// lock-free single-producer/single-consumer ringbuffer // this algorithm is implemented in various projects (linux kernel) // // Copyright (C) 2009-2013 Tim Blechmann // // Distributed under the Boost Software License, Version 1.0. (See // accompanying file LICENSE_1_0.txt or copy at // http://www.boost.org/LICENSE_1_0.txt) #ifndef BOOST_LOCKFREE_SPSC_QUEUE_HPP_INCLUDED #define BOOST_LOCKFREE_SPSC_QUEUE_HPP_INCLUDED #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef BOOST_HAS_PRAGMA_ONCE #pragma once #endif namespace boost { namespace lockfree { namespace detail { typedef parameter::parameters, boost::parameter::optional > ringbuffer_signature; template class ringbuffer_base { #ifndef BOOST_DOXYGEN_INVOKED protected: typedef std::size_t size_t; static const int padding_size = BOOST_LOCKFREE_CACHELINE_BYTES - sizeof(size_t); atomic write_index_; char padding1[padding_size]; /* force read_index and write_index to different cache lines */ atomic read_index_; BOOST_DELETED_FUNCTION(ringbuffer_base(ringbuffer_base const&)) BOOST_DELETED_FUNCTION(ringbuffer_base& operator= (ringbuffer_base const&)) protected: ringbuffer_base(void): write_index_(0), read_index_(0) {} static size_t next_index(size_t arg, size_t max_size) { size_t ret = arg + 1; while (unlikely(ret >= max_size)) ret -= max_size; return ret; } static size_t read_available(size_t write_index, size_t read_index, size_t max_size) { if (write_index >= read_index) return write_index - read_index; const size_t ret = write_index + max_size - read_index; return ret; } static size_t write_available(size_t write_index, size_t read_index, size_t max_size) { size_t ret = read_index - write_index - 1; if (write_index >= read_index) ret += max_size; return ret; } size_t read_available(size_t max_size) const { size_t write_index = write_index_.load(memory_order_acquire); const size_t read_index = read_index_.load(memory_order_relaxed); return read_available(write_index, read_index, max_size); } size_t write_available(size_t max_size) const { size_t write_index = write_index_.load(memory_order_relaxed); const size_t read_index = read_index_.load(memory_order_acquire); return write_available(write_index, read_index, max_size); } bool push(T const & t, T * buffer, size_t max_size) { const size_t write_index = write_index_.load(memory_order_relaxed); // only written from push thread const size_t next = next_index(write_index, max_size); if (next == read_index_.load(memory_order_acquire)) return false; /* ringbuffer is full */ new (buffer + write_index) T(t); // copy-construct write_index_.store(next, memory_order_release); return true; } size_t push(const T * input_buffer, size_t input_count, T * internal_buffer, size_t max_size) { return push(input_buffer, input_buffer + input_count, internal_buffer, max_size) - input_buffer; } template ConstIterator push(ConstIterator begin, ConstIterator end, T * internal_buffer, size_t max_size) { // FIXME: avoid std::distance const size_t write_index = write_index_.load(memory_order_relaxed); // only written from push thread const size_t read_index = read_index_.load(memory_order_acquire); const size_t avail = write_available(write_index, read_index, max_size); if (avail == 0) return begin; size_t input_count = std::distance(begin, end); input_count = (std::min)(input_count, avail); size_t new_write_index = write_index + input_count; const ConstIterator last = boost::next(begin, input_count); if (write_index + input_count > max_size) { /* copy data in two sections */ const size_t count0 = max_size - write_index; const ConstIterator midpoint = boost::next(begin, count0); std::uninitialized_copy(begin, midpoint, internal_buffer + write_index); std::uninitialized_copy(midpoint, last, internal_buffer); new_write_index -= max_size; } else { std::uninitialized_copy(begin, last, internal_buffer + write_index); if (new_write_index == max_size) new_write_index = 0; } write_index_.store(new_write_index, memory_order_release); return last; } template bool consume_one(Functor & functor, T * buffer, size_t max_size) { const size_t write_index = write_index_.load(memory_order_acquire); const size_t read_index = read_index_.load(memory_order_relaxed); // only written from pop thread if ( empty(write_index, read_index) ) return false; T & object_to_consume = buffer[read_index]; functor( object_to_consume ); object_to_consume.~T(); size_t next = next_index(read_index, max_size); read_index_.store(next, memory_order_release); return true; } template bool consume_one(Functor const & functor, T * buffer, size_t max_size) { const size_t write_index = write_index_.load(memory_order_acquire); const size_t read_index = read_index_.load(memory_order_relaxed); // only written from pop thread if ( empty(write_index, read_index) ) return false; T & object_to_consume = buffer[read_index]; functor( object_to_consume ); object_to_consume.