summaryrefslogtreecommitdiff
path: root/mfbt/MathAlgorithms.h
blob: 153564da67cff986fc98aa88dc20786e9b54ce4c (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
/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* This Source Code Form is subject to the terms of the Mozilla Public
 * License, v. 2.0. If a copy of the MPL was not distributed with this
 * file, You can obtain one at http://mozilla.org/MPL/2.0/. */

/* mfbt maths algorithms. */

#ifndef mozilla_MathAlgorithms_h
#define mozilla_MathAlgorithms_h

#include "mozilla/Assertions.h"
#include "mozilla/TypeTraits.h"

#include <cmath>
#include <limits.h>
#include <stdint.h>

namespace mozilla {

// Greatest Common Divisor
template<typename IntegerType>
MOZ_ALWAYS_INLINE IntegerType
EuclidGCD(IntegerType aA, IntegerType aB)
{
  // Euclid's algorithm; O(N) in the worst case.  (There are better
  // ways, but we don't need them for the current use of this algo.)
  MOZ_ASSERT(aA > IntegerType(0));
  MOZ_ASSERT(aB > IntegerType(0));

  while (aA != aB) {
    if (aA > aB) {
      aA = aA - aB;
    } else {
      aB = aB - aA;
    }
  }

  return aA;
}

// Least Common Multiple
template<typename IntegerType>
MOZ_ALWAYS_INLINE IntegerType
EuclidLCM(IntegerType aA, IntegerType aB)
{
  // Divide first to reduce overflow risk.
  return (aA / EuclidGCD(aA, aB)) * aB;
}

namespace detail {

template<typename T>
struct AllowDeprecatedAbsFixed : FalseType {};

template<> struct AllowDeprecatedAbsFixed<int32_t> : TrueType {};
template<> struct AllowDeprecatedAbsFixed<int64_t> : TrueType {};

template<typename T>
struct AllowDeprecatedAbs : AllowDeprecatedAbsFixed<T> {};

template<> struct AllowDeprecatedAbs<int> : TrueType {};
template<> struct AllowDeprecatedAbs<long> : TrueType {};

} // namespace detail

// DO NOT USE DeprecatedAbs.  It exists only until its callers can be converted
// to Abs below, and it will be removed when all callers have been changed.
template<typename T>
inline typename mozilla::EnableIf<detail::AllowDeprecatedAbs<T>::value, T>::Type
DeprecatedAbs(const T aValue)
{
  // The absolute value of the smallest possible value of a signed-integer type
  // won't fit in that type (on twos-complement systems -- and we're blithely
  // assuming we're on such systems, for the non-<stdint.h> types listed above),
  // so assert that the input isn't that value.
  //
  // This is the case if: the value is non-negative; or if adding one (giving a
  // value in the range [-maxvalue, 0]), then negating (giving a value in the
  // range [0, maxvalue]), doesn't produce maxvalue (because in twos-complement,
  // (minvalue + 1) == -maxvalue).
  MOZ_ASSERT(aValue >= 0 ||
             -(aValue + 1) != T((1ULL << (CHAR_BIT * sizeof(T) - 1)) - 1),
             "You can't negate the smallest possible negative integer!");
  return aValue >= 0 ? aValue : -aValue;
}

namespace detail {

// For now mozilla::Abs only takes intN_T, the signed natural types, and
// float/double/long double.  Feel free to add overloads for other standard,
// signed types if you need them.

template<typename T>
struct AbsReturnTypeFixed;

template<> struct AbsReturnTypeFixed<int8_t> { typedef uint8_t Type; };
template<> struct AbsReturnTypeFixed<int16_t> { typedef uint16_t Type; };
template<> struct AbsReturnTypeFixed<int32_t> { typedef uint32_t Type; };
template<> struct AbsReturnTypeFixed<int64_t> { typedef uint64_t Type; };

template<typename T>
struct AbsReturnType : AbsReturnTypeFixed<T> {};

template<> struct AbsReturnType<char> :
  EnableIf<char(-1) < char(0), unsigned char> {};
template<> struct AbsReturnType<signed char> { typedef unsigned char Type; };
template<> struct AbsReturnType<short> { typedef unsigned short Type; };
template<> struct AbsReturnType<int> { typedef unsigned int Type; };
template<> struct AbsReturnType<long> { typedef unsigned long Type; };
template<> struct AbsReturnType<long long> { typedef unsigned long long Type; };
template<> struct AbsReturnType<float> { typedef float Type; };
template<> struct AbsReturnType<double> { typedef double Type; };
template<> struct AbsReturnType<long double> { typedef long double Type; };

