1 // Copyright 2011 the V8 project authors. All rights reserved. 2 // Redistribution and use in source and binary forms, with or without 3 // modification, are permitted provided that the following conditions are 4 // met: 5 // 6 // * Redistributions of source code must retain the above copyright 7 // notice, this list of conditions and the following disclaimer. 8 // * Redistributions in binary form must reproduce the above 9 // copyright notice, this list of conditions and the following 10 // disclaimer in the documentation and/or other materials provided 11 // with the distribution. 12 // * Neither the name of Google Inc. nor the names of its 13 // contributors may be used to endorse or promote products derived 14 // from this software without specific prior written permission. 15 // 16 // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 17 // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 18 // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 19 // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 20 // OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 21 // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 22 // LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 23 // DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 24 // THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 25 // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 26 // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 27 28 #include "../include/v8stdint.h" 29 #include "utils.h" 30 #include "bignum.h" 31 32 namespace v8 { 33 namespace internal { 34 35 Bignum::Bignum() 36 : bigits_(bigits_buffer_, kBigitCapacity), used_digits_(0), exponent_(0) { 37 for (int i = 0; i < kBigitCapacity; ++i) { 38 bigits_[i] = 0; 39 } 40 } 41 42 43 template<typename S> 44 static int BitSize(S value) { 45 return 8 * sizeof(value); 46 } 47 48 // Guaranteed to lie in one Bigit. 49 void Bignum::AssignUInt16(uint16_t value) { 50 ASSERT(kBigitSize >= BitSize(value)); 51 Zero(); 52 if (value == 0) return; 53 54 EnsureCapacity(1); 55 bigits_[0] = value; 56 used_digits_ = 1; 57 } 58 59 60 void Bignum::AssignUInt64(uint64_t value) { 61 const int kUInt64Size = 64; 62 63 Zero(); 64 if (value == 0) return; 65 66 int needed_bigits = kUInt64Size / kBigitSize + 1; 67 EnsureCapacity(needed_bigits); 68 for (int i = 0; i < needed_bigits; ++i) { 69 bigits_[i] = static_cast<Chunk>(value & kBigitMask); 70 value = value >> kBigitSize; 71 } 72 used_digits_ = needed_bigits; 73 Clamp(); 74 } 75 76 77 void Bignum::AssignBignum(const Bignum& other) { 78 exponent_ = other.exponent_; 79 for (int i = 0; i < other.used_digits_; ++i) { 80 bigits_[i] = other.bigits_[i]; 81 } 82 // Clear the excess digits (if there were any). 83 for (int i = other.used_digits_; i < used_digits_; ++i) { 84 bigits_[i] = 0; 85 } 86 used_digits_ = other.used_digits_; 87 } 88 89 90 static uint64_t ReadUInt64(Vector<const char> buffer, 91 int from, 92 int digits_to_read) { 93 uint64_t result = 0; 94 for (int i = from; i < from + digits_to_read; ++i) { 95 int digit = buffer[i] - '0'; 96 ASSERT(0 <= digit && digit <= 9); 97 result = result * 10 + digit; 98 } 99 return result; 100 } 101 102 103 void Bignum::AssignDecimalString(Vector<const char> value) { 104 // 2^64 = 18446744073709551616 > 10^19 105 const int kMaxUint64DecimalDigits = 19; 106 Zero(); 107 int length = value.length(); 108 int pos = 0; 109 // Let's