Added wait strategies to flat combining technique
authorkhizmax <libcds.dev@gmail.com>
Mon, 13 Jun 2016 20:24:09 +0000 (23:24 +0300)
committerkhizmax <libcds.dev@gmail.com>
Mon, 13 Jun 2016 20:24:09 +0000 (23:24 +0300)
31 files changed:
cds/algo/flat_combining.h
cds/algo/flat_combining/defs.h [new file with mode: 0644]
cds/algo/flat_combining/kernel.h [new file with mode: 0644]
cds/algo/flat_combining/wait_strategy.h [new file with mode: 0644]
cds/algo/wait_strategy.h [deleted file]
cds/container/fcdeque.h
cds/container/fcpriority_queue.h
cds/container/fcqueue.h
cds/container/fcstack.h
cds/intrusive/fcqueue.h
cds/intrusive/fcstack.h
projects/Win/vc14/cds.sln
projects/Win/vc14/cds.vcxproj
projects/Win/vc14/cds.vcxproj.filters
projects/Win/vc14/stress-queue.vcxproj
test/include/cds_test/stat_flat_combining_out.h [new file with mode: 0644]
test/stress/pqueue/pqueue_type.h
test/stress/queue/intrusive_push_pop.cpp
test/stress/queue/intrusive_queue_type.h
test/stress/queue/queue_type.h
test/stress/stack/intrusive_stack_type.h
test/stress/stack/stack_type.h
test/unit/deque/fcdeque.cpp
test/unit/pqueue/fcpqueue_boost_stable_vector.cpp
test/unit/pqueue/fcpqueue_deque.cpp
test/unit/pqueue/fcpqueue_vector.cpp
test/unit/queue/fcqueue.cpp
test/unit/queue/intrusive_fcqueue.cpp
test/unit/queue/segmented_queue_dhp.cpp
test/unit/stack/fcstack.cpp
test/unit/stack/intrusive_fcstack.cpp

index c033d6c02e2bcf607b347149cab2e59f4dedc44d..4c3ac202ac2b4de97b505d3f5edfe06e141c7013 100644 (file)
     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.     
+    OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 */
 
 #ifndef CDSLIB_ALGO_FLAT_COMBINING_H
 #define CDSLIB_ALGO_FLAT_COMBINING_H
 
-#include <mutex>
-#include <cds/algo/atomic.h>
-#include <cds/details/allocator.h>
-#include <cds/algo/backoff_strategy.h>
-#include <cds/sync/spinlock.h>
-#include <cds/opt/options.h>
-#include <cds/algo/int_algo.h>
-#include <boost/thread/tss.hpp>     // thread_specific_ptr
-
-namespace cds { namespace algo {
-
-    /// @defgroup cds_flat_combining_intrusive Intrusive flat combining containers
-    /// @defgroup cds_flat_combining_container Non-intrusive flat combining containers
-
-    /// Flat combining
-    /**
-        @anchor cds_flat_combining_description
-        Flat combining (FC) technique is invented by Hendler, Incze, Shavit and Tzafrir in their paper
-        [2010] <i>"Flat Combining and the Synchronization-Parallelism Tradeoff"</i>.
-        The technique converts a sequential data structure to its concurrent implementation.
-        A few structures are added to the sequential implementation: a <i>global lock</i>,
-        a <i>count</i> of the number of combining passes, and a pointer to the <i>head</i>
-        of a <i>publication list</i>. The publication list is a list of thread-local records
-        of a size proportional to the number of threads that are concurrently accessing the shared object.
-
-        Each thread \p t accessing the structure to perform an invocation of some method \p m
-        on the shared object executes the following sequence of steps:
-        <ol>
-        <li>Write the invocation opcode and parameters (if any) of the method \p m to be applied
-        sequentially to the shared object in the <i>request</i> field of your thread local publication
-        record (there is no need to use a load-store memory barrier). The <i>request</i> field will later
-        be used to receive the response. If your thread local publication record is marked as active
-        continue to step 2, otherwise continue to step 5.</li>
-        <li>Check if the global lock is taken. If so (another thread is an active combiner), spin on the <i>request</i>
-        field waiting for a response to the invocation (one can add a yield at this point to allow other threads
-        on the same core to run). Once in a while while spinning check if the lock is still taken and that your
-        record is active. If your record is inactive proceed to step 5. Once the response is available,
-        reset the request field to null and return the response.</li>
-        <li>If the lock is not taken, attempt to acquire it and become a combiner. If you fail,
-        return to spinning in step 2.</li>
-        <li>Otherwise, you hold the lock and are a combiner.
-        <ul>
-            <li>Increment the combining pass count by one.</li>
-            <li>Execute a \p fc_apply() by traversing the publication list from the head,
-            combining all nonnull method call invocations, setting the <i>age</i> of each of these records
-            to the current <i>count</i>, applying the combined method calls to the structure D, and returning
-            responses to all the invocations. This traversal is guaranteed to be wait-free.</li>
-            <li>If the <i>count</i> is such that a cleanup needs to be performed, traverse the publication
-            list from the <i>head</i>. Starting from the second item (we always leave the item pointed to
-            by the head in the list), remove from the publication list all records whose <i>age</i> is
-            much smaller than the current <i>count</i>. This is done by removing the node and marking it
-            as inactive.</li>
-            <li>Release the lock.</li>
-        </ul>
-        <li>If you have no thread local publication record allocate one, marked as active. If you already
-        have one marked as inactive, mark it as active. Execute a store-load memory barrier. Proceed to insert
-        the record into the list with a successful CAS to the <i>head</i>. Then proceed to step 1.</li>
-        </ol>
-
-        As the test results show, the flat combining technique is suitable for non-intrusive containers
-        like stack, queue, deque. For intrusive concurrent containers the flat combining demonstrates
-        less impressive results.
-
-        \ref cds_flat_combining_container "List of FC-based containers" in libcds.
-
-        \ref cds_flat_combining_intrusive "List of intrusive FC-based containers" in libcds.
-    */
-    namespace flat_combining {
-
-        /// Special values of publication_record::nRequest
-        enum request_value
-        {
-            req_EmptyRecord,    ///< Publication record is empty
-            req_Response,       ///< Operation is done
-
-            req_Operation       ///< First operation id for derived classes
-        };
-
-        /// publication_record state
-        enum record_state {
-            inactive,       ///< Record is inactive
-            active,         ///< Record is active
-            removed         ///< Record should be removed
-        };
-
-        /// Record of publication list
-        /**
-            Each data structure based on flat combining contains a class derived from \p %publication_record
-        */
-        struct publication_record {
-            atomics::atomic<unsigned int>    nRequest;   ///< Request field (depends on data structure)
-            atomics::atomic<unsigned int>    nState;     ///< Record state: inactive, active, removed
-            atomics::atomic<unsigned int>    nAge;       ///< Age of the record
-            atomics::atomic<publication_record *> pNext; ///< Next record in publication list
-            void *                              pOwner;    ///< [internal data] Pointer to \ref kernel object that manages the publication list
-
-            /// Initializes publication record
-            publication_record()
-                : nRequest( req_EmptyRecord )
-                , nState( inactive )
-                , nAge(0)
-                , pNext( nullptr )
-                , pOwner( nullptr )
-            {}
-
-            /// Returns the value of \p nRequest field
-            unsigned int op() const
-            {
-                return nRequest.load( atomics::memory_order_relaxed );
-            }
-
-            /// Checks if the operation is done
-            bool is_done() const
-            {
-                return nRequest.load( atomics::memory_order_relaxed ) == req_Response;
-            }
-        };
-
-        /// Flat combining internal statistics
-        template <typename Counter = cds::atomicity::event_counter >
-        struct stat
-        {
-            typedef Counter counter_type;   ///< Event counter type
-
-            counter_type    m_nOperationCount   ;   ///< How many operations have been performed
-            counter_type    m_nCombiningCount   ;   ///< Combining call count
-            counter_type    m_nCompactPublicationList; ///< Count of publication list compacting
-            counter_type    m_nDeactivatePubRecord; ///< How many publication records were deactivated during compacting
-            counter_type    m_nActivatePubRecord;   ///< Count of publication record activating
-            counter_type    m_nPubRecordCreated ;   ///< Count of created publication records
-            counter_type    m_nPubRecordDeteted ;   ///< Count of deleted publication records
-            counter_type    m_nAcquirePubRecCount;  ///< Count of acquiring publication record
-            counter_type    m_nReleasePubRecCount;  ///< Count on releasing publication record
-
-            /// Returns current combining factor
-            /**
-                Combining factor is how many operations perform in one combine pass:
-                <tt>combining_factor := m_nOperationCount / m_nCombiningCount</tt>
-            */
-            double combining_factor() const
-            {
-                return m_nCombiningCount.get() ? double( m_nOperationCount.get()) / m_nCombiningCount.get() : 0.0;
-            }
-
-            //@cond
-            void    onOperation()               { ++m_nOperationCount; }
-            void    onCombining()               { ++m_nCombiningCount; }
-            void    onCompactPublicationList()  { ++m_nCompactPublicationList; }
-            void    onDeactivatePubRecord()     { ++m_nDeactivatePubRecord; }
-            void    onActivatePubRecord()       { ++m_nActivatePubRecord; }
-            void    onCreatePubRecord()         { ++m_nPubRecordCreated; }
-            void    onDeletePubRecord()         { ++m_nPubRecordDeteted; }
-            void    onAcquirePubRecord()        { ++m_nAcquirePubRecCount; }
-            void    onReleasePubRecord()        { ++m_nReleasePubRecCount; }
-            //@endcond
-        };
-
-        /// Flat combining dummy internal statistics
-        struct empty_stat
-        {
-            //@cond
-            void    onOperation()               {}
-            void    onCombining()               {}
-            void    onCompactPublicationList()  {}
-            void    onDeactivatePubRecord()     {}
-            void    onActivatePubRecord()       {}
-            void    onCreatePubRecord()         {}
-            void    onDeletePubRecord()         {}
-            void    onAcquirePubRecord()        {}
-            void    onReleasePubRecord()        {}
-            //@endcond
-        };
-
-        /// Type traits of \ref kernel class
-        /**
-            You can define different type traits for \ref kernel
-            by specifying your struct based on \p %traits
-            or by using \ref make_traits metafunction.
-        */
-        struct traits
-        {
-            typedef cds::sync::spin             lock_type;  ///< Lock type
-            typedef cds::backoff::delay_of<2>   back_off;   ///< Back-off strategy
-            typedef CDS_DEFAULT_ALLOCATOR       allocator;  ///< Allocator used for TLS data (allocating publication_record derivatives)
-            typedef empty_stat                  stat;       ///< Internal statistics
-            typedef opt::v::relaxed_ordering  memory_model; ///< /// C++ memory ordering model
-        };
-
-        /// Metafunction converting option list to traits
-        /**
-            \p Options are:
-            - \p opt::lock_type - mutex type, default is \p cds::sync::spin
-            - \p opt::back_off - back-off strategy, defalt is \p cds::backoff::delay_of<2>
-            - \p opt::allocator - allocator type, default is \ref CDS_DEFAULT_ALLOCATOR
-            - \p opt::stat - internal statistics, possible type: \ref stat, \ref empty_stat (the default)
-            - \p opt::memory_model - C++ memory ordering model.
-                List of all available memory ordering see \p opt::memory_model.
-                Default is \p cds::opt::v::relaxed_ordering
-        */
-        template <typename... Options>
-        struct make_traits {
-#   ifdef CDS_DOXYGEN_INVOKED
-            typedef implementation_defined type ;   ///< Metafunction result
-#   else
-            typedef typename cds::opt::make_options<
-                typename cds::opt::find_type_traits< traits, Options... >::type
-                ,Options...
-            >::type   type;
-#   endif
-        };
-
-        /// The kernel of flat combining
-        /**
-            Template parameters:
-            - \p PublicationRecord - a type derived from \ref publication_record
-            - \p Traits - a type traits of flat combining, default is \p flat_combining::traits.
-                \ref make_traits metafunction can be used to create type traits
-
-            The kernel object should be a member of a container class. The container cooperates with flat combining
-            kernel object. There are two ways to interact with the kernel:
-            - One-by-one processing the active records of the publication list. This mode provides by \p combine() function:
-              the container acquires its publication record by \p acquire_record(), fills its fields and calls
-              \p combine() function of its kernel object. If the current thread becomes a combiner, the kernel
-              calls \p fc_apply() function of the container for each active non-empty record. Then, the container
-              should release its publication record by \p release_record(). Only one pass through the publication
-              list is possible.
-            - Batch processing - \p batch_combine() function. It this mode the container obtains access
-              to entire publication list. This mode allows the container to perform an elimination, for example,
-              the stack can collide \p push() and \p pop() requests. The sequence of invocations is the following:
-              the container acquires its publication record by \p acquire_record(), fills its field and call
-              \p batch_combine() function of its kernel object. If the current thread becomes a combiner,
-              the kernel calls \p fc_process() function of the container passing two iterators pointing to
-              the begin and the end of publication list (see \ref iterator class). The iterators allow
-              multiple pass through active records of publication list. For each processed record the container
-              should call \p operation_done() function. On the end, the container should release
-              its record by \p release_record().
-        */
-        template <
-            typename PublicationRecord
-            ,typename Traits = traits
-        >
-        class kernel
-        {
-        public:
-            typedef PublicationRecord   publication_record_type;   ///< publication record type
-            typedef Traits   traits;                               ///< Type traits
-            typedef typename traits::lock_type global_lock_type;   ///< Global lock type
-            typedef typename traits::back_off  back_off;           ///< back-off strategy type
-            typedef typename traits::allocator allocator;          ///< Allocator type (used for allocating publication_record_type data)
-            typedef typename traits::stat      stat;               ///< Internal statistics
-            typedef typename traits::memory_model memory_model;    ///< C++ memory model
-
-        protected:
-            //@cond
-            typedef cds::details::Allocator< publication_record_type, allocator >   cxx11_allocator; ///< internal helper cds::details::Allocator
-            typedef std::lock_guard<global_lock_type> lock_guard;
-            //@endcond
-
-        protected:
-            atomics::atomic<unsigned int>  m_nCount;   ///< Total count of combining passes. Used as an age.
-            publication_record_type *   m_pHead;    ///< Head of publication list
-            boost::thread_specific_ptr< publication_record_type >   m_pThreadRec;   ///< Thread-local publication record
-            mutable global_lock_type    m_Mutex;    ///< Global mutex
-            mutable stat                m_Stat;     ///< Internal statistics
-            unsigned int const          m_nCompactFactor; ///< Publication list compacting factor (the list will be compacted through \p %m_nCompactFactor combining passes)
-            unsigned int const          m_nCombinePassCount; ///< Number of combining passes
-
-        public:
-            /// Initializes the object
-            /**
-                Compact factor = 64
-
-                Combiner pass count = 8
-            */
-            kernel()
-                : m_nCount(0)
-                , m_pHead( nullptr )
-                , m_pThreadRec( tls_cleanup )
-                , m_nCompactFactor( 64 - 1 ) // binary mask
-                , m_nCombinePassCount( 8 )
-            {
-                init();
-            }
-
-            /// Initializes the object
-            kernel(
-                unsigned int nCompactFactor  ///< Publication list compacting factor (the list will be compacted through \p nCompactFactor combining passes)
-                ,unsigned int nCombinePassCount ///< Number of combining passes for combiner thread
-                )
-                : m_nCount(0)
-                , m_pHead( nullptr )
-                , m_pThreadRec( tls_cleanup )
-                , m_nCompactFactor( (unsigned int)( cds::beans::ceil2( nCompactFactor ) - 1 ))   // binary mask
-                , m_nCombinePassCount( nCombinePassCount )
-            {
-                init();
-            }
-
-            /// Destroys the objects and mark all publication records as inactive
-            ~kernel()
-            {
-                // mark all publication record as detached
-                for ( publication_record * p = m_pHead; p; ) {
-                    p->pOwner = nullptr;
-
-                    publication_record * pRec = p;
-                    p = p->pNext.load( memory_model::memory_order_relaxed );
-                    if ( pRec->nState.load( memory_model::memory_order_acquire ) == removed )
-                        free_publication_record( static_cast<publication_record_type *>( pRec ));
-                }
-            }
-
-            /// Gets publication list record for the current thread
-            /**
-                If there is no publication record for the current thread
-                the function allocates it.
-            */
-            publication_record_type * acquire_record()
-            {
-                publication_record_type * pRec = m_pThreadRec.get();
-                if ( !pRec ) {
-                    // Allocate new publication record
-                    pRec = cxx11_allocator().New();
-                    pRec->pOwner = reinterpret_cast<void *>( this );
-                    m_pThreadRec.reset( pRec );
-                    m_Stat.onCreatePubRecord();
-                }
-
-                if ( pRec->nState.load( memory_model::memory_order_acquire ) != active )
-                    publish( pRec );
-
-                assert( pRec->nRequest.load( memory_model::memory_order_relaxed ) == req_EmptyRecord );
-
-                m_Stat.onAcquirePubRecord();
-                return pRec;
-            }
-
-            /// Marks publication record for the current thread as empty
-            void release_record( publication_record_type * pRec )
-            {
-                assert( pRec->is_done() );
-                pRec->nRequest.store( req_EmptyRecord, memory_model::memory_order_release );
-                m_Stat.onReleasePubRecord();
-            }
-
-            /// Trying to execute operation \p nOpId
-            /**
-                \p pRec is the publication record acquiring by \ref acquire_record earlier.
-                \p owner is a container that is owner of flat combining kernel object.
-                As a result the current thread can become a combiner or can wait for
-                another combiner performs \p pRec operation.
-
-                If the thread becomes a combiner, the kernel calls \p owner.fc_apply
-                for each active non-empty publication record.
-            */
-            template <class Container>
-            void combine( unsigned int nOpId, publication_record_type * pRec, Container& owner )
-            {
-                assert( nOpId >= req_Operation );
-                assert( pRec );
-                //assert( pRec->nState.load( memory_model::memory_order_relaxed ) == active );
-                pRec->nRequest.store( nOpId, memory_model::memory_order_release );
-
-                m_Stat.onOperation();
-
-                try_combining( owner, pRec );
-            }
-
-            /// Trying to execute operation \p nOpId in batch-combine mode
-            /**
-                \p pRec is the publication record acquiring by \ref acquire_record earlier.
-                \p owner is a container that owns flat combining kernel object.
-                As a result the current thread can become a combiner or can wait for
-                another combiner performs \p pRec operation.
-
-                If the thread becomes a combiner, the kernel calls \p owner.fc_process
-                giving the container the full access over publication list. This function
-                is useful for an elimination technique if the container supports any kind of
-                that. The container can perform multiple pass through publication list.
-
-                \p owner.fc_process has two arguments - forward iterators on begin and end of
-                publication list, see \ref iterator class. For each processed record the container
-                should call \ref operation_done function to mark the record as processed.
-
-                On the end of \p %batch_combine the \ref combine function is called
-                to process rest of publication records.
-            */
-            template <class Container>
-            void batch_combine( unsigned int nOpId, publication_record_type * pRec, Container& owner )
-            {
-                assert( nOpId >= req_Operation );
-                assert( pRec );
-                //assert( pRec->nState.load( memory_model::memory_order_relaxed ) == active );
-                pRec->nRequest.store( nOpId, memory_model::memory_order_release );
-
-                m_Stat.onOperation();
-
-                try_batch_combining( owner, pRec );
-            }
-
-            /// Waits for end of combining
-            void wait_while_combining() const
-            {
-                lock_guard l( m_Mutex );
-            }
-
-            /// Marks \p rec as executed
-            /**
-                This function should be called by container if batch_combine mode is used.
-                For usual combining (see \ref combine) this function is excess.
-            */
-            void operation_done( publication_record& rec )
-            {
-                rec.nRequest.store( req_Response, memory_model::memory_order_release );
-            }
-
-            /// Internal statistics
-            stat const& statistics() const
-            {
-                return m_Stat;
-            }
-
-            //@cond
-            // For container classes based on flat combining
-            stat& internal_statistics() const
-            {
-                return m_Stat;
-            }
-            //@endcond
-
-            /// Returns the compact factor
-            unsigned int compact_factor() const
-            {
-                return m_nCompactFactor + 1;
-            }
-
-            /// Returns number of combining passes for combiner thread
-            unsigned int combine_pass_count() const
-            {
-                return m_nCombinePassCount;
-            }
-
-        public:
-            /// Publication list iterator
-            /**
-                Iterators are intended for batch processing by container's
-                \p fc_process function.
-                The iterator allows iterate through active publication list.
-            */
-            class iterator
-            {
-                //@cond
-                friend class kernel;
-                publication_record_type * m_pRec;
-                //@endcond
-
-            protected:
-                //@cond
-                iterator( publication_record_type * pRec )
-                    : m_pRec( pRec )
-                {
-                    skip_inactive();
-                }
-
-                void skip_inactive()
-                {
-                    while ( m_pRec && (m_pRec->nState.load( memory_model::memory_order_acquire ) != active
-                                    || m_pRec->nRequest.load( memory_model::memory_order_relaxed) < req_Operation ))
-                    {
-                        m_pRec = static_cast<publication_record_type *>(m_pRec->pNext.load( memory_model::memory_order_acquire ));
-                    }
-                }
-                //@endcond
-
-            public:
-                /// Initializes an empty iterator object
-                iterator()
-                    : m_pRec( nullptr )
-                {}
-
-                /// Copy ctor
-                iterator( iterator const& src )
-                    : m_pRec( src.m_pRec )
-                {}
-
-                /// Pre-increment
-                iterator& operator++()
-                {
-                    assert( m_pRec );
-                    m_pRec = static_cast<publication_record_type *>( m_pRec->pNext.load( memory_model::memory_order_acquire ));
-                    skip_inactive();
-                    return *this;
-                }
-
-                /// Post-increment
-                iterator operator++(int)
-                {
-                    assert( m_pRec );
-                    iterator it(*this);
-                    ++(*this);
-                    return it;
-                }
-
-                /// Dereference operator, can return \p nullptr
-                publication_record_type * operator ->()
-                {
-                    return m_pRec;
-                }
-
-                /// Dereference operator, the iterator should not be an end iterator
-                publication_record_type& operator*()
-                {
-                    assert( m_pRec );
-                    return *m_pRec;
-                }
-
-                /// Iterator equality
-                friend bool operator==( iterator it1, iterator it2 )
-                {
-                    return it1.m_pRec == it2.m_pRec;
-                }
-
-                /// Iterator inequality
-                friend bool operator!=( iterator it1, iterator it2 )
-                {
-                    return !( it1 == it2 );
-                }
-            };
-
-            /// Returns an iterator to the first active publication record
-            iterator begin()    { return iterator(m_pHead); }
-
-            /// Returns an iterator to the end of publication list. Should not be dereferenced.
-            iterator end()      { return iterator(); }
-
-        private:
-            //@cond
-            static void tls_cleanup( publication_record_type * pRec )
-            {
-                // Thread done
-                // pRec that is TLS data should be excluded from publication list
-                if ( pRec ) {
-                    if ( pRec->pOwner ) {
-                        // kernel is alive
-                        pRec->nState.store( removed, memory_model::memory_order_release );
-                    }
-                    else {
-                        // kernel already deleted
-                        free_publication_record( pRec );
-                    }
-                }
-            }
-
-            static void free_publication_record( publication_record_type * pRec )
-            {
-                cxx11_allocator().Delete( pRec );
-            }
-
-            void init()
-            {
-                assert( m_pThreadRec.get() == nullptr );
-                publication_record_type * pRec = cxx11_allocator().New();
-                m_pHead = pRec;
-                pRec->pOwner = this;
-                m_pThreadRec.reset( pRec );
-                m_Stat.onCreatePubRecord();
-            }
-
-            void publish( publication_record_type * pRec )
-            {
-                assert( pRec->nState.load( memory_model::memory_order_relaxed ) == inactive );
-
-                pRec->nAge.store( m_nCount.load(memory_model::memory_order_acquire), memory_model::memory_order_release );
-                pRec->nState.store( active, memory_model::memory_order_release );
-
-                // Insert record to publication list
-                if ( m_pHead != static_cast<publication_record *>(pRec) ) {
-                    publication_record * p = m_pHead->pNext.load(memory_model::memory_order_relaxed);
-                    if ( p != static_cast<publication_record *>( pRec )) {
-                        do {
-                            pRec->pNext = p;
-                            // Failed CAS changes p
-                        } while ( !m_pHead->pNext.compare_exchange_weak( p, static_cast<publication_record *>(pRec),
-                            memory_model::memory_order_release, atomics::memory_order_relaxed ));
-                        m_Stat.onActivatePubRecord();
-                    }
-                }
-            }
-
-            void republish( publication_record_type * pRec )
-            {
-                if ( pRec->nState.load( memory_model::memory_order_relaxed ) != active ) {
-                    // The record has been excluded from publication list. Reinsert it
-                    publish( pRec );
-                }
-            }
-
-            template <class Container>
-            void try_combining( Container& owner, publication_record_type * pRec )
-            {
-                if ( m_Mutex.try_lock() ) {
-                    // The thread becomes a combiner
-                    lock_guard l( m_Mutex, std::adopt_lock_t() );
-
-                    // The record pRec can be excluded from publication list. Re-publish it
-                    republish( pRec );
-
-                    combining( owner );
-                    assert( pRec->nRequest.load( memory_model::memory_order_relaxed ) == req_Response );
-                }
-                else {
-                    // There is another combiner, wait while it executes our request
-                    if ( !wait_for_combining( pRec ) ) {
-                        // The thread becomes a combiner
-                        lock_guard l( m_Mutex, std::adopt_lock_t() );
-
-                        // The record pRec can be excluded from publication list. Re-publish it
-                        republish( pRec );
-
-                        combining( owner );
-                        assert( pRec->nRequest.load( memory_model::memory_order_relaxed ) == req_Response );
-                    }
-                }
-            }
-
-            template <class Container>
-            void try_batch_combining( Container& owner, publication_record_type * pRec )
-            {
-                if ( m_Mutex.try_lock() ) {
-                    // The thread becomes a combiner
-                    lock_guard l( m_Mutex, std::adopt_lock_t() );
-
-                    // The record pRec can be excluded from publication list. Re-publish it
-                    republish( pRec );
-
-                    batch_combining( owner );
-                    assert( pRec->nRequest.load( memory_model::memory_order_relaxed ) == req_Response );
-                }
-                else {
-                    // There is another combiner, wait while it executes our request
-                    if ( !wait_for_combining( pRec ) ) {
-                        // The thread becomes a combiner
-                        lock_guard l( m_Mutex, std::adopt_lock_t() );
-
-                        // The record pRec can be excluded from publication list. Re-publish it
-                        republish( pRec );
-
-                        batch_combining( owner );
-                        assert( pRec->nRequest.load( memory_model::memory_order_relaxed ) == req_Response );
-                    }
-                }
-            }
-
-            template <class Container>
-            void combining( Container& owner )
-            {
-                // The thread is a combiner
-                assert( !m_Mutex.try_lock() );
-
-                unsigned int const nCurAge = m_nCount.fetch_add( 1, memory_model::memory_order_release ) + 1;
-
-                for ( unsigned int nPass = 0; nPass < m_nCombinePassCount; ++nPass )
-                    if ( !combining_pass( owner, nCurAge ))
-                        break;
-
-                m_Stat.onCombining();
-                if ( (nCurAge & m_nCompactFactor) == 0 )
-                    compact_list( nCurAge );
-            }
-
-            template <class Container>
-            bool combining_pass( Container& owner, unsigned int nCurAge )
-            {
-                publication_record * pPrev = nullptr;
-                publication_record * p = m_pHead;
-                bool bOpDone = false;
-                while ( p ) {
-                    switch ( p->nState.load( memory_model::memory_order_acquire )) {
-                        case active:
-                            if ( p->op() >= req_Operation ) {
-                                p->nAge.store( nCurAge, memory_model::memory_order_release );
-                                owner.fc_apply( static_cast<publication_record_type *>(p) );
-                                operation_done( *p );
-                                bOpDone = true;
-                            }
-                            break;
-                        case inactive:
-                            // Only m_pHead can be inactive in the publication list
-                            assert( p == m_pHead );
-                            break;
-                        case removed:
-                            // The record should be removed
-                            p = unlink_and_delete_record( pPrev, p );
-                            continue;
-                        default:
-                            /// ??? That is impossible
-                            assert(false);
-                    }
-                    pPrev = p;
-                    p = p->pNext.load( memory_model::memory_order_acquire );
-                }
-                return bOpDone;
-            }
-
-            template <class Container>
-            void batch_combining( Container& owner )
-            {
-                // The thread is a combiner
-                assert( !m_Mutex.try_lock() );
-
-                unsigned int const nCurAge = m_nCount.fetch_add( 1, memory_model::memory_order_release ) + 1;
-
-                for ( unsigned int nPass = 0; nPass < m_nCombinePassCount; ++nPass )
-                    owner.fc_process( begin(), end() );
-
-                combining_pass( owner, nCurAge );
-                m_Stat.onCombining();
-                if ( (nCurAge & m_nCompactFactor) == 0 )
-                    compact_list( nCurAge );
-            }
-
-            bool wait_for_combining( publication_record_type * pRec )
-            {
-                back_off bkoff;
-                while ( pRec->nRequest.load( memory_model::memory_order_acquire ) != req_Response ) {
-
-                    // The record can be excluded from publication list. Reinsert it
-                    republish( pRec );
-
-                    bkoff();
-
-                    if ( m_Mutex.try_lock() ) {
-                        if ( pRec->nRequest.load( memory_model::memory_order_acquire ) == req_Response ) {
-                            m_Mutex.unlock();
-                            break;
-                        }
-                        // The thread becomes a combiner
-                        return false;
-                    }
-                }
-                return true;
-            }
-
-            void compact_list( unsigned int const nCurAge )
-            {
-                // Thinning publication list
-                publication_record * pPrev = nullptr;
-                for ( publication_record * p = m_pHead; p; ) {
-                    if ( p->nState.load( memory_model::memory_order_acquire ) == active
-                      && p->nAge.load( memory_model::memory_order_acquire ) + m_nCompactFactor < nCurAge )
-                    {
-                        if ( pPrev ) {
-                            publication_record * pNext = p->pNext.load( memory_model::memory_order_acquire );
-                            if ( pPrev->pNext.compare_exchange_strong( p, pNext,
-                                memory_model::memory_order_release, atomics::memory_order_relaxed ))
-                            {
-                                p->nState.store( inactive, memory_model::memory_order_release );
-                                p = pNext;
-                                m_Stat.onDeactivatePubRecord();
-                                continue;
-                            }
-                        }
-                    }
-                    pPrev = p;
-                    p = p->pNext.load( memory_model::memory_order_acquire );
-                }
-
-                m_Stat.onCompactPublicationList();
-            }
-
-            publication_record * unlink_and_delete_record( publication_record * pPrev, publication_record * p )
-            {
-                if ( pPrev ) {
-                    publication_record * pNext = p->pNext.load( memory_model::memory_order_acquire );
-                    if ( pPrev->pNext.compare_exchange_strong( p, pNext,
-                        memory_model::memory_order_release, atomics::memory_order_relaxed ))
-                    {
-                        free_publication_record( static_cast<publication_record_type *>( p ));
-                        m_Stat.onDeletePubRecord();
-                    }
-                    return pNext;
-                }
-                else {
-                    m_pHead = static_cast<publication_record_type *>( p->pNext.load( memory_model::memory_order_acquire ));
-                    free_publication_record( static_cast<publication_record_type *>( p ));
-                    m_Stat.onDeletePubRecord();
-                    return m_pHead;
-                }
-            }
-            //@endcond
-        };
-
-        //@cond
-        class container
-        {
-        public:
-            template <typename PubRecord>
-            void fc_apply( PubRecord * )
-            {
-                assert( false );
-            }
-
-            template <typename Iterator>
-            void fc_process( Iterator, Iterator )
-            {
-                assert( false );
-            }
-        };
-        //@endcond
-
-    } // namespace flat_combining
-}} // namespace cds::algo
+#include <cds/algo/flat_combining/kernel.h>
 
