]> git.saurik.com Git - apple/icu.git/blame - icuSources/i18n/double-conversion-bignum.cpp
ICU-64252.0.1.tar.gz
[apple/icu.git] / icuSources / i18n / double-conversion-bignum.cpp
CommitLineData
0f5d89e8
A
1// © 2018 and later: Unicode, Inc. and others.
2// License & terms of use: http://www.unicode.org/copyright.html
3//
4// From the double-conversion library. Original license:
5//
6// Copyright 2010 the V8 project authors. All rights reserved.
7// Redistribution and use in source and binary forms, with or without
8// modification, are permitted provided that the following conditions are
9// met:
10//
11// * Redistributions of source code must retain the above copyright
12// notice, this list of conditions and the following disclaimer.
13// * Redistributions in binary form must reproduce the above
14// copyright notice, this list of conditions and the following
15// disclaimer in the documentation and/or other materials provided
16// with the distribution.
17// * Neither the name of Google Inc. nor the names of its
18// contributors may be used to endorse or promote products derived
19// from this software without specific prior written permission.
20//
21// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
24// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
25// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
26// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
27// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
31// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32
33// ICU PATCH: ifdef around UCONFIG_NO_FORMATTING
34#include "unicode/utypes.h"
35#if !UCONFIG_NO_FORMATTING
36
37// ICU PATCH: Customize header file paths for ICU.
38
39#include "double-conversion-bignum.h"
40#include "double-conversion-utils.h"
41
42// ICU PATCH: Wrap in ICU namespace
43U_NAMESPACE_BEGIN
44
45namespace double_conversion {
46
47Bignum::Bignum()
3d1f044b 48 : bigits_buffer_(), bigits_(bigits_buffer_, kBigitCapacity), used_digits_(0), exponent_(0) {
0f5d89e8
A
49 for (int i = 0; i < kBigitCapacity; ++i) {
50 bigits_[i] = 0;
51 }
52}
53
54
55template<typename S>
56static int BitSize(S value) {
57 (void) value; // Mark variable as used.
58 return 8 * sizeof(value);
59}
60
61// Guaranteed to lie in one Bigit.
62void Bignum::AssignUInt16(uint16_t value) {
63 ASSERT(kBigitSize >= BitSize(value));
64 Zero();
65 if (value == 0) return;
66
67 EnsureCapacity(1);
68 bigits_[0] = value;
69 used_digits_ = 1;
70}
71
72
73void Bignum::AssignUInt64(uint64_t value) {
74 const int kUInt64Size = 64;
75
76 Zero();
77 if (value == 0) return;
78
79 int needed_bigits = kUInt64Size / kBigitSize + 1;
80 EnsureCapacity(needed_bigits);
81 for (int i = 0; i < needed_bigits; ++i) {
82 bigits_[i] = value & kBigitMask;
83 value = value >> kBigitSize;
84 }
85 used_digits_ = needed_bigits;
86 Clamp();
87}
88
89
90void Bignum::AssignBignum(const Bignum& other) {
91 exponent_ = other.exponent_;
92 for (int i = 0; i < other.used_digits_; ++i) {
93 bigits_[i] = other.bigits_[i];
94 }
95 // Clear the excess digits (if there were any).
96 for (int i = other.used_digits_; i < used_digits_; ++i) {
97 bigits_[i] = 0;
98 }
99 used_digits_ = other.used_digits_;
100}
101
102
103static uint64_t ReadUInt64(Vector<const char> buffer,
104 int from,
105 int digits_to_read) {
106 uint64_t result = 0;
107 for (int i = from; i < from + digits_to_read; ++i) {
108 int digit = buffer[i] - '0';
109 ASSERT(0 <= digit && digit <= 9);
110 result = result * 10 + digit;
111 }
112 return result;
113}
114
115
116void Bignum::AssignDecimalString(Vector<const char> value) {
117 // 2^64 = 18446744073709551616 > 10^19
118 const int kMaxUint64DecimalDigits = 19;
119 Zero();
120 int length = value.length();
121 unsigned int pos = 0;
122 // Let's just say that each digit needs 4 bits.
