Freeze libcxx for cuda without c++20 support

commit_hash:c2eb32b5cb16638c17c6fc6900738f26ba6eca49

1 files changed, 1103 insertions, 0 deletions

diff --git a/contrib/libs/cxxsupp/libcxxcuda11/include/__bit_reference b/contrib/libs/cxxsupp/libcxxcuda11/include/__bit_reference
new file mode 100644
index 00000000000..107368759c6
--- /dev/null
+++ b/contrib/libs/cxxsupp/libcxxcuda11/include/__bit_reference
@@ -0,0 +1,1103 @@
+// -*- C++ -*-
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCPP___BIT_REFERENCE
+#define _LIBCPP___BIT_REFERENCE
+
+#include <__algorithm/copy_n.h>
+#include <__algorithm/fill_n.h>
+#include <__algorithm/min.h>
+#include <__bit/countr.h>
+#include <__bit/invert_if.h>
+#include <__bit/popcount.h>
+#include <__config>
+#include <__fwd/bit_reference.h>
+#include <__iterator/iterator_traits.h>
+#include <__memory/construct_at.h>
+#include <__memory/pointer_traits.h>
+#include <__type_traits/conditional.h>
+#include <__utility/swap.h>
+#include <cstring>
+
+#if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
+#  pragma GCC system_header
+#endif
+
+_LIBCPP_PUSH_MACROS
+#include <__undef_macros>
+
+_LIBCPP_BEGIN_NAMESPACE_STD
+
+template <class _Cp>
+class __bit_const_reference;
+
+template <class _Tp>
+struct __has_storage_type {
+  static const bool value = false;
+};
+
+template <class _Cp, bool = __has_storage_type<_Cp>::value>
+class __bit_reference {
+  using __storage_type    = typename _Cp::__storage_type;
+  using __storage_pointer = typename _Cp::__storage_pointer;
+
+  __storage_pointer __seg_;
+  __storage_type __mask_;
+
+  friend typename _Cp::__self;
+
+  friend class __bit_const_reference<_Cp>;
+  friend class __bit_iterator<_Cp, false>;
+
+public:
+  using __container = typename _Cp::__self;
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 __bit_reference(const __bit_reference&) = default;
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 operator bool() const _NOEXCEPT {
+    return static_cast<bool>(*__seg_ & __mask_);
+  }
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 bool operator~() const _NOEXCEPT {
+    return !static_cast<bool>(*this);
+  }
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 __bit_reference& operator=(bool __x) _NOEXCEPT {
+    if (__x)
+      *__seg_ |= __mask_;
+    else
+      *__seg_ &= ~__mask_;
+    return *this;
+  }
+
+#if _LIBCPP_STD_VER >= 23
+  _LIBCPP_HIDE_FROM_ABI constexpr const __bit_reference& operator=(bool __x) const noexcept {
+    if (__x)
+      *__seg_ |= __mask_;
+    else
+      *__seg_ &= ~__mask_;
+    return *this;
+  }
+#endif
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 __bit_reference& operator=(const __bit_reference& __x) _NOEXCEPT {
+    return operator=(static_cast<bool>(__x));
+  }
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 void flip() _NOEXCEPT { *__seg_ ^= __mask_; }
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 __bit_iterator<_Cp, false> operator&() const _NOEXCEPT {
+    return __bit_iterator<_Cp, false>(__seg_, static_cast<unsigned>(std::__libcpp_ctz(__mask_)));
+  }
+
+private:
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 explicit __bit_reference(
+      __storage_pointer __s, __storage_type __m) _NOEXCEPT
+      : __seg_(__s),
+        __mask_(__m) {}
+};
+
+template <class _Cp>
+class __bit_reference<_Cp, false> {};
+
+template <class _Cp>
+inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 void
+swap(__bit_reference<_Cp> __x, __bit_reference<_Cp> __y) _NOEXCEPT {
+  bool __t = __x;
+  __x      = __y;
+  __y      = __t;
+}
+
+template <class _Cp, class _Dp>
+inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 void
+swap(__bit_reference<_Cp> __x, __bit_reference<_Dp> __y) _NOEXCEPT {
+  bool __t = __x;
+  __x      = __y;
+  __y      = __t;
+}
+
+template <class _Cp>
+inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 void swap(__bit_reference<_Cp> __x, bool& __y) _NOEXCEPT {
+  bool __t = __x;
+  __x      = __y;
+  __y      = __t;
+}
+
+template <class _Cp>
+inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 void swap(bool& __x, __bit_reference<_Cp> __y) _NOEXCEPT {
+  bool __t = __x;
+  __x      = __y;
+  __y      = __t;
+}
+
+template <class _Cp>
+class __bit_const_reference {
+  using __storage_type    = typename _Cp::__storage_type;
+  using __storage_pointer = typename _Cp::__const_storage_pointer;
+
+  __storage_pointer __seg_;
+  __storage_type __mask_;
+
+  friend typename _Cp::__self;
+  friend class __bit_iterator<_Cp, true>;
+
+public:
+  using __container = typename _Cp::__self;
+
+  _LIBCPP_HIDE_FROM_ABI __bit_const_reference(const __bit_const_reference&) = default;
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 __bit_const_reference(const __bit_reference<_Cp>& __x) _NOEXCEPT
+      : __seg_(__x.__seg_),
+        __mask_(__x.__mask_) {}
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR operator bool() const _NOEXCEPT {
+    return static_cast<bool>(*__seg_ & __mask_);
+  }
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 __bit_iterator<_Cp, true> operator&() const _NOEXCEPT {
+    return __bit_iterator<_Cp, true>(__seg_, static_cast<unsigned>(std::__libcpp_ctz(__mask_)));
+  }
+
+private:
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR explicit __bit_const_reference(
+      __storage_pointer __s, __storage_type __m) _NOEXCEPT
+      : __seg_(__s),
+        __mask_(__m) {}
+
+  __bit_const_reference& operator=(const __bit_const_reference&) = delete;
+};
+
+// count
+
+template <bool _ToCount, class _Cp, bool _IsConst>
+_LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX23 typename __bit_iterator<_Cp, _IsConst>::difference_type
