• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /****************************************************************
2  *
3  * The author of this software is David M. Gay.
4  *
5  * Copyright (c) 1991, 2000, 2001 by Lucent Technologies.
6  * Copyright (C) 2002, 2005, 2006, 2007, 2008, 2010, 2012 Apple Inc. All rights reserved.
7  *
8  * Permission to use, copy, modify, and distribute this software for any
9  * purpose without fee is hereby granted, provided that this entire notice
10  * is included in all copies of any software which is or includes a copy
11  * or modification of this software and in all copies of the supporting
12  * documentation for such software.
13  *
14  * THIS SOFTWARE IS BEING PROVIDED "AS IS", WITHOUT ANY EXPRESS OR IMPLIED
15  * WARRANTY.  IN PARTICULAR, NEITHER THE AUTHOR NOR LUCENT MAKES ANY
16  * REPRESENTATION OR WARRANTY OF ANY KIND CONCERNING THE MERCHANTABILITY
17  * OF THIS SOFTWARE OR ITS FITNESS FOR ANY PARTICULAR PURPOSE.
18  *
19  ***************************************************************/
20 
21 /* Please send bug reports to David M. Gay (dmg at acm dot org,
22  * with " at " changed at "@" and " dot " changed to ".").    */
23 
24 /* On a machine with IEEE extended-precision registers, it is
25  * necessary to specify double-precision (53-bit) rounding precision
26  * before invoking strtod or dtoa.  If the machine uses (the equivalent
27  * of) Intel 80x87 arithmetic, the call
28  *    _control87(PC_53, MCW_PC);
29  * does this with many compilers.  Whether this or another call is
30  * appropriate depends on the compiler; for this to work, it may be
31  * necessary to #include "float.h" or another system-dependent header
32  * file.
33  */
34 
35 #include "config.h"
36 #include "dtoa.h"
37 
38 #include "wtf/CPU.h"
39 #include "wtf/MathExtras.h"
40 #include "wtf/ThreadingPrimitives.h"
41 #include "wtf/Vector.h"
42 
43 #if COMPILER(MSVC)
44 #pragma warning(disable: 4244)
45 #pragma warning(disable: 4245)
46 #pragma warning(disable: 4554)
47 
48 #if _MSC_VER == 1800
49 // TODO(scottmg): VS2013 currently ICEs on a bunch of functions in this file.
50 // Upstream bug fixed in next release. See http://crbug.com/288498.
51 #pragma optimize("", off)
52 #endif
53 
54 #endif
55 
56 namespace WTF {
57 
58 Mutex* s_dtoaP5Mutex;
59 
60 typedef union {
61     double d;
62     uint32_t L[2];
63 } U;
64 
65 #if CPU(BIG_ENDIAN) || CPU(MIDDLE_ENDIAN)
66 #define word0(x) (x)->L[0]
67 #define word1(x) (x)->L[1]
68 #else
69 #define word0(x) (x)->L[1]
70 #define word1(x) (x)->L[0]
71 #endif
72 #define dval(x) (x)->d
73 
74 #define Exp_shift  20
75 #define Exp_shift1 20
76 #define Exp_msk1    0x100000
77 #define Exp_msk11   0x100000
78 #define Exp_mask  0x7ff00000
79 #define P 53
80 #define Bias 1023
81 #define Emin (-1022)
82 #define Exp_1  0x3ff00000
83 #define Exp_11 0x3ff00000
84 #define Ebits 11
85 #define Frac_mask  0xfffff
86 #define Frac_mask1 0xfffff
87 #define Ten_pmax 22
88 #define Bletch 0x10
89 #define Bndry_mask  0xfffff
90 #define Bndry_mask1 0xfffff
91 #define LSB 1
92 #define Sign_bit 0x80000000
93 #define Log2P 1
94 #define Tiny0 0
95 #define Tiny1 1
96 #define Quick_max 14
97 #define Int_max 14
98 
99 #define rounded_product(a, b) a *= b
100 #define rounded_quotient(a, b) a /= b
101 
102 #define Big0 (Frac_mask1 | Exp_msk1 * (DBL_MAX_EXP + Bias - 1))
103 #define Big1 0xffffffff
104 
105 #if CPU(X86_64)
106 // FIXME: should we enable this on all 64-bit CPUs?
107 // 64-bit emulation provided by the compiler is likely to be slower than dtoa own code on 32-bit hardware.
108 #define USE_LONG_LONG
109 #endif
110 
111 #ifndef USE_LONG_LONG
112 /* The following definition of Storeinc is appropriate for MIPS processors.