 #endif // #ifndef CDSLIB_ALGO_FLAT_COMBINING_H
diff --git a/cds/algo/flat_combining/defs.h b/cds/algo/flat_combining/defs.h
new file mode 100644 (file)
index 0000000..d29702d
--- /dev/null
@@ -0,0 +1,91 @@
+/*
+    This file is a part of libcds - Concurrent Data Structures library
+
+    (C) Copyright Maxim Khizhinsky (libcds.dev@gmail.com) 2006-2016
+
+    Source code repo: http://github.com/khizmax/libcds/
+    Download: http://sourceforge.net/projects/libcds/files/
+    
+    Redistribution and use in source and binary forms, with or without
+    modification, are permitted provided that the following conditions are met:
+
+    * Redistributions of source code must retain the above copyright notice, this
+      list of conditions and the following disclaimer.
+
+    * 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 THE COPYRIGHT HOLDERS AND CONTRIBUTORS "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 THE COPYRIGHT HOLDER 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.
+*/
+
+#ifndef CDSLIB_ALGO_FLAT_COMBINING_DEFS_H
+#define CDSLIB_ALGO_FLAT_COMBINING_DEFS_H
+
+#include <cds/algo/atomic.h>
+
+
+namespace cds { namespace algo { namespace flat_combining {
+
+    /// Special values of \p publication_record::nRequest
+    enum request_value
+    {
+        req_EmptyRecord,    ///< Publication record is empty
+        req_Response,       ///< Operation is done
+
+        req_Operation       ///< First operation id for derived classes
+    };
+
+    /// \p publication_record state
+    enum record_state {
+        inactive,       ///< Record is inactive
+        active,         ///< Record is active
+        removed         ///< Record should be removed
+    };
+
+    /// Record of publication list
+    /**
+        Each data structure based on flat combining contains a class derived from \p %publication_record
+    */
+    struct publication_record {
+        atomics::atomic<unsigned int>           nRequest;   ///< Request field (depends on data structure)
+        atomics::atomic<unsigned int>           nState;     ///< Record state: inactive, active, removed
+        atomics::atomic<unsigned int>           nAge;       ///< Age of the record
+        atomics::atomic<publication_record *>   pNext;      ///< Next record in publication list
+        void *                                  pOwner;     ///< [internal data] Pointer to \ref kernel object that manages the publication list
+
+        /// Initializes publication record
+        publication_record()
+            : nRequest( req_EmptyRecord )
+            , nState( inactive )
+            , nAge( 0 )
+            , pNext( nullptr )
+            , pOwner( nullptr )
+        {}
+
+        /// Returns the value of \p nRequest field
+        unsigned int op( atomics::memory_order mo = atomics::memory_order_relaxed ) const
+        {
+            return nRequest.load( mo );
+        }
+
+        /// Checks if the operation is done
+        bool is_done() const
+        {
+            return nRequest.load( atomics::memory_order_relaxed ) == req_Response;
+        }
+    };
+
+
+}}} // namespace cds::algo::flat_combining
+
+#endif // CDSLIB_ALGO_FLAT_COMBINING_DEFS_H
diff --git a/cds/algo/flat_combining/kernel.h b/cds/algo/flat_combining/kernel.h
new file mode 100644 (file)
index 0000000..79bf5e7
--- /dev/null
@@ -0,0 +1,858 @@
+/*
+    This file is a part of libcds - Concurrent Data Structures library
+
+    (C) Copyright Maxim Khizhinsky (libcds.dev@gmail.com) 2006-2016
+
+    Source code repo: http://github.com/khizmax/libcds/
+    Download: http://sourceforge.net/projects/libcds/files/
+    
+    Redistribution and use in source and binary forms, with or without
+    modification, are permitted provided that the following conditions are met:
+
+    * Redistributions of source code must retain the above copyright notice, this
+      list of conditions and the following disclaimer.
+
+    * 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 THE COPYRIGHT HOLDERS AND CONTRIBUTORS "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 THE COPYRIGHT HOLDER 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.     
+*/
+
+#ifndef CDSLIB_ALGO_FLAT_COMBINING_KERNEL_H
+#define CDSLIB_ALGO_FLAT_COMBINING_KERNEL_H
+
+#include <cds/algo/flat_combining/defs.h>
+#include <cds/algo/flat_combining/wait_strategy.h>
+
+#include <cds/sync/spinlock.h>
+#include <cds/details/allocator.h>
+#include <cds/opt/options.h>
+#include <cds/algo/int_algo.h>
+
+namespace cds { namespace algo {
+
+    /// @defgroup cds_flat_combining_intrusive Intrusive flat combining containers
+    /// @defgroup cds_flat_combining_container Non-intrusive flat combining containers
+
+    /// Flat combining
+    /**
+        @anchor cds_flat_combining_description
+        Flat combining (FC) technique is invented by Hendler, Incze, Shavit and Tzafrir in their paper
+        [2010] <i>"Flat Combining and the Synchronization-Parallelism Tradeoff"</i>.
+        The technique converts a sequential data structure to its concurrent implementation.
+        A few structures are added to the sequential implementation: a <i>global lock</i>,
+        a <i>count</i> of the number of combining passes, and a pointer to the <i>head</i>
+        of a <i>publication list</i>. The publication list is a list of thread-local records
+        of a size proportional to the number of threads that are concurrently accessing the shared object.
+
+        Each thread \p t accessing the structure to perform an invocation of some method \p m
+        on the shared object executes the following sequence of steps:
+        <ol>
+        <li>Write the invocation opcode and parameters (if any) of the method \p m to be applied
+        sequentially to the shared object in the <i>request</i> field of your thread local publication
+        record (there is no need to use a load-store memory barrier). The <i>request</i> field will later
+        be used to receive the response. If your thread local publication record is marked as active
+        continue to step 2, otherwise continue to step 5.</li>
+        <li>Check if the global lock is taken. If so (another thread is an active combiner), spin on the <i>request</i>
+        field waiting for a response to the invocation (one can add a yield at this point to allow other threads
+        on the same core to run). Once in a while while spinning check if the lock is still taken and that your
+        record is active. If your record is inactive proceed to step 5. Once the response is available,
+        reset the request field to null and return the response.</li>
+        <li>If the lock is not taken, attempt to acquire it and become a combiner. If you fail,
+        return to spinning in step 2.</li>
+        <li>Otherwise, you hold the lock and are a combiner.
+        <ul>
+            <li>Increment the combining pass count by one.</li>
+            <li>Execute a \p fc_apply() by traversing the publication list from the head,
+            combining all nonnull method call invocations, setting the <i>age</i> of each of these records
+            to the current <i>count</i>, applying the combined method calls to the structure D, and returning
+            responses to all the invocations. This traversal is guaranteed to be wait-free.</li>
+            <li>If the <i>count</i> is such that a cleanup needs to be performed, traverse the publication
+            list from the <i>head</i>. Starting from the second item (we always leave the item pointed to
+            by the head in the list), remove from the publication list all records whose <i>age</i> is
+            much smaller than the current <i>count</i>. This is done by removing the node and marking it
+            as inactive.</li>
+            <li>Release the lock.</li>
+        </ul>
+        <li>If you have no thread local publication record allocate one, marked as active. If you already
+        have one marked as inactive, mark it as active. Execute a store-load memory barrier. Proceed to insert
+        the record into the list with a successful CAS to the <i>head</i>. Then proceed to step 1.</li>
+        </ol>
+
+        As the test results show, the flat combining technique is suitable for non-intrusive containers
+        like stack, queue, deque. For intrusive concurrent containers the flat combining demonstrates
+        less impressive results.
+
+        \ref cds_flat_combining_container "List of FC-based containers" in libcds.
+
+        \ref cds_flat_combining_intrusive "List of intrusive FC-based containers" in libcds.
+    */
+    namespace flat_combining {
+
+        /// Flat combining internal statistics
+        template <typename Counter = cds::atomicity::event_counter >
+        struct stat
+        {
+            typedef Counter counter_type;   ///< Event counter type
+
+            counter_type    m_nOperationCount   ;   ///< How many operations have been performed
+            counter_type    m_nCombiningCount   ;   ///< Combining call count
+            counter_type    m_nCompactPublicationList; ///< Count of publication list compacting
+            counter_type    m_nDeactivatePubRecord; ///< How many publication records were deactivated during compacting
+            counter_type    m_nActivatePubRecord;   ///< Count of publication record activating
+            counter_type    m_nPubRecordCreated ;   ///< Count of created publication records
+            counter_type    m_nPubRecordDeteted ;   ///< Count of deleted publication records
+            counter_type    m_nPassiveWaitCall;     ///< Count of passive waiting call (\p kernel::wait_for_combining())
+            counter_type    m_nPassiveWaitIteration;///< Count of iteration inside passive waiting
+            counter_type    m_nPassiveWaitWakeup;   ///< Count of forcing wake-up of passive wait cycle
+            counter_type    m_nInvokeExclusive;     ///< Count of call \p kernel::invoke_exclusive()
+            counter_type    m_nWakeupByNotifying;   ///< How many times the passive thread be waked up by a notification
+            counter_type    m_nPassiveToCombiner;   ///< How many times the passive thread becomes the combiner
+
+            /// Returns current combining factor
+            /**
+                Combining factor is how many operations perform in one combine pass:
+                <tt>combining_factor := m_nOperationCount / m_nCombiningCount</tt>
+            */
+            double combining_factor() const
+            {
+                return m_nCombiningCount.get() ? double( m_nOperationCount.get()) / m_nCombiningCount.get() : 0.0;
+            }
+
+            //@cond
+            void    onOperation()               { ++m_nOperationCount;          }
+            void    onCombining()               { ++m_nCombiningCount;          }
+            void    onCompactPublicationList()  { ++m_nCompactPublicationList;  }
+            void    onDeactivatePubRecord()     { ++m_nDeactivatePubRecord;     }
+            void    onActivatePubRecord()       { ++m_nActivatePubRecord;       }
+            void    onCreatePubRecord()         { ++m_nPubRecordCreated;        }
+            void    onDeletePubRecord()         { ++m_nPubRecordDeteted;        }
+            void    onPassiveWait()             { ++m_nPassiveWaitCall;         }
+            void    onPassiveWaitIteration()    { ++m_nPassiveWaitIteration;    }
+            void    onPassiveWaitWakeup()       { ++m_nPassiveWaitWakeup;       }
+            void    onInvokeExclusive()         { ++m_nInvokeExclusive;         }
+            void    onWakeupByNotifying()       { ++m_nWakeupByNotifying;       }
+            void    onPassiveToCombiner()       { ++m_nPassiveToCombiner;       }
+            
+            //@endcond
+        };
+
+        /// Flat combining dummy internal statistics
+        struct empty_stat
+        {
+            //@cond
+            void    onOperation()               const {}
+            void    onCombining()               const {}
+            void    onCompactPublicationList()  const {}
+            void    onDeactivatePubRecord()     const {}
+            void    onActivatePubRecord()       const {}
+            void    onCreatePubRecord()         const {}
+            void    onDeletePubRecord()         const {}
+            void    onPassiveWait()             const {}
+            void    onPassiveWaitIteration()    const {}
+            void    onPassiveWaitWakeup()       const {}
+            void    onInvokeExclusive()         const {}
+            void    onWakeupByNotifying()       const {}
+            void    onPassiveToCombiner()       const {}
+            //@endcond
+        };
+
+        /// Type traits of \ref kernel class
+        /**
+            You can define different type traits for \ref kernel
+            by specifying your struct based on \p %traits
+            or by using \ref make_traits metafunction.
+        */
+        struct traits
+        {
+            typedef cds::sync::spin             lock_type;  ///< Lock type
+            typedef cds::algo::flat_combining::wait_strategy::backoff< cds::backoff::delay_of<2>> wait_strategy; ///< Wait strategy
+            typedef CDS_DEFAULT_ALLOCATOR       allocator;  ///< Allocator used for TLS data (allocating \p publication_record derivatives)
+            typedef empty_stat                  stat;       ///< Internal statistics
+            typedef opt::v::relaxed_ordering  memory_model; ///< /// C++ memory ordering model
+        };
+
+        /// Metafunction converting option list to traits
+        /**
+            \p Options are:
+            - \p opt::lock_type - mutex type, default is \p cds::sync::spin
+            - \p opt::wait_strategy - wait strategy, see \p wait_strategy namespace, default is \p wait_strategy::backoff.
+            - \p opt::allocator - allocator type, default is \ref CDS_DEFAULT_ALLOCATOR
+            - \p opt::stat - internal statistics, possible type: \ref stat, \ref empty_stat (the default)
+            - \p opt::memory_model - C++ memory ordering model.
+                List of all available memory ordering see \p opt::memory_model.
+                Default is \p cds::opt::v::relaxed_ordering
+        */
+        template <typename... Options>
+        struct make_traits {
+#   ifdef CDS_DOXYGEN_INVOKED
+            typedef implementation_defined type ;   ///< Metafunction result
+#   else
+            typedef typename cds::opt::make_options<
+                typename cds::opt::find_type_traits< traits, Options... >::type
+                ,Options...
+            >::type   type;
+#   endif
+        };
+
+        /// The kernel of flat combining
+        /**
+            Template parameters:
+            - \p PublicationRecord - a type derived from \ref publication_record
+            - \p Traits - a type traits of flat combining, default is \p flat_combining::traits.
+                \ref make_traits metafunction can be used to create type traits
+
+            The kernel object should be a member of a container class. The container cooperates with flat combining
+            kernel object. There are two ways to interact with the kernel:
+            - One-by-one processing the active records of the publication list. This mode provides by \p combine() function:
+              the container acquires its publication record by \p acquire_record(), fills its fields and calls
+              \p combine() function of its kernel object. If the current thread becomes a combiner, the kernel
+              calls \p fc_apply() function of the container for each active non-empty record. Then, the container
+              should release its publication record by \p release_record(). Only one pass through the publication
+              list is possible.
+            - Batch processing - \p batch_combine() function. It this mode the container obtains access
+              to entire publication list. This mode allows the container to perform an elimination, for example,
+              the stack can collide \p push() and \p pop() requests. The sequence of invocations is the following:
+              the container acquires its publication record by \p acquire_record(), fills its field and call
+              \p batch_combine() function of its kernel object. If the current thread becomes a combiner,
+              the kernel calls \p fc_process() function of the container passing two iterators pointing to
+              the begin and the end of publication list (see \ref iterator class). The iterators allow
+              multiple pass through active records of publication list. For each processed record the container
+              should call \p operation_done() function. On the end, the container should release
+              its record by \p release_record().
+        */
+        template <
+            typename PublicationRecord
+            ,typename Traits = traits
+        >
+        class kernel
+        {
+        public:
+            typedef Traits   traits;                               ///< Type traits
+            typedef typename traits::lock_type global_lock_type;   ///< Global lock type
+            typedef typename traits::wait_strategy wait_strategy;  ///< Wait strategy type
+            typedef typename traits::allocator allocator;          ///< Allocator type (used for allocating publication_record_type data)
+            typedef typename traits::stat      stat;               ///< Internal statistics
+            typedef typename traits::memory_model memory_model;    ///< C++ memory model
+
+            typedef typename wait_strategy::template make_publication_record<PublicationRecord>::type publication_record_type; ///< Publication record type
+
+        protected:
+            //@cond
+            typedef cds::details::Allocator< publication_record_type, allocator >   cxx11_allocator; ///< internal helper cds::details::Allocator
+            typedef std::lock_guard<global_lock_type> lock_guard;
+            //@endcond
+
+        protected:
+            atomics::atomic<unsigned int>  m_nCount;   ///< Total count of combining passes. Used as an age.
+            publication_record_type *   m_pHead;    ///< Head of publication list
+            boost::thread_specific_ptr< publication_record_type > m_pThreadRec;   ///< Thread-local publication record
+            mutable global_lock_type    m_Mutex;    ///< Global mutex
+            mutable stat                m_Stat;     ///< Internal statistics
+            unsigned int const          m_nCompactFactor;    ///< Publication list compacting factor (the list will be compacted through \p %m_nCompactFactor combining passes)
+            unsigned int const          m_nCombinePassCount; ///< Number of combining passes
+            wait_strategy               m_waitStrategy;      ///< Wait strategy
+
+        public:
+            /// Initializes the object
+            /**
+                Compact factor = 64
+
+                Combiner pass count = 8
+            */
+            kernel()
+                : kernel( 64, 8 )
+            {}
+
+            /// Initializes the object
+            kernel(
+                unsigned int nCompactFactor  ///< Publication list compacting factor (the list will be compacted through \p nCompactFactor combining passes)
+                ,unsigned int nCombinePassCount ///< Number of combining passes for combiner thread
+                )
+                : m_nCount(0)
+                , m_pHead( nullptr )
+                , m_pThreadRec( tls_cleanup )
+                , m_nCompactFactor( (unsigned int)( cds::beans::ceil2( nCompactFactor ) - 1 ))   // binary mask
+                , m_nCombinePassCount( nCombinePassCount )
+            {
+                init();
+            }
+
+            /// Destroys the objects and mark all publication records as inactive
+            ~kernel()
+            {
+                // mark all publication record as detached
+                for ( publication_record* p = m_pHead; p; ) {
+                    p->pOwner = nullptr;
+
+                    publication_record * pRec = p;
+                    p = p->pNext.load( memory_model::memory_order_relaxed );
+                    if ( pRec->nState.load( memory_model::memory_order_acquire ) == removed )
+                        free_publication_record( static_cast<publication_record_type *>( pRec ));
+                }
+            }
+
+            /// Gets publication list record for the current thread
+            /**
+                If there is no publication record for the current thread
+                the function allocates it.
+            */
+            publication_record_type * acquire_record()
+            {
+                publication_record_type * pRec = m_pThreadRec.get();
+                if ( !pRec ) {
+                    // Allocate new publication record
+                    pRec = cxx11_allocator().New();
+                    pRec->pOwner = reinterpret_cast<void *>( this );
+                    m_pThreadRec.reset( pRec );
+                    m_Stat.onCreatePubRecord();
+                }
+
+                if ( pRec->nState.load( memory_model::memory_order_acquire ) != active )
+                    publish( pRec );
+
+                assert( pRec->op() == req_EmptyRecord );
+
+                return pRec;
+            }
+
+            /// Marks publication record for the current thread as empty
+            void release_record( publication_record_type * pRec )
+            {
+                assert( pRec->is_done() );
+                pRec->nRequest.store( req_EmptyRecord, memory_model::memory_order_release );
+            }
+
+            /// Trying to execute operation \p nOpId
+            /**
+                \p pRec is the publication record acquiring by \ref acquire_record earlier.
+                \p owner is a container that is owner of flat combining kernel object.
+                As a result the current thread can become a combiner or can wait for
+                another combiner performs \p pRec operation.
+
+                If the thread becomes a combiner, the kernel calls \p owner.fc_apply
+                for each active non-empty publication record.
+            */
+            template <class Container>
+            void combine( unsigned int nOpId, publication_record_type * pRec, Container& owner )
+            {
+                assert( nOpId >= req_Operation );
+                assert( pRec );
+
+                pRec->nRequest.store( nOpId, memory_model::memory_order_release );
+                m_Stat.onOperation();
+
+                try_combining( owner, pRec );
+            }
+
+            /// Trying to execute operation \p nOpId in batch-combine mode
+            /**
+                \p pRec is the publication record acquiring by \p acquire_record() earlier.
+                \p owner is a container that owns flat combining kernel object.
+                As a result the current thread can become a combiner or can wait for
+                another combiner performs \p pRec operation.
+
+                If the thread becomes a combiner, the kernel calls \p owner.fc_process()
+                giving the container the full access over publication list. This function
+                is useful for an elimination technique if the container supports any kind of
+                that. The container can perform multiple pass through publication list.
+
+                \p owner.fc_process() has two arguments - forward iterators on begin and end of
+                publication list, see \ref iterator class. For each processed record the container
+                should call \p operation_done() function to mark the record as processed.
+
+                On the end of \p %batch_combine the \p combine() function is called
+                to process rest of publication records.
+            */
+            template <class Container>
+            void batch_combine( unsigned int nOpId, publication_record_type* pRec, Container& owner )
+            {
+                assert( nOpId >= req_Operation );
+                assert( pRec );
+
+                pRec->nRequest.store( nOpId, memory_model::memory_order_release );
+                m_Stat.onOperation();
+
+                try_batch_combining( owner, pRec );
+            }
+
+            /// Invokes \p Func in exclusive mode
+            /**
+                Some operation in flat combining containers should be called in exclusive mode
+                i.e the current thread should become the combiner to process the operation.
+                The typical example is \p empty() function.
+                
+                \p %invoke_exclusive() allows do that: the current thread becomes the combiner,
+                invokes \p f exclusively but unlike a typical usage the thread does not process any pending request.
+                Instead, after end of \p f call the current thread wakes up a pending thread if any.
+            */
+            template <typename Func>
+            void invoke_exclusive( Func f )
+            {
+                {
+                    lock_guard l( m_Mutex );
+                    f();
+                }
+                m_waitStrategy.wakeup( *this );
+                m_Stat.onInvokeExclusive();
+            }
+
+            /// Marks \p rec as executed
+            /**
+                This function should be called by container if \p batch_combine mode is used.
+                For usual combining (see \p combine() ) this function is excess.
+            */
+            void operation_done( publication_record& rec )
+            {
+                rec.nRequest.store( req_Response, memory_model::memory_order_release );
+                m_waitStrategy.notify( *this, static_cast<publication_record_type&>( rec ));
+            }
+
+            /// Internal statistics
+            stat const& statistics() const
+            {
+                return m_Stat;
+            }
+
+            //@cond
+            // For container classes based on flat combining
+            stat& internal_statistics() const
+            {
+                return m_Stat;
+            }
+            //@endcond
+
+            /// Returns the compact factor
+            unsigned int compact_factor() const
+            {
+                return m_nCompactFactor + 1;
+            }
+
+            /// Returns number of combining passes for combiner thread
+            unsigned int combine_pass_count() const
+            {
+                return m_nCombinePassCount;
+            }
+
+        public:
+            /// Publication list iterator
+            /**
+                Iterators are intended for batch processing by container's
+                \p fc_process function.
+                The iterator allows iterate through active publication list.
+            */
+            class iterator
+            {
+                //@cond
+                friend class kernel;
+                publication_record_type * m_pRec;
+                //@endcond
+
+            protected:
+                //@cond
+                iterator( publication_record_type * pRec )
+                    : m_pRec( pRec )
+                {
+                    skip_inactive();
+                }
+
+                void skip_inactive()
+                {
+                    while ( m_pRec && (m_pRec->nState.load( memory_model::memory_order_acquire ) != active
+                                    || m_pRec->op( memory_model::memory_order_relaxed) < req_Operation ))
+                    {
+                        m_pRec = static_cast<publication_record_type*>(m_pRec->pNext.load( memory_model::memory_order_acquire ));
+                    }
+                }
+                //@endcond
+
+            public:
+                /// Initializes an empty iterator object
+                iterator()
+                    : m_pRec( nullptr )
+                {}
+
+                /// Copy ctor
+                iterator( iterator const& src )
+                    : m_pRec( src.m_pRec )
+                {}
+
+                /// Pre-increment
+                iterator& operator++()
+                {
+                    assert( m_pRec );
+                    m_pRec = static_cast<publication_record_type *>( m_pRec->pNext.load( memory_model::memory_order_acquire ));
+                    skip_inactive();
+                    return *this;
+                }
+
+                /// Post-increment
+                iterator operator++(int)
+                {
+                    assert( m_pRec );
+                    iterator it(*this);
+                    ++(*this);
+                    return it;
+                }
+
+                /// Dereference operator, can return \p nullptr
+                publication_record_type* operator ->()
+                {
+                    return m_pRec;
+                }
+
+                /// Dereference operator, the iterator should not be an end iterator
+                publication_record_type& operator*()
+                {
+                    assert( m_pRec );
+                    return *m_pRec;
+                }
+
+                /// Iterator equality
+                friend bool operator==( iterator it1, iterator it2 )
+                {
+                    return it1.m_pRec == it2.m_pRec;
+                }
+
+                /// Iterator inequality
+                friend bool operator!=( iterator it1, iterator it2 )
+                {
+                    return !( it1 == it2 );
+                }
+            };
+
+            /// Returns an iterator to the first active publication record
+            iterator begin()    { return iterator(m_pHead); }
+
+            /// Returns an iterator to the end of publication list. Should not be dereferenced.
+            iterator end()      { return iterator(); }
+
+        public:
+            /// Gets current value of \p rec.nRequest
+            /**
+                This function is intended for invoking from a wait strategy
+            */
+            int get_operation( publication_record& rec )
+            {
+                return rec.op( memory_model::memory_order_acquire );
+            }
+
+            /// Wakes up any waiting thread
+            /**
+                This function is intended for invoking from a wait strategy
+            */
+            void wakeup_any()
+            {
+                publication_record* pRec = m_pHead;
+                while ( pRec ) {
+                    if ( pRec->nState.load( memory_model::memory_order_acquire ) == active
+                      && pRec->op( memory_model::memory_order_acquire ) >= req_Operation )
+                    {
+                        m_waitStrategy.notify( *this, static_cast<publication_record_type&>( *pRec ));
+                        break;
+                    }
+                    pRec = pRec->pNext.load( memory_model::memory_order_acquire );
+                }
+            }
+
+        private:
+            //@cond
+            static void tls_cleanup( publication_record_type* pRec )
+            {
+                // Thread done
+                // pRec that is TLS data should be excluded from publication list
+                if ( pRec ) {
+                    if ( pRec->pOwner ) {
+                        // kernel is alive
+                        pRec->nState.store( removed, memory_model::memory_order_release );
+                    }
+                    else {
+                        // kernel already deleted
+                        free_publication_record( pRec );
+                    }
+                }
+            }
+
+            static void free_publication_record( publication_record_type* pRec )
+            {
+                cxx11_allocator().Delete( pRec );
+            }
+
+            void init()
+            {
+                assert( m_pThreadRec.get() == nullptr );
+                publication_record_type* pRec = cxx11_allocator().New();
+                m_pHead = pRec;
+                pRec->pOwner = this;
+                m_pThreadRec.reset( pRec );
+                m_Stat.onCreatePubRecord();
+            }
+
+            void publish( publication_record_type* pRec )
+            {
+                assert( pRec->nState.load( memory_model::memory_order_relaxed ) == inactive );
+
+                pRec->nAge.store( m_nCount.load(memory_model::memory_order_acquire), memory_model::memory_order_release );
+                pRec->nState.store( active, memory_model::memory_order_release );
+
+                // Insert record to publication list
+                if ( m_pHead != static_cast<publication_record *>(pRec) ) {
+                    publication_record * p = m_pHead->pNext.load(memory_model::memory_order_relaxed);
+                    if ( p != static_cast<publication_record *>( pRec )) {
+                        do {
+                            pRec->pNext = p;
+                            // Failed CAS changes p
+                        } while ( !m_pHead->pNext.compare_exchange_weak( p, static_cast<publication_record *>(pRec),
+                            memory_model::memory_order_release, atomics::memory_order_relaxed ));
+                        m_Stat.onActivatePubRecord();
