v8  3.11.10(node0.8.26)
V8 is Google's open source JavaScript engine
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Friends Macros Pages
bignum-dtoa.cc
Go to the documentation of this file.
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 <math.h>
29 
30 #include "../include/v8stdint.h"
31 #include "checks.h"
32 #include "utils.h"
33 
34 #include "bignum-dtoa.h"
35 
36 #include "bignum.h"
37 #include "double.h"
38 
39 namespace v8 {
40 namespace internal {
41 
42 static int NormalizedExponent(uint64_t significand, int exponent) {
43  ASSERT(significand != 0);
44  while ((significand & Double::kHiddenBit) == 0) {
45  significand = significand << 1;
46  exponent = exponent - 1;
47  }
48  return exponent;
49 }
50 
51 
52 // Forward declarations:
53 // Returns an estimation of k such that 10^(k-1) <= v < 10^k.
54 static int EstimatePower(int exponent);
55 // Computes v / 10^estimated_power exactly, as a ratio of two bignums, numerator
56 // and denominator.
57 static void InitialScaledStartValues(double v,
58  int estimated_power,
59  bool need_boundary_deltas,
60  Bignum* numerator,
61  Bignum* denominator,
62  Bignum* delta_minus,
63  Bignum* delta_plus);
64 // Multiplies numerator/denominator so that its values lies in the range 1-10.
65 // Returns decimal_point s.t.
66 // v = numerator'/denominator' * 10^(decimal_point-1)
67 // where numerator' and denominator' are the values of numerator and
68 // denominator after the call to this function.
69 static void FixupMultiply10(int estimated_power, bool is_even,
70  int* decimal_point,
71  Bignum* numerator, Bignum* denominator,
72  Bignum* delta_minus, Bignum* delta_plus);
73 // Generates digits from the left to the right and stops when the generated
74 // digits yield the shortest decimal representation of v.
75 static void GenerateShortestDigits(Bignum* numerator, Bignum* denominator,
76  Bignum* delta_minus, Bignum* delta_plus,
77  bool is_even,
78  Vector<char> buffer, int* length);
79 // Generates 'requested_digits' after the decimal point.
80 static void BignumToFixed(int requested_digits, int* decimal_point,
81  Bignum* numerator, Bignum* denominator,
82  Vector<char>(buffer), int* length);
83 // Generates 'count' digits of numerator/denominator.
84 // Once 'count' digits have been produced rounds the result depending on the
85 // remainder (remainders of exactly .5 round upwards). Might update the
86 // decimal_point when rounding up (for example for 0.9999).
87 static void GenerateCountedDigits(int count, int* decimal_point,
88  Bignum* numerator, Bignum* denominator,
89  Vector<char>(buffer), int* length);
90 
91 
92 void BignumDtoa(double v, BignumDtoaMode mode, int requested_digits,
93  Vector<char> buffer, int* length, int* decimal_point) {
94  ASSERT(v > 0);
95  ASSERT(!Double(v).IsSpecial());
96  uint64_t significand = Double(v).Significand();
97  bool is_even = (significand & 1) == 0;
98  int exponent = Double(v).Exponent();
99  int normalized_exponent = NormalizedExponent(significand, exponent);
100  // estimated_power might be too low by 1.
101  int estimated_power = EstimatePower(normalized_exponent);
102 
103  // Shortcut for Fixed.
104  // The requested digits correspond to the digits after the point. If the
105  // number is much too small, then there is no need in trying to get any
106  // digits.
107  if (mode == BIGNUM_DTOA_FIXED && -estimated_power - 1 > requested_digits) {
108  buffer[0] = '\0';
109  *length = 0;
110  // Set decimal-point to -requested_digits. This is what Gay does.
111  // Note that it should not have any effect anyways since the string is
112  // empty.
113  *decimal_point = -requested_digits;
114  return;
115  }
116 
117  Bignum numerator;
118  Bignum denominator;
119  Bignum delta_minus;
120  Bignum delta_plus;
121  // Make sure the bignum can grow large enough. The smallest double equals
122  // 4e-324. In this case the denominator needs fewer than 324*4 binary digits.
123  // The maximum double is 1.7976931348623157e308 which needs fewer than
124  // 308*4 binary digits.
