summaryrefslogtreecommitdiff
path: root/src/strconv/extfloat.go
blob: bed8b16bdaef3ca1f5eb4eb9b08f2a13eb856919 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
// Copyright 2011 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.

package strconv

// An extFloat represents an extended floating-point number, with more
// precision than a float64. It does not try to save bits: the
// number represented by the structure is mant*(2^exp), with a negative
// sign if neg is true.
type extFloat struct {
	mant uint64
	exp  int
	neg  bool
}

// Powers of ten taken from double-conversion library.
// http://code.google.com/p/double-conversion/
const (
	firstPowerOfTen = -348
	stepPowerOfTen  = 8
)

var smallPowersOfTen = [...]extFloat{
	{1 << 63, -63, false},        // 1
	{0xa << 60, -60, false},      // 1e1
	{0x64 << 57, -57, false},     // 1e2
	{0x3e8 << 54, -54, false},    // 1e3
	{0x2710 << 50, -50, false},   // 1e4
	{0x186a0 << 47, -47, false},  // 1e5
	{0xf4240 << 44, -44, false},  // 1e6
	{0x989680 << 40, -40, false}, // 1e7
}

var powersOfTen = [...]extFloat{
	{0xfa8fd5a0081c0288, -1220, false}, // 10^-348
	{0xbaaee17fa23ebf76, -1193, false}, // 10^-340
	{0x8b16fb203055ac76, -1166, false}, // 10^-332
	{0xcf42894a5dce35ea, -1140, false}, // 10^-324
	{0x9a6bb0aa55653b2d, -1113, false}, // 10^-316
	{0xe61acf033d1a45df, -1087, false}, // 10^-308
	{0xab70fe17c79ac6ca, -1060, false}, // 10^-300
	{0xff77b1fcbebcdc4f, -1034, false}, // 10^-292
	{0xbe5691ef416bd60c, -1007, false}, // 10^-284
	{0x8dd01fad907ffc3c, -980, false},  // 10^-276
	{0xd3515c2831559a83, -954, false},  // 10^-268
	{0x9d71ac8fada6c9b5, -927, false},  // 10^-260
	{0xea9c227723ee8bcb, -901, false},  // 10^-252
	{0xaecc49914078536d, -874, false},  // 10^-244
	{0x823c12795db6ce57, -847, false},  // 10^-236
	{0xc21094364dfb5637, -821, false},  // 10^-228
	{0x9096ea6f3848984f, -794, false},  // 10^-220
	{0xd77485cb25823ac7, -768, false},  // 10^-212
	{0xa086cfcd97bf97f4, -741, false},  // 10^-204
	{0xef340a98172aace5, -715, false},  // 10^-196
	{0xb23867fb2a35b28e, -688, false},  // 10^-188
	{0x84c8d4dfd2c63f3b, -661, false},  // 10^-180
	{0xc5dd44271ad3cdba, -635, false},  // 10^-172
	{0x936b9fcebb25c996, -608, false},  // 10^-164
	{0xdbac6c247d62a584, -582, false},  // 10^-156
	{0xa3ab66580d5fdaf6, -555, false},  // 10^-148
	{0xf3e2f893dec3f126, -529, false},  // 10^-140
	{0xb5b5ada8aaff80b8, -502, false},  // 10^-132
	{0x87625f056c7c4a8b, -475, false},  // 10^-124
	{0xc9bcff6034c13053, -449, false},  // 10^-116
	{0x964e858c91ba2655, -422, false},  // 10^-108
	{0xdff9772470297ebd, -396, false},  // 10^-100
	{0xa6dfbd9fb8e5b88f, -369, false},  // 10^-92
	{0xf8a95fcf88747d94, -343, false},  // 10^-84
	{0xb94470938fa89bcf, -316, false},  // 10^-76
	{0x8a08f0f8bf0f156b, -289, false},  // 10^-68
	{0xcdb02555653131b6, -263, false},  // 10^-60
	{0x993fe2c6d07b7fac, -236, false},  // 10^-52
	{0xe45c10c42a2b3b06, -210, false},  // 10^-44
	{0xaa242499697392d3, -183, false},  // 10^-36
	{0xfd87b5f28300ca0e, -157, false},  // 10^-28
	{0xbce5086492111aeb, -130, false},  // 10^-20
	{0x8cbccc096f5088cc, -103, false},  // 10^-12
	{0xd1b71758e219652c, -77, false},   // 10^-4
	{0x9c40000000000000, -50, false},   // 10^4
	{0xe8d4a51000000000, -24, false},   // 10^12
	{0xad78ebc5ac620000, 3, false},     // 10^20
	{0x813f3978f8940984, 30, false},    // 10^28
	{0xc097ce7bc90715b3, 56, false},    // 10^36
	{0x8f7e32ce7bea5c70, 83, false},    // 10^44
	{0xd5d238a4abe98068, 109, false},   // 10^52
	{0x9f4f2726179a2245, 136, false},   // 10^60
	{0xed63a231d4c4fb27, 162, false},   // 10^68
	{0xb0de65388cc8ada8, 189, false},   // 10^76
	{0x83c7088e1aab65db, 216, false},   // 10^84
	{0xc45d1df942711d9a, 242, false},   // 10^92
	{0x924d692ca61be758, 269, false},   // 10^100
	{0xda01ee641a708dea, 295, false},   // 10^108
	{0xa26da3999aef774a, 322, false},   // 10^116
	{0xf209787bb47d6b85, 348, false},   // 10^124
	{0xb454e4a179dd1877, 375, false},   // 10^132
	{0x865b86925b9bc5c2, 402, false},   // 10^140
	{0xc83553c5c8965d3d, 428, false},   // 10^148
	{0x952ab45cfa97a0b3, 455, false},   // 10^156
	{0xde469fbd99a05fe3, 481, false},   // 10^164
	{0xa59bc234db398c25, 508, false},   // 10^172
	{0xf6c69a72a3989f5c, 534, false},   // 10^180
	{0xb7dcbf5354e9bece, 561, false},   // 10^188
	{0x88fcf317f22241e2, 588, false},   // 10^196
	{0xcc20ce9bd35c78a5, 614, false},   // 10^204
	{0x98165af37b2153df, 641, false},   // 10^212
	{0xe2a0b5dc971f303a, 667, false},   // 10^220
	{0xa8d9d1535ce3b396, 694, false},   // 10^228
	{0xfb9b7cd9a4a7443c, 720, false},   // 10^236
	{0xbb764c4ca7a44410, 747, false},   // 10^244
	{0x8bab8eefb6409c1a, 774, false},   // 10^252
	{0xd01fef10a657842c, 800, false},   // 10^260
	{0x9b10a4e5e9913129, 827, false},   // 10^268
	{0xe7109bfba19c0c9d, 853, false},   // 10^276
	{0xac2820d9623bf429, 880, false},   // 10^284
	{0x80444b5e7aa7cf85, 907, false},   // 10^292
	{0xbf21e44003acdd2d, 933, false},   // 10^300
	{0x8e679c2f5e44ff8f, 960, false},   // 10^308
	{0xd433179d9c8cb841, 986, false},   // 10^316
	{0x9e19db92b4e31ba9, 1013, false},  // 10^324
	{0xeb96bf6ebadf77d9, 1039, false},  // 10^332
	{0xaf87023b9bf0ee6b, 1066, false},  // 10^340
}