+__count_bool(__bit_iterator<_Cp, _IsConst> __first, typename _Cp::size_type __n) {
+  using _It             = __bit_iterator<_Cp, _IsConst>;
+  using __storage_type  = typename _It::__storage_type;
+  using difference_type = typename _It::difference_type;
+
+  const int __bits_per_word = _It::__bits_per_word;
+  difference_type __r       = 0;
+  // do first partial word
+  if (__first.__ctz_ != 0) {
+    __storage_type __clz_f = static_cast<__storage_type>(__bits_per_word - __first.__ctz_);
+    __storage_type __dn    = std::min(__clz_f, __n);
+    __storage_type __m     = (~__storage_type(0) << __first.__ctz_) & (~__storage_type(0) >> (__clz_f - __dn));
+    __r                    = std::__libcpp_popcount(std::__invert_if<!_ToCount>(*__first.__seg_) & __m);
+    __n -= __dn;
+    ++__first.__seg_;
+  }
+  // do middle whole words
+  for (; __n >= __bits_per_word; ++__first.__seg_, __n -= __bits_per_word)
+    __r += std::__libcpp_popcount(std::__invert_if<!_ToCount>(*__first.__seg_));
+  // do last partial word
+  if (__n > 0) {
+    __storage_type __m = ~__storage_type(0) >> (__bits_per_word - __n);
+    __r += std::__libcpp_popcount(std::__invert_if<!_ToCount>(*__first.__seg_) & __m);
+  }
+  return __r;
+}
+
+template <class _Cp, bool _IsConst, class _Tp>
+inline _LIBCPP_HIDE_FROM_ABI typename __bit_iterator<_Cp, _IsConst>::difference_type
+count(__bit_iterator<_Cp, _IsConst> __first, __bit_iterator<_Cp, _IsConst> __last, const _Tp& __value) {
+  if (static_cast<bool>(__value))
+    return std::__count_bool<true>(__first, static_cast<typename _Cp::size_type>(__last - __first));
+  return std::__count_bool<false>(__first, static_cast<typename _Cp::size_type>(__last - __first));
+}
+
+// fill_n
+
+template <bool _FillValue, class _Cp>
+_LIBCPP_CONSTEXPR_SINCE_CXX20 _LIBCPP_HIDE_FROM_ABI void
+__fill_n(__bit_iterator<_Cp, false> __first, typename _Cp::size_type __n) {
+  using _It            = __bit_iterator<_Cp, false>;
+  using __storage_type = typename _It::__storage_type;
+
+  const int __bits_per_word = _It::__bits_per_word;
+  // do first partial word
+  if (__first.__ctz_ != 0) {
+    __storage_type __clz_f = static_cast<__storage_type>(__bits_per_word - __first.__ctz_);
+    __storage_type __dn    = std::min(__clz_f, __n);
+    __storage_type __m     = (~__storage_type(0) << __first.__ctz_) & (~__storage_type(0) >> (__clz_f - __dn));
+    if (_FillValue)
+      *__first.__seg_ |= __m;
+    else
+      *__first.__seg_ &= ~__m;
+    __n -= __dn;
+    ++__first.__seg_;
+  }
+  // do middle whole words
+  __storage_type __nw = __n / __bits_per_word;
+  std::fill_n(std::__to_address(__first.__seg_), __nw, _FillValue ? static_cast<__storage_type>(-1) : 0);
+  __n -= __nw * __bits_per_word;
+  // do last partial word
+  if (__n > 0) {
+    __first.__seg_ += __nw;
+    __storage_type __m = ~__storage_type(0) >> (__bits_per_word - __n);
+    if (_FillValue)
+      *__first.__seg_ |= __m;
+    else
+      *__first.__seg_ &= ~__m;
+  }
+}
+
+template <class _Cp>
+inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 void
+fill_n(__bit_iterator<_Cp, false> __first, typename _Cp::size_type __n, bool __value) {
+  if (__n > 0) {
+    if (__value)
+      std::__fill_n<true>(__first, __n);
+    else
+      std::__fill_n<false>(__first, __n);
+  }
+}
+
+// fill
+
+template <class _Cp>
+inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 void
+fill(__bit_iterator<_Cp, false> __first, __bit_iterator<_Cp, false> __last, bool __value) {
+  std::fill_n(__first, static_cast<typename _Cp::size_type>(__last - __first), __value);
+}
+
+// copy
+
+template <class _Cp, bool _IsConst>
+_LIBCPP_CONSTEXPR_SINCE_CXX20 _LIBCPP_HIDE_FROM_ABI __bit_iterator<_Cp, false> __copy_aligned(
+    __bit_iterator<_Cp, _IsConst> __first, __bit_iterator<_Cp, _IsConst> __last, __bit_iterator<_Cp, false> __result) {
+  using _In             = __bit_iterator<_Cp, _IsConst>;
+  using difference_type = typename _In::difference_type;
+  using __storage_type  = typename _In::__storage_type;
+
+  const int __bits_per_word = _In::__bits_per_word;
+  difference_type __n       = __last - __first;
+  if (__n > 0) {
+    // do first word
+    if (__first.__ctz_ != 0) {
+      unsigned __clz       = __bits_per_word - __first.__ctz_;
+      difference_type __dn = std::min(static_cast<difference_type>(__clz), __n);
+      __n -= __dn;
+      __storage_type __m = (~__storage_type(0) << __first.__ctz_) & (~__storage_type(0) >> (__clz - __dn));
+      __storage_type __b = *__first.__seg_ & __m;
+      *__result.__seg_ &= ~__m;
+      *__result.__seg_ |= __b;
+      __result.__seg_ += (__dn + __result.__ctz_) / __bits_per_word;
+      __result.__ctz_ = static_cast<unsigned>((__dn + __result.__ctz_) % __bits_per_word);
+      ++__first.__seg_;
+      // __first.__ctz_ = 0;
+    }
+    // __first.__ctz_ == 0;
+    // do middle words
+    __storage_type __nw = __n / __bits_per_word;
+    std::copy_n(std::__to_address(__first.__seg_), __nw, std::__to_address(__result.__seg_));
+    __n -= __nw * __bits_per_word;
+    __result.__seg_ += __nw;
+    // do last word
+    if (__n > 0) {
+      __first.__seg_ += __nw;
+      __storage_type __m = ~__storage_type(0) >> (__bits_per_word - __n);
+      __storage_type __b = *__first.__seg_ & __m;
+      *__result.__seg_ &= ~__m;
+      *__result.__seg_ |= __b;
+      __result.__ctz_ = static_cast<unsigned>(__n);
+    }
+  }
+  return __result;
+}
+
+template <class _Cp, bool _IsConst>
+_LIBCPP_CONSTEXPR_SINCE_CXX20 _LIBCPP_HIDE_FROM_ABI __bit_iterator<_Cp, false> __copy_unaligned(
+    __bit_iterator<_Cp, _IsConst> __first, __bit_iterator<_Cp, _IsConst> __last, __bit_iterator<_Cp, false> __result) {