123 while (length >= kMaxUint64DecimalDigits) {
124 uint64_t digits = ReadUInt64(value, pos, kMaxUint64DecimalDigits);
125 pos += kMaxUint64DecimalDigits;
126 length -= kMaxUint64DecimalDigits;
127 MultiplyByPowerOfTen(kMaxUint64DecimalDigits);
128 AddUInt64(digits);
129 }
130 uint64_t digits = ReadUInt64(value, pos, length);
131 MultiplyByPowerOfTen(length);
132 AddUInt64(digits);
133 Clamp();
134}
135
136
137static int HexCharValue(char c) {
138 if ('0' <= c && c <= '9') return c - '0';
139 if ('a' <= c && c <= 'f') return 10 + c - 'a';
140 ASSERT('A' <= c && c <= 'F');
141 return 10 + c - 'A';
142}
143
144
145void Bignum::AssignHexString(Vector<const char> value) {
146 Zero();
147 int length = value.length();
148
149 int needed_bigits = length * 4 / kBigitSize + 1;
150 EnsureCapacity(needed_bigits);
151 int string_index = length - 1;
152 for (int i = 0; i < needed_bigits - 1; ++i) {
153 // These bigits are guaranteed to be "full".
154 Chunk current_bigit = 0;
155 for (int j = 0; j < kBigitSize / 4; j++) {
156 current_bigit += HexCharValue(value[string_index--]) << (j * 4);
157 }
158 bigits_[i] = current_bigit;
159 }
160 used_digits_ = needed_bigits - 1;
161
162 Chunk most_significant_bigit = 0; // Could be = 0;
163 for (int j = 0; j <= string_index; ++j) {
164 most_significant_bigit <<= 4;
165 most_significant_bigit += HexCharValue(value[j]);
166 }
167 if (most_significant_bigit != 0) {
168 bigits_[used_digits_] = most_significant_bigit;
169 used_digits_++;
170 }
171 Clamp();
172}
173
174
175void Bignum::AddUInt64(uint64_t operand) {
176 if (operand == 0) return;
177 Bignum other;
178 other.AssignUInt64(operand);
179 AddBignum(other);
180}
181
182
183void Bignum::AddBignum(const Bignum& other) {
184 ASSERT(IsClamped());
185 ASSERT(other.IsClamped());
186
187 // If this has a greater exponent than other append zero-bigits to this.
188 // After this call exponent_ <= other.exponent_.
189 Align(other);
190
191 // There are two possibilities:
192 // aaaaaaaaaaa 0000 (where the 0s represent a's exponent)
193 // bbbbb 00000000
194 // ----------------
195 // ccccccccccc 0000
196 // or
197 // aaaaaaaaaa 0000
198 // bbbbbbbbb 0000000
199 // -----------------
200 // cccccccccccc 0000
201 // In both cases we might need a carry bigit.
202
203 EnsureCapacity(1 + Max(BigitLength(), other.BigitLength()) - exponent_);
204 Chunk carry = 0;
205 int bigit_pos = other.exponent_ - exponent_;
206 ASSERT(bigit_pos >= 0);
207 for (int i = 0; i < other.used_digits_; ++i) {
208 Chunk sum = bigits_[bigit_pos] + other.bigits_[i] + carry;
209 bigits_[bigit_pos] = sum & kBigitMask;
210 carry = sum >> kBigitSize;
211 bigit_pos++;
212 }
213
214 while (carry != 0) {
215 Chunk sum = bigits_[bigit_pos] + carry;
216 bigits_[bigit_pos] = sum & kBigitMask;
217 carry = sum >> kBigitSize;
218 bigit_pos++;
219 }
220 used_digits_ = Max(bigit_pos, used_digits_);
221 ASSERT(IsClamped());
222}
223
224
225void Bignum::SubtractBignum(const Bignum& other) {
226 ASSERT(IsClamped());
227 ASSERT(other.IsClamped());
228 // We require this to be bigger than other.