~T(); size_t next = next_index(read_index, max_size); read_index_.store(next, memory_order_release); return true; } template size_t consume_all (Functor const & functor, T * internal_buffer, size_t max_size) { const size_t write_index = write_index_.load(memory_order_acquire); const size_t read_index = read_index_.load(memory_order_relaxed); // only written from pop thread const size_t avail = read_available(write_index, read_index, max_size); if (avail == 0) return 0; const size_t output_count = avail; size_t new_read_index = read_index + output_count; if (read_index + output_count > max_size) { /* copy data in two sections */ const size_t count0 = max_size - read_index; const size_t count1 = output_count - count0; run_functor_and_delete(internal_buffer + read_index, internal_buffer + max_size, functor); run_functor_and_delete(internal_buffer, internal_buffer + count1, functor); new_read_index -= max_size; } else { run_functor_and_delete(internal_buffer + read_index, internal_buffer + read_index + output_count, functor); if (new_read_index == max_size) new_read_index = 0; } read_index_.store(new_read_index, memory_order_release); return output_count; } template size_t consume_all (Functor & functor, T * internal_buffer, size_t max_size) { const size_t write_index = write_index_.load(memory_order_acquire); const size_t read_index = read_index_.load(memory_order_relaxed); // only written from pop thread const size_t avail = read_available(write_index, read_index, max_size); if (avail == 0) return 0; const size_t output_count = avail; size_t new_read_index = read_index + output_count; if (read_index + output_count > max_size) { /* copy data in two sections */ const size_t count0 = max_size - read_index; const size_t count1 = output_count - count0; run_functor_and_delete(internal_buffer + read_index, internal_buffer + max_size, functor); run_functor_and_delete(internal_buffer, internal_buffer + count1, functor); new_read_index -= max_size; } else { run_functor_and_delete(internal_buffer + read_index, internal_buffer + read_index + output_count, functor); if (new_read_index == max_size) new_read_index = 0; } read_index_.store(new_read_index, memory_order_release); return output_count; } size_t pop (T * output_buffer, size_t output_count, T * internal_buffer, size_t max_size) { const size_t write_index = write_index_.load(memory_order_acquire); const size_t read_index = read_index_.load(memory_order_relaxed); // only written from pop thread const size_t avail = read_available(write_index, read_index, max_size); if (avail == 0) return 0; output_count = (std::min)(output_count, avail); size_t new_read_index = read_index + output_count; if (read_index + output_count > max_size) { /* copy data in two sections */ const size_t count0 = max_size - read_index; const size_t count1 = output_count - count0; copy_and_delete(internal_buffer + read_index, internal_buffer + max_size, output_buffer); copy_and_delete(internal_buffer, internal_buffer + count1, output_buffer + count0); new_read_index -= max_size; } else { copy_and_delete(internal_buffer + read_index, internal_buffer + read_index + output_count, output_buffer); if (new_read_index == max_size) new_read_index = 0; } read_index_.store(new_read_index, memory_order_release); return output_count; } template size_t pop_to_output_iterator (OutputIterator it, T * internal_buffer, size_t max_size) { const size_t write_index = write_index_.load(memory_order_acquire); const size_t read_index = read_index_.load(memory_order_relaxed); // only written from pop thread const size_t avail = read_available(write_index, read_index, max_size); if (avail == 0) return 0; size_t new_read_index = read_index + avail; if (read_index + avail > max_size) { /* copy data in two sections */ const size_t count0 = max_size - read_index; const size_t count1 = avail - count0; it = copy_and_delete(internal_buffer + read_index, internal_buffer + max_size, it); copy_and_delete(internal_buffer, internal_buffer + count1, it); new_read_index -= max_size; } else { copy_and_delete(internal_buffer + read_index, internal_buffer + read_index + avail, it); if (new_read_index == max_size) new_read_index = 0; } read_index_.store(new_read_index, memory_order_release); return avail; } const T& front(const T * internal_buffer) const { const size_t read_index = read_index_.load(memory_order_relaxed); // only written from pop thread return *(internal_buffer + read_index); } T& front(T * internal_buffer) { const size_t read_index = read_index_.load(memory_order_relaxed); // only written from pop thread return *(internal_buffer + read_index); } #endif public: /** reset the ringbuffer * * \note Not thread-safe * */ void reset(void) { if ( !boost::has_trivial_destructor::value ) { // make sure to call all destructors! T dummy_element; while (pop(dummy_element)) {} } else { write_index_.store(0, memory_order_relaxed); read_index_.store(0, memory_order_release); } } /** Check if the ringbuffer is empty * * \return true, if the ringbuffer is empty, false otherwise * \note Due to the concurrent nature of the ringbuffer the result may be inaccurate. * */ bool empty(void) { return empty(write_index_.load(memory_order_relaxed), read_index_.load(memory_order_relaxed)); } /** * \return true, if implementation is lock-free. * * */ bool is_lock_free(void) const { return write_index_.is_lock_free() && read_index_.is_lock_free(); } private: bool empty(size_t write_index, size_t read_index) { return write_index == read_index; } template< class OutputIterator > OutputIterator copy_and_delete( T * first, T * last, OutputIterator out ) { if (boost::has_trivial_destructor::value) { return std::copy(first, last, out); // will use memcpy if possible } else { for (; first != last; ++first, ++out) { *out = *first; first->~T(); } return out; } } template< class Functor > void run_functor_and_delete( T * first, T * last, Functor & functor ) { for (; first != last; ++first) { functor(*first); first->~T(); } } template< class Functor > void run_functor_and_delete( T * first, T * last, Functor const & functor ) { for (; first != last; ++first) { functor(*first); first->~T(); } } }; template class compile_time_sized_ringbuffer: public ringbuffer_base { typedef std::size_t size_type; static const std::size_t max_size = MaxSize + 1; typedef typename boost::aligned_storage::value >::type storage_type; storage_type storage_; T * data() { return static_cast(storage_.address()); } const T * data() const { return static_cast(storage_.address()); } protected: size_type max_number_of_elements() const { return max_size; } public: bool push(T const & t) { return ringbuffer_base::push(t, data(), max_size); } template bool consume_one(Functor & f) { return ringbuffer_base::consume_one(f, data(), max_size); } template bool consume_one(Functor const & f) { return ringbuffer_base::consume_one(f, data(), max_size); } template size_type consume_all(Functor & f) { return ringbuffer_base::consume_all(f, data(), max_size); } template size_type consume_all(Functor const & f) { return ringbuffer_base::consume_all(f, data(), max_size); } size_type push(T const * t, size_type size) { return ringbuffer_base::push(t, size, data(), max_size); } template size_type push(T const (&t)[size]) { return push(t, size); } template ConstIterator push(ConstIterator begin, ConstIterator end) { return ringbuffer_base::push(begin, end, data(), max_size); } size_type pop(T * ret, size_type size) { return ringbuffer_base::pop(ret, size, data(), max_size); } template size_type pop_to_output_iterator(OutputIterator it) { return ringbuffer_base::pop_to_output_iterator(it, data(), max_size); } const T& front(void) const { return ringbuffer_base::front(data()); } T& front(void) { return ringbuffer_base::front(data()); } }; template class runtime_sized_ringbuffer: public ringbuffer_base, private Alloc { typedef std::size_t size_type; size_type max_elements_; typedef typename Alloc::pointer pointer; pointer array_; protected: size_type max_number_of_elements() const { return max_elements_; } public: explicit runtime_sized_ringbuffer(size_type max_elements): max_elements_(max_elements + 1) { array_ = Alloc::allocate(max_elements_); } template runtime_sized_ringbuffer(typename Alloc::template rebind::other const & alloc, size_type max_elements): Alloc(alloc), max_elements_(max_elements + 1) { array_ = Alloc::allocate(max_elements_); } runtime_sized_ringbuffer(Alloc const & alloc, size_type max_elements): Alloc(alloc), max_elements_(max_elements + 1) { array_ = Alloc::allocate(max_elements_); } ~runtime_sized_ringbuffer(void) { // destroy all remaining items T out; while (pop(&out, 1)) {} Alloc::deallocate(array_, max_elements_); } bool push(T const & t) { return ringbuffer_base::push(t, &*array_, max_elements_); } template bool consume_one(Functor & f) { return ringbuffer_base::consume_one(f, &*array_, max_elements_); } template bool consume_one(Functor const & f) { return ringbuffer_base::consume_one(f, &*array_, max_elements_); } template size_type consume_all(Functor & f) { return ringbuffer_base::consume_all(f, &*array_, max_elements_); } template