- //===- llvm/ADT/BitVector.h - Bit vectors -----------------------*- 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 
- // 
- //===----------------------------------------------------------------------===// 
- /// 
- /// \file 
- /// This file implements the BitVector class. 
- /// 
- //===----------------------------------------------------------------------===// 
-   
- #ifndef LLVM_ADT_BITVECTOR_H 
- #define LLVM_ADT_BITVECTOR_H 
-   
- #include "llvm/ADT/ArrayRef.h" 
- #include "llvm/ADT/DenseMapInfo.h" 
- #include "llvm/ADT/iterator_range.h" 
- #include "llvm/Support/MathExtras.h" 
- #include <algorithm> 
- #include <cassert> 
- #include <climits> 
- #include <cstdint> 
- #include <cstdlib> 
- #include <cstring> 
- #include <utility> 
-   
- namespace llvm { 
-   
- /// ForwardIterator for the bits that are set. 
- /// Iterators get invalidated when resize / reserve is called. 
- template <typename BitVectorT> class const_set_bits_iterator_impl { 
-   const BitVectorT &Parent; 
-   int Current = 0; 
-   
-   void advance() { 
-     assert(Current != -1 && "Trying to advance past end."); 
-     Current = Parent.find_next(Current); 
-   } 
-   
- public: 
-   const_set_bits_iterator_impl(const BitVectorT &Parent, int Current) 
-       : Parent(Parent), Current(Current) {} 
-   explicit const_set_bits_iterator_impl(const BitVectorT &Parent) 
-       : const_set_bits_iterator_impl(Parent, Parent.find_first()) {} 
-   const_set_bits_iterator_impl(const const_set_bits_iterator_impl &) = default; 
-   
-   const_set_bits_iterator_impl operator++(int) { 
-     auto Prev = *this; 
-     advance(); 
-     return Prev; 
-   } 
-   
-   const_set_bits_iterator_impl &operator++() { 
-     advance(); 
-     return *this; 
-   } 
-   
-   unsigned operator*() const { return Current; } 
-   
-   bool operator==(const const_set_bits_iterator_impl &Other) const { 
-     assert(&Parent == &Other.Parent && 
-            "Comparing iterators from different BitVectors"); 
-     return Current == Other.Current; 
-   } 
-   
-   bool operator!=(const const_set_bits_iterator_impl &Other) const { 
-     assert(&Parent == &Other.Parent && 
-            "Comparing iterators from different BitVectors"); 
-     return Current != Other.Current; 
-   } 
- }; 
-   
- class BitVector { 
-   typedef uintptr_t BitWord; 
-   
-   enum { BITWORD_SIZE = (unsigned)sizeof(BitWord) * CHAR_BIT }; 
-   
-   static_assert(BITWORD_SIZE == 64 || BITWORD_SIZE == 32, 
-                 "Unsupported word size"); 
-   
-   using Storage = SmallVector<BitWord>; 
-   
-   Storage Bits;  // Actual bits. 
-   unsigned Size = 0; // Size of bitvector in bits. 
-   
- public: 
-   using size_type = unsigned; 
-   
-   // Encapsulation of a single bit. 
-   class reference { 
-   
-     BitWord *WordRef; 
-     unsigned BitPos; 
-   
-   public: 
-     reference(BitVector &b, unsigned Idx) { 
-       WordRef = &b.Bits[Idx / BITWORD_SIZE]; 
-       BitPos = Idx % BITWORD_SIZE; 
-     } 
-   
-     reference() = delete; 
-     reference(const reference&) = default; 
-   
-     reference &operator=(reference t) { 
-       *this = bool(t); 
-       return *this; 
-     } 
-   
-     reference& operator=(bool t) { 
-       if (t) 
-         *WordRef |= BitWord(1) << BitPos; 
-       else 
-         *WordRef &= ~(BitWord(1) << BitPos); 
-       return *this; 
-     } 
-   
-     operator bool() const { 
-       return ((*WordRef) & (BitWord(1) << BitPos)) != 0; 
-     } 
-   }; 
-   
-   typedef const_set_bits_iterator_impl<BitVector> const_set_bits_iterator; 
-   typedef const_set_bits_iterator set_iterator; 
-   
-   const_set_bits_iterator set_bits_begin() const { 
-     return const_set_bits_iterator(*this); 
-   } 
-   const_set_bits_iterator set_bits_end() const { 
-     return const_set_bits_iterator(*this, -1); 
-   } 
-   iterator_range<const_set_bits_iterator> set_bits() const { 
-     return make_range(set_bits_begin(), set_bits_end()); 
-   } 
-   
-   /// BitVector default ctor - Creates an empty bitvector. 
