blob: dd38bd888242171571966043a2a45d608ea602b8 [file] [log] [blame]
/*
* Copyright (C) 2011-2019 Apple Inc. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY APPLE INC. ``AS IS'' AND ANY
* EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL APPLE INC. OR
* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
* PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
* OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include <stdio.h>
#include <wtf/Assertions.h>
#include <wtf/DataLog.h>
#include <wtf/HashFunctions.h>
#include <wtf/HashTraits.h>
#include <wtf/PrintStream.h>
#include <wtf/StdLibExtras.h>
namespace JSC {
class CachedBitVector;
}
namespace WTF {
// This is a space-efficient, resizeable bitvector class. In the common case it
// occupies one word, but if necessary, it will inflate this one word to point
// to a single chunk of out-of-line allocated storage to store an arbitrary number
// of bits.
//
// - The bitvector remembers the bound of how many bits can be stored, but this
// may be slightly greater (by as much as some platform-specific constant)
// than the last argument passed to ensureSize().
//
// - The bitvector can resize itself automatically (set, clear, get) or can be used
// in a manual mode, which is faster (quickSet, quickClear, quickGet, ensureSize).
//
// - Accesses ASSERT that you are within bounds.
//
// - Bits are automatically initialized to zero.
//
// On the other hand, this BitVector class may not be the fastest around, since
// it does conditionals on every get/set/clear. But it is great if you need to
// juggle a lot of variable-length BitVectors and you're worried about wasting
// space.
class BitVector final {
WTF_MAKE_FAST_ALLOCATED;
public:
BitVector()
: m_bitsOrPointer(makeInlineBits(0))
{
}
explicit BitVector(size_t numBits)
: m_bitsOrPointer(makeInlineBits(0))
{
ensureSize(numBits);
}
BitVector(const BitVector& other)
: m_bitsOrPointer(makeInlineBits(0))
{
(*this) = other;
}
~BitVector()
{
if (isInline())
return;
OutOfLineBits::destroy(outOfLineBits());
}
BitVector& operator=(const BitVector& other)
{
if (isInline() && other.isInline())
m_bitsOrPointer = other.m_bitsOrPointer;
else
setSlow(other);
return *this;
}
size_t size() const
{
if (isInline())
return maxInlineBits();
return outOfLineBits()->numBits();
}
void ensureSize(size_t numBits)
{
if (numBits <= size())
return;
resizeOutOfLine(numBits);
}
// Like ensureSize(), but supports reducing the size of the bitvector.
WTF_EXPORT_PRIVATE void resize(size_t numBits);
WTF_EXPORT_PRIVATE void clearAll();
bool quickGet(size_t bit) const
{
ASSERT_WITH_SECURITY_IMPLICATION(bit < size());
return !!(bits()[bit / bitsInPointer()] & (static_cast<uintptr_t>(1) << (bit & (bitsInPointer() - 1))));
}
bool quickSet(size_t bit)
{
ASSERT_WITH_SECURITY_IMPLICATION(bit < size());
uintptr_t& word = bits()[bit / bitsInPointer()];
uintptr_t mask = static_cast<uintptr_t>(1) << (bit & (bitsInPointer() - 1));
bool result = !!(word & mask);
word |= mask;
return result;
}
bool quickClear(size_t bit)
{
ASSERT_WITH_SECURITY_IMPLICATION(bit < size());
uintptr_t& word = bits()[bit / bitsInPointer()];
uintptr_t mask = static_cast<uintptr_t>(1) << (bit & (bitsInPointer() - 1));
bool result = !!(word & mask);
word &= ~mask;
return result;
}
bool quickSet(size_t bit, bool value)
{
if (value)
return quickSet(bit);
return quickClear(bit);
}
bool get(size_t bit) const
{
if (bit >= size())
return false;
return quickGet(bit);
}
bool contains(size_t bit) const
{
return get(bit);
}
bool set(size_t bit)
{
ensureSize(bit + 1);
return quickSet(bit);
}
// This works like the add methods of sets. Instead of returning the previous value, like set(),
// it returns whether the bit transitioned from false to true.
