 #endif // BC_ENABLE_LIBRARY 2587 2588 // If zero, just return because the number should be virgin (already 0). 2589 for (i = 0; zero && i < len; ++i) 2590 { 2591 zero = (val[i] == '.' || val[i] == '0'); 2592 } 2593 if (zero) return; 2594 2595 BC_SIG_LOCK; 2596 2597 bc_num_init(&temp, BC_NUM_BIGDIG_LOG10); 2598 bc_num_init(&mult1, BC_NUM_BIGDIG_LOG10); 2599 2600 BC_SETJMP_LOCKED(vm, int_err); 2601 2602 BC_SIG_UNLOCK; 2603 2604 // We split parsing into parsing the integer and parsing the fractional 2605 // part. 2606 2607 // Parse the integer part. This is the easy part because we just multiply 2608 // the number by the base, then add the digit. 2609 for (i = 0; i < len && (c = val[i]) && c != '.'; ++i) 2610 { 2611 // Convert the character to a digit. 2612 v = bc_num_parseChar(c, base); 2613 2614 // Multiply the number. 2615 bc_num_mulArray(n, base, &mult1); 2616 2617 // Convert the digit to a number and add. 2618 bc_num_bigdig2num(&temp, v); 2619 bc_num_add(&mult1, &temp, n, 0); 2620 } 2621 2622 // If this condition is true, then we are done. We still need to do cleanup 2623 // though. 2624 if (i == len && !val[i]) goto int_err; 2625 2626 // If we get here, we *must* be at the radix point. 2627 assert(val[i] == '.'); 2628 2629 BC_SIG_LOCK; 2630 2631 // Unset the jump to reset in for these new initializations. 2632 BC_UNSETJMP(vm); 2633 2634 bc_num_init(&mult2, BC_NUM_BIGDIG_LOG10); 2635 bc_num_init(&result1, BC_NUM_DEF_SIZE); 2636 bc_num_init(&result2, BC_NUM_DEF_SIZE); 2637 bc_num_one(&mult1); 2638 2639 BC_SETJMP_LOCKED(vm, err); 2640 2641 BC_SIG_UNLOCK; 2642 2643 // Pointers for easy switching. 2644 m1 = &mult1; 2645 m2 = &mult2; 2646 2647 // Parse the fractional part. This is the hard part. 2648 for (i += 1, digs = 0; i < len && (c = val[i]); ++i, ++digs) 2649 { 2650 size_t rdx; 2651 2652 // Convert the character to a digit. 2653 v = bc_num_parseChar(c, base); 2654 2655 // We keep growing result2 according to the base because the more digits 2656 // after the radix, the more significant the digits close to the radix 2657 // should be. 2658 bc_num_mulArray(&result1, base, &result2); 2659 2660 // Convert the digit to a number. 2661 bc_num_bigdig2num(&temp, v); 2662 2663 // Add the digit into the fraction part. 2664 bc_num_add(&result2, &temp, &result1, 0); 2665 2666 // Keep growing m1 and m2 for use after the loop. 2667 bc_num_mulArray(m1, base, m2); 2668 2669 rdx = BC_NUM_RDX_VAL(m2); 2670 2671 if (m2->len < rdx) m2->len = rdx; 2672 2673 // Switch. 2674 ptr = m1; 2675 m1 = m2; 2676 m2 = ptr; 2677 } 2678 2679 // This one cannot be a divide by 0 because mult starts out at 1, then is 2680 // multiplied by base, and base cannot be 0, so mult cannot be 0. And this 2681 // is the reason we keep growing m1 and m2; this division is what converts 2682 // the parsed fractional part from an integer to a fractional part. 2683 bc_num_div(&result1, m1, &result2, digs * 2); 2684 2685 // Pretruncate. 2686 bc_num_truncate(&result2, digs); 2687 2688 // The final add of the integer part to the fractional part. 2689 bc_num_add(n, &result2, n, digs); 2690 2691 // Basic cleanup. 2692 if (BC_NUM_NONZERO(n)) 2693 { 2694 if (n->scale < digs) bc_num_extend(n, digs - n->scale); 2695 } 2696 else bc_num_zero(n); 2697 2698 err: 2699 BC_SIG_MAYLOCK; 2700 bc_num_free(&result2); 2701 bc_num_free(&result1); 2702 bc_num_free(&mult2); 2703 int_err: 2704 BC_SIG_MAYLOCK; 2705 bc_num_free(&mult1); 2706 bc_num_free(&temp); 2707 BC_LONGJMP_CONT(vm); 2708 } 2709 2710 /** 2711 * Prints a backslash+newline combo if the number of characters needs it. This 2712 * is really a convenience function. 2713 */ 2714 static inline void 2715 bc_num_printNewline(void) 2716 { 2717 #if !BC_ENABLE_LIBRARY 2718 if (vm->nchars >= vm->line_len - 1 && vm->line_len) 2719 { 2720 bc_vm_putchar('\\', bc_flush_none); 2721 bc_vm_putchar('\n', bc_flush_err); 2722 } 2723 #endif // !BC_ENABLE_LIBRARY 2724 } 2725 2726 /** 2727 * Prints a character after a backslash+newline, if needed. 2728 * @param c The character to print. 2729 * @param bslash Whether to print a backslash+newline. 2730 */ 2731 static void 2732 bc_num_putchar(int c, bool bslash) 2733 { 2734 if (c != '\n' && bslash) bc_num_printNewline(); 2735 bc_vm_putchar(c, bc_flush_save); 2736 } 2737 2738 #if !BC_ENABLE_LIBRARY 2739 2740 /** 2741 * Prints a character for a number's digit. This is for printing for dc's P 2742 * command. This function does not need to worry about radix points. This is a 2743 * BcNumDigitOp. 2744 * @param n The "digit" to print. 2745 * @param len The "length" of the digit, or number of characters that will 2746 * need to be printed for the digit. 2747 * @param rdx True if a decimal (radix) point should be printed. 2748 * @param bslash True if a backslash+newline should be printed if the character 2749 * limit for the line is reached, false otherwise. 2750 */ 2751 static void 2752 bc_num_printChar(size_t n, size_t len, bool rdx, bool bslash) 2753 { 2754 BC_UNUSED(rdx); 2755 BC_UNUSED(len); 2756 BC_UNUSED(bslash); 2757 assert(len == 1); 2758 bc_vm_putchar((uchar) n, bc_flush_save); 2759 } 2760 2761 #endif // !BC_ENABLE_LIBRARY 2762 2763 /** 2764 * Prints a series of characters for large bases. This is for printing in bases 2765 * above hexadecimal. This is a BcNumDigitOp. 2766 * @param n The "digit" to print. 2767 * @param len The "length" of the digit, or number of characters that will 2768 * need to be printed for the digit. 2769 * @param rdx True if a decimal (radix) point should be printed. 2770 * @param bslash True if a backslash+newline should be printed if the character 2771 * limit for the line is reached, false otherwise. 2772 */ 2773 static void 2774 bc_num_printDigits(size_t n, size_t len, bool rdx, bool bslash) 2775 { 2776 size_t exp, pow; 2777 2778 // If needed, print the radix; otherwise, print a space to separate digits. 2779 bc_num_putchar(rdx ? '.' : ' ', true); 2780 2781 // Calculate the exponent and power. 2782 for (exp = 0, pow = 1; exp < len - 1; ++exp, pow *= BC_BASE) 2783 { 2784 continue; 2785 } 2786 2787 // Print each character individually. 2788 for (exp = 0; exp < len; pow /= BC_BASE, ++exp) 2789 { 2790 // The individual subdigit. 2791 size_t dig = n / pow; 2792 2793 // Take the subdigit away. 2794 n -= dig * pow; 2795 2796 // Print the subdigit. 2797 bc_num_putchar(((uchar) dig) + '0', bslash || exp != len - 1); 2798 } 2799 } 2800 2801 /** 2802 * Prints a character for a number's digit. This is for printing in bases for 2803 * hexadecimal and below because they always print only one character at a time. 2804 * This is a BcNumDigitOp. 2805 * @param n The "digit" to print. 2806 * @param len The "length" of the digit, or number of characters that will 2807 * need to be printed for the digit. 2808 * @param rdx True if a decimal (radix) point should be printed. 2809 * @param bslash True if a backslash+newline should be printed if the character 2810 * limit for the line is reached, false otherwise. 2811 */ 2812 static void 2813 bc_num_printHex(size_t n, size_t len, bool rdx, bool bslash) 2814 { 2815 BC_UNUSED(len); 2816 BC_UNUSED(bslash); 2817 2818 assert(len == 1); 2819 2820 if (rdx) bc_num_putchar('.', true); 2821 2822 bc_num_putchar(bc_num_hex_digits[n], bslash); 2823 } 2824 2825 /** 2826 * Prints a decimal number. This is specially written for optimization since 2827 * this will be used the most and because bc's numbers are already in decimal. 