113  * An alternative that might be better on some machines is
114  *  *p++ = high << 16 | low & 0xffff;
115  */
storeInc(uint32_t * p,uint16_t high,uint16_t low)116 static ALWAYS_INLINE uint32_t* storeInc(uint32_t* p, uint16_t high, uint16_t low)
117 {
118     uint16_t* p16 = reinterpret_cast<uint16_t*>(p);
119 #if CPU(BIG_ENDIAN)
120     p16[0] = high;
121     p16[1] = low;
122 #else
123     p16[1] = high;
124     p16[0] = low;
125 #endif
126     return p + 1;
127 }
128 #endif
129 
130 struct BigInt {
BigIntWTF::BigInt131     BigInt() : sign(0) { }
132     int sign;
133 
clearWTF::BigInt134     void clear()
135     {
136         sign = 0;
137         m_words.clear();
138     }
139 
sizeWTF::BigInt140     size_t size() const
141     {
142         return m_words.size();
143     }
144 
resizeWTF::BigInt145     void resize(size_t s)
146     {
147         m_words.resize(s);
148     }
149 
wordsWTF::BigInt150     uint32_t* words()
151     {
152         return m_words.data();
153     }
154 
wordsWTF::BigInt155     const uint32_t* words() const
156     {
157         return m_words.data();
158     }
159 
appendWTF::BigInt160     void append(uint32_t w)
161     {
162         m_words.append(w);
163     }
164 
165     Vector<uint32_t, 16> m_words;
166 };
167 
multadd(BigInt & b,int m,int a)168 static void multadd(BigInt& b, int m, int a)    /* multiply by m and add a */
169 {
170 #ifdef USE_LONG_LONG
171     unsigned long long carry;
172 #else
173     uint32_t carry;
174 #endif
175 
176     int wds = b.size();
177     uint32_t* x = b.words();
178     int i = 0;
179     carry = a;
180     do {
181 #ifdef USE_LONG_LONG
182         unsigned long long y = *x * (unsigned long long)m + carry;
183         carry = y >> 32;
184         *x++ = (uint32_t)y & 0xffffffffUL;
185 #else
186         uint32_t xi = *x;
187         uint32_t y = (xi & 0xffff) * m + carry;
188         uint32_t z = (xi >> 16) * m + (y >> 16);
189         carry = z >> 16;
190         *x++ = (z << 16) + (y & 0xffff);
191 #endif
192     } while (++i < wds);
193 
194     if (carry)
195         b.append((uint32_t)carry);
196 }
197 
hi0bits(uint32_t x)198 static int hi0bits(uint32_t x)
199 {
200     int k = 0;
201 
202     if (!(x & 0xffff0000)) {
203         k = 16;
204         x <<= 16;
205     }
206     if (!(x & 0xff000000)) {
207         k += 8;
208         x <<= 8;
209     }
210     if (!(x & 0xf0000000)) {
211         k += 4;
212         x <<= 4;
213     }
214     if (!(x & 0xc0000000)) {
215         k += 2;
216         x <<= 2;
217     }
218     if (!(x & 0x80000000)) {
219         k++;
220         if (!(x & 0x40000000))
221             return 32;
222     }
223     return k;
224 }
225 
lo0bits(uint32_t * y)226 static int lo0bits(uint32_t* y)
227 {
228     int k;
229     uint32_t x = *y;
230 
231     if (x & 7) {
232         if (x & 1)
233             return 0;
234         if (x & 2) {
235             *y = x >> 1;
236             return 1;
237         }
238         *y = x >> 2;
239         return 2;
240     }
241     k = 0;
242     if (!(x & 0xffff)) {
243         k = 16;
244         x >>= 16;
245     }
246     if (!(x & 0xff)) {
247         k += 8;
248         x >>= 8;
249     }
250     if (!(x & 0xf)) {
251         k += 4;
252         x >>= 4;
253     }
254     if (!(x & 0x3)) {
255         k += 2;
256         x >>= 2;
257     }
258     if (!(x & 1)) {
259         k++;
260         x >>= 1;
261         if (!x)
262             return 32;
263     }
264     *y = x;
265     return k;
266 }
267 
i2b(BigInt & b,int i)268 static void i2b(BigInt& b, int i)
269 {
270     b.sign = 0;
271     b.resize(1);
272     b.words()[0] = i;
273 }
274 
mult(BigInt & aRef,const BigInt & bRef)275 static void mult(BigInt& aRef, const BigInt& bRef)
276 {
277     const BigInt* a = &aRef;
278     const BigInt* b = &bRef;
279     BigInt c;
280     int wa, wb, wc;
281     const uint32_t* x = 0;
282     const uint32_t* xa;
283     const uint32_t* xb;
284     const uint32_t* xae;
285     const uint32_t* xbe;
286     uint32_t* xc;
287     uint32_t* xc0;
288     uint32_t y;
289 #ifdef USE_LONG_LONG
290     unsigned long long carry, z;
291 #else
292     uint32_t carry, z;
293 #endif
294 
295     if (a->size() < b->size()) {
296         const BigInt* tmp = a;
297         a = b;
298         b = tmp;
299     }
300 
301     wa = a->size();
302     wb = b->size();
303     wc = wa + wb;
304     c.resize(wc);
305 
306     for (xc = c.words(), xa = xc + wc; xc < xa; xc++)
307         *xc = 0;
308     xa = a->words();
309     xae = xa + wa;
310     xb = b->words();
311     xbe = xb + wb;
312     xc0 = c.words();
313 #ifdef USE_LONG_LONG
314     for (; xb < xbe; xc0++) {
315         if ((y = *xb++)) {
316             x = xa;
317             xc = xc0;
318             carry = 0;
319             do {
320                 z = *x++ * (unsigned long long)y + *xc + carry;
321                 carry = z >> 32;
322                 *xc++ = (uint32_t)z & 0xffffffffUL;
323             } while (x < xae);
324             *xc = (uint32_t)carry;
325         }
326     }
327 #else
328     for (; xb < xbe; xb++, xc0++) {
329         if ((y = *xb & 0xffff)) {
330             x = xa;
331             xc = xc0;
332             carry = 0;
333             do {
334                 z = (*x & 0xffff) * y + (*xc & 0xffff) + carry;
335                 carry = z >> 16;
336                 uint32_t z2 = (*x++ >> 16) * y + (*xc >> 16) + carry;