} // namespace detail

template<typename T>
inline typename detail::AbsReturnType<T>::Type
Abs(const T aValue)
{
  typedef typename detail::AbsReturnType<T>::Type ReturnType;
  return aValue >= 0 ? ReturnType(aValue) : ~ReturnType(aValue) + 1;
}

template<>
inline float
Abs<float>(const float aFloat)
{
  return std::fabs(aFloat);
}

template<>
inline double
Abs<double>(const double aDouble)
{
  return std::fabs(aDouble);
}

template<>
inline long double
Abs<long double>(const long double aLongDouble)
{
  return std::fabs(aLongDouble);
}

} // namespace mozilla

#if defined(_MSC_VER) && \
    (defined(_M_IX86) || defined(_M_AMD64) || defined(_M_X64))
#  define MOZ_BITSCAN_WINDOWS

#  include <intrin.h>
#  pragma intrinsic(_BitScanForward, _BitScanReverse)

#  if defined(_M_AMD64) || defined(_M_X64)
#    define MOZ_BITSCAN_WINDOWS64
#   pragma intrinsic(_BitScanForward64, _BitScanReverse64)
#  endif

#endif

namespace mozilla {

namespace detail {

#if defined(MOZ_BITSCAN_WINDOWS)

inline uint_fast8_t
CountLeadingZeroes32(uint32_t aValue)
{
  unsigned long index;
  if (!_BitScanReverse(&index, static_cast<unsigned long>(aValue)))
      return 32;
  return uint_fast8_t(31 - index);
}


inline uint_fast8_t
CountTrailingZeroes32(uint32_t aValue)
{
  unsigned long index;
  if (!_BitScanForward(&index, static_cast<unsigned long>(aValue)))
      return 32;
  return uint_fast8_t(index);
}

inline uint_fast8_t
CountPopulation32(uint32_t aValue)
{
  uint32_t x = aValue - ((aValue >> 1) & 0x55555555);
  x = (x & 0x33333333) + ((x >> 2) & 0x33333333);
  return (((x + (x >> 4)) & 0xf0f0f0f) * 0x1010101) >> 24;
}
inline uint_fast8_t
CountPopulation64(uint64_t aValue)
{
  return uint_fast8_t(CountPopulation32(aValue & 0xffffffff) +
                      CountPopulation32(aValue >> 32));
}

inline uint_fast8_t
CountLeadingZeroes64(uint64_t aValue)
{
#if defined(MOZ_BITSCAN_WINDOWS64)
  unsigned long index;
  if (!_BitScanReverse64(&index, static_cast<unsigned __int64>(aValue)))
      return 64;
  return uint_fast8_t(63 - index);
#else
  uint32_t hi = uint32_t(aValue >> 32);
  if (hi != 0) {
    return CountLeadingZeroes32(hi);
  }
  return 32u + CountLeadingZeroes32(uint32_t(aValue));
#endif
}

inline uint_fast8_t
CountTrailingZeroes64(uint64_t aValue)
{
#if defined(MOZ_BITSCAN_WINDOWS64)
  unsigned long index;
  if (!_BitScanForward64(&index, static_cast<unsigned __int64>(aValue)))
      return 64;
  return uint_fast8_t(index);
#else
  uint32_t lo = uint32_t(aValue);
  if (lo != 0) {
    return CountTrailingZeroes32(lo);
  }
  return 32u + CountTrailingZeroes32(uint32_t(aValue >> 32));
#endif
}

#  ifdef MOZ_HAVE_BITSCAN64
#    undef MOZ_HAVE_BITSCAN64
#  endif

#elif defined(__clang__) || defined(__GNUC__)

#  if defined(__clang__)
#    if !__has_builtin(__builtin_ctz) || !__has_builtin(__builtin_clz)
#      error "A clang providing __builtin_c[lt]z is required to build"
#    endif
#  else
     // gcc has had __builtin_clz and friends since 3.4: no need to check.
#  endif

inline uint_fast8_t
CountLeadingZeroes32(uint32_t aValue)
{
  return __builtin_clz(aValue);
}

inline uint_fast8_t
CountTrailingZeroes32(uint32_t aValue)
{
  return __builtin_ctz(aValue);
}

inline uint_fast8_t
CountPopulation32(uint32_t aValue)
{
  return __builtin_popcount(aValue);
}

inline uint_fast8_t
CountPopulation64(uint64_t aValue)
{
  return __builtin_popcountll(aValue);
}

inline uint_fast8_t
CountLeadingZeroes64(uint64_t aValue)
{
  return __builtin_clzll(aValue);
}

inline uint_fast8_t
CountTrailingZeroes64(uint64_t aValue)
{
  return __builtin_ctzll(aValue);
}