126  bool need_boundary_deltas = (mode == BIGNUM_DTOA_SHORTEST);
127  InitialScaledStartValues(v, estimated_power, need_boundary_deltas,
128  &numerator, &denominator,
129  &delta_minus, &delta_plus);
130  // We now have v = (numerator / denominator) * 10^estimated_power.
131  FixupMultiply10(estimated_power, is_even, decimal_point,
132  &numerator, &denominator,
133  &delta_minus, &delta_plus);
134  // We now have v = (numerator / denominator) * 10^(decimal_point-1), and
135  // 1 <= (numerator + delta_plus) / denominator < 10
136  switch (mode) {
138  GenerateShortestDigits(&numerator, &denominator,
139  &delta_minus, &delta_plus,
140  is_even, buffer, length);
141  break;
142  case BIGNUM_DTOA_FIXED:
143  BignumToFixed(requested_digits, decimal_point,
144  &numerator, &denominator,
145  buffer, length);
146  break;
148  GenerateCountedDigits(requested_digits, decimal_point,
149  &numerator, &denominator,
150  buffer, length);
151  break;
152  default:
153  UNREACHABLE();
154  }
155  buffer[*length] = '\0';
156 }
157 
158 
159 // The procedure starts generating digits from the left to the right and stops
160 // when the generated digits yield the shortest decimal representation of v. A
161 // decimal representation of v is a number lying closer to v than to any other
162 // double, so it converts to v when read.
163 //
164 // This is true if d, the decimal representation, is between m- and m+, the
165 // upper and lower boundaries. d must be strictly between them if !is_even.
166 // m- := (numerator - delta_minus) / denominator
167 // m+ := (numerator + delta_plus) / denominator
168 //
169 // Precondition: 0 <= (numerator+delta_plus) / denominator < 10.
170 // If 1 <= (numerator+delta_plus) / denominator < 10 then no leading 0 digit
171 // will be produced. This should be the standard precondition.
172 static void GenerateShortestDigits(Bignum* numerator, Bignum* denominator,
173  Bignum* delta_minus, Bignum* delta_plus,
174  bool is_even,
175  Vector<char> buffer, int* length) {
176  // Small optimization: if delta_minus and delta_plus are the same just reuse
177  // one of the two bignums.
178  if (Bignum::Equal(*delta_minus, *delta_plus)) {
179  delta_plus = delta_minus;
180  }
181  *length = 0;
182  while (true) {
183  uint16_t digit;
184  digit = numerator->DivideModuloIntBignum(*denominator);
185  ASSERT(digit <= 9); // digit is a uint16_t and therefore always positive.
186  // digit = numerator / denominator (integer division).
187  // numerator = numerator % denominator.
188  buffer[(*length)++] = digit + '0';
189 
190  // Can we stop already?
191  // If the remainder of the division is less than the distance to the lower
192  // boundary we can stop. In this case we simply round down (discarding the
193  // remainder).
194  // Similarly we test if we can round up (using the upper boundary).
195  bool in_delta_room_minus;
196  bool in_delta_room_plus;
197  if (is_even) {
198  in_delta_room_minus = Bignum::LessEqual(*numerator, *delta_minus);
199  } else {
200  in_delta_room_minus = Bignum::Less(*numerator, *delta_minus);
201  }
202  if (is_even) {
203  in_delta_room_plus =
204  Bignum::PlusCompare(*numerator, *delta_plus, *denominator) >= 0;
205  } else {
206  in_delta_room_plus =
207  Bignum::PlusCompare(*numerator, *delta_plus, *denominator) > 0;
208  }
209  if (!in_delta_room_minus && !in_delta_room_plus) {
210  // Prepare for next iteration.
211  numerator->Times10();
212  delta_minus->Times10();
213  // We optimized delta_plus to be equal to delta_minus (if they share the
214  // same value). So don't multiply delta_plus if they point to the same
215  // object.
216  if (delta_minus != delta_plus) {
217  delta_plus->Times10();
218  }
219  } else if (in_delta_room_minus && in_delta_room_plus) {
220  // Let's see if 2*numerator < denominator.
221  // If yes, then the next digit would be < 5 and we can round down.