337                 carry = z2 >> 16;
338                 xc = storeInc(xc, z2, z);
339             } while (x < xae);
340             *xc = carry;
341         }
342         if ((y = *xb >> 16)) {
343             x = xa;
344             xc = xc0;
345             carry = 0;
346             uint32_t z2 = *xc;
347             do {
348                 z = (*x & 0xffff) * y + (*xc >> 16) + carry;
349                 carry = z >> 16;
350                 xc = storeInc(xc, z, z2);
351                 z2 = (*x++ >> 16) * y + (*xc & 0xffff) + carry;
352                 carry = z2 >> 16;
353             } while (x < xae);
354             *xc = z2;
355         }
356     }
357 #endif
358     for (xc0 = c.words(), xc = xc0 + wc; wc > 0 && !*--xc; --wc) { }
359     c.resize(wc);
360     aRef = c;
361 }
362 
363 struct P5Node {
364     WTF_MAKE_NONCOPYABLE(P5Node); WTF_MAKE_FAST_ALLOCATED;
365 public:
P5NodeWTF::P5Node366     P5Node() { }
367     BigInt val;
368     P5Node* next;
369 };
370 
371 static P5Node* p5s;
372 static int p5sCount;
373 
pow5mult(BigInt & b,int k)374 static ALWAYS_INLINE void pow5mult(BigInt& b, int k)
375 {
376     static int p05[3] = { 5, 25, 125 };
377 
378     if (int i = k & 3)
379         multadd(b, p05[i - 1], 0);
380 
381     if (!(k >>= 2))
382         return;
383 
384     s_dtoaP5Mutex->lock();
385     P5Node* p5 = p5s;
386 
387     if (!p5) {
388         /* first time */
389         p5 = new P5Node;
390         i2b(p5->val, 625);
391         p5->next = 0;
392         p5s = p5;
393         p5sCount = 1;
394     }
395 
396     int p5sCountLocal = p5sCount;
397     s_dtoaP5Mutex->unlock();
398     int p5sUsed = 0;
399 
400     for (;;) {
401         if (k & 1)
402             mult(b, p5->val);
403 
404         if (!(k >>= 1))
405             break;
406 
407         if (++p5sUsed == p5sCountLocal) {
408             s_dtoaP5Mutex->lock();
409             if (p5sUsed == p5sCount) {
410                 ASSERT(!p5->next);
411                 p5->next = new P5Node;
412                 p5->next->next = 0;
413                 p5->next->val = p5->val;
414                 mult(p5->next->val, p5->next->val);
415                 ++p5sCount;
416             }
417 
418             p5sCountLocal = p5sCount;
419             s_dtoaP5Mutex->unlock();
420         }
421         p5 = p5->next;
422     }
423 }
424 
lshift(BigInt & b,int k)425 static ALWAYS_INLINE void lshift(BigInt& b, int k)
426 {
427     int n = k >> 5;
428 
429     int origSize = b.size();
430     int n1 = n + origSize + 1;
431 
432     if (k &= 0x1f)
433         b.resize(b.size() + n + 1);
434     else
435         b.resize(b.size() + n);
436 
437     const uint32_t* srcStart = b.words();
438     uint32_t* dstStart = b.words();
439     const uint32_t* src = srcStart + origSize - 1;
440     uint32_t* dst = dstStart + n1 - 1;
441     if (k) {
442         uint32_t hiSubword = 0;
443         int s = 32 - k;
444         for (; src >= srcStart; --src) {
445             *dst-- = hiSubword | *src >> s;
446             hiSubword = *src << k;
447         }
448         *dst = hiSubword;
449         ASSERT(dst == dstStart + n);
450 
451         b.resize(origSize + n + !!b.words()[n1 - 1]);
452     }
453     else {
454         do {
455             *--dst = *src--;
456         } while (src >= srcStart);
457     }
458     for (dst = dstStart + n; dst != dstStart; )
459         *--dst = 0;
460 
461     ASSERT(b.size() <= 1 || b.words()[b.size() - 1]);
462 }
463 
cmp(const BigInt & a,const BigInt & b)464 static int cmp(const BigInt& a, const BigInt& b)
465 {
466     const uint32_t *xa, *xa0, *xb, *xb0;
467     int i, j;
468 
469     i = a.size();
470     j = b.size();
471     ASSERT(i <= 1 || a.words()[i - 1]);
472     ASSERT(j <= 1 || b.words()[j - 1]);
473     if (i -= j)
474         return i;
475     xa0 = a.words();
476     xa = xa0 + j;
477     xb0 = b.words();
478     xb = xb0 + j;
479     for (;;) {
480         if (*--xa != *--xb)
481             return *xa < *xb ? -1 : 1;
482         if (xa <= xa0)
483             break;
484     }
485     return 0;
486 }
487 
diff(BigInt & c,const BigInt & aRef,const BigInt & bRef)488 static ALWAYS_INLINE void diff(BigInt& c, const BigInt& aRef, const BigInt& bRef)
489 {
490     const BigInt* a = &aRef;
491     const BigInt* b = &bRef;
492     int i, wa, wb;
493     uint32_t* xc;
494 
495     i = cmp(*a, *b);
496     if (!i) {
497         c.sign = 0;
498         c.resize(1);
499         c.words()[0] = 0;
500         return;
501     }
502     if (i < 0) {
503         const BigInt* tmp = a;
504         a = b;
505         b = tmp;
506         i = 1;
507     } else
508         i = 0;
509 
510     wa = a->size();
511     const uint32_t* xa = a->words();
512     const uint32_t* xae = xa + wa;
513     wb = b->size();
514     const uint32_t* xb = b->words();
515     const uint32_t* xbe = xb + wb;
516 
517     c.resize(wa);
518     c.sign = i;
519     xc = c.words();
520 #ifdef USE_LONG_LONG
521     unsigned long long borrow = 0;
522     do {
523         unsigned long long y = (unsigned long long)*xa++ - *xb++ - borrow;
524         borrow = y >> 32 & (uint32_t)1;
525         *xc++ = (uint32_t)y & 0xffffffffUL;
526     } while (xb < xbe);
527     while (xa < xae) {
528         unsigned long long y = *xa++ - borrow;
529         borrow = y >> 32 & (uint32_t)1;
530         *xc++ = (uint32_t)y & 0xffffffffUL;