-   BitVector() = default; 
-   
-   /// BitVector ctor - Creates a bitvector of specified number of bits. All 
-   /// bits are initialized to the specified value. 
-   explicit BitVector(unsigned s, bool t = false) 
-       : Bits(NumBitWords(s), 0 - (BitWord)t), Size(s) { 
-     if (t) 
-       clear_unused_bits(); 
-   } 
-   
-   /// empty - Tests whether there are no bits in this bitvector. 
-   bool empty() const { return Size == 0; } 
-   
-   /// size - Returns the number of bits in this bitvector. 
-   size_type size() const { return Size; } 
-   
-   /// count - Returns the number of bits which are set. 
-   size_type count() const { 
-     unsigned NumBits = 0; 
-     for (auto Bit : Bits) 
-       NumBits += llvm::popcount(Bit); 
-     return NumBits; 
-   } 
-   
-   /// any - Returns true if any bit is set. 
-   bool any() const { 
-     return any_of(Bits, [](BitWord Bit) { return Bit != 0; }); 
-   } 
-   
-   /// all - Returns true if all bits are set. 
-   bool all() const { 
-     for (unsigned i = 0; i < Size / BITWORD_SIZE; ++i) 
-       if (Bits[i] != ~BitWord(0)) 
-         return false; 
-   
-     // If bits remain check that they are ones. The unused bits are always zero. 
-     if (unsigned Remainder = Size % BITWORD_SIZE) 
-       return Bits[Size / BITWORD_SIZE] == (BitWord(1) << Remainder) - 1; 
-   
-     return true; 
-   } 
-   
-   /// none - Returns true if none of the bits are set. 
-   bool none() const { 
-     return !any(); 
-   } 
-   
-   /// find_first_in - Returns the index of the first set / unset bit, 
-   /// depending on \p Set, in the range [Begin, End). 
-   /// Returns -1 if all bits in the range are unset / set. 
-   int find_first_in(unsigned Begin, unsigned End, bool Set = true) const { 
-     assert(Begin <= End && End <= Size); 
-     if (Begin == End) 
-       return -1; 
-   
-     unsigned FirstWord = Begin / BITWORD_SIZE; 
-     unsigned LastWord = (End - 1) / BITWORD_SIZE; 
-   
-     // Check subsequent words. 
-     // The code below is based on search for the first _set_ bit. If 
-     // we're searching for the first _unset_, we just take the 
-     // complement of each word before we use it and apply 
-     // the same method. 
-     for (unsigned i = FirstWord; i <= LastWord; ++i) { 
-       BitWord Copy = Bits[i]; 
-       if (!Set) 
-         Copy = ~Copy; 
-   
-       if (i == FirstWord) { 
-         unsigned FirstBit = Begin % BITWORD_SIZE; 
-         Copy &= maskTrailingZeros<BitWord>(FirstBit); 
-       } 
-   
-       if (i == LastWord) { 
-         unsigned LastBit = (End - 1) % BITWORD_SIZE; 
-         Copy &= maskTrailingOnes<BitWord>(LastBit + 1); 
-       } 
-       if (Copy != 0) 
-         return i * BITWORD_SIZE + countTrailingZeros(Copy); 
-     } 
-     return -1; 
-   } 
-   
-   /// find_last_in - Returns the index of the last set bit in the range 
-   /// [Begin, End).  Returns -1 if all bits in the range are unset. 
-   int find_last_in(unsigned Begin, unsigned End) const { 
-     assert(Begin <= End && End <= Size); 
-     if (Begin == End) 
-       return -1; 
-   
-     unsigned LastWord = (End - 1) / BITWORD_SIZE; 
-     unsigned FirstWord = Begin / BITWORD_SIZE; 
-   
-     for (unsigned i = LastWord + 1; i >= FirstWord + 1; --i) { 
-       unsigned CurrentWord = i - 1; 
-   
-       BitWord Copy = Bits[CurrentWord]; 
-       if (CurrentWord == LastWord) { 
-         unsigned LastBit = (End - 1) % BITWORD_SIZE; 
-         Copy &= maskTrailingOnes<BitWord>(LastBit + 1); 
-       } 
-   
-       if (CurrentWord == FirstWord) { 
-         unsigned FirstBit = Begin % BITWORD_SIZE; 
-         Copy &= maskTrailingZeros<BitWord>(FirstBit); 
-       } 
-   
-       if (Copy != 0) 
-         return (CurrentWord + 1) * BITWORD_SIZE - countLeadingZeros(Copy) - 1; 
-     } 
-   
-     return -1; 
-   } 
-   
-   /// find_first_unset_in - Returns the index of the first unset bit in the 
-   /// range [Begin, End).  Returns -1 if all bits in the range are set. 