222  int compare = Bignum::PlusCompare(*numerator, *numerator, *denominator);
223  if (compare < 0) {
224  // Remaining digits are less than .5. -> Round down (== do nothing).
225  } else if (compare > 0) {
226  // Remaining digits are more than .5 of denominator. -> Round up.
227  // Note that the last digit could not be a '9' as otherwise the whole
228  // loop would have stopped earlier.
229  // We still have an assert here in case the preconditions were not
230  // satisfied.
231  ASSERT(buffer[(*length) - 1] != '9');
232  buffer[(*length) - 1]++;
233  } else {
234  // Halfway case.
235  // TODO(floitsch): need a way to solve half-way cases.
236  // For now let's round towards even (since this is what Gay seems to
237  // do).
238 
239  if ((buffer[(*length) - 1] - '0') % 2 == 0) {
240  // Round down => Do nothing.
241  } else {
242  ASSERT(buffer[(*length) - 1] != '9');
243  buffer[(*length) - 1]++;
244  }
245  }
246  return;
247  } else if (in_delta_room_minus) {
248  // Round down (== do nothing).
249  return;
250  } else { // in_delta_room_plus
251  // Round up.
252  // Note again that the last digit could not be '9' since this would have
253  // stopped the loop earlier.
254  // We still have an ASSERT here, in case the preconditions were not
255  // satisfied.
256  ASSERT(buffer[(*length) -1] != '9');
257  buffer[(*length) - 1]++;
258  return;
259  }
260  }
261 }
262 
263 
264 // Let v = numerator / denominator < 10.
265 // Then we generate 'count' digits of d = x.xxxxx... (without the decimal point)
266 // from left to right. Once 'count' digits have been produced we decide wether
267 // to round up or down. Remainders of exactly .5 round upwards. Numbers such
268 // as 9.999999 propagate a carry all the way, and change the
269 // exponent (decimal_point), when rounding upwards.
270 static void GenerateCountedDigits(int count, int* decimal_point,
271  Bignum* numerator, Bignum* denominator,
272  Vector<char>(buffer), int* length) {
273  ASSERT(count >= 0);
274  for (int i = 0; i < count - 1; ++i) {
275  uint16_t digit;
276  digit = numerator->DivideModuloIntBignum(*denominator);
277  ASSERT(digit <= 9); // digit is a uint16_t and therefore always positive.
278  // digit = numerator / denominator (integer division).
279  // numerator = numerator % denominator.
280  buffer[i] = digit + '0';
281  // Prepare for next iteration.
282  numerator->Times10();
283  }
284  // Generate the last digit.
285  uint16_t digit;
286  digit = numerator->DivideModuloIntBignum(*denominator);
287  if (Bignum::PlusCompare(*numerator, *numerator, *denominator) >= 0) {
288  digit++;
289  }
290  buffer[count - 1] = digit + '0';
291  // Correct bad digits (in case we had a sequence of '9's). Propagate the
292  // carry until we hat a non-'9' or til we reach the first digit.
293  for (int i = count - 1; i > 0; --i) {
294  if (buffer[i] != '0' + 10) break;
295  buffer[i] = '0';
296  buffer[i - 1]++;
297  }
298  if (buffer[0] == '0' + 10) {
299  // Propagate a carry past the top place.
300  buffer[0] = '1';
301  (*decimal_point)++;
302  }
303  *length = count;
304 }
305 
306 
307 // Generates 'requested_digits' after the decimal point. It might omit
308 // trailing '0's. If the input number is too small then no digits at all are
309 // generated (ex.: 2 fixed digits for 0.00001).
310 //
311 // Input verifies: 1 <= (numerator + delta) / denominator < 10.
312 static void BignumToFixed(int requested_digits, int* decimal_point,
313  Bignum* numerator, Bignum* denominator,
314  Vector<char>(buffer), int* length) {
315  // Note that we have to look at more than just the requested_digits, since
316  // a number could be rounded up. Example: v=0.5 with requested_digits=0.
317  // Even though the power of v equals 0 we can't just stop here.
318  if (-(*decimal_point) > requested_digits) {
319  // The number is definitively too small.
320  // Ex: 0.001 with requested_digits == 1.