2828 * @param n The number to print. 2829 * @param newline Whether to print backslash+newlines on long enough lines. 2830 */ 2831 static void 2832 bc_num_printDecimal(const BcNum* restrict n, bool newline) 2833 { 2834 size_t i, j, rdx = BC_NUM_RDX_VAL(n); 2835 bool zero = true; 2836 size_t buffer[BC_BASE_DIGS]; 2837 2838 // Print loop. 2839 for (i = n->len - 1; i < n->len; --i) 2840 { 2841 BcDig n9 = n->num[i]; 2842 size_t temp; 2843 bool irdx = (i == rdx - 1); 2844 2845 // Calculate the number of digits in the limb. 2846 zero = (zero & !irdx); 2847 temp = n->scale % BC_BASE_DIGS; 2848 temp = i || !temp ? 0 : BC_BASE_DIGS - temp; 2849 2850 // NOLINTNEXTLINE 2851 memset(buffer, 0, BC_BASE_DIGS * sizeof(size_t)); 2852 2853 // Fill the buffer with individual digits. 2854 for (j = 0; n9 && j < BC_BASE_DIGS; ++j) 2855 { 2856 buffer[j] = ((size_t) n9) % BC_BASE; 2857 n9 /= BC_BASE; 2858 } 2859 2860 // Print the digits in the buffer. 2861 for (j = BC_BASE_DIGS - 1; j < BC_BASE_DIGS && j >= temp; --j) 2862 { 2863 // Figure out whether to print the decimal point. 2864 bool print_rdx = (irdx & (j == BC_BASE_DIGS - 1)); 2865 2866 // The zero variable helps us skip leading zero digits in the limb. 2867 zero = (zero && buffer[j] == 0); 2868 2869 if (!zero) 2870 { 2871 // While the first three arguments should be self-explanatory, 2872 // the last needs explaining. I don't want to print a newline 2873 // when the last digit to be printed could take the place of the 2874 // backslash rather than being pushed, as a single character, to 2875 // the next line. That's what that last argument does for bc. 2876 bc_num_printHex(buffer[j], 1, print_rdx, 2877 !newline || (j > temp || i != 0)); 2878 } 2879 } 2880 } 2881 } 2882 2883 #if BC_ENABLE_EXTRA_MATH 2884 2885 /** 2886 * Prints a number in scientific or engineering format. When doing this, we are 2887 * always printing in decimal. 2888 * @param n The number to print. 2889 * @param eng True if we are in engineering mode. 2890 * @param newline Whether to print backslash+newlines on long enough lines. 2891 */ 2892 static void 2893 bc_num_printExponent(const BcNum* restrict n, bool eng, bool newline) 2894 { 2895 size_t places, mod, nrdx = BC_NUM_RDX_VAL(n); 2896 bool neg = (n->len <= nrdx); 2897 BcNum temp, exp; 2898 BcDig digs[BC_NUM_BIGDIG_LOG10]; 2899 #if BC_ENABLE_LIBRARY 2900 BcVm* vm = bcl_getspecific(); 2901 #endif // BC_ENABLE_LIBRARY 2902 2903 BC_SIG_LOCK; 2904 2905 bc_num_createCopy(&temp, n); 2906 2907 BC_SETJMP_LOCKED(vm, exit); 2908 2909 BC_SIG_UNLOCK; 2910 2911 // We need to calculate the exponents, and they change based on whether the 2912 // number is all fractional or not, obviously. 2913 if (neg) 2914 { 2915 // Figure out the negative power of 10. 2916 places = bc_num_negPow10(n); 2917 2918 // Figure out how many digits mod 3 there are (important for 2919 // engineering mode). 2920 mod = places % 3; 2921 2922 // Calculate places if we are in engineering mode. 2923 if (eng && mod != 0) places += 3 - mod; 2924 2925 // Shift the temp to the right place. 2926 bc_num_shiftLeft(&temp, places); 2927 } 2928 else 2929 { 2930 // This is the number of digits that we are supposed to put behind the 2931 // decimal point. 2932 places = bc_num_intDigits(n) - 1; 2933 2934 // Calculate the true number based on whether engineering mode is 2935 // activated. 2936 mod = places % 3; 2937 if (eng && mod != 0) places -= 3 - (3 - mod); 2938 2939 // Shift the temp to the right place. 2940 bc_num_shiftRight(&temp, places); 2941 } 2942 2943 // Print the shifted number. 2944 bc_num_printDecimal(&temp, newline); 2945 2946 // Print the e. 2947 bc_num_putchar('e', !newline); 2948 2949 // Need to explicitly print a zero exponent. 2950 if (!places) 2951 { 2952 bc_num_printHex(0, 1, false, !newline); 2953 goto exit; 2954 } 2955 2956 // Need to print sign for the exponent. 2957 if (neg) bc_num_putchar('-', true); 2958 2959 // Create a temporary for the exponent... 2960 bc_num_setup(&exp, digs, BC_NUM_BIGDIG_LOG10); 2961 bc_num_bigdig2num(&exp, (BcBigDig) places); 2962 2963 /// ..and print it. 2964 bc_num_printDecimal(&exp, newline); 2965 2966 exit: 2967 BC_SIG_MAYLOCK; 2968 bc_num_free(&temp); 2969 BC_LONGJMP_CONT(vm); 2970 } 2971 #endif // BC_ENABLE_EXTRA_MATH 2972 2973 /** 2974 * Takes a number with limbs with base BC_BASE_POW and converts the limb at the 2975 * given index to base @a pow, where @a pow is obase^N. 2976 * @param n The number to convert. 2977 * @param rem BC_BASE_POW - @a pow. 2978 * @param pow The power of obase we will convert the number to. 2979 * @param idx The index of the number to start converting at. Doing the 2980 * conversion is O(n^2); we have to sweep through starting at the 2981 * least significant limb. 2982 */ 2983 static void 2984 bc_num_printFixup(BcNum* restrict n, BcBigDig rem, BcBigDig pow, size_t idx) 2985 { 2986 size_t i, len = n->len - idx; 2987 BcBigDig acc; 2988 BcDig* a = n->num + idx; 2989 2990 // Ignore if there's just one limb left. This is the part that requires the 2991 // extra loop after the one calling this function in bc_num_printPrepare(). 2992 if (len < 2) return; 2993 2994 // Loop through the remaining limbs and convert. We start at the second limb 2995 // because we pull the value from the previous one as well. 2996 for (i = len - 1; i > 0; --i) 2997 { 2998 // Get the limb and add it to the previous, along with multiplying by 2999 // the remainder because that's the proper overflow. "acc" means 3000 // "accumulator," by the way. 3001 acc = ((BcBigDig) a[i]) * rem + ((BcBigDig) a[i - 1]); 3002 3003 // Store a value in base pow in the previous limb. 3004 a[i - 1] = (BcDig) (acc % pow); 3005 3006 // Divide by the base and accumulate the remaining value in the limb. 3007 acc /= pow; 3008 acc += (BcBigDig) a[i]; 3009 3010 // If the accumulator is greater than the base... 3011 if (acc >= BC_BASE_POW) 3012 { 3013 // Do we need to grow? 3014 if (i == len - 1) 3015 { 3016 // Grow. 3017 len = bc_vm_growSize(len, 1); 3018 bc_num_expand(n, bc_vm_growSize(len, idx)); 3019 3020 // Update the pointer because it may have moved. 3021 a = n->num + idx; 3022 3023 // Zero out the last limb. 3024 a[len - 1] = 0; 3025 } 3026 3027 // Overflow into the next limb since we are over the base. 3028 a[i + 1] += acc / BC_BASE_POW; 3029 acc %= BC_BASE_POW; 3030 } 3031 3032 assert(acc < BC_BASE_POW); 3033 3034 // Set the limb. 3035 a[i] = (BcDig) acc; 3036 } 3037 3038 // We may have grown the number, so adjust the length. 3039 n->len = len + idx; 3040 } 3041 3042 /** 3043 * Prepares a number for printing in a base that does not have BC_BASE_POW as a 3044 * power. This basically converts the number from having limbs of base 3045 * BC_BASE_POW to limbs of pow, where pow is obase^N. 3046 * @param n The number to prepare for printing. 