boost/lockfree/stack.hpp

//  Copyright (C) 2008-2013 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_STACK_HPP_INCLUDED
#define BOOST_LOCKFREE_STACK_HPP_INCLUDED

#include <boost/assert.hpp>
#include <boost/checked_delete.hpp>
#include <boost/core/no_exceptions_support.hpp>
#include <boost/integer_traits.hpp>
#include <boost/static_assert.hpp>
#include <boost/tuple/tuple.hpp>
#include <boost/type_traits/is_copy_constructible.hpp>

#include <boost/lockfree/detail/allocator_rebind_helper.hpp>
#include <boost/lockfree/detail/atomic.hpp>
#include <boost/lockfree/detail/copy_payload.hpp>
#include <boost/lockfree/detail/freelist.hpp>
#include <boost/lockfree/detail/parameter.hpp>
#include <boost/lockfree/detail/tagged_ptr.hpp>

#include <boost/lockfree/lockfree_forward.hpp>

#ifdef BOOST_HAS_PRAGMA_ONCE
#pragma once
#endif

namespace boost    {
namespace lockfree {
namespace detail   {

typedef parameter::parameters<boost::parameter::optional<tag::allocator>,
                              boost::parameter::optional<tag::capacity>
                             > stack_signature;

}

/** The stack class provides a multi-writer/multi-reader stack, pushing and popping is lock-free,
 *  construction/destruction has to be synchronized. It uses a freelist for memory management,
 *  freed nodes are pushed to the freelist and not returned to the OS before the stack is destroyed.
 *
 *  \b Policies:
 *
 *  - \c boost::lockfree::fixed_sized<>, defaults to \c boost::lockfree::fixed_sized<false> <br>
 *    Can be used to completely disable dynamic memory allocations during push in order to ensure lockfree behavior.<br>
 *    If the data structure is configured as fixed-sized, the internal nodes are stored inside an array and they are addressed
 *    by array indexing. This limits the possible size of the stack to the number of elements that can be addressed by the index
 *    type (usually 2**16-2), but on platforms that lack double-width compare-and-exchange instructions, this is the best way
 *    to achieve lock-freedom.
 *
 *  - \c boost::lockfree::capacity<>, optional <br>
 *    If this template argument is passed to the options, the size of the stack is set at compile-time. <br>
 *    It this option implies \c fixed_sized<true>
 *
 *  - \c boost::lockfree::allocator<>, defaults to \c boost::lockfree::allocator<std::allocator<void>> <br>
 *    Specifies the allocator that is used for the internal freelist
 *
 *  \b Requirements:
 *  - T must have a copy constructor
 * */
#ifdef BOOST_NO_CXX11_VARIADIC_TEMPLATES
template <typename T, class A0, class A1, class A2>
#else
template <typename T, typename ...Options>
#endif
class stack
{
private:
#ifndef BOOST_DOXYGEN_INVOKED
    BOOST_STATIC_ASSERT(boost::is_copy_constructible<T>::value);

#ifdef BOOST_NO_CXX11_VARIADIC_TEMPLATES
    typedef typename detail::stack_signature::bind<A0, A1, A2>::type bound_args;
#else
    typedef typename detail::stack_signature::bind<Options...>::type bound_args;
#endif

    static const bool has_capacity = detail::extract_capacity<bound_args>::has_capacity;
    static const size_t capacity = detail::extract_capacity<bound_args>::capacity;
    static const bool fixed_sized = detail::extract_fixed_sized<bound_args>::value;
    static const bool node_based = !(has_capacity || fixed_sized);
    static const bool compile_time_sized = has_capacity;

    struct node
    {
        node(T const & val):
            v(val)
        {}

        typedef typename detail::select_tagged_handle<node, node_based>::handle_type handle_t;
        handle_t next;
        const T v;
    };

    typedef typename detail::extract_allocator<bound_args, node>::type node_allocator;
    typedef typename detail::select_freelist<node, node_allocator, compile_time_sized, fixed_sized, capacity>::type pool_t;
    typedef typename pool_t::tagged_node_handle tagged_node_handle;