#else
#  error "Implement these!"
inline uint_fast8_t CountLeadingZeroes32(uint32_t aValue) = delete;
inline uint_fast8_t CountTrailingZeroes32(uint32_t aValue) = delete;
inline uint_fast8_t CountPopulation32(uint32_t aValue) = delete;
inline uint_fast8_t CountPopulation64(uint64_t aValue) = delete;
inline uint_fast8_t CountLeadingZeroes64(uint64_t aValue) = delete;
inline uint_fast8_t CountTrailingZeroes64(uint64_t aValue) = delete;
#endif

} // namespace detail

/**
 * Compute the number of high-order zero bits in the NON-ZERO number |aValue|.
 * That is, looking at the bitwise representation of the number, with the
 * highest- valued bits at the start, return the number of zeroes before the
 * first one is observed.
 *
 * CountLeadingZeroes32(0xF0FF1000) is 0;
 * CountLeadingZeroes32(0x7F8F0001) is 1;
 * CountLeadingZeroes32(0x3FFF0100) is 2;
 * CountLeadingZeroes32(0x1FF50010) is 3; and so on.
 */
inline uint_fast8_t
CountLeadingZeroes32(uint32_t aValue)
{
  MOZ_ASSERT(aValue != 0);
  return detail::CountLeadingZeroes32(aValue);
}

/**
 * Compute the number of low-order zero bits in the NON-ZERO number |aValue|.
 * That is, looking at the bitwise representation of the number, with the
 * lowest- valued bits at the start, return the number of zeroes before the
 * first one is observed.
 *
 * CountTrailingZeroes32(0x0100FFFF) is 0;
 * CountTrailingZeroes32(0x7000FFFE) is 1;
 * CountTrailingZeroes32(0x0080FFFC) is 2;
 * CountTrailingZeroes32(0x0080FFF8) is 3; and so on.
 */
inline uint_fast8_t
CountTrailingZeroes32(uint32_t aValue)
{
  MOZ_ASSERT(aValue != 0);
  return detail::CountTrailingZeroes32(aValue);
}

/**
 * Compute the number of one bits in the number |aValue|,
 */
inline uint_fast8_t
CountPopulation32(uint32_t aValue)
{
  return detail::CountPopulation32(aValue);
}

/** Analogous to CountPopulation32, but for 64-bit numbers */
inline uint_fast8_t
CountPopulation64(uint64_t aValue)
{
  return detail::CountPopulation64(aValue);
}

/** Analogous to CountLeadingZeroes32, but for 64-bit numbers. */
inline uint_fast8_t
CountLeadingZeroes64(uint64_t aValue)
{
  MOZ_ASSERT(aValue != 0);
  return detail::CountLeadingZeroes64(aValue);
}

/** Analogous to CountTrailingZeroes32, but for 64-bit numbers. */
inline uint_fast8_t
CountTrailingZeroes64(uint64_t aValue)
{
  MOZ_ASSERT(aValue != 0);
  return detail::CountTrailingZeroes64(aValue);
}

namespace detail {

template<typename T, size_t Size = sizeof(T)>
class CeilingLog2;

template<typename T>
class CeilingLog2<T, 4>
{
public:
  static uint_fast8_t compute(const T aValue)
  {
    // Check for <= 1 to avoid the == 0 undefined case.
    return aValue <= 1 ? 0u : 32u - CountLeadingZeroes32(aValue - 1);
  }
};

template<typename T>
class CeilingLog2<T, 8>
{
public:
  static uint_fast8_t compute(const T aValue)
  {
    // Check for <= 1 to avoid the == 0 undefined case.
    return aValue <= 1 ? 0u : 64u - CountLeadingZeroes64(aValue - 1);
  }
};

} // namespace detail

/**
 * Compute the log of the least power of 2 greater than or equal to |aValue|.
 *
 * CeilingLog2(0..1) is 0;
 * CeilingLog2(2) is 1;
 * CeilingLog2(3..4) is 2;
 * CeilingLog2(5..8) is 3;
 * CeilingLog2(9..16) is 4; and so on.
 */
template<typename T>
inline uint_fast8_t
CeilingLog2(const T aValue)
{
  return detail::CeilingLog2<T>::compute(aValue);
}