// floatBits returns the bits of the float64 that best approximates
// the extFloat passed as receiver. Overflow is set to true if
// the resulting float64 is ±Inf.
func (f *extFloat) floatBits(flt *floatInfo) (bits uint64, overflow bool) {
	f.Normalize()

	exp := f.exp + 63

	// Exponent too small.
	if exp < flt.bias+1 {
		n := flt.bias + 1 - exp
		f.mant >>= uint(n)
		exp += n
	}

	// Extract 1+flt.mantbits bits from the 64-bit mantissa.
	mant := f.mant >> (63 - flt.mantbits)
	if f.mant&(1<<(62-flt.mantbits)) != 0 {
		// Round up.
		mant += 1
	}

	// Rounding might have added a bit; shift down.
	if mant == 2<<flt.mantbits {
		mant >>= 1
		exp++
	}

	// Infinities.
	if exp-flt.bias >= 1<<flt.expbits-1 {
		// ±Inf
		mant = 0
		exp = 1<<flt.expbits - 1 + flt.bias
		overflow = true
	} else if mant&(1<<flt.mantbits) == 0 {
		// Denormalized?
		exp = flt.bias
	}
	// Assemble bits.
	bits = mant & (uint64(1)<<flt.mantbits - 1)
	bits |= uint64((exp-flt.bias)&(1<<flt.expbits-1)) << flt.mantbits
	if f.neg {
		bits |= 1 << (flt.mantbits + flt.expbits)
	}
	return
}