+  using _In             = __bit_iterator<_Cp, _IsConst>;
+  using difference_type = typename _In::difference_type;
+  using __storage_type  = typename _In::__storage_type;
+
+  const int __bits_per_word = _In::__bits_per_word;
+  difference_type __n       = __last - __first;
+  if (__n > 0) {
+    // do first word
+    if (__first.__ctz_ != 0) {
+      unsigned __clz_f     = __bits_per_word - __first.__ctz_;
+      difference_type __dn = std::min(static_cast<difference_type>(__clz_f), __n);
+      __n -= __dn;
+      __storage_type __m   = (~__storage_type(0) << __first.__ctz_) & (~__storage_type(0) >> (__clz_f - __dn));
+      __storage_type __b   = *__first.__seg_ & __m;
+      unsigned __clz_r     = __bits_per_word - __result.__ctz_;
+      __storage_type __ddn = std::min<__storage_type>(__dn, __clz_r);
+      __m                  = (~__storage_type(0) << __result.__ctz_) & (~__storage_type(0) >> (__clz_r - __ddn));
+      *__result.__seg_ &= ~__m;
+      if (__result.__ctz_ > __first.__ctz_)
+        *__result.__seg_ |= __b << (__result.__ctz_ - __first.__ctz_);
+      else
+        *__result.__seg_ |= __b >> (__first.__ctz_ - __result.__ctz_);
+      __result.__seg_ += (__ddn + __result.__ctz_) / __bits_per_word;
+      __result.__ctz_ = static_cast<unsigned>((__ddn + __result.__ctz_) % __bits_per_word);
+      __dn -= __ddn;
+      if (__dn > 0) {
+        __m = ~__storage_type(0) >> (__bits_per_word - __dn);
+        *__result.__seg_ &= ~__m;
+        *__result.__seg_ |= __b >> (__first.__ctz_ + __ddn);
+        __result.__ctz_ = static_cast<unsigned>(__dn);
+      }
+      ++__first.__seg_;
+      // __first.__ctz_ = 0;
+    }
+    // __first.__ctz_ == 0;
+    // do middle words
+    unsigned __clz_r   = __bits_per_word - __result.__ctz_;
+    __storage_type __m = ~__storage_type(0) << __result.__ctz_;
+    for (; __n >= __bits_per_word; __n -= __bits_per_word, ++__first.__seg_) {
+      __storage_type __b = *__first.__seg_;
+      *__result.__seg_ &= ~__m;
+      *__result.__seg_ |= __b << __result.__ctz_;
+      ++__result.__seg_;
+      *__result.__seg_ &= __m;
+      *__result.__seg_ |= __b >> __clz_r;
+    }
+    // do last word
+    if (__n > 0) {
+      __m                 = ~__storage_type(0) >> (__bits_per_word - __n);
+      __storage_type __b  = *__first.__seg_ & __m;
+      __storage_type __dn = std::min(__n, static_cast<difference_type>(__clz_r));
+      __m                 = (~__storage_type(0) << __result.__ctz_) & (~__storage_type(0) >> (__clz_r - __dn));
+      *__result.__seg_ &= ~__m;
+      *__result.__seg_ |= __b << __result.__ctz_;
+      __result.__seg_ += (__dn + __result.__ctz_) / __bits_per_word;
+      __result.__ctz_ = static_cast<unsigned>((__dn + __result.__ctz_) % __bits_per_word);
+      __n -= __dn;
+      if (__n > 0) {
+        __m = ~__storage_type(0) >> (__bits_per_word - __n);
+        *__result.__seg_ &= ~__m;
+        *__result.__seg_ |= __b >> __dn;
+        __result.__ctz_ = static_cast<unsigned>(__n);
+      }
+    }
+  }
+  return __result;
+}
+
+template <class _Cp, bool _IsConst>
+inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 __bit_iterator<_Cp, false>
+copy(__bit_iterator<_Cp, _IsConst> __first, __bit_iterator<_Cp, _IsConst> __last, __bit_iterator<_Cp, false> __result) {
+  if (__first.__ctz_ == __result.__ctz_)
+    return std::__copy_aligned(__first, __last, __result);
+  return std::__copy_unaligned(__first, __last, __result);
+}
+
+// copy_backward
+
+template <class _Cp, bool _IsConst>
+_LIBCPP_CONSTEXPR_SINCE_CXX20 _LIBCPP_HIDE_FROM_ABI __bit_iterator<_Cp, false> __copy_backward_aligned(
+    __bit_iterator<_Cp, _IsConst> __first, __bit_iterator<_Cp, _IsConst> __last, __bit_iterator<_Cp, false> __result) {
+  using _In             = __bit_iterator<_Cp, _IsConst>;
+  using difference_type = typename _In::difference_type;
+  using __storage_type  = typename _In::__storage_type;
+
+  const int __bits_per_word = _In::__bits_per_word;
+  difference_type __n       = __last - __first;
+  if (__n > 0) {
+    // do first word
+    if (__last.__ctz_ != 0) {
+      difference_type __dn = std::min(static_cast<difference_type>(__last.__ctz_), __n);
+      __n -= __dn;
+      unsigned __clz     = __bits_per_word - __last.__ctz_;
+      __storage_type __m = (~__storage_type(0) << (__last.__ctz_ - __dn)) & (~__storage_type(0) >> __clz);
+      __storage_type __b = *__last.__seg_ & __m;
+      *__result.__seg_ &= ~__m;
+      *__result.__seg_ |= __b;
+      __result.__ctz_ = static_cast<unsigned>(((-__dn & (__bits_per_word - 1)) + __result.__ctz_) % __bits_per_word);
+      // __last.__ctz_ = 0
+    }
+    // __last.__ctz_ == 0 || __n == 0
+    // __result.__ctz_ == 0 || __n == 0
+    // do middle words
+    __storage_type __nw = __n / __bits_per_word;
+    __result.__seg_ -= __nw;
+    __last.__seg_ -= __nw;
+    std::copy_n(std::__to_address(__last.__seg_), __nw, std::__to_address(__result.__seg_));
+    __n -= __nw * __bits_per_word;
+    // do last word
+    if (__n > 0) {
+      __storage_type __m = ~__storage_type(0) << (__bits_per_word - __n);
+      __storage_type __b = *--__last.__seg_ & __m;
+      *--__result.__seg_ &= ~__m;
+      *__result.__seg_ |= __b;
+      __result.__ctz_ = static_cast<unsigned>(-__n & (__bits_per_word - 1));
+    }
+  }
+  return __result;
+}
+
+template <class _Cp, bool _IsConst>
+_LIBCPP_CONSTEXPR_SINCE_CXX20 _LIBCPP_HIDE_FROM_ABI __bit_iterator<_Cp, false> __copy_backward_unaligned(
+    __bit_iterator<_Cp, _IsConst> __first, __bit_iterator<_Cp, _IsConst> __last, __bit_iterator<_Cp, false> __result) {