229 ASSERT(LessEqual(other, *this));
230
231 Align(other);
232
233 int offset = other.exponent_ - exponent_;
234 Chunk borrow = 0;
235 int i;
236 for (i = 0; i < other.used_digits_; ++i) {
237 ASSERT((borrow == 0) || (borrow == 1));
238 Chunk difference = bigits_[i + offset] - other.bigits_[i] - borrow;
239 bigits_[i + offset] = difference & kBigitMask;
240 borrow = difference >> (kChunkSize - 1);
241 }
242 while (borrow != 0) {
243 Chunk difference = bigits_[i + offset] - borrow;
244 bigits_[i + offset] = difference & kBigitMask;
245 borrow = difference >> (kChunkSize - 1);
246 ++i;
247 }
248 Clamp();
249}
250
251
252void Bignum::ShiftLeft(int shift_amount) {
253 if (used_digits_ == 0) return;
254 exponent_ += shift_amount / kBigitSize;
255 int local_shift = shift_amount % kBigitSize;
256 EnsureCapacity(used_digits_ + 1);
257 BigitsShiftLeft(local_shift);
258}
259
260
261void Bignum::MultiplyByUInt32(uint32_t factor) {
262 if (factor == 1) return;
263 if (factor == 0) {
264 Zero();
265 return;
266 }
267 if (used_digits_ == 0) return;
268
269 // The product of a bigit with the factor is of size kBigitSize + 32.
270 // Assert that this number + 1 (for the carry) fits into double chunk.
271 ASSERT(kDoubleChunkSize >= kBigitSize + 32 + 1);
272 DoubleChunk carry = 0;
273 for (int i = 0; i < used_digits_; ++i) {
274 DoubleChunk product = static_cast<DoubleChunk>(factor) * bigits_[i] + carry;
275 bigits_[i] = static_cast<Chunk>(product & kBigitMask);
276 carry = (product >> kBigitSize);
277 }
278 while (carry != 0) {
279 EnsureCapacity(used_digits_ + 1);
280 bigits_[used_digits_] = carry & kBigitMask;
281 used_digits_++;
282 carry >>= kBigitSize;
283 }
284}
285
286
287void Bignum::MultiplyByUInt64(uint64_t factor) {
288 if (factor == 1) return;
289 if (factor == 0) {
290 Zero();
291 return;
292 }
293 ASSERT(kBigitSize < 32);
294 uint64_t carry = 0;
295 uint64_t low = factor & 0xFFFFFFFF;
296 uint64_t high = factor >> 32;
297 for (int i = 0; i < used_digits_; ++i) {
298 uint64_t product_low = low * bigits_[i];
299 uint64_t product_high = high * bigits_[i];
300 uint64_t tmp = (carry & kBigitMask) + product_low;
301 bigits_[i] = tmp & kBigitMask;
302 carry = (carry >> kBigitSize) + (tmp >> kBigitSize) +
303 (product_high << (32 - kBigitSize));
304 }
305 while (carry != 0) {
306 EnsureCapacity(used_digits_ + 1);
307 bigits_[used_digits_] = carry & kBigitMask;
308 used_digits_++;
309 carry >>= kBigitSize;
310 }
311}
312
313
314void Bignum::MultiplyByPowerOfTen(int exponent) {
315 const uint64_t kFive27 = UINT64_2PART_C(0x6765c793, fa10079d);
316 const uint16_t kFive1 = 5;
317 const uint16_t kFive2 = kFive1 * 5;
318 const uint16_t kFive3 = kFive2 * 5;
319 const uint16_t kFive4 = kFive3 * 5;
320 const uint16_t kFive5 = kFive4 * 5;
321 const uint16_t kFive6 = kFive5 * 5;
322 const uint32_t kFive7 = kFive6 * 5;
323 const uint32_t kFive8 = kFive7 * 5;
324 const uint32_t kFive9 = kFive8 * 5;
325 const uint32_t kFive10 = kFive9 * 5;
326 const uint32_t kFive11 = kFive10 * 5;
327 const uint32_t kFive12 = kFive11 * 5;
328 const uint32_t kFive13 = kFive12 * 5;
329 const uint32_t kFive1_to_12[] =
330 { kFive1, kFive2, kFive3, kFive4, kFive5, kFive6,
331 kFive7, kFive8, kFive9, kFive10, kFive11, kFive12 };
332
333 ASSERT(exponent >= 0);
334 if (exponent == 0) return;
335 if (used_digits_ == 0) return;
336
337 // We shift by exponent at the end just before returning.
