/* * Copyright 2016 Facebook, Inc. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. * You may obtain a copy of the License at * * http://www.apache.org/licenses/LICENSE-2.0 * * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * See the License for the specific language governing permissions and * limitations under the License. */ /** * AtomicHashArray is the building block for AtomicHashMap. It provides the * core lock-free functionality, but is limited by the fact that it cannot * grow past its initialization size and is a little more awkward (no public * constructor, for example). If you're confident that you won't run out of * space, don't mind the awkardness, and really need bare-metal performance, * feel free to use AHA directly. * * Check out AtomicHashMap.h for more thorough documentation on perf and * general pros and cons relative to other hash maps. * * @author Spencer Ahrens * @author Jordan DeLong */ #pragma once #define FOLLY_ATOMICHASHARRAY_H_ #include #include #include #include #include namespace folly { struct AtomicHashArrayLinearProbeFcn { inline size_t operator()(size_t idx, size_t /* numProbes */, size_t capacity) const { idx += 1; // linear probing // Avoid modulus because it's slow return LIKELY(idx < capacity) ? idx : (idx - capacity); } }; struct AtomicHashArrayQuadraticProbeFcn { inline size_t operator()(size_t idx, size_t numProbes, size_t capacity) const{ idx += numProbes; // quadratic probing // Avoid modulus because it's slow return LIKELY(idx < capacity) ? idx : (idx - capacity); } }; // Enables specializing checkLegalKey without specializing its class. namespace detail { // Local copy of folly::gen::Identity, to avoid heavy dependencies. class AHAIdentity { public: template auto operator()(Value&& value) const -> decltype(std::forward(value)) { return std::forward(value); } }; template inline void checkLegalKeyIfKeyTImpl(NotKeyT /* ignored */, KeyT /* emptyKey */, KeyT /* lockedKey */, KeyT /* erasedKey */) {} template inline void checkLegalKeyIfKeyTImpl(KeyT key_in, KeyT emptyKey, KeyT lockedKey, KeyT erasedKey) { DCHECK_NE(key_in, emptyKey); DCHECK_NE(key_in, lockedKey); DCHECK_NE(key_in, erasedKey); } } // namespace detail template , class EqualFcn = std::equal_to, class Allocator = std::allocator, class ProbeFcn = AtomicHashArrayLinearProbeFcn, class KeyConvertFcn = detail::AHAIdentity> class AtomicHashMap; template , class EqualFcn = std::equal_to, class Allocator = std::allocator, class ProbeFcn = AtomicHashArrayLinearProbeFcn, class KeyConvertFcn = detail::AHAIdentity> class AtomicHashArray : boost::noncopyable { static_assert((std::is_convertible::value || std::is_convertible::value || std::is_convertible::value), "You are trying to use AtomicHashArray with disallowed key " "types. You must use atomically compare-and-swappable integer " "keys, or a different container class."); public: typedef KeyT key_type; typedef ValueT mapped_type; typedef HashFcn hasher; typedef EqualFcn key_equal; typedef KeyConvertFcn key_convert; typedef std::pair value_type; typedef std::size_t size_type; typedef std::ptrdiff_t difference_type; typedef value_type& reference; typedef const value_type& const_reference; typedef value_type* pointer; typedef const value_type* const_pointer; const size_t capacity_; const size_t maxEntries_; const KeyT kEmptyKey_; const KeyT kLockedKey_; const KeyT kErasedKey_; template struct aha_iterator; typedef aha_iterator const_iterator; typedef aha_iterator iterator; // You really shouldn't need this if you use the SmartPtr provided by create, // but if you really want to do something crazy like stick the released // pointer into a DescriminatedPtr or something, you'll need this to clean up // after yourself. static void destroy(AtomicHashArray*); private: const size_t kAnchorMask_; struct Deleter { void operator()(AtomicHashArray* ptr) { AtomicHashArray::destroy(ptr); } }; public: typedef std::unique_ptr SmartPtr; /* * create -- * * Creates AtomicHashArray objects. Use instead of constructor/destructor. * * We do things this way in order to avoid the perf penalty of a second * pointer indirection when composing these into AtomicHashMap, which needs * to store an array of pointers so that it can perform atomic operations on * them when growing. * * Instead of a mess of arguments, we take a max size and a Config struct to * simulate named ctor parameters. The Config struct has sensible defaults * for everything, but is overloaded - if you specify a positive capacity, * that will be used directly instead of computing it based on * maxLoadFactor. * * Create returns an AHA::SmartPtr which is a unique_ptr with a custom * deleter to make sure everything is cleaned up properly. */ struct Config { KeyT emptyKey; KeyT lockedKey; KeyT erasedKey; double maxLoadFactor; double growthFactor; int entryCountThreadCacheSize; size_t capacity; // if positive, overrides maxLoadFactor public: // Cannot have constexpr ctor because some compilers rightly complain. Config() : emptyKey((KeyT)-1), lockedKey((KeyT)-2), erasedKey((KeyT)-3), maxLoadFactor(0.8), growthFactor(-1), entryCountThreadCacheSize(1000), capacity(0) {} }; // Cannot have pre-instantiated const Config instance because of SIOF. static SmartPtr create(size_t maxSize, const Config& c = Config()); /* * find -- * * * Returns the iterator to the element if found, otherwise end(). * * As an optional feature, the type of the key to look up (LookupKeyT) is * allowed to be different from the type of keys actually stored (KeyT). * * This enables use cases where materializing the key is costly and usually * redudant, e.g., canonicalizing/interning a set of strings and being able * to look