+  using _In             = __bit_iterator<_Cp, _IsConst>;
+  using difference_type = typename _In::difference_type;
+  using __storage_type  = typename _In::__storage_type;
+
+  const int __bits_per_word = _In::__bits_per_word;
+  difference_type __n       = __last - __first;
+  if (__n > 0) {
+    // do first word
+    if (__last.__ctz_ != 0) {
+      difference_type __dn = std::min(static_cast<difference_type>(__last.__ctz_), __n);
+      __n -= __dn;
+      unsigned __clz_l     = __bits_per_word - __last.__ctz_;
+      __storage_type __m   = (~__storage_type(0) << (__last.__ctz_ - __dn)) & (~__storage_type(0) >> __clz_l);
+      __storage_type __b   = *__last.__seg_ & __m;
+      unsigned __clz_r     = __bits_per_word - __result.__ctz_;
+      __storage_type __ddn = std::min(__dn, static_cast<difference_type>(__result.__ctz_));
+      if (__ddn > 0) {
+        __m = (~__storage_type(0) << (__result.__ctz_ - __ddn)) & (~__storage_type(0) >> __clz_r);
+        *__result.__seg_ &= ~__m;
+        if (__result.__ctz_ > __last.__ctz_)
+          *__result.__seg_ |= __b << (__result.__ctz_ - __last.__ctz_);
+        else
+          *__result.__seg_ |= __b >> (__last.__ctz_ - __result.__ctz_);
+        __result.__ctz_ = static_cast<unsigned>(((-__ddn & (__bits_per_word - 1)) + __result.__ctz_) % __bits_per_word);
+        __dn -= __ddn;
+      }
+      if (__dn > 0) {
+        // __result.__ctz_ == 0
+        --__result.__seg_;
+        __result.__ctz_ = static_cast<unsigned>(-__dn & (__bits_per_word - 1));
+        __m             = ~__storage_type(0) << __result.__ctz_;
+        *__result.__seg_ &= ~__m;
+        __last.__ctz_ -= __dn + __ddn;
+        *__result.__seg_ |= __b << (__result.__ctz_ - __last.__ctz_);
+      }
+      // __last.__ctz_ = 0
+    }
+    // __last.__ctz_ == 0 || __n == 0
+    // __result.__ctz_ != 0 || __n == 0
+    // do middle words
+    unsigned __clz_r   = __bits_per_word - __result.__ctz_;
+    __storage_type __m = ~__storage_type(0) >> __clz_r;
+    for (; __n >= __bits_per_word; __n -= __bits_per_word) {
+      __storage_type __b = *--__last.__seg_;
+      *__result.__seg_ &= ~__m;
+      *__result.__seg_ |= __b >> __clz_r;
+      *--__result.__seg_ &= __m;
+      *__result.__seg_ |= __b << __result.__ctz_;
+    }
+    // do last word
+    if (__n > 0) {
+      __m                 = ~__storage_type(0) << (__bits_per_word - __n);
+      __storage_type __b  = *--__last.__seg_ & __m;
+      __clz_r             = __bits_per_word - __result.__ctz_;
+      __storage_type __dn = std::min(__n, static_cast<difference_type>(__result.__ctz_));
+      __m                 = (~__storage_type(0) << (__result.__ctz_ - __dn)) & (~__storage_type(0) >> __clz_r);
+      *__result.__seg_ &= ~__m;
+      *__result.__seg_ |= __b >> (__bits_per_word - __result.__ctz_);
+      __result.__ctz_ = static_cast<unsigned>(((-__dn & (__bits_per_word - 1)) + __result.__ctz_) % __bits_per_word);
+      __n -= __dn;
+      if (__n > 0) {
+        // __result.__ctz_ == 0
+        --__result.__seg_;
+        __result.__ctz_ = static_cast<unsigned>(-__n & (__bits_per_word - 1));
+        __m             = ~__storage_type(0) << __result.__ctz_;
+        *__result.__seg_ &= ~__m;
+        *__result.__seg_ |= __b << (__result.__ctz_ - (__bits_per_word - __n - __dn));
+      }
+    }
+  }
+  return __result;
+}
+
+template <class _Cp, bool _IsConst>
+inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 __bit_iterator<_Cp, false> copy_backward(
+    __bit_iterator<_Cp, _IsConst> __first, __bit_iterator<_Cp, _IsConst> __last, __bit_iterator<_Cp, false> __result) {
+  if (__last.__ctz_ == __result.__ctz_)
+    return std::__copy_backward_aligned(__first, __last, __result);
+  return std::__copy_backward_unaligned(__first, __last, __result);
+}
+
+// move
+
+template <class _Cp, bool _IsConst>
+inline _LIBCPP_HIDE_FROM_ABI __bit_iterator<_Cp, false>
+move(__bit_iterator<_Cp, _IsConst> __first, __bit_iterator<_Cp, _IsConst> __last, __bit_iterator<_Cp, false> __result) {
+  return std::copy(__first, __last, __result);
+}
+
+// move_backward
+
+template <class _Cp, bool _IsConst>
+inline _LIBCPP_HIDE_FROM_ABI __bit_iterator<_Cp, false> move_backward(
+    __bit_iterator<_Cp, _IsConst> __first, __bit_iterator<_Cp, _IsConst> __last, __bit_iterator<_Cp, false> __result) {
+  return std::copy_backward(__first, __last, __result);
+}
+
+// swap_ranges
+
+template <class _Cl, class _Cr>
+_LIBCPP_HIDE_FROM_ABI __bit_iterator<_Cr, false> __swap_ranges_aligned(
+    __bit_iterator<_Cl, false> __first, __bit_iterator<_Cl, false> __last, __bit_iterator<_Cr, false> __result) {
+  using _I1             = __bit_iterator<_Cl, false>;
+  using difference_type = typename _I1::difference_type;
+  using __storage_type  = typename _I1::__storage_type;
+
+  const int __bits_per_word = _I1::__bits_per_word;
+  difference_type __n       = __last - __first;
+  if (__n > 0) {
+    // do first word
+    if (__first.__ctz_ != 0) {
+      unsigned __clz       = __bits_per_word - __first.__ctz_;
+      difference_type __dn = std::min(static_cast<difference_type>(__clz), __n);
+      __n -= __dn;
+      __storage_type __m  = (~__storage_type(0) << __first.__ctz_) & (~__storage_type(0) >> (__clz - __dn));
+      __storage_type __b1 = *__first.__seg_ & __m;
+      *__first.__seg_ &= ~__m;
+      __storage_type __b2 = *__result.__seg_ & __m;
+      *__result.__seg_ &= ~__m;
+      *__result.__seg_ |= __b1;
+      *__first.__seg_ |= __b2;
+      __result.__seg_ += (__dn + __result.__ctz_) / __bits_per_word;
+      __result.__ctz_ = static_cast<unsigned>((__dn + __result.__ctz_) % __bits_per_word);
+      ++__first.__seg_;
