libstdc++
bit
Go to the documentation of this file.
1 // <bit> -*- C++ -*-
2 
3 // Copyright (C) 2018-2024 Free Software Foundation, Inc.
4 //
5 // This file is part of the GNU ISO C++ Library. This library is free
6 // software; you can redistribute it and/or modify it under the
7 // terms of the GNU General Public License as published by the
8 // Free Software Foundation; either version 3, or (at your option)
9 // any later version.
10 
11 // This library is distributed in the hope that it will be useful,
12 // but WITHOUT ANY WARRANTY; without even the implied warranty of
13 // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 // GNU General Public License for more details.
15 
16 // Under Section 7 of GPL version 3, you are granted additional
17 // permissions described in the GCC Runtime Library Exception, version
18 // 3.1, as published by the Free Software Foundation.
19 
20 // You should have received a copy of the GNU General Public License and
21 // a copy of the GCC Runtime Library Exception along with this program;
22 // see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
23 // <http://www.gnu.org/licenses/>.
24 
25 /** @file include/bit
26  * This is a Standard C++ Library header.
27  */
28 
29 #ifndef _GLIBCXX_BIT
30 #define _GLIBCXX_BIT 1
31 
32 #pragma GCC system_header
33 
34 #if __cplusplus >= 201402L
35 
36 #include <concepts> // for std::integral
37 #include <type_traits>
38 
39 #if _GLIBCXX_HOSTED || __has_include(<ext/numeric_traits.h>)
40 # include <ext/numeric_traits.h>
41 #else
42 # include <limits>
43 /// @cond undocumented
44 namespace __gnu_cxx
45 {
46  template<typename _Tp>
47  struct __int_traits
48  {
49  static constexpr int __digits = std::numeric_limits<_Tp>::digits;
50  static constexpr _Tp __max = std::numeric_limits<_Tp>::max();
51  };
52 }
53 /// @endcond
54 #endif
55 
56 #define __glibcxx_want_bit_cast
57 #define __glibcxx_want_byteswap
58 #define __glibcxx_want_bitops
59 #define __glibcxx_want_int_pow2
60 #define __glibcxx_want_endian
61 #include <bits/version.h>
62 
63 namespace std _GLIBCXX_VISIBILITY(default)
64 {
65 _GLIBCXX_BEGIN_NAMESPACE_VERSION
66 
67  /**
68  * @defgroup bit_manip Bit manipulation
69  * @ingroup numerics
70  *
71  * Utilities for examining and manipulating individual bits.
72  *
73  * @{
74  */
75 
76 #ifdef __cpp_lib_bit_cast // C++ >= 20
77 
78  /// Create a value of type `To` from the bits of `from`.
79  /**
80  * @tparam _To A trivially-copyable type.
81  * @param __from A trivially-copyable object of the same size as `_To`.
82  * @return An object of type `_To`.
83  * @since C++20
84  */
85  template<typename _To, typename _From>
86  [[nodiscard]]
87  constexpr _To
88  bit_cast(const _From& __from) noexcept
89 #ifdef __cpp_concepts
90  requires (sizeof(_To) == sizeof(_From))
91  && is_trivially_copyable_v<_To> && is_trivially_copyable_v<_From>
92 #endif
93  {
94  return __builtin_bit_cast(_To, __from);
95  }
96 #endif // __cpp_lib_bit_cast
97 
98 #ifdef __cpp_lib_byteswap // C++ >= 23
99 
100  /// Reverse order of bytes in the object representation of `value`.
101  /**
102  * @tparam _Tp An integral type.
103  * @param __value An object of integer type.
104  * @return An object of the same type, with the bytes reversed.
105  * @since C++23
106  */
107  template<integral _Tp>
108  [[nodiscard]]
109  constexpr _Tp
110  byteswap(_Tp __value) noexcept
111  {
112  if constexpr (sizeof(_Tp) == 1)
113  return __value;
114 #if __cpp_if_consteval >= 202106L && __CHAR_BIT__ == 8
115  if !consteval
116  {
117  if constexpr (sizeof(_Tp) == 2)
118  return __builtin_bswap16(__value);
119  if constexpr (sizeof(_Tp) == 4)
120  return __builtin_bswap32(__value);
121  if constexpr (sizeof(_Tp) == 8)
122  return __builtin_bswap64(__value);
123  if constexpr (sizeof(_Tp) == 16)
124 #if __has_builtin(__builtin_bswap128)
125  return __builtin_bswap128(__value);
126 #else
127  return (__builtin_bswap64(__value >> 64)
128  | (static_cast<_Tp>(__builtin_bswap64(__value)) << 64));
129 #endif
130  }
131 #endif
132 
133  // Fallback implementation that handles even __int24 etc.