+                    }
+                }
+            }
+
+            void republish( publication_record_type* pRec )
+            {
+                if ( pRec->nState.load( memory_model::memory_order_relaxed ) != active ) {
+                    // The record has been excluded from publication list. Reinsert it
+                    publish( pRec );
+                }
+            }
+
+            template <class Container>
+            void try_combining( Container& owner, publication_record_type* pRec )
+            {
+                if ( m_Mutex.try_lock() ) {
+                    // The thread becomes a combiner
+                    lock_guard l( m_Mutex, std::adopt_lock_t() );
+
+                    // The record pRec can be excluded from publication list. Re-publish it
+                    republish( pRec );
+
+                    combining( owner );
+                    assert( pRec->op( memory_model::memory_order_relaxed ) == req_Response );
+                }
+                else {
+                    // There is another combiner, wait while it executes our request
+                    if ( !wait_for_combining( pRec ) ) {
+                        // The thread becomes a combiner
+                        lock_guard l( m_Mutex, std::adopt_lock_t() );
+
+                        // The record pRec can be excluded from publication list. Re-publish it
+                        republish( pRec );
+
+                        combining( owner );
+                        assert( pRec->op( memory_model::memory_order_relaxed ) == req_Response );
+                    }
+                }
+            }
+
+            template <class Container>
+            void try_batch_combining( Container& owner, publication_record_type * pRec )
+            {
+                if ( m_Mutex.try_lock() ) {
+                    // The thread becomes a combiner
+                    lock_guard l( m_Mutex, std::adopt_lock_t() );
+
+                    // The record pRec can be excluded from publication list. Re-publish it
+                    republish( pRec );
+
+                    batch_combining( owner );
+                    assert( pRec->op( memory_model::memory_order_relaxed ) == req_Response );
+                }
+                else {
+                    // There is another combiner, wait while it executes our request
+                    if ( !wait_for_combining( pRec ) ) {
+                        // The thread becomes a combiner
+                        lock_guard l( m_Mutex, std::adopt_lock_t() );
+
+                        // The record pRec can be excluded from publication list. Re-publish it
+                        republish( pRec );
+
+                        batch_combining( owner );
+                        assert( pRec->op( memory_model::memory_order_relaxed ) == req_Response );
+                    }
+                }
+            }
+
+            template <class Container>
+            void combining( Container& owner )
+            {
+                // The thread is a combiner
+                assert( !m_Mutex.try_lock() );
+
+                unsigned int const nCurAge = m_nCount.fetch_add( 1, memory_model::memory_order_release ) + 1;
+
+                for ( unsigned int nPass = 0; nPass < m_nCombinePassCount; ++nPass )
+                    if ( !combining_pass( owner, nCurAge ))
+                        break;
+
+                m_Stat.onCombining();
+                if ( (nCurAge & m_nCompactFactor) == 0 )
+                    compact_list( nCurAge );
+            }
+
+            template <class Container>
+            bool combining_pass( Container& owner, unsigned int nCurAge )
+            {
+                publication_record* pPrev = nullptr;
+                publication_record* p = m_pHead;
+                bool bOpDone = false;
+                while ( p ) {
+                    switch ( p->nState.load( memory_model::memory_order_acquire )) {
+                        case active:
+                            if ( p->op() >= req_Operation ) {
+                                p->nAge.store( nCurAge, memory_model::memory_order_release );
+                                owner.fc_apply( static_cast<publication_record_type*>(p) );
+                                operation_done( *p );
+                                bOpDone = true;
+                            }
+                            break;
+                        case inactive:
+                            // Only m_pHead can be inactive in the publication list
+                            assert( p == m_pHead );
+                            break;
+                        case removed:
+                            // The record should be removed
+                            p = unlink_and_delete_record( pPrev, p );
+                            continue;
+                        default:
+                            /// ??? That is impossible
+                            assert(false);
+                    }
+                    pPrev = p;
+                    p = p->pNext.load( memory_model::memory_order_acquire );
+                }
+                return bOpDone;
+            }
+
+            template <class Container>
+            void batch_combining( Container& owner )
+            {
+                // The thread is a combiner
+                assert( !m_Mutex.try_lock() );
+
+                unsigned int const nCurAge = m_nCount.fetch_add( 1, memory_model::memory_order_release ) + 1;
+
+                for ( unsigned int nPass = 0; nPass < m_nCombinePassCount; ++nPass )
+                    owner.fc_process( begin(), end() );
+
+                combining_pass( owner, nCurAge );
+                m_Stat.onCombining();
+                if ( (nCurAge & m_nCompactFactor) == 0 )
+                    compact_list( nCurAge );
+            }
+
+            bool wait_for_combining( publication_record_type * pRec )
+            {
+                m_waitStrategy.prepare( *pRec );
+                m_Stat.onPassiveWait();
+
+                while ( pRec->op( memory_model::memory_order_acquire ) != req_Response ) {
+                    // The record can be excluded from publication list. Reinsert it
+                    republish( pRec );
+
+                    m_Stat.onPassiveWaitIteration();
+
+                    // Wait while operation processing
+                    if ( m_waitStrategy.wait( *this, *pRec ))
+                        m_Stat.onWakeupByNotifying();
+
+                    if ( m_Mutex.try_lock() ) {
+                        if ( pRec->op( memory_model::memory_order_acquire ) == req_Response ) {
+                            // Operation is done
+                            m_Mutex.unlock();
+
+                            // Wake up a pending threads
+                            m_waitStrategy.wakeup( *this );
+                            m_Stat.onPassiveWaitWakeup();
+
+                            break;
+                        }
+                        // The thread becomes a combiner
+                        m_Stat.onPassiveToCombiner();
+                        return false;
+                    }
+                }
+                return true;
+            }
+
+            void compact_list( unsigned int const nCurAge )
+            {
+                // Thinning publication list
+                publication_record * pPrev = nullptr;
+                for ( publication_record * p = m_pHead; p; ) {
+                    if ( p->nState.load( memory_model::memory_order_acquire ) == active
+                      && p->nAge.load( memory_model::memory_order_acquire ) + m_nCompactFactor < nCurAge )
+                    {
+                        if ( pPrev ) {
+                            publication_record * pNext = p->pNext.load( memory_model::memory_order_acquire );
+                            if ( pPrev->pNext.compare_exchange_strong( p, pNext,
+                                memory_model::memory_order_release, atomics::memory_order_relaxed ))
+                            {
+                                p->nState.store( inactive, memory_model::memory_order_release );
+                                p = pNext;
+                                m_Stat.onDeactivatePubRecord();
+                                continue;
+                            }
+                        }
+                    }
+                    pPrev = p;
+                    p = p->pNext.load( memory_model::memory_order_acquire );
+                }
+
+                m_Stat.onCompactPublicationList();
+            }
+
+            publication_record * unlink_and_delete_record( publication_record * pPrev, publication_record * p )
+            {
+                if ( pPrev ) {
+                    publication_record * pNext = p->pNext.load( memory_model::memory_order_acquire );
+                    if ( pPrev->pNext.compare_exchange_strong( p, pNext,
+                        memory_model::memory_order_release, atomics::memory_order_relaxed ))
+                    {
+                        free_publication_record( static_cast<publication_record_type *>( p ));
+                        m_Stat.onDeletePubRecord();
+                    }
+                    return pNext;
+                }
+                else {
+                    m_pHead = static_cast<publication_record_type *>( p->pNext.load( memory_model::memory_order_acquire ));
+                    free_publication_record( static_cast<publication_record_type *>( p ));
+                    m_Stat.onDeletePubRecord();
+                    return m_pHead;
+                }
+            }
+            //@endcond
+        };
+
+        //@cond
+        class container
+        {
+        public:
+            template <typename PubRecord>
+            void fc_apply( PubRecord * )
+            {
+                assert( false );
+            }
+
+            template <typename Iterator>
+            void fc_process( Iterator, Iterator )
+            {
+                assert( false );
+            }
+        };
+        //@endcond
+
+    } // namespace flat_combining
+}} // namespace cds::algo
+
+#endif // #ifndef CDSLIB_ALGO_FLAT_COMBINING_KERNEL_H
diff --git a/cds/algo/flat_combining/wait_strategy.h b/cds/algo/flat_combining/wait_strategy.h
new file mode 100644 (file)
index 0000000..c8ec6f8
--- /dev/null
@@ -0,0 +1,297 @@
+/*
+    This file is a part of libcds - Concurrent Data Structures library
+
+    (C) Copyright Maxim Khizhinsky (libcds.dev@gmail.com) 2006-2016
+
+    Source code repo: http://github.com/khizmax/libcds/
+    Download: http://sourceforge.net/projects/libcds/files/
+
+    Redistribution and use in source and binary forms, with or without
+    modification, are permitted provided that the following conditions are met:
+
+    * Redistributions of source code must retain the above copyright notice, this
+      list of conditions and the following disclaimer.
+
+    * 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 THE COPYRIGHT HOLDERS AND CONTRIBUTORS "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 THE COPYRIGHT HOLDER 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.
+*/
+
+#ifndef CDSLIB_ALGO_FLAT_COMBINING_WAIT_STRATEGY_H
+#define CDSLIB_ALGO_FLAT_COMBINING_WAIT_STRATEGY_H
+
+#include <cds/algo/flat_combining/defs.h>
+#include <cds/algo/backoff_strategy.h>
+#include <mutex>
+#include <condition_variable>
+#include <boost/thread/tss.hpp>  // thread_specific_ptr
+
+
+namespace cds { namespace opt {
+
+    /// Wait strategy option for \p flat_combining::kernel
+    template <typename Strategy>
+    struct wait_strategy {
+        //@cond
+        template <typename Base> struct pack: public Base
+        {
+            typedef Strategy wait_strategy;
+        };
+        //@endcond
+    };
+
+}} // namespace cds::opt
+
+namespace cds { namespace algo { namespace flat_combining {
+
+    /// Wait strategies for \p flat_combining technique
+    namespace wait_strategy {
+
+        /// Empty wait strategy
+        struct empty
+        {
+        //@cond
+            template <typename PublicationRecord>
+            struct make_publication_record {
+                typedef PublicationRecord type;
+            };
+
+            template <typename PublicationRecord>
+            void prepare( PublicationRecord& /*rec*/ )
+            {}
+
+            template <typename FCKernel, typename PublicationRecord>
+            bool wait( FCKernel& /*fc*/, PublicationRecord& /*rec*/ )
+            {
+                return false;
+            }
+
+            template <typename FCKernel, typename PublicationRecord>
+            void notify( FCKernel& /*fc*/, PublicationRecord& /*rec*/ )
+            {}
+
+            template <typename FCKernel>
+            void wakeup( FCKernel& /*fc*/ )
+            {}
+        //@endcond
+        };
+
+        /// Back-off wait strategy
+        template <typename BackOff = cds::backoff::delay_of<2>>
+        struct backoff
+        {
+        //@cond
+            typedef BackOff back_off;
+
+            template <typename PublicationRecord>
+            struct make_publication_record {
+                struct type: public PublicationRecord
+                {
+                    back_off bkoff;
+                };
+            };
+
+            template <typename PublicationRecord>
+            void prepare( PublicationRecord& rec )
+            {
+                rec.bkoff.reset();
+            }
+
+            template <typename FCKernel, typename PublicationRecord>
+            bool wait( FCKernel& /*fc*/, PublicationRecord& rec )
+            {
+                rec.bkoff();
+                return false;
+            }
+
+            template <typename FCKernel, typename PublicationRecord>
+            void notify( FCKernel& /*fc*/, PublicationRecord& /*rec*/ )
+            {}
+
+            template <typename FCKernel>
+            void wakeup( FCKernel& )
+            {}
+        //@endcond
+        };
+
+        /// Wait strategy based on the single mutex and the condition variable
+        /**
+            The strategy shares the mutex and conditional variable for all thread.
+
+            Template parameter \p Milliseconds specifies waiting duration;
+            the minimal value is 1.
+        */
+        template <int Milliseconds = 2>
+        class single_mutex_single_condvar
+        {
+        //@cond
+            std::mutex m_mutex;
+            std::condition_variable m_condvar;
+
+            typedef std::unique_lock< std::mutex > unique_lock;
+
+        public:
+            enum {
+                c_nWaitMilliseconds = Milliseconds < 1 ? 1 : Milliseconds
+            };
+
+            template <typename PublicationRecord>
+            struct make_publication_record {
+                typedef PublicationRecord type;
+            };
+
+            template <typename PublicationRecord>
+            void prepare( PublicationRecord& /*rec*/ )
+            {}
+
+            template <typename FCKernel, typename PublicationRecord>
+            bool wait( FCKernel& fc, PublicationRecord& rec )
+            {
+                if ( fc.get_operation( rec ) >= req_Operation ) {
+                    unique_lock lock( m_mutex );
+                    if ( fc.get_operation( rec ) >= req_Operation )
+                        return m_condvar.wait_for( lock, std::chrono::milliseconds( c_nWaitMilliseconds )) == std::cv_status::no_timeout;
+                }
+                return false;
+            }
+
+            template <typename FCKernel, typename PublicationRecord>
+            void notify( FCKernel& fc, PublicationRecord& rec )
+            {
+                m_condvar.notify_all();
+            }
+
+            template <typename FCKernel>
+            void wakeup( FCKernel& /*fc*/ )
+            {
+                m_condvar.notify_all();
+            }
+        //@endcond
+        };
+
+        /// Wait strategy based on the single mutex and thread-local condition variables
+        /**
+            The strategy shares the mutex, but each thread has its own conditional variable
+
+            Template parameter \p Milliseconds specifies waiting duration;
+            the minimal value is 1.
+        */
+        template <int Milliseconds = 2>
+        class single_mutex_multi_condvar
+        {
+        //@cond
+            std::mutex m_mutex;
+
+            typedef std::unique_lock< std::mutex > unique_lock;
+
+        public:
+            enum {
+                c_nWaitMilliseconds = Milliseconds < 1 ? 1 : Milliseconds
+            };
+
+            template <typename PublicationRecord>
+            struct make_publication_record {
+                struct type: public PublicationRecord
+                {
+                    std::condition_variable m_condvar;
+                };
+            };
+
+            template <typename PublicationRecord>
+            void prepare( PublicationRecord& /*rec*/ )
+            {}
+
+            template <typename FCKernel, typename PublicationRecord>
+            bool wait( FCKernel& fc, PublicationRecord& rec )
+            {
+                if ( fc.get_operation( rec ) >= req_Operation ) {
+                    unique_lock lock( m_mutex );
+                    if ( fc.get_operation( rec ) >= req_Operation )
+                        return rec.m_condvar.wait_for( lock, std::chrono::milliseconds( c_nWaitMilliseconds )) == std::cv_status::no_timeout;
+                }
+                return false;
+            }
+
+            template <typename FCKernel, typename PublicationRecord>
+            void notify( FCKernel& fc, PublicationRecord& rec )
+            {
+                rec.m_condvar.notify_one();
+            }
+
+            template <typename FCKernel>
+            void wakeup( FCKernel& fc )
+            {
+                fc.wakeup_any();
+            }
+        //@endcond
+        };
+
+        /// Wait strategy where each thread has a mutex and a condition variable
+        /**
+            Template parameter \p Milliseconds specifies waiting duration;
+            the minimal value is 1.
+        */
+        template <int Milliseconds = 2>
+        class multi_mutex_multi_condvar
+        {
+        //@cond
+            typedef std::unique_lock< std::mutex > unique_lock;
+
+        public:
+            enum {
+                c_nWaitMilliseconds = Milliseconds < 1 ? 1 : Milliseconds
+            };
+
+            template <typename PublicationRecord>
+            struct make_publication_record {
+                struct type: public PublicationRecord
+                {
+                    std::mutex              m_mutex;
+                    std::condition_variable m_condvar;
+                };
+            };
+
+            template <typename PublicationRecord>
+            void prepare( PublicationRecord& /*rec*/ )
+            {}
+
+            template <typename FCKernel, typename PublicationRecord>
+            bool wait( FCKernel& fc, PublicationRecord& rec )
+            {
+                if ( fc.get_operation( rec ) >= req_Operation ) {
+                    unique_lock lock( rec.m_mutex );
+                    if ( fc.get_operation( rec ) >= req_Operation )
+                        return rec.m_condvar.wait_for( lock, std::chrono::milliseconds( c_nWaitMilliseconds )) == std::cv_status::no_timeout;
+                }
+                return false;
+            }
+
+            template <typename FCKernel, typename PublicationRecord>
+            void notify( FCKernel& /*fc*/, PublicationRecord& rec )
+            {
+                rec.m_condvar.notify_one();
+            }
+
+            template <typename FCKernel>
+            void wakeup( FCKernel& fc )
+            {
+                fc.wakeup_any();
+            }
+        //@endcond
+        };
+
+    } // namespace wait_strategy
+}}} // namespace cds::algo::flat_combining
+
+#endif //CDSLIB_ALGO_FLAT_COMBINING_WAIT_STRATEGY_H
diff --git a/cds/algo/wait_strategy.h b/cds/algo/wait_strategy.h
deleted file mode 100644 (file)
index 422e12e..0000000
+++ /dev/null
@@ -1,267 +0,0 @@
-/*\r
-    This file is a part of libcds - Concurrent Data Structures library\r
-\r
-    (C) Copyright Maxim Khizhinsky (libcds.dev@gmail.com) 2006-2016\r
-\r
-    Source code repo: http://github.com/khizmax/libcds/\r
-    Download: http://sourceforge.net/projects/libcds/files/\r
-\r
-    Redistribution and use in source and binary forms, with or without\r
-    modification, are permitted provided that the following conditions are met:\r
-\r
-    * Redistributions of source code must retain the above copyright notice, this\r
-      list of conditions and the following disclaimer.\r
-\r
-    * Redistributions in binary form must reproduce the above copyright notice,\r
-      this list of conditions and the following disclaimer in the documentation\r
-      and/or other materials provided with the distribution.\r
-\r
-    THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"\r
-    AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\r
-    IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE\r
-    DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE\r
-    FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL\r
-    DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR\r
-    SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER\r
-    CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,\r
-    OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE\r
-    OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.\r
-*/\r
-\r
-#ifndef CDSLIB_ALGO_FLAT_COMBINING_WAIT_STRATEGY_H\r
-#define CDSLIB_ALGO_FLAT_COMBINING_WAIT_STRATEGY_H\r
-\r
-#include <iostream>\r
-\r
-#include <boost/thread.hpp>\r
-#include <cds/algo/backoff_strategy.h>\r
-\r
-namespace cds {  namespace algo {  namespace flat_combining {\r
-       /// Waiting strategy for flat combining\r
-\r
-    // Special values of publication_record::nRequest\r
-    enum request_value\r
-    {\r
-        req_EmptyRecord,    ///< Publication record is empty\r
-        req_Response,       ///< Operation is done\r
-\r
-        req_Operation       ///< First operation id for derived classes\r
-    };\r
-\r
-    /// publication_record state\r
-    enum record_state {\r
-        inactive,       ///< Record is inactive\r
-        active,         ///< Record is active\r
-        removed         ///< Record should be removed\r
-    };\r
-\r
-    /// do-nothing strategy\r
-    template<typename UserPublicationRecord, typename Traits>\r
-    struct BareWaitStartegy\r
-    {\r
-        struct ExtendedPublicationRecord: public UserPublicationRecord\r
-        {\r
-        };\r
-\r
-        void wait(ExtendedPublicationRecord * pRec){}\r
-        void notify(ExtendedPublicationRecord* pRec){\r
-            pRec->nRequest.store( req_Response, Traits::memory_model::memory_order_release);\r
-        }\r
-    };\r
-\r
-    /// Wait/notify strategy based on thread-local mutex and thread-local condition variable\r
-    template<typename UserPublicationRecord, typename Traits>\r
-    struct WaitStartegyMultMutexMultCondVar\r
-    {\r
-        struct ExtendedPublicationRecord: public UserPublicationRecord\r
-        {\r
-            boost::mutex                _waitMutex;\r
-            boost::condition_variable   _condVar;\r
-        };\r
-\r
-        void wait(ExtendedPublicationRecord * pRec){\r
-            boost::unique_lock<boost::mutex> lock(pRec->_waitMutex);\r
-            if (pRec->nRequest.load( Traits::memory_model::memory_order_acquire ) >= req_Operation)\r
-                               pRec->_condVar.wait(lock);\r
-        }\r
-\r
-        void notify(ExtendedPublicationRecord* pRec){\r
-            boost::unique_lock<boost::mutex> lock(pRec->_waitMutex);\r
-            pRec->nRequest.store( req_Response, Traits::memory_model::memory_order_release);\r
-            pRec->_condVar.notify_one();\r
-        }\r
-    };\r
-\r
-    /// Back-off strategy\r
-    /**\r
-     * When N threads compete for the critical resource that can be accessed with the help of CAS-operations,\r
-     * only one of them gets an access. Other N–1 threads interrupt each other and consume process time in vain.\r
-     */\r
-    template<typename UserPublicationRecord, typename Traits>\r
-       struct WaitBakkOffStrategy\r
-       {\r
-        struct ExtendedPublicationRecord: public UserPublicationRecord\r
-        {\r
-        };\r
-        void wait(ExtendedPublicationRecord * pRec){\r
-                             cds::backoff::delay_of<2>   back_off;\r
-            back_off();\r
-        }\r
-        void notify(ExtendedPublicationRecord* pRec){\r
-            pRec->nRequest.store( req_Response, Traits::memory_model::memory_order_release);\r
-        }\r
-    };\r
-\r
-    /// Wait/notify strategy based on the global mutex and the condition variable\r
-    /**\r
-     *  The strategy is based on the usage of synchronization primitives of\r
-     *  the FC core which are shared by all threads.\r
-     */\r
-    template<typename UserPublicationRecord, typename Traits>\r
-    struct WaitOneMutexOneCondVarStrategy\r
-       {\r
-        boost::mutex              _globalMutex;\r
-        boost::condition_variable _globalCondVar;\r
-\r
-        struct ExtendedPublicationRecord: public UserPublicationRecord\r
-        {\r
-        };\r
-        void wait(ExtendedPublicationRecord * pRec){\r
-            boost::unique_lock<boost::mutex> lock(_globalMutex);\r
-            if (pRec->nRequest.load( Traits::memory_model::memory_order_acquire ) >= req_Operation)\r
-                               //_globalCondVar.timed_wait(lock, static_cast<boost::posix_time::seconds>(1));\r
-                               _globalCondVar.wait(lock);\r
-        }\r
-\r
-        void notify(ExtendedPublicationRecord* pRec){\r
-            boost::unique_lock<boost::mutex> lock(_globalMutex);\r
-            pRec->nRequest.store( req_Response, Traits::memory_model::memory_order_release);\r
-            _globalCondVar.notify_all();\r
-        }\r
-    };\r
-\r
-    /// Wait/notify strategy based on the global mutex and the thread-local condition variable\r
-    /**\r
-     * It uses one extra mutex aggregated in the FC core and a condition variable for every\r
-     * thread aggregated in thread's publication record in the publication list\r
-     */\r
-       template<typename UserPublicationRecord, typename Traits>\r
-       struct WaitStratygyBaseOnSingleMutexMutlLocalCondVars\r
-       {\r
-               boost::mutex              _globalMutex;\r
-\r
-               struct ExtendedPublicationRecord : public UserPublicationRecord\r
-               {\r
-                       boost::condition_variable _globalCondVar;\r
-               };\r
-               void wait(ExtendedPublicationRecord * pRec){\r
-                       boost::unique_lock<boost::mutex> lock(_globalMutex);\r
-                       if (pRec->nRequest.load(Traits::memory_model::memory_order_acquire) >= req_Operation)\r
-                               //_globalCondVar.timed_wait(lock, static_cast<boost::posix_time::seconds>(1));\r
-                               pRec->_globalCondVar.wait(lock);\r
-               }\r
-\r
-               void notify(ExtendedPublicationRecord* pRec){\r
-                       boost::unique_lock<boost::mutex> lock(_globalMutex);\r
-                       pRec->nRequest.store(req_Response, Traits::memory_model::memory_order_release);\r
-                       pRec->_globalCondVar.notify_one();\r
-               }\r
-       };\r
-\r
-       //===================================================================\r
-    template<typename UserPublicationRecord, typename Traits>\r
-    struct TimedWaitMultMutexMultCondVar\r
-    {\r
-        struct ExtendedPublicationRecord: public UserPublicationRecord\r
-        {\r
-            boost::mutex                _waitMutex;\r
-            boost::condition_variable   _condVar;\r
-        };\r
-\r
-        void wait(ExtendedPublicationRecord * pRec){\r
-            boost::unique_lock<boost::mutex> lock(pRec->_waitMutex);\r
-            if (pRec->nRequest.load( Traits::memory_model::memory_order_acquire ) >= req_Operation)\r
-                pRec->_condVar.timed_wait(lock, boost::posix_time::millisec(2));\r
-        }\r
-\r
-        void notify(ExtendedPublicationRecord* pRec){\r
-            pRec->nRequest.store(req_Response, Traits::memory_model::memory_order_release);\r
-            pRec->_condVar.notify_one();\r
-        }\r
-    };\r
-    //===================================================================\r
-    template<typename UserPublicationRecord, typename Traits>\r
-    struct TimedWaitGlobalMutexAndCondVar\r
-    {\r
-        boost::mutex              _globalMutex;\r
-        boost::condition_variable _globalCondVar;\r
-\r
-        struct ExtendedPublicationRecord: public UserPublicationRecord\r
-        {\r
-        };\r
-        void wait(ExtendedPublicationRecord * pRec){\r
-            boost::unique_lock<boost::mutex> lock(_globalMutex);\r
-            if (pRec->nRequest.load( Traits::memory_model::memory_order_acquire ) >= req_Operation)\r
-                _globalCondVar.timed_wait(lock, boost::posix_time::millisec(2));\r
-        }\r
-\r
-        void notify(ExtendedPublicationRecord* pRec){\r
-            pRec->nRequest.store( req_Response, Traits::memory_model::memory_order_release);\r
-            _globalCondVar.notify_all();\r
-        }\r
-    };\r
-    //===================================================================\r
-    template <int v>\r
-    struct Int2Type{\r
-        enum {value = v};\r
-    };\r
-\r
-    ///Adaptive strategy\r
-    /**\r
-     * It works like “back-off”-strategy with “light” elements with small size\r
-     * and like Wait/notify strategy based on thread-local mutex and thread-local condition variable\r
-     * with “heavy” elements.\r
-     */\r
-    template<typename UserPublicationRecord, typename Traits>\r
-    struct AutoWaitStrategy\r
-    {\r
-        struct ExtendedPublicationRecord: public UserPublicationRecord\r
-        {\r
-            boost::mutex                _waitMutex;\r
-            boost::condition_variable   _condVar;\r
-        };\r
-\r
-        void wait(ExtendedPublicationRecord * pRec){\r
-            doWait(pRec, Int2Type<sizeof(typename UserPublicationRecord::value_type) <= 4*sizeof(int) >());\r
-        }\r
-\r
-        void notify(ExtendedPublicationRecord* pRec){\r
-            doNotify(pRec, Int2Type<sizeof(typename UserPublicationRecord::value_type) <= 4*sizeof(int) >());\r
-        }\r
-\r
-     private:\r
-        //The container consists a small data\r
-        void doWait(ExtendedPublicationRecord * pRec, Int2Type<true>){\r
-            cds::backoff::delay_of<2>   back_off;\r
-            back_off();\r
-        }\r
-\r
-        void doNotify(ExtendedPublicationRecord * pRec, Int2Type<true>){\r
-            pRec->nRequest.store( req_Response, Traits::memory_model::memory_order_release);\r
-        }\r
-\r
-        //The container consists a big data\r
-        void doWait(ExtendedPublicationRecord * pRec, Int2Type<false>){\r
-            boost::unique_lock<boost::mutex> lock(pRec->_waitMutex);\r
-            if (pRec->nRequest.load( Traits::memory_model::memory_order_acquire ) >= req_Operation)\r
-                pRec->_condVar.timed_wait(lock, boost::posix_time::millisec(2));\r
-        }\r
-\r
-        void doNotify(ExtendedPublicationRecord * pRec, Int2Type<false>){\r
-            pRec->nRequest.store(req_Response, Traits::memory_model::memory_order_release);\r
-            pRec->_condVar.notify_one();\r
-        }\r
-    };//class AutoWaitStrategy\r
-}}}//end namespace cds::algo::flat_combining\r
-\r
-#endif //CDSLIB_ALGO_FLAT_COMBINING_WAIT_STRATEGY_H\r
index 7dccc157cdedf821e3aa36bfb77e26bc17f9b586..fc28a24e554954166b73a63c0379b9f7e0979e59 100644 (file)
@@ -94,13 +94,8 @@ namespace cds { namespace container {
         /// Metafunction converting option list to traits
         /**
             \p Options are:
-            - \p opt::lock_type - mutex type, default is \p cds::sync::spin
-            - \p opt::back_off - back-off strategy, defalt is \p cds::backoff::delay_of<2>
-            - \p opt::allocator - allocator type, default is \ref CDS_DEFAULT_ALLOCATOR
+            - any \p cds::algo::flat_combining::make_traits options
             - \p opt::stat - internal statistics, possible type: \ref stat, \ref empty_stat (the default)
-            - \p opt::memory_model - C++ memory ordering model.
-                List of all available memory ordering see opt::memory_model.
-                Default if cds::opt::v:relaxed_ordering
             - \p opt::enable_elimination - enable/disable operation \ref cds_elimination_description "elimination"
                 By default, the elimination is disabled. For queue, the elimination is possible if the queue
                 is empty.
@@ -161,8 +156,7 @@ namespace cds { namespace container {
             op_push_back_move,      ///< Push back (move semantics)
             op_pop_front,           ///< Pop front
             op_pop_back,            ///< Pop back
-            op_clear,               ///< Clear
-            op_empty                ///< Empty
+            op_clear                ///< Clear
         };
 
