Coverage Report

Created: 2026-04-12 06:54

next uncovered line (L), next uncovered region (R), next uncovered branch (B)
/src/cpython/Python/dtoa.c
Line
Count
Source
1
/****************************************************************
2
 *
3
 * The author of this software is David M. Gay.
4
 *
5
 * Copyright (c) 1991, 2000, 2001 by Lucent Technologies.
6
 *
7
 * Permission to use, copy, modify, and distribute this software for any
8
 * purpose without fee is hereby granted, provided that this entire notice
9
 * is included in all copies of any software which is or includes a copy
10
 * or modification of this software and in all copies of the supporting
11
 * documentation for such software.
12
 *
13
 * THIS SOFTWARE IS BEING PROVIDED "AS IS", WITHOUT ANY EXPRESS OR IMPLIED
14
 * WARRANTY.  IN PARTICULAR, NEITHER THE AUTHOR NOR LUCENT MAKES ANY
15
 * REPRESENTATION OR WARRANTY OF ANY KIND CONCERNING THE MERCHANTABILITY
16
 * OF THIS SOFTWARE OR ITS FITNESS FOR ANY PARTICULAR PURPOSE.
17
 *
18
 ***************************************************************/
19
20
/****************************************************************
21
 * This is dtoa.c by David M. Gay, downloaded from
22
 * http://www.netlib.org/fp/dtoa.c on April 15, 2009 and modified for
23
 * inclusion into the Python core by Mark E. T. Dickinson and Eric V. Smith.
24
 *
25
 * Please remember to check http://www.netlib.org/fp regularly (and especially
26
 * before any Python release) for bugfixes and updates.
27
 *
28
 * The major modifications from Gay's original code are as follows:
29
 *
30
 *  0. The original code has been specialized to Python's needs by removing
31
 *     many of the #ifdef'd sections.  In particular, code to support VAX and
32
 *     IBM floating-point formats, hex NaNs, hex floats, locale-aware
33
 *     treatment of the decimal point, and setting of the inexact flag have
34
 *     been removed.
35
 *
36
 *  1. We use PyMem_Malloc and PyMem_Free in place of malloc and free.
37
 *
38
 *  2. The public functions strtod, dtoa and freedtoa all now have
39
 *     a _Py_dg_ prefix.
40
 *
41
 *  3. Instead of assuming that PyMem_Malloc always succeeds, we thread
42
 *     PyMem_Malloc failures through the code.  The functions
43
 *
44
 *       Balloc, multadd, s2b, i2b, mult, pow5mult, lshift, diff, d2b
45
 *
46
 *     of return type *Bigint all return NULL to indicate a malloc failure.
47
 *     Similarly, rv_alloc and nrv_alloc (return type char *) return NULL on
48
 *     failure.  bigcomp now has return type int (it used to be void) and
49
 *     returns -1 on failure and 0 otherwise.  _Py_dg_dtoa returns NULL
50
 *     on failure.  _Py_dg_strtod indicates failure due to malloc failure
51
 *     by returning -1.0, setting errno=ENOMEM and *se to s00.
52
 *
53
 *  4. The static variable dtoa_result has been removed.  Callers of
54
 *     _Py_dg_dtoa are expected to call _Py_dg_freedtoa to free
55
 *     the memory allocated by _Py_dg_dtoa.
56
 *
57
 *  5. The code has been reformatted to better fit with Python's
58
 *     C style guide (PEP 7).
59
 *
60
 *  6. A bug in the memory allocation has been fixed: to avoid FREEing memory
61
 *     that hasn't been MALLOC'ed, private_mem should only be used when k <=
62
 *     Kmax.
63
 *
64
 *  7. _Py_dg_strtod has been modified so that it doesn't accept strings with
65
 *     leading whitespace.
66
 *
67
 *  8. A corner case where _Py_dg_dtoa didn't strip trailing zeros has been
68
 *     fixed. (bugs.python.org/issue40780)
69
 *
70
 ***************************************************************/
71
72
/* Please send bug reports for the original dtoa.c code to David M. Gay (dmg
73
 * at acm dot org, with " at " changed at "@" and " dot " changed to ".").
74
 * Please report bugs for this modified version using the Python issue tracker
75
 * as detailed at (https://devguide.python.org/triage/issue-tracker/). */
76
77
/* On a machine with IEEE extended-precision registers, it is
78
 * necessary to specify double-precision (53-bit) rounding precision
79
 * before invoking strtod or dtoa.  If the machine uses (the equivalent
80
 * of) Intel 80x87 arithmetic, the call
81
 *      _control87(PC_53, MCW_PC);
82
 * does this with many compilers.  Whether this or another call is
83
 * appropriate depends on the compiler; for this to work, it may be
84
 * necessary to #include "float.h" or another system-dependent header
85
 * file.
86
 */
87
88
/* strtod for IEEE-, VAX-, and IBM-arithmetic machines.
89
 *
90
 * This strtod returns a nearest machine number to the input decimal
91
 * string (or sets errno to ERANGE).  With IEEE arithmetic, ties are
92
 * broken by the IEEE round-even rule.  Otherwise ties are broken by
93
 * biased rounding (add half and chop).
94
 *
95
 * Inspired loosely by William D. Clinger's paper "How to Read Floating
96
 * Point Numbers Accurately" [Proc. ACM SIGPLAN '90, pp. 92-101].
97
 *
98
 * Modifications:
99
 *
100
 *      1. We only require IEEE, IBM, or VAX double-precision
101
 *              arithmetic (not IEEE double-extended).
102
 *      2. We get by with floating-point arithmetic in a case that
103
 *              Clinger missed -- when we're computing d * 10^n
104
 *              for a small integer d and the integer n is not too
105
 *              much larger than 22 (the maximum integer k for which
106
 *              we can represent 10^k exactly), we may be able to
107
 *              compute (d*10^k) * 10^(e-k) with just one roundoff.
108
 *      3. Rather than a bit-at-a-time adjustment of the binary
109
 *              result in the hard case, we use floating-point
110
 *              arithmetic to determine the adjustment to within
111
 *              one bit; only in really hard cases do we need to
112
 *              compute a second residual.
113
 *      4. Because of 3., we don't need a large table of powers of 10
114
 *              for ten-to-e (just some small tables, e.g. of 10^k
115
 *              for 0 <= k <= 22).
116
 */
117
118
/* Linking of Python's #defines to Gay's #defines starts here. */
119
120
#include "Python.h"
121
#include "pycore_dtoa.h"          // _PY_SHORT_FLOAT_REPR
122
#include "pycore_interp_structs.h"// struct Bigint
123
#include "pycore_pystate.h"       // _PyInterpreterState_GET()
124
#include <stdlib.h>               // exit()
125
126
127
/* if _PY_SHORT_FLOAT_REPR == 0, then don't even try to compile
128
   the following code */
129
#if _PY_SHORT_FLOAT_REPR == 1
130
131
#include "float.h"
132
133
59
#define MALLOC PyMem_Malloc
134
0
#define FREE PyMem_Free
135
136
/* This code should also work for ARM mixed-endian format on little-endian
137
   machines, where doubles have byte order 45670123 (in increasing address
138
   order, 0 being the least significant byte). */
139
#ifdef DOUBLE_IS_LITTLE_ENDIAN_IEEE754
140
#  define IEEE_8087
141
#endif
142
#if defined(DOUBLE_IS_BIG_ENDIAN_IEEE754)
143
#  define IEEE_MC68k
144
#endif
145
#if defined(IEEE_8087) + defined(IEEE_MC68k) != 1
146
#error "Exactly one of IEEE_8087 or IEEE_MC68k should be defined."
147
#endif
148
149
/* The code below assumes that the endianness of integers matches the
150
   endianness of the two 32-bit words of a double.  Check this. */
151
#if defined(WORDS_BIGENDIAN) && defined(DOUBLE_IS_LITTLE_ENDIAN_IEEE754)
152
#error "doubles and ints have incompatible endianness"
153
#endif
154
155
#if !defined(WORDS_BIGENDIAN) && defined(DOUBLE_IS_BIG_ENDIAN_IEEE754)
156
#error "doubles and ints have incompatible endianness"
157
#endif
158
159
160
typedef uint32_t ULong;
161
typedef int32_t Long;
162
typedef uint64_t ULLong;
163
164
#undef DEBUG
165
#ifdef Py_DEBUG
166
#define DEBUG
167
#endif
168
169
/* End Python #define linking */
170
171
#ifdef DEBUG
172
#define Bug(x) {fprintf(stderr, "%s\n", x); exit(1);}
173
#endif
174
175
typedef union { double d; ULong L[2]; } U;
176
177
#ifdef IEEE_8087
178
5.52M
#define word0(x) (x)->L[1]
179
3.82M
#define word1(x) (x)->L[0]
180
#else
181
#define word0(x) (x)->L[0]
182
#define word1(x) (x)->L[1]
183
#endif
184
9.73M
#define dval(x) (x)->d
185
186
#ifndef STRTOD_DIGLIM
187
447k
#define STRTOD_DIGLIM 40
188
#endif
189
190
/* maximum permitted exponent value for strtod; exponents larger than
191
   MAX_ABS_EXP in absolute value get truncated to +-MAX_ABS_EXP.  MAX_ABS_EXP
192
   should fit into an int. */
193
#ifndef MAX_ABS_EXP
194
1.13M
#define MAX_ABS_EXP 1100000000U
195
#endif
196
/* Bound on length of pieces of input strings in _Py_dg_strtod; specifically,
197
   this is used to bound the total number of digits ignoring leading zeros and
198
   the number of digits that follow the decimal point.  Ideally, MAX_DIGITS
199
   should satisfy MAX_DIGITS + 400 < MAX_ABS_EXP; that ensures that the
200
   exponent clipping in _Py_dg_strtod can't affect the value of the output. */
201
#ifndef MAX_DIGITS
202
3.65M
#define MAX_DIGITS 1000000000U
203
#endif
204
205
/* Guard against trying to use the above values on unusual platforms with ints
206
 * of width less than 32 bits. */
207
#if MAX_ABS_EXP > INT_MAX
208
#error "MAX_ABS_EXP should fit in an int"
209
#endif
210
#if MAX_DIGITS > INT_MAX
211
#error "MAX_DIGITS should fit in an int"
212
#endif
213
214
/* The following definition of Storeinc is appropriate for MIPS processors.
215
 * An alternative that might be better on some machines is
216
 * #define Storeinc(a,b,c) (*a++ = b << 16 | c & 0xffff)
217
 */
218
#if defined(IEEE_8087)
219
#define Storeinc(a,b,c) (((unsigned short *)a)[1] = (unsigned short)b,  \
220
                         ((unsigned short *)a)[0] = (unsigned short)c, a++)
221
#else
222
#define Storeinc(a,b,c) (((unsigned short *)a)[0] = (unsigned short)b,  \
223
                         ((unsigned short *)a)[1] = (unsigned short)c, a++)
224
#endif
225
226
/* #define P DBL_MANT_DIG */
227
/* Ten_pmax = floor(P*log(2)/log(5)) */
228
/* Bletch = (highest power of 2 < DBL_MAX_10_EXP) / 16 */
229
/* Quick_max = floor((P-1)*log(FLT_RADIX)/log(10) - 1) */
230
/* Int_max = floor(P*log(FLT_RADIX)/log(10) - 1) */
231
232
582k
#define Exp_shift  20
233
95.4k
#define Exp_shift1 20
234
3.11M
#define Exp_msk1    0x100000
235
#define Exp_msk11   0x100000
236
2.67M
#define Exp_mask  0x7ff00000
237
2.12M
#define P 53
238
#define Nbits 53
239
1.19M
#define Bias 1023
240
#define Emax 1023
241
#define Emin (-1022)
242
1.11M
#define Etiny (-1074)  /* smallest denormal is 2**Etiny */
243
627k
#define Exp_1  0x3ff00000
244
44.2k
#define Exp_11 0x3ff00000
245
1.94M
#define Ebits 11
246
559k
#define Frac_mask  0xfffff
247
46.6k
#define Frac_mask1 0xfffff
248
2.18M
#define Ten_pmax 22
249
83
#define Bletch 0x10
250
179k
#define Bndry_mask  0xfffff
251
8.87k
#define Bndry_mask1 0xfffff
252
75.3k
#define Sign_bit 0x80000000
253
8.09k
#define Log2P 1
254
#define Tiny0 0
255
489k
#define Tiny1 1
256
47.8k
#define Quick_max 14
257
30.3k
#define Int_max 14
258
259
#ifndef Flt_Rounds
260
#ifdef FLT_ROUNDS
261
1.17M
#define Flt_Rounds FLT_ROUNDS
262
#else
263
#define Flt_Rounds 1
264
#endif
265
#endif /*Flt_Rounds*/
266
267
#define Rounding Flt_Rounds
268
269
3.65k
#define Big0 (Frac_mask1 | Exp_msk1*(DBL_MAX_EXP+Bias-1))
270
2.30k
#define Big1 0xffffffff
271
272
/* Bits of the representation of positive infinity. */
273
274
#define POSINF_WORD0 0x7ff00000
275
#define POSINF_WORD1 0
276
277
/* struct BCinfo is used to pass information from _Py_dg_strtod to bigcomp */
278
279
typedef struct BCinfo BCinfo;
280
struct
281
BCinfo {
282
    int e0, nd, nd0, scale;
283
};
284
285
32.2M
#define FFFFFFFF 0xffffffffUL
286
287
/* struct Bigint is used to represent arbitrary-precision integers.  These
288
   integers are stored in sign-magnitude format, with the magnitude stored as
289
   an array of base 2**32 digits.  Bigints are always normalized: if x is a
290
   Bigint then x->wds >= 1, and either x->wds == 1 or x[wds-1] is nonzero.
291
292
   The Bigint fields are as follows:
293
294
     - next is a header used by Balloc and Bfree to keep track of lists
295
         of freed Bigints;  it's also used for the linked list of
296
         powers of 5 of the form 5**2**i used by pow5mult.
297
     - k indicates which pool this Bigint was allocated from
298
     - maxwds is the maximum number of words space was allocated for
299
       (usually maxwds == 2**k)
300
     - sign is 1 for negative Bigints, 0 for positive.  The sign is unused
301
       (ignored on inputs, set to 0 on outputs) in almost all operations
302
       involving Bigints: a notable exception is the diff function, which
303
       ignores signs on inputs but sets the sign of the output correctly.
304
     - wds is the actual number of significant words
305
     - x contains the vector of words (digits) for this Bigint, from least
306
       significant (x[0]) to most significant (x[wds-1]).
307
*/
308
309
// struct Bigint is defined in pycore_dtoa.h.
310
typedef struct Bigint Bigint;
311
312
#if !defined(Py_GIL_DISABLED) && !defined(Py_USING_MEMORY_DEBUGGER)
313
314
/* Memory management: memory is allocated from, and returned to, Kmax+1 pools
315
   of memory, where pool k (0 <= k <= Kmax) is for Bigints b with b->maxwds ==
316
   1 << k.  These pools are maintained as linked lists, with freelist[k]
317
   pointing to the head of the list for pool k.