338 int remaining_exponent = exponent;
339 while (remaining_exponent >= 27) {
340 MultiplyByUInt64(kFive27);
341 remaining_exponent -= 27;
342 }
343 while (remaining_exponent >= 13) {
344 MultiplyByUInt32(kFive13);
345 remaining_exponent -= 13;
346 }
347 if (remaining_exponent > 0) {
348 MultiplyByUInt32(kFive1_to_12[remaining_exponent - 1]);
349 }
350 ShiftLeft(exponent);
351}
352
353
354void Bignum::Square() {
355 ASSERT(IsClamped());
356 int product_length = 2 * used_digits_;
357 EnsureCapacity(product_length);
358
359 // Comba multiplication: compute each column separately.
360 // Example: r = a2a1a0 * b2b1b0.
361 // r = 1 * a0b0 +
362 // 10 * (a1b0 + a0b1) +
363 // 100 * (a2b0 + a1b1 + a0b2) +
364 // 1000 * (a2b1 + a1b2) +
365 // 10000 * a2b2
366 //
367 // In the worst case we have to accumulate nb-digits products of digit*digit.
368 //
369 // Assert that the additional number of bits in a DoubleChunk are enough to
370 // sum up used_digits of Bigit*Bigit.
371 if ((1 << (2 * (kChunkSize - kBigitSize))) <= used_digits_) {
372 UNIMPLEMENTED();
373 }
374 DoubleChunk accumulator = 0;
375 // First shift the digits so we don't overwrite them.
376 int copy_offset = used_digits_;
377 for (int i = 0; i < used_digits_; ++i) {
378 bigits_[copy_offset + i] = bigits_[i];
379 }
380 // We have two loops to avoid some 'if's in the loop.
381 for (int i = 0; i < used_digits_; ++i) {
382 // Process temporary digit i with power i.
383 // The sum of the two indices must be equal to i.
384 int bigit_index1 = i;
385 int bigit_index2 = 0;
386 // Sum all of the sub-products.
387 while (bigit_index1 >= 0) {
388 Chunk chunk1 = bigits_[copy_offset + bigit_index1];
389 Chunk chunk2 = bigits_[copy_offset + bigit_index2];
390 accumulator += static_cast<DoubleChunk>(chunk1) * chunk2;
391 bigit_index1--;
392 bigit_index2++;
393 }
394 bigits_[i] = static_cast<Chunk>(accumulator) & kBigitMask;
395 accumulator >>= kBigitSize;
396 }
397 for (int i = used_digits_; i < product_length; ++i) {
398 int bigit_index1 = used_digits_ - 1;
399 int bigit_index2 = i - bigit_index1;
400 // Invariant: sum of both indices is again equal to i.
401 // Inner loop runs 0 times on last iteration, emptying accumulator.
402 while (bigit_index2 < used_digits_) {
403 Chunk chunk1 = bigits_[copy_offset + bigit_index1];
404 Chunk chunk2 = bigits_[copy_offset + bigit_index2];
405 accumulator += static_cast<DoubleChunk>(chunk1) * chunk2;
406 bigit_index1--;
407 bigit_index2++;
408 }
409 // The overwritten bigits_[i] will never be read in further loop iterations,
410 // because bigit_index1 and bigit_index2 are always greater
411 // than i - used_digits_.
412 bigits_[i] = static_cast<Chunk>(accumulator) & kBigitMask;
413 accumulator >>= kBigitSize;
414 }
415 // Since the result was guaranteed to lie inside the number the
416 // accumulator must be 0 now.
417 ASSERT(accumulator == 0);
418
419 // Don't forget to update the used_digits and the exponent.
420 used_digits_ = product_length;
421 exponent_ *= 2;
422 Clamp();
423}
424
425
426void Bignum::AssignPowerUInt16(uint16_t base, int power_exponent) {
427 ASSERT(base != 0);
428 ASSERT(power_exponent >= 0);
429 if (power_exponent == 0) {
430 AssignUInt16(1);
431 return;
432 }
433 Zero();
434 int shifts = 0;
435 // We expect base to be in range 2-32, and most often to be 10.
436 // It does not make much sense to implement different algorithms for counting
437 // the bits.
438 while ((base & 1) == 0) {
439 base >>= 1;
440 shifts++;
441 }
442 int bit_size = 0;
443 int tmp_base = base;
444 while (tmp_base != 0) {
445 tmp_base >>= 1;
446 bit_size++;
447 }
448 int final_size = bit_size * power_exponent;
449 // 1 extra bigit for the shifting, and one for rounded final_size.
450 EnsureCapacity(final_size / kBigitSize + 2);
451
452 // Left to Right exponentiation.