321  // Set decimal-point to -requested_digits. This is what Gay does.
322  // Note that it should not have any effect anyways since the string is
323  // empty.
324  *decimal_point = -requested_digits;
325  *length = 0;
326  return;
327  } else if (-(*decimal_point) == requested_digits) {
328  // We only need to verify if the number rounds down or up.
329  // Ex: 0.04 and 0.06 with requested_digits == 1.
330  ASSERT(*decimal_point == -requested_digits);
331  // Initially the fraction lies in range (1, 10]. Multiply the denominator
332  // by 10 so that we can compare more easily.
333  denominator->Times10();
334  if (Bignum::PlusCompare(*numerator, *numerator, *denominator) >= 0) {
335  // If the fraction is >= 0.5 then we have to include the rounded
336  // digit.
337  buffer[0] = '1';
338  *length = 1;
339  (*decimal_point)++;
340  } else {
341  // Note that we caught most of similar cases earlier.
342  *length = 0;
343  }
344  return;
345  } else {
346  // The requested digits correspond to the digits after the point.
347  // The variable 'needed_digits' includes the digits before the point.
348  int needed_digits = (*decimal_point) + requested_digits;
349  GenerateCountedDigits(needed_digits, decimal_point,
350  numerator, denominator,
351  buffer, length);
352  }
353 }
354 
355 
356 // Returns an estimation of k such that 10^(k-1) <= v < 10^k where
357 // v = f * 2^exponent and 2^52 <= f < 2^53.
358 // v is hence a normalized double with the given exponent. The output is an
359 // approximation for the exponent of the decimal approimation .digits * 10^k.
360 //
361 // The result might undershoot by 1 in which case 10^k <= v < 10^k+1.
362 // Note: this property holds for v's upper boundary m+ too.
363 // 10^k <= m+ < 10^k+1.
364 // (see explanation below).
365 //
366 // Examples:
367 // EstimatePower(0) => 16
368 // EstimatePower(-52) => 0
369 //
370 // Note: e >= 0 => EstimatedPower(e) > 0. No similar claim can be made for e<0.
371 static int EstimatePower(int exponent) {
372  // This function estimates log10 of v where v = f*2^e (with e == exponent).
373  // Note that 10^floor(log10(v)) <= v, but v <= 10^ceil(log10(v)).
374  // Note that f is bounded by its container size. Let p = 53 (the double's
375  // significand size). Then 2^(p-1) <= f < 2^p.
376  //
377  // Given that log10(v) == log2(v)/log2(10) and e+(len(f)-1) is quite close
378  // to log2(v) the function is simplified to (e+(len(f)-1)/log2(10)).
379  // The computed number undershoots by less than 0.631 (when we compute log3
380  // and not log10).
381  //
382  // Optimization: since we only need an approximated result this computation
383  // can be performed on 64 bit integers. On x86/x64 architecture the speedup is
384  // not really measurable, though.
385  //
386  // Since we want to avoid overshooting we decrement by 1e10 so that
387  // floating-point imprecisions don't affect us.
388  //
389  // Explanation for v's boundary m+: the computation takes advantage of
390  // the fact that 2^(p-1) <= f < 2^p. Boundaries still satisfy this requirement
391  // (even for denormals where the delta can be much more important).
392 
393  const double k1Log10 = 0.30102999566398114; // 1/lg(10)
394 
395  // For doubles len(f) == 53 (don't forget the hidden bit).
396  const int kSignificandSize = 53;
397  double estimate = ceil((exponent + kSignificandSize - 1) * k1Log10 - 1e-10);
398  return static_cast<int>(estimate);
399 }
400 
401 
402 // See comments for InitialScaledStartValues.
403 static void InitialScaledStartValuesPositiveExponent(
404  double v, int estimated_power, bool need_boundary_deltas,
405  Bignum* numerator, Bignum* denominator,
406  Bignum* delta_minus, Bignum* delta_plus) {
407  // A positive exponent implies a positive power.
408  ASSERT(estimated_power >= 0);
409  // Since the estimated_power is positive we simply multiply the denominator
410  // by 10^estimated_power.
411 
412  // numerator = v.
413  numerator->AssignUInt64(Double(v).Significand());
414  numerator->ShiftLeft(Double(v).Exponent());
415  // denominator = 10^estimated_power.