+      // __first.__ctz_ = 0;
+    }
+    // __first.__ctz_ == 0;
+    // do middle words
+    for (; __n >= __bits_per_word; __n -= __bits_per_word, ++__first.__seg_, ++__result.__seg_)
+      swap(*__first.__seg_, *__result.__seg_);
+    // do last word
+    if (__n > 0) {
+      __storage_type __m  = ~__storage_type(0) >> (__bits_per_word - __n);
+      __storage_type __b1 = *__first.__seg_ & __m;
+      *__first.__seg_ &= ~__m;
+      __storage_type __b2 = *__result.__seg_ & __m;
+      *__result.__seg_ &= ~__m;
+      *__result.__seg_ |= __b1;
+      *__first.__seg_ |= __b2;
+      __result.__ctz_ = static_cast<unsigned>(__n);
+    }
+  }
+  return __result;
+}
+
+template <class _Cl, class _Cr>
+_LIBCPP_HIDE_FROM_ABI __bit_iterator<_Cr, false> __swap_ranges_unaligned(
+    __bit_iterator<_Cl, false> __first, __bit_iterator<_Cl, false> __last, __bit_iterator<_Cr, false> __result) {
+  using _I1             = __bit_iterator<_Cl, false>;
+  using difference_type = typename _I1::difference_type;
+  using __storage_type  = typename _I1::__storage_type;
+
+  const int __bits_per_word = _I1::__bits_per_word;
+  difference_type __n       = __last - __first;
+  if (__n > 0) {
+    // do first word
+    if (__first.__ctz_ != 0) {
+      unsigned __clz_f     = __bits_per_word - __first.__ctz_;
+      difference_type __dn = std::min(static_cast<difference_type>(__clz_f), __n);
+      __n -= __dn;
+      __storage_type __m  = (~__storage_type(0) << __first.__ctz_) & (~__storage_type(0) >> (__clz_f - __dn));
+      __storage_type __b1 = *__first.__seg_ & __m;
+      *__first.__seg_ &= ~__m;
+      unsigned __clz_r     = __bits_per_word - __result.__ctz_;
+      __storage_type __ddn = std::min<__storage_type>(__dn, __clz_r);
+      __m                  = (~__storage_type(0) << __result.__ctz_) & (~__storage_type(0) >> (__clz_r - __ddn));
+      __storage_type __b2  = *__result.__seg_ & __m;
+      *__result.__seg_ &= ~__m;
+      if (__result.__ctz_ > __first.__ctz_) {
+        unsigned __s = __result.__ctz_ - __first.__ctz_;
+        *__result.__seg_ |= __b1 << __s;
+        *__first.__seg_ |= __b2 >> __s;
+      } else {
+        unsigned __s = __first.__ctz_ - __result.__ctz_;
+        *__result.__seg_ |= __b1 >> __s;
+        *__first.__seg_ |= __b2 << __s;
+      }
+      __result.__seg_ += (__ddn + __result.__ctz_) / __bits_per_word;
+      __result.__ctz_ = static_cast<unsigned>((__ddn + __result.__ctz_) % __bits_per_word);
+      __dn -= __ddn;
+      if (__dn > 0) {
+        __m  = ~__storage_type(0) >> (__bits_per_word - __dn);
+        __b2 = *__result.__seg_ & __m;
+        *__result.__seg_ &= ~__m;
+        unsigned __s = __first.__ctz_ + __ddn;
+        *__result.__seg_ |= __b1 >> __s;
+        *__first.__seg_ |= __b2 << __s;
+        __result.__ctz_ = static_cast<unsigned>(__dn);
+      }
+      ++__first.__seg_;
+      // __first.__ctz_ = 0;
+    }
+    // __first.__ctz_ == 0;
+    // do middle words
+    __storage_type __m = ~__storage_type(0) << __result.__ctz_;
+    unsigned __clz_r   = __bits_per_word - __result.__ctz_;
+    for (; __n >= __bits_per_word; __n -= __bits_per_word, ++__first.__seg_) {
+      __storage_type __b1 = *__first.__seg_;
+      __storage_type __b2 = *__result.__seg_ & __m;
+      *__result.__seg_ &= ~__m;
+      *__result.__seg_ |= __b1 << __result.__ctz_;
+      *__first.__seg_ = __b2 >> __result.__ctz_;
+      ++__result.__seg_;
+      __b2 = *__result.__seg_ & ~__m;
+      *__result.__seg_ &= __m;
+      *__result.__seg_ |= __b1 >> __clz_r;
+      *__first.__seg_ |= __b2 << __clz_r;
+    }
+    // do last word
+    if (__n > 0) {
+      __m                 = ~__storage_type(0) >> (__bits_per_word - __n);
+      __storage_type __b1 = *__first.__seg_ & __m;
+      *__first.__seg_ &= ~__m;
+      __storage_type __dn = std::min<__storage_type>(__n, __clz_r);
+      __m                 = (~__storage_type(0) << __result.__ctz_) & (~__storage_type(0) >> (__clz_r - __dn));
+      __storage_type __b2 = *__result.__seg_ & __m;
+      *__result.__seg_ &= ~__m;
+      *__result.__seg_ |= __b1 << __result.__ctz_;
+      *__first.__seg_ |= __b2 >> __result.__ctz_;
+      __result.__seg_ += (__dn + __result.__ctz_) / __bits_per_word;
+      __result.__ctz_ = static_cast<unsigned>((__dn + __result.__ctz_) % __bits_per_word);
+      __n -= __dn;
+      if (__n > 0) {
+        __m  = ~__storage_type(0) >> (__bits_per_word - __n);
+        __b2 = *__result.__seg_ & __m;
+        *__result.__seg_ &= ~__m;
+        *__result.__seg_ |= __b1 >> __dn;
+        *__first.__seg_ |= __b2 << __dn;
+        __result.__ctz_ = static_cast<unsigned>(__n);
+      }
+    }
+  }
+  return __result;
+}
+
+template <class _Cl, class _Cr>
+inline _LIBCPP_HIDE_FROM_ABI __bit_iterator<_Cr, false> swap_ranges(
+    __bit_iterator<_Cl, false> __first1, __bit_iterator<_Cl, false> __last1, __bit_iterator<_Cr, false> __first2) {
+  if (__first1.__ctz_ == __first2.__ctz_)
+    return std::__swap_ranges_aligned(__first1, __last1, __first2);
+  return std::__swap_ranges_unaligned(__first1, __last1, __first2);
+}
+
+// rotate
+
+template <class _Cp>
+struct __bit_array {
+  using difference_type   = typename _Cp::difference_type;
+  using __storage_type    = typename _Cp::__storage_type;
+  using __storage_pointer = typename _Cp::__storage_pointer;
+  using iterator          = typename _Cp::iterator;
+
+  static const unsigned __bits_per_word = _Cp::__bits_per_word;
+  static const unsigned _Np             = 4;
+
+  difference_type __size_;
+  __storage_type __word_[_Np];
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 static difference_type capacity() {