134  using _Up = typename __make_unsigned<__remove_cv_t<_Tp>>::__type;
135  size_t __diff = __CHAR_BIT__ * (sizeof(_Tp) - 1);
136  _Up __mask1 = static_cast<unsigned char>(~0);
137  _Up __mask2 = __mask1 << __diff;
138  _Up __val = __value;
139  for (size_t __i = 0; __i < sizeof(_Tp) / 2; ++__i)
140  {
141  _Up __byte1 = __val & __mask1;
142  _Up __byte2 = __val & __mask2;
143  __val = (__val ^ __byte1 ^ __byte2
144  ^ (__byte1 << __diff) ^ (__byte2 >> __diff));
145  __mask1 <<= __CHAR_BIT__;
146  __mask2 >>= __CHAR_BIT__;
147  __diff -= 2 * __CHAR_BIT__;
148  }
149  return __val;
150  }
151 #endif // __cpp_lib_byteswap
152 
153  /// @cond undocumented
154 
155  template<typename _Tp>
156  constexpr _Tp
157  __rotl(_Tp __x, int __s) noexcept
158  {
159  constexpr auto _Nd = __gnu_cxx::__int_traits<_Tp>::__digits;
160  if _GLIBCXX17_CONSTEXPR ((_Nd & (_Nd - 1)) == 0)
161  {
162  // Variant for power of two _Nd which the compiler can
163  // easily pattern match.
164  constexpr unsigned __uNd = _Nd;
165  const unsigned __r = __s;
166  return (__x << (__r % __uNd)) | (__x >> ((-__r) % __uNd));
167  }
168  const int __r = __s % _Nd;
169  if (__r == 0)
170  return __x;
171  else if (__r > 0)
172  return (__x << __r) | (__x >> ((_Nd - __r) % _Nd));
173  else
174  return (__x >> -__r) | (__x << ((_Nd + __r) % _Nd)); // rotr(x, -r)
175  }
176 
177  template<typename _Tp>
178  constexpr _Tp
179  __rotr(_Tp __x, int __s) noexcept
180  {
181  constexpr auto _Nd = __gnu_cxx::__int_traits<_Tp>::__digits;
182  if _GLIBCXX17_CONSTEXPR ((_Nd & (_Nd - 1)) == 0)
183  {
184  // Variant for power of two _Nd which the compiler can
185  // easily pattern match.
186  constexpr unsigned __uNd = _Nd;
187  const unsigned __r = __s;
188  return (__x >> (__r % __uNd)) | (__x << ((-__r) % __uNd));
189  }
190  const int __r = __s % _Nd;
191  if (__r == 0)
192  return __x;
193  else if (__r > 0)
194  return (__x >> __r) | (__x << ((_Nd - __r) % _Nd));
195  else
196  return (__x << -__r) | (__x >> ((_Nd + __r) % _Nd)); // rotl(x, -r)
197  }
198 
199  template<typename _Tp>
200  constexpr int
201  __countl_zero(_Tp __x) noexcept
202  {
203  using __gnu_cxx::__int_traits;
204  constexpr auto _Nd = __int_traits<_Tp>::__digits;
205 
206  if (__x == 0)
207  return _Nd;
208 
209  constexpr auto _Nd_ull = __int_traits<unsigned long long>::__digits;
210  constexpr auto _Nd_ul = __int_traits<unsigned long>::__digits;
211  constexpr auto _Nd_u = __int_traits<unsigned>::__digits;
212 
213  if _GLIBCXX17_CONSTEXPR (_Nd <= _Nd_u)
214  {
215  constexpr int __diff = _Nd_u - _Nd;
216  return __builtin_clz(__x) - __diff;
217  }
218  else if _GLIBCXX17_CONSTEXPR (_Nd <= _Nd_ul)
219  {
220  constexpr int __diff = _Nd_ul - _Nd;
221  return __builtin_clzl(__x) - __diff;
222  }
223  else if _GLIBCXX17_CONSTEXPR (_Nd <= _Nd_ull)
224  {
225  constexpr int __diff = _Nd_ull - _Nd;
226  return __builtin_clzll(__x) - __diff;
227  }
228  else // (_Nd > _Nd_ull)
229  {
230  static_assert(_Nd <= (2 * _Nd_ull),
231  "Maximum supported integer size is 128-bit");
232 
233  unsigned long long __high = __x >> _Nd_ull;
234  if (__high != 0)
235  {
236  constexpr int __diff = (2 * _Nd_ull) - _Nd;
237  return __builtin_clzll(__high) - __diff;
238  }
239  constexpr auto __max_ull = __int_traits<unsigned long long>::__max;
240  unsigned long long __low = __x & __max_ull;
241  return (_Nd - _Nd_ull) + __builtin_clzll(__low);
242  }
243  }
244 
245  template<typename _Tp>
246  constexpr int
247  __countl_one(_Tp __x) noexcept
248  {
249  return std::__countl_zero<_Tp>((_Tp)~__x);