up by StringPiece. To use this feature, LookupHashFcn must take * a LookupKeyT, and LookupEqualFcn must take KeyT and LookupKeyT as first * and second parameter, respectively. * * See folly/test/ArrayHashArrayTest.cpp for sample usage. */ template iterator find(LookupKeyT k) { return iterator(this, findInternal(k).idx); } template const_iterator find(LookupKeyT k) const { return const_cast(this)-> find(k); } /* * insert -- * * Returns a pair with iterator to the element at r.first and bool success. * Retrieve the index with ret.first.getIndex(). * * Fails on key collision (does not overwrite) or if map becomes * full, at which point no element is inserted, iterator is set to end(), * and success is set false. On collisions, success is set false, but the * iterator is set to the existing entry. */ std::pair insert(const value_type& r) { return emplace(r.first, r.second); } std::pair insert(value_type&& r) { return emplace(r.first, std::move(r.second)); } /* * emplace -- * * Same contract as insert(), but performs in-place construction * of the value type using the specified arguments. * * Also, like find(), this method optionally allows 'key_in' to have a type * different from that stored in the table; see find(). If and only if no * equal key is already present, this method converts 'key_in' to a key of * type KeyT using the provided LookupKeyToKeyFcn. */ template std::pair emplace(LookupKeyT key_in, ArgTs&&... vCtorArgs) { SimpleRetT ret = insertInternal( key_in, std::forward(vCtorArgs)...); return std::make_pair(iterator(this, ret.idx), ret.success); } // returns the number of elements erased - should never exceed 1 size_t erase(KeyT k); // clears all keys and values in the map and resets all counters. Not thread // safe. void clear(); // Exact number of elements in the map - note that readFull() acquires a // mutex. See folly/ThreadCachedInt.h for more details. size_t size() const { return numEntries_.readFull() - numErases_.load(std::memory_order_relaxed); } bool empty() const { return size() == 0; } iterator begin() { iterator it(this, 0); it.advancePastEmpty(); return it; } const_iterator begin() const { const_iterator it(this, 0); it.advancePastEmpty(); return it; } iterator end() { return iterator(this, capacity_); } const_iterator end() const { return const_iterator(this, capacity_); } // See AtomicHashMap::findAt - access elements directly // WARNING: The following 2 functions will fail silently for hashtable // with capacity > 2^32 iterator findAt(uint32_t idx) { DCHECK_LT(idx, capacity_); return iterator(this, idx); } const_iterator findAt(uint32_t idx) const { return const_cast(this)->findAt(idx); } iterator makeIter(size_t idx) { return iterator(this, idx); } const_iterator makeIter(size_t idx) const { return const_iterator(this, idx); } // The max load factor allowed for this map double maxLoadFactor() const { return ((double) maxEntries_) / capacity_; } void setEntryCountThreadCacheSize(uint32_t newSize) { numEntries_.setCacheSize(newSize); numPendingEntries_.setCacheSize(newSize); } int getEntryCountThreadCacheSize() const { return numEntries_.getCacheSize(); } /* Private data and helper functions... */ private: friend class AtomicHashMap; struct SimpleRetT { size_t idx; bool success; SimpleRetT(size_t i, bool s) : idx(i), success(s) {} SimpleRetT() = default; }; template SimpleRetT insertInternal(LookupKeyT key, ArgTs&&... vCtorArgs); template SimpleRetT findInternal(const LookupKeyT key); template void checkLegalKeyIfKey(MaybeKeyT key) { detail::checkLegalKeyIfKeyTImpl(key, kEmptyKey_, kLockedKey_, kErasedKey_); } static std::atomic* cellKeyPtr(const value_type& r) { // We need some illegal casting here in order to actually store // our value_type as a std::pair. But a little bit of // undefined behavior never hurt anyone ... static_assert(sizeof(std::atomic) == sizeof(KeyT), "std::atomic is implemented in an unexpected way for AHM"); return const_cast*>( reinterpret_cast const*>(&r.first)); } static KeyT relaxedLoadKey(const value_type& r) { return cellKeyPtr(r)->load(std::memory_order_relaxed); } static KeyT acquireLoadKey(const value_type& r) { return cellKeyPtr(r)->load(std::memory_order_acquire); } // Fun with thread local storage - atomic increment is expensive // (relatively), so we accumulate in the thread cache and periodically // flush to the actual variable, and walk through the unflushed counts when // reading the value, so be careful of calling size() too frequently. This // increases insertion throughput several times over while keeping the count // accurate. ThreadCachedInt numEntries_; // Successful key inserts ThreadCachedInt numPendingEntries_; // Used by insertInternal std::atomic isFull_; // Used by insertInternal std::atomic numErases_; // Successful key erases value_type cells_[0]; // This must be the last field of this class // Force constructor/destructor private since create/destroy should be // used externally instead AtomicHashArray(size_t capacity, KeyT emptyKey, KeyT lockedKey, KeyT erasedKey, double maxLoadFactor, size_t cacheSize); ~AtomicHashArray() = default; inline void unlockCell(value_type* const cell, KeyT newKey) { cellKeyPtr(*cell)->store(newKey, std::memory_order_release); } inline bool tryLockCell(value_type* const cell) { KeyT expect = kEmptyKey_; return cellKeyPtr(*cell)->compare_exchange_strong(expect, kLockedKey_, std::memory_order_acq_rel); } template inline size_t keyToAnchorIdx(const LookupKeyT k) const { const size_t hashVal = LookupHashFcn()(k); const size_t probe = hashVal & kAnchorMask_; return LIKELY(probe < capacity_) ? probe : hashVal % capacity_; } }; // AtomicHashArray } // namespace folly #include