just say that each digit needs 4 bits. 110 while (length >= kMaxUint64DecimalDigits) { 111 uint64_t digits = ReadUInt64(value, pos, kMaxUint64DecimalDigits); 112 pos += kMaxUint64DecimalDigits; 113 length -= kMaxUint64DecimalDigits; 114 MultiplyByPowerOfTen(kMaxUint64DecimalDigits); 115 AddUInt64(digits); 116 } 117 uint64_t digits = ReadUInt64(value, pos, length); 118 MultiplyByPowerOfTen(length); 119 AddUInt64(digits); 120 Clamp(); 121 } 122 123 124 static int HexCharValue(char c) { 125 if ('0' <= c && c <= '9') return c - '0'; 126 if ('a' <= c && c <= 'f') return 10 + c - 'a'; 127 if ('A' <= c && c <= 'F') return 10 + c - 'A'; 128 UNREACHABLE(); 129 return 0; // To make compiler happy. 130 } 131 132 133 void Bignum::AssignHexString(Vector<const char> value) { 134 Zero(); 135 int length = value.length(); 136 137 int needed_bigits = length * 4 / kBigitSize + 1; 138 EnsureCapacity(needed_bigits); 139 int string_index = length - 1; 140 for (int i = 0; i < needed_bigits - 1; ++i) { 141 // These bigits are guaranteed to be "full". 142 Chunk current_bigit = 0; 143 for (int j = 0; j < kBigitSize / 4; j++) { 144 current_bigit += HexCharValue(value[string_index--]) << (j * 4); 145 } 146 bigits_[i] = current_bigit; 147 } 148 used_digits_ = needed_bigits - 1; 149 150 Chunk most_significant_bigit = 0; // Could be = 0; 151 for (int j = 0; j <= string_index; ++j) { 152 most_significant_bigit <<= 4; 153 most_significant_bigit += HexCharValue(value[j]); 154 } 155 if (most_significant_bigit != 0) { 156 bigits_[used_digits_] = most_significant_bigit; 157 used_digits_++; 158 } 159 Clamp(); 160 } 161 162 163 void Bignum::AddUInt64(uint64_t operand) { 164 if (operand == 0) return; 165 Bignum other; 166 other.AssignUInt64(operand); 167 AddBignum(other); 168 } 169 170 171 void Bignum::AddBignum(const Bignum& other) { 172 ASSERT(IsClamped()); 173 ASSERT(other.IsClamped()); 174 175 // If this has a greater exponent than other append zero-bigits to this. 176 // After this call exponent_ <= other.exponent_. 177 Align(other); 178 179 // There are two possibilities: 180 // aaaaaaaaaaa 0000 (where the 0s represent a's exponent) 181 // bbbbb 00000000 182 // ---------------- 183 // ccccccccccc 0000 184 // or 185 // aaaaaaaaaa 0000 186 // bbbbbbbbb 0000000 187 // ----------------- 188 // cccccccccccc 0000 189 // In both cases we might need a carry bigit. 190 191 EnsureCapacity(1 + Max(BigitLength(), other.BigitLength()) - exponent_); 192 Chunk carry = 0; 193 int bigit_pos = other.exponent_ - exponent_; 194 ASSERT(bigit_pos >= 0); 195 for (int i = 0; i < other.used_digits_; ++i) { 196 Chunk sum = bigits_[bigit_pos] + other.bigits_[i] + carry; 197 bigits_[bigit_pos] = sum & kBigitMask; 198 carry = sum >> kBigitSize; 199 bigit_pos++; 200 } 201 202 while (carry != 0) { 203 Chunk sum = bigits_[bigit_pos] + carry; 204 bigits_[bigit_pos] = sum & kBigitMask; 205 carry = sum >> kBigitSize; 206 bigit_pos++; 207 } 208 used_digits_ = Max(bigit_pos, used_digits_); 209 ASSERT(IsClamped()); 210 } 211 212 213 void Bignum::SubtractBignum(const Bignum& other) { 214 ASSERT(IsClamped()); 215 ASSERT(other.IsClamped()); 216 // We require this to be bigger than other. 