3047 * @param rem The remainder of BC_BASE_POW when divided by a power of the base. 3048 * @param pow The power of the base. 3049 */ 3050 static void 3051 bc_num_printPrepare(BcNum* restrict n, BcBigDig rem, BcBigDig pow) 3052 { 3053 size_t i; 3054 3055 // Loop from the least significant limb to the most significant limb and 3056 // convert limbs in each pass. 3057 for (i = 0; i < n->len; ++i) 3058 { 3059 bc_num_printFixup(n, rem, pow, i); 3060 } 3061 3062 // bc_num_printFixup() does not do everything it is supposed to, so we do 3063 // the last bit of cleanup here. That cleanup is to ensure that each limb 3064 // is less than pow and to expand the number to fit new limbs as necessary. 3065 for (i = 0; i < n->len; ++i) 3066 { 3067 assert(pow == ((BcBigDig) ((BcDig) pow))); 3068 3069 // If the limb needs fixing... 3070 if (n->num[i] >= (BcDig) pow) 3071 { 3072 // Do we need to grow? 3073 if (i + 1 == n->len) 3074 { 3075 // Grow the number. 3076 n->len = bc_vm_growSize(n->len, 1); 3077 bc_num_expand(n, n->len); 3078 3079 // Without this, we might use uninitialized data. 3080 n->num[i + 1] = 0; 3081 } 3082 3083 assert(pow < BC_BASE_POW); 3084 3085 // Overflow into the next limb. 3086 n->num[i + 1] += n->num[i] / ((BcDig) pow); 3087 n->num[i] %= (BcDig) pow; 3088 } 3089 } 3090 } 3091 3092 static void 3093 bc_num_printNum(BcNum* restrict n, BcBigDig base, size_t len, 3094 BcNumDigitOp print, bool newline) 3095 { 3096 BcVec stack; 3097 BcNum intp, fracp1, fracp2, digit, flen1, flen2; 3098 BcNum* n1; 3099 BcNum* n2; 3100 BcNum* temp; 3101 BcBigDig dig = 0, acc, exp; 3102 BcBigDig* ptr; 3103 size_t i, j, nrdx, idigits; 3104 bool radix; 3105 BcDig digit_digs[BC_NUM_BIGDIG_LOG10 + 1]; 3106 #if BC_ENABLE_LIBRARY 3107 BcVm* vm = bcl_getspecific(); 3108 #endif // BC_ENABLE_LIBRARY 3109 3110 assert(base > 1); 3111 3112 // Easy case. Even with scale, we just print this. 3113 if (BC_NUM_ZERO(n)) 3114 { 3115 print(0, len, false, !newline); 3116 return; 3117 } 3118 3119 // This function uses an algorithm that Stefan Esser <se@freebsd.org> came 3120 // up with to print the integer part of a number. What it does is convert 3121 // intp into a number of the specified base, but it does it directly, 3122 // instead of just doing a series of divisions and printing the remainders 3123 // in reverse order. 3124 // 3125 // Let me explain in a bit more detail: 3126 // 3127 // The algorithm takes the current least significant limb (after intp has 3128 // been converted to an integer) and the next to least significant limb, and 3129 // it converts the least significant limb into one of the specified base, 3130 // putting any overflow into the next to least significant limb. It iterates 3131 // through the whole number, from least significant to most significant, 3132 // doing this conversion. At the end of that iteration, the least 3133 // significant limb is converted, but the others are not, so it iterates 3134 // again, starting at the next to least significant limb. It keeps doing 3135 // that conversion, skipping one more limb than the last time, until all 3136 // limbs have been converted. Then it prints them in reverse order. 3137 // 3138 // That is the gist of the algorithm. It leaves out several things, such as 3139 // the fact that limbs are not always converted into the specified base, but 3140 // into something close, basically a power of the specified base. In 3141 // Stefan's words, "You could consider BcDigs to be of base 10^BC_BASE_DIGS 3142 // in the normal case and obase^N for the largest value of N that satisfies 3143 // obase^N <= 10^BC_BASE_DIGS. [This means that] the result is not in base 3144 // "obase", but in base "obase^N", which happens to be printable as a number 3145 // of base "obase" without consideration for neighbouring BcDigs." This fact 3146 // is what necessitates the existence of the loop later in this function. 3147 // 3148 // The conversion happens in bc_num_printPrepare() where the outer loop 3149 // happens and bc_num_printFixup() where the inner loop, or actual 3150 // conversion, happens. In other words, bc_num_printPrepare() is where the 3151 // loop that starts at the least significant limb and goes to the most 3152 // significant limb. Then, on every iteration of its loop, it calls 3153 // bc_num_printFixup(), which has the inner loop of actually converting 3154 // the limbs it passes into limbs of base obase^N rather than base 3155 // BC_BASE_POW. 3156 3157 nrdx = BC_NUM_RDX_VAL(n); 3158 3159 BC_SIG_LOCK; 3160 3161 // The stack is what allows us to reverse the digits for printing. 3162 bc_vec_init(&stack, sizeof(BcBigDig), BC_DTOR_NONE); 3163 bc_num_init(&fracp1, nrdx); 3164 3165 // intp will be the "integer part" of the number, so copy it. 3166 bc_num_createCopy(&intp, n); 3167 3168 BC_SETJMP_LOCKED(vm, err); 3169 3170 BC_SIG_UNLOCK; 3171 3172 // Make intp an integer. 3173 bc_num_truncate(&intp, intp.scale); 3174 3175 // Get the fractional part out. 3176 bc_num_sub(n, &intp, &fracp1, 0); 3177 3178 // If the base is not the same as the last base used for printing, we need 3179 // to update the cached exponent and power. Yes, we cache the values of the 3180 // exponent and power. That is to prevent us from calculating them every 3181 // time because printing will probably happen multiple times on the same 3182 // base. 3183 if (base != vm->last_base) 3184 { 3185 vm->last_pow = 1; 3186 vm->last_exp = 0; 3187 3188 // Calculate the exponent and power. 3189 while (vm->last_pow * base <= BC_BASE_POW) 3190 { 3191 vm->last_pow *= base; 3192 vm->last_exp += 1; 3193 } 3194 3195 // Also, the remainder and base itself. 3196 vm->last_rem = BC_BASE_POW - vm->last_pow; 3197 vm->last_base = base; 3198 } 3199 3200 exp = vm->last_exp; 3201 3202 // If vm->last_rem is 0, then the base we are printing in is a divisor of 3203 // BC_BASE_POW, which is the easy case because it means that BC_BASE_POW is 3204 // a power of obase, and no conversion is needed. If it *is* 0, then we have 3205 // the hard case, and we have to prepare the number for the base. 3206 if (vm->last_rem != 0) 3207 { 3208 bc_num_printPrepare(&intp, vm->last_rem, vm->last_pow); 3209 } 3210 3211 // After the conversion comes the surprisingly easy part. From here on out, 3212 // this is basically naive code that I wrote, adjusted for the larger bases. 3213 3214 // Fill the stack of digits for the integer part. 3215 for (i = 0; i < intp.len; ++i) 3216 { 3217 // Get the limb. 3218 acc = (BcBigDig) intp.num[i]; 3219 3220 // Turn the limb into digits of base obase. 3221 for (j = 0; j < exp && (i < intp.len - 1 || acc != 0); ++j) 3222 { 3223 // This condition is true if we are not at the last digit. 3224 if (j != exp - 1) 3225 { 3226 dig = acc % base; 3227 acc /= base; 3228 } 3229 else 3230 { 3231 dig = acc; 3232 acc = 0; 3233 } 3234 3235 assert(dig < base); 3236 3237 // Push the digit onto the stack. 3238 bc_vec_push(&stack, &dig); 3239 } 3240 3241 assert(acc == 0); 3242 } 3243 3244 // Go through the stack backwards and print each digit. 3245 for (i = 0; i < stack.len; ++i) 3246 { 3247 ptr = bc_vec_item_rev(&stack, i); 3248 3249 assert(ptr != NULL); 3250 3251 // While the first three arguments should be self-explanatory, the last 3252 // needs explaining. I don't want to print a backslash+newline when the 3253 // last digit to be printed could take the place of the backslash rather 3254 // than being pushed, as a single character, to the next line. That's 3255 // what that last argument does for bc. 3256 // 3257 // First, it needs to check if newlines are completely disabled. If they 3258 // are not disabled, it needs to check the next part. 3259 // 3260 // If the number has a scale, then because we are printing just the 3261 // integer part, there will be at least two more characters (a radix 3262 // point plus at least one digit). So if there is a scale, a backslash 3263 // is necessary. 