318
319
   On allocation, if there's no free slot in the appropriate pool, MALLOC is
320
   called to get more memory.  This memory is not returned to the system until
321
   Python quits.  There's also a private memory pool that's allocated from
322
   in preference to using MALLOC.
323
324
   For Bigints with more than (1 << Kmax) digits (which implies at least 1233
325
   decimal digits), memory is directly allocated using MALLOC, and freed using
326
   FREE.
327
328
   XXX: it would be easy to bypass this memory-management system and
329
   translate each call to Balloc into a call to PyMem_Malloc, and each
330
   Bfree to PyMem_Free.  Investigate whether this has any significant
331
   performance on impact. */
332
333
22.1M
#define freelist interp->dtoa.freelist
334
432
#define private_mem interp->dtoa.preallocated
335
1.17k
#define pmem_next interp->dtoa.preallocated_next
336
337
/* Allocate space for a Bigint with up to 1<<k digits */
338
339
static Bigint *
340
Balloc(int k)
341
5.54M
{
342
5.54M
    int x;
343
5.54M
    Bigint *rv;
344
5.54M
    unsigned int len;
345
5.54M
    PyInterpreterState *interp = _PyInterpreterState_GET();
346
347
5.54M
    if (k <= Bigint_Kmax && (rv = freelist[k]))
348
5.54M
        freelist[k] = rv->next;
349
432
    else {
350
432
        x = 1 << k;
351
432
        len = (sizeof(Bigint) + (x-1)*sizeof(ULong) + sizeof(double) - 1)
352
432
            /sizeof(double);
353
432
        if (k <= Bigint_Kmax &&
354
432
            pmem_next - private_mem + len <= (Py_ssize_t)Bigint_PREALLOC_SIZE
355
432
        ) {
356
373
            rv = (Bigint*)pmem_next;
357
373
            pmem_next += len;
358
373
        }
359
59
        else {
360
59
            rv = (Bigint*)MALLOC(len*sizeof(double));
361
59
            if (rv == NULL)
362
0
                return NULL;
363
59
        }
364
432
        rv->k = k;
365
432
        rv->maxwds = x;
366
432
    }
367
5.54M
    rv->sign = rv->wds = 0;
368
5.54M
    return rv;
369
5.54M
}
370
371
/* Free a Bigint allocated with Balloc */
372
373
static void
374
Bfree(Bigint *v)
375
9.66M
{
376
9.66M
    if (v) {
377
5.54M
        if (v->k > Bigint_Kmax)
378
0
            FREE((void*)v);
379
5.54M
        else {
380
5.54M
            PyInterpreterState *interp = _PyInterpreterState_GET();
381
5.54M
            v->next = freelist[v->k];
382
5.54M
            freelist[v->k] = v;
383
5.54M
        }
384
5.54M
    }
385
9.66M
}
386
387
#undef pmem_next
388
#undef private_mem
389
#undef freelist
390
391
#else
392
393
/* Alternative versions of Balloc and Bfree that use PyMem_Malloc and
394
   PyMem_Free directly in place of the custom memory allocation scheme above.
395
   These are provided for the benefit of memory debugging tools like
396
   Valgrind. */
397
398
/* Allocate space for a Bigint with up to 1<<k digits */
399
400
static Bigint *
401
Balloc(int k)
402
{
403
    int x;
404
    Bigint *rv;
405
    unsigned int len;
406
407
    x = 1 << k;
408
    len = (sizeof(Bigint) + (x-1)*sizeof(ULong) + sizeof(double) - 1)
409
        /sizeof(double);
410
411
    rv = (Bigint*)MALLOC(len*sizeof(double));
412
    if (rv == NULL)
413
        return NULL;
414
415
    rv->k = k;
416
    rv->maxwds = x;
417
    rv->sign = rv->wds = 0;
418
    return rv;
419
}
420
421
/* Free a Bigint allocated with Balloc */
422
423
static void
424
Bfree(Bigint *v)
425
{
426
    if (v) {
427
        FREE((void*)v);
428
    }
429
}
430
431
#endif /* !defined(Py_GIL_DISABLED) && !defined(Py_USING_MEMORY_DEBUGGER) */
432
433
491k
#define Bcopy(x,y) memcpy((char *)&x->sign, (char *)&y->sign,   \
434
491k
                          y->wds*sizeof(Long) + 2*sizeof(int))
435
436
/* Multiply a Bigint b by m and add a.  Either modifies b in place and returns
437
   a pointer to the modified b, or Bfrees b and returns a pointer to a copy.
438
   On failure, return NULL.  In this case, b will have been already freed. */
439
440
static Bigint *
441
multadd(Bigint *b, int m, int a)       /* multiply by m and add a */
442
1.22M
{
443
1.22M
    int i, wds;
444
1.22M
    ULong *x;
445
1.22M
    ULLong carry, y;
446
1.22M
    Bigint *b1;
447
448
1.22M
    wds = b->wds;
449
1.22M
    x = b->x;
450
1.22M
    i = 0;
451
1.22M
    carry = a;
452
4.14M
    do {
453
4.14M
        y = *x * (ULLong)m + carry;
454
4.14M
        carry = y >> 32;
455
4.14M
        *x++ = (ULong)(y & FFFFFFFF);
456
4.14M
    }
457
4.14M
    while(++i < wds);
458
1.22M
    if (carry) {
459
74.4k
        if (wds >= b->maxwds) {
460
3.67k
            b1 = Balloc(b->k+1);
461
3.67k
            if (b1 == NULL){
462
0
                Bfree(b);
463
0
                return NULL;
464
0
            }
465
3.67k
            Bcopy(b1, b);
466
3.67k
            Bfree(b);
467
3.67k
            b = b1;
468
3.67k
        }
469
74.4k
        b->x[wds++] = (ULong)carry;
470
74.4k
        b->wds = wds;
471
74.4k
    }
472
1.22M
    return b;
473
1.22M
}
474
475
/* convert a string s containing nd decimal digits (possibly containing a
476
   decimal separator at position nd0, which is ignored) to a Bigint.  This
477
   function carries on where the parsing code in _Py_dg_strtod leaves off: on
478
   entry, y9 contains the result of converting the first 9 digits.  Returns
479
   NULL on failure. */
480
481
static Bigint *
482
s2b(const char *s, int nd0, int nd, ULong y9)
483
447k
{
484
447k
    Bigint *b;
485
447k
    int i, k;
486
447k
    Long x, y;
487
488
447k
    x = (nd + 8) / 9;
489
505k
    for(k = 0, y = 1; x > y; y <<= 1, k++) ;
490
447k
    b = Balloc(k);
491
447k
    if (b == NULL)
492
0
        return NULL;
493
447k
    b->x[0] = y9;
494
447k
    b->wds = 1;
495
496
447k
    if (nd <= 9)
497
401k
      return b;
498
499
45.7k
    s += 9;
500
352k
    for (i = 9; i < nd0; i++) {
501
306k
        b = multadd(b, 10, *s++ - '0');
502
306k
        if (b == NULL)
503
0
            return NULL;
504
306k
    }
505
45.7k
    s++;
506
155k
    for(; i < nd; i++) {
507
110k
        b = multadd(b, 10, *s++ - '0');
508
110k
        if (b == NULL)
509
0
            return NULL;
510
110k
    }
511
45.7k
    return b;
512
45.7k
}
513
514
/* count leading 0 bits in the 32-bit integer x. */
515
516
static int
517
hi0bits(ULong x)
518
670k
{
519
670k
    int k = 0;
520
521
670k
    if (!(x & 0xffff0000)) {
522
249k
        k = 16;
523
249k
        x <<= 16;
524
249k
    }
525
670k
    if (!(x & 0xff000000)) {
526
263k
        k += 8;
527
263k
        x <<= 8;
528
263k
    }
529
670k
    if (!(x & 0xf0000000)) {
530
469k
        k += 4;
531
469k
        x <<= 4;
532
469k
    }
533
670k
    if (!(x & 0xc0000000)) {
534
314k
        k += 2;
535
314k
        x <<= 2;
536
314k
    }
537
670k
    if (!(x & 0x80000000)) {
538
160k
        k++;
539
160k
        if (!(x & 0x40000000))
540
0
            return 32;
541
160k
    }
542
670k
    return k;
543
670k
}
544
545
/* count trailing 0 bits in the 32-bit integer y, and shift y right by that
546
   number of bits. */
547
548
static int
549
lo0bits(ULong *y)
550
47.7k
{
551
47.7k
    int k;
552
47.7k
    ULong x = *y;
553
554
47.7k
    if (x & 7) {
555
29.5k
        if (x & 1)
556
14.2k
            return 0;
557
15.3k
        if (x & 2) {
558
6.69k
            *y = x >> 1;
559
6.69k
            return 1;
560
6.69k
        }
561
8.60k
        *y = x >> 2;
562
8.60k
        return 2;
563
15.3k
    }
564
18.1k
    k = 0;
565
18.1k
    if (!(x & 0xffff)) {
566
6.44k
        k = 16;
567
6.44k
        x >>= 16;
568
6.44k
    }
569
18.1k
    if (!(x & 0xff)) {
570
3.09k
        k += 8;
571
3.09k
        x >>= 8;
572
3.09k
    }
573
18.1k
    if (!(x & 0xf)) {
574
9.66k
        k += 4;
575
9.66k
        x >>= 4;
576
9.66k
    }
577
18.1k
    if (!(x & 0x3)) {
578
9.61k
        k += 2;
579
9.61k
        x >>= 2;
580
9.61k
    }
581
18.1k
    if (!(x & 1)) {
582
13.1k
        k++;
583
13.1k
        x >>= 1;
584
13.1k
        if (!x)
585
0
            return 32;
586
13.1k
    }
587
18.1k
    *y = x;
588
18.1k
    return k;
589
18.1k
}
590
591
/* convert a small nonnegative integer to a Bigint */
592
593
static Bigint *
594
i2b(int i)
595
584k
{
596
584k
    Bigint *b;
597
598
584k
    b = Balloc(1);
599
584k
    if (b == NULL)
600
0
        return NULL;
601
584k
    b->x[0] = i;
602
584k
    b->wds = 1;
603
584k
    return b;
604
584k
}
605
606
/* multiply two Bigints.  Returns a new Bigint, or NULL on failure.  Ignores
607
   the signs of a and b. */
608
609
static Bigint *
610
mult(Bigint *a, Bigint *b)
611
1.64M
{
612
1.64M
    Bigint *c;
613
1.64M
    int k, wa, wb, wc;
614
1.64M
    ULong *x, *xa, *xae, *xb, *xbe, *xc, *xc0;
615
1.64M
    ULong y;
616
1.64M
    ULLong carry, z;
617
618
1.64M
    if ((!a->x[0] && a->wds == 1) || (!b->x[0] && b->wds == 1)) {
619
7.12k
        c = Balloc(0);
620
7.12k
        if (c == NULL)
621
0
            return NULL;
622
7.12k
        c->wds = 1;
623
7.12k
        c->x[0] = 0;
624
7.12k
        return c;
625
7.12k
    }
626
627
1.63M
    if (a->wds < b->wds) {
628
894k
        c = a;
629
894k
        a = b;
630
894k
        b = c;
631
894k
    }
632
1.63M
    k = a->k;
633
1.63M
    wa = a->wds;
634
1.63M
    wb = b->wds;
635
1.63M
    wc = wa + wb;
636
1.63M
    if (wc > a->maxwds)
637
1.05M
        k++;
638
1.63M
    c = Balloc(k);
639
1.63M
    if (c == NULL)
640
0
        return NULL;
641
10.8M
    for(x = c->x, xa = x + wc; x < xa; x++)
642
9.19M
        *x = 0;
643
1.63M
    xa = a->x;
644
1.63M
    xae = xa + wa;
645
1.63M
    xb = b->x;
646
1.63M
    xbe = xb + wb;
647
1.63M
    xc0 = c->x;
648
4.85M
    for(; xb < xbe; xc0++) {
649
3.21M
        if ((y = *xb++)) {
650
3.20M
            x = xa;
651
3.20M
            xc = xc0;
652
3.20M
            carry = 0;
653
19.0M
            do {
654
19.0M
                z = *x++ * (ULLong)y + *xc + carry;
655
19.0M
                carry = z >> 32;
656
19.0M
                *xc++ = (ULong)(z & FFFFFFFF);
657
19.0M
            }
658
19.0M
            while(x < xae);
659
3.20M
            *xc = (ULong)carry;
660
3.20M
        }
661
3.21M
    }
662
2.73M
    for(xc0 = c->x, xc = xc0 + wc; wc > 0 && !*--xc; --wc) ;
663
1.63M
    c->wds = wc;
664
1.63M
    return c;
665
1.63M
}
666
667
#ifndef Py_USING_MEMORY_DEBUGGER
668
669
/* multiply the Bigint b by 5**k.  Returns a pointer to the result, or NULL on
670
   failure; if the returned pointer is distinct from b then the original
671
   Bigint b will have been Bfree'd.   Ignores the sign of b. */
672
673
static Bigint *
674
pow5mult(Bigint *b, int k)
675
534k
{
676
534k
    Bigint *b1, *p5, **p5s;
677
534k
    int i;
678
534k
    static const int p05[3] = { 5, 25, 125 };
679
680
    // For double-to-string conversion, the maximum value of k is limited by
681
    // DBL_MAX_10_EXP (308), the maximum decimal base-10 exponent for binary64.
682
    // For string-to-double conversion, the extreme case is constrained by our
683
    // hardcoded exponent limit before we underflow of -512, adjusted by
684
    // STRTOD_DIGLIM-DBL_DIG-1, giving a maximum of k=535.