217 ASSERT(LessEqual(other, *this)); 218 219 Align(other); 220 221 int offset = other.exponent_ - exponent_; 222 Chunk borrow = 0; 223 int i; 224 for (i = 0; i < other.used_digits_; ++i) { 225 ASSERT((borrow == 0) || (borrow == 1)); 226 Chunk difference = bigits_[i + offset] - other.bigits_[i] - borrow; 227 bigits_[i + offset] = difference & kBigitMask; 228 borrow = difference >> (kChunkSize - 1); 229 } 230 while (borrow != 0) { 231 Chunk difference = bigits_[i + offset] - borrow; 232 bigits_[i + offset] = difference & kBigitMask; 233 borrow = difference >> (kChunkSize - 1); 234 ++i; 235 } 236 Clamp(); 237 } 238 239 240 void Bignum::ShiftLeft(int shift_amount) { 241 if (used_digits_ == 0) return; 242 exponent_ += shift_amount / kBigitSize; 243 int local_shift = shift_amount % kBigitSize; 244 EnsureCapacity(used_digits_ + 1); 245 BigitsShiftLeft(local_shift); 246 } 247 248 249 void Bignum::MultiplyByUInt32(uint32_t factor) { 250 if (factor == 1) return; 251 if (factor == 0) { 252 Zero(); 253 return; 254 } 255 if (used_digits_ == 0) return; 256 257 // The product of a bigit with the factor is of size kBigitSize + 32. 258 // Assert that this number + 1 (for the carry) fits into double chunk. 259 ASSERT(kDoubleChunkSize >= kBigitSize + 32 + 1); 260 DoubleChunk carry = 0; 261 for (int i = 0; i < used_digits_; ++i) { 262 DoubleChunk product = static_cast<DoubleChunk>(factor) * bigits_[i] + carry; 263 bigits_[i] = static_cast<Chunk>(product & kBigitMask); 264 carry = (product >> kBigitSize); 265 } 266 while (carry != 0) { 267 EnsureCapacity(used_digits_ + 1); 268 bigits_[used_digits_] = static_cast<Chunk>(carry & kBigitMask); 269 used_digits_++; 270 carry >>= kBigitSize; 271 } 272 } 273 274 275 void Bignum::MultiplyByUInt64(uint64_t factor) { 276 if (factor == 1) return; 277 if (factor == 0) { 278 Zero(); 279 return; 280 } 281 ASSERT(kBigitSize < 32); 282 uint64_t carry = 0; 283 uint64_t low = factor & 0xFFFFFFFF; 284 uint64_t high = factor >> 32; 285 for (int i = 0; i < used_digits_; ++i) { 286 uint64_t product_low = low * bigits_[i]; 287 uint64_t product_high = high * bigits_[i]; 288 uint64_t tmp = (carry & kBigitMask) + product_low; 289 bigits_[i] = static_cast<Chunk>(tmp & kBigitMask); 290 carry = (carry >> kBigitSize) + (tmp >> kBigitSize) + 291 (product_high << (32 - kBigitSize)); 292 } 293 while (carry != 0) { 294 EnsureCapacity(used_digits_ + 1); 295 bigits_[used_digits_] = static_cast<Chunk>(carry & kBigitMask); 296 used_digits_++; 297 carry >>= kBigitSize; 298 } 299 } 300 301 302 void Bignum::MultiplyByPowerOfTen(int exponent) { 303 const uint64_t kFive27 = V8_2PART_UINT64_C(0x6765c793, fa10079d); 304 const uint16_t kFive1 = 5; 305 const uint16_t kFive2 = kFive1 * 5; 306 const uint16_t kFive3 = kFive2 * 5; 307 const uint16_t kFive4 = kFive3 * 5; 308 const uint16_t kFive5 = kFive4 * 5; 309 const uint16_t kFive6 = kFive5 * 5; 310 const uint32_t kFive7 = kFive6 * 5; 311 const uint32_t kFive8 = kFive7 * 5; 312 const uint32_t kFive9 = kFive8 * 5; 313 const uint32_t kFive10 = kFive9 * 5; 314 const uint32_t kFive11 = kFive10 * 5; 315 const uint32_t kFive12 = kFive11 * 5; 316 const uint32_t kFive13 = kFive12 * 5; 317 const uint32_t kFive1_to_12[] = 318 { kFive1, kFive2, kFive3, kFive4, kFive5, kFive6, 319 kFive7, kFive8, kFive9, kFive10, kFive11, kFive12 }; 320 321 ASSERT(exponent >= 0); 322 if (exponent == 0) return; 323 if (used_digits_ == 0) return; 324 325 // We shift by exponent at the end just before returning. 