+    return static_cast<difference_type>(_Np * __bits_per_word);
+  }
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 explicit __bit_array(difference_type __s) : __size_(__s) {
+    if (__libcpp_is_constant_evaluated()) {
+      for (size_t __i = 0; __i != __bit_array<_Cp>::_Np; ++__i)
+        std::__construct_at(__word_ + __i, 0);
+    }
+  }
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 iterator begin() {
+    return iterator(pointer_traits<__storage_pointer>::pointer_to(__word_[0]), 0);
+  }
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 iterator end() {
+    return iterator(pointer_traits<__storage_pointer>::pointer_to(__word_[0]) + __size_ / __bits_per_word,
+                    static_cast<unsigned>(__size_ % __bits_per_word));
+  }
+};
+
+template <class _Cp>
+_LIBCPP_CONSTEXPR_SINCE_CXX20 _LIBCPP_HIDE_FROM_ABI __bit_iterator<_Cp, false>
+rotate(__bit_iterator<_Cp, false> __first, __bit_iterator<_Cp, false> __middle, __bit_iterator<_Cp, false> __last) {
+  using _I1             = __bit_iterator<_Cp, false>;
+  using difference_type = typename _I1::difference_type;
+
+  difference_type __d1 = __middle - __first;
+  difference_type __d2 = __last - __middle;
+  _I1 __r              = __first + __d2;
+  while (__d1 != 0 && __d2 != 0) {
+    if (__d1 <= __d2) {
+      if (__d1 <= __bit_array<_Cp>::capacity()) {
+        __bit_array<_Cp> __b(__d1);
+        std::copy(__first, __middle, __b.begin());
+        std::copy(__b.begin(), __b.end(), std::copy(__middle, __last, __first));
+        break;
+      } else {
+        __bit_iterator<_Cp, false> __mp = std::swap_ranges(__first, __middle, __middle);
+        __first                         = __middle;
+        __middle                        = __mp;
+        __d2 -= __d1;
+      }
+    } else {
+      if (__d2 <= __bit_array<_Cp>::capacity()) {
+        __bit_array<_Cp> __b(__d2);
+        std::copy(__middle, __last, __b.begin());
+        std::copy_backward(__b.begin(), __b.end(), std::copy_backward(__first, __middle, __last));
+        break;
+      } else {
+        __bit_iterator<_Cp, false> __mp = __first + __d2;
+        std::swap_ranges(__first, __mp, __middle);
+        __first = __mp;
+        __d1 -= __d2;
+      }
+    }
+  }
+  return __r;
+}
+
+// equal
+
+template <class _Cp, bool _IC1, bool _IC2>
+_LIBCPP_CONSTEXPR_SINCE_CXX20 _LIBCPP_HIDE_FROM_ABI bool __equal_unaligned(
+    __bit_iterator<_Cp, _IC1> __first1, __bit_iterator<_Cp, _IC1> __last1, __bit_iterator<_Cp, _IC2> __first2) {
+  using _It             = __bit_iterator<_Cp, _IC1>;
+  using difference_type = typename _It::difference_type;
+  using __storage_type  = typename _It::__storage_type;
+
+  const int __bits_per_word = _It::__bits_per_word;
+  difference_type __n       = __last1 - __first1;
+  if (__n > 0) {
+    // do first word
+    if (__first1.__ctz_ != 0) {
+      unsigned __clz_f     = __bits_per_word - __first1.__ctz_;
+      difference_type __dn = std::min(static_cast<difference_type>(__clz_f), __n);
+      __n -= __dn;
+      __storage_type __m   = (~__storage_type(0) << __first1.__ctz_) & (~__storage_type(0) >> (__clz_f - __dn));
+      __storage_type __b   = *__first1.__seg_ & __m;
+      unsigned __clz_r     = __bits_per_word - __first2.__ctz_;
+      __storage_type __ddn = std::min<__storage_type>(__dn, __clz_r);
+      __m                  = (~__storage_type(0) << __first2.__ctz_) & (~__storage_type(0) >> (__clz_r - __ddn));
+      if (__first2.__ctz_ > __first1.__ctz_) {
+        if ((*__first2.__seg_ & __m) != (__b << (__first2.__ctz_ - __first1.__ctz_)))
+          return false;
+      } else {
+        if ((*__first2.__seg_ & __m) != (__b >> (__first1.__ctz_ - __first2.__ctz_)))
+          return false;
+      }
+      __first2.__seg_ += (__ddn + __first2.__ctz_) / __bits_per_word;
+      __first2.__ctz_ = static_cast<unsigned>((__ddn + __first2.__ctz_) % __bits_per_word);
+      __dn -= __ddn;
+      if (__dn > 0) {
+        __m = ~__storage_type(0) >> (__bits_per_word - __dn);
+        if ((*__first2.__seg_ & __m) != (__b >> (__first1.__ctz_ + __ddn)))
+          return false;
+        __first2.__ctz_ = static_cast<unsigned>(__dn);
+      }
+      ++__first1.__seg_;
+      // __first1.__ctz_ = 0;
+    }
+    // __first1.__ctz_ == 0;
+    // do middle words
+    unsigned __clz_r   = __bits_per_word - __first2.__ctz_;
+    __storage_type __m = ~__storage_type(0) << __first2.__ctz_;
+    for (; __n >= __bits_per_word; __n -= __bits_per_word, ++__first1.__seg_) {
+      __storage_type __b = *__first1.__seg_;
+      if ((*__first2.__seg_ & __m) != (__b << __first2.__ctz_))
+        return false;
+      ++__first2.__seg_;
+      if ((*__first2.__seg_ & ~__m) != (__b >> __clz_r))
+        return false;
+    }
+    // do last word
+    if (__n > 0) {
+      __m                 = ~__storage_type(0) >> (__bits_per_word - __n);
+      __storage_type __b  = *__first1.__seg_ & __m;
+      __storage_type __dn = std::min(__n, static_cast<difference_type>(__clz_r));
+      __m                 = (~__storage_type(0) << __first2.__ctz_) & (~__storage_type(0) >> (__clz_r - __dn));
+      if ((*__first2.__seg_ & __m) != (__b << __first2.__ctz_))
+        return false;
+      __first2.__seg_ += (__dn + __first2.__ctz_) / __bits_per_word;
+      __first2.__ctz_ = static_cast<unsigned>((__dn + __first2.__ctz_) % __bits_per_word);
+      __n -= __dn;
+      if (__n > 0) {
+        __m = ~__storage_type(0) >> (__bits_per_word - __n);
+        if ((*__first2.__seg_ & __m) != (__b >> __dn))
+          return false;
+      }
+    }
+  }
+  return true;
+}
+
+template <class _Cp, bool _IC1, bool _IC2>