bool add(size_t bit)
{
return !set(bit);
}
bool ensureSizeAndSet(size_t bit, size_t size)
{
ensureSize(size);
return quickSet(bit);
}
bool clear(size_t bit)
{
if (bit >= size())
return false;
return quickClear(bit);
}
bool remove(size_t bit)
{
return clear(bit);
}
bool set(size_t bit, bool value)
{
if (value)
return set(bit);
return clear(bit);
}
void merge(const BitVector& other)
{
if (!isInline() || !other.isInline()) {
mergeSlow(other);
return;
}
m_bitsOrPointer |= other.m_bitsOrPointer;
ASSERT(isInline());
}
void filter(const BitVector& other)
{
if (!isInline() || !other.isInline()) {
filterSlow(other);
return;
}
m_bitsOrPointer &= other.m_bitsOrPointer;
ASSERT(isInline());
}
void exclude(const BitVector& other)
{
if (!isInline() || !other.isInline()) {
excludeSlow(other);
return;
}
m_bitsOrPointer &= ~other.m_bitsOrPointer;
m_bitsOrPointer |= (static_cast<uintptr_t>(1) << maxInlineBits());
ASSERT(isInline());
}
size_t bitCount() const
{
if (isInline())
return bitCount(cleanseInlineBits(m_bitsOrPointer));
return bitCountSlow();
}
bool isEmpty() const
{
if (isInline())
return !cleanseInlineBits(m_bitsOrPointer);
return isEmptySlow();
}
size_t findBit(size_t index, bool value) const
{
size_t result = findBitFast(index, value);
if (ASSERT_ENABLED) {
size_t expectedResult = findBitSimple(index, value);
if (result != expectedResult) {
dataLog("findBit(", index, ", ", value, ") on ", *this, " should have gotten ", expectedResult, " but got ", result, "\n");
ASSERT_NOT_REACHED();
}
}
return result;
}
WTF_EXPORT_PRIVATE void dump(PrintStream& out) const;
enum EmptyValueTag { EmptyValue };
enum DeletedValueTag { DeletedValue };
BitVector(EmptyValueTag)
: m_bitsOrPointer(0)
{
}
BitVector(DeletedValueTag)
: m_bitsOrPointer(1)
{
}
bool isEmptyValue() const { return !m_bitsOrPointer; }
bool isDeletedValue() const { return m_bitsOrPointer == 1; }
bool isEmptyOrDeletedValue() const { return m_bitsOrPointer <= 1; }
bool operator==(const BitVector& other) const
{
if (isInline() && other.isInline())
return m_bitsOrPointer == other.m_bitsOrPointer;
return equalsSlowCase(other);
}
unsigned hash() const
{
// This is a very simple hash. Just xor together the words that hold the various
// bits and then compute the hash. This makes it very easy to deal with bitvectors
// that have a lot of trailing zero's.
uintptr_t value;
if (isInline())
value = cleanseInlineBits(m_bitsOrPointer);
else
value = hashSlowCase();
return IntHash<uintptr_t>::hash(value);
}
class iterator {
WTF_MAKE_FAST_ALLOCATED;
public:
iterator()
: m_bitVector(nullptr)
, m_index(0)
{
}
iterator(const BitVector& bitVector, size_t index)
: m_bitVector(&bitVector)
, m_index(index)
{
}
size_t operator*() const { return m_index; }
iterator& operator++()
{
m_index = m_bitVector->findBit(m_index + 1, true);
return *this;
}
iterator operator++(int)
{
iterator result = *this;
++(*this);
return result;
}
bool isAtEnd() const
{
return m_index >= m_bitVector->size();
}
bool operator==(const iterator& other) const
{
return m_index == other.m_index;
}
bool operator!=(const iterator& other) const
{
return !(*this == other);
}
private:
const BitVector* m_bitVector;
size_t m_index;
};
// Use this to iterate over set bits.