326 int remaining_exponent = exponent; 327 while (remaining_exponent >= 27) { 328 MultiplyByUInt64(kFive27); 329 remaining_exponent -= 27; 330 } 331 while (remaining_exponent >= 13) { 332 MultiplyByUInt32(kFive13); 333 remaining_exponent -= 13; 334 } 335 if (remaining_exponent > 0) { 336 MultiplyByUInt32(kFive1_to_12[remaining_exponent - 1]); 337 } 338 ShiftLeft(exponent); 339 } 340 341 342 void Bignum::Square() { 343 ASSERT(IsClamped()); 344 int product_length = 2 * used_digits_; 345 EnsureCapacity(product_length); 346 347 // Comba multiplication: compute each column separately. 348 // Example: r = a2a1a0 * b2b1b0. 349 // r = 1 * a0b0 + 350 // 10 * (a1b0 + a0b1) + 351 // 100 * (a2b0 + a1b1 + a0b2) + 352 // 1000 * (a2b1 + a1b2) + 353 // 10000 * a2b2 354 // 355 // In the worst case we have to accumulate nb-digits products of digit*digit. 356 // 357 // Assert that the additional number of bits in a DoubleChunk are enough to 358 // sum up used_digits of Bigit*Bigit. 359 if ((1 << (2 * (kChunkSize - kBigitSize))) <= used_digits_) { 360 UNIMPLEMENTED(); 361 } 362 DoubleChunk accumulator = 0; 363 // First shift the digits so we don't overwrite them. 364 int copy_offset = used_digits_; 365 for (int i = 0; i < used_digits_; ++i) { 366 bigits_[copy_offset + i] = bigits_[i]; 367 } 368 // We have two loops to avoid some 'if's in the loop. 369 for (int i = 0; i < used_digits_; ++i) { 370 // Process temporary digit i with power i. 371 // The sum of the two indices must be equal to i. 372 int bigit_index1 = i; 373 int bigit_index2 = 0; 374 // Sum all of the sub-products. 375 while (bigit_index1 >= 0) { 376 Chunk chunk1 = bigits_[copy_offset + bigit_index1]; 377 Chunk chunk2 = bigits_[copy_offset + bigit_index2]; 378 accumulator += static_cast<DoubleChunk>(chunk1) * chunk2; 379 bigit_index1--; 380 bigit_index2++; 381 } 382 bigits_[i] = static_cast<Chunk>(accumulator) & kBigitMask; 383 accumulator >>= kBigitSize; 384 } 385 for (int i = used_digits_; i < product_length; ++i) { 386 int bigit_index1 = used_digits_ - 1; 387 int bigit_index2 = i - bigit_index1; 388 // Invariant: sum of both indices is again equal to i. 389 // Inner loop runs 0 times on last iteration, emptying accumulator. 390 while (bigit_index2 < used_digits_) { 391 Chunk chunk1 = bigits_[copy_offset + bigit_index1]; 392 Chunk chunk2 = bigits_[copy_offset + bigit_index2]; 393 accumulator += static_cast<DoubleChunk>(chunk1) * chunk2; 394 bigit_index1--; 395 bigit_index2++; 396 } 397 // The overwritten bigits_[i] will never be read in further loop iterations, 398 // because bigit_index1 and bigit_index2 are always greater 399 // than i - used_digits_. 400 bigits_[i] = static_cast<Chunk>(accumulator) & kBigitMask; 401 accumulator >>= kBigitSize; 402 } 403 // Since the result was guaranteed to lie inside the number the 404 // accumulator must be 0 now. 405 ASSERT(accumulator == 0); 406 407 // Don't forget to update the used_digits and the exponent. 408 used_digits_ = product_length; 409 exponent_ *= 2; 410 Clamp(); 411 } 412 413 414 void Bignum::AssignPowerUInt16(uint16_t base, int power_exponent) { 415 ASSERT(base != 0); 416 ASSERT(power_exponent >= 0); 417 if (power_exponent == 0) { 418 AssignUInt16(1); 419 return; 420 } 421 Zero(); 422 int shifts = 0; 423 // We expect base to be in range 2-32, and most often to be 10. 