         /// Flat combining publication list record
@@ -181,8 +175,8 @@ namespace cds { namespace container {
 
     protected:
         //@cond
-        fc_kernel   m_FlatCombining;
-        deque_type  m_Deque;
+        mutable fc_kernel m_FlatCombining;
+        deque_type        m_Deque;
         //@endcond
 
     public:
@@ -206,7 +200,7 @@ namespace cds { namespace container {
             value_type const& val ///< Value to be copied to inserted element
         )
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
             pRec->pValPush = &val;
 
             if ( c_bEliminationEnabled )
@@ -228,7 +222,7 @@ namespace cds { namespace container {
             value_type&& val ///< Value to be moved to inserted element
         )
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
             pRec->pValPush = &val;
 
             if ( c_bEliminationEnabled )
@@ -250,7 +244,7 @@ namespace cds { namespace container {
             value_type const& val ///< Value to be copied to inserted element
         )
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
             pRec->pValPush = &val;
 
             if ( c_bEliminationEnabled )
@@ -272,7 +266,7 @@ namespace cds { namespace container {
             value_type&& val ///< Value to be moved to inserted element
         )
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
             pRec->pValPush = &val;
 
             if ( c_bEliminationEnabled )
@@ -295,7 +289,7 @@ namespace cds { namespace container {
             value_type& val ///< Target to be received the copy of removed element
         )
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
             pRec->pValPop = &val;
 
             if ( c_bEliminationEnabled )
@@ -318,7 +312,7 @@ namespace cds { namespace container {
             value_type& val ///< Target to be received the copy of removed element
         )
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
             pRec->pValPop = &val;
 
             if ( c_bEliminationEnabled )
@@ -335,7 +329,7 @@ namespace cds { namespace container {
         /// Clears the deque
         void clear()
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
 
             if ( c_bEliminationEnabled )
                 m_FlatCombining.batch_combine( op_clear, pRec, *this );
@@ -361,18 +355,12 @@ namespace cds { namespace container {
         /**
             If the combining is in process the function waits while combining done.
         */
-        bool empty()
+        bool empty() const
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
-
-            if ( c_bEliminationEnabled )
-                m_FlatCombining.batch_combine( op_empty, pRec, *this );
-            else
-                m_FlatCombining.combine( op_empty, pRec, *this );
-
-            assert( pRec->is_done() );
-            m_FlatCombining.release_record( pRec );
-            return pRec->bEmpty;
+            bool bRet = false;
+            auto const& deq = m_Deque;
+            m_FlatCombining.invoke_exclusive( [&deq, &bRet]() { bRet = deq.empty(); } );
+            return bRet;
         }
 
         /// Internal statistics
@@ -433,9 +421,6 @@ namespace cds { namespace container {
                 while ( !m_Deque.empty() )
                     m_Deque.pop_front();
                 break;
-            case op_empty:
-                pRec->bEmpty = m_Deque.empty();
-                break;
             default:
                 assert(false);
                 break;
index 779849b5fd73c438af7c0d7a084304707f36ed3a..c06a9ce960a65bfa56047462342a9453cff99b36 100644 (file)
@@ -80,13 +80,8 @@ namespace cds { namespace container {
         /// Metafunction converting option list to traits
         /**
             \p Options are:
-            - \p opt::lock_type - mutex type, default is \p cds::sync::spin
-            - \p opt::back_off - back-off strategy, defalt is \p cds::backoff::delay_of<2>
-            - \p opt::allocator - allocator type, default is \ref CDS_DEFAULT_ALLOCATOR
+            - any \p cds::algo::flat_combining::make_traits options
             - \p opt::stat - internal statistics, possible type: \p fcpqueue::stat, \p fcpqueue::empty_stat (the default)
-            - \p opt::memory_model - C++ memory ordering model.
-                List of all available memory ordering see \p opt::memory_model.
-                Default is \p cds::opt::v:relaxed_ordering
         */
         template <typename... Options>
         struct make_traits {
@@ -139,8 +134,7 @@ namespace cds { namespace container {
             op_push = cds::algo::flat_combining::req_Operation,
             op_push_move,
             op_pop,
-            op_clear,
-            op_empty
+            op_clear
         };
 
         // Flat combining publication list record
@@ -159,8 +153,8 @@ namespace cds { namespace container {
 
     protected:
         //@cond
-        fc_kernel               m_FlatCombining;
-        priority_queue_type     m_PQueue;
+        mutable fc_kernel   m_FlatCombining;
+        priority_queue_type m_PQueue;
         //@endcond
 
     public:
@@ -184,7 +178,7 @@ namespace cds { namespace container {
             value_type const& val ///< Value to be copied to inserted element
         )
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
             pRec->pValPush = &val;
 
             m_FlatCombining.combine( op_push, pRec, *this );
@@ -203,7 +197,7 @@ namespace cds { namespace container {
             value_type&& val ///< Value to be moved to inserted element
         )
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
             pRec->pValPush = &val;
 
             m_FlatCombining.combine( op_push_move, pRec, *this );
@@ -223,7 +217,7 @@ namespace cds { namespace container {
             value_type& val ///< Target to be received the copy of top element
         )
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
             pRec->pValPop = &val;
 
             m_FlatCombining.combine( op_pop, pRec, *this );
@@ -237,7 +231,7 @@ namespace cds { namespace container {
         /// Clears the priority queue
         void clear()
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
 
            m_FlatCombining.combine( op_clear, pRec, *this );
 
@@ -262,12 +256,10 @@ namespace cds { namespace container {
         */
         bool empty()
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
-
-            m_FlatCombining.combine( op_empty, pRec, *this );
-            assert( pRec->is_done() );
-            m_FlatCombining.release_record( pRec );
-            return pRec->bEmpty;
+            bool bRet = false;
+            auto const& pq = m_PQueue;
+            m_FlatCombining.invoke_exclusive( [&pq, &bRet]() { bRet = pq.empty(); } );
+            return bRet;
         }
 