3264 // 3265 // Finally, the last condition checks to see if we are at the end of the 3266 // stack. If we are *not* (i.e., the index is not one less than the 3267 // stack length), then a backslash is necessary because there is at 3268 // least one more character for at least one more digit). Otherwise, if 3269 // the index is equal to one less than the stack length, we want to 3270 // disable backslash printing. 3271 // 3272 // The function that prints bases 17 and above will take care of not 3273 // printing a backslash in the right case. 3274 print(*ptr, len, false, 3275 !newline || (n->scale != 0 || i < stack.len - 1)); 3276 } 3277 3278 // We are done if there is no fractional part. 3279 if (!n->scale) goto err; 3280 3281 BC_SIG_LOCK; 3282 3283 // Reset the jump because some locals are changing. 3284 BC_UNSETJMP(vm); 3285 3286 bc_num_init(&fracp2, nrdx); 3287 bc_num_setup(&digit, digit_digs, sizeof(digit_digs) / sizeof(BcDig)); 3288 bc_num_init(&flen1, BC_NUM_BIGDIG_LOG10); 3289 bc_num_init(&flen2, BC_NUM_BIGDIG_LOG10); 3290 3291 BC_SETJMP_LOCKED(vm, frac_err); 3292 3293 BC_SIG_UNLOCK; 3294 3295 bc_num_one(&flen1); 3296 3297 radix = true; 3298 3299 // Pointers for easy switching. 3300 n1 = &flen1; 3301 n2 = &flen2; 3302 3303 fracp2.scale = n->scale; 3304 BC_NUM_RDX_SET_NP(fracp2, BC_NUM_RDX(fracp2.scale)); 3305 3306 // As long as we have not reached the scale of the number, keep printing. 3307 while ((idigits = bc_num_intDigits(n1)) <= n->scale) 3308 { 3309 // These numbers will keep growing. 3310 bc_num_expand(&fracp2, fracp1.len + 1); 3311 bc_num_mulArray(&fracp1, base, &fracp2); 3312 3313 nrdx = BC_NUM_RDX_VAL_NP(fracp2); 3314 3315 // Ensure an invariant. 3316 if (fracp2.len < nrdx) fracp2.len = nrdx; 3317 3318 // fracp is guaranteed to be non-negative and small enough. 3319 dig = bc_num_bigdig2(&fracp2); 3320 3321 // Convert the digit to a number and subtract it from the number. 3322 bc_num_bigdig2num(&digit, dig); 3323 bc_num_sub(&fracp2, &digit, &fracp1, 0); 3324 3325 // While the first three arguments should be self-explanatory, the last 3326 // needs explaining. I don't want to print a newline when the last digit 3327 // to be printed could take the place of the backslash rather than being 3328 // pushed, as a single character, to the next line. That's what that 3329 // last argument does for bc. 3330 print(dig, len, radix, !newline || idigits != n->scale); 3331 3332 // Update the multipliers. 3333 bc_num_mulArray(n1, base, n2); 3334 3335 radix = false; 3336 3337 // Switch. 3338 temp = n1; 3339 n1 = n2; 3340 n2 = temp; 3341 } 3342 3343 frac_err: 3344 BC_SIG_MAYLOCK; 3345 bc_num_free(&flen2); 3346 bc_num_free(&flen1); 3347 bc_num_free(&fracp2); 3348 err: 3349 BC_SIG_MAYLOCK; 3350 bc_num_free(&fracp1); 3351 bc_num_free(&intp); 3352 bc_vec_free(&stack); 3353 BC_LONGJMP_CONT(vm); 3354 } 3355 3356 /** 3357 * Prints a number in the specified base, or rather, figures out which function 3358 * to call to print the number in the specified base and calls it. 3359 * @param n The number to print. 3360 * @param base The base to print in. 3361 * @param newline Whether to print backslash+newlines on long enough lines. 3362 */ 3363 static void 3364 bc_num_printBase(BcNum* restrict n, BcBigDig base, bool newline) 3365 { 3366 size_t width; 3367 BcNumDigitOp print; 3368 bool neg = BC_NUM_NEG(n); 3369 3370 // Clear the sign because it makes the actual printing easier when we have 3371 // to do math. 3372 BC_NUM_NEG_CLR(n); 3373 3374 // Bases at hexadecimal and below are printed as one character, larger bases 3375 // are printed as a series of digits separated by spaces. 3376 if (base <= BC_NUM_MAX_POSIX_IBASE) 3377 { 3378 width = 1; 3379 print = bc_num_printHex; 3380 } 3381 else 3382 { 3383 assert(base <= BC_BASE_POW); 3384 width = bc_num_log10(base - 1); 3385 print = bc_num_printDigits; 3386 } 3387 3388 // Print. 3389 bc_num_printNum(n, base, width, print, newline); 3390 3391 // Reset the sign. 3392 n->rdx = BC_NUM_NEG_VAL(n, neg); 3393 } 3394 3395 #if !BC_ENABLE_LIBRARY 3396 3397 void 3398 bc_num_stream(BcNum* restrict n) 3399 { 3400 bc_num_printNum(n, BC_NUM_STREAM_BASE, 1, bc_num_printChar, false); 3401 } 3402 3403 #endif // !BC_ENABLE_LIBRARY 3404 3405 void 3406 bc_num_setup(BcNum* restrict n, BcDig* restrict num, size_t cap) 3407 { 3408 assert(n != NULL); 3409 n->num = num; 3410 n->cap = cap; 3411 bc_num_zero(n); 3412 } 3413 3414 void 3415 bc_num_init(BcNum* restrict n, size_t req) 3416 { 3417 BcDig* num; 3418 3419 BC_SIG_ASSERT_LOCKED; 3420 3421 assert(n != NULL); 3422 3423 // BC_NUM_DEF_SIZE is set to be about the smallest allocation size that 3424 // malloc() returns in practice, so just use it. 3425 req = req >= BC_NUM_DEF_SIZE ? req : BC_NUM_DEF_SIZE; 3426 3427 // If we can't use a temp, allocate. 3428 if (req != BC_NUM_DEF_SIZE) num = bc_vm_malloc(BC_NUM_SIZE(req)); 3429 else 3430 { 3431 num = bc_vm_getTemp() == NULL ? bc_vm_malloc(BC_NUM_SIZE(req)) : 3432 bc_vm_takeTemp(); 3433 } 3434 3435 bc_num_setup(n, num, req); 3436 } 3437 3438 void 3439 bc_num_clear(BcNum* restrict n) 3440 { 3441 n->num = NULL; 3442 n->cap = 0; 3443 } 3444 3445 void 3446 bc_num_free(void* num) 3447 { 3448 BcNum* n = (BcNum*) num; 3449 3450 BC_SIG_ASSERT_LOCKED; 3451 3452 assert(n != NULL); 3453 3454 if (n->cap == BC_NUM_DEF_SIZE) bc_vm_addTemp(n->num); 3455 else free(n->num); 3456 } 3457 3458 void 3459 bc_num_copy(BcNum* d, const BcNum* s) 3460 { 3461 assert(d != NULL && s != NULL); 3462 3463 if (d == s) return; 3464 3465 bc_num_expand(d, s->len); 3466 d->len = s->len; 3467 3468 // I can just copy directly here because the sign *and* rdx will be 3469 // properly preserved. 3470 d->rdx = s->rdx; 3471 d->scale = s->scale; 3472 // NOLINTNEXTLINE 3473 memcpy(d->num, s->num, BC_NUM_SIZE(d->len)); 3474 } 3475 3476 void 3477 bc_num_createCopy(BcNum* d, const BcNum* s) 3478 { 3479 BC_SIG_ASSERT_LOCKED; 3480 bc_num_init(d, s->len); 3481 bc_num_copy(d, s); 3482 } 3483 3484 void 3485 bc_num_createFromBigdig(BcNum* restrict n, BcBigDig val) 3486 { 3487 BC_SIG_ASSERT_LOCKED; 3488 bc_num_init(n, BC_NUM_BIGDIG_LOG10); 3489 bc_num_bigdig2num(n, val); 3490 } 3491 3492 size_t 3493 bc_num_scale(const BcNum* restrict n) 3494 { 3495 return n->scale; 3496 } 3497 3498 size_t 3499 bc_num_len(const BcNum* restrict n) 3500 { 3501 size_t len = n->len; 3502 3503 // Always return at least 1. 3504 if (BC_NUM_ZERO(n)) return n->scale ? n->scale : 1; 3505 3506 // If this is true, there is no integer portion of the number. 3507 if (BC_NUM_RDX_VAL(n) == len) 3508 { 3509 // We have to take into account the fact that some of the digits right 3510 // after the decimal could be zero. If that is the case, we need to 3511 // ignore them until we hit the first non-zero digit. 3512 3513 size_t zero, scale; 3514 3515 // The number of limbs with non-zero digits. 