424 // It does not make much sense to implement different algorithms for counting 425 // the bits. 426 while ((base & 1) == 0) { 427 base >>= 1; 428 shifts++; 429 } 430 int bit_size = 0; 431 int tmp_base = base; 432 while (tmp_base != 0) { 433 tmp_base >>= 1; 434 bit_size++; 435 } 436 int final_size = bit_size * power_exponent; 437 // 1 extra bigit for the shifting, and one for rounded final_size. 438 EnsureCapacity(final_size / kBigitSize + 2); 439 440 // Left to Right exponentiation. 441 int mask = 1; 442 while (power_exponent >= mask) mask <<= 1; 443 444 // The mask is now pointing to the bit above the most significant 1-bit of 445 // power_exponent. 446 // Get rid of first 1-bit; 447 mask >>= 2; 448 uint64_t this_value = base; 449 450 bool delayed_multipliciation = false; 451 const uint64_t max_32bits = 0xFFFFFFFF; 452 while (mask != 0 && this_value <= max_32bits) { 453 this_value = this_value * this_value; 454 // Verify that there is enough space in this_value to perform the 455 // multiplication. The first bit_size bits must be 0. 456 if ((power_exponent & mask) != 0) { 457 uint64_t base_bits_mask = 458 ~((static_cast<uint64_t>(1) << (64 - bit_size)) - 1); 459 bool high_bits_zero = (this_value & base_bits_mask) == 0; 460 if (high_bits_zero) { 461 this_value *= base; 462 } else { 463 delayed_multipliciation = true; 464 } 465 } 466 mask >>= 1; 467 } 468 AssignUInt64(this_value); 469 if (delayed_multipliciation) { 470 MultiplyByUInt32(base); 471 } 472 473 // Now do the same thing as a bignum. 474 while (mask != 0) { 475 Square(); 476 if ((power_exponent & mask) != 0) { 477 MultiplyByUInt32(base); 478 } 479 mask >>= 1; 480 } 481 482 // And finally add the saved shifts. 483 ShiftLeft(shifts * power_exponent); 484 } 485 486 487 // Precondition: this/other < 16bit. 488 uint16_t Bignum::DivideModuloIntBignum(const Bignum& other) { 489 ASSERT(IsClamped()); 490 ASSERT(other.IsClamped()); 491 ASSERT(other.used_digits_ > 0); 492 493 // Easy case: if we have less digits than the divisor than the result is 0. 494 // Note: this handles the case where this == 0, too. 495 if (BigitLength() < other.BigitLength()) { 496 return 0; 497 } 498 499 Align(other); 500 501 uint16_t result = 0; 502 503 // Start by removing multiples of 'other' until both numbers have the same 504 // number of digits. 505 while (BigitLength() > other.BigitLength()) { 506 // This naive approach is extremely inefficient if the this divided other 507 // might be big. This function is implemented for doubleToString where 508 // the result should be small (less than 10). 509 ASSERT(other.bigits_[other.used_digits_ - 1] >= ((1 << kBigitSize) / 16)); 510 // Remove the multiples of the first digit. 511 // Example this = 23 and other equals 9. -> Remove 2 multiples. 512 result += bigits_[used_digits_ - 1]; 513 SubtractTimes(other, bigits_[used_digits_ - 1]); 514 } 515 516 ASSERT(BigitLength() == other.BigitLength()); 517 518 // Both bignums are at the same length now. 519 // Since other has more than 0 digits we know that the access to 520 // bigits_[used_digits_ - 1] is safe. 521 Chunk this_bigit = bigits_[used_digits_ - 1]; 522 Chunk other_bigit = other.bigits_[other.used_digits_ - 1]; 523 524 if (other.used_digits_ == 1) { 525 // Shortcut for easy (and common) case. 526 int quotient = this_bigit / other_bigit; 527 bigits_[used_digits_ - 1] = this_bigit - other_bigit * quotient; 528 result += quotient; 529 Clamp(); 530 return result; 531 } 532 533 int division_estimate = this_bigit / (other_bigit + 1); 534 result += division_estimate; 535 SubtractTimes(other, division_estimate); 536 537 if (other_bigit * (division_estimate + 1) > this_bigit) { 538 // No need to even try to subtract. Even if other's remaining digits were 0 539 // another subtraction would be too much. 540 return result; 541 } 542 543 while (LessEqual(other, *this)) { 544 SubtractBignum(other); 545 result++; 546 } 547 return result; 548 } 549 550 551 template<typename S> 552 static int SizeInHexChars(S number) { 553 ASSERT(number > 0); 554 int result = 0; 555 while (number != 0) { 556 number >>= 4; 557 result++; 558 } 559 return result; 560 } 561 562 563 static char HexCharOfValue(int value) { 564 ASSERT(0 <= value && value <= 16); 565 if (value < 10) return value + '0'; 566 return value - 10 + 'A'; 567 } 568 569 570 bool Bignum::ToHexString(char* buffer, int buffer_size) const { 571 ASSERT(IsClamped()); 572 // Each bigit must be printable as separate hex-character. 573 ASSERT(kBigitSize % 4 == 0); 574 const int kHexCharsPerBigit = kBigitSize / 4; 575 576 if (used_digits_ == 0) { 577 if (buffer_size < 2) return false; 578 buffer[0] = '0'; 579 buffer[1] = '\0'; 580 return true; 581 } 582 // We add 1 for the terminating '\0' character. 583 int needed_chars = (BigitLength() - 1) * kHexCharsPerBigit + 584 SizeInHexChars(bigits_[used_digits_ - 1]) + 1; 585 if (needed_chars > buffer_size) return false; 586 int string_index = needed_chars - 1; 587 buffer[string_index--] = '\0'; 588 for (int i = 0; i < exponent_; ++i) { 589 for (int j = 0; j < kHexCharsPerBigit; ++j) { 590 buffer[string_index--] = '0'; 591 } 592 } 593 for (int i = 0; i < used_digits_ - 1; ++i) { 594 Chunk current_bigit = bigits_[i]; 595 for (int j = 0; j < kHexCharsPerBigit; ++j) { 596 buffer[string_index--] = HexCharOfValue(current_bigit & 0xF); 597 current_bigit >>= 4; 598 } 599 } 600 // And finally the last bigit. 601 Chunk most_significant_bigit = bigits_[used_digits_ - 1]; 602 while (most_significant_bigit != 0) { 603 buffer[string_index--] = HexCharOfValue(most_significant_bigit & 0xF); 604 most_significant_bigit >>= 4; 605 } 606 return true; 607 } 608 609 610 Bignum::Chunk Bignum::BigitAt(int index) const { 611 if (index >= BigitLength()) return 0; 612 if (index < exponent_) return 0; 613 return bigits_[index - exponent_]; 614 } 615 616 617 int Bignum::Compare(const Bignum& a, const Bignum& b) { 618 ASSERT(a.IsClamped()); 619 ASSERT(b.IsClamped()); 620 int bigit_length_a = a.BigitLength(); 621 int bigit_length_b = b.BigitLength(); 622 if (bigit_length_a < bigit_length_b) return -1; 623 if (bigit_length_a > bigit_length_b) return +1; 624 for (int i = bigit_length_a - 1; i >= Min(a.exponent_, b.exponent_); --i) { 625 Chunk bigit_a = a.BigitAt(i); 626 Chunk bigit_b = b.BigitAt(i); 627 if (bigit_a < bigit_b) return -1; 628 if (bigit_a > bigit_b) return +1; 629 // Otherwise they are equal up to this digit. Try the next digit. 630 } 631 return 0; 632 } 633 634 635 int Bignum::PlusCompare(const Bignum& a, const Bignum& b, const Bignum& c) { 636 ASSERT(a.IsClamped()); 637 ASSERT(b.IsClamped()); 638 ASSERT(c.IsClamped()); 639 if (a.BigitLength() < b.BigitLength()) { 640 return PlusCompare(b, a, c); 641 } 642 if (a.BigitLength() + 1 < c.BigitLength()) return -1; 643 if (a.BigitLength() > c.BigitLength()) return +1; 644 // The exponent encodes 0-bigits. So if there are more 0-digits in 'a' than 645 // 'b' has digits, then the bigit-length of 'a'+'b' must be equal to the one 646 // of 'a'. 