         /// Internal statistics
@@ -311,9 +303,6 @@ namespace cds { namespace container {
                 while ( !m_PQueue.empty() )
                     m_PQueue.pop();
                 break;
-            case op_empty:
-                pRec->bEmpty = m_PQueue.empty();
-                break;
             default:
                 assert(false);
                 break;
index dff95edea937173e8b044efd083dffcfdbf567ad..1f7c7346bc32fe5896b1d53f95311c0bf00bd0a4 100644 (file)
@@ -84,13 +84,8 @@ namespace cds { namespace container {
         /// Metafunction converting option list to traits
         /**
             \p Options are:
-            - \p opt::lock_type - mutex type, default is \p cds::sync::spin
-            - \p opt::back_off - back-off strategy, defalt is \p cds::backoff::delay_of<2>
-            - \p opt::allocator - allocator type, default is \ref CDS_DEFAULT_ALLOCATOR
+            - any \p cds::algo::flat_combining::make_traits options
             - \p opt::stat - internal statistics, possible type: \p fcqueue::stat, \p fcqueue::empty_stat (the default)
-            - \p opt::memory_model - C++ memory ordering model.
-                List of all available memory ordering see \p opt::memory_model.
-                Default is \p cds::opt::v:relaxed_ordering
             - \p opt::enable_elimination - enable/disable operation \ref cds_elimination_description "elimination"
                 By default, the elimination is disabled. For queue, the elimination is possible if the queue
                 is empty.
@@ -147,8 +142,7 @@ namespace cds { namespace container {
             op_enq = cds::algo::flat_combining::req_Operation, ///< Enqueue
             op_enq_move,    ///< Enqueue (move semantics)
             op_deq,         ///< Dequeue
-            op_clear,       ///< Clear
-            op_empty        ///< Empty
+            op_clear        ///< Clear
         };
 
         /// Flat combining publication list record
@@ -167,8 +161,8 @@ namespace cds { namespace container {
 
     protected:
         //@cond
-        fc_kernel   m_FlatCombining;
-        queue_type  m_Queue;
+        mutable fc_kernel m_FlatCombining;
+        queue_type        m_Queue;
         //@endcond
 
     public:
@@ -192,7 +186,7 @@ namespace cds { namespace container {
         */
         bool enqueue( value_type const& val )
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
             pRec->pValEnq = &val;
 
             if ( c_bEliminationEnabled )
@@ -218,7 +212,7 @@ namespace cds { namespace container {
         */
         bool enqueue( value_type&& val )
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
             pRec->pValEnq = &val;
 
             if ( c_bEliminationEnabled )
@@ -245,7 +239,7 @@ namespace cds { namespace container {
         */
         bool dequeue( value_type& val )
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
             pRec->pValDeq = &val;
 
             if ( c_bEliminationEnabled )
@@ -269,7 +263,7 @@ namespace cds { namespace container {
         /// Clears the queue
         void clear()
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
 
             if ( c_bEliminationEnabled )
                 m_FlatCombining.batch_combine( op_clear, pRec, *this );
@@ -295,18 +289,12 @@ namespace cds { namespace container {
         /**
             If the combining is in process the function waits while combining done.
         */
-        bool empty()
+        bool empty() const
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
-
-            if ( c_bEliminationEnabled )
-                m_FlatCombining.batch_combine( op_empty, pRec, *this );
-            else
-                m_FlatCombining.combine( op_empty, pRec, *this );
-
-            assert( pRec->is_done() );
-            m_FlatCombining.release_record( pRec );
-            return pRec->bEmpty;
+            bool bRet = false;
+            auto const& queue = m_Queue;
+            m_FlatCombining.invoke_exclusive( [&queue, &bRet]() { bRet = queue.empty(); } );
+            return bRet;
         }
 
         /// Internal statistics
@@ -351,9 +339,6 @@ namespace cds { namespace container {
                 while ( !m_Queue.empty() )
                     m_Queue.pop();
                 break;
-            case op_empty:
-                pRec->bEmpty = m_Queue.empty();
-                break;
             default:
                 assert(false);
                 break;
index eb7c8de513e99abe743e272f2142065c0d3b7a6b..ba95875ba53ca7e3dfe438e8e84c033d2ff91974 100644 (file)
@@ -84,13 +84,8 @@ namespace cds { namespace container {
         /// Metafunction converting option list to traits
         /**
             \p Options are:
-            - \p opt::lock_type - mutex type, default is \p cds::sync::spin
-            - \p opt::back_off - back-off strategy, defalt is \p cds::backoff::Default
-            - \p opt::allocator - allocator type, default is \ref CDS_DEFAULT_ALLOCATOR
+            - any \p cds::algo::flat_combining::make_traits options
             - \p opt::stat - internal statistics, possible type: \p fcstack::stat, \p fcstack::empty_stat (the default)
-            - \p opt::memory_model - C++ memory ordering model.
-                List of all available memory ordering see \p opt::memory_model.
-                Default is \p cds::opt::v:relaxed_ordering
             - \p opt::enable_elimination - enable/disable operation \ref cds_elimination_description "elimination"
                 By default, the elimination is disabled.
         */
@@ -165,8 +160,8 @@ namespace cds { namespace container {
 
     protected:
         //@cond
-        fc_kernel   m_FlatCombining;
-        stack_type  m_Stack;
+        mutable fc_kernel m_FlatCombining;
+        stack_type        m_Stack;
         //@endcond
 
     public:
@@ -188,7 +183,7 @@ namespace cds { namespace container {
         */
         bool push( value_type const& val )
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
             pRec->pValPush = &val;
 
             if ( c_bEliminationEnabled )
@@ -208,7 +203,7 @@ namespace cds { namespace container {
         */
         bool push( value_type&& val )
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
             pRec->pValPush = &val;
 
             if ( c_bEliminationEnabled )
@@ -229,7 +224,7 @@ namespace cds { namespace container {
         */
         bool pop( value_type& val )
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
             pRec->pValPop = &val;
 
             if ( c_bEliminationEnabled )
@@ -247,7 +242,7 @@ namespace cds { namespace container {
         /// Clears the stack
         void clear()
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
 
             if ( c_bEliminationEnabled )
                 m_FlatCombining.batch_combine( op_clear, pRec, *this );
@@ -275,7 +270,7 @@ namespace cds { namespace container {
         */
         bool empty()
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
 
             if ( c_bEliminationEnabled )
                 m_FlatCombining.batch_combine( op_empty, pRec, *this );
index dae8166584fe2295628fe1db996fe7acc5dae597..412bf57f37efac2d67dda8dbcf54e84d3135e50e 100644 (file)
@@ -81,15 +81,10 @@ namespace cds { namespace intrusive {
         /// Metafunction converting option list to traits
         /**
             \p Options are:
-            - \p opt::lock_type - mutex type, default is \p cds::sync::spin
-            - \p opt::back_off - back-off strategy, defalt is \p cds::backoff::Default
+            - any \p cds::algo::flat_combining::make_traits options
             - \p opt::disposer - the functor used to dispose removed items. Default is \p opt::intrusive::v::empty_disposer.
                 This option is used only in \p FCQueue::clear() function.
-            - \p opt::allocator - allocator type, default is \ref CDS_DEFAULT_ALLOCATOR
             - \p opt::stat - internal statistics, possible type: \p fcqueue::stat, \p fcqueue::empty_stat (the default)
-            - \p opt::memory_model - C++ memory ordering model.
-                List of all available memory ordering see \p opt::memory_model.
-                Default is \p cds::opt::v:relaxed_ordering
             - \p opt::enable_elimination - enable/disable operation \ref cds_elimination_description "elimination"
                 By default, the elimination is disabled (\p false)
         */
@@ -161,8 +156,8 @@ namespace cds { namespace intrusive {
 
     protected:
         //@cond
-        fc_kernel       m_FlatCombining;
-        container_type  m_Queue;
+        mutable fc_kernel m_FlatCombining;
+        container_type    m_Queue;
         //@endcond
 
     public:
@@ -184,7 +179,7 @@ namespace cds { namespace intrusive {
         */
         bool enqueue( value_type& val )
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
             pRec->pVal = &val;
 
             if ( c_bEliminationEnabled )
@@ -210,7 +205,7 @@ namespace cds { namespace intrusive {
         */
         value_type * dequeue()
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
             pRec->pVal = nullptr;
 
             if ( c_bEliminationEnabled )
@@ -238,7 +233,7 @@ namespace cds { namespace intrusive {
         */
         void clear( bool bDispose = false )
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
 
             if ( c_bEliminationEnabled )
                 m_FlatCombining.batch_combine( bDispose ? op_clear_and_dispose : op_clear, pRec, *this );
@@ -266,8 +261,10 @@ namespace cds { namespace intrusive {
         */
         bool empty() const
         {
-            m_FlatCombining.wait_while_combining();
-            return m_Queue.empty();
+            bool bRet = false;
+            auto const& queue = m_Queue;
+            m_FlatCombining.invoke_exclusive([&queue, &bRet]() { bRet = queue.empty(); });
+            return bRet;
         }
 
         /// Internal statistics
index cfb1626eb64fcce1f2014a81066d57326c56ecd2..553ac903c928912eccbc52443ed8c6229202ffc1 100644 (file)
@@ -81,15 +81,10 @@ namespace cds { namespace intrusive {
         /// Metafunction converting option list to traits
         /**
             \p Options are:
-            - \p opt::lock_type - mutex type, default is \p cds::sync::spin
-            - \p opt::back_off - back-off strategy, defalt is \p cds::backoff::Default
+            - any \p cds::algo::flat_combining::make_traits options
             - \p opt::disposer - the functor used for dispose removed items. Default is \p opt::intrusive::v::empty_disposer.
                 This option is used only in \p FCStack::clear() function.
-            - \p opt::allocator - allocator type, default is \ref CDS_DEFAULT_ALLOCATOR
             - \p opt::stat - internal statistics, possible type: \p fcstack::stat, \p fcstack::empty_stat (the default)
-            - \p opt::memory_model - C++ memory ordering model.
-                List of all available memory ordering see \p opt::memory_model.
-                Default is \p cds::opt::v:relaxed_ordering
             - \p opt::enable_elimination - enable/disable operation \ref cds_elimination_description "elimination"
                 By default, the elimination is disabled.
         */
@@ -162,8 +157,8 @@ namespace cds { namespace intrusive {
 
     protected:
         //@cond
-        fc_kernel       m_FlatCombining;
-        container_type  m_Stack;
+        mutable fc_kernel m_FlatCombining;
+        container_type    m_Stack;
         //@endcond
 
     public:
@@ -185,7 +180,7 @@ namespace cds { namespace intrusive {
         */
         bool push( value_type& val )
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
             pRec->pVal = &val;
 
             if ( c_bEliminationEnabled )
@@ -202,7 +197,7 @@ namespace cds { namespace intrusive {
         /// Removes the element on top of the stack
         value_type * pop()
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
             pRec->pVal = nullptr;
 
             if ( c_bEliminationEnabled )
@@ -224,7 +219,7 @@ namespace cds { namespace intrusive {
         */
         void clear( bool bDispose = false )
         {
-            fc_record * pRec = m_FlatCombining.acquire_record();
+            auto pRec = m_FlatCombining.acquire_record();
 
             if ( c_bEliminationEnabled )
                 m_FlatCombining.batch_combine( bDispose ? op_clear_and_dispose : op_clear, pRec, *this );
@@ -252,8 +247,10 @@ namespace cds { namespace intrusive {
         */
         bool empty() const
         {
-            m_FlatCombining.wait_while_combining();
-            return m_Stack.empty();
+            bool bRet = false;
+            auto const& stack = m_Stack;
+            m_FlatCombining.invoke_exclusive( [&stack, &bRet]() { bRet = stack.empty(); } );
+            return bRet;
         }
 
         /// Internal statistics
@@ -271,7 +268,7 @@ namespace cds { namespace intrusive {
             object if the current thread becomes a combiner. Invocation of the function means that
             the stack should perform an action recorded in \p pRec.
         */
-        void fc_apply( fc_record * pRec )
+        void fc_apply( fc_record* pRec )
         {
             assert( pRec );
 
index c15b6f4a57129794d011b2ac43e14d23ad46603c..4864a992a54bdb35a0d7529355ef67cfe70eaf9b 100644 (file)
@@ -23,6 +23,7 @@ Project("{2150E333-8FDC-42A3-9474-1A3956D46DE8}") = "cds_test", "cds_test", "{3A
                ..\..\..\test\include\cds_test\stat_cuckoo_out.h = ..\..\..\test\include\cds_test\stat_cuckoo_out.h\r
                ..\..\..\test\include\cds_test\stat_ellenbintree_out.h = ..\..\..\test\include\cds_test\stat_ellenbintree_out.h\r
                ..\..\..\test\include\cds_test\stat_feldman_hashset_out.h = ..\..\..\test\include\cds_test\stat_feldman_hashset_out.h\r
+               ..\..\..\test\include\cds_test\stat_flat_combining_out.h = ..\..\..\test\include\cds_test\stat_flat_combining_out.h\r
                ..\..\..\test\include\cds_test\stat_skiplist_out.h = ..\..\..\test\include\cds_test\stat_skiplist_out.h\r
                ..\..\..\test\include\cds_test\stat_splitlist_out.h = ..\..\..\test\include\cds_test\stat_splitlist_out.h\r
                ..\..\..\test\include\cds_test\stat_sync_monitor_out.h = ..\..\..\test\include\cds_test\stat_sync_monitor_out.h\r
index 2a78696d82fdd8cedf491dae80df8d70a1497870..9694c1e025ef3fd2bb868783e4112e555aabcd65 100644 (file)
     <ClInclude Include="..\..\..\cds\algo\backoff_strategy.h" />\r
     <ClInclude Include="..\..\..\cds\algo\base.h" />\r
     <ClInclude Include="..\..\..\cds\algo\bitop.h" />\r
+    <ClInclude Include="..\..\..\cds\algo\flat_combining\defs.h" />\r
+    <ClInclude Include="..\..\..\cds\algo\flat_combining\kernel.h" />\r
+    <ClInclude Include="..\..\..\cds\algo\flat_combining\wait_strategy.h" />\r
     <ClInclude Include="..\..\..\cds\algo\split_bitstring.h" />\r
     <ClInclude Include="..\..\..\cds\algo\elimination.h" />\r
     <ClInclude Include="..\..\..\cds\algo\elimination_opt.h" />\r
index 3e4963d997aaf7cdbe34b26b0a4b7f45afcbeb58..b9a9ef05b862f083df6779cadd730ea6da53ff19 100644 (file)
     <Filter Include="Header Files\cds\sync">\r
       <UniqueIdentifier>{03d212fb-73f8-4f0e-9aff-f22b0783fee8}</UniqueIdentifier>\r
     </Filter>\r
+    <Filter Include="Header Files\cds\algo\flat_combining">\r
+      <UniqueIdentifier>{fe703227-44ad-4ad6-bae4-b6c9f5c65355}</UniqueIdentifier>\r
+    </Filter>\r
   </ItemGroup>\r
   <ItemGroup>\r
     <ClCompile Include="..\..\..\src\dllmain.cpp">\r
     <ClInclude Include="..\..\..\cds\intrusive\free_list_tagged.h">\r
       <Filter>Header Files\cds\intrusive</Filter>\r
     </ClInclude>\r
+    <ClInclude Include="..\..\..\cds\algo\flat_combining\defs.h">\r
+      <Filter>Header Files\cds\algo\flat_combining</Filter>\r
+    </ClInclude>\r
+    <ClInclude Include="..\..\..\cds\algo\flat_combining\wait_strategy.h">\r
+      <Filter>Header Files\cds\algo\flat_combining</Filter>\r
+    </ClInclude>\r
+    <ClInclude Include="..\..\..\cds\algo\flat_combining\kernel.h">\r
+      <Filter>Header Files\cds\algo\flat_combining</Filter>\r
+    </ClInclude>\r
   </ItemGroup>\r
 </Project>
\ No newline at end of file
index 5dc030055b86362602f60bd358c4d8e444d41f3a..e2184cd888bdc64cb660f320abe8f282efeeb115 100644 (file)
     <ClCompile Include="..\..\..\test\stress\main.cpp" />\r
     <ClCompile Include="..\..\..\test\stress\queue\bounded_queue_fulness.cpp" />\r
     <ClCompile Include="..\..\..\test\stress\queue\intrusive_push_pop.cpp" />\r
-    <ClCompile Include="..\..\..\test\stress\queue\pop.cpp" />\r
-    <ClCompile Include="..\..\..\test\stress\queue\push.cpp" />\r
+    <ClCompile Include="..\..\..\test\stress\queue\pop.cpp">\r
+      <AdditionalOptions Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">/bigobj %(AdditionalOptions)</AdditionalOptions>\r
+      <AdditionalOptions Condition="'$(Configuration)|$(Platform)'=='DebugVLD|Win32'">/bigobj %(AdditionalOptions)</AdditionalOptions>\r
+      <AdditionalOptions Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">/bigobj %(AdditionalOptions)</AdditionalOptions>\r
+      <AdditionalOptions Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">/bigobj %(AdditionalOptions)</AdditionalOptions>\r
+      <AdditionalOptions Condition="'$(Configuration)|$(Platform)'=='DebugVLD|x64'">/bigobj %(AdditionalOptions)</AdditionalOptions>\r
+      <AdditionalOptions Condition="'$(Configuration)|$(Platform)'=='Release|x64'">/bigobj %(AdditionalOptions)</AdditionalOptions>\r
+    </ClCompile>\r
+    <ClCompile Include="..\..\..\test\stress\queue\push.cpp">\r
+      <AdditionalOptions Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">/bigobj %(AdditionalOptions)</AdditionalOptions>\r
+      <AdditionalOptions Condition="'$(Configuration)|$(Platform)'=='DebugVLD|Win32'">/bigobj %(AdditionalOptions)</AdditionalOptions>\r
+      <AdditionalOptions Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">/bigobj %(AdditionalOptions)</AdditionalOptions>\r
+      <AdditionalOptions Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">/bigobj %(AdditionalOptions)</AdditionalOptions>\r
+      <AdditionalOptions Condition="'$(Configuration)|$(Platform)'=='DebugVLD|x64'">/bigobj %(AdditionalOptions)</AdditionalOptions>\r
+      <AdditionalOptions Condition="'$(Configuration)|$(Platform)'=='Release|x64'">/bigobj %(AdditionalOptions)</AdditionalOptions>\r
+    </ClCompile>\r
     <ClCompile Include="..\..\..\test\stress\queue\push_pop.cpp">\r
       <AdditionalOptions Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">/bigobj %(AdditionalOptions)</AdditionalOptions>\r
       <AdditionalOptions Condition="'$(Configuration)|$(Platform)'=='DebugVLD|Win32'">/bigobj %(AdditionalOptions)</AdditionalOptions>\r
       <AdditionalOptions Condition="'$(Configuration)|$(Platform)'=='DebugVLD|x64'">/bigobj %(AdditionalOptions)</AdditionalOptions>\r
       <AdditionalOptions Condition="'$(Configuration)|$(Platform)'=='Release|x64'">/bigobj %(AdditionalOptions)</AdditionalOptions>\r
     </ClCompile>\r
-    <ClCompile Include="..\..\..\test\stress\queue\random.cpp" />\r
+    <ClCompile Include="..\..\..\test\stress\queue\random.cpp">\r
+      <AdditionalOptions Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">/bigobj %(AdditionalOptions)</AdditionalOptions>\r
+      <AdditionalOptions Condition="'$(Configuration)|$(Platform)'=='DebugVLD|Win32'">/bigobj %(AdditionalOptions)</AdditionalOptions>\r
+      <AdditionalOptions Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">/bigobj %(AdditionalOptions)</AdditionalOptions>\r
+      <AdditionalOptions Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">/bigobj %(AdditionalOptions)</AdditionalOptions>\r
+      <AdditionalOptions Condition="'$(Configuration)|$(Platform)'=='DebugVLD|x64'">/bigobj %(AdditionalOptions)</AdditionalOptions>\r
+      <AdditionalOptions Condition="'$(Configuration)|$(Platform)'=='Release|x64'">/bigobj %(AdditionalOptions)</AdditionalOptions>\r
+    </ClCompile>\r
   </ItemGroup>\r
   <ItemGroup>\r
     <ClInclude Include="..\..\..\test\stress\queue\intrusive_queue_type.h" />\r
diff --git a/test/include/cds_test/stat_flat_combining_out.h b/test/include/cds_test/stat_flat_combining_out.h
new file mode 100644 (file)
index 0000000..d18b0a9
--- /dev/null
@@ -0,0 +1,64 @@
+/*
+    This file is a part of libcds - Concurrent Data Structures library
+
+    (C) Copyright Maxim Khizhinsky (libcds.dev@gmail.com) 2006-2016
+
+    Source code repo: http://github.com/khizmax/libcds/
+    Download: http://sourceforge.net/projects/libcds/files/
+    
+    Redistribution and use in source and binary forms, with or without
+    modification, are permitted provided that the following conditions are met:
+
+    * Redistributions of source code must retain the above copyright notice, this
+      list of conditions and the following disclaimer.
+
+    * 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 THE COPYRIGHT HOLDERS AND CONTRIBUTORS "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 THE COPYRIGHT HOLDER 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.     
+*/
+
+#ifndef CDSTEST_STAT_FLAT_COMBINING_OUT_H
+#define CDSTEST_STAT_FLAT_COMBINING_OUT_H
+
+#include <cds/algo/flat_combining.h>
+
+namespace cds_test {
+
+    static inline property_stream& operator <<( property_stream& o, cds::algo::flat_combining::empty_stat const& /*s*/ )
+    {
+        return o;
+    }
+
+    static inline property_stream& operator <<( property_stream& o, cds::algo::flat_combining::stat<> const& s )
+    {
+        return o
+            << CDSSTRESS_STAT_OUT_( "combining_factor", s.combining_factor() )
+            << CDSSTRESS_STAT_OUT( s, m_nOperationCount )
+            << CDSSTRESS_STAT_OUT( s, m_nCombiningCount )
+            << CDSSTRESS_STAT_OUT( s, m_nCompactPublicationList )
+            << CDSSTRESS_STAT_OUT( s, m_nDeactivatePubRecord )
+            << CDSSTRESS_STAT_OUT( s, m_nActivatePubRecord )
+            << CDSSTRESS_STAT_OUT( s, m_nPubRecordCreated )
+            << CDSSTRESS_STAT_OUT( s, m_nPubRecordDeteted )
+            << CDSSTRESS_STAT_OUT( s, m_nPassiveWaitCall )
+            << CDSSTRESS_STAT_OUT( s, m_nPassiveWaitIteration )
+            << CDSSTRESS_STAT_OUT( s, m_nPassiveWaitWakeup )
+            << CDSSTRESS_STAT_OUT( s, m_nInvokeExclusive )
+            << CDSSTRESS_STAT_OUT( s, m_nWakeupByNotifying )
+            << CDSSTRESS_STAT_OUT( s, m_nPassiveToCombiner );
+    }
+
+} // namespace cds_test
+
+#endif // #ifndef CDSTEST_STAT_SPLITLIST_OUT_H
index b776a60191272f7637b6df5ddf20b0636bac0165..1df733d17b31f028f867d6146616b567f77f3471 100644 (file)
@@ -61,6 +61,7 @@
 #include <cds_test/stress_test.h>
 #include <cds_test/stat_ellenbintree_out.h>
 #include <cds_test/stat_skiplist_out.h>
+#include <cds_test/stat_flat_combining_out.h>
 
 namespace pqueue {
     namespace cc = cds::container;
@@ -599,16 +600,6 @@ namespace pqueue {
         typedef details::StdPQueue< Value, std::deque<Value>,  std::mutex >  StdPQueue_deque_mutex;
     };
 
-
-    //template <typename Stat>
-    //static inline void check_statistics( Stat const& /*s*/ )
-    //{}
-
-    //static inline void check_statistics( cds::container::ellen_bintree::stat<> const& s )
-    //{
-    //    CPPUNIT_CHECK_CURRENT( s.m_nInternalNodeCreated.get() == s.m_nInternalNodeDeleted.get() );
-    //    CPPUNIT_CHECK_CURRENT( s.m_nUpdateDescCreated.get() == s.m_nUpdateDescDeleted.get() );
-    //}
 }   // namespace pqueue
 