-   int find_first_unset_in(unsigned Begin, unsigned End) const { 
-     return find_first_in(Begin, End, /* Set = */ false); 
-   } 
-   
-   /// find_last_unset_in - Returns the index of the last unset bit in the 
-   /// range [Begin, End).  Returns -1 if all bits in the range are set. 
-   int find_last_unset_in(unsigned Begin, unsigned End) const { 
-     assert(Begin <= End && End <= Size); 
-     if (Begin == End) 
-       return -1; 
-   
-     unsigned LastWord = (End - 1) / BITWORD_SIZE; 
-     unsigned FirstWord = Begin / BITWORD_SIZE; 
-   
-     for (unsigned i = LastWord + 1; i >= FirstWord + 1; --i) { 
-       unsigned CurrentWord = i - 1; 
-   
-       BitWord Copy = Bits[CurrentWord]; 
-       if (CurrentWord == LastWord) { 
-         unsigned LastBit = (End - 1) % BITWORD_SIZE; 
-         Copy |= maskTrailingZeros<BitWord>(LastBit + 1); 
-       } 
-   
-       if (CurrentWord == FirstWord) { 
-         unsigned FirstBit = Begin % BITWORD_SIZE; 
-         Copy |= maskTrailingOnes<BitWord>(FirstBit); 
-       } 
-   
-       if (Copy != ~BitWord(0)) { 
-         unsigned Result = 
-             (CurrentWord + 1) * BITWORD_SIZE - countLeadingOnes(Copy) - 1; 
-         return Result < Size ? Result : -1; 
-       } 
-     } 
-     return -1; 
-   } 
-   
-   /// find_first - Returns the index of the first set bit, -1 if none 
-   /// of the bits are set. 
-   int find_first() const { return find_first_in(0, Size); } 
-   
-   /// find_last - Returns the index of the last set bit, -1 if none of the bits 
-   /// are set. 
-   int find_last() const { return find_last_in(0, Size); } 
-   
-   /// find_next - Returns the index of the next set bit following the 
-   /// "Prev" bit. Returns -1 if the next set bit is not found. 
-   int find_next(unsigned Prev) const { return find_first_in(Prev + 1, Size); } 
-   
-   /// find_prev - Returns the index of the first set bit that precedes the 
-   /// the bit at \p PriorTo.  Returns -1 if all previous bits are unset. 
-   int find_prev(unsigned PriorTo) const { return find_last_in(0, PriorTo); } 
-   
-   /// find_first_unset - Returns the index of the first unset bit, -1 if all 
-   /// of the bits are set. 
-   int find_first_unset() const { return find_first_unset_in(0, Size); } 
-   
-   /// find_next_unset - Returns the index of the next unset bit following the 
-   /// "Prev" bit.  Returns -1 if all remaining bits are set. 
-   int find_next_unset(unsigned Prev) const { 
-     return find_first_unset_in(Prev + 1, Size); 
-   } 
-   
-   /// find_last_unset - Returns the index of the last unset bit, -1 if all of 
-   /// the bits are set. 
-   int find_last_unset() const { return find_last_unset_in(0, Size); } 
-   
-   /// find_prev_unset - Returns the index of the first unset bit that precedes 
-   /// the bit at \p PriorTo.  Returns -1 if all previous bits are set. 
-   int find_prev_unset(unsigned PriorTo) { 
-     return find_last_unset_in(0, PriorTo); 
-   } 
-   
-   /// clear - Removes all bits from the bitvector. 
-   void clear() { 
-     Size = 0; 
-     Bits.clear(); 
-   } 
-   
-   /// resize - Grow or shrink the bitvector. 