3516 len = bc_num_nonZeroLen(n); 3517 3518 // Get the number of digits in the last limb. 3519 scale = n->scale % BC_BASE_DIGS; 3520 scale = scale ? scale : BC_BASE_DIGS; 3521 3522 // Get the number of zero digits. 3523 zero = bc_num_zeroDigits(n->num + len - 1); 3524 3525 // Calculate the true length. 3526 len = len * BC_BASE_DIGS - zero - (BC_BASE_DIGS - scale); 3527 } 3528 // Otherwise, count the number of int digits and return that plus the scale. 3529 else len = bc_num_intDigits(n) + n->scale; 3530 3531 return len; 3532 } 3533 3534 void 3535 bc_num_parse(BcNum* restrict n, const char* restrict val, BcBigDig base) 3536 { 3537 #if BC_DEBUG 3538 #if BC_ENABLE_LIBRARY 3539 BcVm* vm = bcl_getspecific(); 3540 #endif // BC_ENABLE_LIBRARY 3541 #endif // BC_DEBUG 3542 3543 assert(n != NULL && val != NULL && base); 3544 assert(base >= BC_NUM_MIN_BASE && base <= vm->maxes[BC_PROG_GLOBALS_IBASE]); 3545 assert(bc_num_strValid(val)); 3546 3547 // A one character number is *always* parsed as though the base was the 3548 // maximum allowed ibase, per the bc spec. 3549 if (!val[1]) 3550 { 3551 BcBigDig dig = bc_num_parseChar(val[0], BC_NUM_MAX_LBASE); 3552 bc_num_bigdig2num(n, dig); 3553 } 3554 else if (base == BC_BASE) bc_num_parseDecimal(n, val); 3555 else bc_num_parseBase(n, val, base); 3556 3557 assert(BC_NUM_RDX_VALID(n)); 3558 } 3559 3560 void 3561 bc_num_print(BcNum* restrict n, BcBigDig base, bool newline) 3562 { 3563 assert(n != NULL); 3564 assert(BC_ENABLE_EXTRA_MATH || base >= BC_NUM_MIN_BASE); 3565 3566 // We may need a newline, just to start. 3567 bc_num_printNewline(); 3568 3569 if (BC_NUM_NONZERO(n)) 3570 { 3571 #if BC_ENABLE_LIBRARY 3572 BcVm* vm = bcl_getspecific(); 3573 #endif // BC_ENABLE_LIBRARY 3574 3575 // Print the sign. 3576 if (BC_NUM_NEG(n)) bc_num_putchar('-', true); 3577 3578 // Print the leading zero if necessary. We don't print when using 3579 // scientific or engineering modes. 3580 if (BC_Z && BC_NUM_RDX_VAL(n) == n->len && base != 0 && base != 1) 3581 { 3582 bc_num_printHex(0, 1, false, !newline); 3583 } 3584 } 3585 3586 // Short-circuit 0. 3587 if (BC_NUM_ZERO(n)) bc_num_printHex(0, 1, false, !newline); 3588 else if (base == BC_BASE) bc_num_printDecimal(n, newline); 3589 #if BC_ENABLE_EXTRA_MATH 3590 else if (base == 0 || base == 1) 3591 { 3592 bc_num_printExponent(n, base != 0, newline); 3593 } 3594 #endif // BC_ENABLE_EXTRA_MATH 3595 else bc_num_printBase(n, base, newline); 3596 3597 if (newline) bc_num_putchar('\n', false); 3598 } 3599 3600 BcBigDig 3601 bc_num_bigdig2(const BcNum* restrict n) 3602 { 3603 #if BC_DEBUG 3604 #if BC_ENABLE_LIBRARY 3605 BcVm* vm = bcl_getspecific(); 3606 #endif // BC_ENABLE_LIBRARY 3607 #endif // BC_DEBUG 3608 3609 // This function returns no errors because it's guaranteed to succeed if 3610 // its preconditions are met. Those preconditions include both n needs to 3611 // be non-NULL, n being non-negative, and n being less than vm->max. If all 3612 // of that is true, then we can just convert without worrying about negative 3613 // errors or overflow. 3614 3615 BcBigDig r = 0; 3616 size_t nrdx = BC_NUM_RDX_VAL(n); 3617 3618 assert(n != NULL); 3619 assert(!BC_NUM_NEG(n)); 3620 assert(bc_num_cmp(n, &vm->max) < 0); 3621 assert(n->len - nrdx <= 3); 3622 3623 // There is a small speed win from unrolling the loop here, and since it 3624 // only adds 53 bytes, I decided that it was worth it. 3625 switch (n->len - nrdx) 3626 { 3627 case 3: 3628 { 3629 r = (BcBigDig) n->num[nrdx + 2]; 3630 3631 // Fallthrough. 3632 BC_FALLTHROUGH 3633 } 3634 3635 case 2: 3636 { 3637 r = r * BC_BASE_POW + (BcBigDig) n->num[nrdx + 1]; 3638 3639 // Fallthrough. 3640 BC_FALLTHROUGH 3641 } 3642 3643 case 1: 3644 { 3645 r = r * BC_BASE_POW + (BcBigDig) n->num[nrdx]; 3646 } 3647 } 3648 3649 return r; 3650 } 3651 3652 BcBigDig 3653 bc_num_bigdig(const BcNum* restrict n) 3654 { 3655 #if BC_ENABLE_LIBRARY 3656 BcVm* vm = bcl_getspecific(); 3657 #endif // BC_ENABLE_LIBRARY 3658 3659 assert(n != NULL); 3660 3661 // This error checking is extremely important, and if you do not have a 3662 // guarantee that converting a number will always succeed in a particular 3663 // case, you *must* call this function to get these error checks. This 3664 // includes all instances of numbers inputted by the user or calculated by 3665 // the user. Otherwise, you can call the faster bc_num_bigdig2(). 3666 if (BC_ERR(BC_NUM_NEG(n))) bc_err(BC_ERR_MATH_NEGATIVE); 3667 if (BC_ERR(bc_num_cmp(n, &vm->max) >= 0)) bc_err(BC_ERR_MATH_OVERFLOW); 3668 3669 return bc_num_bigdig2(n); 3670 } 3671 3672 void 3673 bc_num_bigdig2num(BcNum* restrict n, BcBigDig val) 3674 { 3675 BcDig* ptr; 3676 size_t i; 3677 3678 assert(n != NULL); 3679 3680 bc_num_zero(n); 3681 3682 // Already 0. 3683 if (!val) return; 3684 3685 // Expand first. This is the only way this function can fail, and it's a 3686 // fatal error. 3687 bc_num_expand(n, BC_NUM_BIGDIG_LOG10); 3688 3689 // The conversion is easy because numbers are laid out in little-endian 3690 // order. 3691 for (ptr = n->num, i = 0; val; ++i, val /= BC_BASE_POW) 3692 { 3693 ptr[i] = val % BC_BASE_POW; 3694 } 3695 3696 n->len = i; 3697 } 3698 3699 #if BC_ENABLE_EXTRA_MATH 3700 3701 void 3702 bc_num_rng(const BcNum* restrict n, BcRNG* rng) 3703 { 3704 BcNum temp, temp2, intn, frac; 3705 BcRand state1, state2, inc1, inc2; 3706 size_t nrdx = BC_NUM_RDX_VAL(n); 3707 #if BC_ENABLE_LIBRARY 3708 BcVm* vm = bcl_getspecific(); 3709 #endif // BC_ENABLE_LIBRARY 3710 3711 // This function holds the secret of how I interpret a seed number for the 3712 // PRNG. Well, it's actually in the development manual 3713 // (manuals/development.md#pseudo-random-number-generator), so look there 3714 // before you try to understand this. 3715 3716 BC_SIG_LOCK; 3717 3718 bc_num_init(&temp, n->len); 3719 bc_num_init(&temp2, n->len); 3720 bc_num_init(&frac, nrdx); 3721 bc_num_init(&intn, bc_num_int(n)); 3722 3723 BC_SETJMP_LOCKED(vm, err); 3724 3725 BC_SIG_UNLOCK; 3726 3727 assert(BC_NUM_RDX_VALID_NP(vm->max)); 3728 3729 // NOLINTNEXTLINE 3730 memcpy(frac.num, n->num, BC_NUM_SIZE(nrdx)); 3731 frac.len = nrdx; 3732 BC_NUM_RDX_SET_NP(frac, nrdx); 3733 frac.scale = n->scale; 3734 3735 assert(BC_NUM_RDX_VALID_NP(frac)); 3736 assert(BC_NUM_RDX_VALID_NP(vm->max2)); 3737 3738 // Multiply the fraction and truncate so that it's an integer. The 3739 // truncation is what clamps it, by the way. 3740 bc_num_mul(&frac, &vm->max2, &temp, 0); 3741 bc_num_truncate(&temp, temp.scale); 3742 bc_num_copy(&frac, &temp); 3743 3744 // Get the integer. 3745 // NOLINTNEXTLINE 3746 memcpy(intn.num, n->num + nrdx, BC_NUM_SIZE(bc_num_int(n))); 3747 intn.len = bc_num_int(n); 3748 3749 // This assert is here because it has to be true. It is also here to justify 3750 // some optimizations. 3751 assert(BC_NUM_NONZERO(&vm->max)); 3752 3753 // If there *was* a fractional part... 3754 if (BC_NUM_NONZERO(&frac)) 3755 { 3756 // This divmod splits frac into the two state parts. 3757 bc_num_divmod(&frac, &vm->max, &temp, &temp2, 0); 3758 3759 // frac is guaranteed to be smaller than vm->max * vm->max (pow). 3760 // This means that when dividing frac by vm->max, as above, the 3761 // quotient and remainder are both guaranteed to be less than vm->max, 3762 // which means we can use bc_num_bigdig2() here and not worry about 3763 // overflow. 