    // check compile-time capacity
    BOOST_STATIC_ASSERT((mpl::if_c<has_capacity,
                                   mpl::bool_<capacity - 1 < boost::integer_traits<boost::uint16_t>::const_max>,
                                   mpl::true_
                                  >::type::value));

    struct implementation_defined
    {
        typedef node_allocator allocator;
        typedef std::size_t size_type;
    };

#endif

    BOOST_DELETED_FUNCTION(stack(stack const&))
    BOOST_DELETED_FUNCTION(stack& operator= (stack const&))

public:
    typedef T value_type;
    typedef typename implementation_defined::allocator allocator;
    typedef typename implementation_defined::size_type size_type;

    /**
     * \return true, if implementation is lock-free.
     *
     * \warning It only checks, if the top stack node and the freelist can be modified in a lock-free manner.
     *          On most platforms, the whole implementation is lock-free, if this is true. Using c++0x-style atomics,
     *          there is no possibility to provide a completely accurate implementation, because one would need to test
     *          every internal node, which is impossible if further nodes will be allocated from the operating system.
     *
     * */
    bool is_lock_free (void) const
    {
        return tos.is_lock_free() && pool.is_lock_free();
    }

    //! Construct stack
    // @{
    stack(void):
        pool(node_allocator(), capacity)
    {
        BOOST_ASSERT(has_capacity);
        initialize();
    }

    template <typename U>
    explicit stack(typename detail::allocator_rebind_helper<node_allocator, U>::type const & alloc):
        pool(alloc, capacity)
    {
        BOOST_STATIC_ASSERT(has_capacity);
        initialize();
    }

    explicit stack(allocator const & alloc):
        pool(alloc, capacity)
    {
        BOOST_ASSERT(has_capacity);
        initialize();
    }
    // @}

    //! Construct stack, allocate n nodes for the freelist.
    // @{
    explicit stack(size_type n):
        pool(node_allocator(), n)
    {
        BOOST_ASSERT(!has_capacity);
        initialize();
    }

    template <typename U>
    stack(size_type n, typename detail::allocator_rebind_helper<node_allocator, U>::type const & alloc):
        pool(alloc, n)
    {
        BOOST_STATIC_ASSERT(!has_capacity);
        initialize();
    }
    // @}

    /** Allocate n nodes for freelist
     *
     *  \pre  only valid if no capacity<> argument given
     *  \note thread-safe, may block if memory allocator blocks
     *
     * */
    void reserve(size_type n)
    {
        BOOST_STATIC_ASSERT(!has_capacity);
        pool.template reserve<true>(n);
    }

    /** Allocate n nodes for freelist
     *
     *  \pre  only valid if no capacity<> argument given
     *  \note not thread-safe, may block if memory allocator blocks
     *
     * */
    void reserve_unsafe(size_type n)
    {
        BOOST_STATIC_ASSERT(!has_capacity);
        pool.template reserve<false>(n);
    }

    /** Destroys stack, free all nodes from freelist.
     *
     *  \note not thread-safe
     *
     * */
    ~stack(void)
    {
        T dummy;
        while(unsynchronized_pop(dummy))
        {}
    }

private:
#ifndef BOOST_DOXYGEN_INVOKED
    void initialize(void)
    {
        tos.store(tagged_node_handle(pool.null_handle(), 0));
    }

    void link_nodes_atomic(node * new_top_node, node * end_node)
    {
        tagged_node_handle old_tos = tos.load(detail::memory_order_relaxed);
        for (;;) {
            tagged_node_handle new_tos (pool.get_handle(new_top_node), old_tos.get_tag());
            end_node->next = pool.get_handle(old_tos);

            if (tos.compare_exchange_weak(old_tos, new_tos))
                break;
        }
    }

    void link_nodes_unsafe(node * new_top_node, node * end_node)
    {
        tagged_node_handle old_tos = tos.load(detail::memory_order_relaxed);

        tagged_node_handle new_tos (pool.get_handle(new_top_node), old_tos.get_tag());
        end_node->next = pool.get_pointer(old_tos);

        tos.store(new_tos, memory_order_relaxed);
    }

    template <bool Threadsafe, bool Bounded, typename ConstIterator>
    tuple<node*, node*> prepare_node_list(ConstIterator begin, ConstIterator end, ConstIterator & ret)
    {
        ConstIterator it = begin;
        node * end_node = pool.template construct<Threadsafe, Bounded>(*it++);
        if (end_node == NULL) {
            ret = begin;
            return make_tuple<node*, node*>(NULL, NULL);
        }

        node * new_top_node = end_node;
        end_node->next = NULL;