250  }
251 
252  template<typename _Tp>
253  constexpr int
254  __countr_zero(_Tp __x) noexcept
255  {
256  using __gnu_cxx::__int_traits;
257  constexpr auto _Nd = __int_traits<_Tp>::__digits;
258 
259  if (__x == 0)
260  return _Nd;
261 
262  constexpr auto _Nd_ull = __int_traits<unsigned long long>::__digits;
263  constexpr auto _Nd_ul = __int_traits<unsigned long>::__digits;
264  constexpr auto _Nd_u = __int_traits<unsigned>::__digits;
265 
266  if _GLIBCXX17_CONSTEXPR (_Nd <= _Nd_u)
267  return __builtin_ctz(__x);
268  else if _GLIBCXX17_CONSTEXPR (_Nd <= _Nd_ul)
269  return __builtin_ctzl(__x);
270  else if _GLIBCXX17_CONSTEXPR (_Nd <= _Nd_ull)
271  return __builtin_ctzll(__x);
272  else // (_Nd > _Nd_ull)
273  {
274  static_assert(_Nd <= (2 * _Nd_ull),
275  "Maximum supported integer size is 128-bit");
276 
277  constexpr auto __max_ull = __int_traits<unsigned long long>::__max;
278  unsigned long long __low = __x & __max_ull;
279  if (__low != 0)
280  return __builtin_ctzll(__low);
281  unsigned long long __high = __x >> _Nd_ull;
282  return __builtin_ctzll(__high) + _Nd_ull;
283  }
284  }
285 
286  template<typename _Tp>
287  constexpr int
288  __countr_one(_Tp __x) noexcept
289  {
290  return std::__countr_zero((_Tp)~__x);
291  }
292 
293  template<typename _Tp>
294  constexpr int
295  __popcount(_Tp __x) noexcept
296  {
297  using __gnu_cxx::__int_traits;
298  constexpr auto _Nd = __int_traits<_Tp>::__digits;
299 
300  constexpr auto _Nd_ull = __int_traits<unsigned long long>::__digits;
301  constexpr auto _Nd_ul = __int_traits<unsigned long>::__digits;
302  constexpr auto _Nd_u = __int_traits<unsigned>::__digits;
303 
304  if _GLIBCXX17_CONSTEXPR (_Nd <= _Nd_u)
305  return __builtin_popcount(__x);
306  else if _GLIBCXX17_CONSTEXPR (_Nd <= _Nd_ul)
307  return __builtin_popcountl(__x);
308  else if _GLIBCXX17_CONSTEXPR (_Nd <= _Nd_ull)
309  return __builtin_popcountll(__x);
310  else // (_Nd > _Nd_ull)
311  {
312  static_assert(_Nd <= (2 * _Nd_ull),
313  "Maximum supported integer size is 128-bit");
314 
315  constexpr auto __max_ull = __int_traits<unsigned long long>::__max;
316  unsigned long long __low = __x & __max_ull;
317  unsigned long long __high = __x >> _Nd_ull;
318  return __builtin_popcountll(__low) + __builtin_popcountll(__high);
319  }
320  }
321 
322  template<typename _Tp>
323  constexpr bool
324  __has_single_bit(_Tp __x) noexcept
325  { return std::__popcount(__x) == 1; }
326 
327  template<typename _Tp>
328  constexpr _Tp
329  __bit_ceil(_Tp __x) noexcept
330  {
331  using __gnu_cxx::__int_traits;
332  constexpr auto _Nd = __int_traits<_Tp>::__digits;
333  if (__x == 0 || __x == 1)
334  return 1;
335  auto __shift_exponent = _Nd - std::__countl_zero((_Tp)(__x - 1u));
336  // If the shift exponent equals _Nd then the correct result is not
337  // representable as a value of _Tp, and so the result is undefined.
338  // Want that undefined behaviour to be detected in constant expressions,
339  // by UBSan, and by debug assertions.
340  if (!std::__is_constant_evaluated())
341  {
342  __glibcxx_assert( __shift_exponent != __int_traits<_Tp>::__digits );
343  }
344 
345  using __promoted_type = decltype(__x << 1);
346  if _GLIBCXX17_CONSTEXPR (!is_same<__promoted_type, _Tp>::value)
347  {
348  // If __x undergoes integral promotion then shifting by _Nd is
349  // not undefined. In order to make the shift undefined, so that
350  // it is diagnosed in constant expressions and by UBsan, we also
351  // need to "promote" the shift exponent to be too large for the
352  // promoted type.