531     }
532 #else
533     uint32_t borrow = 0;
534     do {
535         uint32_t y = (*xa & 0xffff) - (*xb & 0xffff) - borrow;
536         borrow = (y & 0x10000) >> 16;
537         uint32_t z = (*xa++ >> 16) - (*xb++ >> 16) - borrow;
538         borrow = (z & 0x10000) >> 16;
539         xc = storeInc(xc, z, y);
540     } while (xb < xbe);
541     while (xa < xae) {
542         uint32_t y = (*xa & 0xffff) - borrow;
543         borrow = (y & 0x10000) >> 16;
544         uint32_t z = (*xa++ >> 16) - borrow;
545         borrow = (z & 0x10000) >> 16;
546         xc = storeInc(xc, z, y);
547     }
548 #endif
549     while (!*--xc)
550         wa--;
551     c.resize(wa);
552 }
553 
d2b(BigInt & b,U * d,int * e,int * bits)554 static ALWAYS_INLINE void d2b(BigInt& b, U* d, int* e, int* bits)
555 {
556     int de, k;
557     uint32_t* x;
558     uint32_t y, z;
559     int i;
560 #define d0 word0(d)
561 #define d1 word1(d)
562 
563     b.sign = 0;
564     b.resize(1);
565     x = b.words();
566 
567     z = d0 & Frac_mask;
568     d0 &= 0x7fffffff;    /* clear sign bit, which we ignore */
569     if ((de = (int)(d0 >> Exp_shift)))
570         z |= Exp_msk1;
571     if ((y = d1)) {
572         if ((k = lo0bits(&y))) {
573             x[0] = y | (z << (32 - k));
574             z >>= k;
575         } else
576             x[0] = y;
577         if (z) {
578             b.resize(2);
579             x[1] = z;
580         }
581 
582         i = b.size();
583     } else {
584         k = lo0bits(&z);
585         x[0] = z;
586         i = 1;
587         b.resize(1);
588         k += 32;
589     }
590     if (de) {
591         *e = de - Bias - (P - 1) + k;
592         *bits = P - k;
593     } else {
594         *e = 0 - Bias - (P - 1) + 1 + k;
595         *bits = (32 * i) - hi0bits(x[i - 1]);
596     }
597 }
598 #undef d0
599 #undef d1
600 
601 static const double tens[] = {
602     1e0, 1e1, 1e2, 1e3, 1e4, 1e5, 1e6, 1e7, 1e8, 1e9,
603     1e10, 1e11, 1e12, 1e13, 1e14, 1e15, 1e16, 1e17, 1e18, 1e19,
604     1e20, 1e21, 1e22
605 };
606 
607 static const double bigtens[] = { 1e16, 1e32, 1e64, 1e128, 1e256 };
608 
609 #define Scale_Bit 0x10
610 #define n_bigtens 5
611 
quorem(BigInt & b,BigInt & S)612 static ALWAYS_INLINE int quorem(BigInt& b, BigInt& S)
613 {
614     size_t n;
615     uint32_t* bx;
616     uint32_t* bxe;
617     uint32_t q;
618     uint32_t* sx;
619     uint32_t* sxe;
620 #ifdef USE_LONG_LONG
621     unsigned long long borrow, carry, y, ys;
622 #else
623     uint32_t borrow, carry, y, ys;
624     uint32_t si, z, zs;
625 #endif
626     ASSERT(b.size() <= 1 || b.words()[b.size() - 1]);
627     ASSERT(S.size() <= 1 || S.words()[S.size() - 1]);
628 
629     n = S.size();
630     ASSERT_WITH_MESSAGE(b.size() <= n, "oversize b in quorem");
631     if (b.size() < n)
632         return 0;
633     sx = S.words();
634     sxe = sx + --n;
635     bx = b.words();
636     bxe = bx + n;
637     q = *bxe / (*sxe + 1);    /* ensure q <= true quotient */
638     ASSERT_WITH_MESSAGE(q <= 9, "oversized quotient in quorem");
639     if (q) {
640         borrow = 0;
641         carry = 0;
642         do {
643 #ifdef USE_LONG_LONG
644             ys = *sx++ * (unsigned long long)q + carry;
645             carry = ys >> 32;
646             y = *bx - (ys & 0xffffffffUL) - borrow;
647             borrow = y >> 32 & (uint32_t)1;
648             *bx++ = (uint32_t)y & 0xffffffffUL;
649 #else
650             si = *sx++;
651             ys = (si & 0xffff) * q + carry;
652             zs = (si >> 16) * q + (ys >> 16);
653             carry = zs >> 16;
654             y = (*bx & 0xffff) - (ys & 0xffff) - borrow;
655             borrow = (y & 0x10000) >> 16;
656             z = (*bx >> 16) - (zs & 0xffff) - borrow;
657             borrow = (z & 0x10000) >> 16;
658             bx = storeInc(bx, z, y);
659 #endif
660         } while (sx <= sxe);
661         if (!*bxe) {
662             bx = b.words();
663             while (--bxe > bx && !*bxe)
664                 --n;
665             b.resize(n);
666         }
667     }
668     if (cmp(b, S) >= 0) {
669         q++;
670         borrow = 0;
671         carry = 0;
672         bx = b.words();
673         sx = S.words();
674         do {
675 #ifdef USE_LONG_LONG
676             ys = *sx++ + carry;
677             carry = ys >> 32;
678             y = *bx - (ys & 0xffffffffUL) - borrow;
679             borrow = y >> 32 & (uint32_t)1;
680             *bx++ = (uint32_t)y & 0xffffffffUL;
681 #else
682             si = *sx++;
683             ys = (si & 0xffff) + carry;
684             zs = (si >> 16) + (ys >> 16);
685             carry = zs >> 16;
686             y = (*bx & 0xffff) - (ys & 0xffff) - borrow;
687             borrow = (y & 0x10000) >> 16;
688             z = (*bx >> 16) - (zs & 0xffff) - borrow;
689             borrow = (z & 0x10000) >> 16;
690             bx = storeInc(bx, z, y);
691 #endif
692         } while (sx <= sxe);
693         bx = b.words();
694         bxe = bx + n;
695         if (!*bxe) {
696             while (--bxe > bx && !*bxe)
697                 --n;
698             b.resize(n);
699         }
700     }
701     return q;
702 }
703 
704 /* dtoa for IEEE arithmetic (dmg): convert double to ASCII string.
705  *
706  * Inspired by "How to Print Floating-Point Numbers Accurately" by
707  * Guy L. Steele, Jr. and Jon L. White [Proc. ACM SIGPLAN '90, pp. 112-126].