        BOOST_TRY {
            /* link nodes */
            for (; it != end; ++it) {
                node * newnode = pool.template construct<Threadsafe, Bounded>(*it);
                if (newnode == NULL)
                    break;
                newnode->next = new_top_node;
                new_top_node = newnode;
            }
        } BOOST_CATCH (...) {
            for (node * current_node = new_top_node; current_node != NULL;) {
                node * next = current_node->next;
                pool.template destruct<Threadsafe>(current_node);
                current_node = next;
            }
            BOOST_RETHROW;
        }
        BOOST_CATCH_END

        ret = it;
        return make_tuple(new_top_node, end_node);
    }
#endif

public:
    /** Pushes object t to the stack.
     *
     * \post object will be pushed to the stack, if internal node can be allocated
     * \returns true, if the push operation is successful.
     *
     * \note Thread-safe. If internal memory pool is exhausted and the memory pool is not fixed-sized, a new node will be allocated
     *                    from the OS. This may not be lock-free.
     * \throws if memory allocator throws
     * */
    bool push(T const & v)
    {
        return do_push<false>(v);
    }

    /** Pushes object t to the stack.
     *
     * \post object will be pushed to the stack, if internal node can be allocated
     * \returns true, if the push operation is successful.
     *
     * \note Thread-safe and non-blocking. If internal memory pool is exhausted, the push operation will fail
     * */
    bool bounded_push(T const & v)
    {
        return do_push<true>(v);
    }

#ifndef BOOST_DOXYGEN_INVOKED
private:
    template <bool Bounded>
    bool do_push(T const & v)
    {
        node * newnode = pool.template construct<true, Bounded>(v);
        if (newnode == 0)
            return false;

        link_nodes_atomic(newnode, newnode);
        return true;
    }

    template <bool Bounded, typename ConstIterator>
    ConstIterator do_push(ConstIterator begin, ConstIterator end)
    {
        node * new_top_node;
        node * end_node;
        ConstIterator ret;

        tie(new_top_node, end_node) = prepare_node_list<true, Bounded>(begin, end, ret);
        if (new_top_node)
            link_nodes_atomic(new_top_node, end_node);

        return ret;
    }

public:
#endif

    /** Pushes as many objects from the range [begin, end) as freelist node can be allocated.
     *
     * \return iterator to the first element, which has not been pushed
     *
     * \note Operation is applied atomically
     * \note Thread-safe. If internal memory pool is exhausted and the memory pool is not fixed-sized, a new node will be allocated
     *                    from the OS. This may not be lock-free.
     * \throws if memory allocator throws
     */
    template <typename ConstIterator>
    ConstIterator push(ConstIterator begin, ConstIterator end)
    {
        return do_push<false, ConstIterator>(begin, end);
    }

    /** Pushes as many objects from the range [begin, end) as freelist node can be allocated.
     *
     * \return iterator to the first element, which has not been pushed
     *
     * \note Operation is applied atomically
     * \note Thread-safe and non-blocking. If internal memory pool is exhausted, the push operation will fail
     * \throws if memory allocator throws
     */
    template <typename ConstIterator>
    ConstIterator bounded_push(ConstIterator begin, ConstIterator end)
    {
        return do_push<true, ConstIterator>(begin, end);
    }