353  const int __extra_exp = sizeof(__promoted_type) / sizeof(_Tp) / 2;
354  __shift_exponent |= (__shift_exponent & _Nd) << __extra_exp;
355  }
356  return (_Tp)1u << __shift_exponent;
357  }
358 
359  template<typename _Tp>
360  constexpr _Tp
361  __bit_floor(_Tp __x) noexcept
362  {
363  constexpr auto _Nd = __gnu_cxx::__int_traits<_Tp>::__digits;
364  if (__x == 0)
365  return 0;
366  return (_Tp)1u << (_Nd - std::__countl_zero((_Tp)(__x >> 1)));
367  }
368 
369  template<typename _Tp>
370  constexpr int
371  __bit_width(_Tp __x) noexcept
372  {
373  constexpr auto _Nd = __gnu_cxx::__int_traits<_Tp>::__digits;
374  return _Nd - std::__countl_zero(__x);
375  }
376 
377  /// @endcond
378 
379 #ifdef __cpp_lib_bitops // C++ >= 20
380 
381  /// @cond undocumented
382  template<typename _Tp>
383  concept __unsigned_integer = __is_unsigned_integer<_Tp>::value;
384  /// @endcond
385 
386  // [bit.rot], rotating
387 
388  /// Rotate `x` to the left by `s` bits.
389  template<__unsigned_integer _Tp>
390  [[nodiscard]] constexpr _Tp
391  rotl(_Tp __x, int __s) noexcept
392  { return std::__rotl(__x, __s); }
393 
394  /// Rotate `x` to the right by `s` bits.
395  template<__unsigned_integer _Tp>
396  [[nodiscard]] constexpr _Tp
397  rotr(_Tp __x, int __s) noexcept
398  { return std::__rotr(__x, __s); }
399 
400  // [bit.count], counting
401 
402  /// The number of contiguous zero bits, starting from the highest bit.
403  template<__unsigned_integer _Tp>
404  constexpr int
405  countl_zero(_Tp __x) noexcept
406  { return std::__countl_zero(__x); }
407 
408  /// The number of contiguous one bits, starting from the highest bit.
409  template<__unsigned_integer _Tp>
410  constexpr int
411  countl_one(_Tp __x) noexcept
412  { return std::__countl_one(__x); }
413 
414  /// The number of contiguous zero bits, starting from the lowest bit.
415  template<__unsigned_integer _Tp>
416  constexpr int
417  countr_zero(_Tp __x) noexcept
418  { return std::__countr_zero(__x); }
419 
420  /// The number of contiguous one bits, starting from the lowest bit.
421  template<__unsigned_integer _Tp>
422  constexpr int
423  countr_one(_Tp __x) noexcept
424  { return std::__countr_one(__x); }
425 
426  /// The number of bits set in `x`.
427  template<__unsigned_integer _Tp>
428  constexpr int
429  popcount(_Tp __x) noexcept
430  { return std::__popcount(__x); }
431 #endif // __cpp_lib_bitops
432 
433 #ifdef __cpp_lib_int_pow2 // C++ >= 20
434  // [bit.pow.two], integral powers of 2
435 
436  /// True if `x` is a power of two, false otherwise.
437  template<__unsigned_integer _Tp>
438  constexpr bool
439  has_single_bit(_Tp __x) noexcept
440  { return std::__has_single_bit(__x); }
441 
442  /// The smallest power-of-two not less than `x`.
443  template<__unsigned_integer _Tp>
444  constexpr _Tp
445  bit_ceil(_Tp __x) noexcept
446  { return std::__bit_ceil(__x); }
447 
448  /// The largest power-of-two not greater than `x`.
449  template<__unsigned_integer _Tp>
450  constexpr _Tp
451  bit_floor(_Tp __x) noexcept
452  { return std::__bit_floor(__x); }
453 
454  // _GLIBCXX_RESOLVE_LIB_DEFECTS
455  // 3656. Inconsistent bit operations returning a count
456  /// The smallest integer greater than the base-2 logarithm of `x`.
457  template<__unsigned_integer _Tp>
458  constexpr int
459  bit_width(_Tp __x) noexcept
460  { return std::__bit_width(__x); }
461 #endif // defined (__cpp_lib_int_pow2)
462 
463 #ifdef __cpp_lib_endian // C++ >= 20
464 
465  /// Byte order constants
466  /**
467  * The platform endianness can be checked by comparing `std::endian::native`
468  * to one of `std::endian::big` or `std::endian::little`.
469  *
470  * @since C++20
471  */
472  enum class endian
473  {
474  little = __ORDER_LITTLE_ENDIAN__,
475  big = __ORDER_BIG_ENDIAN__,
476  native = __BYTE_ORDER__
477  };
478 #endif // __cpp_lib_endian
479 
480  /// @}
481 
482 _GLIBCXX_END_NAMESPACE_VERSION
483 } // namespace std
484 
485 #endif // C++14
486 #endif // _GLIBCXX_BIT