iterator begin() const { return iterator(*this, findBit(0, true)); }
iterator end() const { return iterator(*this, size()); }
private:
friend class JSC::CachedBitVector;
static unsigned bitsInPointer()
{
return sizeof(void*) << 3;
}
static unsigned maxInlineBits()
{
return bitsInPointer() - 1;
}
static size_t byteCount(size_t bitCount)
{
return (bitCount + 7) >> 3;
}
static uintptr_t makeInlineBits(uintptr_t bits)
{
ASSERT(!(bits & (static_cast<uintptr_t>(1) << maxInlineBits())));
return bits | (static_cast<uintptr_t>(1) << maxInlineBits());
}
static uintptr_t cleanseInlineBits(uintptr_t bits)
{
return bits & ~(static_cast<uintptr_t>(1) << maxInlineBits());
}
static size_t bitCount(uintptr_t bits)
{
if (sizeof(uintptr_t) == 4)
return WTF::bitCount(static_cast<unsigned>(bits));
return WTF::bitCount(static_cast<uint64_t>(bits));
}
size_t findBitFast(size_t startIndex, bool value) const
{
if (isInline()) {
size_t index = startIndex;
findBitInWord(m_bitsOrPointer, index, maxInlineBits(), value);
return index;
}
const OutOfLineBits* bits = outOfLineBits();
// value = true: casts to 1, then xors to 0, then negates to 0.
// value = false: casts to 0, then xors to 1, then negates to -1 (i.e. all one bits).
uintptr_t skipValue = -(static_cast<uintptr_t>(value) ^ 1);
size_t numWords = bits->numWords();
size_t wordIndex = startIndex / bitsInPointer();
size_t startIndexInWord = startIndex - wordIndex * bitsInPointer();
while (wordIndex < numWords) {
uintptr_t word = bits->bits()[wordIndex];
if (word != skipValue) {
size_t index = startIndexInWord;
if (findBitInWord(word, index, bitsInPointer(), value))
return wordIndex * bitsInPointer() + index;
}
wordIndex++;
startIndexInWord = 0;
}
return bits->numBits();
}
size_t findBitSimple(size_t index, bool value) const
{
while (index < size()) {
if (get(index) == value)
return index;
index++;
}
return size();
}
class OutOfLineBits {
public:
size_t numBits() const { return m_numBits; }
size_t numWords() const { return (m_numBits + bitsInPointer() - 1) / bitsInPointer(); }
uintptr_t* bits() { return bitwise_cast<uintptr_t*>(this + 1); }
const uintptr_t* bits() const { return bitwise_cast<const uintptr_t*>(this + 1); }
static WTF_EXPORT_PRIVATE OutOfLineBits* create(size_t numBits);
static WTF_EXPORT_PRIVATE void destroy(OutOfLineBits*);
private:
OutOfLineBits(size_t numBits)
: m_numBits(numBits)
{
}
size_t m_numBits;
};
bool isInline() const { return m_bitsOrPointer >> maxInlineBits(); }
const OutOfLineBits* outOfLineBits() const { return bitwise_cast<const OutOfLineBits*>(m_bitsOrPointer << 1); }
OutOfLineBits* outOfLineBits() { return bitwise_cast<OutOfLineBits*>(m_bitsOrPointer << 1); }
WTF_EXPORT_PRIVATE void resizeOutOfLine(size_t numBits);
WTF_EXPORT_PRIVATE void setSlow(const BitVector& other);
WTF_EXPORT_PRIVATE void mergeSlow(const BitVector& other);
WTF_EXPORT_PRIVATE void filterSlow(const BitVector& other);
WTF_EXPORT_PRIVATE void excludeSlow(const BitVector& other);
WTF_EXPORT_PRIVATE size_t bitCountSlow() const;
WTF_EXPORT_PRIVATE bool isEmptySlow() const;
WTF_EXPORT_PRIVATE bool equalsSlowCase(const BitVector& other) const;
bool equalsSlowCaseFast(const BitVector& other) const;
bool equalsSlowCaseSimple(const BitVector& other) const;
WTF_EXPORT_PRIVATE uintptr_t hashSlowCase() const;
uintptr_t* bits()
{
if (isInline())
return &m_bitsOrPointer;
return outOfLineBits()->bits();
}
const uintptr_t* bits() const
{
if (isInline())
return &m_bitsOrPointer;
return outOfLineBits()->bits();
}
uintptr_t m_bitsOrPointer;
};
struct BitVectorHash {
static unsigned hash(const BitVector& vector) { return vector.hash(); }
static bool equal(const BitVector& a, const BitVector& b) { return a == b; }
static constexpr bool safeToCompareToEmptyOrDeleted = false;
};
template<typename T> struct DefaultHash;
template<> struct DefaultHash<BitVector> : BitVectorHash { };
template<> struct HashTraits<BitVector> : public CustomHashTraits<BitVector> { };
} // namespace WTF
using WTF::BitVector;