size_type consume_all(Functor const & f) { return ringbuffer_base::consume_all(f, &*array_, max_elements_); } size_type push(T const * t, size_type size) { return ringbuffer_base::push(t, size, &*array_, max_elements_); } template size_type push(T const (&t)[size]) { return push(t, size); } template ConstIterator push(ConstIterator begin, ConstIterator end) { return ringbuffer_base::push(begin, end, array_, max_elements_); } size_type pop(T * ret, size_type size) { return ringbuffer_base::pop(ret, size, array_, max_elements_); } template size_type pop_to_output_iterator(OutputIterator it) { return ringbuffer_base::pop_to_output_iterator(it, array_, max_elements_); } const T& front(void) const { return ringbuffer_base::front(array_); } T& front(void) { return ringbuffer_base::front(array_); } }; template struct make_ringbuffer { typedef typename ringbuffer_signature::bind::type bound_args; typedef extract_capacity extract_capacity_t; static const bool runtime_sized = !extract_capacity_t::has_capacity; static const size_t capacity = extract_capacity_t::capacity; typedef extract_allocator extract_allocator_t; typedef typename extract_allocator_t::type allocator; // allocator argument is only sane, for run-time sized ringbuffers BOOST_STATIC_ASSERT((mpl::if_, mpl::bool_, mpl::true_ >::type::value)); typedef typename mpl::if_c, compile_time_sized_ringbuffer >::type ringbuffer_type; }; } /* namespace detail */ /** The spsc_queue class provides a single-writer/single-reader fifo queue, pushing and popping is wait-free. * * \b Policies: * - \c boost::lockfree::capacity<>, optional
* If this template argument is passed to the options, the size of the ringbuffer is set at compile-time. * * - \c boost::lockfree::allocator<>, defaults to \c boost::lockfree::allocator>
* Specifies the allocator that is used to allocate the ringbuffer. This option is only valid, if the ringbuffer is configured * to be sized at run-time * * \b Requirements: * - T must have a default constructor * - T must be copyable * */ #ifndef BOOST_DOXYGEN_INVOKED template #else template #endif class spsc_queue: public detail::make_ringbuffer::ringbuffer_type { private: #ifndef BOOST_DOXYGEN_INVOKED typedef typename detail::make_ringbuffer::ringbuffer_type base_type; static const bool runtime_sized = detail::make_ringbuffer::runtime_sized; typedef typename detail::make_ringbuffer::allocator allocator_arg; struct implementation_defined { typedef allocator_arg allocator; typedef std::size_t size_type; }; #endif public: typedef T value_type; typedef typename implementation_defined::allocator allocator; typedef typename implementation_defined::size_type size_type; /** Constructs a spsc_queue * * \pre spsc_queue must be configured to be sized at compile-time */ // @{ spsc_queue(void) { BOOST_ASSERT(!runtime_sized); } template explicit spsc_queue(typename allocator::template rebind::other const & alloc) { // just for API compatibility: we don't actually need an allocator BOOST_STATIC_ASSERT(!runtime_sized); } explicit spsc_queue(allocator const & alloc) { // just for API compatibility: we don't actually need an allocator BOOST_ASSERT(!runtime_sized); } // @} /** Constructs a spsc_queue for element_count elements * * \pre spsc_queue must be configured to be sized at run-time */ // @{ explicit spsc_queue(size_type element_count): base_type(element_count) { BOOST_ASSERT(runtime_sized); } template spsc_queue(size_type element_count, typename allocator::template rebind::other const & alloc): base_type(alloc, element_count) { BOOST_STATIC_ASSERT(runtime_sized); } spsc_queue(size_type element_count, allocator_arg const & alloc): base_type(alloc, element_count) { BOOST_ASSERT(runtime_sized); } // @} /** Pushes object t to the ringbuffer. * * \pre only one thread is allowed to push data to the spsc_queue * \post object will be pushed to the spsc_queue, unless it is full. * \return true, if the push operation is successful. * * \note Thread-safe and wait-free * */ bool push(T const & t) { return base_type::push(t); } /** Pops one object from ringbuffer. * * \pre only one thread is allowed to pop data to the spsc_queue * \post if ringbuffer is not empty, object will be discarded. * \return true, if the pop operation is successful, false if ringbuffer was empty. * * \note Thread-safe and wait-free */ bool pop () { detail::consume_noop consume_functor; return consume_one( consume_functor ); } /** Pops one object from ringbuffer. * * \pre only one thread is allowed to pop data to the spsc_queue * \post if ringbuffer is not empty, object will be copied to ret. * \return true, if the pop operation is successful, false