3764 state1 = (BcRand) bc_num_bigdig2(&temp2); 3765 state2 = (BcRand) bc_num_bigdig2(&temp); 3766 } 3767 else state1 = state2 = 0; 3768 3769 // If there *was* an integer part... 3770 if (BC_NUM_NONZERO(&intn)) 3771 { 3772 // This divmod splits intn into the two inc parts. 3773 bc_num_divmod(&intn, &vm->max, &temp, &temp2, 0); 3774 3775 // Because temp2 is the mod of vm->max, from above, it is guaranteed 3776 // to be small enough to use bc_num_bigdig2(). 3777 inc1 = (BcRand) bc_num_bigdig2(&temp2); 3778 3779 // Clamp the second inc part. 3780 if (bc_num_cmp(&temp, &vm->max) >= 0) 3781 { 3782 bc_num_copy(&temp2, &temp); 3783 bc_num_mod(&temp2, &vm->max, &temp, 0); 3784 } 3785 3786 // The if statement above ensures that temp is less than vm->max, which 3787 // means that we can use bc_num_bigdig2() here. 3788 inc2 = (BcRand) bc_num_bigdig2(&temp); 3789 } 3790 else inc1 = inc2 = 0; 3791 3792 bc_rand_seed(rng, state1, state2, inc1, inc2); 3793 3794 err: 3795 BC_SIG_MAYLOCK; 3796 bc_num_free(&intn); 3797 bc_num_free(&frac); 3798 bc_num_free(&temp2); 3799 bc_num_free(&temp); 3800 BC_LONGJMP_CONT(vm); 3801 } 3802 3803 void 3804 bc_num_createFromRNG(BcNum* restrict n, BcRNG* rng) 3805 { 3806 BcRand s1, s2, i1, i2; 3807 BcNum conv, temp1, temp2, temp3; 3808 BcDig temp1_num[BC_RAND_NUM_SIZE], temp2_num[BC_RAND_NUM_SIZE]; 3809 BcDig conv_num[BC_NUM_BIGDIG_LOG10]; 3810 #if BC_ENABLE_LIBRARY 3811 BcVm* vm = bcl_getspecific(); 3812 #endif // BC_ENABLE_LIBRARY 3813 3814 BC_SIG_LOCK; 3815 3816 bc_num_init(&temp3, 2 * BC_RAND_NUM_SIZE); 3817 3818 BC_SETJMP_LOCKED(vm, err); 3819 3820 BC_SIG_UNLOCK; 3821 3822 bc_num_setup(&temp1, temp1_num, sizeof(temp1_num) / sizeof(BcDig)); 3823 bc_num_setup(&temp2, temp2_num, sizeof(temp2_num) / sizeof(BcDig)); 3824 bc_num_setup(&conv, conv_num, sizeof(conv_num) / sizeof(BcDig)); 3825 3826 // This assert is here because it has to be true. It is also here to justify 3827 // the assumption that vm->max is not zero. 3828 assert(BC_NUM_NONZERO(&vm->max)); 3829 3830 // Because this is true, we can just ignore math errors that would happen 3831 // otherwise. 3832 assert(BC_NUM_NONZERO(&vm->max2)); 3833 3834 bc_rand_getRands(rng, &s1, &s2, &i1, &i2); 3835 3836 // Put the second piece of state into a number. 3837 bc_num_bigdig2num(&conv, (BcBigDig) s2); 3838 3839 assert(BC_NUM_RDX_VALID_NP(conv)); 3840 3841 // Multiply by max to make room for the first piece of state. 3842 bc_num_mul(&conv, &vm->max, &temp1, 0); 3843 3844 // Add in the first piece of state. 3845 bc_num_bigdig2num(&conv, (BcBigDig) s1); 3846 bc_num_add(&conv, &temp1, &temp2, 0); 3847 3848 // Divide to make it an entirely fractional part. 3849 bc_num_div(&temp2, &vm->max2, &temp3, BC_RAND_STATE_BITS); 3850 3851 // Now start on the increment parts. It's the same process without the 3852 // divide, so put the second piece of increment into a number. 3853 bc_num_bigdig2num(&conv, (BcBigDig) i2); 3854 3855 assert(BC_NUM_RDX_VALID_NP(conv)); 3856 3857 // Multiply by max to make room for the first piece of increment. 3858 bc_num_mul(&conv, &vm->max, &temp1, 0); 3859 3860 // Add in the first piece of increment. 3861 bc_num_bigdig2num(&conv, (BcBigDig) i1); 3862 bc_num_add(&conv, &temp1, &temp2, 0); 3863 3864 // Now add the two together. 3865 bc_num_add(&temp2, &temp3, n, 0); 3866 3867 assert(BC_NUM_RDX_VALID(n)); 3868 3869 err: 3870 BC_SIG_MAYLOCK; 3871 bc_num_free(&temp3); 3872 BC_LONGJMP_CONT(vm); 3873 } 3874 3875 void 3876 bc_num_irand(BcNum* restrict a, BcNum* restrict b, BcRNG* restrict rng) 3877 { 3878 BcNum atemp; 3879 size_t i; 3880 3881 assert(a != b); 3882 3883 if (BC_ERR(BC_NUM_NEG(a))) bc_err(BC_ERR_MATH_NEGATIVE); 3884 3885 // If either of these are true, then the numbers are integers. 3886 if (BC_NUM_ZERO(a) || BC_NUM_ONE(a)) return; 3887 3888 #if BC_GCC 3889 // This is here in GCC to quiet the "maybe-uninitialized" warning. 3890 atemp.num = NULL; 3891 atemp.len = 0; 3892 #endif // BC_GCC 3893 3894 if (BC_ERR(bc_num_nonInt(a, &atemp))) bc_err(BC_ERR_MATH_NON_INTEGER); 3895 3896 assert(atemp.num != NULL); 3897 assert(atemp.len); 3898 3899 if (atemp.len > 2) 3900 { 3901 size_t len; 3902 3903 len = atemp.len - 2; 3904 3905 // Just generate a random number for each limb. 3906 for (i = 0; i < len; i += 2) 3907 { 3908 BcRand dig; 3909 3910 dig = bc_rand_bounded(rng, BC_BASE_RAND_POW); 3911 3912 b->num[i] = (BcDig) (dig % BC_BASE_POW); 3913 b->num[i + 1] = (BcDig) (dig / BC_BASE_POW); 3914 } 3915 } 3916 else 3917 { 3918 // We need this set. 3919 i = 0; 3920 } 3921 3922 // This will be true if there's one full limb after the two limb groups. 3923 if (i == atemp.len - 2) 3924 { 3925 // Increment this for easy use. 3926 i += 1; 3927 3928 // If the last digit is not one, we need to set a bound for it 3929 // explicitly. Since there's still an empty limb, we need to fill that. 3930 if (atemp.num[i] != 1) 3931 { 3932 BcRand dig; 3933 BcRand bound; 3934 3935 // Set the bound to the bound of the last limb times the amount 3936 // needed to fill the second-to-last limb as well. 3937 bound = ((BcRand) atemp.num[i]) * BC_BASE_POW; 3938 3939 dig = bc_rand_bounded(rng, bound); 3940 3941 // Fill the last two. 3942 b->num[i - 1] = (BcDig) (dig % BC_BASE_POW); 3943 b->num[i] = (BcDig) (dig / BC_BASE_POW); 3944 3945 // Ensure that the length will be correct. If the last limb is zero, 3946 // then the length needs to be one less than the bound. 3947 b->len = atemp.len - (b->num[i] == 0); 3948 } 3949 // Here the last limb *is* one, which means the last limb does *not* 3950 // need to be filled. Also, the length needs to be one less because the 3951 // last limb is 0. 3952 else 3953 { 3954 b->num[i - 1] = (BcDig) bc_rand_bounded(rng, BC_BASE_POW); 3955 b->len = atemp.len - 1; 3956 } 3957 } 3958 // Here, there is only one limb to fill. 3959 else 3960 { 3961 // See above for how this works. 3962 if (atemp.num[i] != 1) 3963 { 3964 b->num[i] = (BcDig) bc_rand_bounded(rng, (BcRand) atemp.num[i]); 3965 b->len = atemp.len - (b->num[i] == 0); 3966 } 3967 else b->len = atemp.len - 1; 3968 } 3969 3970 bc_num_clean(b); 3971 3972 assert(BC_NUM_RDX_VALID(b)); 3973 } 3974 #endif // BC_ENABLE_EXTRA_MATH 3975 3976 size_t 3977 bc_num_addReq(const BcNum* a, const BcNum* b, size_t scale) 3978 { 3979 size_t aint, bint, ardx, brdx; 3980 3981 // Addition and subtraction require the max of the length of the two numbers 3982 // plus 1. 3983 3984 BC_UNUSED(scale); 3985 3986 ardx = BC_NUM_RDX_VAL(a); 3987 aint = bc_num_int(a); 3988 assert(aint <= a->len && ardx <= a->len); 3989 3990 brdx = BC_NUM_RDX_VAL(b); 3991 bint = bc_num_int(b); 3992 assert(bint <= b->len && brdx <= b->len); 3993 3994 ardx = BC_MAX(ardx, brdx); 3995 aint = BC_MAX(aint, bint); 3996 3997 return bc_vm_growSize(bc_vm_growSize(ardx, aint), 1); 3998 } 3999 4000 size_t 4001 bc_num_mulReq(const BcNum* a, const BcNum* b, size_t scale) 4002 { 4003 size_t max, rdx; 4004 4005 // Multiplication requires the sum of the lengths of the numbers. 4006 4007 rdx = bc_vm_growSize(BC_NUM_RDX_VAL(a), BC_NUM_RDX_VAL(b)); 4008 4009 max = BC_NUM_RDX(scale); 4010 4011 max = bc_vm_growSize(BC_MAX(max, rdx), 1); 4012 rdx = bc_vm_growSize(bc_vm_growSize(bc_num_int(a), bc_num_int(b)), max); 4013 4014 return rdx; 4015 } 4016 4017 size_t 4018 bc_num_divReq(const BcNum* a, const BcNum* b, size_t scale) 4019 { 4020 size_t max, rdx; 4021 4022 // Division requires the length of the dividend plus the scale. 