/** A CeilingLog2 variant that accepts only size_t. */
inline uint_fast8_t
CeilingLog2Size(size_t aValue)
{
  return CeilingLog2(aValue);
}

namespace detail {

template<typename T, size_t Size = sizeof(T)>
class FloorLog2;

template<typename T>
class FloorLog2<T, 4>
{
public:
  static uint_fast8_t compute(const T aValue)
  {
    return 31u - CountLeadingZeroes32(aValue | 1);
  }
};

template<typename T>
class FloorLog2<T, 8>
{
public:
  static uint_fast8_t compute(const T aValue)
  {
    return 63u - CountLeadingZeroes64(aValue | 1);
  }
};

} // namespace detail

/**
 * Compute the log of the greatest power of 2 less than or equal to |aValue|.
 *
 * FloorLog2(0..1) is 0;
 * FloorLog2(2..3) is 1;
 * FloorLog2(4..7) is 2;
 * FloorLog2(8..15) is 3; and so on.
 */
template<typename T>
inline uint_fast8_t
FloorLog2(const T aValue)
{
  return detail::FloorLog2<T>::compute(aValue);
}

/** A FloorLog2 variant that accepts only size_t. */
inline uint_fast8_t
FloorLog2Size(size_t aValue)
{
  return FloorLog2(aValue);
}

/*
 * Compute the smallest power of 2 greater than or equal to |x|.  |x| must not
 * be so great that the computed value would overflow |size_t|.
 */
inline size_t
RoundUpPow2(size_t aValue)
{
  MOZ_ASSERT(aValue <= (size_t(1) << (sizeof(size_t) * CHAR_BIT - 1)),
             "can't round up -- will overflow!");
  return size_t(1) << CeilingLog2(aValue);
}

/**
 * Rotates the bits of the given value left by the amount of the shift width.
 */
template<typename T>
inline T
RotateLeft(const T aValue, uint_fast8_t aShift)
{
  MOZ_ASSERT(aShift < sizeof(T) * CHAR_BIT, "Shift value is too large!");
  MOZ_ASSERT(aShift > 0,
             "Rotation by value length is undefined behavior, but compilers "
             "do not currently fold a test into the rotate instruction. "
             "Please remove this restriction when compilers optimize the "
             "zero case (http://blog.regehr.org/archives/1063).");
  static_assert(IsUnsigned<T>::value, "Rotates require unsigned values");
  return (aValue << aShift) | (aValue >> (sizeof(T) * CHAR_BIT - aShift));
}

/**
 * Rotates the bits of the given value right by the amount of the shift width.
 */
template<typename T>
inline T
RotateRight(const T aValue, uint_fast8_t aShift)
{
  MOZ_ASSERT(aShift < sizeof(T) * CHAR_BIT, "Shift value is too large!");
  MOZ_ASSERT(aShift > 0,
             "Rotation by value length is undefined behavior, but compilers "
             "do not currently fold a test into the rotate instruction. "
             "Please remove this restriction when compilers optimize the "
             "zero case (http://blog.regehr.org/archives/1063).");
  static_assert(IsUnsigned<T>::value, "Rotates require unsigned values");
  return (aValue >> aShift) | (aValue << (sizeof(T) * CHAR_BIT - aShift));
}

/**
 * Returns true if |x| is a power of two.
 * Zero is not an integer power of two. (-Inf is not an integer)
 */
template<typename T>
constexpr bool
IsPowerOfTwo(T x)
{
    static_assert(IsUnsigned<T>::value,
                  "IsPowerOfTwo requires unsigned values");
    return x && (x & (x - 1)) == 0;
}

template<typename T>
inline T
Clamp(const T aValue, const T aMin, const T aMax)
{
    static_assert(IsIntegral<T>::value,
                  "Clamp accepts only integral types, so that it doesn't have"
                  " to distinguish differently-signed zeroes (which users may"
                  " or may not care to distinguish, likely at a perf cost) or"
                  " to decide how to clamp NaN or a range with a NaN"
                  " endpoint.");
    MOZ_ASSERT(aMin <= aMax);

    if (aValue <= aMin)
        return aMin;
    if (aValue >= aMax)
        return aMax;
    return aValue;
}

} /* namespace mozilla */

#endif /* mozilla_MathAlgorithms_h */