+_LIBCPP_CONSTEXPR_SINCE_CXX20 _LIBCPP_HIDE_FROM_ABI bool __equal_aligned(
+    __bit_iterator<_Cp, _IC1> __first1, __bit_iterator<_Cp, _IC1> __last1, __bit_iterator<_Cp, _IC2> __first2) {
+  using _It             = __bit_iterator<_Cp, _IC1>;
+  using difference_type = typename _It::difference_type;
+  using __storage_type  = typename _It::__storage_type;
+
+  const int __bits_per_word = _It::__bits_per_word;
+  difference_type __n       = __last1 - __first1;
+  if (__n > 0) {
+    // do first word
+    if (__first1.__ctz_ != 0) {
+      unsigned __clz       = __bits_per_word - __first1.__ctz_;
+      difference_type __dn = std::min(static_cast<difference_type>(__clz), __n);
+      __n -= __dn;
+      __storage_type __m = (~__storage_type(0) << __first1.__ctz_) & (~__storage_type(0) >> (__clz - __dn));
+      if ((*__first2.__seg_ & __m) != (*__first1.__seg_ & __m))
+        return false;
+      ++__first2.__seg_;
+      ++__first1.__seg_;
+      // __first1.__ctz_ = 0;
+      // __first2.__ctz_ = 0;
+    }
+    // __first1.__ctz_ == 0;
+    // __first2.__ctz_ == 0;
+    // do middle words
+    for (; __n >= __bits_per_word; __n -= __bits_per_word, ++__first1.__seg_, ++__first2.__seg_)
+      if (*__first2.__seg_ != *__first1.__seg_)
+        return false;
+    // do last word
+    if (__n > 0) {
+      __storage_type __m = ~__storage_type(0) >> (__bits_per_word - __n);
+      if ((*__first2.__seg_ & __m) != (*__first1.__seg_ & __m))
+        return false;
+    }
+  }
+  return true;
+}
+
+template <class _Cp, bool _IC1, bool _IC2>
+inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 bool
+equal(__bit_iterator<_Cp, _IC1> __first1, __bit_iterator<_Cp, _IC1> __last1, __bit_iterator<_Cp, _IC2> __first2) {
+  if (__first1.__ctz_ == __first2.__ctz_)
+    return std::__equal_aligned(__first1, __last1, __first2);
+  return std::__equal_unaligned(__first1, __last1, __first2);
+}
+
+template <class _Cp, bool _IsConst, typename _Cp::__storage_type>
+class __bit_iterator {
+public:
+  using difference_type = typename _Cp::difference_type;
+  using value_type      = bool;
+  using pointer         = __bit_iterator;
+#ifndef _LIBCPP_ABI_BITSET_VECTOR_BOOL_CONST_SUBSCRIPT_RETURN_BOOL
+  using reference = __conditional_t<_IsConst, __bit_const_reference<_Cp>, __bit_reference<_Cp> >;
+#else
+  using reference = __conditional_t<_IsConst, bool, __bit_reference<_Cp> >;
+#endif
+  using iterator_category = random_access_iterator_tag;
+
+private:
+  using __storage_type = typename _Cp::__storage_type;
+  using __storage_pointer =
+      __conditional_t<_IsConst, typename _Cp::__const_storage_pointer, typename _Cp::__storage_pointer>;
+
+  static const unsigned __bits_per_word = _Cp::__bits_per_word;
+
+  __storage_pointer __seg_;
+  unsigned __ctz_;
+
+public:
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 __bit_iterator() _NOEXCEPT
+#if _LIBCPP_STD_VER >= 14
+      : __seg_(nullptr),
+        __ctz_(0)
+#endif
+  {
+  }
+
+  // When _IsConst=false, this is the copy constructor.
+  // It is non-trivial. Making it trivial would break ABI.
+  // When _IsConst=true, this is a converting constructor;
+  // the copy and move constructors are implicitly generated
+  // and trivial.
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 __bit_iterator(const __bit_iterator<_Cp, false>& __it) _NOEXCEPT
+      : __seg_(__it.__seg_),
+        __ctz_(__it.__ctz_) {}
+
+  // When _IsConst=false, we have a user-provided copy constructor,
+  // so we must also provide a copy assignment operator because
+  // the implicit generation of a defaulted one is deprecated.
+  // When _IsConst=true, the assignment operators are
+  // implicitly generated and trivial.
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 __bit_iterator&
+  operator=(const _If<_IsConst, struct __private_nat, __bit_iterator>& __it) {
+    __seg_ = __it.__seg_;
+    __ctz_ = __it.__ctz_;
+    return *this;
+  }
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 reference operator*() const _NOEXCEPT {
+    return __conditional_t<_IsConst, __bit_const_reference<_Cp>, __bit_reference<_Cp> >(
+        __seg_, __storage_type(1) << __ctz_);
+  }
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 __bit_iterator& operator++() {
+    if (__ctz_ != __bits_per_word - 1)
+      ++__ctz_;
+    else {
+      __ctz_ = 0;
+      ++__seg_;
+    }
+    return *this;
+  }
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 __bit_iterator operator++(int) {
+    __bit_iterator __tmp = *this;
+    ++(*this);
+    return __tmp;
+  }
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 __bit_iterator& operator--() {
+    if (__ctz_ != 0)
+      --__ctz_;
+    else {
+      __ctz_ = __bits_per_word - 1;
+      --__seg_;
+    }
+    return *this;
+  }
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 __bit_iterator operator--(int) {
+    __bit_iterator __tmp = *this;
+    --(*this);
+    return __tmp;
+  }
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 __bit_iterator& operator+=(difference_type __n) {
+    if (__n >= 0)
+      __seg_ += (__n + __ctz_) / __bits_per_word;
+    else
+      __seg_ += static_cast<difference_type>(__n - __bits_per_word + __ctz_ + 1) /
+                static_cast<difference_type>(__bits_per_word);
+    __n &= (__bits_per_word - 1);
+    __ctz_ = static_cast<unsigned>((__n + __ctz_) % __bits_per_word);
+    return *this;
+  }
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 __bit_iterator& operator-=(difference_type __n) {
+    return *this += -__n;
+  }
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 __bit_iterator operator+(difference_type __n) const {
+    __bit_iterator __t(*this);