-   void resize(unsigned N, bool t = false) { 
-     set_unused_bits(t); 
-     Size = N; 
-     Bits.resize(NumBitWords(N), 0 - BitWord(t)); 
-     clear_unused_bits(); 
-   } 
-   
-   void reserve(unsigned N) { Bits.reserve(NumBitWords(N)); } 
-   
-   // Set, reset, flip 
-   BitVector &set() { 
-     init_words(true); 
-     clear_unused_bits(); 
-     return *this; 
-   } 
-   
-   BitVector &set(unsigned Idx) { 
-     assert(Idx < Size && "access in bound"); 
-     Bits[Idx / BITWORD_SIZE] |= BitWord(1) << (Idx % BITWORD_SIZE); 
-     return *this; 
-   } 
-   
-   /// set - Efficiently set a range of bits in [I, E) 
-   BitVector &set(unsigned I, unsigned E) { 
-     assert(I <= E && "Attempted to set backwards range!"); 
-     assert(E <= size() && "Attempted to set out-of-bounds range!"); 
-   
-     if (I == E) return *this; 
-   
-     if (I / BITWORD_SIZE == E / BITWORD_SIZE) { 
-       BitWord EMask = BitWord(1) << (E % BITWORD_SIZE); 
-       BitWord IMask = BitWord(1) << (I % BITWORD_SIZE); 
-       BitWord Mask = EMask - IMask; 
-       Bits[I / BITWORD_SIZE] |= Mask; 
-       return *this; 
-     } 
-   
-     BitWord PrefixMask = ~BitWord(0) << (I % BITWORD_SIZE); 
-     Bits[I / BITWORD_SIZE] |= PrefixMask; 
-     I = alignTo(I, BITWORD_SIZE); 
-   
-     for (; I + BITWORD_SIZE <= E; I += BITWORD_SIZE) 
-       Bits[I / BITWORD_SIZE] = ~BitWord(0); 
-   
-     BitWord PostfixMask = (BitWord(1) << (E % BITWORD_SIZE)) - 1; 
-     if (I < E) 
-       Bits[I / BITWORD_SIZE] |= PostfixMask; 
-   
-     return *this; 
-   } 
-   
-   BitVector &reset() { 
-     init_words(false); 
-     return *this; 
-   } 
-   
-   BitVector &reset(unsigned Idx) { 
-     Bits[Idx / BITWORD_SIZE] &= ~(BitWord(1) << (Idx % BITWORD_SIZE)); 
-     return *this; 
-   } 
-   
-   /// reset - Efficiently reset a range of bits in [I, E) 
-   BitVector &reset(unsigned I, unsigned E) { 
-     assert(I <= E && "Attempted to reset backwards range!"); 
-     assert(E <= size() && "Attempted to reset out-of-bounds range!"); 
-   
-     if (I == E) return *this; 
-   
-     if (I / BITWORD_SIZE == E / BITWORD_SIZE) { 
-       BitWord EMask = BitWord(1) << (E % BITWORD_SIZE); 
-       BitWord IMask = BitWord(1) << (I % BITWORD_SIZE); 
-       BitWord Mask = EMask - IMask; 
-       Bits[I / BITWORD_SIZE] &= ~Mask; 
-       return *this; 
-     } 
-   
-     BitWord PrefixMask = ~BitWord(0) << (I % BITWORD_SIZE); 
-     Bits[I / BITWORD_SIZE] &= ~PrefixMask; 
-     I = alignTo(I, BITWORD_SIZE); 
-   
-     for (; I + BITWORD_SIZE <= E; I += BITWORD_SIZE) 
-       Bits[I / BITWORD_SIZE] = BitWord(0); 
-   
-     BitWord PostfixMask = (BitWord(1) << (E % BITWORD_SIZE)) - 1; 
-     if (I < E) 
-       Bits[I / BITWORD_SIZE] &= ~PostfixMask; 
-   
-     return *this; 
-   } 
-   
-   BitVector &flip() { 
-     for (auto &Bit : Bits) 
-       Bit = ~Bit; 
-     clear_unused_bits(); 
-     return *this; 
-   } 
-   
-   BitVector &flip(unsigned Idx) { 
-     Bits[Idx / BITWORD_SIZE] ^= BitWord(1) << (Idx % BITWORD_SIZE); 
-     return *this; 
-   } 
-   
-   // Indexing. 
-   reference operator[](unsigned Idx) { 
-     assert (Idx < Size && "Out-of-bounds Bit access."); 
-     return reference(*this, Idx); 
-   } 
-   
-   bool operator[](unsigned Idx) const { 
-     assert (Idx < Size && "Out-of-bounds Bit access."); 
-     BitWord Mask = BitWord(1) << (Idx % BITWORD_SIZE); 
-     return (Bits[Idx / BITWORD_SIZE] & Mask) != 0; 
-   } 
-   
-   /// Return the last element in the vector. 