 
@@ -633,16 +624,7 @@ namespace cds_test {
             << CDSSTRESS_STAT_OUT( s, m_nPushMove )
             << CDSSTRESS_STAT_OUT( s, m_nPop )
             << CDSSTRESS_STAT_OUT( s, m_nFailedPop )
-            << CDSSTRESS_STAT_OUT_( "combining_factor", s.combining_factor() )
-            << CDSSTRESS_STAT_OUT( s, m_nOperationCount )
-            << CDSSTRESS_STAT_OUT( s, m_nCombiningCount )
-            << CDSSTRESS_STAT_OUT( s, m_nCompactPublicationList )
-            << CDSSTRESS_STAT_OUT( s, m_nDeactivatePubRecord )
-            << CDSSTRESS_STAT_OUT( s, m_nActivatePubRecord )
-            << CDSSTRESS_STAT_OUT( s, m_nPubRecordCreated )
-            << CDSSTRESS_STAT_OUT( s, m_nPubRecordDeteted )
-            << CDSSTRESS_STAT_OUT( s, m_nAcquirePubRecCount )
-            << CDSSTRESS_STAT_OUT( s, m_nReleasePubRecCount );
+            << static_cast<cds::algo::flat_combining::stat<> const&>(s);
     }
 
     static inline property_stream& operator <<( property_stream& o, cds::container::mspriority_queue::empty_stat const& /*s*/ )
index 35d705ede8c733b51455066d6a24eb46fa7fd90b..ea229eaea9173d574cbebbccbbe3216fdc72e5f4 100644 (file)
@@ -359,76 +359,6 @@ namespace {
 
             propout() << q.statistics();
         }
-
-/*
-        template <typename Queue>
-        void test( Queue& q )
-        {
-            value_array<typename Queue::value_type> arrValue( s_nQueueSize );
-            {
-                {
-                    Queue q;
-                    test_with( q, arrValue, 0, 0 );
-                }
-                Queue::gc::force_dispose();
-            }
-        }
-
-        template <typename Queue>
-        void test_boost()
-        {
-            value_array<typename Queue::value_type> arrValue( s_nQueueSize );
-            {
-                Queue q;
-                test_with(q, arrValue, 0, 0);
-            }
-        }
-
-        template <typename Queue>
-        void test_bounded()
-        {
-            value_array<typename Queue::value_type> arrValue( s_nQueueSize );
-            Queue q;
-            test_with(q, arrValue, 0, 0);
-        }
-
-        template <typename Queue>
-        void test_fcqueue()
-        {
-            value_array<typename Queue::value_type> arrValue( s_nQueueSize );
-            CPPUNIT_MSG( "Combining pass count: " << s_nFCPassCount << ", compact factor: " << s_nFCCompactFactor );
-            Queue q( s_nFCCompactFactor, s_nFCPassCount );
-            test_with(q, arrValue, 0, 0);
-        }
-
-        template <typename Queue>
-        void test_segmented()
-        {
-            value_array<typename Queue::value_type> arrValue( s_nQueueSize );
-            for ( size_t nSegmentSize = 4; nSegmentSize <= 256; nSegmentSize *= 4 ) {
-                CPPUNIT_MSG( "Segment size: " << nSegmentSize );
-                {
-                    Queue q( nSegmentSize );
-                    test_with( q, arrValue, nSegmentSize * 2, nSegmentSize );
-                }
-                Queue::gc::force_dispose();
-            }
-        }
-
-        template <typename Queue>
-        void test_spqueue()
-        {
-            value_array<typename Queue::value_type> arrValue( s_nQueueSize );
-            for ( size_t nArraySize = 2; nArraySize <= 64; nArraySize *= 2 ) {
-                CPPUNIT_MSG( "Array size: " << nArraySize );
-                {
-                    Queue q( nArraySize );
-                    test_with( q, arrValue, 0, 0 );
-                }
-                Queue::gc::force_dispose();
-            }
-        }
-*/
     };
 
 #define CDSSTRESS_QUEUE_F( QueueType, NodeType ) \
@@ -489,6 +419,12 @@ namespace {
     CDSSTRESS_QUEUE_F(FCQueue_list_delay2_elimination_stat,     boost::intrusive::list_base_hook<> )
     CDSSTRESS_QUEUE_F(FCQueue_list_expbackoff_elimination,      boost::intrusive::list_base_hook<> )
     CDSSTRESS_QUEUE_F(FCQueue_list_expbackoff_elimination_stat, boost::intrusive::list_base_hook<> )
+    CDSSTRESS_QUEUE_F(FCQueue_list_wait_ss,                     boost::intrusive::list_base_hook<> )
+    CDSSTRESS_QUEUE_F(FCQueue_list_wait_ss_stat,                boost::intrusive::list_base_hook<> )
+    CDSSTRESS_QUEUE_F(FCQueue_list_wait_sm,                     boost::intrusive::list_base_hook<> )
+    CDSSTRESS_QUEUE_F(FCQueue_list_wait_sm_stat,                boost::intrusive::list_base_hook<> )
+    CDSSTRESS_QUEUE_F(FCQueue_list_wait_mm,                     boost::intrusive::list_base_hook<> )
+    CDSSTRESS_QUEUE_F(FCQueue_list_wait_mm_stat,                boost::intrusive::list_base_hook<> )
 #undef CDSSTRESS_QUEUE_F
 
 
index fd380311abb2e7a6c0be5c1aca014a4aff0ea115..c213343150f10ef9ba01185706bd73f2d3852d23 100644 (file)
@@ -46,6 +46,7 @@
 #include <boost/intrusive/slist.hpp>
 
 #include <cds_test/stress_test.h>
+#include <cds_test/stat_flat_combining_out.h>
 #include "print_stat.h"
 
 namespace queue {
@@ -412,18 +413,18 @@ namespace queue {
         // FCQueue
         class traits_FCQueue_delay2:
             public cds::intrusive::fcqueue::make_traits<
-                cds::opt::back_off< cds::backoff::delay_of<2> >
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::backoff< cds::backoff::delay_of<2>>>
             >::type
         {};
         class traits_FCQueue_delay2_elimination:
             public cds::intrusive::fcqueue::make_traits<
-                cds::opt::back_off< cds::backoff::delay_of<2> >
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::backoff< cds::backoff::delay_of<2>>>
                 ,cds::opt::enable_elimination< true >
             >::type
         {};
         class traits_FCQueue_delay2_elimination_stat:
             public cds::intrusive::fcqueue::make_traits<
-                cds::opt::back_off< cds::backoff::delay_of<2> >
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::backoff< cds::backoff::delay_of<2>>>
                 ,cds::opt::stat< cds::intrusive::fcqueue::stat<> >
                 ,cds::opt::enable_elimination< true >
             >::type
@@ -442,11 +443,45 @@ namespace queue {
             >::type
         {};
 
+        class traits_FCQueue_wait_ss:
+            public cds::intrusive::fcqueue::make_traits<
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_single_condvar<>>
+            >::type
+        {};
+        struct traits_FCQueue_wait_ss_stat: traits_FCQueue_wait_ss
+        {
+            typedef cds::intrusive::fcqueue::stat<> stat;
+        };
+        class traits_FCQueue_wait_sm:
+            public cds::intrusive::fcqueue::make_traits<
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_multi_condvar<>>
+            >::type
+        {};
+        struct traits_FCQueue_wait_sm_stat: traits_FCQueue_wait_sm
+        {
+            typedef cds::intrusive::fcqueue::stat<> stat;
+        };
+        class traits_FCQueue_wait_mm:
+            public cds::intrusive::fcqueue::make_traits<
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::multi_mutex_multi_condvar<>>
+            >::type
+        {};
+        struct traits_FCQueue_wait_mm_stat: traits_FCQueue_wait_mm
+        {
+            typedef cds::intrusive::fcqueue::stat<> stat;
+        };
+
         typedef cds::intrusive::FCQueue< T, boost::intrusive::list<T>, traits_FCQueue_delay2 > FCQueue_list_delay2;
         typedef cds::intrusive::FCQueue< T, boost::intrusive::list<T>, traits_FCQueue_delay2_elimination > FCQueue_list_delay2_elimination;
         typedef cds::intrusive::FCQueue< T, boost::intrusive::list<T>, traits_FCQueue_delay2_elimination_stat > FCQueue_list_delay2_elimination_stat;
         typedef cds::intrusive::FCQueue< T, boost::intrusive::list<T>, traits_FCQueue_expbackoff_elimination > FCQueue_list_expbackoff_elimination;
         typedef cds::intrusive::FCQueue< T, boost::intrusive::list<T>, traits_FCQueue_expbackoff_elimination_stat > FCQueue_list_expbackoff_elimination_stat;
+        typedef cds::intrusive::FCQueue< T, boost::intrusive::list<T>, traits_FCQueue_wait_ss > FCQueue_list_wait_ss;
+        typedef cds::intrusive::FCQueue< T, boost::intrusive::list<T>, traits_FCQueue_wait_ss_stat > FCQueue_list_wait_ss_stat;
+        typedef cds::intrusive::FCQueue< T, boost::intrusive::list<T>, traits_FCQueue_wait_sm > FCQueue_list_wait_sm;
+        typedef cds::intrusive::FCQueue< T, boost::intrusive::list<T>, traits_FCQueue_wait_sm_stat > FCQueue_list_wait_sm_stat;
+        typedef cds::intrusive::FCQueue< T, boost::intrusive::list<T>, traits_FCQueue_wait_mm > FCQueue_list_wait_mm;
+        typedef cds::intrusive::FCQueue< T, boost::intrusive::list<T>, traits_FCQueue_wait_mm_stat > FCQueue_list_wait_mm_stat;
 
         // SegmentedQueue
         class traits_SegmentedQueue_spin_stat:
@@ -508,16 +543,7 @@ namespace cds_test {
             << CDSSTRESS_STAT_OUT( s, m_nDequeue )
             << CDSSTRESS_STAT_OUT( s, m_nFailedDeq )
             << CDSSTRESS_STAT_OUT( s, m_nCollided )
-            << CDSSTRESS_STAT_OUT_( "combining_factor", s.combining_factor() )
-            << CDSSTRESS_STAT_OUT( s, m_nOperationCount )
-            << CDSSTRESS_STAT_OUT( s, m_nCombiningCount )
-            << CDSSTRESS_STAT_OUT( s, m_nCompactPublicationList )
-            << CDSSTRESS_STAT_OUT( s, m_nDeactivatePubRecord )
-            << CDSSTRESS_STAT_OUT( s, m_nActivatePubRecord )
-            << CDSSTRESS_STAT_OUT( s, m_nPubRecordCreated )
-            << CDSSTRESS_STAT_OUT( s, m_nPubRecordDeteted )
-            << CDSSTRESS_STAT_OUT( s, m_nAcquirePubRecCount )
-            << CDSSTRESS_STAT_OUT( s, m_nReleasePubRecCount );
+            << static_cast<cds::algo::flat_combining::stat<> const&>(s);
     }
 
     static inline property_stream& operator <<( property_stream& o, cds::intrusive::fcqueue::empty_stat const& /*s*/ )
index 0db165ec50d24152c3f8fe2b6fa0ae195eafbd1f..c4178fef20214b4fb0bcfe8d6a61b6d6593b54a7 100644 (file)
@@ -52,6 +52,7 @@
 #include <boost/container/deque.hpp>
 
 #include <cds_test/stress_test.h>
+#include <cds_test/stat_flat_combining_out.h>
 #include "print_stat.h"
 
 namespace queue {
@@ -417,12 +418,39 @@ namespace queue {
         typedef cds::container::RWQueue< Value, traits_RWQueue_mutex > RWQueue_mutex;
 
         // FCQueue
-        class traits_FCQueue_elimination:
+        struct traits_FCQueue_single_mutex_single_condvar:
+            public cds::container::fcqueue::make_traits<
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_single_condvar<>>
+            >::type
+        {};
+        struct traits_FCQueue_single_mutex_single_condvar_stat: traits_FCQueue_single_mutex_single_condvar
+        {
+            typedef cds::container::fcqueue::stat<> stat;
+        };
+        struct traits_FCQueue_single_mutex_multi_condvar:
+            public cds::container::fcqueue::make_traits<
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_multi_condvar<>>
+            >::type
+        {};
+        struct traits_FCQueue_single_mutex_multi_condvar_stat: traits_FCQueue_single_mutex_multi_condvar
+        {
+            typedef cds::container::fcqueue::stat<> stat;
+        };
+        struct traits_FCQueue_multi_mutex_multi_condvar:
+            public cds::container::fcqueue::make_traits<
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::multi_mutex_multi_condvar<>>
+            >::type
+        {};
+        struct traits_FCQueue_multi_mutex_multi_condvar_stat: traits_FCQueue_multi_mutex_multi_condvar
+        {
+            typedef cds::container::fcqueue::stat<> stat;
+        };
+        struct traits_FCQueue_elimination:
             public cds::container::fcqueue::make_traits<
                 cds::opt::enable_elimination< true >
             >::type
         {};
-        class traits_FCQueue_elimination_stat:
+        struct traits_FCQueue_elimination_stat:
             public cds::container::fcqueue::make_traits<
                 cds::opt::enable_elimination< true >
                 ,cds::opt::stat< cds::container::fcqueue::stat<> >
@@ -430,10 +458,24 @@ namespace queue {
         {};
 
         typedef cds::container::FCQueue< Value > FCQueue_deque;
+        typedef cds::container::FCQueue< Value, std::queue<Value>, traits_FCQueue_single_mutex_single_condvar> FCQueue_deque_wait_ss;
+        typedef cds::container::FCQueue< Value, std::queue<Value>, traits_FCQueue_single_mutex_single_condvar_stat> FCQueue_deque_wait_ss_stat;
+        typedef cds::container::FCQueue< Value, std::queue<Value>, traits_FCQueue_single_mutex_multi_condvar> FCQueue_deque_wait_sm;
+        typedef cds::container::FCQueue< Value, std::queue<Value>, traits_FCQueue_single_mutex_multi_condvar_stat> FCQueue_deque_wait_sm_stat;
+        typedef cds::container::FCQueue< Value, std::queue<Value>, traits_FCQueue_multi_mutex_multi_condvar> FCQueue_deque_wait_mm;
+        typedef cds::container::FCQueue< Value, std::queue<Value>, traits_FCQueue_multi_mutex_multi_condvar_stat> FCQueue_deque_wait_mm_stat;
+
         typedef cds::container::FCQueue< Value, std::queue<Value>, traits_FCQueue_elimination > FCQueue_deque_elimination;
         typedef cds::container::FCQueue< Value, std::queue<Value>, traits_FCQueue_elimination_stat > FCQueue_deque_elimination_stat;
 
-        typedef cds::container::FCQueue< Value, std::queue<Value, std::list<Value> > > FCQueue_list;
+        typedef cds::container::FCQueue< Value, std::queue<Value, std::list<Value> >> FCQueue_list;
+        typedef cds::container::FCQueue< Value, std::queue<Value, std::list<Value>>, traits_FCQueue_single_mutex_single_condvar> FCQueue_list_wait_ss;
+        typedef cds::container::FCQueue< Value, std::queue<Value, std::list<Value>>, traits_FCQueue_single_mutex_single_condvar_stat> FCQueue_list_wait_ss_stat;
+        typedef cds::container::FCQueue< Value, std::queue<Value, std::list<Value>>, traits_FCQueue_single_mutex_multi_condvar> FCQueue_list_wait_sm;
+        typedef cds::container::FCQueue< Value, std::queue<Value, std::list<Value>>, traits_FCQueue_single_mutex_multi_condvar_stat> FCQueue_list_wait_sm_stat;
+        typedef cds::container::FCQueue< Value, std::queue<Value, std::list<Value>>, traits_FCQueue_multi_mutex_multi_condvar> FCQueue_list_wait_mm;
+        typedef cds::container::FCQueue< Value, std::queue<Value, std::list<Value>>, traits_FCQueue_multi_mutex_multi_condvar_stat> FCQueue_list_wait_mm_stat;
+
         typedef cds::container::FCQueue< Value, std::queue<Value, std::list<Value> >, traits_FCQueue_elimination > FCQueue_list_elimination;
         typedef cds::container::FCQueue< Value, std::queue<Value, std::list<Value> >, traits_FCQueue_elimination_stat > FCQueue_list_elimination_stat;
 
@@ -461,9 +503,40 @@ namespace queue {
             >::type
         {};
 
+        struct traits_FCDeque_wait_ss: cds::container::fcdeque::traits
+        {
+            typedef cds::algo::flat_combining::wait_strategy::single_mutex_single_condvar<> wait_strategy;
+        };
+        struct traits_FCDeque_wait_ss_stat: traits_FCDeque_wait_ss
+        {
+            typedef cds::container::fcdeque::stat<> stat;
+        };
+        struct traits_FCDeque_wait_sm: cds::container::fcdeque::traits
+        {
+            typedef cds::algo::flat_combining::wait_strategy::single_mutex_multi_condvar<> wait_strategy;
+        };
+        struct traits_FCDeque_wait_sm_stat: traits_FCDeque_wait_sm
+        {
+            typedef cds::container::fcdeque::stat<> stat;
+        };
+        struct traits_FCDeque_wait_mm: cds::container::fcdeque::traits
+        {
+            typedef cds::algo::flat_combining::wait_strategy::multi_mutex_multi_condvar<> wait_strategy;
+        };
+        struct traits_FCDeque_wait_mm_stat: traits_FCDeque_wait_mm
+        {
+            typedef cds::container::fcdeque::stat<> stat;
+        };
+
         typedef details::FCDequeL< Value > FCDequeL_default;
         typedef details::FCDequeL< Value, traits_FCDeque_mutex > FCDequeL_mutex;
         typedef details::FCDequeL< Value, traits_FCDeque_stat > FCDequeL_stat;
+        typedef details::FCDequeL< Value, traits_FCDeque_wait_ss > FCDequeL_wait_ss;
+        typedef details::FCDequeL< Value, traits_FCDeque_wait_ss_stat > FCDequeL_wait_ss_stat;
+        typedef details::FCDequeL< Value, traits_FCDeque_wait_sm > FCDequeL_wait_sm;
+        typedef details::FCDequeL< Value, traits_FCDeque_wait_sm_stat > FCDequeL_wait_sm_stat;
+        typedef details::FCDequeL< Value, traits_FCDeque_wait_mm > FCDequeL_wait_mm;
+        typedef details::FCDequeL< Value, traits_FCDeque_wait_mm_stat > FCDequeL_wait_mm_stat;
         typedef details::FCDequeL< Value, traits_FCDeque_elimination > FCDequeL_elimination;
         typedef details::FCDequeL< Value, traits_FCDeque_elimination_stat > FCDequeL_elimination_stat;
 
@@ -475,6 +548,12 @@ namespace queue {
         typedef details::FCDequeR< Value > FCDequeR_default;
         typedef details::FCDequeR< Value, traits_FCDeque_mutex > FCDequeR_mutex;
         typedef details::FCDequeR< Value, traits_FCDeque_stat > FCDequeR_stat;
+        typedef details::FCDequeR< Value, traits_FCDeque_wait_ss > FCDequeR_wait_ss;
+        typedef details::FCDequeR< Value, traits_FCDeque_wait_ss_stat > FCDequeR_wait_ss_stat;
+        typedef details::FCDequeR< Value, traits_FCDeque_wait_sm > FCDequeR_wait_sm;
+        typedef details::FCDequeR< Value, traits_FCDeque_wait_sm_stat > FCDequeR_wait_sm_stat;
+        typedef details::FCDequeR< Value, traits_FCDeque_wait_mm > FCDequeR_wait_mm;
+        typedef details::FCDequeR< Value, traits_FCDeque_wait_mm_stat > FCDequeR_wait_mm_stat;
         typedef details::FCDequeR< Value, traits_FCDeque_elimination > FCDequeR_elimination;
         typedef details::FCDequeR< Value, traits_FCDeque_elimination_stat > FCDequeR_elimination_stat;
 
@@ -483,6 +562,7 @@ namespace queue {
         typedef details::FCDequeR< Value, traits_FCDeque_elimination, boost::container::deque<Value> > FCDequeR_boost_elimination;
         typedef details::FCDequeR< Value, traits_FCDeque_elimination_stat, boost::container::deque<Value> > FCDequeR_boost_elimination_stat;
 
+        // STL
         typedef StdQueue_deque<Value>               StdQueue_deque_Spinlock;
         typedef StdQueue_list<Value>                StdQueue_list_Spinlock;
         typedef StdQueue_deque<Value, std::mutex>   StdQueue_deque_Mutex;
@@ -553,16 +633,7 @@ namespace cds_test {
                 << CDSSTRESS_STAT_OUT( s, m_nDequeue )
                 << CDSSTRESS_STAT_OUT( s, m_nFailedDeq )
                 << CDSSTRESS_STAT_OUT( s, m_nCollided )
-                << CDSSTRESS_STAT_OUT_( "combining_factor", s.combining_factor() )
-                << CDSSTRESS_STAT_OUT( s, m_nOperationCount )
-                << CDSSTRESS_STAT_OUT( s, m_nCombiningCount )
-                << CDSSTRESS_STAT_OUT( s, m_nCompactPublicationList )
-                << CDSSTRESS_STAT_OUT( s, m_nDeactivatePubRecord )
-                << CDSSTRESS_STAT_OUT( s, m_nActivatePubRecord )
-                << CDSSTRESS_STAT_OUT( s, m_nPubRecordCreated )
-                << CDSSTRESS_STAT_OUT( s, m_nPubRecordDeteted )
-                << CDSSTRESS_STAT_OUT( s, m_nAcquirePubRecCount )
-                << CDSSTRESS_STAT_OUT( s, m_nReleasePubRecCount );
+                << static_cast<cds::algo::flat_combining::stat<> const&>(s);
     }
 
     static inline property_stream& operator <<( property_stream& o, cds::container::fcqueue::empty_stat const& /*s*/ )
@@ -587,16 +658,7 @@ namespace cds_test {
             << CDSSTRESS_STAT_OUT( s, m_nPopBack )
             << CDSSTRESS_STAT_OUT( s, m_nFailedPopBack )
             << CDSSTRESS_STAT_OUT( s, m_nCollided )
-            << CDSSTRESS_STAT_OUT_( "combining_factor", s.combining_factor() )
-            << CDSSTRESS_STAT_OUT( s, m_nOperationCount )
-            << CDSSTRESS_STAT_OUT( s, m_nCombiningCount )
-            << CDSSTRESS_STAT_OUT( s, m_nCompactPublicationList )
-            << CDSSTRESS_STAT_OUT( s, m_nDeactivatePubRecord )
-            << CDSSTRESS_STAT_OUT( s, m_nActivatePubRecord )
-            << CDSSTRESS_STAT_OUT( s, m_nPubRecordCreated )
-            << CDSSTRESS_STAT_OUT( s, m_nPubRecordDeteted )
-            << CDSSTRESS_STAT_OUT( s, m_nAcquirePubRecCount )
-            << CDSSTRESS_STAT_OUT( s, m_nReleasePubRecCount );
+            << static_cast<cds::algo::flat_combining::stat<> const&>(s);
     }
 
 } // namespace cds_test
@@ -659,9 +721,21 @@ namespace cds_test {
 
 #define CDSSTRESS_FCQueue( test_fixture ) \
     CDSSTRESS_Queue_F( test_fixture, FCQueue_deque ) \
+    CDSSTRESS_Queue_F( test_fixture, FCQueue_deque_wait_ss ) \
+    CDSSTRESS_Queue_F( test_fixture, FCQueue_deque_wait_ss_stat ) \
+    CDSSTRESS_Queue_F( test_fixture, FCQueue_deque_wait_sm ) \
+    CDSSTRESS_Queue_F( test_fixture, FCQueue_deque_wait_sm_stat ) \
+    CDSSTRESS_Queue_F( test_fixture, FCQueue_deque_wait_mm ) \
+    CDSSTRESS_Queue_F( test_fixture, FCQueue_deque_wait_mm_stat ) \
     CDSSTRESS_Queue_F( test_fixture, FCQueue_deque_elimination ) \
     CDSSTRESS_Queue_F( test_fixture, FCQueue_deque_elimination_stat ) \
     CDSSTRESS_Queue_F( test_fixture, FCQueue_list ) \
+    CDSSTRESS_Queue_F( test_fixture, FCQueue_list_wait_ss ) \
+    CDSSTRESS_Queue_F( test_fixture, FCQueue_list_wait_ss_stat ) \
+    CDSSTRESS_Queue_F( test_fixture, FCQueue_list_wait_sm ) \
+    CDSSTRESS_Queue_F( test_fixture, FCQueue_list_wait_sm_stat ) \
+    CDSSTRESS_Queue_F( test_fixture, FCQueue_list_wait_mm ) \
+    CDSSTRESS_Queue_F( test_fixture, FCQueue_list_wait_mm_stat ) \
     CDSSTRESS_Queue_F( test_fixture, FCQueue_list_elimination ) \
     CDSSTRESS_Queue_F( test_fixture, FCQueue_list_elimination_stat )
 