// AssignComputeBounds sets f to the floating point value
// defined by mant, exp and precision given by flt. It returns
// lower, upper such that any number in the closed interval
// [lower, upper] is converted back to the same floating point number.
func (f *extFloat) AssignComputeBounds(mant uint64, exp int, neg bool, flt *floatInfo) (lower, upper extFloat) {
	f.mant = mant
	f.exp = exp - int(flt.mantbits)
	f.neg = neg
	if f.exp <= 0 && mant == (mant>>uint(-f.exp))<<uint(-f.exp) {
		// An exact integer
		f.mant >>= uint(-f.exp)
		f.exp = 0
		return *f, *f
	}
	expBiased := exp - flt.bias

	upper = extFloat{mant: 2*f.mant + 1, exp: f.exp - 1, neg: f.neg}
	if mant != 1<<flt.mantbits || expBiased == 1 {
		lower = extFloat{mant: 2*f.mant - 1, exp: f.exp - 1, neg: f.neg}
	} else {
		lower = extFloat{mant: 4*f.mant - 1, exp: f.exp - 2, neg: f.neg}
	}
	return
}

// Normalize normalizes f so that the highest bit of the mantissa is
// set, and returns the number by which the mantissa was left-shifted.
func (f *extFloat) Normalize() (shift uint) {
	mant, exp := f.mant, f.exp
	if mant == 0 {
		return 0
	}
	if mant>>(64-32) == 0 {
		mant <<= 32
		exp -= 32
	}
	if mant>>(64-16) == 0 {
		mant <<= 16
		exp -= 16
	}
	if mant>>(64-8) == 0 {
		mant <<= 8
		exp -= 8
	}
	if mant>>(64-4) == 0 {
		mant <<= 4
		exp -= 4
	}
	if mant>>(64-2) == 0 {
		mant <<= 2
		exp -= 2
	}
	if mant>>(64-1) == 0 {
		mant <<= 1
		exp -= 1
	}
	shift = uint(f.exp - exp)
	f.mant, f.exp = mant, exp
	return
}

// Multiply sets f to the product f*g: the result is correctly rounded,
// but not normalized.
func (f *extFloat) Multiply(g extFloat) {
	fhi, flo := f.mant>>32, uint64(uint32(f.mant))
	ghi, glo := g.mant>>32, uint64(uint32(g.mant))

	// Cross products.
	cross1 := fhi * glo
	cross2 := flo * ghi

	// f.mant*g.mant is fhi*ghi << 64 + (cross1+cross2) << 32 + flo*glo
	f.mant = fhi*ghi + (cross1 >> 32) + (cross2 >> 32)
	rem := uint64(uint32(cross1)) + uint64(uint32(cross2)) + ((flo * glo) >> 32)
	// Round up.
	rem += (1 << 31)

	f.mant += (rem >> 32)
	f.exp = f.exp + g.exp + 64
}

var uint64pow10 = [...]uint64{
	1, 1e1, 1e2, 1e3, 1e4, 1e5, 1e6, 1e7, 1e8, 1e9,
	1e10, 1e11, 1e12, 1e13, 1e14, 1e15, 1e16, 1e17, 1e18, 1e19,
}

// AssignDecimal sets f to an approximate value mantissa*10^exp. It
// returns true if the value represented by f is guaranteed to be the
// best approximation of d after being rounded to a float64 or
// float32 depending on flt.
func (f *extFloat) AssignDecimal(mantissa uint64, exp10 int, neg bool, trunc bool, flt *floatInfo) (ok bool) {
	const uint64digits = 19
	const errorscale = 8
	errors := 0 // An upper bound for error, computed in errorscale*ulp.
	if trunc {
		// the decimal number was truncated.
		errors += errorscale / 2
	}

	f.mant = mantissa
	f.exp = 0
	f.neg = neg

	// Multiply by powers of ten.
	i := (exp10 - firstPowerOfTen) / stepPowerOfTen
	if exp10 < firstPowerOfTen || i >= len(powersOfTen) {
		return false
	}
	adjExp := (exp10 - firstPowerOfTen) % stepPowerOfTen