416  denominator->AssignPowerUInt16(10, estimated_power);
417 
418  if (need_boundary_deltas) {
419  // Introduce a common denominator so that the deltas to the boundaries are
420  // integers.
421  denominator->ShiftLeft(1);
422  numerator->ShiftLeft(1);
423  // Let v = f * 2^e, then m+ - v = 1/2 * 2^e; With the common
424  // denominator (of 2) delta_plus equals 2^e.
425  delta_plus->AssignUInt16(1);
426  delta_plus->ShiftLeft(Double(v).Exponent());
427  // Same for delta_minus (with adjustments below if f == 2^p-1).
428  delta_minus->AssignUInt16(1);
429  delta_minus->ShiftLeft(Double(v).Exponent());
430 
431  // If the significand (without the hidden bit) is 0, then the lower
432  // boundary is closer than just half a ulp (unit in the last place).
433  // There is only one exception: if the next lower number is a denormal then
434  // the distance is 1 ulp. This cannot be the case for exponent >= 0 (but we
435  // have to test it in the other function where exponent < 0).
436  uint64_t v_bits = Double(v).AsUint64();
437  if ((v_bits & Double::kSignificandMask) == 0) {
438  // The lower boundary is closer at half the distance of "normal" numbers.
439  // Increase the common denominator and adapt all but the delta_minus.
440  denominator->ShiftLeft(1); // *2
441  numerator->ShiftLeft(1); // *2
442  delta_plus->ShiftLeft(1); // *2
443  }
444  }
445 }
446 
447 
448 // See comments for InitialScaledStartValues
449 static void InitialScaledStartValuesNegativeExponentPositivePower(
450  double v, int estimated_power, bool need_boundary_deltas,
451  Bignum* numerator, Bignum* denominator,
452  Bignum* delta_minus, Bignum* delta_plus) {
453  uint64_t significand = Double(v).Significand();
454  int exponent = Double(v).Exponent();
455  // v = f * 2^e with e < 0, and with estimated_power >= 0.
456  // This means that e is close to 0 (have a look at how estimated_power is
457  // computed).
458 
459  // numerator = significand
460  // since v = significand * 2^exponent this is equivalent to
461  // numerator = v * / 2^-exponent
462  numerator->AssignUInt64(significand);
463  // denominator = 10^estimated_power * 2^-exponent (with exponent < 0)
464  denominator->AssignPowerUInt16(10, estimated_power);
465  denominator->ShiftLeft(-exponent);
466 
467  if (need_boundary_deltas) {
468  // Introduce a common denominator so that the deltas to the boundaries are
469  // integers.
470  denominator->ShiftLeft(1);
471  numerator->ShiftLeft(1);
472  // Let v = f * 2^e, then m+ - v = 1/2 * 2^e; With the common
473  // denominator (of 2) delta_plus equals 2^e.
474  // Given that the denominator already includes v's exponent the distance
475  // to the boundaries is simply 1.
476  delta_plus->AssignUInt16(1);
477  // Same for delta_minus (with adjustments below if f == 2^p-1).
478  delta_minus->AssignUInt16(1);
479 
480  // If the significand (without the hidden bit) is 0, then the lower
481  // boundary is closer than just one ulp (unit in the last place).
482  // There is only one exception: if the next lower number is a denormal
483  // then the distance is 1 ulp. Since the exponent is close to zero
484  // (otherwise estimated_power would have been negative) this cannot happen
485  // here either.
486  uint64_t v_bits = Double(v).AsUint64();
487  if ((v_bits & Double::kSignificandMask) == 0) {
488  // The lower boundary is closer at half the distance of "normal" numbers.
489  // Increase the denominator and adapt all but the delta_minus.
490  denominator->ShiftLeft(1); // *2
491  numerator->ShiftLeft(1); // *2
492  delta_plus->ShiftLeft(1); // *2
493  }
494  }
495 }
496 
497 
498 // See comments for InitialScaledStartValues
499 static void InitialScaledStartValuesNegativeExponentNegativePower(
500  double v, int estimated_power, bool need_boundary_deltas,
501  Bignum* numerator, Bignum* denominator,
502  Bignum* delta_minus, Bignum* delta_plus) {
503  const uint64_t kMinimalNormalizedExponent =
504  V8_2PART_UINT64_C(0x00100000, 00000000);
505  uint64_t significand = Double(v).Significand();
506  int exponent = Double(v).Exponent();
507  // Instead of multiplying the denominator with 10^estimated_power we
508  // multiply all values (numerator and deltas) by 10^-estimated_power.