@@ -669,6 +743,12 @@ namespace cds_test {
     CDSSTRESS_Queue_F( test_fixture, FCDequeL_default ) \
     CDSSTRESS_Queue_F( test_fixture, FCDequeL_mutex ) \
     CDSSTRESS_Queue_F( test_fixture, FCDequeL_stat ) \
+    CDSSTRESS_Queue_F( test_fixture, FCDequeL_wait_ss ) \
+    CDSSTRESS_Queue_F( test_fixture, FCDequeL_wait_ss_stat ) \
+    CDSSTRESS_Queue_F( test_fixture, FCDequeL_wait_sm ) \
+    CDSSTRESS_Queue_F( test_fixture, FCDequeL_wait_sm_stat ) \
+    CDSSTRESS_Queue_F( test_fixture, FCDequeL_wait_mm ) \
+    CDSSTRESS_Queue_F( test_fixture, FCDequeL_wait_mm_stat ) \
     CDSSTRESS_Queue_F( test_fixture, FCDequeL_elimination ) \
     CDSSTRESS_Queue_F( test_fixture, FCDequeL_elimination_stat ) \
     CDSSTRESS_Queue_F( test_fixture, FCDequeL_boost ) \
@@ -678,6 +758,12 @@ namespace cds_test {
     CDSSTRESS_Queue_F( test_fixture, FCDequeR_default ) \
     CDSSTRESS_Queue_F( test_fixture, FCDequeR_mutex ) \
     CDSSTRESS_Queue_F( test_fixture, FCDequeR_stat ) \
+    CDSSTRESS_Queue_F( test_fixture, FCDequeR_wait_ss ) \
+    CDSSTRESS_Queue_F( test_fixture, FCDequeR_wait_ss_stat ) \
+    CDSSTRESS_Queue_F( test_fixture, FCDequeR_wait_sm ) \
+    CDSSTRESS_Queue_F( test_fixture, FCDequeR_wait_sm_stat ) \
+    CDSSTRESS_Queue_F( test_fixture, FCDequeR_wait_mm ) \
+    CDSSTRESS_Queue_F( test_fixture, FCDequeR_wait_mm_stat ) \
     CDSSTRESS_Queue_F( test_fixture, FCDequeR_elimination ) \
     CDSSTRESS_Queue_F( test_fixture, FCDequeR_elimination_stat ) \
     CDSSTRESS_Queue_F( test_fixture, FCDequeR_boost ) \
index ffcb39313e13463fbf63132ca1f939ec6cfd2e71..5740e2f96a8a9cf7e9faab81107cb61a179114f3 100644 (file)
@@ -45,6 +45,7 @@
 #include <boost/intrusive/list.hpp>
 
 #include <cds_test/stress_test.h>
+#include <cds_test/stat_flat_combining_out.h>
 
 
 namespace istack {
@@ -411,16 +412,7 @@ namespace cds_test {
             << CDSSTRESS_STAT_OUT( s, m_nPop )
             << CDSSTRESS_STAT_OUT( s, m_nFailedPop )
             << CDSSTRESS_STAT_OUT( s, m_nCollided )
-            << CDSSTRESS_STAT_OUT_( "combining_factor", s.combining_factor() )
-            << CDSSTRESS_STAT_OUT( s, m_nOperationCount )
-            << CDSSTRESS_STAT_OUT( s, m_nCombiningCount )
-            << CDSSTRESS_STAT_OUT( s, m_nCompactPublicationList )
-            << CDSSTRESS_STAT_OUT( s, m_nDeactivatePubRecord )
-            << CDSSTRESS_STAT_OUT( s, m_nActivatePubRecord )
-            << CDSSTRESS_STAT_OUT( s, m_nPubRecordCreated )
-            << CDSSTRESS_STAT_OUT( s, m_nPubRecordDeteted )
-            << CDSSTRESS_STAT_OUT( s, m_nAcquirePubRecCount )
-            << CDSSTRESS_STAT_OUT( s, m_nReleasePubRecCount );
+            << static_cast< cds::algo::flat_combining::stat<> const&>( s );
     }
 
 } // namespace cds_test
index 39d476d47f2813d42265fecfa366b2097df8bf6d..6a14e748a7f477a1d329179f5f21807e3bcd9241 100644 (file)
@@ -45,6 +45,7 @@
 #include <vector>
 
 #include <cds_test/stress_test.h>
+#include <cds_test/stat_flat_combining_out.h>
 
 namespace stack {
 
@@ -455,16 +456,7 @@ namespace cds_test {
             << CDSSTRESS_STAT_OUT( s, m_nPop )
             << CDSSTRESS_STAT_OUT( s, m_nFailedPop )
             << CDSSTRESS_STAT_OUT( s, m_nCollided )
-            << CDSSTRESS_STAT_OUT_( "combining_factor", s.combining_factor())
-            << CDSSTRESS_STAT_OUT( s, m_nOperationCount )
-            << CDSSTRESS_STAT_OUT( s, m_nCombiningCount )
-            << CDSSTRESS_STAT_OUT( s, m_nCompactPublicationList )
-            << CDSSTRESS_STAT_OUT( s, m_nDeactivatePubRecord )
-            << CDSSTRESS_STAT_OUT( s, m_nActivatePubRecord )
-            << CDSSTRESS_STAT_OUT( s, m_nPubRecordCreated )
-            << CDSSTRESS_STAT_OUT( s, m_nPubRecordDeteted )
-            << CDSSTRESS_STAT_OUT( s, m_nAcquirePubRecCount )
-            << CDSSTRESS_STAT_OUT( s, m_nReleasePubRecCount );
+            << static_cast<cds::algo::flat_combining::stat<> const&>( s );
     }
 
     static inline property_stream& operator <<( property_stream& o, cds::container::fcdeque::empty_stat const& /*s*/ )
@@ -484,16 +476,7 @@ namespace cds_test {
             << CDSSTRESS_STAT_OUT( s, m_nPopBack )
             << CDSSTRESS_STAT_OUT( s, m_nFailedPopBack )
             << CDSSTRESS_STAT_OUT( s, m_nCollided )
-            << CDSSTRESS_STAT_OUT_( "combining_factor", s.combining_factor() )
-            << CDSSTRESS_STAT_OUT( s, m_nOperationCount )
-            << CDSSTRESS_STAT_OUT( s, m_nCombiningCount )
-            << CDSSTRESS_STAT_OUT( s, m_nCompactPublicationList )
-            << CDSSTRESS_STAT_OUT( s, m_nDeactivatePubRecord )
-            << CDSSTRESS_STAT_OUT( s, m_nActivatePubRecord )
-            << CDSSTRESS_STAT_OUT( s, m_nPubRecordCreated )
-            << CDSSTRESS_STAT_OUT( s, m_nPubRecordDeteted )
-            << CDSSTRESS_STAT_OUT( s, m_nAcquirePubRecCount )
-            << CDSSTRESS_STAT_OUT( s, m_nReleasePubRecCount );
+            << static_cast<cds::algo::flat_combining::stat<> const&>(s);
     }
 } // namespace cds_test
 
index 3c6ad1ab5ced905bf2590283ae265451e3f2b438..154d0bbc0438715e8a6b40d9dc577b4aa929a60a 100644 (file)
@@ -114,6 +114,30 @@ namespace {
         test( dq );
     }
 
+    TEST_F( FCDeque, std_empty_wait_strategy )
+    {
+        typedef cds::container::FCDeque<int, std::deque<int>,
+            cds::container::fcdeque::make_traits<
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::empty >
+            >::type
+        > deque_type;
+
+        deque_type dq;
+        test( dq );
+    }
+
+    TEST_F( FCDeque, std_multi_mutex_multi_condvar )
+    {
+        typedef cds::container::FCDeque<int, std::deque<int>,
+            cds::container::fcdeque::make_traits<
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::multi_mutex_multi_condvar<>>
+            >::type
+        > deque_type;
+
+        deque_type dq;
+        test( dq );
+    }
+
     TEST_F( FCDeque, std_elimination )
     {
         typedef cds::container::FCDeque<int, std::deque<int>,
@@ -126,6 +150,19 @@ namespace {
         test( dq );
     }
 
+    TEST_F( FCDeque, std_elimination_single_mutex_single_condvar )
+    {
+        typedef cds::container::FCDeque<int, std::deque<int>,
+            cds::container::fcdeque::make_traits<
+                cds::opt::enable_elimination< true >
+                , cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_single_condvar<3>>
+            >::type
+        > deque_type;
+
+        deque_type dq;
+        test( dq );
+    }
+
     TEST_F( FCDeque, std_statistics )
     {
         typedef cds::container::FCDeque<int, std::deque<int>,
@@ -138,6 +175,19 @@ namespace {
         test( dq );
     }
 
+    TEST_F( FCDeque, std_stat_single_mutex_multi_condvar )
+    {
+        typedef cds::container::FCDeque<int, std::deque<int>,
+            cds::container::fcdeque::make_traits<
+                cds::opt::stat< cds::container::fcdeque::stat<> >
+                , cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_multi_condvar<2>>
+            >::type
+        > deque_type;
+
+        deque_type dq;
+        test( dq );
+    }
+
     TEST_F( FCDeque, std_mutex )
     {
         struct deque_traits : public
@@ -161,6 +211,30 @@ namespace {
         test( dq );
     }
 
+    TEST_F( FCDeque, boost_empty_wait_strategy )
+    {
+        typedef cds::container::FCDeque<int, boost::container::deque<int>,
+            cds::container::fcdeque::make_traits<
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::empty >
+            >::type
+        > deque_type;
+
+        deque_type dq;
+        test( dq );
+    }
+
+    TEST_F( FCDeque, boost_single_mutex_single_condvar )
+    {
+        typedef cds::container::FCDeque<int, boost::container::deque<int>,
+            cds::container::fcdeque::make_traits<
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_single_condvar<>>
+            >::type
+        > deque_type;
+
+        deque_type dq;
+        test( dq );
+    }
+
     TEST_F( FCDeque, boost_elimination )
     {
         typedef cds::container::FCDeque<int, boost::container::deque<int>,
@@ -173,6 +247,19 @@ namespace {
         test( dq );
     }
 
+    TEST_F( FCDeque, boost_elimination_single_mutex_multi_condvar )
+    {
+        typedef cds::container::FCDeque<int, boost::container::deque<int>,
+            cds::container::fcdeque::make_traits<
+                cds::opt::enable_elimination< true >
+                ,cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_multi_condvar<5>>
+            >::type
+        > deque_type;
+
+        deque_type dq;
+        test( dq );
+    }
+
     TEST_F( FCDeque, boost_statistics )
     {
         typedef cds::container::FCDeque<int, boost::container::deque<int>,
@@ -198,4 +285,18 @@ namespace {
         test( dq );
     }
 
+    TEST_F( FCDeque, boost_mutex_multi_mutex_multi_condvar )
+    {
+        typedef cds::container::FCDeque<int, boost::container::deque<int>,
+            cds::container::fcdeque::make_traits<
+                cds::opt::enable_elimination< true >
+                , cds::opt::lock_type< std::mutex >
+                , cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::multi_mutex_multi_condvar<>>
+            >::type
+        > deque_type;
+
+        deque_type dq;
+        test( dq );
+    }
+
 } // namespace
index 91044795a052865e746b07529fbe36b0dac08398..cb51b44527bb234d2f75326ae22b6bb7505acf34 100644 (file)
@@ -68,6 +68,63 @@ namespace cds_test {
         test( pq );
     }
 
+    TEST_F( FCPQueue, stable_vector_empty_wait_strategy )
+    {
+        typedef cds::container::FCPriorityQueue<
+            value_type
+            ,std::priority_queue<
+                value_type
+                ,boost::container::stable_vector<value_type>
+                ,less
+            >
+            ,cds::container::fcpqueue::make_traits<
+                cds::opt::stat< cds::container::fcpqueue::stat<> >
+                , cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::empty >
+            >::type
+        > pqueue_type;
+
+        pqueue_type pq;
+        test( pq );
+    }
+
+    TEST_F( FCPQueue, stable_vector_single_mutex_single_condvar )
+    {
+        typedef cds::container::FCPriorityQueue<
+            value_type
+            ,std::priority_queue<
+                value_type
+                ,boost::container::stable_vector<value_type>
+                ,less
+            >
+            ,cds::container::fcpqueue::make_traits<
+                cds::opt::stat< cds::container::fcpqueue::stat<> >
+                , cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_single_condvar<>>
+            >::type
+        > pqueue_type;
+
+        pqueue_type pq;
+        test( pq );
+    }
+
+    TEST_F( FCPQueue, stable_vector_single_mutex_multi_condvar )
+    {
+        typedef cds::container::FCPriorityQueue<
+            value_type
+            ,std::priority_queue<
+                value_type
+                ,boost::container::stable_vector<value_type>
+                ,less
+            >
+            ,cds::container::fcpqueue::make_traits<
+                cds::opt::stat< cds::container::fcpqueue::stat<> >
+                , cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_multi_condvar<4>>
+            >::type
+        > pqueue_type;
+
+        pqueue_type pq;
+        test( pq );
+    }
+
     TEST_F( FCPQueue, boost_deque )
     {
         typedef cds::container::FCPriorityQueue<
@@ -101,4 +158,42 @@ namespace cds_test {
         test( pq );
     }
 
+    TEST_F( FCPQueue, boost_deque_empty_wait_strategy )
+    {
+        typedef cds::container::FCPriorityQueue<
+            value_type
+            ,std::priority_queue<
+                value_type
+                ,boost::container::deque<value_type>
+                ,less
+            >
+            ,cds::container::fcpqueue::make_traits<
+                cds::opt::stat< cds::container::fcpqueue::stat<> >
+                ,cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::empty >
+            >::type
+        > pqueue_type;
+
+        pqueue_type pq;
+        test( pq );
+    }
+
+    TEST_F( FCPQueue, boost_deque_multi_mutex_multi_condvar )
+    {
+        typedef cds::container::FCPriorityQueue<
+            value_type
+            ,std::priority_queue<
+                value_type
+                ,boost::container::deque<value_type>
+                ,less
+            >
+            ,cds::container::fcpqueue::make_traits<
+                cds::opt::stat< cds::container::fcpqueue::stat<> >
+                ,cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::multi_mutex_multi_condvar<6>>
+            >::type
+        > pqueue_type;
+
+        pqueue_type pq;
+        test( pq );
+    }
+
 } // namespace cds_test
index 0f7e0b9369fbfb8bd515e6b82ddeec08c3159f87..1da9be981e3c99449bc7b9d5f3f2524660de855b 100644 (file)
@@ -67,6 +67,63 @@ namespace cds_test {
         test( pq );
     }
 
+    TEST_F( FCPQueue, deque_stat_single_mutex_single_condvar )
+    {
+        typedef cds::container::FCPriorityQueue<
+            value_type
+            ,std::priority_queue<
+                value_type
+                ,std::deque<value_type>
+                ,less
+            >
+            ,cds::container::fcpqueue::make_traits<
+                cds::opt::stat< cds::container::fcpqueue::stat<> >
+                , cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_single_condvar<>>
+            >::type
+        > pqueue_type;
+
+        pqueue_type pq;
+        test( pq );
+    }
+
+    TEST_F( FCPQueue, deque_empty_wait_strategy )
+    {
+        typedef cds::container::FCPriorityQueue<
+            value_type
+            ,std::priority_queue<
+                value_type
+                ,std::deque<value_type>
+                ,less
+            >
+            ,cds::container::fcpqueue::make_traits<
+                cds::opt::stat< cds::container::fcpqueue::stat<> >
+                , cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::empty >
+            >::type
+        > pqueue_type;
+
+        pqueue_type pq;
+        test( pq );
+    }
+
+    TEST_F( FCPQueue, deque_single_mutex_multi_condvar )
+    {
+        typedef cds::container::FCPriorityQueue<
+            value_type
+            ,std::priority_queue<
+                value_type
+                ,std::deque<value_type>
+                ,less
+            >
+            ,cds::container::fcpqueue::make_traits<
+                cds::opt::stat< cds::container::fcpqueue::stat<> >
+                , cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_multi_condvar<2>>
+            >::type
+        > pqueue_type;
+
+        pqueue_type pq;
+        test( pq );
+    }
+
     TEST_F( FCPQueue, deque_mutex )
     {
         typedef cds::container::FCPriorityQueue<
@@ -84,4 +141,22 @@ namespace cds_test {
         test( pq );
     }
 
+    TEST_F( FCPQueue, deque_multi_mutex_multi_condvar )
+    {
+        typedef cds::container::FCPriorityQueue<
+            value_type
+            ,std::priority_queue<
+                value_type
+                ,std::deque<value_type>
+            >
+            ,cds::container::fcpqueue::make_traits<
+                cds::opt::lock_type< std::mutex >
+                , cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::multi_mutex_multi_condvar<1000>>
+            >::type
+        > pqueue_type;
+
+        pqueue_type pq;
+        test( pq );
+    }
+
 } // namespace cds_test
index 6de4855a4bb5f703317af82cfc7e938c40131f22..b1c9ef0f95a847051863c49231c8af8398f43d8a 100644 (file)
@@ -40,6 +40,50 @@ namespace cds_test {
         test( pq );
     }
 
+    TEST_F( FCPQueue, vector_empty_wait_strategy )
+    {
+        struct pqueue_traits : public cds::container::fcpqueue::traits
+        {
+            typedef cds::container::fcpqueue::stat<> stat;
+            typedef cds::algo::flat_combining::wait_strategy::empty wait_strategy;
+        };
+
+        typedef cds::container::FCPriorityQueue<
+            value_type
+            , std::priority_queue<
+                value_type
+                , std::vector<value_type>
+                , less
+            >
+            , pqueue_traits
+        > pqueue_type;
+
+        pqueue_type pq;
+        test( pq );
+    }
+
+    TEST_F( FCPQueue, vector_multi_mutex_multi_condvar )
+    {
+        struct pqueue_traits : public cds::container::fcpqueue::traits
+        {
+            typedef cds::container::fcpqueue::stat<> stat;
+            typedef cds::algo::flat_combining::wait_strategy::multi_mutex_multi_condvar<> wait_strategy;
+        };
+
+        typedef cds::container::FCPriorityQueue<
+            value_type
+            , std::priority_queue<
+                value_type
+                , std::vector<value_type>
+                , less
+            >
+            , pqueue_traits
+        > pqueue_type;
+
+        pqueue_type pq;
+        test( pq );
+    }
+
     TEST_F( FCPQueue, vector_stat )
     {
         struct pqueue_traits : public cds::container::fcpqueue::traits
@@ -61,6 +105,28 @@ namespace cds_test {
         test( pq );
     }
 
+    TEST_F( FCPQueue, vector_stat_single_mutex_multi_condvar )
+    {
+        struct pqueue_traits : public cds::container::fcpqueue::traits
+        {
+            typedef cds::container::fcpqueue::stat<> stat;
+            typedef cds::algo::flat_combining::wait_strategy::single_mutex_multi_condvar<42> wait_strategy;
+        };
+
+        typedef cds::container::FCPriorityQueue<
+            value_type
+            , std::priority_queue<
+                value_type
+                , std::vector<value_type>
+                , less
+            >
+            , pqueue_traits
+        > pqueue_type;
+
+        pqueue_type pq;
+        test( pq );
+    }
+
     TEST_F( FCPQueue, vector_mutex )
     {
         typedef cds::container::FCPriorityQueue<
@@ -75,4 +141,19 @@ namespace cds_test {
         test( pq );
     }
 
+    TEST_F( FCPQueue, vector_single_mutex_single_condvar )
+    {
+        typedef cds::container::FCPriorityQueue<
+            value_type
+            ,std::priority_queue< value_type >
+            ,cds::container::fcpqueue::make_traits<
+                cds::opt::lock_type< std::mutex >
+                , cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_single_condvar<1000>>
+            >::type
+        > pqueue_type;
+
+        pqueue_type pq;
+        test( pq );
+    }
+
 } // namespace cds_test
index 7681ddeb6946bd0c5531e999e32ba41160cdae4f..99f7b0c32b65a3c6c3e3048670ff51444989f22c 100644 (file)
@@ -154,6 +154,30 @@ namespace {
         test_string( q );
     }
 
+    TEST_F( FCQueue, std_empty_wait_strategy )
+    {
+        typedef cds::container::FCQueue<int, std::queue< int, std::deque<int>>,
+            cds::container::fcqueue::make_traits<
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::empty >
+            >::type
+        > queue_type;
+
+        queue_type q;
+        test( q );
+    }
+
+    TEST_F( FCQueue, std_single_mutex_single_condvar )
+    {
+        typedef cds::container::FCQueue<int, std::queue< int, std::deque<int>>,
+            cds::container::fcqueue::make_traits<
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_single_condvar<>>
+            >::type
+        > queue_type;
+
+        queue_type q;
+        test( q );
+    }
+
     TEST_F( FCQueue, std_deque_elimination )
     {
         typedef cds::container::FCQueue<int, std::queue< int, std::deque<int>>,
@@ -166,6 +190,19 @@ namespace {
         test( q );
     }
 
+    TEST_F( FCQueue, std_deque_elimination_single_mutex_multi_condvar )
+    {
+        typedef cds::container::FCQueue<int, std::queue< int, std::deque<int>>,
+            cds::container::fcqueue::make_traits<
+                cds::opt::enable_elimination< true >
+                , cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_multi_condvar<2>>
+            >::type
+        > queue_type;
+
+        queue_type q;
+        test( q );
+    }
+
     TEST_F( FCQueue, std_deque_elimination_move )
     {
         typedef cds::container::FCQueue<std::string, std::queue< std::string, std::deque<std::string>>,
@@ -178,6 +215,19 @@ namespace {
         test_string( q );
     }
 
+    TEST_F( FCQueue, std_deque_elimination_move_multi_mutex_multi_condvar )
+    {
+        typedef cds::container::FCQueue<std::string, std::queue< std::string, std::deque<std::string>>,
+            cds::container::fcqueue::make_traits<
+                cds::opt::enable_elimination< true >
+                , cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::multi_mutex_multi_condvar<>>
+            >::type
+        > queue_type;
+
+        queue_type q;
+        test_string( q );
+    }
+
     TEST_F( FCQueue, std_deque_mutex )
     {
         typedef cds::container::FCQueue<int, std::queue< int, std::deque<int>>,
@@ -206,6 +256,30 @@ namespace {
         test_string( q );
     }
 
+    TEST_F( FCQueue, std_list_empty_wait_strategy )
+    {
+        typedef cds::container::FCQueue<int, std::queue< int, std::list<int> >,
+            cds::container::fcqueue::make_traits<
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::empty >
+            >::type
+        > queue_type;
+
+        queue_type q;
+        test( q );
+    }
+
+    TEST_F( FCQueue, std_list_single_mutex_single_condvar )
+    {
+        typedef cds::container::FCQueue<int, std::queue< int, std::list<int> >,
+            cds::container::fcqueue::make_traits<
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_single_condvar<5>>
+            >::type
+        > queue_type;
+
+        queue_type q;
+        test( q );
+    }
+
     TEST_F( FCQueue, std_list_elimination )
     {
         typedef cds::container::FCQueue<int, std::queue< int, std::list<int> >,
@@ -218,6 +292,19 @@ namespace {
         test( q );
     }
 