708  *
709  * Modifications:
710  *    1. Rather than iterating, we use a simple numeric overestimate
711  *       to determine k = floor(log10(d)).  We scale relevant
712  *       quantities using O(log2(k)) rather than O(k) multiplications.
713  *    2. For some modes > 2 (corresponding to ecvt and fcvt), we don't
714  *       try to generate digits strictly left to right.  Instead, we
715  *       compute with fewer bits and propagate the carry if necessary
716  *       when rounding the final digit up.  This is often faster.
717  *    3. Under the assumption that input will be rounded nearest,
718  *       mode 0 renders 1e23 as 1e23 rather than 9.999999999999999e22.
719  *       That is, we allow equality in stopping tests when the
720  *       round-nearest rule will give the same floating-point value
721  *       as would satisfaction of the stopping test with strict
722  *       inequality.
723  *    4. We remove common factors of powers of 2 from relevant
724  *       quantities.
725  *    5. When converting floating-point integers less than 1e16,
726  *       we use floating-point arithmetic rather than resorting
727  *       to multiple-precision integers.
728  *    6. When asked to produce fewer than 15 digits, we first try
729  *       to get by with floating-point arithmetic; we resort to
730  *       multiple-precision integer arithmetic only if we cannot
731  *       guarantee that the floating-point calculation has given
732  *       the correctly rounded result.  For k requested digits and
733  *       "uniformly" distributed input, the probability is
734  *       something like 10^(k-15) that we must resort to the int32_t
735  *       calculation.
736  *
737  * Note: 'leftright' translates to 'generate shortest possible string'.
738  */
739 template<bool roundingNone, bool roundingSignificantFigures, bool roundingDecimalPlaces, bool leftright>
dtoa(DtoaBuffer result,double dd,int ndigits,bool & signOut,int & exponentOut,unsigned & precisionOut)740 void dtoa(DtoaBuffer result, double dd, int ndigits, bool& signOut, int& exponentOut, unsigned& precisionOut)
741 {
742     // Exactly one rounding mode must be specified.
743     ASSERT(roundingNone + roundingSignificantFigures + roundingDecimalPlaces == 1);
744     // roundingNone only allowed (only sensible?) with leftright set.
745     ASSERT(!roundingNone || leftright);
746 
747     ASSERT(std::isfinite(dd));
748 
749     int bbits, b2, b5, be, dig, i, ieps, ilim = 0, ilim0, ilim1 = 0,
750         j, j1, k, k0, k_check, m2, m5, s2, s5,
751         spec_case;
752     int32_t L;
753     int denorm;
754     uint32_t x;
755     BigInt b, delta, mlo, mhi, S;
756     U d2, eps, u;
757     double ds;
758     char* s;
759     char* s0;
760 
761     u.d = dd;
762 
763     /* Infinity or NaN */
764     ASSERT((word0(&u) & Exp_mask) != Exp_mask);
765 
766     // JavaScript toString conversion treats -0 as 0.
767     if (!dval(&u)) {
768         signOut = false;
769         exponentOut = 0;
770         precisionOut = 1;
771         result[0] = '0';
772         result[1] = '\0';
773         return;
774     }
775 
776     if (word0(&u) & Sign_bit) {
777         signOut = true;
778         word0(&u) &= ~Sign_bit; // clear sign bit
779     } else
780         signOut = false;
781 
782     d2b(b, &u, &be, &bbits);
783     if ((i = (int)(word0(&u) >> Exp_shift1 & (Exp_mask >> Exp_shift1)))) {
784         dval(&d2) = dval(&u);
785         word0(&d2) &= Frac_mask1;
786         word0(&d2) |= Exp_11;
787 
788         /* log(x)    ~=~ log(1.5) + (x-1.5)/1.5
789          * log10(x)     =  log(x) / log(10)
790          *        ~=~ log(1.5)/log(10) + (x-1.5)/(1.5*log(10))
791          * log10(d) = (i-Bias)*log(2)/log(10) + log10(d2)
792          *
793          * This suggests computing an approximation k to log10(d) by
794          *
795          * k = (i - Bias)*0.301029995663981
796          *    + ( (d2-1.5)*0.289529654602168 + 0.176091259055681 );
797          *
798          * We want k to be too large rather than too small.
799          * The error in the first-order Taylor series approximation
800          * is in our favor, so we just round up the constant enough
801          * to compensate for any error in the multiplication of
802          * (i - Bias) by 0.301029995663981; since |i - Bias| <= 1077,
803          * and 1077 * 0.30103 * 2^-52 ~=~ 7.2e-14,
804          * adding 1e-13 to the constant term more than suffices.
805          * Hence we adjust the constant term to 0.1760912590558.
806          * (We could get a more accurate k by invoking log10,
807          *  but this is probably not worthwhile.)