	// We multiply by exp%step
	if adjExp < uint64digits && mantissa < uint64pow10[uint64digits-adjExp] {
		// We can multiply the mantissa exactly.
		f.mant *= uint64pow10[adjExp]
		f.Normalize()
	} else {
		f.Normalize()
		f.Multiply(smallPowersOfTen[adjExp])
		errors += errorscale / 2
	}

	// We multiply by 10 to the exp - exp%step.
	f.Multiply(powersOfTen[i])
	if errors > 0 {
		errors += 1
	}
	errors += errorscale / 2

	// Normalize
	shift := f.Normalize()
	errors <<= shift

	// Now f is a good approximation of the decimal.
	// Check whether the error is too large: that is, if the mantissa
	// is perturbated by the error, the resulting float64 will change.
	// The 64 bits mantissa is 1 + 52 bits for float64 + 11 extra bits.
	//
	// In many cases the approximation will be good enough.
	denormalExp := flt.bias - 63
	var extrabits uint
	if f.exp <= denormalExp {
		// f.mant * 2^f.exp is smaller than 2^(flt.bias+1).
		extrabits = uint(63 - flt.mantbits + 1 + uint(denormalExp-f.exp))
	} else {
		extrabits = uint(63 - flt.mantbits)
	}

	halfway := uint64(1) << (extrabits - 1)
	mant_extra := f.mant & (1<<extrabits - 1)

	// Do a signed comparison here! If the error estimate could make
	// the mantissa round differently for the conversion to double,
	// then we can't give a definite answer.
	if int64(halfway)-int64(errors) < int64(mant_extra) &&
		int64(mant_extra) < int64(halfway)+int64(errors) {
		return false
	}
	return true
}

// Frexp10 is an analogue of math.Frexp for decimal powers. It scales
// f by an approximate power of ten 10^-exp, and returns exp10, so
// that f*10^exp10 has the same value as the old f, up to an ulp,
// as well as the index of 10^-exp in the powersOfTen table.
func (f *extFloat) frexp10() (exp10, index int) {
	// The constants expMin and expMax constrain the final value of the
	// binary exponent of f. We want a small integral part in the result
	// because finding digits of an integer requires divisions, whereas
	// digits of the fractional part can be found by repeatedly multiplying
	// by 10.
	const expMin = -60
	const expMax = -32
	// Find power of ten such that x * 10^n has a binary exponent
	// between expMin and expMax.
	approxExp10 := ((expMin+expMax)/2 - f.exp) * 28 / 93 // log(10)/log(2) is close to 93/28.
	i := (approxExp10 - firstPowerOfTen) / stepPowerOfTen
Loop:
	for {
		exp := f.exp + powersOfTen[i].exp + 64
		switch {
		case exp < expMin:
			i++
		case exp > expMax:
			i--
		default:
			break Loop
		}
	}
	// Apply the desired decimal shift on f. It will have exponent
	// in the desired range. This is multiplication by 10^-exp10.
	f.Multiply(powersOfTen[i])

	return -(firstPowerOfTen + i*stepPowerOfTen), i
}

// frexp10Many applies a common shift by a power of ten to a, b, c.
func frexp10Many(a, b, c *extFloat) (exp10 int) {
	exp10, i := c.frexp10()
	a.Multiply(powersOfTen[i])
	b.Multiply(powersOfTen[i])
	return
}

// FixedDecimal stores in d the first n significant digits
// of the decimal representation of f. It returns false
// if it cannot be sure of the answer.
func (f *extFloat) FixedDecimal(d *decimalSlice, n int) bool {
	if f.mant == 0 {
		d.nd = 0
		d.dp = 0
		d.neg = f.neg
		return true
	}
	if n == 0 {
		panic("strconv: internal error: extFloat.FixedDecimal called with n == 0")
	}
	// Multiply by an appropriate power of ten to have a reasonable
	// number to process.
	f.Normalize()
	exp10, _ := f.frexp10()

	shift := uint(-f.exp)
	integer := uint32(f.mant >> shift)
	fraction := f.mant - (uint64(integer) << shift)
	ε := uint64(1) // ε is the uncertainty we have on the mantissa of f.