453 int mask = 1;
454 while (power_exponent >= mask) mask <<= 1;
455
456 // The mask is now pointing to the bit above the most significant 1-bit of
457 // power_exponent.
458 // Get rid of first 1-bit;
459 mask >>= 2;
460 uint64_t this_value = base;
461
3d1f044b 462 bool delayed_multiplication = false;
0f5d89e8
A
463 const uint64_t max_32bits = 0xFFFFFFFF;
464 while (mask != 0 && this_value <= max_32bits) {
465 this_value = this_value * this_value;
466 // Verify that there is enough space in this_value to perform the
467 // multiplication. The first bit_size bits must be 0.
468 if ((power_exponent & mask) != 0) {
3d1f044b 469 ASSERT(bit_size > 0);
0f5d89e8
A
470 uint64_t base_bits_mask =
471 ~((static_cast<uint64_t>(1) << (64 - bit_size)) - 1);
472 bool high_bits_zero = (this_value & base_bits_mask) == 0;
473 if (high_bits_zero) {
474 this_value *= base;
475 } else {
3d1f044b 476 delayed_multiplication = true;
0f5d89e8
A
477 }
478 }
479 mask >>= 1;
480 }
481 AssignUInt64(this_value);
3d1f044b 482 if (delayed_multiplication) {
0f5d89e8
A
483 MultiplyByUInt32(base);
484 }
485
486 // Now do the same thing as a bignum.
487 while (mask != 0) {
488 Square();
489 if ((power_exponent & mask) != 0) {
490 MultiplyByUInt32(base);
491 }
492 mask >>= 1;
493 }
494
495 // And finally add the saved shifts.
496 ShiftLeft(shifts * power_exponent);
497}
498
499
500// Precondition: this/other < 16bit.
501uint16_t Bignum::DivideModuloIntBignum(const Bignum& other) {
502 ASSERT(IsClamped());
503 ASSERT(other.IsClamped());
504 ASSERT(other.used_digits_ > 0);
505
506 // Easy case: if we have less digits than the divisor than the result is 0.
507 // Note: this handles the case where this == 0, too.
508 if (BigitLength() < other.BigitLength()) {
509 return 0;
510 }
511
512 Align(other);
513
514 uint16_t result = 0;
515
516 // Start by removing multiples of 'other' until both numbers have the same
517 // number of digits.
518 while (BigitLength() > other.BigitLength()) {
519 // This naive approach is extremely inefficient if `this` divided by other
520 // is big. This function is implemented for doubleToString where
521 // the result should be small (less than 10).
522 ASSERT(other.bigits_[other.used_digits_ - 1] >= ((1 << kBigitSize) / 16));
523 ASSERT(bigits_[used_digits_ - 1] < 0x10000);
524 // Remove the multiples of the first digit.
525 // Example this = 23 and other equals 9. -> Remove 2 multiples.
526 result += static_cast<uint16_t>(bigits_[used_digits_ - 1]);
527 SubtractTimes(other, bigits_[used_digits_ - 1]);
528 }
529
530 ASSERT(BigitLength() == other.BigitLength());
531
532 // Both bignums are at the same length now.
533 // Since other has more than 0 digits we know that the access to
534 // bigits_[used_digits_ - 1] is safe.
535 Chunk this_bigit = bigits_[used_digits_ - 1];
536 Chunk other_bigit = other.bigits_[other.used_digits_ - 1];
537
538 if (other.used_digits_ == 1) {
539 // Shortcut for easy (and common) case.
540 int quotient = this_bigit / other_bigit;
541 bigits_[used_digits_ - 1] = this_bigit - other_bigit * quotient;
542 ASSERT(quotient < 0x10000);
543 result += static_cast<uint16_t>(quotient);
544 Clamp();
545 return result;
546 }
547
548 int division_estimate = this_bigit / (other_bigit + 1);
549 ASSERT(division_estimate < 0x10000);
550 result += static_cast<uint16_t>(division_estimate);
551 SubtractTimes(other, division_estimate);
552
553 if (other_bigit * (division_estimate + 1) > this_bigit) {
554 // No need to even try to subtract. Even if other's remaining digits were 0
555 // another subtraction would be too much.