+    __t += __n;
+    return __t;
+  }
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 __bit_iterator operator-(difference_type __n) const {
+    __bit_iterator __t(*this);
+    __t -= __n;
+    return __t;
+  }
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 friend __bit_iterator
+  operator+(difference_type __n, const __bit_iterator& __it) {
+    return __it + __n;
+  }
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 friend difference_type
+  operator-(const __bit_iterator& __x, const __bit_iterator& __y) {
+    return (__x.__seg_ - __y.__seg_) * __bits_per_word + __x.__ctz_ - __y.__ctz_;
+  }
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 reference operator[](difference_type __n) const {
+    return *(*this + __n);
+  }
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 friend bool
+  operator==(const __bit_iterator& __x, const __bit_iterator& __y) {
+    return __x.__seg_ == __y.__seg_ && __x.__ctz_ == __y.__ctz_;
+  }
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 friend bool
+  operator!=(const __bit_iterator& __x, const __bit_iterator& __y) {
+    return !(__x == __y);
+  }
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 friend bool
+  operator<(const __bit_iterator& __x, const __bit_iterator& __y) {
+    return __x.__seg_ < __y.__seg_ || (__x.__seg_ == __y.__seg_ && __x.__ctz_ < __y.__ctz_);
+  }
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 friend bool
+  operator>(const __bit_iterator& __x, const __bit_iterator& __y) {
+    return __y < __x;
+  }
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 friend bool
+  operator<=(const __bit_iterator& __x, const __bit_iterator& __y) {
+    return !(__y < __x);
+  }
+
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 friend bool
+  operator>=(const __bit_iterator& __x, const __bit_iterator& __y) {
+    return !(__x < __y);
+  }
+
+private:
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 explicit __bit_iterator(
+      __storage_pointer __s, unsigned __ctz) _NOEXCEPT
+      : __seg_(__s),
+        __ctz_(__ctz) {}
+
+  friend typename _Cp::__self;
+
+  friend class __bit_reference<_Cp>;
+  friend class __bit_const_reference<_Cp>;
+  friend class __bit_iterator<_Cp, true>;
+  template <class _Dp>
+  friend struct __bit_array;
+  template <bool _FillValue, class _Dp>
+  _LIBCPP_CONSTEXPR_SINCE_CXX20 friend void __fill_n(__bit_iterator<_Dp, false> __first, typename _Dp::size_type __n);
+
+  template <class _Dp, bool _IC>
+  _LIBCPP_CONSTEXPR_SINCE_CXX20 friend __bit_iterator<_Dp, false> __copy_aligned(
+      __bit_iterator<_Dp, _IC> __first, __bit_iterator<_Dp, _IC> __last, __bit_iterator<_Dp, false> __result);
+  template <class _Dp, bool _IC>
+  _LIBCPP_CONSTEXPR_SINCE_CXX20 friend __bit_iterator<_Dp, false> __copy_unaligned(
+      __bit_iterator<_Dp, _IC> __first, __bit_iterator<_Dp, _IC> __last, __bit_iterator<_Dp, false> __result);
+  template <class _Dp, bool _IC>
+  _LIBCPP_CONSTEXPR_SINCE_CXX20 friend __bit_iterator<_Dp, false>
+  copy(__bit_iterator<_Dp, _IC> __first, __bit_iterator<_Dp, _IC> __last, __bit_iterator<_Dp, false> __result);
+  template <class _Dp, bool _IC>
+  _LIBCPP_CONSTEXPR_SINCE_CXX20 friend __bit_iterator<_Dp, false> __copy_backward_aligned(
+      __bit_iterator<_Dp, _IC> __first, __bit_iterator<_Dp, _IC> __last, __bit_iterator<_Dp, false> __result);
+  template <class _Dp, bool _IC>
+  _LIBCPP_CONSTEXPR_SINCE_CXX20 friend __bit_iterator<_Dp, false> __copy_backward_unaligned(
+      __bit_iterator<_Dp, _IC> __first, __bit_iterator<_Dp, _IC> __last, __bit_iterator<_Dp, false> __result);
+  template <class _Dp, bool _IC>
+  _LIBCPP_CONSTEXPR_SINCE_CXX20 friend __bit_iterator<_Dp, false>
+  copy_backward(__bit_iterator<_Dp, _IC> __first, __bit_iterator<_Dp, _IC> __last, __bit_iterator<_Dp, false> __result);
+  template <class _Cl, class _Cr>
+  friend __bit_iterator<_Cr, false>
+      __swap_ranges_aligned(__bit_iterator<_Cl, false>, __bit_iterator<_Cl, false>, __bit_iterator<_Cr, false>);
+  template <class _Cl, class _Cr>
+  friend __bit_iterator<_Cr, false>
+      __swap_ranges_unaligned(__bit_iterator<_Cl, false>, __bit_iterator<_Cl, false>, __bit_iterator<_Cr, false>);
+  template <class _Cl, class _Cr>
+  friend __bit_iterator<_Cr, false>
+      swap_ranges(__bit_iterator<_Cl, false>, __bit_iterator<_Cl, false>, __bit_iterator<_Cr, false>);
+  template <class _Dp>
+  _LIBCPP_CONSTEXPR_SINCE_CXX20 friend __bit_iterator<_Dp, false>
+      rotate(__bit_iterator<_Dp, false>, __bit_iterator<_Dp, false>, __bit_iterator<_Dp, false>);
+  template <class _Dp, bool _IC1, bool _IC2>
+  _LIBCPP_CONSTEXPR_SINCE_CXX20 friend bool
+      __equal_aligned(__bit_iterator<_Dp, _IC1>, __bit_iterator<_Dp, _IC1>, __bit_iterator<_Dp, _IC2>);
+  template <class _Dp, bool _IC1, bool _IC2>
+  _LIBCPP_CONSTEXPR_SINCE_CXX20 friend bool
+      __equal_unaligned(__bit_iterator<_Dp, _IC1>, __bit_iterator<_Dp, _IC1>, __bit_iterator<_Dp, _IC2>);
+  template <class _Dp, bool _IC1, bool _IC2>
+  _LIBCPP_CONSTEXPR_SINCE_CXX20 friend bool
+      equal(__bit_iterator<_Dp, _IC1>, __bit_iterator<_Dp, _IC1>, __bit_iterator<_Dp, _IC2>);
+  template <bool _ToFind, class _Dp, bool _IC>
+  _LIBCPP_CONSTEXPR_SINCE_CXX20 friend __bit_iterator<_Dp, _IC>
+      __find_bool(__bit_iterator<_Dp, _IC>, typename _Dp::size_type);
+  template <bool _ToCount, class _Dp, bool _IC>
+  friend typename __bit_iterator<_Dp, _IC>::difference_type _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX23
+      __count_bool(__bit_iterator<_Dp, _IC>, typename _Dp::size_type);
+};
+
+_LIBCPP_END_NAMESPACE_STD
+
+_LIBCPP_POP_MACROS
+
+#endif // _LIBCPP___BIT_REFERENCE