685
534k
    assert(0 <= k && k < 1024);
686
687
534k
    if ((i = k & 3)) {
688
180k
        b = multadd(b, p05[i-1], 0);
689
180k
        if (b == NULL)
690
0
            return NULL;
691
180k
    }
692
693
534k
    if (!(k >>= 2))
694
13.3k
        return b;
695
521k
    PyInterpreterState *interp = _PyInterpreterState_GET();
696
521k
    p5s = interp->dtoa.p5s;
697
2.42M
    for(;;) {
698
2.42M
        assert(p5s != interp->dtoa.p5s + Bigint_Pow5size);
699
2.42M
        p5 = *p5s;
700
2.42M
        p5s++;
701
2.42M
        if (k & 1) {
702
1.55M
            b1 = mult(b, p5);
703
1.55M
            Bfree(b);
704
1.55M
            b = b1;
705
1.55M
            if (b == NULL)
706
0
                return NULL;
707
1.55M
        }
708
2.42M
        if (!(k >>= 1))
709
521k
            break;
710
2.42M
    }
711
521k
    return b;
712
521k
}
713
714
#else
715
716
/* Version of pow5mult that doesn't cache powers of 5. Provided for
717
   the benefit of memory debugging tools like Valgrind. */
718
719
static Bigint *
720
pow5mult(Bigint *b, int k)
721
{
722
    Bigint *b1, *p5, *p51;
723
    int i;
724
    static const int p05[3] = { 5, 25, 125 };
725
726
    if ((i = k & 3)) {
727
        b = multadd(b, p05[i-1], 0);
728
        if (b == NULL)
729
            return NULL;
730
    }
731
732
    if (!(k >>= 2))
733
        return b;
734
    p5 = i2b(625);
735
    if (p5 == NULL) {
736
        Bfree(b);
737
        return NULL;
738
    }
739
740
    for(;;) {
741
        if (k & 1) {
742
            b1 = mult(b, p5);
743
            Bfree(b);
744
            b = b1;
745
            if (b == NULL) {
746
                Bfree(p5);
747
                return NULL;
748
            }
749
        }
750
        if (!(k >>= 1))
751
            break;
752
        p51 = mult(p5, p5);
753
        Bfree(p5);
754
        p5 = p51;
755
        if (p5 == NULL) {
756
            Bfree(b);
757
            return NULL;
758
        }
759
    }
760
    Bfree(p5);
761
    return b;
762
}
763
764
#endif /* Py_USING_MEMORY_DEBUGGER */
765
766
/* shift a Bigint b left by k bits.  Return a pointer to the shifted result,
767
   or NULL on failure.  If the returned pointer is distinct from b then the
768
   original b will have been Bfree'd.   Ignores the sign of b. */
769
770
static Bigint *
771
lshift(Bigint *b, int k)
772
1.15M
{
773
1.15M
    int i, k1, n, n1;
774
1.15M
    Bigint *b1;
775
1.15M
    ULong *x, *x1, *xe, z;
776
777
1.15M
    if (!k || (!b->x[0] && b->wds == 1))
778
7.64k
        return b;
779
780
1.15M
    n = k >> 5;
781
1.15M
    k1 = b->k;
782
1.15M
    n1 = n + b->wds + 1;
783
3.12M
    for(i = b->maxwds; n1 > i; i <<= 1)
784
1.97M
        k1++;
785
1.15M
    b1 = Balloc(k1);
786
1.15M
    if (b1 == NULL) {
787
0
        Bfree(b);
788
0
        return NULL;
789
0
    }
790
1.15M
    x1 = b1->x;
791
6.20M
    for(i = 0; i < n; i++)
792
5.05M
        *x1++ = 0;
793
1.15M
    x = b->x;
794
1.15M
    xe = x + b->wds;
795
1.15M
    if (k &= 0x1f) {
796
1.12M
        k1 = 32 - k;
797
1.12M
        z = 0;
798
2.96M
        do {
799
2.96M
            *x1++ = *x << k | z;
800
2.96M
            z = *x++ >> k1;
801
2.96M
        }
802
2.96M
        while(x < xe);
803
1.12M
        if ((*x1 = z))
804
301k
            ++n1;
805
1.12M
    }
806
22.1k
    else do
807
41.2k
             *x1++ = *x++;
808
41.2k
        while(x < xe);
809
1.15M
    b1->wds = n1 - 1;
810
1.15M
    Bfree(b);
811
1.15M
    return b1;
812
1.15M
}
813
814
/* Do a three-way compare of a and b, returning -1 if a < b, 0 if a == b and
815
   1 if a > b.  Ignores signs of a and b. */
816
817
static int
818
cmp(Bigint *a, Bigint *b)
819
1.95M
{
820
1.95M
    ULong *xa, *xa0, *xb, *xb0;
821
1.95M
    int i, j;
822
823
1.95M
    i = a->wds;
824
1.95M
    j = b->wds;
825
#ifdef DEBUG
826
    if (i > 1 && !a->x[i-1])
827
        Bug("cmp called with a->x[a->wds-1] == 0");
828
    if (j > 1 && !b->x[j-1])
829
        Bug("cmp called with b->x[b->wds-1] == 0");
830
#endif
831
1.95M
    if (i -= j)
832
198k
        return i;
833
1.75M
    xa0 = a->x;
834
1.75M
    xa = xa0 + j;
835
1.75M
    xb0 = b->x;
836
1.75M
    xb = xb0 + j;
837
2.60M
    for(;;) {
838
2.60M
        if (*--xa != *--xb)
839
1.73M
            return *xa < *xb ? -1 : 1;
840
861k
        if (xa <= xa0)
841
18.4k
            break;
842
861k
    }
843
18.4k
    return 0;
844
1.75M
}
845
846
/* Take the difference of Bigints a and b, returning a new Bigint.  Returns
847
   NULL on failure.  The signs of a and b are ignored, but the sign of the
848
   result is set appropriately. */
849
850
static Bigint *
851
diff(Bigint *a, Bigint *b)
852
613k
{
853
613k
    Bigint *c;
854
613k
    int i, wa, wb;
855
613k
    ULong *xa, *xae, *xb, *xbe, *xc;
856
613k
    ULLong borrow, y;
857
858
613k
    i = cmp(a,b);
859
613k
    if (!i) {
860
3.07k
        c = Balloc(0);
861
3.07k
        if (c == NULL)
862
0
            return NULL;
863
3.07k
        c->wds = 1;
864
3.07k
        c->x[0] = 0;
865
3.07k
        return c;
866
3.07k
    }
867
610k
    if (i < 0) {
868
88.1k
        c = a;
869
88.1k
        a = b;
870
88.1k
        b = c;
871
88.1k
        i = 1;
872
88.1k
    }
873
521k
    else
874
521k
        i = 0;
875
610k
    c = Balloc(a->k);
876
610k
    if (c == NULL)
877
0
        return NULL;
878
610k
    c->sign = i;
879
610k
    wa = a->wds;
880
610k
    xa = a->x;
881
610k
    xae = xa + wa;
882
610k
    wb = b->wds;
883
610k
    xb = b->x;
884
610k
    xbe = xb + wb;
885
610k
    xc = c->x;
886
610k
    borrow = 0;
887
4.46M
    do {
888
4.46M
        y = (ULLong)*xa++ - *xb++ - borrow;
889
4.46M
        borrow = y >> 32 & (ULong)1;
890
4.46M
        *xc++ = (ULong)(y & FFFFFFFF);
891
4.46M
    }
892
4.46M
    while(xb < xbe);
893
905k
    while(xa < xae) {
894
295k
        y = *xa++ - borrow;
895
295k
        borrow = y >> 32 & (ULong)1;
896
295k
        *xc++ = (ULong)(y & FFFFFFFF);
897
295k
    }
898
1.39M
    while(!*--xc)
899
781k
        wa--;
900
610k
    c->wds = wa;
901
610k
    return c;
902
610k
}
903
904
/* Given a positive normal double x, return the difference between x and the
905
   next double up.  Doesn't give correct results for subnormals. */
906
907
static double
908
ulp(U *x)
909
308k
{
910
308k
    Long L;
911
308k
    U u;
912
913
308k
    L = (word0(x) & Exp_mask) - (P-1)*Exp_msk1;
914
308k
    word0(&u) = L;
915
308k
    word1(&u) = 0;
916
308k
    return dval(&u);
917
308k
}
918
919
/* Convert a Bigint to a double plus an exponent */
920
921
static double
922
b2d(Bigint *a, int *e)
923
606k
{
924
606k
    ULong *xa, *xa0, w, y, z;
925
606k
    int k;
926
606k
    U d;
927
928
606k
    xa0 = a->x;
929
606k
    xa = xa0 + a->wds;
930
606k
    y = *--xa;
931
#ifdef DEBUG
932
    if (!y) Bug("zero y in b2d");
933
#endif
934
606k
    k = hi0bits(y);
935
606k
    *e = 32 - k;
936
606k
    if (k < Ebits) {
937
365k
        word0(&d) = Exp_1 | y >> (Ebits - k);
938
365k
        w = xa > xa0 ? *--xa : 0;
939
365k
        word1(&d) = y << ((32-Ebits) + k) | w >> (Ebits - k);
940
365k
        goto ret_d;
941
365k
    }
942
240k
    z = xa > xa0 ? *--xa : 0;
943
240k
    if (k -= Ebits) {
944
235k
        word0(&d) = Exp_1 | y << k | z >> (32 - k);
945
235k
        y = xa > xa0 ? *--xa : 0;
946
235k
        word1(&d) = z << k | y >> (32 - k);
947
235k
    }
948
5.39k
    else {
949
5.39k
        word0(&d) = Exp_1 | y;
950
5.39k
        word1(&d) = z;
951
5.39k
    }
952
606k
  ret_d:
953
606k
    return dval(&d);
954
240k
}
955
956
/* Convert a scaled double to a Bigint plus an exponent.  Similar to d2b,
957
   except that it accepts the scale parameter used in _Py_dg_strtod (which
958
   should be either 0 or 2*P), and the normalization for the return value is
959
   different (see below).  On input, d should be finite and nonnegative, and d
960
   / 2**scale should be exactly representable as an IEEE 754 double.
961
962
   Returns a Bigint b and an integer e such that
963
964
     dval(d) / 2**scale = b * 2**e.
965
966
   Unlike d2b, b is not necessarily odd: b and e are normalized so
967
   that either 2**(P-1) <= b < 2**P and e >= Etiny, or b < 2**P
968
   and e == Etiny.  This applies equally to an input of 0.0: in that
969
   case the return values are b = 0 and e = Etiny.
970
971
   The above normalization ensures that for all possible inputs d,
972
   2**e gives ulp(d/2**scale).
973
974
   Returns NULL on failure.
975
*/
976
977
static Bigint *
978
sd2b(U *d, int scale, int *e)
979
511k
{
980
511k
    Bigint *b;
981
982
511k
    b = Balloc(1);
983
511k
    if (b == NULL)
984
0
        return NULL;
985
986
    /* First construct b and e assuming that scale == 0. */
987
511k
    b->wds = 2;
988
511k
    b->x[0] = word1(d);
989
511k
    b->x[1] = word0(d) & Frac_mask;
990
511k
    *e = Etiny - 1 + (int)((word0(d) & Exp_mask) >> Exp_shift);
991
511k
    if (*e < Etiny)
992
7.64k
        *e = Etiny;
993
503k
    else
994
503k
        b->x[1] |= Exp_msk1;
995
996
    /* Now adjust for scale, provided that b != 0. */
997
511k
    if (scale && (b->x[0] || b->x[1])) {
998
31.6k
        *e -= scale;
999
31.6k
        if (*e < Etiny) {
1000
27.3k
            scale = Etiny - *e;
1001
27.3k
            *e = Etiny;
1002
            /* We can't shift more than P-1 bits without shifting out a 1. */
1003
27.3k
            assert(0 < scale && scale <= P - 1);
1004
27.3k
            if (scale >= 32) {
1005
                /* The bits shifted out should all be zero. */
1006
15.4k
                assert(b->x[0] == 0);
1007
15.4k
                b->x[0] = b->x[1];
1008
15.4k
                b->x[1] = 0;
1009
15.4k
                scale -= 32;
1010
15.4k
            }
1011
27.3k
            if (scale) {
1012
                /* The bits shifted out should all be zero. */
1013
25.4k
                assert(b->x[0] << (32 - scale) == 0);
1014
25.4k
                b->x[0] = (b->x[0] >> scale) | (b->x[1] << (32 - scale));
1015
25.4k
                b->x[1] >>= scale;
1016
25.4k
            }
1017
27.3k
        }
1018
31.6k
    }
1019
    /* Ensure b is normalized. */
1020
511k
    if (!b->x[1])
1021
27.0k
        b->wds = 1;
1022
1023
511k
    return b;
1024
511k
}
1025
1026
/* Convert a double to a Bigint plus an exponent.  Return NULL on failure.
1027
1028
   Given a finite nonzero double d, return an odd Bigint b and exponent *e
1029
   such that fabs(d) = b * 2**e.  On return, *bbits gives the number of
1030
   significant bits of b; that is, 2**(*bbits-1) <= b < 2**(*bbits).
1031
1032
   If d is zero, then b == 0, *e == -1010, *bbits = 0.
1033
 */
1034
1035
static Bigint *
1036
d2b(U *d, int *e, int *bits)
1037
47.7k
{
1038
47.7k
    Bigint *b;
1039
47.7k
    int de, k;
1040
47.7k
    ULong *x, y, z;
1041
47.7k
    int i;
1042
1043
47.7k
    b = Balloc(1);
1044
47.7k
    if (b == NULL)
1045
0
        return NULL;
1046
47.7k
    x = b->x;
1047
1048
47.7k
    z = word0(d) & Frac_mask;
1049
47.7k
    word0(d) &= 0x7fffffff;   /* clear sign bit, which we ignore */
1050
47.7k
    if ((de = (int)(word0(d) >> Exp_shift)))
1051
44.2k
        z |= Exp_msk1;
1052
47.7k
    if ((y = word1(d))) {
1053
36.1k
        if ((k = lo0bits(&y))) {
1054
22.7k
            x[0] = y | z << (32 - k);
1055
22.7k
            z >>= k;
1056
22.7k
        }
1057
13.4k
        else
1058
13.4k
            x[0] = y;
1059
36.1k
        i =
1060
36.1k
            b->wds = (x[1] = z) ? 2 : 1;
1061
36.1k
    }
1062
11.5k
    else {
1063
11.5k
        k = lo0bits(&z);
1064
11.5k
        x[0] = z;
1065
11.5k
        i =
1066
11.5k
            b->wds = 1;
1067
11.5k
        k += 32;
1068
11.5k
    }
1069
47.7k
    if (de) {
1070
44.2k
        *e = de - Bias - (P-1) + k;
1071
44.2k
        *bits = P - k;
1072
44.2k
    }
1073
3.41k
    else {
1074
3.41k
        *e = de - Bias - (P-1) + 1 + k;
1075
3.41k
        *bits = 32*i - hi0bits(x[i-1]);
1076
3.41k
    }
1077
47.7k
    return b;
1078
47.7k
}
1079
1080
/* Compute the ratio of two Bigints, as a double.  The result may have an
1081
   error of up to 2.5 ulps. */
1082
1083
static double
1084
ratio(Bigint *a, Bigint *b)
1085
303k
{
1086
303k
    U da, db;
1087
303k
    int k, ka, kb;
1088
1089
303k
    dval(&da) = b2d(a, &ka);
1090
303k
    dval(&db) = b2d(b, &kb);
1091
303k
    k = ka - kb + 32*(a->wds - b->wds);
1092
303k
    if (k > 0)
1093
46.5k
        word0(&da) += k*Exp_msk1;
1094
256k
    else {
1095
256k
        k = -k;
1096
256k
        word0(&db) += k*Exp_msk1;
1097
256k
    }
1098
303k
    return dval(&da) / dval(&db);
1099
303k
}
1100
1101
static const double
1102
tens[] = {
1103
    1e0, 1e1, 1e2, 1e3, 1e4, 1e5, 1e6, 1e7, 1e8, 1e9,
1104
    1e10, 1e11, 1e12, 1e13, 1e14, 1e15, 1e16, 1e17, 1e18, 1e19,
1105
    1e20, 1e21, 1e22
1106
};
1107
1108
static const double
1109
bigtens[] = { 1e16, 1e32, 1e64, 1e128, 1e256 };
1110
static const double tinytens[] = { 1e-16, 1e-32, 1e-64, 1e-128,
1111
                                   9007199254740992.*9007199254740992.e-256
1112
                                   /* = 2^106 * 1e-256 */
1113
};
1114
/* The factor of 2^53 in tinytens[4] helps us avoid setting the underflow */
1115
/* flag unnecessarily.  It leads to a song and dance at the end of strtod. */
1116
36.3k
#define Scale_Bit 0x10
1117
41.8k
#define n_bigtens 5
1118
1119
#define ULbits 32
1120
#define kshift 5
1121
61.0k
#define kmask 31
1122
1123
1124
static int
1125
dshift(Bigint *b, int p2)
1126
61.0k
{
1127
61.0k
    int rv = hi0bits(b->x[b->wds-1]) - 4;
1128
61.0k
    if (p2 > 0)
1129
24.7k
        rv -= p2;
1130
61.0k
    return rv & kmask;
1131
61.0k
}
1132
1133
/* special case of Bigint division.  The quotient is always in the range 0 <=
1134
   quotient < 10, and on entry the divisor S is normalized so that its top 4
1135
   bits (28--31) are zero and bit 27 is set. */
1136
1137
static int
1138
quorem(Bigint *b, Bigint *S)
1139
561k
{
1140
561k
    int n;
1141
561k
    ULong *bx, *bxe, q, *sx, *sxe;
1142
561k
    ULLong borrow, carry, y, ys;
1143
1144
561k
    n = S->wds;
1145
#ifdef DEBUG
1146
    /*debug*/ if (b->wds > n)
1147
        /*debug*/       Bug("oversize b in quorem");
1148
#endif
1149
561k
    if (b->wds < n)
1150
16.9k
        return 0;
1151
544k
    sx = S->x;
1152
544k
    sxe = sx + --n;
1153
544k
    bx = b->x;
1154
544k
    bxe = bx + n;
1155
544k
    q = *bxe / (*sxe + 1);      /* ensure q <= true quotient */
1156
#ifdef DEBUG
1157
    /*debug*/ if (q > 9)
1158
        /*debug*/       Bug("oversized quotient in quorem");
1159
#endif
1160
544k
    if (q) {
1161
374k
        borrow = 0;
1162
374k
        carry = 0;
1163
1.96M
        do {
1164
1.96M
            ys = *sx++ * (ULLong)q + carry;
1165
1.96M
            carry = ys >> 32;
1166
1.96M
            y = *bx - (ys & FFFFFFFF) - borrow;
1167
1.96M
            borrow = y >> 32 & (ULong)1;
1168
1.96M
            *bx++ = (ULong)(y & FFFFFFFF);
1169
1.96M
        }
1170
1.96M
        while(sx <= sxe);
1171
374k
        if (!*bxe) {
1172
1.94k
            bx = b->x;
1173
1.94k
            while(--bxe > bx && !*bxe)
1174
0
                --n;
1175
1.94k
            b->wds = n;
1176
1.94k
        }
1177
374k
    }
1178
544k
    if (cmp(b, S) >= 0) {
1179
31.4k
        q++;
1180
31.4k
        borrow = 0;
1181
31.4k
        carry = 0;
1182
31.4k
        bx = b->x;
1183
31.4k
        sx = S->x;
1184
179k
        do {
1185
179k
            ys = *sx++ + carry;
1186
179k
            carry = ys >> 32;
1187
179k
            y = *bx - (ys & FFFFFFFF) - borrow;
1188
179k
            borrow = y >> 32 & (ULong)1;
1189
179k
            *bx++ = (ULong)(y & FFFFFFFF);
1190
179k
        }
1191
179k
        while(sx <= sxe);
1192
31.4k
        bx = b->x;
1193
31.4k
        bxe = bx + n;
1194
31.4k
        if (!*bxe) {
1195
23.3k
            while(--bxe > bx && !*bxe)
1196
1.69k
                --n;
1197
21.6k
            b->wds = n;
1198
21.6k
        }
1199
31.4k
    }
1200
544k
    return q;
1201
561k
}
1202
1203
/* sulp(x) is a version of ulp(x) that takes bc.scale into account.