808          */
809 
810         i -= Bias;
811         denorm = 0;
812     } else {
813         /* d is denormalized */
814 
815         i = bbits + be + (Bias + (P - 1) - 1);
816         x = (i > 32) ? (word0(&u) << (64 - i)) | (word1(&u) >> (i - 32))
817                 : word1(&u) << (32 - i);
818         dval(&d2) = x;
819         word0(&d2) -= 31 * Exp_msk1; /* adjust exponent */
820         i -= (Bias + (P - 1) - 1) + 1;
821         denorm = 1;
822     }
823     ds = (dval(&d2) - 1.5) * 0.289529654602168 + 0.1760912590558 + (i * 0.301029995663981);
824     k = (int)ds;
825     if (ds < 0. && ds != k)
826         k--;    /* want k = floor(ds) */
827     k_check = 1;
828     if (k >= 0 && k <= Ten_pmax) {
829         if (dval(&u) < tens[k])
830             k--;
831         k_check = 0;
832     }
833     j = bbits - i - 1;
834     if (j >= 0) {
835         b2 = 0;
836         s2 = j;
837     } else {
838         b2 = -j;
839         s2 = 0;
840     }
841     if (k >= 0) {
842         b5 = 0;
843         s5 = k;
844         s2 += k;
845     } else {
846         b2 -= k;
847         b5 = -k;
848         s5 = 0;
849     }
850 
851     if (roundingNone) {
852         ilim = ilim1 = -1;
853         i = 18;
854         ndigits = 0;
855     }
856     if (roundingSignificantFigures) {
857         if (ndigits <= 0)
858             ndigits = 1;
859         ilim = ilim1 = i = ndigits;
860     }
861     if (roundingDecimalPlaces) {
862         i = ndigits + k + 1;
863         ilim = i;
864         ilim1 = i - 1;
865         if (i <= 0)
866             i = 1;
867     }
868 
869     s = s0 = result;
870 
871     if (ilim >= 0 && ilim <= Quick_max) {
872         /* Try to get by with floating-point arithmetic. */
873 
874         i = 0;
875         dval(&d2) = dval(&u);
876         k0 = k;
877         ilim0 = ilim;
878         ieps = 2; /* conservative */
879         if (k > 0) {
880             ds = tens[k & 0xf];
881             j = k >> 4;
882             if (j & Bletch) {
883                 /* prevent overflows */
884                 j &= Bletch - 1;
885                 dval(&u) /= bigtens[n_bigtens - 1];
886                 ieps++;
887             }
888             for (; j; j >>= 1, i++) {
889                 if (j & 1) {
890                     ieps++;
891                     ds *= bigtens[i];
892                 }
893             }
894             dval(&u) /= ds;
895         } else if ((j1 = -k)) {
896             dval(&u) *= tens[j1 & 0xf];
897             for (j = j1 >> 4; j; j >>= 1, i++) {
898                 if (j & 1) {
899                     ieps++;
900                     dval(&u) *= bigtens[i];
901                 }
902             }
903         }
904         if (k_check && dval(&u) < 1. && ilim > 0) {
905             if (ilim1 <= 0)
906                 goto fastFailed;
907             ilim = ilim1;
908             k--;
909             dval(&u) *= 10.;
910             ieps++;
911         }
912         dval(&eps) = (ieps * dval(&u)) + 7.;
913         word0(&eps) -= (P - 1) * Exp_msk1;
914         if (!ilim) {
915             S.clear();
916             mhi.clear();
917             dval(&u) -= 5.;
918             if (dval(&u) > dval(&eps))
919                 goto oneDigit;
920             if (dval(&u) < -dval(&eps))
921                 goto noDigits;
922             goto fastFailed;
923         }
924         if (leftright) {
925             /* Use Steele & White method of only
926              * generating digits needed.
927              */
928             dval(&eps) = (0.5 / tens[ilim - 1]) - dval(&eps);
929             for (i = 0;;) {
930                 L = (long int)dval(&u);
931                 dval(&u) -= L;
932                 *s++ = '0' + (int)L;
933                 if (dval(&u) < dval(&eps))
934                     goto ret;
935                 if (1. - dval(&u) < dval(&eps))
936                     goto bumpUp;
937                 if (++i >= ilim)
938                     break;
939                 dval(&eps) *= 10.;
940                 dval(&u) *= 10.;
941             }
942         } else {
943             /* Generate ilim digits, then fix them up. */
944             dval(&eps) *= tens[ilim - 1];
945             for (i = 1;; i++, dval(&u) *= 10.) {
946                 L = (int32_t)(dval(&u));
947                 if (!(dval(&u) -= L))
948                     ilim = i;
949                 *s++ = '0' + (int)L;
950                 if (i == ilim) {
951                     if (dval(&u) > 0.5 + dval(&eps))
952                         goto bumpUp;
953                     if (dval(&u) < 0.5 - dval(&eps)) {
954                         while (*--s == '0') { }
955                         s++;
956                         goto ret;
957                     }
958                     break;
959                 }
960             }
961         }
962 fastFailed:
963         s = s0;
964         dval(&u) = dval(&d2);
965         k = k0;
966         ilim = ilim0;
967     }
968 
969     /* Do we have a "small" integer? */
970 
971     if (be >= 0 && k <= Int_max) {
972         /* Yes. */
973         ds = tens[k];
974         if (ndigits < 0 && ilim <= 0) {
975             S.clear();
976             mhi.clear();
977             if (ilim < 0 || dval(&u) <= 5 * ds)
978                 goto noDigits;
979             goto oneDigit;
980         }
981         for (i = 1;; i++, dval(&u) *= 10.) {
982             L = (int32_t)(dval(&u) / ds);
983             dval(&u) -= L * ds;
984             *s++ = '0' + (int)L;
985             if (!dval(&u)) {
986                 break;
987             }
988             if (i == ilim) {
989                 dval(&u) += dval(&u);
990                 if (dval(&u) > ds || (dval(&u) == ds && (L & 1))) {
991 bumpUp:
992                     while (*--s == '9')
993                         if (s == s0) {
994                             k++;
995                             *s = '0';
996                             break;
997                         }
998                     ++*s++;
999                 }
1000                 break;
1001             }
1002         }
1003         goto ret;
1004     }
1005 
1006     m2 = b2;
1007     m5 = b5;
1008     mhi.clear();
1009     mlo.clear();
1010     if (leftright) {
1011         i = denorm ? be + (Bias + (P - 1) - 1 + 1) : 1 + P - bbits;
1012         b2 += i;
1013         s2 += i;
1014         i2b(mhi, 1);
1015     }
1016     if (m2 > 0 && s2 > 0) {
1017         i = m2 < s2 ? m2 : s2;
1018         b2 -= i;
1019         m2 -= i;
1020         s2 -= i;
1021     }
1022     if (b5 > 0) {
1023         if (leftright) {
1024             if (m5 > 0) {
1025                 pow5mult(mhi, m5);
1026                 mult(b, mhi);
1027             }
1028             if ((j = b5 - m5))
1029                 pow5mult(b, j);
1030         } else
1031             pow5mult(b, b5);
1032     }
1033     i2b(S, 1);
1034     if (s5 > 0)
1035         pow5mult(S, s5);
1036 
1037     /* Check for special case that d is a normalized power of 2. */
1038 
1039     spec_case = 0;
1040     if ((roundingNone || leftright) && (!word1(&u) && !(word0(&u) & Bndry_mask) && word0(&u) & (Exp_mask & ~Exp_msk1))) {
1041         /* The special case */
1042         b2 += Log2P;
1043         s2 += Log2P;
1044         spec_case = 1;
1045     }
1046 
1047     /* Arrange for convenient computation of quotients:
1048      * shift left if necessary so divisor has 4 leading 0 bits.