556 return result;
557 }
558
559 while (LessEqual(other, *this)) {
560 SubtractBignum(other);
561 result++;
562 }
563 return result;
564}
565
566
567template<typename S>
568static int SizeInHexChars(S number) {
569 ASSERT(number > 0);
570 int result = 0;
571 while (number != 0) {
572 number >>= 4;
573 result++;
574 }
575 return result;
576}
577
578
579static char HexCharOfValue(int value) {
580 ASSERT(0 <= value && value <= 16);
581 if (value < 10) return static_cast<char>(value + '0');
582 return static_cast<char>(value - 10 + 'A');
583}
584
585
586bool Bignum::ToHexString(char* buffer, int buffer_size) const {
587 ASSERT(IsClamped());
588 // Each bigit must be printable as separate hex-character.
589 ASSERT(kBigitSize % 4 == 0);
590 const int kHexCharsPerBigit = kBigitSize / 4;
591
592 if (used_digits_ == 0) {
593 if (buffer_size < 2) return false;
594 buffer[0] = '0';
595 buffer[1] = '\0';
596 return true;
597 }
598 // We add 1 for the terminating '\0' character.
599 int needed_chars = (BigitLength() - 1) * kHexCharsPerBigit +
600 SizeInHexChars(bigits_[used_digits_ - 1]) + 1;
601 if (needed_chars > buffer_size) return false;
602 int string_index = needed_chars - 1;
603 buffer[string_index--] = '\0';
604 for (int i = 0; i < exponent_; ++i) {
605 for (int j = 0; j < kHexCharsPerBigit; ++j) {
606 buffer[string_index--] = '0';
607 }
608 }
609 for (int i = 0; i < used_digits_ - 1; ++i) {
610 Chunk current_bigit = bigits_[i];
611 for (int j = 0; j < kHexCharsPerBigit; ++j) {
612 buffer[string_index--] = HexCharOfValue(current_bigit & 0xF);
613 current_bigit >>= 4;
614 }
615 }
616 // And finally the last bigit.
617 Chunk most_significant_bigit = bigits_[used_digits_ - 1];
618 while (most_significant_bigit != 0) {
619 buffer[string_index--] = HexCharOfValue(most_significant_bigit & 0xF);
620 most_significant_bigit >>= 4;
621 }
622 return true;
623}
624
625
626Bignum::Chunk Bignum::BigitAt(int index) const {
627 if (index >= BigitLength()) return 0;
628 if (index < exponent_) return 0;
629 return bigits_[index - exponent_];
630}
631
632
633int Bignum::Compare(const Bignum& a, const Bignum& b) {
634 ASSERT(a.IsClamped());
635 ASSERT(b.IsClamped());
636 int bigit_length_a = a.BigitLength();
637 int bigit_length_b = b.BigitLength();
638 if (bigit_length_a < bigit_length_b) return -1;
639 if (bigit_length_a > bigit_length_b) return +1;
640 for (int i = bigit_length_a - 1; i >= Min(a.exponent_, b.exponent_); --i) {
641 Chunk bigit_a = a.BigitAt(i);
642 Chunk bigit_b = b.BigitAt(i);
643 if (bigit_a < bigit_b) return -1;
644 if (bigit_a > bigit_b) return +1;
645 // Otherwise they are equal up to this digit. Try the next digit.
646 }
647 return 0;
648}
649
650
651int Bignum::PlusCompare(const Bignum& a, const Bignum& b, const Bignum& c) {
652 ASSERT(a.IsClamped());
653 ASSERT(b.IsClamped());
654 ASSERT(c.IsClamped());
655 if (a.BigitLength() < b.BigitLength()) {
656 return PlusCompare(b, a, c);
657 }
658 if (a.BigitLength() + 1 < c.BigitLength()) return -1;
659 if (a.BigitLength() > c.BigitLength()) return +1;
660 // The exponent encodes 0-bigits. So if there are more 0-digits in 'a' than
661 // 'b' has digits, then the bigit-length of 'a'+'b' must be equal to the one
662 // of 'a'.
663 if (a.exponent_ >= b.BigitLength() && a.BigitLength() < c.BigitLength()) {
664 return -1;
665 }
666
667 Chunk borrow = 0;
668 // Starting at min_exponent all digits are == 0. So no need to compare them.