+    TEST_F( FCQueue, std_list_elimination_multi_mutex_multi_condvar )
+    {
+        typedef cds::container::FCQueue<int, std::queue< int, std::list<int> >,
+            cds::container::fcqueue::make_traits<
+                cds::opt::enable_elimination< true >
+                ,cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::multi_mutex_multi_condvar<5>>
+            >::type
+        > queue_type;
+
+        queue_type q;
+        test( q );
+    }
+
     TEST_F( FCQueue, std_list_elimination_move )
     {
         typedef cds::container::FCQueue<std::string, std::queue< std::string, std::list<std::string> >,
index 189cf0978f2386481ef82833e5e205b96200ae08..aa2db27120bd924a1b2e5e2445e017b7782100c1 100644 (file)
@@ -169,6 +169,36 @@ namespace {
         test( q );
     }
 
+    TEST_F( IntrusiveFCQueue, base_empty_wait_strategy )
+    {
+        typedef base_hook_item< boost::intrusive::list_base_hook<> > value_type;
+        struct traits: public cds::intrusive::fcqueue::traits
+        {
+            typedef IntrusiveFCQueue::disposer disposer;
+            typedef cds::algo::flat_combining::wait_strategy::empty wait_strategy;
+            typedef cds::intrusive::fcqueue::stat<> stat;
+        };
+        typedef cds::intrusive::FCQueue< value_type, boost::intrusive::list< value_type >, traits > queue_type;
+
+        queue_type q;
+        test( q );
+    }
+
+    TEST_F( IntrusiveFCQueue, base_single_mutex_single_condvar )
+    {
+        typedef base_hook_item< boost::intrusive::list_base_hook<> > value_type;
+        struct traits: public cds::intrusive::fcqueue::traits
+        {
+            typedef IntrusiveFCQueue::disposer disposer;
+            typedef cds::algo::flat_combining::wait_strategy::single_mutex_single_condvar<> wait_strategy;
+            typedef cds::intrusive::fcqueue::stat<> stat;
+        };
+        typedef cds::intrusive::FCQueue< value_type, boost::intrusive::list< value_type >, traits > queue_type;
+
+        queue_type q;
+        test( q );
+    }
+
     TEST_F( IntrusiveFCQueue, base_mutex )
     {
         typedef base_hook_item< boost::intrusive::list_base_hook<> > value_type;
@@ -184,6 +214,22 @@ namespace {
         test( q );
     }
 
+    TEST_F( IntrusiveFCQueue, base_mutex_single_mutex_multi_condvar )
+    {
+        typedef base_hook_item< boost::intrusive::list_base_hook<> > value_type;
+        struct traits: public cds::intrusive::fcqueue::traits
+        {
+            typedef IntrusiveFCQueue::disposer disposer;
+            typedef std::mutex lock_type;
+            typedef cds::intrusive::fcqueue::stat<> stat;
+            typedef cds::algo::flat_combining::wait_strategy::single_mutex_multi_condvar<> wait_strategy;
+        };
+        typedef cds::intrusive::FCQueue< value_type, boost::intrusive::list< value_type >, traits > queue_type;
+
+        queue_type q;
+        test( q );
+    }
+
     TEST_F( IntrusiveFCQueue, base_elimination )
     {
         typedef base_hook_item< boost::intrusive::list_base_hook<> > value_type;
@@ -199,6 +245,22 @@ namespace {
         test( q );
     }
 
+    TEST_F( IntrusiveFCQueue, base_elimination_multi_mutex_multi_condvar )
+    {
+        typedef base_hook_item< boost::intrusive::list_base_hook<> > value_type;
+        struct traits: public
+            cds::intrusive::fcqueue::make_traits <
+                cds::intrusive::opt::disposer< disposer >
+                , cds::opt::enable_elimination < true >
+                , cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_multi_condvar<>>
+            > ::type
+        {};
+        typedef cds::intrusive::FCQueue< value_type, boost::intrusive::list< value_type >, traits > queue_type;
+
+        queue_type q;
+        test( q );
+    }
+
     TEST_F( IntrusiveFCQueue, member )
     {
         typedef member_hook_item< boost::intrusive::list_member_hook<> > value_type;
@@ -231,6 +293,54 @@ namespace {
         test( q );
     }
 
+    TEST_F( IntrusiveFCQueue, member_empty_wait_strategy )
+    {
+        typedef member_hook_item< boost::intrusive::list_member_hook<> > value_type;
+        typedef boost::intrusive::member_hook<value_type, boost::intrusive::list_member_hook<>, &value_type::hMember> member_option;
+
+        typedef cds::intrusive::FCQueue< value_type, boost::intrusive::list< value_type, member_option >,
+            cds::intrusive::fcqueue::make_traits<
+                cds::intrusive::opt::disposer< disposer >
+                , cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::empty >
+            >::type
+        > queue_type;
+
+        queue_type q;
+        test( q );
+    }
+
+    TEST_F( IntrusiveFCQueue, member_single_mutex_single_condvar )
+    {
+        typedef member_hook_item< boost::intrusive::list_member_hook<> > value_type;
+        typedef boost::intrusive::member_hook<value_type, boost::intrusive::list_member_hook<>, &value_type::hMember> member_option;
+
+        typedef cds::intrusive::FCQueue< value_type, boost::intrusive::list< value_type, member_option >,
+            cds::intrusive::fcqueue::make_traits<
+                cds::intrusive::opt::disposer< disposer >
+                , cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_single_condvar<2>>
+            >::type
+        > queue_type;
+
+        queue_type q;
+        test( q );
+    }
+
+    TEST_F( IntrusiveFCQueue, member_multi_mutex_multi_condvar )
+    {
+        typedef member_hook_item< boost::intrusive::list_member_hook<> > value_type;
+        typedef boost::intrusive::member_hook<value_type, boost::intrusive::list_member_hook<>, &value_type::hMember> member_option;
+
+        typedef cds::intrusive::FCQueue< value_type, boost::intrusive::list< value_type, member_option >,
+            cds::intrusive::fcqueue::make_traits<
+                cds::intrusive::opt::disposer< disposer >
+                , cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::multi_mutex_multi_condvar<>>
+            >::type
+        > queue_type;
+
+        queue_type q;
+        test( q );
+    }
+
     TEST_F( IntrusiveFCQueue, member_elimination )
     {
         typedef member_hook_item< boost::intrusive::list_member_hook<> > value_type;
@@ -247,6 +357,23 @@ namespace {
         test( q );
     }
 
+    TEST_F( IntrusiveFCQueue, member_elimination_single_mutex_multi_condvar )
+    {
+        typedef member_hook_item< boost::intrusive::list_member_hook<> > value_type;
+        typedef boost::intrusive::member_hook<value_type, boost::intrusive::list_member_hook<>, &value_type::hMember> member_option;
+
+        typedef cds::intrusive::FCQueue< value_type, boost::intrusive::list< value_type, member_option >,
+            cds::intrusive::fcqueue::make_traits<
+                cds::intrusive::opt::disposer< disposer >
+                ,cds::opt::enable_elimination< true >
+                , cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_multi_condvar<2>>
+            >::type
+        > queue_type;
+
+        queue_type q;
+        test( q );
+    }
+
     TEST_F( IntrusiveFCQueue, slist_base )
     {
         typedef base_hook_item< boost::intrusive::slist_base_hook<>> value_type;
index fff7e296d3d8bb1ac79c3c72028e9736a9f6fbc0..6637004395dcff72f07d958141c6e955cc6e2700 100644 (file)
@@ -53,7 +53,7 @@ namespace {
         void TearDown()
         {
             cds::threading::Manager::detachThread();
-            cds::gc::hp::GarbageCollector::Destruct();
+            cds::gc::dhp::GarbageCollector::Destruct();
         }
     };
 
index 03c178a7d78a1c49a9c96675a382e279f4a7fffd..9e612af4fa28636c1d5515ef0a9938778955e29a 100644 (file)
@@ -101,6 +101,83 @@ namespace {
         test<stack_type>();
     }
 
+    TEST_F( FCStack, deque_empty_wait_strategy )
+    {
+        struct stack_traits: public
+            cds::container::fcstack::make_traits <
+                cds::opt::wait_strategy<cds::algo::flat_combining::wait_strategy::empty>
+            > ::type
+        {};
+        typedef cds::container::FCStack< unsigned int, std::stack<unsigned int, std::deque<unsigned int>>, stack_traits > stack_type;
+        test<stack_type>();
+    }
+
+    TEST_F( FCStack, deque_single_mutex_single_condvar )
+    {
+        struct stack_traits: public
+            cds::container::fcstack::make_traits <
+            cds::opt::wait_strategy<cds::algo::flat_combining::wait_strategy::single_mutex_single_condvar<>>
+            > ::type
+        {};
+        typedef cds::container::FCStack< unsigned int, std::stack<unsigned int, std::deque<unsigned int>>, stack_traits > stack_type;
+        test<stack_type>();
+    }
+
+    TEST_F( FCStack, deque_single_mutex_multi_condvar )
+    {
+        struct stack_traits: public
+            cds::container::fcstack::make_traits <
+            cds::opt::wait_strategy<cds::algo::flat_combining::wait_strategy::single_mutex_multi_condvar<>>
+            > ::type
+        {};
+        typedef cds::container::FCStack< unsigned int, std::stack<unsigned int, std::deque<unsigned int>>, stack_traits > stack_type;
+        test<stack_type>();
+    }
+
+    TEST_F( FCStack, deque_multi_mutex_multi_condvar )
+    {
+        struct stack_traits: public
+            cds::container::fcstack::make_traits <
+            cds::opt::wait_strategy<cds::algo::flat_combining::wait_strategy::multi_mutex_multi_condvar<>>
+            > ::type
+        {};
+        typedef cds::container::FCStack< unsigned int, std::stack<unsigned int, std::deque<unsigned int>>, stack_traits > stack_type;
+        test<stack_type>();
+    }
+
+    TEST_F( FCStack, deque_single_mutex_single_condvar_2ms )
+    {
+        struct stack_traits: public
+            cds::container::fcstack::make_traits <
+            cds::opt::wait_strategy<cds::algo::flat_combining::wait_strategy::single_mutex_single_condvar<2>>
+            > ::type
+        {};
+        typedef cds::container::FCStack< unsigned int, std::stack<unsigned int, std::deque<unsigned int>>, stack_traits > stack_type;
+        test<stack_type>();
+    }
+
+    TEST_F( FCStack, deque_single_mutex_multi_condvar_2ms )
+    {
+        struct stack_traits: public
+            cds::container::fcstack::make_traits <
+            cds::opt::wait_strategy<cds::algo::flat_combining::wait_strategy::single_mutex_multi_condvar<2>>
+            > ::type
+        {};
+        typedef cds::container::FCStack< unsigned int, std::stack<unsigned int, std::deque<unsigned int>>, stack_traits > stack_type;
+        test<stack_type>();
+    }
+
+    TEST_F( FCStack, deque_multi_mutex_multi_condvar_3ms )
+    {
+        struct stack_traits: public
+            cds::container::fcstack::make_traits <
+            cds::opt::wait_strategy<cds::algo::flat_combining::wait_strategy::multi_mutex_multi_condvar<3>>
+            > ::type
+        {};
+        typedef cds::container::FCStack< unsigned int, std::stack<unsigned int, std::deque<unsigned int>>, stack_traits > stack_type;
+        test<stack_type>();
+    }
+
     TEST_F( FCStack, deque_elimination )
     {
         struct stack_traits : public
@@ -118,6 +195,46 @@ namespace {
         test<stack_type>();
     }
 
+    TEST_F( FCStack, vector_empty_wait_strategy )
+    {
+        typedef cds::container::FCStack< unsigned int, std::stack<unsigned int, std::vector<unsigned int>>,
+            cds::container::fcstack::make_traits<
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::empty >
+            >::type
+        > stack_type;
+        test<stack_type>();
+    }
+
+    TEST_F( FCStack, vector_multi_mutex_multi_condvar )
+    {
+        typedef cds::container::FCStack< unsigned int, std::stack<unsigned int, std::vector<unsigned int>>,
+            cds::container::fcstack::make_traits<
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::multi_mutex_multi_condvar<>>
+            >::type
+        > stack_type;
+        test<stack_type>();
+    }
+
+    TEST_F( FCStack, vector_single_mutex_multi_condvar )
+    {
+        typedef cds::container::FCStack< unsigned int, std::stack<unsigned int, std::vector<unsigned int>>,
+            cds::container::fcstack::make_traits<
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_multi_condvar<>>
+            >::type
+        > stack_type;
+        test<stack_type>();
+    }
+
+    TEST_F( FCStack, vector_single_mutex_single_condvar )
+    {
+        typedef cds::container::FCStack< unsigned int, std::stack<unsigned int, std::vector<unsigned int>>,
+            cds::container::fcstack::make_traits<
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_single_condvar<>>
+            >::type
+        > stack_type;
+        test<stack_type>();
+    }
+
     TEST_F( FCStack, vector_elimination )
     {
         typedef cds::container::FCStack< unsigned int, std::stack<unsigned int, std::vector<unsigned int>>,
@@ -134,11 +251,52 @@ namespace {
         test<stack_type>();
     }
 
+    TEST_F( FCStack, list_empty_wait_strategy )
+    {
+        typedef cds::container::FCStack< unsigned int, std::stack<unsigned int, std::list<unsigned int>>,
+            cds::container::fcstack::make_traits<
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::empty >
+            >::type
+        > stack_type;
+        test<stack_type>();
+    }
+
+    TEST_F( FCStack, list_single_mutex_single_condvar )
+    {
+        typedef cds::container::FCStack< unsigned int, std::stack<unsigned int, std::list<unsigned int>>,
+            cds::container::fcstack::make_traits<
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_single_condvar<>>
+            >::type
+        > stack_type;
+        test<stack_type>();
+    }
+
+    TEST_F( FCStack, list_single_mutex_multi_condvar )
+    {
+        typedef cds::container::FCStack< unsigned int, std::stack<unsigned int, std::list<unsigned int>>,
+            cds::container::fcstack::make_traits<
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_multi_condvar<>>
+            >::type
+        > stack_type;
+        test<stack_type>();
+    }
+
+    TEST_F( FCStack, list_multi_mutex_multi_condvar )
+    {
+        typedef cds::container::FCStack< unsigned int, std::stack<unsigned int, std::list<unsigned int>>,
+            cds::container::fcstack::make_traits<
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::multi_mutex_multi_condvar<>>
+            >::type
+        > stack_type;
+        test<stack_type>();
+    }
+
     TEST_F( FCStack, list_elimination )
     {
         typedef cds::container::FCStack< unsigned int, std::stack<unsigned int, std::list<unsigned int>>,
             cds::container::fcstack::make_traits<
-            cds::opt::enable_elimination< true >
+                cds::opt::enable_elimination< true >
+                , cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::multi_mutex_multi_condvar<2>>
             >::type
         > stack_type;
         test<stack_type>();
index d4cce2b235d2c7b0aea53d20ed651799209893a7..32b25bbbd18c347882179105adc00bbf666dc41c 100644 (file)
@@ -143,6 +143,83 @@ namespace {
         test<stack_type>();
     }
 
+    TEST_F( IntrusiveFCStack, slist_empty_wait_strategy )
+    {
+        typedef base_hook_item< boost::intrusive::slist_base_hook<> > value_type;
+        struct stack_traits: public cds::intrusive::fcstack::traits
+        {
+            typedef cds::algo::flat_combining::wait_strategy::empty wait_strategy;
+        };
+        typedef cds::intrusive::FCStack< value_type, boost::intrusive::slist< value_type >, stack_traits > stack_type;
+        test<stack_type>();
+    }
+
+    TEST_F( IntrusiveFCStack, slist_single_mutex_single_condvar )
+    {
+        typedef base_hook_item< boost::intrusive::slist_base_hook<> > value_type;
+        struct stack_traits: public cds::intrusive::fcstack::traits
+        {
+            typedef cds::algo::flat_combining::wait_strategy::single_mutex_single_condvar<> wait_strategy;
+        };
+        typedef cds::intrusive::FCStack< value_type, boost::intrusive::slist< value_type >, stack_traits > stack_type;
+        test<stack_type>();
+    }
+
+    TEST_F( IntrusiveFCStack, slist_single_mutex_multi_condvar )
+    {
+        typedef base_hook_item< boost::intrusive::slist_base_hook<> > value_type;
+        struct stack_traits: public cds::intrusive::fcstack::traits
+        {
+            typedef cds::algo::flat_combining::wait_strategy::single_mutex_multi_condvar<> wait_strategy;
+        };
+        typedef cds::intrusive::FCStack< value_type, boost::intrusive::slist< value_type >, stack_traits > stack_type;
+        test<stack_type>();
+    }
+
+    TEST_F( IntrusiveFCStack, slist_multi_mutex_multi_condvar )
+    {
+        typedef base_hook_item< boost::intrusive::slist_base_hook<> > value_type;
+        struct stack_traits: public cds::intrusive::fcstack::traits
+        {
+            typedef cds::algo::flat_combining::wait_strategy::multi_mutex_multi_condvar<> wait_strategy;
+        };
+        typedef cds::intrusive::FCStack< value_type, boost::intrusive::slist< value_type >, stack_traits > stack_type;
+        test<stack_type>();
+    }
+
+    TEST_F( IntrusiveFCStack, slist_single_mutex_single_condvar_2ms )
+    {
+        typedef base_hook_item< boost::intrusive::slist_base_hook<> > value_type;
+        struct stack_traits: public cds::intrusive::fcstack::traits
+        {
+            typedef cds::algo::flat_combining::wait_strategy::single_mutex_single_condvar<2> wait_strategy;
+        };
+        typedef cds::intrusive::FCStack< value_type, boost::intrusive::slist< value_type >, stack_traits > stack_type;
+        test<stack_type>();
+    }
+
+    TEST_F( IntrusiveFCStack, slist_single_mutex_multi_condvar_3ms )
+    {
+        typedef base_hook_item< boost::intrusive::slist_base_hook<> > value_type;
+        struct stack_traits: public cds::intrusive::fcstack::traits
+        {
+            typedef cds::algo::flat_combining::wait_strategy::single_mutex_multi_condvar<3> wait_strategy;
+        };
+        typedef cds::intrusive::FCStack< value_type, boost::intrusive::slist< value_type >, stack_traits > stack_type;
+        test<stack_type>();
+    }
+
+    TEST_F( IntrusiveFCStack, slist_multi_mutex_multi_condvar_2ms )
+    {
+        typedef base_hook_item< boost::intrusive::slist_base_hook<> > value_type;
+        struct stack_traits: public cds::intrusive::fcstack::traits
+        {
+            typedef cds::algo::flat_combining::wait_strategy::multi_mutex_multi_condvar<2> wait_strategy;
+        };
+        typedef cds::intrusive::FCStack< value_type, boost::intrusive::slist< value_type >, stack_traits > stack_type;
+        test<stack_type>();
+    }
+
     TEST_F( IntrusiveFCStack, slist_disposer )
     {
         typedef base_hook_item< boost::intrusive::slist_base_hook<> > value_type;
@@ -183,7 +260,8 @@ namespace {
         struct stack_traits : public
             cds::intrusive::fcstack::make_traits <
                 cds::opt::enable_elimination < true >,
-                cds::intrusive::opt::disposer< mock_disposer >
+                cds::intrusive::opt::disposer< mock_disposer >,
+                cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::multi_mutex_multi_condvar<>>
             > ::type
         {};
         typedef cds::intrusive::FCStack< value_type, boost::intrusive::slist< value_type >, stack_traits > stack_type;
@@ -195,8 +273,9 @@ namespace {
         typedef base_hook_item< boost::intrusive::slist_base_hook<> > value_type;
         typedef cds::intrusive::FCStack< value_type, boost::intrusive::slist< value_type >,
             cds::intrusive::fcstack::make_traits<
-            cds::opt::enable_elimination< true >
-            , cds::opt::stat< cds::intrusive::fcstack::stat<> >
+                cds::opt::enable_elimination< true >
+                , cds::opt::stat< cds::intrusive::fcstack::stat<> >
+                , cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_multi_condvar<>>
             >::type
         > stack_type;
         test<stack_type>();
@@ -211,6 +290,58 @@ namespace {
         test<stack_type>();
     }
 
+    TEST_F( IntrusiveFCStack, slist_member_empty_wait_strategy )
+    {
+        typedef member_hook_item< boost::intrusive::slist_member_hook<> > value_type;
+        typedef boost::intrusive::member_hook<value_type, boost::intrusive::slist_member_hook<>, &value_type::hMember> member_option;
+        struct stack_traits: public cds::intrusive::fcstack::traits
+        {
+            typedef cds::algo::flat_combining::wait_strategy::empty wait_strategy;
+        };
+
+        typedef cds::intrusive::FCStack< value_type, boost::intrusive::slist< value_type, member_option >, stack_traits > stack_type;
+        test<stack_type>();
+    }
+
+    TEST_F( IntrusiveFCStack, slist_member_single_mutex_single_condvar )
+    {
+        typedef member_hook_item< boost::intrusive::slist_member_hook<> > value_type;
+        typedef boost::intrusive::member_hook<value_type, boost::intrusive::slist_member_hook<>, &value_type::hMember> member_option;
+        struct stack_traits: public cds::intrusive::fcstack::traits
+        {
+            typedef cds::algo::flat_combining::wait_strategy::single_mutex_single_condvar<> wait_strategy;
+        };
+
+        typedef cds::intrusive::FCStack< value_type, boost::intrusive::slist< value_type, member_option >, stack_traits > stack_type;
+        test<stack_type>();
+    }
+
+    TEST_F( IntrusiveFCStack, slist_member_single_mutex_multi_condvar )
+    {
+        typedef member_hook_item< boost::intrusive::slist_member_hook<> > value_type;
+        typedef boost::intrusive::member_hook<value_type, boost::intrusive::slist_member_hook<>, &value_type::hMember> member_option;
+        struct stack_traits: public cds::intrusive::fcstack::traits
+        {
+            typedef cds::algo::flat_combining::wait_strategy::single_mutex_multi_condvar<> wait_strategy;
+        };
+
+        typedef cds::intrusive::FCStack< value_type, boost::intrusive::slist< value_type, member_option >, stack_traits > stack_type;
+        test<stack_type>();
+    }
+
+    TEST_F( IntrusiveFCStack, slist_member_multi_mutex_multi_condvar )
+    {
+        typedef member_hook_item< boost::intrusive::slist_member_hook<> > value_type;
+        typedef boost::intrusive::member_hook<value_type, boost::intrusive::slist_member_hook<>, &value_type::hMember> member_option;
+        struct stack_traits: public cds::intrusive::fcstack::traits
+        {
+            typedef cds::algo::flat_combining::wait_strategy::multi_mutex_multi_condvar<> wait_strategy;
+        };
+
+        typedef cds::intrusive::FCStack< value_type, boost::intrusive::slist< value_type, member_option >, stack_traits > stack_type;
+        test<stack_type>();
+    }
+
     TEST_F( IntrusiveFCStack, slist_member_disposer )
     {
         typedef member_hook_item< boost::intrusive::slist_member_hook<> > value_type;
@@ -246,6 +377,7 @@ namespace {
             cds::intrusive::fcstack::make_traits<
             cds::opt::enable_elimination< true >
             , cds::opt::stat< cds::intrusive::fcstack::stat<> >
+            , cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_multi_condvar<>>
             >::type
         > stack_type;
         test<stack_type>();
@@ -288,6 +420,7 @@ namespace {
             cds::intrusive::fcstack::make_traits<
             cds::opt::enable_elimination< true >
             , cds::opt::stat< cds::intrusive::fcstack::stat<> >
+            , cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::multi_mutex_multi_condvar<>>
             >::type
         > stack_type;
         test<stack_type>();
@@ -324,6 +457,7 @@ namespace {
             cds::intrusive::fcstack::make_traits<
             cds::opt::enable_elimination< true >
             , cds::opt::stat< cds::intrusive::fcstack::stat<> >
+            , cds::opt::wait_strategy< cds::algo::flat_combining::wait_strategy::single_mutex_single_condvar<>>
             >::type
         > stack_type;
         test<stack_type>();