1049      *
1050      * Perhaps we should just compute leading 28 bits of S once
1051      * and for all and pass them and a shift to quorem, so it
1052      * can do shifts and ors to compute the numerator for q.
1053      */
1054     if ((i = ((s5 ? 32 - hi0bits(S.words()[S.size() - 1]) : 1) + s2) & 0x1f))
1055         i = 32 - i;
1056     if (i > 4) {
1057         i -= 4;
1058         b2 += i;
1059         m2 += i;
1060         s2 += i;
1061     } else if (i < 4) {
1062         i += 28;
1063         b2 += i;
1064         m2 += i;
1065         s2 += i;
1066     }
1067     if (b2 > 0)
1068         lshift(b, b2);
1069     if (s2 > 0)
1070         lshift(S, s2);
1071     if (k_check) {
1072         if (cmp(b, S) < 0) {
1073             k--;
1074             multadd(b, 10, 0);    /* we botched the k estimate */
1075             if (leftright)
1076                 multadd(mhi, 10, 0);
1077             ilim = ilim1;
1078         }
1079     }
1080     if (ilim <= 0 && roundingDecimalPlaces) {
1081         if (ilim < 0)
1082             goto noDigits;
1083         multadd(S, 5, 0);
1084         // For IEEE-754 unbiased rounding this check should be <=, such that 0.5 would flush to zero.
1085         if (cmp(b, S) < 0)
1086             goto noDigits;
1087         goto oneDigit;
1088     }
1089     if (leftright) {
1090         if (m2 > 0)
1091             lshift(mhi, m2);
1092 
1093         /* Compute mlo -- check for special case
1094          * that d is a normalized power of 2.
1095          */
1096 
1097         mlo = mhi;
1098         if (spec_case)
1099             lshift(mhi, Log2P);
1100 
1101         for (i = 1;;i++) {
1102             dig = quorem(b, S) + '0';
1103             /* Do we yet have the shortest decimal string
1104              * that will round to d?
1105              */
1106             j = cmp(b, mlo);
1107             diff(delta, S, mhi);
1108             j1 = delta.sign ? 1 : cmp(b, delta);
1109 #ifdef DTOA_ROUND_BIASED
1110             if (j < 0 || !j) {
1111 #else
1112             // FIXME: ECMA-262 specifies that equidistant results round away from
1113             // zero, which probably means we shouldn't be on the unbiased code path
1114             // (the (word1(&u) & 1) clause is looking highly suspicious). I haven't
1115             // yet understood this code well enough to make the call, but we should
1116             // probably be enabling DTOA_ROUND_BIASED. I think the interesting corner
1117             // case to understand is probably "Math.pow(0.5, 24).toString()".
1118             // I believe this value is interesting because I think it is precisely
1119             // representable in binary floating point, and its decimal representation
1120             // has a single digit that Steele & White reduction can remove, with the
1121             // value 5 (thus equidistant from the next numbers above and below).
1122             // We produce the correct answer using either codepath, and I don't as
1123             // yet understand why. :-)
1124             if (!j1 && !(word1(&u) & 1)) {
1125                 if (dig == '9')
1126                     goto round9up;
1127                 if (j > 0)
1128                     dig++;
1129                 *s++ = dig;
1130                 goto ret;
1131             }
1132             if (j < 0 || (!j && !(word1(&u) & 1))) {
1133 #endif
1134                 if ((b.words()[0] || b.size() > 1) && (j1 > 0)) {
1135                     lshift(b, 1);
1136                     j1 = cmp(b, S);
1137                     // For IEEE-754 round-to-even, this check should be (j1 > 0 || (!j1 && (dig & 1))),
1138                     // but ECMA-262 specifies that equidistant values (e.g. (.5).toFixed()) should
1139                     // be rounded away from zero.
1140                     if (j1 >= 0) {
1141                         if (dig == '9')
1142                             goto round9up;
1143                         dig++;
1144                     }
1145                 }
1146                 *s++ = dig;
1147                 goto ret;
1148             }
1149             if (j1 > 0) {
1150                 if (dig == '9') { /* possible if i == 1 */
1151 round9up:
1152                     *s++ = '9';
1153                     goto roundoff;
1154                 }
1155                 *s++ = dig + 1;
1156                 goto ret;
1157             }
1158             *s++ = dig;
1159             if (i == ilim)
1160                 break;
1161             multadd(b, 10, 0);
1162             multadd(mlo, 10, 0);
1163             multadd(mhi, 10, 0);
1164         }
1165     } else {
1166         for (i = 1;; i++) {
1167             *s++ = dig = quorem(b, S) + '0';
1168             if (!b.words()[0] && b.size() <= 1)
1169                 goto ret;
1170             if (i >= ilim)
1171                 break;
1172             multadd(b, 10, 0);
1173         }
1174     }
1175 
1176     /* Round off last digit */
1177 
1178     lshift(b, 1);
1179     j = cmp(b, S);
1180     // For IEEE-754 round-to-even, this check should be (j > 0 || (!j && (dig & 1))),
1181     // but ECMA-262 specifies that equidistant values (e.g. (.5).toFixed()) should
1182     // be rounded away from zero.