-   bool back() const { 
-     assert(!empty() && "Getting last element of empty vector."); 
-     return (*this)[size() - 1]; 
-   } 
-   
-   bool test(unsigned Idx) const { 
-     return (*this)[Idx]; 
-   } 
-   
-   // Push single bit to end of vector. 
-   void push_back(bool Val) { 
-     unsigned OldSize = Size; 
-     unsigned NewSize = Size + 1; 
-   
-     // Resize, which will insert zeros. 
-     // If we already fit then the unused bits will be already zero. 
-     if (NewSize > getBitCapacity()) 
-       resize(NewSize, false); 
-     else 
-       Size = NewSize; 
-   
-     // If true, set single bit. 
-     if (Val) 
-       set(OldSize); 
-   } 
-   
-   /// Pop one bit from the end of the vector. 
-   void pop_back() { 
-     assert(!empty() && "Empty vector has no element to pop."); 
-     resize(size() - 1); 
-   } 
-   
-   /// Test if any common bits are set. 
-   bool anyCommon(const BitVector &RHS) const { 
-     unsigned ThisWords = Bits.size(); 
-     unsigned RHSWords = RHS.Bits.size(); 
-     for (unsigned i = 0, e = std::min(ThisWords, RHSWords); i != e; ++i) 
-       if (Bits[i] & RHS.Bits[i]) 
-         return true; 
-     return false; 
-   } 
-   
-   // Comparison operators. 
-   bool operator==(const BitVector &RHS) const { 
-     if (size() != RHS.size()) 
-       return false; 
-     unsigned NumWords = Bits.size(); 
-     return std::equal(Bits.begin(), Bits.begin() + NumWords, RHS.Bits.begin()); 
-   } 
-   
-   bool operator!=(const BitVector &RHS) const { return !(*this == RHS); } 
-   
-   /// Intersection, union, disjoint union. 
-   BitVector &operator&=(const BitVector &RHS) { 
-     unsigned ThisWords = Bits.size(); 
-     unsigned RHSWords = RHS.Bits.size(); 
-     unsigned i; 
-     for (i = 0; i != std::min(ThisWords, RHSWords); ++i) 
-       Bits[i] &= RHS.Bits[i]; 
-   
-     // Any bits that are just in this bitvector become zero, because they aren't 
-     // in the RHS bit vector.  Any words only in RHS are ignored because they 
-     // are already zero in the LHS. 
-     for (; i != ThisWords; ++i) 
-       Bits[i] = 0; 
-   
-     return *this; 
-   } 
-   
-   /// reset - Reset bits that are set in RHS. Same as *this &= ~RHS. 
-   BitVector &reset(const BitVector &RHS) { 
-     unsigned ThisWords = Bits.size(); 
-     unsigned RHSWords = RHS.Bits.size(); 
-     for (unsigned i = 0; i != std::min(ThisWords, RHSWords); ++i) 
-       Bits[i] &= ~RHS.Bits[i]; 
-     return *this; 
-   } 
-   
-   /// test - Check if (This - RHS) is zero. 
-   /// This is the same as reset(RHS) and any(). 
-   bool test(const BitVector &RHS) const { 
-     unsigned ThisWords = Bits.size(); 
-     unsigned RHSWords = RHS.Bits.size(); 
-     unsigned i; 
-     for (i = 0; i != std::min(ThisWords, RHSWords); ++i) 
-       if ((Bits[i] & ~RHS.Bits[i]) != 0) 
-         return true; 
-   
-     for (; i != ThisWords ; ++i) 
-       if (Bits[i] != 0) 
-         return true; 
-   
-     return false; 
-   } 
-   
-   template <class F, class... ArgTys> 
-   static BitVector &apply(F &&f, BitVector &Out, BitVector const &Arg, 
-                           ArgTys const &...Args) { 
-     assert(llvm::all_of( 
-                std::initializer_list<unsigned>{Args.size()...}, 
-                [&Arg](auto const &BV) { return Arg.size() == BV; }) && 
-            "consistent sizes"); 
-     Out.resize(Arg.size()); 
-     for (size_type I = 0, E = Arg.Bits.size(); I != E; ++I) 
-       Out.Bits[I] = f(Arg.Bits[I], Args.Bits[I]...); 
-     Out.clear_unused_bits(); 
-     return Out; 
-   } 
-   
-   BitVector &operator|=(const BitVector &RHS) { 
-     if (size() < RHS.size()) 
-       resize(RHS.size()); 
-     for (size_type I = 0, E = RHS.Bits.size(); I != E; ++I) 
-       Bits[I] |= RHS.Bits[I]; 
-     return *this; 
-   } 
-   
-   BitVector &operator^=(const BitVector &RHS) { 
-     if (size() < RHS.size()) 
-       resize(RHS.size()); 
-     for (size_type I = 0, E = RHS.Bits.size(); I != E; ++I) 
-       Bits[I] ^= RHS.Bits[I]; 
-     return *this; 
-   } 
-   
-   BitVector &operator>>=(unsigned N) { 
-     assert(N <= Size); 
-     if (LLVM_UNLIKELY(empty() || N == 0)) 
-       return *this; 
-   
-     unsigned NumWords = Bits.size(); 
-     assert(NumWords >= 1); 
-   
-     wordShr(N / BITWORD_SIZE); 
-   
-     unsigned BitDistance = N % BITWORD_SIZE; 
-     if (BitDistance == 0) 
-       return *this; 
-   
-     // When the shift size is not a multiple of the word size, then we have 
-     // a tricky situation where each word in succession needs to extract some 
-     // of the bits from the next word and or them into this word while 
-     // shifting this word to make room for the new bits.  This has to be done 
-     // for every word in the array. 