1204
1205
   Assuming that x is finite and nonnegative (positive zero is fine
1206
   here) and x / 2^bc.scale is exactly representable as a double,
1207
   sulp(x) is equivalent to 2^bc.scale * ulp(x / 2^bc.scale). */
1208
1209
static double
1210
sulp(U *x, BCinfo *bc)
1211
5.09k
{
1212
5.09k
    U u;
1213
1214
5.09k
    if (bc->scale && 2*P + 1 > (int)((word0(x) & Exp_mask) >> Exp_shift)) {
1215
        /* rv/2^bc->scale is subnormal */
1216
267
        word0(&u) = (P+2)*Exp_msk1;
1217
267
        word1(&u) = 0;
1218
267
        return u.d;
1219
267
    }
1220
4.82k
    else {
1221
4.82k
        assert(word0(x) || word1(x)); /* x != 0.0 */
1222
4.82k
        return ulp(x);
1223
4.82k
    }
1224
5.09k
}
1225
1226
/* The bigcomp function handles some hard cases for strtod, for inputs
1227
   with more than STRTOD_DIGLIM digits.  It's called once an initial
1228
   estimate for the double corresponding to the input string has
1229
   already been obtained by the code in _Py_dg_strtod.
1230
1231
   The bigcomp function is only called after _Py_dg_strtod has found a
1232
   double value rv such that either rv or rv + 1ulp represents the
1233
   correctly rounded value corresponding to the original string.  It
1234
   determines which of these two values is the correct one by
1235
   computing the decimal digits of rv + 0.5ulp and comparing them with
1236
   the corresponding digits of s0.
1237
1238
   In the following, write dv for the absolute value of the number represented
1239
   by the input string.
1240
1241
   Inputs:
1242
1243
     s0 points to the first significant digit of the input string.
1244
1245
     rv is a (possibly scaled) estimate for the closest double value to the
1246
        value represented by the original input to _Py_dg_strtod.  If
1247
        bc->scale is nonzero, then rv/2^(bc->scale) is the approximation to
1248
        the input value.
1249
1250
     bc is a struct containing information gathered during the parsing and
1251
        estimation steps of _Py_dg_strtod.  Description of fields follows:
1252
1253
        bc->e0 gives the exponent of the input value, such that dv = (integer
1254
           given by the bd->nd digits of s0) * 10**e0
1255
1256
        bc->nd gives the total number of significant digits of s0.  It will
1257
           be at least 1.
1258
1259
        bc->nd0 gives the number of significant digits of s0 before the
1260
           decimal separator.  If there's no decimal separator, bc->nd0 ==
1261
           bc->nd.
1262
1263
        bc->scale is the value used to scale rv to avoid doing arithmetic with
1264
           subnormal values.  It's either 0 or 2*P (=106).
1265
1266
   Outputs:
1267
1268
     On successful exit, rv/2^(bc->scale) is the closest double to dv.
1269
1270
     Returns 0 on success, -1 on failure (e.g., due to a failed malloc call). */
1271
1272
static int
1273
bigcomp(U *rv, const char *s0, BCinfo *bc)
1274
24.5k
{
1275
24.5k
    Bigint *b, *d;
1276
24.5k
    int b2, d2, dd, i, nd, nd0, odd, p2, p5;
1277
1278
24.5k
    nd = bc->nd;
1279
24.5k
    nd0 = bc->nd0;
1280
24.5k
    p5 = nd + bc->e0;
1281
24.5k
    b = sd2b(rv, bc->scale, &p2);
1282
24.5k
    if (b == NULL)
1283
0
        return -1;
1284
1285
    /* record whether the lsb of rv/2^(bc->scale) is odd:  in the exact halfway
1286
       case, this is used for round to even. */
1287
24.5k
    odd = b->x[0] & 1;
1288
1289
    /* left shift b by 1 bit and or a 1 into the least significant bit;
1290
       this gives us b * 2**p2 = rv/2^(bc->scale) + 0.5 ulp. */
1291
24.5k
    b = lshift(b, 1);
1292
24.5k
    if (b == NULL)
1293
0
        return -1;
1294
24.5k
    b->x[0] |= 1;
1295
24.5k
    p2--;
1296
1297
24.5k
    p2 -= p5;
1298
24.5k
    d = i2b(1);
1299
24.5k
    if (d == NULL) {
1300
0
        Bfree(b);
1301
0
        return -1;
1302
0
    }
1303
    /* Arrange for convenient computation of quotients:
1304
     * shift left if necessary so divisor has 4 leading 0 bits.
1305
     */
1306
24.5k
    if (p5 > 0) {
1307
21.5k
        d = pow5mult(d, p5);
1308
21.5k
        if (d == NULL) {
1309
0
            Bfree(b);
1310
0
            return -1;
1311
0
        }
1312
21.5k
    }
1313
2.94k
    else if (p5 < 0) {
1314
1.49k
        b = pow5mult(b, -p5);
1315
1.49k
        if (b == NULL) {
1316
0
            Bfree(d);
1317
0
            return -1;
1318
0
        }
1319
1.49k
    }
1320
24.5k
    if (p2 > 0) {
1321
19.6k
        b2 = p2;
1322
19.6k
        d2 = 0;
1323
19.6k
    }
1324
4.89k
    else {
1325
4.89k
        b2 = 0;
1326
4.89k
        d2 = -p2;
1327
4.89k
    }
1328
24.5k
    i = dshift(d, d2);
1329
24.5k
    if ((b2 += i) > 0) {
1330
24.2k
        b = lshift(b, b2);
1331
24.2k
        if (b == NULL) {
1332
0
            Bfree(d);
1333
0
            return -1;
1334
0
        }
1335
24.2k
    }
1336
24.5k
    if ((d2 += i) > 0) {
1337
23.3k
        d = lshift(d, d2);
1338
23.3k
        if (d == NULL) {
1339
0
            Bfree(b);
1340
0
            return -1;
1341
0
        }
1342
23.3k
    }
1343
1344
    /* Compare s0 with b/d: set dd to -1, 0, or 1 according as s0 < b/d, s0 ==
1345
     * b/d, or s0 > b/d.  Here the digits of s0 are thought of as representing
1346
     * a number in the range [0.1, 1). */
1347
24.5k
    if (cmp(b, d) >= 0)
1348
        /* b/d >= 1 */
1349
845
        dd = -1;
1350
23.6k
    else {
1351
23.6k
        i = 0;
1352
435k
        for(;;) {
1353
435k
            b = multadd(b, 10, 0);
1354
435k
            if (b == NULL) {
1355
0
                Bfree(d);
1356
0
                return -1;
1357
0
            }
1358
435k
            dd = s0[i < nd0 ? i : i+1] - '0' - quorem(b, d);
1359
435k
            i++;
1360
1361
435k
            if (dd)
1362
21.4k
                break;
1363
413k
            if (!b->x[0] && b->wds == 1) {
1364
                /* b/d == 0 */
1365
1.28k
                dd = i < nd;
1366
1.28k
                break;
1367
1.28k
            }
1368
412k
            if (!(i < nd)) {
1369
                /* b/d != 0, but digits of s0 exhausted */
1370
928
                dd = -1;
1371
928
                break;
1372
928
            }
1373
412k
        }
1374
23.6k
    }
1375
24.5k
    Bfree(b);
1376
24.5k
    Bfree(d);
1377
24.5k
    if (dd > 0 || (dd == 0 && odd))
1378
2.78k
        dval(rv) += sulp(rv, bc);
1379
24.5k
    return 0;
1380
24.5k
}
1381
1382
1383
double
1384
_Py_dg_strtod(const char *s00, char **se)
1385
1.27M
{
1386
1.27M
    int bb2, bb5, bbe, bd2, bd5, bs2, c, dsign, e, e1, error;
1387
1.27M
    int esign, i, j, k, lz, nd, nd0, odd, sign;
1388
1.27M
    const char *s, *s0, *s1;
1389
1.27M
    double aadj, aadj1;
1390
1.27M
    U aadj2, adj, rv, rv0;
1391
1.27M
    ULong y, z, abs_exp;
1392
1.27M
    Long L;
1393
1.27M
    BCinfo bc;
1394
1.27M
    Bigint *bb = NULL, *bd = NULL, *bd0 = NULL, *bs = NULL, *delta = NULL;
1395
1.27M
    size_t ndigits, fraclen;
1396
1.27M
    double result;
1397
1398
1.27M
    dval(&rv) = 0.;
1399
1400
    /* Start parsing. */
1401
1.27M
    c = *(s = s00);
1402
1403
    /* Parse optional sign, if present. */
1404
1.27M
    sign = 0;
1405
1.27M
    switch (c) {
1406
641k
    case '-':
1407
641k
        sign = 1;
1408
641k
        _Py_FALLTHROUGH;
1409
641k
    case '+':
1410
641k
        c = *++s;
1411
1.27M
    }
1412
1413
    /* Skip leading zeros: lz is true iff there were leading zeros. */
1414
1.27M
    s1 = s;
1415
1.30M
    while (c == '0')
1416
28.8k
        c = *++s;
1417
1.27M
    lz = s != s1;
1418
1419
    /* Point s0 at the first nonzero digit (if any).  fraclen will be the
1420
       number of digits between the decimal point and the end of the
1421
       digit string.  ndigits will be the total number of digits ignoring
1422
       leading zeros. */
1423
1.27M
    s0 = s1 = s;
1424
7.97M
    while ('0' <= c && c <= '9')
1425
6.70M
        c = *++s;
1426
1.27M
    ndigits = s - s1;
1427
1.27M
    fraclen = 0;
1428
1429
    /* Parse decimal point and following digits. */
1430
1.27M
    if (c == '.') {
1431
88.3k
        c = *++s;
1432
88.3k
        if (!ndigits) {
1433
34.2k
            s1 = s;
1434
1.18M
            while (c == '0')
1435
1.15M
                c = *++s;
1436
34.2k
            lz = lz || s != s1;
1437
34.2k
            fraclen += (s - s1);
1438
34.2k
            s0 = s;
1439
34.2k
        }
1440
88.3k
        s1 = s;
1441
32.4M
        while ('0' <= c && c <= '9')
1442
32.3M
            c = *++s;
1443
88.3k
        ndigits += s - s1;
1444
88.3k
        fraclen += s - s1;
1445
88.3k
    }
1446
1447
    /* Now lz is true if and only if there were leading zero digits, and
1448
       ndigits gives the total number of digits ignoring leading zeros.  A
1449
       valid input must have at least one digit. */
1450
1.27M
    if (!ndigits && !lz) {
1451
55.9k
        if (se)
1452
55.9k
            *se = (char *)s00;
1453
55.9k
        goto parse_error;
1454
55.9k
    }
1455
1456
    /* Range check ndigits and fraclen to make sure that they, and values
1457
       computed with them, can safely fit in an int. */
1458
1.21M
    if (ndigits > MAX_DIGITS || fraclen > MAX_DIGITS) {
1459
0
        if (se)
1460
0
            *se = (char *)s00;
1461
0
        goto parse_error;
1462
0
    }
1463
1.21M
    nd = (int)ndigits;
1464
1.21M
    nd0 = (int)ndigits - (int)fraclen;
1465
1466
    /* Parse exponent. */
1467
1.21M
    e = 0;
1468
1.21M
    if (c == 'e' || c == 'E') {
1469
1.13M
        s00 = s;
1470
1.13M
        c = *++s;
1471
1472
        /* Exponent sign. */
1473
1.13M
        esign = 0;
1474
1.13M
        switch (c) {
1475
42.6k
        case '-':
1476
42.6k
            esign = 1;
1477
42.6k
            _Py_FALLTHROUGH;
1478
60.4k
        case '+':
1479
60.4k
            c = *++s;
1480
1.13M
        }
1481
1482
        /* Skip zeros.  lz is true iff there are leading zeros. */
1483
1.13M
        s1 = s;
1484
1.37M
        while (c == '0')
1485
239k
            c = *++s;
1486
1.13M
        lz = s != s1;
1487
1488
        /* Get absolute value of the exponent. */
1489
1.13M
        s1 = s;
1490
1.13M
        abs_exp = 0;
1491
14.8M
        while ('0' <= c && c <= '9') {
1492
13.7M
            abs_exp = 10*abs_exp + (c - '0');
1493
13.7M
            c = *++s;
1494
13.7M
        }
1495
1496
        /* abs_exp will be correct modulo 2**32.  But 10**9 < 2**32, so if
1497
           there are at most 9 significant exponent digits then overflow is
1498
           impossible. */
1499
1.13M
        if (s - s1 > 9 || abs_exp > MAX_ABS_EXP)
1500
10.5k
            e = (int)MAX_ABS_EXP;
1501
1.12M
        else
1502
1.12M
            e = (int)abs_exp;
1503
1.13M
        if (esign)
1504
42.6k
            e = -e;
1505
1506
        /* A valid exponent must have at least one digit. */
1507
1.13M
        if (s == s1 && !lz)
1508
0
            s = s00;
1509
1.13M
    }
1510
1511
    /* Adjust exponent to take into account position of the point. */
1512
1.21M
    e -= nd - nd0;
1513
1.21M
    if (nd0 <= 0)
1514
39.9k
        nd0 = nd;
1515
1516
    /* Finished parsing.  Set se to indicate how far we parsed */
1517
1.21M
    if (se)
1518
1.21M
        *se = (char *)s;
1519
1520
    /* If all digits were zero, exit with return value +-0.0.  Otherwise,
1521
       strip trailing zeros: scan back until we hit a nonzero digit. */
1522
1.21M
    if (!nd)
1523
15.7k
        goto ret;
1524
10.5M
    for (i = nd; i > 0; ) {
1525
10.5M
        --i;
1526
10.5M
        if (s0[i < nd0 ? i : i+1] != '0') {
1527
1.20M
            ++i;
1528
1.20M
            break;
1529
1.20M
        }
1530
10.5M
    }
1531
1.20M
    e += nd - i;
1532
1.20M
    nd = i;
1533
1.20M
    if (nd0 > nd)
1534
23.2k
        nd0 = nd;
1535
1536
    /* Summary of parsing results.  After parsing, and dealing with zero
1537
     * inputs, we have values s0, nd0, nd, e, sign, where:
1538
     *
1539
     *  - s0 points to the first significant digit of the input string
1540
     *
1541
     *  - nd is the total number of significant digits (here, and
1542
     *    below, 'significant digits' means the set of digits of the
1543
     *    significand of the input that remain after ignoring leading
1544
     *    and trailing zeros).