    /** Pushes object t to the stack.
     *
     * \post object will be pushed to the stack, if internal node can be allocated
     * \returns true, if the push operation is successful.
     *
     * \note Not thread-safe. If internal memory pool is exhausted and the memory pool is not fixed-sized, a new node will be allocated
     *       from the OS. This may not be lock-free.
     * \throws if memory allocator throws
     * */
    bool unsynchronized_push(T const & v)
    {
        node * newnode = pool.template construct<false, false>(v);
        if (newnode == 0)
            return false;

        link_nodes_unsafe(newnode, newnode);
        return true;
    }

    /** Pushes as many objects from the range [begin, end) as freelist node can be allocated.
     *
     * \return iterator to the first element, which has not been pushed
     *
     * \note Not thread-safe. If internal memory pool is exhausted and the memory pool is not fixed-sized, a new node will be allocated
     *       from the OS. This may not be lock-free.
     * \throws if memory allocator throws
     */
    template <typename ConstIterator>
    ConstIterator unsynchronized_push(ConstIterator begin, ConstIterator end)
    {
        node * new_top_node;
        node * end_node;
        ConstIterator ret;

        tie(new_top_node, end_node) = prepare_node_list<false, false>(begin, end, ret);
        if (new_top_node)
            link_nodes_unsafe(new_top_node, end_node);

        return ret;
    }


    /** Pops object from stack.
     *
     * \post if pop operation is successful, object will be copied to ret.
     * \returns true, if the pop operation is successful, false if stack was empty.
     *
     * \note Thread-safe and non-blocking
     *
     * */
    bool pop(T & ret)
    {
        return pop<T>(ret);
    }

    /** Pops object from stack.
     *
     * \pre type T must be convertible to U
     * \post if pop operation is successful, object will be copied to ret.
     * \returns true, if the pop operation is successful, false if stack was empty.
     *
     * \note Thread-safe and non-blocking
     *
     * */
    template <typename U>
    bool pop(U & ret)
    {
        BOOST_STATIC_ASSERT((boost::is_convertible<T, U>::value));
        detail::consume_via_copy<U> consumer(ret);

        return consume_one(consumer);
    }


    /** Pops object from stack.
     *
     * \post if pop operation is successful, object will be copied to ret.
     * \returns true, if the pop operation is successful, false if stack was empty.
     *
     * \note Not thread-safe, but non-blocking
     *
     * */
    bool unsynchronized_pop(T & ret)
    {
        return unsynchronized_pop<T>(ret);
    }

    /** Pops object from stack.
     *
     * \pre type T must be convertible to U
     * \post if pop operation is successful, object will be copied to ret.
     * \returns true, if the pop operation is successful, false if stack was empty.
     *
     * \note Not thread-safe, but non-blocking
     *
     * */
    template <typename U>
    bool unsynchronized_pop(U & ret)
    {
        BOOST_STATIC_ASSERT((boost::is_convertible<T, U>::value));
        tagged_node_handle old_tos = tos.load(detail::memory_order_relaxed);
        node * old_tos_pointer = pool.get_pointer(old_tos);

        if (!pool.get_pointer(old_tos))
            return false;

        node * new_tos_ptr = pool.get_pointer(old_tos_pointer->next);
        tagged_node_handle new_tos(pool.get_handle(new_tos_ptr), old_tos.get_next_tag());

        tos.store(new_tos, memory_order_relaxed);
        detail::copy_payload(old_tos_pointer->v, ret);
        pool.template destruct<false>(old_tos);
        return true;
    }

    /** consumes one element via a functor
     *
     *  pops one element from the stack 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 <typename Functor>
    bool consume_one(Functor & f)
    {
        tagged_node_handle old_tos = tos.load(detail::memory_order_consume);

        for (;;) {
            node * old_tos_pointer = pool.get_pointer(old_tos);
            if (!old_tos_pointer)
                return false;

            tagged_node_handle new_tos(old_tos_pointer->next, old_tos.get_next_tag());

            if (tos.compare_exchange_weak(old_tos, new_tos)) {
                f(old_tos_pointer->v);
                pool.template destruct<true>(old_tos);
                return true;
            }
        }
    }