	// Write exactly n digits to d.
	needed := n        // how many digits are left to write.
	integerDigits := 0 // the number of decimal digits of integer.
	pow10 := uint64(1) // the power of ten by which f was scaled.
	for i, pow := 0, uint64(1); i < 20; i++ {
		if pow > uint64(integer) {
			integerDigits = i
			break
		}
		pow *= 10
	}
	rest := integer
	if integerDigits > needed {
		// the integral part is already large, trim the last digits.
		pow10 = uint64pow10[integerDigits-needed]
		integer /= uint32(pow10)
		rest -= integer * uint32(pow10)
	} else {
		rest = 0
	}

	// Write the digits of integer: the digits of rest are omitted.
	var buf [32]byte
	pos := len(buf)
	for v := integer; v > 0; {
		v1 := v / 10
		v -= 10 * v1
		pos--
		buf[pos] = byte(v + '0')
		v = v1
	}
	for i := pos; i < len(buf); i++ {
		d.d[i-pos] = buf[i]
	}
	nd := len(buf) - pos
	d.nd = nd
	d.dp = integerDigits + exp10
	needed -= nd

	if needed > 0 {
		if rest != 0 || pow10 != 1 {
			panic("strconv: internal error, rest != 0 but needed > 0")
		}
		// Emit digits for the fractional part. Each time, 10*fraction
		// fits in a uint64 without overflow.
		for needed > 0 {
			fraction *= 10
			ε *= 10 // the uncertainty scales as we multiply by ten.
			if 2*ε > 1<<shift {
				// the error is so large it could modify which digit to write, abort.
				return false
			}
			digit := fraction >> shift
			d.d[nd] = byte(digit + '0')
			fraction -= digit << shift
			nd++
			needed--
		}
		d.nd = nd
	}

	// We have written a truncation of f (a numerator / 10^d.dp). The remaining part
	// can be interpreted as a small number (< 1) to be added to the last digit of the
	// numerator.
	//
	// If rest > 0, the amount is:
	//    (rest<<shift | fraction) / (pow10 << shift)
	//    fraction being known with a ±ε uncertainty.
	//    The fact that n > 0 guarantees that pow10 << shift does not overflow a uint64.
	//
	// If rest = 0, pow10 == 1 and the amount is
	//    fraction / (1 << shift)
	//    fraction being known with a ±ε uncertainty.
	//
	// We pass this information to the rounding routine for adjustment.

	ok := adjustLastDigitFixed(d, uint64(rest)<<shift|fraction, pow10, shift, ε)
	if !ok {
		return false
	}
	// Trim trailing zeros.
	for i := d.nd - 1; i >= 0; i-- {
		if d.d[i] != '0' {
			d.nd = i + 1
			break
		}
	}
	return true
}

// adjustLastDigitFixed assumes d contains the representation of the integral part
// of some number, whose fractional part is num / (den << shift). The numerator
// num is only known up to an uncertainty of size ε, assumed to be less than
// (den << shift)/2.
//
// It will increase the last digit by one to account for correct rounding, typically
// when the fractional part is greater than 1/2, and will return false if ε is such
// that no correct answer can be given.
func adjustLastDigitFixed(d *decimalSlice, num, den uint64, shift uint, ε uint64) bool {
	if num > den<<shift {
		panic("strconv: num > den<<shift in adjustLastDigitFixed")
	}
	if 2*ε > den<<shift {
		panic("strconv: ε > (den<<shift)/2")
	}
	if 2*(num+ε) < den<<shift {
		return true
	}
	if 2*(num-ε) > den<<shift {
		// increment d by 1.
		i := d.nd - 1
		for ; i >= 0; i-- {
			if d.d[i] == '9' {
				d.nd--
			} else {
				break
			}
		}
		if i < 0 {
			d.d[0] = '1'
			d.nd = 1
			d.dp++
		} else {
			d.d[i]++
		}
		return true
	}
	return false
}