509 
510  // Use numerator as temporary container for power_ten.
511  Bignum* power_ten = numerator;
512  power_ten->AssignPowerUInt16(10, -estimated_power);
513 
514  if (need_boundary_deltas) {
515  // Since power_ten == numerator we must make a copy of 10^estimated_power
516  // before we complete the computation of the numerator.
517  // delta_plus = delta_minus = 10^estimated_power
518  delta_plus->AssignBignum(*power_ten);
519  delta_minus->AssignBignum(*power_ten);
520  }
521 
522  // numerator = significand * 2 * 10^-estimated_power
523  // since v = significand * 2^exponent this is equivalent to
524  // numerator = v * 10^-estimated_power * 2 * 2^-exponent.
525  // Remember: numerator has been abused as power_ten. So no need to assign it
526  // to itself.
527  ASSERT(numerator == power_ten);
528  numerator->MultiplyByUInt64(significand);
529 
530  // denominator = 2 * 2^-exponent with exponent < 0.
531  denominator->AssignUInt16(1);
532  denominator->ShiftLeft(-exponent);
533 
534  if (need_boundary_deltas) {
535  // Introduce a common denominator so that the deltas to the boundaries are
536  // integers.
537  numerator->ShiftLeft(1);
538  denominator->ShiftLeft(1);
539  // With this shift the boundaries have their correct value, since
540  // delta_plus = 10^-estimated_power, and
541  // delta_minus = 10^-estimated_power.
542  // These assignments have been done earlier.
543 
544  // The special case where the lower boundary is twice as close.
545  // This time we have to look out for the exception too.
546  uint64_t v_bits = Double(v).AsUint64();
547  if ((v_bits & Double::kSignificandMask) == 0 &&
548  // The only exception where a significand == 0 has its boundaries at
549  // "normal" distances:
550  (v_bits & Double::kExponentMask) != kMinimalNormalizedExponent) {
551  numerator->ShiftLeft(1); // *2
552  denominator->ShiftLeft(1); // *2
553  delta_plus->ShiftLeft(1); // *2
554  }
555  }
556 }
557 
558 
559 // Let v = significand * 2^exponent.
560 // Computes v / 10^estimated_power exactly, as a ratio of two bignums, numerator
561 // and denominator. The functions GenerateShortestDigits and
562 // GenerateCountedDigits will then convert this ratio to its decimal
563 // representation d, with the required accuracy.
564 // Then d * 10^estimated_power is the representation of v.
565 // (Note: the fraction and the estimated_power might get adjusted before
566 // generating the decimal representation.)
567 //
568 // The initial start values consist of:
569 // - a scaled numerator: s.t. numerator/denominator == v / 10^estimated_power.
570 // - a scaled (common) denominator.
571 // optionally (used by GenerateShortestDigits to decide if it has the shortest
572 // decimal converting back to v):
573 // - v - m-: the distance to the lower boundary.
574 // - m+ - v: the distance to the upper boundary.
575 //
576 // v, m+, m-, and therefore v - m- and m+ - v all share the same denominator.
577 //
578 // Let ep == estimated_power, then the returned values will satisfy:
579 // v / 10^ep = numerator / denominator.
580 // v's boundarys m- and m+:
581 // m- / 10^ep == v / 10^ep - delta_minus / denominator
582 // m+ / 10^ep == v / 10^ep + delta_plus / denominator
583 // Or in other words:
584 // m- == v - delta_minus * 10^ep / denominator;
585 // m+ == v + delta_plus * 10^ep / denominator;
586 //
587 // Since 10^(k-1) <= v < 10^k (with k == estimated_power)
588 // or 10^k <= v < 10^(k+1)
589 // we then have 0.1 <= numerator/denominator < 1
590 // or 1 <= numerator/denominator < 10
591 //
592 // It is then easy to kickstart the digit-generation routine.