1545
     *
1546
     *  - nd0 indicates the position of the decimal point, if present; it
1547
     *    satisfies 1 <= nd0 <= nd.  The nd significant digits are in
1548
     *    s0[0:nd0] and s0[nd0+1:nd+1] using the usual Python half-open slice
1549
     *    notation.  (If nd0 < nd, then s0[nd0] contains a '.'  character; if
1550
     *    nd0 == nd, then s0[nd0] could be any non-digit character.)
1551
     *
1552
     *  - e is the adjusted exponent: the absolute value of the number
1553
     *    represented by the original input string is n * 10**e, where
1554
     *    n is the integer represented by the concatenation of
1555
     *    s0[0:nd0] and s0[nd0+1:nd+1]
1556
     *
1557
     *  - sign gives the sign of the input:  1 for negative, 0 for positive
1558
     *
1559
     *  - the first and last significant digits are nonzero
1560
     */
1561
1562
    /* put first DBL_DIG+1 digits into integer y and z.
1563
     *
1564
     *  - y contains the value represented by the first min(9, nd)
1565
     *    significant digits
1566
     *
1567
     *  - if nd > 9, z contains the value represented by significant digits
1568
     *    with indices in [9, min(16, nd)).  So y * 10**(min(16, nd) - 9) + z
1569
     *    gives the value represented by the first min(16, nd) sig. digits.
1570
     */
1571
1572
1.20M
    bc.e0 = e1 = e;
1573
1.20M
    y = z = 0;
1574
4.71M
    for (i = 0; i < nd; i++) {
1575
3.56M
        if (i < 9)
1576
3.06M
            y = 10*y + s0[i < nd0 ? i : i+1] - '0';
1577
505k
        else if (i < DBL_DIG+1)
1578
453k
            z = 10*z + s0[i < nd0 ? i : i+1] - '0';
1579
51.7k
        else
1580
51.7k
            break;
1581
3.56M
    }
1582
1583
1.20M
    k = nd < DBL_DIG + 1 ? nd : DBL_DIG + 1;
1584
1.20M
    dval(&rv) = y;
1585
1.20M
    if (k > 9) {
1586
79.9k
        dval(&rv) = tens[k - 9] * dval(&rv) + z;
1587
79.9k
    }
1588
1.20M
    if (nd <= DBL_DIG
1589
1.14M
        && Flt_Rounds == 1
1590
1.20M
        ) {
1591
1.14M
        if (!e)
1592
12.5k
            goto ret;
1593
1.12M
        if (e > 0) {
1594
1.06M
            if (e <= Ten_pmax) {
1595
37.4k
                dval(&rv) *= tens[e];
1596
37.4k
                goto ret;
1597
37.4k
            }
1598
1.02M
            i = DBL_DIG - nd;
1599
1.02M
            if (e <= Ten_pmax + i) {
1600
                /* A fancier test would sometimes let us do
1601
                 * this for larger i values.
1602
                 */
1603
3.81k
                e -= i;
1604
3.81k
                dval(&rv) *= tens[i];
1605
3.81k
                dval(&rv) *= tens[e];
1606
3.81k
                goto ret;
1607
3.81k
            }
1608
1.02M
        }
1609
66.8k
        else if (e >= -Ten_pmax) {
1610
33.2k
            dval(&rv) /= tens[-e];
1611
33.2k
            goto ret;
1612
33.2k
        }
1613
1.12M
    }
1614
1.11M
    e1 += nd - k;
1615
1616
1.11M
    bc.scale = 0;
1617
1618
    /* Get starting approximation = rv * 10**e1 */
1619
1620
1.11M
    if (e1 > 0) {
1621
1.05M
        if ((i = e1 & 15))
1622
929k
            dval(&rv) *= tens[i];
1623
1.05M
        if (e1 &= ~15) {
1624
1.04M
            if (e1 > DBL_MAX_10_EXP)
1625
658k
                goto ovfl;
1626
389k
            e1 >>= 4;
1627
1.01M
            for(j = 0; e1 > 1; j++, e1 >>= 1)
1628
622k
                if (e1 & 1)
1629
338k
                    dval(&rv) *= bigtens[j];
1630
            /* The last multiplication could overflow. */
1631
389k
            word0(&rv) -= P*Exp_msk1;
1632
389k
            dval(&rv) *= bigtens[j];
1633
389k
            if ((z = word0(&rv) & Exp_mask)
1634
389k
                > Exp_msk1*(DBL_MAX_EXP+Bias-P))
1635
2.11k
                goto ovfl;
1636
387k
            if (z > Exp_msk1*(DBL_MAX_EXP+Bias-1-P)) {
1637
                /* set to largest number */
1638
                /* (Can't trust DBL_MAX) */
1639
571
                word0(&rv) = Big0;
1640
571
                word1(&rv) = Big1;
1641
571
            }
1642
386k
            else
1643
386k
                word0(&rv) += P*Exp_msk1;
1644
387k
        }
1645
1.05M
    }
1646
54.9k
    else if (e1 < 0) {
1647
        /* The input decimal value lies in [10**e1, 10**(e1+16)).
1648
1649
           If e1 <= -512, underflow immediately.
1650
           If e1 <= -256, set bc.scale to 2*P.
1651
1652
           So for input value < 1e-256, bc.scale is always set;
1653
           for input value >= 1e-240, bc.scale is never set.
1654
           For input values in [1e-256, 1e-240), bc.scale may or may
1655
           not be set. */
1656
1657
51.5k
        e1 = -e1;
1658
51.5k
        if ((i = e1 & 15))
1659
39.6k
            dval(&rv) /= tens[i];
1660
51.5k
        if (e1 >>= 4) {
1661
41.8k
            if (e1 >= 1 << n_bigtens)
1662
5.56k
                goto undfl;
1663
36.3k
            if (e1 & Scale_Bit)
1664
20.7k
                bc.scale = 2*P;
1665
172k
            for(j = 0; e1 > 0; j++, e1 >>= 1)
1666
136k
                if (e1 & 1)
1667
82.7k
                    dval(&rv) *= tinytens[j];
1668
36.3k
            if (bc.scale && (j = 2*P + 1 - ((word0(&rv) & Exp_mask)
1669
20.7k
                                            >> Exp_shift)) > 0) {
1670
                /* scaled rv is denormal; clear j low bits */
1671
18.8k
                if (j >= 32) {
1672
12.3k
                    word1(&rv) = 0;
1673
12.3k
                    if (j >= 53)
1674
7.72k
                        word0(&rv) = (P+2)*Exp_msk1;
1675
4.63k
                    else
1676
4.63k
                        word0(&rv) &= 0xffffffff << (j-32);
1677
12.3k
                }
1678
6.46k
                else
1679
6.46k
                    word1(&rv) &= 0xffffffff << j;
1680
18.8k
            }
1681
36.3k
            if (!dval(&rv))
1682
0
                goto undfl;
1683
36.3k
        }
1684
51.5k
    }
1685
1686
    /* Now the hard part -- adjusting rv to the correct value.*/
1687
1688
    /* Put digits into bd: true value = bd * 10^e */
1689
1690
447k
    bc.nd = nd;
1691
447k
    bc.nd0 = nd0;       /* Only needed if nd > STRTOD_DIGLIM, but done here */
1692
                        /* to silence an erroneous warning about bc.nd0 */
1693
                        /* possibly not being initialized. */
1694
447k
    if (nd > STRTOD_DIGLIM) {
1695
        /* ASSERT(STRTOD_DIGLIM >= 18); 18 == one more than the */
1696
        /* minimum number of decimal digits to distinguish double values */
1697
        /* in IEEE arithmetic. */
1698
1699
        /* Truncate input to 18 significant digits, then discard any trailing
1700
           zeros on the result by updating nd, nd0, e and y suitably. (There's
1701
           no need to update z; it's not reused beyond this point.) */
1702
251k
        for (i = 18; i > 0; ) {
1703
            /* scan back until we hit a nonzero digit.  significant digit 'i'
1704
            is s0[i] if i < nd0, s0[i+1] if i >= nd0. */
1705
251k
            --i;
1706
251k
            if (s0[i < nd0 ? i : i+1] != '0') {
1707
31.3k
                ++i;
1708
31.3k
                break;
1709
31.3k
            }
1710
251k
        }
1711
31.3k
        e += nd - i;
1712
31.3k
        nd = i;
1713
31.3k
        if (nd0 > nd)
1714
27.0k
            nd0 = nd;
1715
31.3k
        if (nd < 9) { /* must recompute y */
1716
19.1k
            y = 0;
1717
146k
            for(i = 0; i < nd0; ++i)
1718
127k
                y = 10*y + s0[i] - '0';
1719
26.5k
            for(; i < nd; ++i)
1720
7.43k
                y = 10*y + s0[i+1] - '0';
1721
19.1k
        }
1722
31.3k
    }
1723
447k
    bd0 = s2b(s0, nd0, nd, y);
1724
447k
    if (bd0 == NULL)
1725
0
        goto failed_malloc;
1726
1727
    /* Notation for the comments below.  Write:
1728
1729
         - dv for the absolute value of the number represented by the original
1730
           decimal input string.
1731
1732
         - if we've truncated dv, write tdv for the truncated value.
1733
           Otherwise, set tdv == dv.
1734
1735
         - srv for the quantity rv/2^bc.scale; so srv is the current binary
1736
           approximation to tdv (and dv).  It should be exactly representable
1737
           in an IEEE 754 double.
1738
    */
1739
1740
486k
    for(;;) {
1741
1742
        /* This is the main correction loop for _Py_dg_strtod.
1743
1744
           We've got a decimal value tdv, and a floating-point approximation
1745
           srv=rv/2^bc.scale to tdv.  The aim is to determine whether srv is
1746
           close enough (i.e., within 0.5 ulps) to tdv, and to compute a new
1747
           approximation if not.
1748
1749
           To determine whether srv is close enough to tdv, compute integers
1750
           bd, bb and bs proportional to tdv, srv and 0.5 ulp(srv)
1751
           respectively, and then use integer arithmetic to determine whether
1752
           |tdv - srv| is less than, equal to, or greater than 0.5 ulp(srv).
1753
        */
1754
1755
486k
        bd = Balloc(bd0->k);
1756
486k
        if (bd == NULL) {
1757
0
            goto failed_malloc;
1758
0
        }
1759
486k
        Bcopy(bd, bd0);
1760
486k
        bb = sd2b(&rv, bc.scale, &bbe);   /* srv = bb * 2^bbe */
1761
486k
        if (bb == NULL) {
1762
0
            goto failed_malloc;
1763
0
        }
1764
        /* Record whether lsb of bb is odd, in case we need this
1765
           for the round-to-even step later. */
1766
486k
        odd = bb->x[0] & 1;
1767
1768
        /* tdv = bd * 10**e;  srv = bb * 2**bbe */
1769
486k
        bs = i2b(1);
1770
486k
        if (bs == NULL) {
1771
0
            goto failed_malloc;
1772
0
        }
1773
1774
486k
        if (e >= 0) {
1775
418k
            bb2 = bb5 = 0;
1776
418k
            bd2 = bd5 = e;
1777
418k
        }
1778
68.5k
        else {
1779
68.5k
            bb2 = bb5 = -e;
1780
68.5k
            bd2 = bd5 = 0;
1781
68.5k
        }
1782
486k
        if (bbe >= 0)
1783
420k
            bb2 += bbe;
1784
66.3k
        else
1785
66.3k
            bd2 -= bbe;
1786
486k
        bs2 = bb2;
1787
486k
        bb2++;
1788
486k
        bd2++;
1789
1790
        /* At this stage bd5 - bb5 == e == bd2 - bb2 + bbe, bb2 - bs2 == 1,
1791
           and bs == 1, so:
1792
1793
              tdv == bd * 10**e = bd * 2**(bbe - bb2 + bd2) * 5**(bd5 - bb5)
1794
              srv == bb * 2**bbe = bb * 2**(bbe - bb2 + bb2)
1795
              0.5 ulp(srv) == 2**(bbe-1) = bs * 2**(bbe - bb2 + bs2)
1796
1797
           It follows that:
1798
1799
              M * tdv = bd * 2**bd2 * 5**bd5
1800
              M * srv = bb * 2**bb2 * 5**bb5
1801
              M * 0.5 ulp(srv) = bs * 2**bs2 * 5**bb5
1802
1803
           for some constant M.  (Actually, M == 2**(bb2 - bbe) * 5**bb5, but
1804
           this fact is not needed below.)