669 int min_exponent = Min(Min(a.exponent_, b.exponent_), c.exponent_);
670 for (int i = c.BigitLength() - 1; i >= min_exponent; --i) {
671 Chunk chunk_a = a.BigitAt(i);
672 Chunk chunk_b = b.BigitAt(i);
673 Chunk chunk_c = c.BigitAt(i);
674 Chunk sum = chunk_a + chunk_b;
675 if (sum > chunk_c + borrow) {
676 return +1;
677 } else {
678 borrow = chunk_c + borrow - sum;
679 if (borrow > 1) return -1;
680 borrow <<= kBigitSize;
681 }
682 }
683 if (borrow == 0) return 0;
684 return -1;
685}
686
687
688void Bignum::Clamp() {
689 while (used_digits_ > 0 && bigits_[used_digits_ - 1] == 0) {
690 used_digits_--;
691 }
692 if (used_digits_ == 0) {
693 // Zero.
694 exponent_ = 0;
695 }
696}
697
698
699bool Bignum::IsClamped() const {
700 return used_digits_ == 0 || bigits_[used_digits_ - 1] != 0;
701}
702
703
704void Bignum::Zero() {
705 for (int i = 0; i < used_digits_; ++i) {
706 bigits_[i] = 0;
707 }
708 used_digits_ = 0;
709 exponent_ = 0;
710}
711
712
713void Bignum::Align(const Bignum& other) {
714 if (exponent_ > other.exponent_) {
715 // If "X" represents a "hidden" digit (by the exponent) then we are in the
716 // following case (a == this, b == other):
717 // a: aaaaaaXXXX or a: aaaaaXXX
718 // b: bbbbbbX b: bbbbbbbbXX
719 // We replace some of the hidden digits (X) of a with 0 digits.
720 // a: aaaaaa000X or a: aaaaa0XX
721 int zero_digits = exponent_ - other.exponent_;
722 EnsureCapacity(used_digits_ + zero_digits);
723 for (int i = used_digits_ - 1; i >= 0; --i) {
724 bigits_[i + zero_digits] = bigits_[i];
725 }
726 for (int i = 0; i < zero_digits; ++i) {
727 bigits_[i] = 0;
728 }
729 used_digits_ += zero_digits;
730 exponent_ -= zero_digits;
731 ASSERT(used_digits_ >= 0);
732 ASSERT(exponent_ >= 0);
733 }
734}
735
736
737void Bignum::BigitsShiftLeft(int shift_amount) {
738 ASSERT(shift_amount < kBigitSize);
739 ASSERT(shift_amount >= 0);
740 Chunk carry = 0;
741 for (int i = 0; i < used_digits_; ++i) {
742 Chunk new_carry = bigits_[i] >> (kBigitSize - shift_amount);
743 bigits_[i] = ((bigits_[i] << shift_amount) + carry) & kBigitMask;
744 carry = new_carry;
745 }
746 if (carry != 0) {
747 bigits_[used_digits_] = carry;
748 used_digits_++;
749 }
750}
751
752
753void Bignum::SubtractTimes(const Bignum& other, int factor) {
754 ASSERT(exponent_ <= other.exponent_);
755 if (factor < 3) {
756 for (int i = 0; i < factor; ++i) {
757 SubtractBignum(other);
758 }
759 return;
760 }
761 Chunk borrow = 0;
762 int exponent_diff = other.exponent_ - exponent_;
763 for (int i = 0; i < other.used_digits_; ++i) {
764 DoubleChunk product = static_cast<DoubleChunk>(factor) * other.bigits_[i];
765 DoubleChunk remove = borrow + product;
766 Chunk difference = bigits_[i + exponent_diff] - (remove & kBigitMask);
767 bigits_[i + exponent_diff] = difference & kBigitMask;
768 borrow = static_cast<Chunk>((difference >> (kChunkSize - 1)) +
769 (remove >> kBigitSize));
770 }
771 for (int i = other.used_digits_ + exponent_diff; i < used_digits_; ++i) {
772 if (borrow == 0) return;
773 Chunk difference = bigits_[i] - borrow;
774 bigits_[i] = difference & kBigitMask;
775 borrow = difference >> (kChunkSize - 1);
776 }
777 Clamp();
778}
779
780
781} // namespace double_conversion
782
783// ICU PATCH: Close ICU namespace
784U_NAMESPACE_END
785#endif // ICU PATCH: close #if !UCONFIG_NO_FORMATTING