// ShortestDecimal stores in d the shortest decimal representation of f
// which belongs to the open interval (lower, upper), where f is supposed
// to lie. It returns false whenever the result is unsure. The implementation
// uses the Grisu3 algorithm.
func (f *extFloat) ShortestDecimal(d *decimalSlice, lower, upper *extFloat) bool {
	if f.mant == 0 {
		d.nd = 0
		d.dp = 0
		d.neg = f.neg
		return true
	}
	if f.exp == 0 && *lower == *f && *lower == *upper {
		// an exact integer.
		var buf [24]byte
		n := len(buf) - 1
		for v := f.mant; v > 0; {
			v1 := v / 10
			v -= 10 * v1
			buf[n] = byte(v + '0')
			n--
			v = v1
		}
		nd := len(buf) - n - 1
		for i := 0; i < nd; i++ {
			d.d[i] = buf[n+1+i]
		}
		d.nd, d.dp = nd, nd
		for d.nd > 0 && d.d[d.nd-1] == '0' {
			d.nd--
		}
		if d.nd == 0 {
			d.dp = 0
		}
		d.neg = f.neg
		return true
	}
	upper.Normalize()
	// Uniformize exponents.
	if f.exp > upper.exp {
		f.mant <<= uint(f.exp - upper.exp)
		f.exp = upper.exp
	}
	if lower.exp > upper.exp {
		lower.mant <<= uint(lower.exp - upper.exp)
		lower.exp = upper.exp
	}

	exp10 := frexp10Many(lower, f, upper)
	// Take a safety margin due to rounding in frexp10Many, but we lose precision.
	upper.mant++
	lower.mant--

	// The shortest representation of f is either rounded up or down, but
	// in any case, it is a truncation of upper.
	shift := uint(-upper.exp)
	integer := uint32(upper.mant >> shift)
	fraction := upper.mant - (uint64(integer) << shift)

	// How far we can go down from upper until the result is wrong.
	allowance := upper.mant - lower.mant
	// How far we should go to get a very precise result.
	targetDiff := upper.mant - f.mant

	// Count integral digits: there are at most 10.
	var integerDigits int
	for i, pow := 0, uint64(1); i < 20; i++ {
		if pow > uint64(integer) {
			integerDigits = i
			break
		}
		pow *= 10
	}
	for i := 0; i < integerDigits; i++ {
		pow := uint64pow10[integerDigits-i-1]
		digit := integer / uint32(pow)
		d.d[i] = byte(digit + '0')
		integer -= digit * uint32(pow)
		// evaluate whether we should stop.
		if currentDiff := uint64(integer)<<shift + fraction; currentDiff < allowance {
			d.nd = i + 1
			d.dp = integerDigits + exp10
			d.neg = f.neg
			// Sometimes allowance is so large the last digit might need to be
			// decremented to get closer to f.
			return adjustLastDigit(d, currentDiff, targetDiff, allowance, pow<<shift, 2)
		}
	}
	d.nd = integerDigits
	d.dp = d.nd + exp10
	d.neg = f.neg

	// Compute digits of the fractional part. At each step fraction does not
	// overflow. The choice of minExp implies that fraction is less than 2^60.
	var digit int
	multiplier := uint64(1)
	for {
		fraction *= 10
		multiplier *= 10
		digit = int(fraction >> shift)
		d.d[d.nd] = byte(digit + '0')
		d.nd++
		fraction -= uint64(digit) << shift
		if fraction < allowance*multiplier {
			// We are in the admissible range. Note that if allowance is about to
			// overflow, that is, allowance > 2^64/10, the condition is automatically
			// true due to the limited range of fraction.
			return adjustLastDigit(d,
				fraction, targetDiff*multiplier, allowance*multiplier,
				1<<shift, multiplier*2)
		}
	}
}

// adjustLastDigit modifies d = x-currentDiff*ε, to get closest to
// d = x-targetDiff*ε, without becoming smaller than x-maxDiff*ε.
// It assumes that a decimal digit is worth ulpDecimal*ε, and that
// all data is known with a error estimate of ulpBinary*ε.
func adjustLastDigit(d *decimalSlice, currentDiff, targetDiff, maxDiff, ulpDecimal, ulpBinary uint64) bool {
	if ulpDecimal < 2*ulpBinary {
		// Approximation is too wide.
		return false
	}
	for currentDiff+ulpDecimal/2+ulpBinary < targetDiff {
		d.d[d.nd-1]--
		currentDiff += ulpDecimal
	}
	if currentDiff+ulpDecimal <= targetDiff+ulpDecimal/2+ulpBinary {
		// we have two choices, and don't know what to do.
		return false
	}
	if currentDiff < ulpBinary || currentDiff > maxDiff-ulpBinary {
		// we went too far
		return false
	}
	if d.nd == 1 && d.d[0] == '0' {
		// the number has actually reached zero.
		d.nd = 0
		d.dp = 0
	}
	return true
}