1183     if (j >= 0) {
1184 roundoff:
1185         while (*--s == '9')
1186             if (s == s0) {
1187                 k++;
1188                 *s++ = '1';
1189                 goto ret;
1190             }
1191         ++*s++;
1192     } else {
1193         while (*--s == '0') { }
1194         s++;
1195     }
1196     goto ret;
1197 noDigits:
1198     exponentOut = 0;
1199     precisionOut = 1;
1200     result[0] = '0';
1201     result[1] = '\0';
1202     return;
1203 oneDigit:
1204     *s++ = '1';
1205     k++;
1206     goto ret;
1207 ret:
1208     ASSERT(s > result);
1209     *s = 0;
1210     exponentOut = k;
1211     precisionOut = s - result;
1212 }
1213 
1214 void dtoa(DtoaBuffer result, double dd, bool& sign, int& exponent, unsigned& precision)
1215 {
1216     // flags are roundingNone, leftright.
1217     dtoa<true, false, false, true>(result, dd, 0, sign, exponent, precision);
1218 }
1219 
1220 void dtoaRoundSF(DtoaBuffer result, double dd, int ndigits, bool& sign, int& exponent, unsigned& precision)
1221 {
1222     // flag is roundingSignificantFigures.
1223     dtoa<false, true, false, false>(result, dd, ndigits, sign, exponent, precision);
1224 }
1225 
1226 void dtoaRoundDP(DtoaBuffer result, double dd, int ndigits, bool& sign, int& exponent, unsigned& precision)
1227 {
1228     // flag is roundingDecimalPlaces.
1229     dtoa<false, false, true, false>(result, dd, ndigits, sign, exponent, precision);
1230 }
1231 
1232 const char* numberToString(double d, NumberToStringBuffer buffer)
1233 {
1234     double_conversion::StringBuilder builder(buffer, NumberToStringBufferLength);
1235     const double_conversion::DoubleToStringConverter& converter = double_conversion::DoubleToStringConverter::EcmaScriptConverter();
1236     converter.ToShortest(d, &builder);
1237     return builder.Finalize();
1238 }
1239 
1240 static inline const char* formatStringTruncatingTrailingZerosIfNeeded(NumberToStringBuffer buffer, double_conversion::StringBuilder& builder)
1241 {
1242     size_t length = builder.position();
1243     size_t decimalPointPosition = 0;
1244     for (; decimalPointPosition < length; ++decimalPointPosition) {
1245         if (buffer[decimalPointPosition] == '.')
1246             break;
1247     }
1248 
1249     // No decimal seperator found, early exit.
1250     if (decimalPointPosition == length)
1251         return builder.Finalize();
1252 
1253     size_t truncatedLength = length - 1;
1254     for (; truncatedLength > decimalPointPosition; --truncatedLength) {
1255         if (buffer[truncatedLength] != '0')
1256             break;
1257     }
1258 
1259     // No trailing zeros found to strip.
1260     if (truncatedLength == length - 1)
1261         return builder.Finalize();
1262 
1263     // If we removed all trailing zeros, remove the decimal point as well.
1264     if (truncatedLength == decimalPointPosition) {
1265         ASSERT(truncatedLength > 0);
1266         --truncatedLength;
1267     }
1268 
1269     // Truncate the StringBuilder, and return the final result.
1270     builder.SetPosition(truncatedLength + 1);
1271     return builder.Finalize();
1272 }
1273 
1274 const char* numberToFixedPrecisionString(double d, unsigned significantFigures, NumberToStringBuffer buffer, bool truncateTrailingZeros)
1275 {
1276     // Mimic String::format("%.[precision]g", ...), but use dtoas rounding facilities.
1277     // "g": Signed value printed in f or e format, whichever is more compact for the given value and precision.
1278     // The e format is used only when the exponent of the value is less than –4 or greater than or equal to the
1279     // precision argument. Trailing zeros are truncated, and the decimal point appears only if one or more digits follow it.
1280     // "precision": The precision specifies the maximum number of significant digits printed.
1281     double_conversion::StringBuilder builder(buffer, NumberToStringBufferLength);
1282     const double_conversion::DoubleToStringConverter& converter = double_conversion::DoubleToStringConverter::EcmaScriptConverter();
1283     converter.ToPrecision(d, significantFigures, &builder);
1284     if (!truncateTrailingZeros)
1285         return builder.Finalize();
1286     return formatStringTruncatingTrailingZerosIfNeeded(buffer, builder);
1287 }
1288 
1289 const char* numberToFixedWidthString(double d, unsigned decimalPlaces, NumberToStringBuffer buffer)
1290 {
1291     // Mimic String::format("%.[precision]f", ...), but use dtoas rounding facilities.
1292     // "f": Signed value having the form [ – ]dddd.dddd, where dddd is one or more decimal digits.
1293     // The number of digits before the decimal point depends on the magnitude of the number, and
1294     // the number of digits after the decimal point depends on the requested precision.
1295     // "precision": The precision value specifies the number of digits after the decimal point.
1296     // If a decimal point appears, at least one digit appears before it.
1297     // The value is rounded to the appropriate number of digits.
1298     double_conversion::StringBuilder builder(buffer, NumberToStringBufferLength);
1299     const double_conversion::DoubleToStringConverter& converter = double_conversion::DoubleToStringConverter::EcmaScriptConverter();
1300     converter.ToFixed(d, decimalPlaces, &builder);
1301     return builder.Finalize();
1302 }
1303 
1304 namespace Internal {
1305 
1306 double parseDoubleFromLongString(const UChar* string, size_t length, size_t& parsedLength)
1307 {
1308     Vector<LChar> conversionBuffer(length);
1309     for (size_t i = 0; i < length; ++i)
1310         conversionBuffer[i] = isASCII(string[i]) ? string[i] : 0;
1311     return parseDouble(conversionBuffer.data(), length, parsedLength);
1312 }
1313 
1314 } // namespace Internal
1315 
1316 } // namespace WTF
1317