4023 4024 rdx = bc_vm_growSize(BC_NUM_RDX_VAL(a), BC_NUM_RDX_VAL(b)); 4025 4026 max = BC_NUM_RDX(scale); 4027 4028 max = bc_vm_growSize(BC_MAX(max, rdx), 1); 4029 rdx = bc_vm_growSize(bc_num_int(a), max); 4030 4031 return rdx; 4032 } 4033 4034 size_t 4035 bc_num_powReq(const BcNum* a, const BcNum* b, size_t scale) 4036 { 4037 BC_UNUSED(scale); 4038 return bc_vm_growSize(bc_vm_growSize(a->len, b->len), 1); 4039 } 4040 4041 #if BC_ENABLE_EXTRA_MATH 4042 size_t 4043 bc_num_placesReq(const BcNum* a, const BcNum* b, size_t scale) 4044 { 4045 BC_UNUSED(scale); 4046 return a->len + b->len - BC_NUM_RDX_VAL(a) - BC_NUM_RDX_VAL(b); 4047 } 4048 #endif // BC_ENABLE_EXTRA_MATH 4049 4050 void 4051 bc_num_add(BcNum* a, BcNum* b, BcNum* c, size_t scale) 4052 { 4053 assert(BC_NUM_RDX_VALID(a)); 4054 assert(BC_NUM_RDX_VALID(b)); 4055 bc_num_binary(a, b, c, false, bc_num_as, bc_num_addReq(a, b, scale)); 4056 } 4057 4058 void 4059 bc_num_sub(BcNum* a, BcNum* b, BcNum* c, size_t scale) 4060 { 4061 assert(BC_NUM_RDX_VALID(a)); 4062 assert(BC_NUM_RDX_VALID(b)); 4063 bc_num_binary(a, b, c, true, bc_num_as, bc_num_addReq(a, b, scale)); 4064 } 4065 4066 void 4067 bc_num_mul(BcNum* a, BcNum* b, BcNum* c, size_t scale) 4068 { 4069 assert(BC_NUM_RDX_VALID(a)); 4070 assert(BC_NUM_RDX_VALID(b)); 4071 bc_num_binary(a, b, c, scale, bc_num_m, bc_num_mulReq(a, b, scale)); 4072 } 4073 4074 void 4075 bc_num_div(BcNum* a, BcNum* b, BcNum* c, size_t scale) 4076 { 4077 assert(BC_NUM_RDX_VALID(a)); 4078 assert(BC_NUM_RDX_VALID(b)); 4079 bc_num_binary(a, b, c, scale, bc_num_d, bc_num_divReq(a, b, scale)); 4080 } 4081 4082 void 4083 bc_num_mod(BcNum* a, BcNum* b, BcNum* c, size_t scale) 4084 { 4085 assert(BC_NUM_RDX_VALID(a)); 4086 assert(BC_NUM_RDX_VALID(b)); 4087 bc_num_binary(a, b, c, scale, bc_num_rem, bc_num_divReq(a, b, scale)); 4088 } 4089 4090 void 4091 bc_num_pow(BcNum* a, BcNum* b, BcNum* c, size_t scale) 4092 { 4093 assert(BC_NUM_RDX_VALID(a)); 4094 assert(BC_NUM_RDX_VALID(b)); 4095 bc_num_binary(a, b, c, scale, bc_num_p, bc_num_powReq(a, b, scale)); 4096 } 4097 4098 #if BC_ENABLE_EXTRA_MATH 4099 void 4100 bc_num_places(BcNum* a, BcNum* b, BcNum* c, size_t scale) 4101 { 4102 assert(BC_NUM_RDX_VALID(a)); 4103 assert(BC_NUM_RDX_VALID(b)); 4104 bc_num_binary(a, b, c, scale, bc_num_place, bc_num_placesReq(a, b, scale)); 4105 } 4106 4107 void 4108 bc_num_lshift(BcNum* a, BcNum* b, BcNum* c, size_t scale) 4109 { 4110 assert(BC_NUM_RDX_VALID(a)); 4111 assert(BC_NUM_RDX_VALID(b)); 4112 bc_num_binary(a, b, c, scale, bc_num_left, bc_num_placesReq(a, b, scale)); 4113 } 4114 4115 void 4116 bc_num_rshift(BcNum* a, BcNum* b, BcNum* c, size_t scale) 4117 { 4118 assert(BC_NUM_RDX_VALID(a)); 4119 assert(BC_NUM_RDX_VALID(b)); 4120 bc_num_binary(a, b, c, scale, bc_num_right, bc_num_placesReq(a, b, scale)); 4121 } 4122 #endif // BC_ENABLE_EXTRA_MATH 4123 4124 void 4125 bc_num_sqrt(BcNum* restrict a, BcNum* restrict b, size_t scale) 4126 { 4127 BcNum num1, num2, half, f, fprime; 4128 BcNum* x0; 4129 BcNum* x1; 4130 BcNum* temp; 4131 // realscale is meant to quiet a warning on GCC about longjmp() clobbering. 4132 // This one is real. 4133 size_t pow, len, rdx, req, resscale, realscale; 4134 BcDig half_digs[1]; 4135 #if BC_ENABLE_LIBRARY 4136 BcVm* vm = bcl_getspecific(); 4137 #endif // BC_ENABLE_LIBRARY 4138 4139 assert(a != NULL && b != NULL && a != b); 4140 4141 if (BC_ERR(BC_NUM_NEG(a))) bc_err(BC_ERR_MATH_NEGATIVE); 4142 4143 // We want to calculate to a's scale if it is bigger so that the result will 4144 // truncate properly. 4145 if (a->scale > scale) realscale = a->scale; 4146 else realscale = scale; 4147 4148 // Set parameters for the result. 4149 len = bc_vm_growSize(bc_num_intDigits(a), 1); 4150 rdx = BC_NUM_RDX(realscale); 4151 4152 // Square root needs half of the length of the parameter. 4153 req = bc_vm_growSize(BC_MAX(rdx, BC_NUM_RDX_VAL(a)), len >> 1); 4154 req = bc_vm_growSize(req, 1); 4155 4156 BC_SIG_LOCK; 4157 4158 // Unlike the binary operators, this function is the only single parameter 4159 // function and is expected to initialize the result. This means that it 4160 // expects that b is *NOT* preallocated. We allocate it here. 4161 bc_num_init(b, req); 4162 4163 BC_SIG_UNLOCK; 4164 4165 assert(a != NULL && b != NULL && a != b); 4166 assert(a->num != NULL && b->num != NULL); 4167 4168 // Easy case. 4169 if (BC_NUM_ZERO(a)) 4170 { 4171 bc_num_setToZero(b, realscale); 4172 return; 4173 } 4174 4175 // Another easy case. 4176 if (BC_NUM_ONE(a)) 4177 { 4178 bc_num_one(b); 4179 bc_num_extend(b, realscale); 4180 return; 4181 } 4182 4183 // Set the parameters again. 4184 rdx = BC_NUM_RDX(realscale); 4185 rdx = BC_MAX(rdx, BC_NUM_RDX_VAL(a)); 4186 len = bc_vm_growSize(a->len, rdx); 4187 4188 BC_SIG_LOCK; 4189 4190 bc_num_init(&num1, len); 4191 bc_num_init(&num2, len); 4192 bc_num_setup(&half, half_digs, sizeof(half_digs) / sizeof(BcDig)); 4193 4194 // There is a division by two in the formula. We set up a number that's 1/2 4195 // so that we can use multiplication instead of heavy division. 4196 bc_num_setToZero(&half, 1); 4197 half.num[0] = BC_BASE_POW / 2; 4198 half.len = 1; 4199 BC_NUM_RDX_SET_NP(half, 1); 4200 4201 bc_num_init(&f, len); 4202 bc_num_init(&fprime, len); 4203 4204 BC_SETJMP_LOCKED(vm, err); 4205 4206 BC_SIG_UNLOCK; 4207 4208 // Pointers for easy switching. 4209 x0 = &num1; 4210 x1 = &num2; 4211 4212 // Start with 1. 4213 bc_num_one(x0); 4214 4215 // The power of the operand is needed for the estimate. 4216 pow = bc_num_intDigits(a); 4217 4218 // The code in this if statement calculates the initial estimate. First, if 4219 // a is less than 1, then 0 is a good estimate. Otherwise, we want something 4220 // in the same ballpark. That ballpark is half of pow because the result 4221 // will have half the digits. 4222 if (pow) 4223 { 4224 // An odd number is served by starting with 2^((pow-1)/2), and an even 4225 // number is served by starting with 6^((pow-2)/2). Why? Because math. 4226 if (pow & 1) x0->num[0] = 2; 4227 else x0->num[0] = 6; 4228 4229 pow -= 2 - (pow & 1); 4230 bc_num_shiftLeft(x0, pow / 2); 4231 } 4232 4233 // I can set the rdx here directly because neg should be false. 4234 x0->scale = x0->rdx = 0; 4235 resscale = (realscale + BC_BASE_DIGS) + 2; 4236 4237 // This is the calculation loop. This compare goes to 0 eventually as the 4238 // difference between the two numbers gets smaller than resscale. 4239 while (bc_num_cmp(x1, x0)) 4240 { 4241 assert(BC_NUM_NONZERO(x0)); 4242 4243 // This loop directly corresponds to the iteration in Newton's method. 4244 // If you know the formula, this loop makes sense. Go study the formula. 4245 4246 bc_num_div(a, x0, &f, resscale); 4247 bc_num_add(x0, &f, &fprime, resscale); 4248 4249 assert(BC_NUM_RDX_VALID_NP(fprime)); 4250 assert(BC_NUM_RDX_VALID_NP(half)); 4251 4252 bc_num_mul(&fprime, &half, x1, resscale); 4253 4254 // Switch. 4255 temp = x0; 4256 x0 = x1; 4257 x1 = temp; 4258 } 4259 4260 // Copy to the result and truncate. 