1805
        */
1806
1807
        /* Remove factor of 2**i, where i = min(bb2, bd2, bs2). */
1808
486k
        i = bb2 < bd2 ? bb2 : bd2;
1809
486k
        if (i > bs2)
1810
65.1k
            i = bs2;
1811
486k
        if (i > 0) {
1812
485k
            bb2 -= i;
1813
485k
            bd2 -= i;
1814
485k
            bs2 -= i;
1815
485k
        }
1816
1817
        /* Scale bb, bd, bs by the appropriate powers of 2 and 5. */
1818
486k
        if (bb5 > 0) {
1819
68.5k
            bs = pow5mult(bs, bb5);
1820
68.5k
            if (bs == NULL) {
1821
0
                goto failed_malloc;
1822
0
            }
1823
68.5k
            Bigint *bb1 = mult(bs, bb);
1824
68.5k
            Bfree(bb);
1825
68.5k
            bb = bb1;
1826
68.5k
            if (bb == NULL) {
1827
0
                goto failed_malloc;
1828
0
            }
1829
68.5k
        }
1830
486k
        if (bb2 > 0) {
1831
486k
            bb = lshift(bb, bb2);
1832
486k
            if (bb == NULL) {
1833
0
                goto failed_malloc;
1834
0
            }
1835
486k
        }
1836
486k
        if (bd5 > 0) {
1837
408k
            bd = pow5mult(bd, bd5);
1838
408k
            if (bd == NULL) {
1839
0
                goto failed_malloc;
1840
0
            }
1841
408k
        }
1842
486k
        if (bd2 > 0) {
1843
65.1k
            bd = lshift(bd, bd2);
1844
65.1k
            if (bd == NULL) {
1845
0
                goto failed_malloc;
1846
0
            }
1847
65.1k
        }
1848
486k
        if (bs2 > 0) {
1849
416k
            bs = lshift(bs, bs2);
1850
416k
            if (bs == NULL) {
1851
0
                goto failed_malloc;
1852
0
            }
1853
416k
        }
1854
1855
        /* Now bd, bb and bs are scaled versions of tdv, srv and 0.5 ulp(srv),
1856
           respectively.  Compute the difference |tdv - srv|, and compare
1857
           with 0.5 ulp(srv). */
1858
1859
486k
        delta = diff(bb, bd);
1860
486k
        if (delta == NULL) {
1861
0
            goto failed_malloc;
1862
0
        }
1863
486k
        dsign = delta->sign;
1864
486k
        delta->sign = 0;
1865
486k
        i = cmp(delta, bs);
1866
486k
        if (bc.nd > nd && i <= 0) {
1867
31.3k
            if (dsign)
1868
22.9k
                break;  /* Must use bigcomp(). */
1869
1870
            /* Here rv overestimates the truncated decimal value by at most
1871
               0.5 ulp(rv).  Hence rv either overestimates the true decimal
1872
               value by <= 0.5 ulp(rv), or underestimates it by some small
1873
               amount (< 0.1 ulp(rv)); either way, rv is within 0.5 ulps of
1874
               the true decimal value, so it's possible to exit.
1875
1876
               Exception: if scaled rv is a normal exact power of 2, but not
1877
               DBL_MIN, then rv - 0.5 ulp(rv) takes us all the way down to the
1878
               next double, so the correctly rounded result is either rv - 0.5
1879
               ulp(rv) or rv; in this case, use bigcomp to distinguish. */
1880
1881
8.34k
            if (!word1(&rv) && !(word0(&rv) & Bndry_mask)) {
1882
                /* rv can't be 0, since it's an overestimate for some
1883
                   nonzero value.  So rv is a normal power of 2. */
1884
1.82k
                j = (int)(word0(&rv) & Exp_mask) >> Exp_shift;
1885
                /* rv / 2^bc.scale = 2^(j - 1023 - bc.scale); use bigcomp if
1886
                   rv / 2^bc.scale >= 2^-1021. */
1887
1.82k
                if (j - bc.scale >= 2) {
1888
1.56k
                    dval(&rv) -= 0.5 * sulp(&rv, &bc);
1889
1.56k
                    break; /* Use bigcomp. */
1890
1.56k
                }
1891
1.82k
            }
1892
1893
6.78k
            {
1894
6.78k
                bc.nd = nd;
1895
6.78k
                i = -1; /* Discarded digits make delta smaller. */
1896
6.78k
            }
1897
6.78k
        }
1898
1899
462k
        if (i < 0) {
1900
            /* Error is less than half an ulp -- check for
1901
             * special case of mantissa a power of two.
1902
             */
1903
153k
            if (dsign || word1(&rv) || word0(&rv) & Bndry_mask
1904
6.95k
                || (word0(&rv) & Exp_mask) <= (2*P+1)*Exp_msk1
1905
153k
                ) {
1906
148k
                break;
1907
148k
            }
1908
5.56k
            if (!delta->x[0] && delta->wds <= 1) {
1909
                /* exact result */
1910
709
                break;
1911
709
            }
1912
4.85k
            delta = lshift(delta,Log2P);
1913
4.85k
            if (delta == NULL) {
1914
0
                goto failed_malloc;
1915
0
            }
1916
4.85k
            if (cmp(delta, bs) > 0)
1917
1.44k
                goto drop_down;
1918
3.41k
            break;
1919
4.85k
        }
1920
308k
        if (i == 0) {
1921
            /* exactly half-way between */
1922
5.19k
            if (dsign) {
1923
2.47k
                if ((word0(&rv) & Bndry_mask1) == Bndry_mask1
1924
842
                    &&  word1(&rv) == (
1925
842
                        (bc.scale &&
1926
0
                         (y = word0(&rv) & Exp_mask) <= 2*P*Exp_msk1) ?
1927
0
                        (0xffffffff & (0xffffffff << (2*P+1-(y>>Exp_shift)))) :
1928
842
                        0xffffffff)) {
1929
                    /*boundary case -- increment exponent*/
1930
515
                    word0(&rv) = (word0(&rv) & Exp_mask)
1931
515
                        + Exp_msk1
1932
515
                        ;
1933
515
                    word1(&rv) = 0;
1934
                    /* dsign = 0; */
1935
515
                    break;
1936
515
                }
1937
2.47k
            }
1938
2.71k
            else if (!(word0(&rv) & Bndry_mask) && !word1(&rv)) {
1939
1.44k
              drop_down:
1940
                /* boundary case -- decrement exponent */
1941
1.44k
                if (bc.scale) {
1942
0
                    L = word0(&rv) & Exp_mask;
1943
0
                    if (L <= (2*P+1)*Exp_msk1) {
1944
0
                        if (L > (P+2)*Exp_msk1)
1945
                            /* round even ==> */
1946
                            /* accept rv */
1947
0
                            break;
1948
                        /* rv = smallest denormal */
1949
0
                        if (bc.nd > nd)
1950
0
                            break;
1951
0
                        goto undfl;
1952
0
                    }
1953
0
                }
1954
1.44k
                L = (word0(&rv) & Exp_mask) - Exp_msk1;
1955
1.44k
                word0(&rv) = L | Bndry_mask1;
1956
1.44k
                word1(&rv) = 0xffffffff;
1957
1.44k
                break;
1958
1.44k
            }
1959
4.67k
            if (!odd)
1960
3.93k
                break;
1961
745
            if (dsign)
1962
600
                dval(&rv) += sulp(&rv, &bc);
1963
145
            else {
1964
145
                dval(&rv) -= sulp(&rv, &bc);
1965
145
                if (!dval(&rv)) {
1966
0
                    if (bc.nd >nd)
1967
0
                        break;
1968
0
                    goto undfl;
1969
0
                }
1970
145
            }
1971
            /* dsign = 1 - dsign; */
1972
745
            break;
1973
745
        }
1974
303k
        if ((aadj = ratio(delta, bs)) <= 2.) {
1975
272k
            if (dsign)
1976
20.3k
                aadj = aadj1 = 1.;
1977
252k
            else if (word1(&rv) || word0(&rv) & Bndry_mask) {
1978
244k
                if (word1(&rv) == Tiny1 && !word0(&rv)) {
1979
0
                    if (bc.nd >nd)
1980
0
                        break;
1981
0
                    goto undfl;
1982
0
                }
1983
244k
                aadj = 1.;
1984
244k
                aadj1 = -1.;
1985
244k
            }
1986
7.12k
            else {
1987
                /* special case -- power of FLT_RADIX to be */
1988
                /* rounded down... */
1989
1990
7.12k
                if (aadj < 2./FLT_RADIX)
1991
0
                    aadj = 1./FLT_RADIX;
1992
7.12k
                else
1993
7.12k
                    aadj *= 0.5;
1994
7.12k
                aadj1 = -aadj;
1995
7.12k
            }
1996
272k
        }
1997
30.8k
        else {
1998
30.8k
            aadj *= 0.5;
1999
30.8k
            aadj1 = dsign ? aadj : -aadj;
2000
30.8k
            if (Flt_Rounds == 0)
2001
0
                aadj1 += 0.5;
2002
30.8k
        }
2003
303k
        y = word0(&rv) & Exp_mask;
2004
2005
        /* Check for overflow */
2006
2007
303k
        if (y == Exp_msk1*(DBL_MAX_EXP+Bias-1)) {
2008
6.13k
            dval(&rv0) = dval(&rv);
2009
6.13k
            word0(&rv) -= P*Exp_msk1;
2010
6.13k
            adj.d = aadj1 * ulp(&rv);
2011
6.13k
            dval(&rv) += adj.d;
2012
6.13k
            if ((word0(&rv) & Exp_mask) >=
2013
6.13k
                Exp_msk1*(DBL_MAX_EXP+Bias-P)) {
2014
1.34k
                if (word0(&rv0) == Big0 && word1(&rv0) == Big1) {
2015
959
                    goto ovfl;
2016
959
                }
2017
388
                word0(&rv) = Big0;
2018
388
                word1(&rv) = Big1;
2019
388
                goto cont;
2020
1.34k
            }
2021
4.78k
            else
2022
4.78k
                word0(&rv) += P*Exp_msk1;
2023
6.13k
        }
2024
297k
        else {
2025
297k
            if (bc.scale && y <= 2*P*Exp_msk1) {
2026
15.9k
                if (aadj <= 0x7fffffff) {
2027
15.9k
                    if ((z = (ULong)aadj) <= 0)
2028
1.81k
                        z = 1;
2029
15.9k
                    aadj = z;
2030
15.9k
                    aadj1 = dsign ? aadj : -aadj;
2031
15.9k
                }
2032
15.9k
                dval(&aadj2) = aadj1;
2033
15.9k
                word0(&aadj2) += (2*P+1)*Exp_msk1 - y;
2034
15.9k
                aadj1 = dval(&aadj2);
2035
15.9k
            }
2036
297k
            adj.d = aadj1 * ulp(&rv);
2037
297k
            dval(&rv) += adj.d;
2038
297k
        }
2039
301k
        z = word0(&rv) & Exp_mask;
2040
301k
        if (bc.nd == nd) {
2041
283k
            if (!bc.scale)
2042
266k
                if (y == z) {
2043
                    /* Can we stop now? */
2044
265k
                    L = (Long)aadj;
2045
265k
                    aadj -= L;
2046
                    /* The tolerances below are conservative. */
2047
265k
                    if (dsign || word1(&rv) || word0(&rv) & Bndry_mask) {
2048
265k
                        if (aadj < .4999999 || aadj > .5000001)
2049
262k
                            break;
2050
265k
                    }
2051
23
                    else if (aadj < .4999999/FLT_RADIX)
2052
23
                        break;
2053
265k
                }
2054
283k
        }
2055
39.3k
      cont:
2056
39.3k
        Bfree(bb); bb = NULL;
2057
39.3k
        Bfree(bd); bd = NULL;
2058
39.3k
        Bfree(bs); bs = NULL;
2059
39.3k
        Bfree(delta); delta = NULL;
2060
39.3k
    }
2061
446k
    if (bc.nd > nd) {
2062
24.5k
        error = bigcomp(&rv, s0, &bc);
2063
24.5k
        if (error)
2064
0
            goto failed_malloc;
2065
24.5k
    }
2066
2067
446k
    if (bc.scale) {
2068
20.7k
        word0(&rv0) = Exp_1 - 2*P*Exp_msk1;
2069
20.7k
        word1(&rv0) = 0;
2070
20.7k
        dval(&rv) *= dval(&rv0);
2071
20.7k
    }
2072
2073
549k
  ret:
2074
549k
    result = sign ? -dval(&rv) : dval(&rv);
2075
549k
    goto done;
2076
2077
55.9k
  parse_error:
2078
55.9k
    result = 0.0;
2079
55.9k
    goto done;
2080
2081
0
  failed_malloc:
2082
0
    errno = ENOMEM;
2083
0
    result = -1.0;
2084
0
    goto done;
2085
2086
5.56k
  undfl:
2087
5.56k
    result = sign ? -0.0 : 0.0;
2088
5.56k
    goto done;
2089
2090
661k
  ovfl:
2091
661k
    errno = ERANGE;
2092
    /* Can't trust HUGE_VAL */
2093
661k
    word0(&rv) = Exp_mask;
2094
661k
    word1(&rv) = 0;
2095
661k
    result = sign ? -dval(&rv) : dval(&rv);
2096
661k
    goto done;
2097
2098
1.27M
  done:
2099
1.27M
    Bfree(bb);
2100
1.27M
    Bfree(bd);
2101
1.27M
    Bfree(bs);
2102
1.27M
    Bfree(bd0);
2103
1.27M
    Bfree(delta);
2104
1.27M
    return result;
2105
2106
446k
}
2107
2108
static char *
2109
rv_alloc(int i)
2110
54.7k
{
2111
54.7k
    int j, k, *r;
2112
2113
54.7k
    j = sizeof(ULong);
2114
54.7k
    for(k = 0;
2115
54.7k
        sizeof(Bigint) - sizeof(ULong) - sizeof(int) + j <= (unsigned)i;
2116
54.7k
        j <<= 1)
2117
0
        k++;
2118
54.7k
    r = (int*)Balloc(k);
2119
54.7k
    if (r == NULL)
2120
0
        return NULL;
2121
54.7k
    *r = k;
2122
54.7k
    return (char *)(r+1);
2123
54.7k
}
2124
2125
static char *
2126
nrv_alloc(const char *s, char **rve, int n)
2127
7.01k
{
2128
7.01k
    char *rv, *t;
2129
2130
7.01k
    rv = rv_alloc(n);
2131
7.01k
    if (rv == NULL)
2132
0
        return NULL;
2133
7.01k
    t = rv;
2134
18.6k
    while((*t = *s++)) t++;
2135
7.01k
    if (rve)
2136
7.01k
        *rve = t;
2137
7.01k
    return rv;
2138
7.01k
}
2139
2140
/* freedtoa(s) must be used to free values s returned by dtoa
2141
 * when MULTIPLE_THREADS is #defined.  It should be used in all cases,
2142
 * but for consistency with earlier versions of dtoa, it is optional
2143
 * when MULTIPLE_THREADS is not defined.