-   
-     // Since we're shifting each word right, some bits will fall off the end 
-     // of each word to the right, and empty space will be created on the left. 
-     // The final word in the array will lose bits permanently, so starting at 
-     // the beginning, work forwards shifting each word to the right, and 
-     // OR'ing in the bits from the end of the next word to the beginning of 
-     // the current word. 
-   
-     // Example: 
-     //   Starting with {0xAABBCCDD, 0xEEFF0011, 0x22334455} and shifting right 
-     //   by 4 bits. 
-     // Step 1: Word[0] >>= 4           ; 0x0ABBCCDD 
-     // Step 2: Word[0] |= 0x10000000   ; 0x1ABBCCDD 
-     // Step 3: Word[1] >>= 4           ; 0x0EEFF001 
-     // Step 4: Word[1] |= 0x50000000   ; 0x5EEFF001 
-     // Step 5: Word[2] >>= 4           ; 0x02334455 
-     // Result: { 0x1ABBCCDD, 0x5EEFF001, 0x02334455 } 
-     const BitWord Mask = maskTrailingOnes<BitWord>(BitDistance); 
-     const unsigned LSH = BITWORD_SIZE - BitDistance; 
-   
-     for (unsigned I = 0; I < NumWords - 1; ++I) { 
-       Bits[I] >>= BitDistance; 
-       Bits[I] |= (Bits[I + 1] & Mask) << LSH; 
-     } 
-   
-     Bits[NumWords - 1] >>= BitDistance; 
-   
-     return *this; 
-   } 
-   
-   BitVector &operator<<=(unsigned N) { 
-     assert(N <= Size); 
-     if (LLVM_UNLIKELY(empty() || N == 0)) 
-       return *this; 
-   
-     unsigned NumWords = Bits.size(); 
-     assert(NumWords >= 1); 
-   
-     wordShl(N / BITWORD_SIZE); 
-   
-     unsigned BitDistance = N % BITWORD_SIZE; 
-     if (BitDistance == 0) 
-       return *this; 
-   
-     // When the shift size is not a multiple of the word size, then we have 
-     // a tricky situation where each word in succession needs to extract some 
-     // of the bits from the previous word and or them into this word while 
-     // shifting this word to make room for the new bits.  This has to be done 
-     // for every word in the array.  This is similar to the algorithm outlined 
-     // in operator>>=, but backwards. 
-   
-     // Since we're shifting each word left, some bits will fall off the end 
-     // of each word to the left, and empty space will be created on the right. 
-     // The first word in the array will lose bits permanently, so starting at 
-     // the end, work backwards shifting each word to the left, and OR'ing 
-     // in the bits from the end of the next word to the beginning of the 
-     // current word. 
-   
-     // Example: 
-     //   Starting with {0xAABBCCDD, 0xEEFF0011, 0x22334455} and shifting left 
-     //   by 4 bits. 