647 if (a.exponent_ >= b.BigitLength() && a.BigitLength() < c.BigitLength()) { 648 return -1; 649 } 650 651 Chunk borrow = 0; 652 // Starting at min_exponent all digits are == 0. So no need to compare them. 653 int min_exponent = Min(Min(a.exponent_, b.exponent_), c.exponent_); 654 for (int i = c.BigitLength() - 1; i >= min_exponent; --i) { 655 Chunk chunk_a = a.BigitAt(i); 656 Chunk chunk_b = b.BigitAt(i); 657 Chunk chunk_c = c.BigitAt(i); 658 Chunk sum = chunk_a + chunk_b; 659 if (sum > chunk_c + borrow) { 660 return +1; 661 } else { 662 borrow = chunk_c + borrow - sum; 663 if (borrow > 1) return -1; 664 borrow <<= kBigitSize; 665 } 666 } 667 if (borrow == 0) return 0; 668 return -1; 669 } 670 671 672 void Bignum::Clamp() { 673 while (used_digits_ > 0 && bigits_[used_digits_ - 1] == 0) { 674 used_digits_--; 675 } 676 if (used_digits_ == 0) { 677 // Zero. 678 exponent_ = 0; 679 } 680 } 681 682 683 bool Bignum::IsClamped() const { 684 return used_digits_ == 0 || bigits_[used_digits_ - 1] != 0; 685 } 686 687 688 void Bignum::Zero() { 689 for (int i = 0; i < used_digits_; ++i) { 690 bigits_[i] = 0; 691 } 692 used_digits_ = 0; 693 exponent_ = 0; 694 } 695 696 697 void Bignum::Align(const Bignum& other) { 698 if (exponent_ > other.exponent_) { 699 // If "X" represents a "hidden" digit (by the exponent) then we are in the 700 // following case (a == this, b == other): 701 // a: aaaaaaXXXX or a: aaaaaXXX 702 // b: bbbbbbX b: bbbbbbbbXX 703 // We replace some of the hidden digits (X) of a with 0 digits. 704 // a: aaaaaa000X or a: aaaaa0XX 705 int zero_digits = exponent_ - other.exponent_; 706 EnsureCapacity(used_digits_ + zero_digits); 707 for (int i = used_digits_ - 1; i >= 0; --i) { 708 bigits_[i + zero_digits] = bigits_[i]; 709 } 710 for (int i = 0; i < zero_digits; ++i) { 711 bigits_[i] = 0; 712 } 713 used_digits_ += zero_digits; 714 exponent_ -= zero_digits; 715 ASSERT(used_digits_ >= 0); 716 ASSERT(exponent_ >= 0); 717 } 718 } 719 720 721 void Bignum::BigitsShiftLeft(int shift_amount) { 722 ASSERT(shift_amount < kBigitSize); 723 ASSERT(shift_amount >= 0); 724 Chunk carry = 0; 725 for (int i = 0; i < used_digits_; ++i) { 726 Chunk new_carry = bigits_[i] >> (kBigitSize - shift_amount); 727 bigits_[i] = ((bigits_[i] << shift_amount) + carry) & kBigitMask; 728 carry = new_carry; 729 } 730 if (carry != 0) { 731 bigits_[used_digits_] = carry; 732 used_digits_++; 733 } 734 } 735 736 737 void Bignum::SubtractTimes(const Bignum& other, int factor) { 738 ASSERT(exponent_ <= other.exponent_); 739 if (factor < 3) { 740 for (int i = 0; i < factor; ++i) { 741 SubtractBignum(other); 742 } 743 return; 744 } 745 Chunk borrow = 0; 746 int exponent_diff = other.exponent_ - exponent_; 747 for (int i = 0; i < other.used_digits_; ++i) { 748 DoubleChunk product = static_cast<DoubleChunk>(factor) * other.bigits_[i]; 749 DoubleChunk remove = borrow + product; 750 Chunk difference = 751 bigits_[i + exponent_diff] - static_cast<Chunk>(remove & kBigitMask); 752 bigits_[i + exponent_diff] = difference & kBigitMask; 753 borrow = static_cast<Chunk>((difference >> (kChunkSize - 1)) + 754 (remove >> kBigitSize)); 755 } 756 for (int i = other.used_digits_ + exponent_diff; i < used_digits_; ++i) { 757 if (borrow == 0) return; 758 Chunk difference = bigits_[i] - borrow; 759 bigits_[i] = difference & kBigitMask; 760 borrow = difference >> (kChunkSize - 1); 761 ++i; 762 } 763 Clamp(); 764 } 765 766 767 } } // namespace v8::internal 768