    /// \copydoc boost::lockfree::stack::consume_one(Functor & rhs)
    template <typename Functor>
    bool consume_one(Functor const & f)
    {
        tagged_node_handle old_tos = tos.load(detail::memory_order_consume);

        for (;;) {
            node * old_tos_pointer = pool.get_pointer(old_tos);
            if (!old_tos_pointer)
                return false;

            tagged_node_handle new_tos(old_tos_pointer->next, old_tos.get_next_tag());

            if (tos.compare_exchange_weak(old_tos, new_tos)) {
                f(old_tos_pointer->v);
                pool.template destruct<true>(old_tos);
                return true;
            }
        }
    }

    /** consumes all elements via a functor
     *
     * sequentially pops all elements from the stack 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 <typename Functor>
    size_t consume_all(Functor & f)
    {
        size_t element_count = 0;
        while (consume_one(f))
            element_count += 1;

        return element_count;
    }

    /// \copydoc boost::lockfree::stack::consume_all(Functor & rhs)
    template <typename Functor>
    size_t consume_all(Functor const & f)
    {
        size_t element_count = 0;
        while (consume_one(f))
            element_count += 1;

        return element_count;
    }

    /** consumes all elements via a functor
     *
     * atomically pops all elements from the stack 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 <typename Functor>
    size_t consume_all_atomic(Functor & f)
    {
        size_t element_count = 0;
        tagged_node_handle old_tos = tos.load(detail::memory_order_consume);

        for (;;) {
            node * old_tos_pointer = pool.get_pointer(old_tos);
            if (!old_tos_pointer)
                return 0;

            tagged_node_handle new_tos(pool.null_handle(), old_tos.get_next_tag());

            if (tos.compare_exchange_weak(old_tos, new_tos))
                break;
        }

        tagged_node_handle nodes_to_consume = old_tos;

        for(;;) {
            node * node_pointer = pool.get_pointer(nodes_to_consume);
            f(node_pointer->v);
            element_count += 1;

            node * next_node = pool.get_pointer(node_pointer->next);

            if (!next_node) {
                pool.template destruct<true>(nodes_to_consume);
                break;
            }

            tagged_node_handle next(pool.get_handle(next_node), nodes_to_consume.get_next_tag());
            pool.template destruct<true>(nodes_to_consume);
            nodes_to_consume = next;
        }

        return element_count;
    }

    /// \copydoc boost::lockfree::stack::consume_all_atomic(Functor & rhs)
    template <typename Functor>
    size_t consume_all_atomic(Functor const & f)
    {
        size_t element_count = 0;
        tagged_node_handle old_tos = tos.load(detail::memory_order_consume);

        for (;;) {
            node * old_tos_pointer = pool.get_pointer(old_tos);
            if (!old_tos_pointer)
                return 0;

            tagged_node_handle new_tos(pool.null_handle(), old_tos.get_next_tag());

            if (tos.compare_exchange_weak(old_tos, new_tos))
                break;
        }

        tagged_node_handle nodes_to_consume = old_tos;

        for(;;) {
            node * node_pointer = pool.get_pointer(nodes_to_consume);
            f(node_pointer->v);
            element_count += 1;

            node * next_node = pool.get_pointer(node_pointer->next);

            if (!next_node) {
                pool.template destruct<true>(nodes_to_consume);
                break;
            }

            tagged_node_handle next(pool.get_handle(next_node), nodes_to_consume.get_next_tag());
            pool.template destruct<true>(nodes_to_consume);
            nodes_to_consume = next;
        }

        return element_count;
    }

    /** consumes all elements via a functor
     *
     * atomically pops all elements from the stack and applies the functor on each object in reversed order
     *
     * \returns number of elements that are consumed
     *
     * \note Thread-safe and non-blocking, if functor is thread-safe and non-blocking
     * */
    template <typename Functor>
    size_t consume_all_atomic_reversed(Functor & f)
    {
        size_t element_count = 0;
        tagged_node_handle old_tos = tos.load(detail::memory_order_consume);

        for (;;) {
            node * old_tos_pointer = pool.get_pointer(old_tos);
            if (!old_tos_pointer)
                return 0;