-     // Step 1: Word[2] <<= 4           ; 0x23344550 
-     // Step 2: Word[2] |= 0x0000000E   ; 0x2334455E 
-     // Step 3: Word[1] <<= 4           ; 0xEFF00110 
-     // Step 4: Word[1] |= 0x0000000A   ; 0xEFF0011A 
-     // Step 5: Word[0] <<= 4           ; 0xABBCCDD0 
-     // Result: { 0xABBCCDD0, 0xEFF0011A, 0x2334455E } 
-     const BitWord Mask = maskLeadingOnes<BitWord>(BitDistance); 
-     const unsigned RSH = BITWORD_SIZE - BitDistance; 
-   
-     for (int I = NumWords - 1; I > 0; --I) { 
-       Bits[I] <<= BitDistance; 
-       Bits[I] |= (Bits[I - 1] & Mask) >> RSH; 
-     } 
-     Bits[0] <<= BitDistance; 
-     clear_unused_bits(); 
-   
-     return *this; 
-   } 
-   
-   void swap(BitVector &RHS) { 
-     std::swap(Bits, RHS.Bits); 
-     std::swap(Size, RHS.Size); 
-   } 
-   
-   void invalid() { 
-     assert(!Size && Bits.empty()); 
-     Size = (unsigned)-1; 
-   } 
-   bool isInvalid() const { return Size == (unsigned)-1; } 
-   
-   ArrayRef<BitWord> getData() const { return {&Bits[0], Bits.size()}; } 
-   
-   //===--------------------------------------------------------------------===// 
-   // Portable bit mask operations. 
-   //===--------------------------------------------------------------------===// 
-   // 
-   // These methods all operate on arrays of uint32_t, each holding 32 bits. The 
-   // fixed word size makes it easier to work with literal bit vector constants 
-   // in portable code. 
-   // 
-   // The LSB in each word is the lowest numbered bit.  The size of a portable 
-   // bit mask is always a whole multiple of 32 bits.  If no bit mask size is 
-   // given, the bit mask is assumed to cover the entire BitVector. 
-   
-   /// setBitsInMask - Add '1' bits from Mask to this vector. Don't resize. 
-   /// This computes "*this |= Mask". 
-   void setBitsInMask(const uint32_t *Mask, unsigned MaskWords = ~0u) { 
-     applyMask<true, false>(Mask, MaskWords); 
-   } 
-   
-   /// clearBitsInMask - Clear any bits in this vector that are set in Mask. 
-   /// Don't resize. This computes "*this &= ~Mask". 
-   void clearBitsInMask(const uint32_t *Mask, unsigned MaskWords = ~0u) { 
-     applyMask<false, false>(Mask, MaskWords); 
-   } 
-   
-   /// setBitsNotInMask - Add a bit to this vector for every '0' bit in Mask. 
-   /// Don't resize.  This computes "*this |= ~Mask". 
-   void setBitsNotInMask(const uint32_t *Mask, unsigned MaskWords = ~0u) { 
-     applyMask<true, true>(Mask, MaskWords); 
-   } 
-   
-   /// clearBitsNotInMask - Clear a bit in this vector for every '0' bit in Mask. 
-   /// Don't resize.  This computes "*this &= Mask". 
-   void clearBitsNotInMask(const uint32_t *Mask, unsigned MaskWords = ~0u) { 
-     applyMask<false, true>(Mask, MaskWords); 
-   } 
-   
- private: 
-   /// Perform a logical left shift of \p Count words by moving everything 
-   /// \p Count words to the right in memory. 
-   /// 
-   /// While confusing, words are stored from least significant at Bits[0] to 
-   /// most significant at Bits[NumWords-1].  A logical shift left, however, 
-   /// moves the current least significant bit to a higher logical index, and 
-   /// fills the previous least significant bits with 0.  Thus, we actually 
-   /// need to move the bytes of the memory to the right, not to the left. 
-   /// Example: 
-   ///   Words = [0xBBBBAAAA, 0xDDDDFFFF, 0x00000000, 0xDDDD0000] 
-   /// represents a BitVector where 0xBBBBAAAA contain the least significant 
-   /// bits.  So if we want to shift the BitVector left by 2 words, we need 
-   /// to turn this into 0x00000000 0x00000000 0xBBBBAAAA 0xDDDDFFFF by using a 
-   /// memmove which moves right, not left. 
-   void wordShl(uint32_t Count) { 
-     if (Count == 0) 
-       return; 
-   
-     uint32_t NumWords = Bits.size(); 
-   
-     // Since we always move Word-sized chunks of data with src and dest both 
-     // aligned to a word-boundary, we don't need to worry about endianness 
-     // here. 
-     std::copy(Bits.begin(), Bits.begin() + NumWords - Count, 
-               Bits.begin() + Count); 
-     std::fill(Bits.begin(), Bits.begin() + Count, 0); 
-     clear_unused_bits(); 
-   } 
-   
-   /// Perform a logical right shift of \p Count words by moving those 
-   /// words to the left in memory.  See wordShl for more information. 