593 //
594 // The boundary-deltas are only filled if need_boundary_deltas is set.
595 static void InitialScaledStartValues(double v,
596  int estimated_power,
597  bool need_boundary_deltas,
598  Bignum* numerator,
599  Bignum* denominator,
600  Bignum* delta_minus,
601  Bignum* delta_plus) {
602  if (Double(v).Exponent() >= 0) {
603  InitialScaledStartValuesPositiveExponent(
604  v, estimated_power, need_boundary_deltas,
605  numerator, denominator, delta_minus, delta_plus);
606  } else if (estimated_power >= 0) {
607  InitialScaledStartValuesNegativeExponentPositivePower(
608  v, estimated_power, need_boundary_deltas,
609  numerator, denominator, delta_minus, delta_plus);
610  } else {
611  InitialScaledStartValuesNegativeExponentNegativePower(
612  v, estimated_power, need_boundary_deltas,
613  numerator, denominator, delta_minus, delta_plus);
614  }
615 }
616 
617 
618 // This routine multiplies numerator/denominator so that its values lies in the
619 // range 1-10. That is after a call to this function we have:
620 // 1 <= (numerator + delta_plus) /denominator < 10.
621 // Let numerator the input before modification and numerator' the argument
622 // after modification, then the output-parameter decimal_point is such that
623 // numerator / denominator * 10^estimated_power ==
624 // numerator' / denominator' * 10^(decimal_point - 1)
625 // In some cases estimated_power was too low, and this is already the case. We
626 // then simply adjust the power so that 10^(k-1) <= v < 10^k (with k ==
627 // estimated_power) but do not touch the numerator or denominator.
628 // Otherwise the routine multiplies the numerator and the deltas by 10.
629 static void FixupMultiply10(int estimated_power, bool is_even,
630  int* decimal_point,
631  Bignum* numerator, Bignum* denominator,
632  Bignum* delta_minus, Bignum* delta_plus) {
633  bool in_range;
634  if (is_even) {
635  // For IEEE doubles half-way cases (in decimal system numbers ending with 5)
636  // are rounded to the closest floating-point number with even significand.
637  in_range = Bignum::PlusCompare(*numerator, *delta_plus, *denominator) >= 0;
638  } else {
639  in_range = Bignum::PlusCompare(*numerator, *delta_plus, *denominator) > 0;
640  }
641  if (in_range) {
642  // Since numerator + delta_plus >= denominator we already have
643  // 1 <= numerator/denominator < 10. Simply update the estimated_power.
644  *decimal_point = estimated_power + 1;
645  } else {
646  *decimal_point = estimated_power;
647  numerator->Times10();
648  if (Bignum::Equal(*delta_minus, *delta_plus)) {
649  delta_minus->Times10();
650  delta_plus->AssignBignum(*delta_minus);
651  } else {
652  delta_minus->Times10();
653  delta_plus->Times10();
654  }
655  }
656 }
657 
658 } } // namespace v8::internal
uint64_t Significand() const
Definition: double.h:110
static const uint64_t kExponentMask
Definition: double.h:44
static const uint64_t kHiddenBit
Definition: double.h:47
int Exponent() const
Definition: double.h:101
#define ASSERT(condition)
Definition: checks.h:270
unsigned short uint16_t
Definition: unicode.cc:46
#define UNREACHABLE()
Definition: checks.h:50
static int PlusCompare(const Bignum &a, const Bignum &b, const Bignum &c)
Definition: bignum.cc:635
static bool Equal(const Bignum &a, const Bignum &b)
Definition: bignum.h:72
void BignumDtoa(double v, BignumDtoaMode mode, int requested_digits, Vector< char > buffer, int *length, int *decimal_point)
Definition: bignum-dtoa.cc:92
static bool Less(const Bignum &a, const Bignum &b)
Definition: bignum.h:78
#define V8_2PART_UINT64_C(a, b)
Definition: globals.h:202
static const uint64_t kSignificandMask
Definition: double.h:45
static bool LessEqual(const Bignum &a, const Bignum &b)
Definition: bignum.h:75
static const int kMaxSignificantBits
Definition: bignum.h:39