2144
 */
2145
2146
void
2147
_Py_dg_freedtoa(char *s)
2148
54.7k
{
2149
54.7k
    Bigint *b = (Bigint *)((int *)s - 1);
2150
54.7k
    b->maxwds = 1 << (b->k = *(int*)b);
2151
54.7k
    Bfree(b);
2152
54.7k
}
2153
2154
/* dtoa for IEEE arithmetic (dmg): convert double to ASCII string.
2155
 *
2156
 * Inspired by "How to Print Floating-Point Numbers Accurately" by
2157
 * Guy L. Steele, Jr. and Jon L. White [Proc. ACM SIGPLAN '90, pp. 112-126].
2158
 *
2159
 * Modifications:
2160
 *      1. Rather than iterating, we use a simple numeric overestimate
2161
 *         to determine k = floor(log10(d)).  We scale relevant
2162
 *         quantities using O(log2(k)) rather than O(k) multiplications.
2163
 *      2. For some modes > 2 (corresponding to ecvt and fcvt), we don't
2164
 *         try to generate digits strictly left to right.  Instead, we
2165
 *         compute with fewer bits and propagate the carry if necessary
2166
 *         when rounding the final digit up.  This is often faster.
2167
 *      3. Under the assumption that input will be rounded nearest,
2168
 *         mode 0 renders 1e23 as 1e23 rather than 9.999999999999999e22.
2169
 *         That is, we allow equality in stopping tests when the
2170
 *         round-nearest rule will give the same floating-point value
2171
 *         as would satisfaction of the stopping test with strict
2172
 *         inequality.
2173
 *      4. We remove common factors of powers of 2 from relevant
2174
 *         quantities.
2175
 *      5. When converting floating-point integers less than 1e16,
2176
 *         we use floating-point arithmetic rather than resorting
2177
 *         to multiple-precision integers.
2178
 *      6. When asked to produce fewer than 15 digits, we first try
2179
 *         to get by with floating-point arithmetic; we resort to
2180
 *         multiple-precision integer arithmetic only if we cannot
2181
 *         guarantee that the floating-point calculation has given
2182
 *         the correctly rounded result.  For k requested digits and
2183
 *         "uniformly" distributed input, the probability is
2184
 *         something like 10^(k-15) that we must resort to the Long
2185
 *         calculation.
2186
 */
2187
2188
/* Additional notes (METD): (1) returns NULL on failure.  (2) to avoid memory
2189
   leakage, a successful call to _Py_dg_dtoa should always be matched by a
2190
   call to _Py_dg_freedtoa. */
2191
2192
char *
2193
_Py_dg_dtoa(double dd, int mode, int ndigits,
2194
            int *decpt, int *sign, char **rve)
2195
54.7k
{
2196
    /*  Arguments ndigits, decpt, sign are similar to those
2197
        of ecvt and fcvt; trailing zeros are suppressed from
2198
        the returned string.  If not null, *rve is set to point
2199
        to the end of the return value.  If d is +-Infinity or NaN,
2200
        then *decpt is set to 9999.
2201
2202
        mode:
2203
        0 ==> shortest string that yields d when read in
2204
        and rounded to nearest.
2205
        1 ==> like 0, but with Steele & White stopping rule;
2206
        e.g. with IEEE P754 arithmetic , mode 0 gives
2207
        1e23 whereas mode 1 gives 9.999999999999999e22.
2208
        2 ==> max(1,ndigits) significant digits.  This gives a
2209
        return value similar to that of ecvt, except
2210
        that trailing zeros are suppressed.
2211
        3 ==> through ndigits past the decimal point.  This
2212
        gives a return value similar to that from fcvt,
2213
        except that trailing zeros are suppressed, and
2214
        ndigits can be negative.
2215
        4,5 ==> similar to 2 and 3, respectively, but (in
2216
        round-nearest mode) with the tests of mode 0 to
2217
        possibly return a shorter string that rounds to d.
2218
        With IEEE arithmetic and compilation with
2219
        -DHonor_FLT_ROUNDS, modes 4 and 5 behave the same
2220
        as modes 2 and 3 when FLT_ROUNDS != 1.
2221
        6-9 ==> Debugging modes similar to mode - 4:  don't try
2222
        fast floating-point estimate (if applicable).
2223
2224
        Values of mode other than 0-9 are treated as mode 0.
2225
2226
        Sufficient space is allocated to the return value
2227
        to hold the suppressed trailing zeros.
2228
    */
2229
2230
54.7k
    int bbits, b2, b5, be, dig, i, ieps, ilim, ilim0, ilim1,
2231
54.7k
        j, j1, k, k0, k_check, leftright, m2, m5, s2, s5,
2232
54.7k
        spec_case, try_quick;
2233
54.7k
    Long L;
2234
54.7k
    int denorm;
2235
54.7k
    ULong x;
2236
54.7k
    Bigint *b, *b1, *delta, *mlo, *mhi, *S;
2237
54.7k
    U d2, eps, u;
2238
54.7k
    double ds;
2239
54.7k
    char *s, *s0;
2240
2241
    /* set pointers to NULL, to silence gcc compiler warnings and make
2242
       cleanup easier on error */
2243
54.7k
    mlo = mhi = S = 0;
2244
54.7k
    s0 = 0;
2245
2246
54.7k
    u.d = dd;
2247
54.7k
    if (word0(&u) & Sign_bit) {
2248
        /* set sign for everything, including 0's and NaNs */
2249
20.6k
        *sign = 1;
2250
20.6k
        word0(&u) &= ~Sign_bit; /* clear sign bit */
2251
20.6k
    }
2252
34.0k
    else
2253
34.0k
        *sign = 0;
2254
2255
    /* quick return for Infinities, NaNs and zeros */
2256
54.7k
    if ((word0(&u) & Exp_mask) == Exp_mask)
2257
664
    {
2258
        /* Infinity or NaN */
2259
664
        *decpt = 9999;
2260
664
        if (!word1(&u) && !(word0(&u) & 0xfffff))
2261
664
            return nrv_alloc("Infinity", rve, 8);
2262
0
        return nrv_alloc("NaN", rve, 3);
2263
664
    }
2264
54.0k
    if (!dval(&u)) {
2265
6.34k
        *decpt = 1;
2266
6.34k
        return nrv_alloc("0", rve, 1);
2267
6.34k
    }
2268
2269
    /* compute k = floor(log10(d)).  The computation may leave k
2270
       one too large, but should never leave k too small. */
2271
47.7k
    b = d2b(&u, &be, &bbits);
2272
47.7k
    if (b == NULL)
2273
0
        goto failed_malloc;
2274
47.7k
    if ((i = (int)(word0(&u) >> Exp_shift1 & (Exp_mask>>Exp_shift1)))) {
2275
44.2k
        dval(&d2) = dval(&u);
2276
44.2k
        word0(&d2) &= Frac_mask1;
2277
44.2k
        word0(&d2) |= Exp_11;
2278
2279
        /* log(x)       ~=~ log(1.5) + (x-1.5)/1.5
2280
         * log10(x)      =  log(x) / log(10)
2281
         *              ~=~ log(1.5)/log(10) + (x-1.5)/(1.5*log(10))
2282
         * log10(d) = (i-Bias)*log(2)/log(10) + log10(d2)
2283
         *
2284
         * This suggests computing an approximation k to log10(d) by
2285
         *
2286
         * k = (i - Bias)*0.301029995663981
2287
         *      + ( (d2-1.5)*0.289529654602168 + 0.176091259055681 );
2288
         *
2289
         * We want k to be too large rather than too small.
2290
         * The error in the first-order Taylor series approximation
2291
         * is in our favor, so we just round up the constant enough
2292
         * to compensate for any error in the multiplication of
2293
         * (i - Bias) by 0.301029995663981; since |i - Bias| <= 1077,
2294
         * and 1077 * 0.30103 * 2^-52 ~=~ 7.2e-14,
2295
         * adding 1e-13 to the constant term more than suffices.
2296
         * Hence we adjust the constant term to 0.1760912590558.
2297
         * (We could get a more accurate k by invoking log10,
2298
         *  but this is probably not worthwhile.)
2299
         */
2300
2301
44.2k
        i -= Bias;
2302
44.2k
        denorm = 0;
2303
44.2k
    }
2304
3.41k
    else {
2305
        /* d is denormalized */
2306
2307
3.41k
        i = bbits + be + (Bias + (P-1) - 1);
2308
3.41k
        x = i > 32  ? word0(&u) << (64 - i) | word1(&u) >> (i - 32)
2309
3.41k
            : word1(&u) << (32 - i);
2310
3.41k
        dval(&d2) = x;
2311
3.41k
        word0(&d2) -= 31*Exp_msk1; /* adjust exponent */
2312
3.41k
        i -= (Bias + (P-1) - 1) + 1;
2313
3.41k
        denorm = 1;
2314
3.41k
    }
2315
47.7k
    ds = (dval(&d2)-1.5)*0.289529654602168 + 0.1760912590558 +
2316
47.7k
        i*0.301029995663981;
2317
47.7k
    k = (int)ds;
2318
47.7k
    if (ds < 0. && ds != k)
2319
13.0k
        k--;    /* want k = floor(ds) */
2320
47.7k
    k_check = 1;
2321
47.7k
    if (k >= 0 && k <= Ten_pmax) {
2322
18.3k
        if (dval(&u) < tens[k])
2323
2.02k
            k--;
2324
18.3k
        k_check = 0;
2325
18.3k
    }
2326
47.7k
    j = bbits - i - 1;
2327
47.7k
    if (j >= 0) {
2328
18.6k
        b2 = 0;
2329
18.6k
        s2 = j;
2330
18.6k
    }
2331
29.0k
    else {
2332
29.0k
        b2 = -j;
2333
29.0k
        s2 = 0;
2334
29.0k
    }
2335
47.7k
    if (k >= 0) {
2336
33.7k
        b5 = 0;
2337
33.7k
        s5 = k;
2338
33.7k
        s2 += k;
2339
33.7k
    }
2340
13.9k
    else {
2341
13.9k
        b2 -= k;
2342
13.9k
        b5 = -k;
2343
13.9k
        s5 = 0;
2344
13.9k
    }
2345
47.7k
    if (mode < 0 || mode > 9)
2346
0
        mode = 0;
2347
2348
47.7k
    try_quick = 1;
2349
2350
47.7k
    if (mode > 5) {
2351
0
        mode -= 4;
2352
0
        try_quick = 0;
2353
0
    }
2354
47.7k
    leftright = 1;
2355
47.7k
    ilim = ilim1 = -1;  /* Values for cases 0 and 1; done here to */
2356
    /* silence erroneous "gcc -Wall" warning. */
2357
47.7k
    switch(mode) {
2358
47.6k
    case 0:
2359
47.6k
    case 1:
2360
47.6k
        i = 18;
2361
47.6k
        ndigits = 0;
2362
47.6k
        break;
2363
0
    case 2:
2364
0
        leftright = 0;
2365
0
        _Py_FALLTHROUGH;
2366
0
    case 4:
2367
0
        if (ndigits <= 0)
2368
0
            ndigits = 1;
2369
0
        ilim = ilim1 = i = ndigits;
2370
0
        break;
2371
104
    case 3:
2372
104
        leftright = 0;
2373
104
        _Py_FALLTHROUGH;
2374
104
    case 5:
2375
104
        i = ndigits + k + 1;
2376
104
        ilim = i;
2377
104
        ilim1 = i - 1;
2378
104
        if (i <= 0)
2379
0
            i = 1;
2380
47.7k
    }
2381
47.7k
    s0 = rv_alloc(i);
2382
47.7k
    if (s0 == NULL)
2383
0
        goto failed_malloc;
2384
47.7k
    s = s0;
2385
2386
2387
47.7k
    if (ilim >= 0 && ilim <= Quick_max && try_quick) {
2388
2389
        /* Try to get by with floating-point arithmetic. */
2390
2391
104
        i = 0;
2392
104
        dval(&d2) = dval(&u);
2393
104
        k0 = k;
2394
104
        ilim0 = ilim;
2395
104
        ieps = 2; /* conservative */
2396
104
        if (k > 0) {
2397
83
            ds = tens[k&0xf];
2398
83
            j = k >> 4;
2399
83
            if (j & Bletch) {
2400
                /* prevent overflows */
2401
0
                j &= Bletch - 1;
2402
0
                dval(&u) /= bigtens[n_bigtens-1];
2403
0
                ieps++;
2404
0
            }
2405
83
            for(; j; j >>= 1, i++)
2406
0
                if (j & 1) {
2407
0
                    ieps++;
2408
0
                    ds *= bigtens[i];
2409
0
                }
2410
83
            dval(&u) /= ds;
2411
83
        }
2412
21
        else if ((j1 = -k)) {
2413
0
            dval(&u) *= tens[j1 & 0xf];
2414
0
            for(j = j1 >> 4; j; j >>= 1, i++)
2415
0
                if (j & 1) {
2416
0
                    ieps++;
2417
0
                    dval(&u) *= bigtens[i];
2418
0
                }
2419
0
        }
2420
104
        if (k_check && dval(&u) < 1. && ilim > 0) {
2421
0
            if (ilim1 <= 0)
2422
0
                goto fast_failed;
2423
0
            ilim = ilim1;
2424
0
            k--;
2425
0
            dval(&u) *= 10.;
2426
0
            ieps++;
2427
0
        }
2428
104
        dval(&eps) = ieps*dval(&u) + 7.;
2429
104
        word0(&eps) -= (P-1)*Exp_msk1;
2430
104
        if (ilim == 0) {
2431
0
            S = mhi = 0;
2432
0
            dval(&u) -= 5.;
2433
0
            if (dval(&u) > dval(&eps))
2434
0
                goto one_digit;
2435
0
            if (dval(&u) < -dval(&eps))
2436
0
                goto no_digits;
2437
0
            goto fast_failed;
2438
0
        }
2439
104
        if (leftright) {
2440
            /* Use Steele & White method of only
2441
             * generating digits needed.