if ringbuffer was empty. * * \note Thread-safe and wait-free */ template typename boost::enable_if::type, bool>::type pop (U & ret) { detail::consume_via_copy consume_functor(ret); return consume_one( consume_functor ); } /** Pushes as many objects from the array t as there is space. * * \pre only one thread is allowed to push data to the spsc_queue * \return number of pushed items * * \note Thread-safe and wait-free */ size_type push(T const * t, size_type size) { return base_type::push(t, size); } /** Pushes as many objects from the array t as there is space available. * * \pre only one thread is allowed to push data to the spsc_queue * \return number of pushed items * * \note Thread-safe and wait-free */ template size_type push(T const (&t)[size]) { return push(t, size); } /** Pushes as many objects from the range [begin, end) as there is space . * * \pre only one thread is allowed to push data to the spsc_queue * \return iterator to the first element, which has not been pushed * * \note Thread-safe and wait-free */ template ConstIterator push(ConstIterator begin, ConstIterator end) { return base_type::push(begin, end); } /** Pops a maximum of size objects from ringbuffer. * * \pre only one thread is allowed to pop data to the spsc_queue * \return number of popped items * * \note Thread-safe and wait-free * */ size_type pop(T * ret, size_type size) { return base_type::pop(ret, size); } /** Pops a maximum of size objects from spsc_queue. * * \pre only one thread is allowed to pop data to the spsc_queue * \return number of popped items * * \note Thread-safe and wait-free * */ template size_type pop(T (&ret)[size]) { return pop(ret, size); } /** Pops objects to the output iterator it * * \pre only one thread is allowed to pop data to the spsc_queue * \return number of popped items * * \note Thread-safe and wait-free * */ template typename boost::disable_if::type, size_type>::type pop(OutputIterator it) { return base_type::pop_to_output_iterator(it); } /** consumes one element via a functor * * pops one element from the queue and applies the functor on this object * * \returns true, if one element was consumed * * \note Thread-safe and non-blocking, if functor is thread-safe and non-blocking * */ template bool consume_one(Functor & f) { return base_type::consume_one(f); } /// \copydoc boost::lockfree::spsc_queue::consume_one(Functor & rhs) template bool consume_one(Functor const & f) { return base_type::consume_one(f); } /** consumes all elements via a functor * * sequentially pops all elements from the queue and applies the functor on each object * * \returns number of elements that are consumed * * \note Thread-safe and non-blocking, if functor is thread-safe and non-blocking * */ template size_type consume_all(Functor & f) { return base_type::consume_all(f); } /// \copydoc boost::lockfree::spsc_queue::consume_all(Functor & rhs) template size_type consume_all(Functor const & f) { return base_type::consume_all(f); } /** get number of elements that are available for read * * \return number of available elements that can be popped from the spsc_queue * * \note Thread-safe and wait-free, should only be called from the consumer thread * */ size_type read_available() const { return base_type::read_available(base_type::max_number_of_elements()); } /** get write space to write elements * * \return number of elements that can be pushed to the spsc_queue * * \note Thread-safe and wait-free, should only be called from the producer thread * */ size_type write_available() const { return base_type::write_available(base_type::max_number_of_elements()); } /** get reference to element in the front of the queue * * Availability of front element can be checked using read_available(). * * \pre only a consuming thread is allowed to check front element * \pre read_available() > 0. If ringbuffer is empty, it's undefined behaviour to invoke this method. * \return reference to the first element in the queue * * \note Thread-safe and wait-free */ const T& front() const { BOOST_ASSERT(read_available() > 0); return base_type::front(); } /// \copydoc boost::lockfree::spsc_queue::front() const T& front() { BOOST_ASSERT(read_available() > 0); return base_type::front(); } /** reset the ringbuffer * * \note Not thread-safe * */ void reset(void) { if ( !boost::has_trivial_destructor::value ) { // make sure to call all destructors! T dummy_element; while (pop(dummy_element)) {} } else { base_type::write_index_.store(0, memory_order_relaxed); base_type::read_index_.store(0, memory_order_release); } } }; } /* namespace lockfree */ } /* namespace boost */ #endif /* BOOST_LOCKFREE_SPSC_QUEUE_HPP_INCLUDED */