-   /// 
-   void wordShr(uint32_t Count) { 
-     if (Count == 0) 
-       return; 
-   
-     uint32_t NumWords = Bits.size(); 
-   
-     std::copy(Bits.begin() + Count, Bits.begin() + NumWords, Bits.begin()); 
-     std::fill(Bits.begin() + NumWords - Count, Bits.begin() + NumWords, 0); 
-   } 
-   
-   int next_unset_in_word(int WordIndex, BitWord Word) const { 
-     unsigned Result = WordIndex * BITWORD_SIZE + countTrailingOnes(Word); 
-     return Result < size() ? Result : -1; 
-   } 
-   
-   unsigned NumBitWords(unsigned S) const { 
-     return (S + BITWORD_SIZE-1) / BITWORD_SIZE; 
-   } 
-   
-   // Set the unused bits in the high words. 
-   void set_unused_bits(bool t = true) { 
-     //  Then set any stray high bits of the last used word. 
-     if (unsigned ExtraBits = Size % BITWORD_SIZE) { 
-       BitWord ExtraBitMask = ~BitWord(0) << ExtraBits; 
-       if (t) 
-         Bits.back() |= ExtraBitMask; 
-       else 
-         Bits.back() &= ~ExtraBitMask; 
-     } 
-   } 
-   
-   // Clear the unused bits in the high words. 
-   void clear_unused_bits() { 
-     set_unused_bits(false); 
-   } 
-   
-   void init_words(bool t) { 
-     std::fill(Bits.begin(), Bits.end(), 0 - (BitWord)t); 
-   } 
-   
-   template<bool AddBits, bool InvertMask> 
-   void applyMask(const uint32_t *Mask, unsigned MaskWords) { 
-     static_assert(BITWORD_SIZE % 32 == 0, "Unsupported BitWord size."); 
-     MaskWords = std::min(MaskWords, (size() + 31) / 32); 
-     const unsigned Scale = BITWORD_SIZE / 32; 
-     unsigned i; 
-     for (i = 0; MaskWords >= Scale; ++i, MaskWords -= Scale) { 
-       BitWord BW = Bits[i]; 
-       // This inner loop should unroll completely when BITWORD_SIZE > 32. 
-       for (unsigned b = 0; b != BITWORD_SIZE; b += 32) { 
-         uint32_t M = *Mask++; 
-         if (InvertMask) M = ~M; 
-         if (AddBits) BW |=   BitWord(M) << b; 
-         else         BW &= ~(BitWord(M) << b); 
-       } 
-       Bits[i] = BW; 
-     } 
-     for (unsigned b = 0; MaskWords; b += 32, --MaskWords) { 
-       uint32_t M = *Mask++; 
-       if (InvertMask) M = ~M; 
-       if (AddBits) Bits[i] |=   BitWord(M) << b; 
-       else         Bits[i] &= ~(BitWord(M) << b); 
-     } 
-     if (AddBits) 
-       clear_unused_bits(); 
-   } 
-   
- public: 
-   /// Return the size (in bytes) of the bit vector. 
-   size_type getMemorySize() const { return Bits.size() * sizeof(BitWord); } 
-   size_type getBitCapacity() const { return Bits.size() * BITWORD_SIZE; } 
- }; 
-   
- inline BitVector::size_type capacity_in_bytes(const BitVector &X) { 
-   return X.getMemorySize(); 
- } 
-   
- template <> struct DenseMapInfo<BitVector> { 
-   static inline BitVector getEmptyKey() { return {}; } 
-   static inline BitVector getTombstoneKey() { 
-     BitVector V; 
-     V.invalid(); 
-     return V; 
-   } 
-   static unsigned getHashValue(const BitVector &V) { 
-     return DenseMapInfo<std::pair<BitVector::size_type, ArrayRef<uintptr_t>>>:: 
-         getHashValue(std::make_pair(V.size(), V.getData())); 
-   } 
-   static bool isEqual(const BitVector &LHS, const BitVector &RHS) { 
-     if (LHS.isInvalid() || RHS.isInvalid()) 
-       return LHS.isInvalid() == RHS.isInvalid(); 
-     return LHS == RHS; 
-   } 
- }; 
- } // end namespace llvm 
-   
- namespace std { 
-   /// Implement std::swap in terms of BitVector swap. 
- inline void swap(llvm::BitVector &LHS, llvm::BitVector &RHS) { LHS.swap(RHS); } 
- } // end namespace std 
-   
- #endif // LLVM_ADT_BITVECTOR_H 
-