            tagged_node_handle new_tos(pool.null_handle(), old_tos.get_next_tag());

            if (tos.compare_exchange_weak(old_tos, new_tos))
                break;
        }

        tagged_node_handle nodes_to_consume = old_tos;

        node * last_node_pointer = NULL;
        tagged_node_handle nodes_in_reversed_order;
        for(;;) {
            node * node_pointer = pool.get_pointer(nodes_to_consume);
            node * next_node    = pool.get_pointer(node_pointer->next);

            node_pointer->next  = pool.get_handle(last_node_pointer);
            last_node_pointer   = node_pointer;

            if (!next_node) {
                nodes_in_reversed_order = nodes_to_consume;
                break;
            }

            tagged_node_handle next(pool.get_handle(next_node), nodes_to_consume.get_next_tag());
            nodes_to_consume = next;
        }

        for(;;) {
            node * node_pointer = pool.get_pointer(nodes_in_reversed_order);
            f(node_pointer->v);
            element_count += 1;

            node * next_node = pool.get_pointer(node_pointer->next);

            if (!next_node) {
                pool.template destruct<true>(nodes_in_reversed_order);
                break;
            }

            tagged_node_handle next(pool.get_handle(next_node), nodes_in_reversed_order.get_next_tag());
            pool.template destruct<true>(nodes_in_reversed_order);
            nodes_in_reversed_order = next;
        }

        return element_count;
    }

    /// \copydoc boost::lockfree::stack::consume_all_atomic_reversed(Functor & rhs)
    template <typename Functor>
    size_t consume_all_atomic_reversed(Functor const & f)
    {
        size_t element_count = 0;
        tagged_node_handle old_tos = tos.load(detail::memory_order_consume);

        for (;;) {
            node * old_tos_pointer = pool.get_pointer(old_tos);
            if (!old_tos_pointer)
                return 0;

            tagged_node_handle new_tos(pool.null_handle(), old_tos.get_next_tag());

            if (tos.compare_exchange_weak(old_tos, new_tos))
                break;
        }

        tagged_node_handle nodes_to_consume = old_tos;

        node * last_node_pointer = NULL;
        tagged_node_handle nodes_in_reversed_order;
        for(;;) {
            node * node_pointer = pool.get_pointer(nodes_to_consume);
            node * next_node    = pool.get_pointer(node_pointer->next);

            node_pointer->next  = pool.get_handle(last_node_pointer);
            last_node_pointer   = node_pointer;

            if (!next_node) {
                nodes_in_reversed_order = nodes_to_consume;
                break;
            }

            tagged_node_handle next(pool.get_handle(next_node), nodes_to_consume.get_next_tag());
            nodes_to_consume = next;
        }

        for(;;) {
            node * node_pointer = pool.get_pointer(nodes_in_reversed_order);
            f(node_pointer->v);
            element_count += 1;

            node * next_node = pool.get_pointer(node_pointer->next);

            if (!next_node) {
                pool.template destruct<true>(nodes_in_reversed_order);
                break;
            }

            tagged_node_handle next(pool.get_handle(next_node), nodes_in_reversed_order.get_next_tag());
            pool.template destruct<true>(nodes_in_reversed_order);
            nodes_in_reversed_order = next;
        }

        return element_count;
    }
    /**
     * \return true, if stack is empty.
     *
     * \note It only guarantees that at some point during the execution of the function the stack has been empty.
     *       It is rarely practical to use this value in program logic, because the stack can be modified by other threads.
     * */
    bool empty(void) const
    {
        return pool.get_pointer(tos.load()) == NULL;
    }

private:
#ifndef BOOST_DOXYGEN_INVOKED
    detail::atomic<tagged_node_handle> tos;

    static const int padding_size = BOOST_LOCKFREE_CACHELINE_BYTES - sizeof(tagged_node_handle);
    char padding[padding_size];

    pool_t pool;
#endif
};

} /* namespace lockfree */
} /* namespace boost */

#endif /* BOOST_LOCKFREE_STACK_HPP_INCLUDED */