4261 bc_num_copy(b, x0); 4262 if (b->scale > realscale) bc_num_truncate(b, b->scale - realscale); 4263 4264 assert(!BC_NUM_NEG(b) || BC_NUM_NONZERO(b)); 4265 assert(BC_NUM_RDX_VALID(b)); 4266 assert(BC_NUM_RDX_VAL(b) <= b->len || !b->len); 4267 assert(!b->len || b->num[b->len - 1] || BC_NUM_RDX_VAL(b) == b->len); 4268 4269 err: 4270 BC_SIG_MAYLOCK; 4271 bc_num_free(&fprime); 4272 bc_num_free(&f); 4273 bc_num_free(&num2); 4274 bc_num_free(&num1); 4275 BC_LONGJMP_CONT(vm); 4276 } 4277 4278 void 4279 bc_num_divmod(BcNum* a, BcNum* b, BcNum* c, BcNum* d, size_t scale) 4280 { 4281 size_t ts, len; 4282 BcNum *ptr_a, num2; 4283 // This is volatile to quiet a warning on GCC about clobbering with 4284 // longjmp(). 4285 volatile bool init = false; 4286 #if BC_ENABLE_LIBRARY 4287 BcVm* vm = bcl_getspecific(); 4288 #endif // BC_ENABLE_LIBRARY 4289 4290 // The bulk of this function is just doing what bc_num_binary() does for the 4291 // binary operators. However, it assumes that only c and a can be equal. 4292 4293 // Set up the parameters. 4294 ts = BC_MAX(scale + b->scale, a->scale); 4295 len = bc_num_mulReq(a, b, ts); 4296 4297 assert(a != NULL && b != NULL && c != NULL && d != NULL); 4298 assert(c != d && a != d && b != d && b != c); 4299 4300 // Initialize or expand as necessary. 4301 if (c == a) 4302 { 4303 // NOLINTNEXTLINE 4304 memcpy(&num2, c, sizeof(BcNum)); 4305 ptr_a = &num2; 4306 4307 BC_SIG_LOCK; 4308 4309 bc_num_init(c, len); 4310 4311 init = true; 4312 4313 BC_SETJMP_LOCKED(vm, err); 4314 4315 BC_SIG_UNLOCK; 4316 } 4317 else 4318 { 4319 ptr_a = a; 4320 bc_num_expand(c, len); 4321 } 4322 4323 // Do the quick version if possible. 4324 if (BC_NUM_NONZERO(a) && !BC_NUM_RDX_VAL(a) && !BC_NUM_RDX_VAL(b) && 4325 b->len == 1 && !scale) 4326 { 4327 BcBigDig rem; 4328 4329 bc_num_divArray(ptr_a, (BcBigDig) b->num[0], c, &rem); 4330 4331 assert(rem < BC_BASE_POW); 4332 4333 d->num[0] = (BcDig) rem; 4334 d->len = (rem != 0); 4335 } 4336 // Do the slow method. 4337 else bc_num_r(ptr_a, b, c, d, scale, ts); 4338 4339 assert(!BC_NUM_NEG(c) || BC_NUM_NONZERO(c)); 4340 assert(BC_NUM_RDX_VALID(c)); 4341 assert(BC_NUM_RDX_VAL(c) <= c->len || !c->len); 4342 assert(!c->len || c->num[c->len - 1] || BC_NUM_RDX_VAL(c) == c->len); 4343 assert(!BC_NUM_NEG(d) || BC_NUM_NONZERO(d)); 4344 assert(BC_NUM_RDX_VALID(d)); 4345 assert(BC_NUM_RDX_VAL(d) <= d->len || !d->len); 4346 assert(!d->len || d->num[d->len - 1] || BC_NUM_RDX_VAL(d) == d->len); 4347 4348 err: 4349 // Only cleanup if we initialized. 4350 if (init) 4351 { 4352 BC_SIG_MAYLOCK; 4353 bc_num_free(&num2); 4354 BC_LONGJMP_CONT(vm); 4355 } 4356 } 4357 4358 void 4359 bc_num_modexp(BcNum* a, BcNum* b, BcNum* c, BcNum* restrict d) 4360 { 4361 BcNum base, exp, two, temp, atemp, btemp, ctemp; 4362 BcDig two_digs[2]; 4363 #if BC_ENABLE_LIBRARY 4364 BcVm* vm = bcl_getspecific(); 4365 #endif // BC_ENABLE_LIBRARY 4366 4367 assert(a != NULL && b != NULL && c != NULL && d != NULL); 4368 assert(a != d && b != d && c != d); 4369 4370 if (BC_ERR(BC_NUM_ZERO(c))) bc_err(BC_ERR_MATH_DIVIDE_BY_ZERO); 4371 if (BC_ERR(BC_NUM_NEG(b))) bc_err(BC_ERR_MATH_NEGATIVE); 4372 4373 #if BC_DEBUG || BC_GCC 4374 // This is entirely for quieting a useless scan-build error. 4375 btemp.len = 0; 4376 ctemp.len = 0; 4377 #endif // BC_DEBUG || BC_GCC 4378 4379 // Eliminate fractional parts that are zero or error if they are not zero. 4380 if (BC_ERR(bc_num_nonInt(a, &atemp) || bc_num_nonInt(b, &btemp) || 4381 bc_num_nonInt(c, &ctemp))) 4382 { 4383 bc_err(BC_ERR_MATH_NON_INTEGER); 4384 } 4385 4386 bc_num_expand(d, ctemp.len); 4387 4388 BC_SIG_LOCK; 4389 4390 bc_num_init(&base, ctemp.len); 4391 bc_num_setup(&two, two_digs, sizeof(two_digs) / sizeof(BcDig)); 4392 bc_num_init(&temp, btemp.len + 1); 4393 bc_num_createCopy(&exp, &btemp); 4394 4395 BC_SETJMP_LOCKED(vm, err); 4396 4397 BC_SIG_UNLOCK; 4398 4399 bc_num_one(&two); 4400 two.num[0] = 2; 4401 bc_num_one(d); 4402 4403 // We already checked for 0. 4404 bc_num_rem(&atemp, &ctemp, &base, 0); 4405 4406 // If you know the algorithm I used, the memory-efficient method, then this 4407 // loop should be self-explanatory because it is the calculation loop. 4408 while (BC_NUM_NONZERO(&exp)) 4409 { 4410 // Num two cannot be 0, so no errors. 4411 bc_num_divmod(&exp, &two, &exp, &temp, 0); 4412 4413 if (BC_NUM_ONE(&temp) && !BC_NUM_NEG_NP(temp)) 4414 { 4415 assert(BC_NUM_RDX_VALID(d)); 4416 assert(BC_NUM_RDX_VALID_NP(base)); 4417 4418 bc_num_mul(d, &base, &temp, 0); 4419 4420 // We already checked for 0. 4421 bc_num_rem(&temp, &ctemp, d, 0); 4422 } 4423 4424 assert(BC_NUM_RDX_VALID_NP(base)); 4425 4426 bc_num_mul(&base, &base, &temp, 0); 4427 4428 // We already checked for 0. 4429 bc_num_rem(&temp, &ctemp, &base, 0); 4430 } 4431 4432 err: 4433 BC_SIG_MAYLOCK; 4434 bc_num_free(&exp); 4435 bc_num_free(&temp); 4436 bc_num_free(&base); 4437 BC_LONGJMP_CONT(vm); 4438 assert(!BC_NUM_NEG(d) || d->len); 4439 assert(BC_NUM_RDX_VALID(d)); 4440 assert(!d->len || d->num[d->len - 1] || BC_NUM_RDX_VAL(d) == d->len); 4441 } 4442 4443 #if BC_DEBUG_CODE 4444 void 4445 bc_num_printDebug(const BcNum* n, const char* name, bool emptyline) 4446 { 4447 bc_file_puts(&vm->fout, bc_flush_none, name); 4448 bc_file_puts(&vm->fout, bc_flush_none, ": "); 4449 bc_num_printDecimal(n, true); 4450 bc_file_putchar(&vm->fout, bc_flush_err, '\n'); 4451 if (emptyline) bc_file_putchar(&vm->fout, bc_flush_err, '\n'); 4452 vm->nchars = 0; 4453 } 4454 4455 void 4456 bc_num_printDigs(const BcDig* n, size_t len, bool emptyline) 4457 { 4458 size_t i; 4459 4460 for (i = len - 1; i < len; --i) 4461 { 4462 bc_file_printf(&vm->fout, " %lu", (unsigned long) n[i]); 4463 } 4464 4465 bc_file_putchar(&vm->fout, bc_flush_err, '\n'); 4466 if (emptyline) bc_file_putchar(&vm->fout, bc_flush_err, '\n'); 4467 vm->nchars = 0; 4468 } 4469 4470 void 4471 bc_num_printWithDigs(const BcNum* n, const char* name, bool emptyline) 4472 { 4473 bc_file_puts(&vm->fout, bc_flush_none, name); 4474 bc_file_printf(&vm->fout, " len: %zu, rdx: %zu, scale: %zu\n", name, n->len, 4475 BC_NUM_RDX_VAL(n), n->scale); 4476 bc_num_printDigs(n->num, n->len, emptyline); 4477 } 4478 4479 void 4480 bc_num_dump(const char* varname, const BcNum* n) 4481 { 4482 ulong i, scale = n->scale; 4483 4484 bc_file_printf(&vm->ferr, "\n%s = %s", varname, 4485 n->len ? (BC_NUM_NEG(n) ? "-" : "+") : "0 "); 4486 4487 for (i = n->len - 1; i < n->len; --i) 4488 { 4489 if (i + 1 == BC_NUM_RDX_VAL(n)) 4490 { 4491 bc_file_puts(&vm->ferr, bc_flush_none, ". "); 4492 } 4493 4494 if (scale / BC_BASE_DIGS != BC_NUM_RDX_VAL(n) - i - 1) 4495 { 4496 bc_file_printf(&vm->ferr, "%lu ", (unsigned long) n->num[i]); 4497 } 4498 else 4499 { 4500 int mod = scale % BC_BASE_DIGS; 4501 int d = BC_BASE_DIGS - mod; 4502 BcDig div; 4503 4504 if (mod != 0) 4505 { 4506 div = n->num[i] / ((BcDig) bc_num_pow10[(ulong) d]); 4507 bc_file_printf(&vm->ferr, "%lu", (unsigned long) div); 4508 } 4509 4510 div = n->num[i] % ((BcDig) bc_num_pow10[(ulong) d]); 4511 bc_file_printf(&vm->ferr, " ' %lu ", (unsigned long) div); 4512 } 4513 } 4514 4515 bc_file_printf(&vm->ferr, "(%zu | %zu.%zu / %zu) %lu\n", n->scale, n->len, 4516 BC_NUM_RDX_VAL(n), n->cap, (unsigned long) (void*) n->num); 4517 4518 bc_file_flush(&vm->ferr, bc_flush_err); 4519 } 4520 #endif // BC_DEBUG_CODE 4521