2442
             */
2443
0
            dval(&eps) = 0.5/tens[ilim-1] - dval(&eps);
2444
0
            for(i = 0;;) {
2445
0
                L = (Long)dval(&u);
2446
0
                dval(&u) -= L;
2447
0
                *s++ = '0' + (int)L;
2448
0
                if (dval(&u) < dval(&eps))
2449
0
                    goto ret1;
2450
0
                if (1. - dval(&u) < dval(&eps))
2451
0
                    goto bump_up;
2452
0
                if (++i >= ilim)
2453
0
                    break;
2454
0
                dval(&eps) *= 10.;
2455
0
                dval(&u) *= 10.;
2456
0
            }
2457
0
        }
2458
104
        else {
2459
            /* Generate ilim digits, then fix them up. */
2460
104
            dval(&eps) *= tens[ilim-1];
2461
284
            for(i = 1;; i++, dval(&u) *= 10.) {
2462
284
                L = (Long)(dval(&u));
2463
284
                if (!(dval(&u) -= L))
2464
13
                    ilim = i;
2465
284
                *s++ = '0' + (int)L;
2466
284
                if (i == ilim) {
2467
104
                    if (dval(&u) > 0.5 + dval(&eps))
2468
53
                        goto bump_up;
2469
51
                    else if (dval(&u) < 0.5 - dval(&eps)) {
2470
60
                        while(*--s == '0');
2471
51
                        s++;
2472
51
                        goto ret1;
2473
51
                    }
2474
0
                    break;
2475
104
                }
2476
284
            }
2477
104
        }
2478
0
      fast_failed:
2479
0
        s = s0;
2480
0
        dval(&u) = dval(&d2);
2481
0
        k = k0;
2482
0
        ilim = ilim0;
2483
0
    }
2484
2485
    /* Do we have a "small" integer? */
2486
2487
47.6k
    if (be >= 0 && k <= Int_max) {
2488
        /* Yes. */
2489
11.1k
        ds = tens[k];
2490
11.1k
        if (ndigits < 0 && ilim <= 0) {
2491
0
            S = mhi = 0;
2492
0
            if (ilim < 0 || dval(&u) <= 5*ds)
2493
0
                goto no_digits;
2494
0
            goto one_digit;
2495
0
        }
2496
15.7k
        for(i = 1;; i++, dval(&u) *= 10.) {
2497
15.7k
            L = (Long)(dval(&u) / ds);
2498
15.7k
            dval(&u) -= L*ds;
2499
15.7k
            *s++ = '0' + (int)L;
2500
15.7k
            if (!dval(&u)) {
2501
11.1k
                break;
2502
11.1k
            }
2503
4.64k
            if (i == ilim) {
2504
0
                dval(&u) += dval(&u);
2505
0
                if (dval(&u) > ds || (dval(&u) == ds && L & 1)) {
2506
53
                  bump_up:
2507
56
                    while(*--s == '9')
2508
3
                        if (s == s0) {
2509
0
                            k++;
2510
0
                            *s = '0';
2511
0
                            break;
2512
0
                        }
2513
53
                    ++*s++;
2514
53
                }
2515
0
                else {
2516
                    /* Strip trailing zeros. This branch was missing from the
2517
                       original dtoa.c, leading to surplus trailing zeros in
2518
                       some cases. See bugs.python.org/issue40780. */
2519
0
                    while (s > s0 && s[-1] == '0') {
2520
0
                        --s;
2521
0
                    }
2522
0
                }
2523
53
                break;
2524
0
            }
2525
4.64k
        }
2526
11.1k
        goto ret1;
2527
11.1k
    }
2528
2529
36.4k
    m2 = b2;
2530
36.4k
    m5 = b5;
2531
36.4k
    if (leftright) {
2532
36.4k
        i =
2533
36.4k
            denorm ? be + (Bias + (P-1) - 1 + 1) :
2534
36.4k
            1 + P - bbits;
2535
36.4k
        b2 += i;
2536
36.4k
        s2 += i;
2537
36.4k
        mhi = i2b(1);
2538
36.4k
        if (mhi == NULL)
2539
0
            goto failed_malloc;
2540
36.4k
    }
2541
36.4k
    if (m2 > 0 && s2 > 0) {
2542
33.1k
        i = m2 < s2 ? m2 : s2;
2543
33.1k
        b2 -= i;
2544
33.1k
        m2 -= i;
2545
33.1k
        s2 -= i;
2546
33.1k
    }
2547
36.4k
    if (b5 > 0) {
2548
13.9k
        if (leftright) {
2549
13.9k
            if (m5 > 0) {
2550
13.9k
                mhi = pow5mult(mhi, m5);
2551
13.9k
                if (mhi == NULL)
2552
0
                    goto failed_malloc;
2553
13.9k
                b1 = mult(mhi, b);
2554
13.9k
                Bfree(b);
2555
13.9k
                b = b1;
2556
13.9k
                if (b == NULL)
2557
0
                    goto failed_malloc;
2558
13.9k
            }
2559
13.9k
            if ((j = b5 - m5)) {
2560
0
                b = pow5mult(b, j);
2561
0
                if (b == NULL)
2562
0
                    goto failed_malloc;
2563
0
            }
2564
13.9k
        }
2565
0
        else {
2566
0
            b = pow5mult(b, b5);
2567
0
            if (b == NULL)
2568
0
                goto failed_malloc;
2569
0
        }
2570
13.9k
    }
2571
36.4k
    S = i2b(1);
2572
36.4k
    if (S == NULL)
2573
0
        goto failed_malloc;
2574
36.4k
    if (s5 > 0) {
2575
20.7k
        S = pow5mult(S, s5);
2576
20.7k
        if (S == NULL)
2577
0
            goto failed_malloc;
2578
20.7k
    }
2579
2580
    /* Check for special case that d is a normalized power of 2. */
2581
2582
36.4k
    spec_case = 0;
2583
36.4k
    if ((mode < 2 || leftright)
2584
36.4k
        ) {
2585
36.4k
        if (!word1(&u) && !(word0(&u) & Bndry_mask)
2586
1.36k
            && word0(&u) & (Exp_mask & ~Exp_msk1)
2587
36.4k
            ) {
2588
            /* The special case */
2589
1.08k
            b2 += Log2P;
2590
1.08k
            s2 += Log2P;
2591
1.08k
            spec_case = 1;
2592
1.08k
        }
2593
36.4k
    }
2594
2595
    /* Arrange for convenient computation of quotients:
2596
     * shift left if necessary so divisor has 4 leading 0 bits.
2597
     *
2598
     * Perhaps we should just compute leading 28 bits of S once
2599
     * and for all and pass them and a shift to quorem, so it
2600
     * can do shifts and ors to compute the numerator for q.
2601
     */
2602
36.4k
#define iInc 28
2603
36.4k
    i = dshift(S, s2);
2604
36.4k
    b2 += i;
2605
36.4k
    m2 += i;
2606
36.4k
    s2 += i;
2607
36.4k
    if (b2 > 0) {
2608
36.4k
        b = lshift(b, b2);
2609
36.4k
        if (b == NULL)
2610
0
            goto failed_malloc;
2611
36.4k
    }
2612
36.4k
    if (s2 > 0) {
2613
36.0k
        S = lshift(S, s2);
2614
36.0k
        if (S == NULL)
2615
0
            goto failed_malloc;
2616
36.0k
    }
2617
36.4k
    if (k_check) {
2618
29.3k
        if (cmp(b,S) < 0) {
2619
4.47k
            k--;
2620
4.47k
            b = multadd(b, 10, 0);      /* we botched the k estimate */
2621
4.47k
            if (b == NULL)
2622
0
                goto failed_malloc;
2623
4.47k
            if (leftright) {
2624
4.47k
                mhi = multadd(mhi, 10, 0);
2625
4.47k
                if (mhi == NULL)
2626
0
                    goto failed_malloc;
2627
4.47k
            }
2628
4.47k
            ilim = ilim1;
2629
4.47k
        }
2630
29.3k
    }
2631
36.4k
    if (ilim <= 0 && (mode == 3 || mode == 5)) {
2632
0
        if (ilim < 0) {
2633
            /* no digits, fcvt style */
2634
0
          no_digits:
2635
0
            k = -1 - ndigits;
2636
0
            goto ret;
2637
0
        }
2638
0
        else {
2639
0
            S = multadd(S, 5, 0);
2640
0
            if (S == NULL)
2641
0
                goto failed_malloc;
2642
0
            if (cmp(b, S) <= 0)
2643
0
                goto no_digits;
2644
0
        }
2645
0
      one_digit:
2646
0
        *s++ = '1';
2647
0
        k++;
2648
0
        goto ret;
2649
0
    }
2650
36.4k
    if (leftright) {
2651
36.4k
        if (m2 > 0) {
2652
35.7k
            mhi = lshift(mhi, m2);
2653
35.7k
            if (mhi == NULL)
2654
0
                goto failed_malloc;
2655
35.7k
        }
2656
2657
        /* Compute mlo -- check for special case
2658
         * that d is a normalized power of 2.
2659
         */
2660
2661
36.4k
        mlo = mhi;
2662
36.4k
        if (spec_case) {
2663
1.08k
            mhi = Balloc(mhi->k);
2664
1.08k
            if (mhi == NULL)
2665
0
                goto failed_malloc;
2666
1.08k
            Bcopy(mhi, mlo);
2667
1.08k
            mhi = lshift(mhi, Log2P);
2668
1.08k
            if (mhi == NULL)
2669
0
                goto failed_malloc;
2670
1.08k
        }
2671
2672
126k
        for(i = 1;;i++) {
2673
126k
            dig = quorem(b,S) + '0';
2674
            /* Do we yet have the shortest decimal string
2675
             * that will round to d?
2676
             */
2677
126k
            j = cmp(b, mlo);
2678
126k
            delta = diff(S, mhi);
2679
126k
            if (delta == NULL)
2680
0
                goto failed_malloc;
2681
126k
            j1 = delta->sign ? 1 : cmp(b, delta);
2682
126k
            Bfree(delta);
2683
126k
            if (j1 == 0 && mode != 1 && !(word1(&u) & 1)
2684
126k
                ) {
2685
1.94k
                if (dig == '9')
2686
282
                    goto round_9_up;
2687
1.66k
                if (j > 0)
2688
1.03k
                    dig++;
2689
1.66k
                *s++ = dig;
2690
1.66k
                goto ret;
2691
1.94k
            }
2692
124k
            if (j < 0 || (j == 0 && mode != 1
2693
1.80k
                          && !(word1(&u) & 1)
2694
106k
                    )) {
2695
18.1k
                if (!b->x[0] && b->wds <= 1) {
2696
2.36k
                    goto accept_dig;
2697
2.36k
                }
2698
15.7k
                if (j1 > 0) {
2699
2.70k
                    b = lshift(b, 1);
2700
2.70k
                    if (b == NULL)
2701
0
                        goto failed_malloc;
2702
2.70k
                    j1 = cmp(b, S);
2703
2.70k
                    if ((j1 > 0 || (j1 == 0 && dig & 1))
2704
1.68k
                        && dig++ == '9')
2705
274
                        goto round_9_up;
2706
2.70k
                }
2707
17.8k
              accept_dig:
2708
17.8k
                *s++ = dig;
2709
17.8k
                goto ret;
2710
15.7k
            }
2711
106k
            if (j1 > 0) {
2712
16.4k
                if (dig == '9') { /* possible if i == 1 */
2713
4.42k
                  round_9_up:
2714
4.42k
                    *s++ = '9';
2715
4.42k
                    goto roundoff;
2716
3.86k
                }
2717
12.5k
                *s++ = dig + 1;
2718
12.5k
                goto ret;
2719
16.4k
            }
2720
89.7k
            *s++ = dig;
2721
89.7k
            if (i == ilim)
2722
0
                break;
2723
89.7k
            b = multadd(b, 10, 0);
2724
89.7k
            if (b == NULL)
2725
0
                goto failed_malloc;
2726
89.7k
            if (mlo == mhi) {
2727
86.2k
                mlo = mhi = multadd(mhi, 10, 0);
2728
86.2k
                if (mlo == NULL)
2729
0
                    goto failed_malloc;
2730
86.2k
            }
2731
3.49k
            else {
2732
3.49k
                mlo = multadd(mlo, 10, 0);
2733
3.49k
                if (mlo == NULL)
2734
0
                    goto failed_malloc;
2735
3.49k
                mhi = multadd(mhi, 10, 0);
2736
3.49k
                if (mhi == NULL)
2737
0
                    goto failed_malloc;
2738
3.49k
            }
2739
89.7k
        }
2740
36.4k
    }
2741
0
    else
2742
0
        for(i = 1;; i++) {
2743
0
            *s++ = dig = quorem(b,S) + '0';
2744
0
            if (!b->x[0] && b->wds <= 1) {
2745
0
                goto ret;
2746
0
            }
2747
0
            if (i >= ilim)
2748
0
                break;
2749
0
            b = multadd(b, 10, 0);
2750
0
            if (b == NULL)
2751
0
                goto failed_malloc;
2752
0
        }
2753
2754
    /* Round off last digit */
2755
2756
0
    b = lshift(b, 1);
2757
0
    if (b == NULL)
2758
0
        goto failed_malloc;
2759
0
    j = cmp(b, S);
2760
0
    if (j > 0 || (j == 0 && dig & 1)) {
2761
4.42k
      roundoff:
2762
4.42k
        while(*--s == '9')
2763
4.42k
            if (s == s0) {
2764
4.42k
                k++;
2765
4.42k
                *s++ = '1';
2766
4.42k
                goto ret;
2767
4.42k
            }
2768
0
        ++*s++;
2769
0
    }
2770
0
    else {
2771
0
        while(*--s == '0');
2772
0
        s++;
2773
0
    }
2774
36.4k
  ret:
2775
36.4k
    Bfree(S);
2776
36.4k
    if (mhi) {
2777
36.4k
        if (mlo && mlo != mhi)
2778
1.08k
            Bfree(mlo);
2779
36.4k
        Bfree(mhi);
2780
36.4k
    }
2781
47.7k
  ret1:
2782
47.7k
    Bfree(b);
2783
47.7k
    *s = 0;
2784
47.7k
    *decpt = k + 1;
2785
47.7k
    if (rve)
2786
47.7k
        *rve = s;
2787
47.7k
    return s0;
2788
0
  failed_malloc:
2789
0
    if (S)
2790
0
        Bfree(S);
2791
0
    if (mlo && mlo != mhi)
2792
0
        Bfree(mlo);
2793
0
    if (mhi)
2794
0
        Bfree(mhi);
2795
0
    if (b)
2796
0
        Bfree(b);
2797
0
    if (s0)
2798
0
        _Py_dg_freedtoa(s0);
2799
0
    return NULL;
2800
36.4k
}
2801
2802
#endif  // _PY_SHORT_FLOAT_REPR == 1
2803
2804
PyStatus
2805
_PyDtoa_Init(PyInterpreterState *interp)
2806
36
{
2807
36
#if _PY_SHORT_FLOAT_REPR == 1 && !defined(Py_USING_MEMORY_DEBUGGER)
2808
36
    Bigint **p5s = interp->dtoa.p5s;
2809
2810
    // 5**4 = 625
2811
36
    Bigint *p5 = i2b(625);
2812
36
    if (p5 == NULL) {
2813
0
        return PyStatus_NoMemory();
2814
0
    }
2815
36
    p5s[0] = p5;
2816
2817
    // compute 5**8, 5**16, 5**32, ..., 5**512
2818
288
    for (Py_ssize_t i = 1; i < Bigint_Pow5size; i++) {
2819
252
        p5 = mult(p5, p5);
2820
252
        if (p5 == NULL) {
2821
0
            return PyStatus_NoMemory();
2822
0
        }
2823
252
        p5s[i] = p5;
2824
252
    }
2825
2826
36
#endif
2827
36
    return PyStatus_Ok();
2828
36
}
2829
2830
void
2831
_PyDtoa_Fini(PyInterpreterState *interp)
2832
0
{
2833
0
#if _PY_SHORT_FLOAT_REPR == 1 && !defined(Py_USING_MEMORY_DEBUGGER)
2834
0
    Bigint **p5s = interp->dtoa.p5s;
2835
0
    for (Py_ssize_t i = 0; i < Bigint_Pow5size; i++) {
2836
0
        Bigint *p5 = p5s[i];
2837
        p5s[i] = NULL;
2838
0
        Bfree(p5);
2839
0
    }
2840
0
#endif
2841
0
}