parallel_scan.h

Go to the documentation of this file.

/*
    Copyright 2005-2010 Intel Corporation.  All Rights Reserved.

    This file is part of Threading Building Blocks.

    Threading Building Blocks is free software; you can redistribute it
    and/or modify it under the terms of the GNU General Public License
    version 2 as published by the Free Software Foundation.

    Threading Building Blocks is distributed in the hope that it will be
    useful, but WITHOUT ANY WARRANTY; without even the implied warranty
    of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
    GNU General Public License for more details.

    You should have received a copy of the GNU General Public License
    along with Threading Building Blocks; if not, write to the Free Software
    Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA

    As a special exception, you may use this file as part of a free software
    library without restriction.  Specifically, if other files instantiate
    templates or use macros or inline functions from this file, or you compile
    this file and link it with other files to produce an executable, this
    file does not by itself cause the resulting executable to be covered by
    the GNU General Public License.  This exception does not however
    invalidate any other reasons why the executable file might be covered by
    the GNU General Public License.
*/

#ifndef __TBB_parallel_scan_H
#define __TBB_parallel_scan_H

#include "task.h"
#include "aligned_space.h"
#include <new>
#include "partitioner.h"

namespace tbb {


struct pre_scan_tag {
    static bool is_final_scan() {return false;}
};


struct final_scan_tag {
    static bool is_final_scan() {return true;}
};

namespace internal {


    template<typename Range, typename Body>
    class final_sum: public task {
    public:
        Body body;
    private:
        aligned_space<Range,1> range;
        Body* stuff_last;
    public:
        final_sum( Body& body_ ) :
            body(body_,split())
        {
            poison_pointer(stuff_last);
        }
        ~final_sum() {
            range.begin()->~Range();
        }     
        void finish_construction( const Range& range_, Body* stuff_last_ ) {
            new( range.begin() ) Range(range_);
            stuff_last = stuff_last_;
        }
    private:
        /*override*/ task* execute() {
            body( *range.begin(), final_scan_tag() );
            if( stuff_last )
                stuff_last->assign(body);
            return NULL;
        }
    };       


    template<typename Range, typename Body>
    class sum_node: public task {
        typedef final_sum<Range,Body> final_sum_type;
    public:
        final_sum_type *incoming; 
        final_sum_type *body;
        Body *stuff_last;
    private:
        final_sum_type *left_sum;
        sum_node *left;
        sum_node *right;     
        bool left_is_final;
        Range range;
        sum_node( const Range range_, bool left_is_final_ ) : 
            left_sum(NULL), 
            left(NULL), 
            right(NULL), 
            left_is_final(left_is_final_), 
            range(range_)
        {
            // Poison fields that will be set by second pass.
            poison_pointer(body);
            poison_pointer(incoming);
        }
        task* create_child( const Range& range_, final_sum_type& f, sum_node* n, final_sum_type* incoming_, Body* stuff_last_ ) {
            if( !n ) {
                f.recycle_as_child_of( *this );
                f.finish_construction( range_, stuff_last_ );
                return &f;
            } else {
                n->body = &f;
                n->incoming = incoming_;
                n->stuff_last = stuff_last_;
                return n;
            }
        }
        /*override*/ task* execute() {
            if( body ) {
                if( incoming )
                    left_sum->body.reverse_join( incoming->body );
                recycle_as_continuation();
                sum_node& c = *this;
                task* b = c.create_child(Range(range,split()),*left_sum,right,left_sum,stuff_last);
                task* a = left_is_final ? NULL : c.create_child(range,*body,left,incoming,NULL);
                set_ref_count( (a!=NULL)+(b!=NULL) );
                body = NULL; 
                if( a ) spawn(*b);
                else a = b;
                return a;
            } else {
                return NULL;
            }
        }
        template<typename Range_,typename Body_,typename Partitioner_>
        friend class start_scan;

        template<typename Range_,typename Body_>
        friend class finish_scan;
    };


    template<typename Range, typename Body>
    class finish_scan: public task {
        typedef sum_node<Range,Body> sum_node_type;
        typedef final_sum<Range,Body> final_sum_type;
        final_sum_type** const sum;
        sum_node_type*& return_slot;
    public:
        final_sum_type* right_zombie;
        sum_node_type& result;

        /*override*/ task* execute() {
            __TBB_ASSERT( result.ref_count()==(result.left!=NULL)+(result.right!=NULL), NULL );
            if( result.left )
                result.left_is_final = false;
            if( right_zombie && sum ) 
                ((*sum)->body).reverse_join(result.left_sum->body);
            __TBB_ASSERT( !return_slot, NULL );
            if( right_zombie || result.right ) {
                return_slot = &result;
            } else {
                destroy( result );
            }
            if( right_zombie && !sum && !result.right ) destroy(*right_zombie);
            return NULL;
        }

        finish_scan( sum_node_type*& return_slot_, final_sum_type** sum_, sum_node_type& result_ ) : 
            sum(sum_),
            return_slot(return_slot_), 
            right_zombie(NULL),
            result(result_)
        {
            __TBB_ASSERT( !return_slot, NULL );
        }
    };


    template<typename Range, typename Body, typename Partitioner=simple_partitioner>
    class start_scan: public task {
        typedef sum_node<Range,Body> sum_node_type;
        typedef final_sum<Range,Body> final_sum_type;
        final_sum_type* body;
        final_sum_type** sum; 
        sum_node_type** return_slot;
        sum_node_type* parent_sum;
        bool is_final;
        bool is_right_child;
        Range range;
        typename Partitioner::partition_type partition;
        /*override*/ task* execute();
    public:
        start_scan( sum_node_type*& return_slot_, start_scan& parent_, sum_node_type* parent_sum_ ) :
            body(parent_.body),
            sum(parent_.sum),
            return_slot(&return_slot_),
            parent_sum(parent_sum_),
            is_final(parent_.is_final),
            is_right_child(false),
            range(parent_.range,split()),
            partition(parent_.partition,split())
        {
            __TBB_ASSERT( !*return_slot, NULL );
        }

        start_scan( sum_node_type*& return_slot_, const Range& range_, final_sum_type& body_, const Partitioner& partitioner_) :
            body(&body_),
            sum(NULL),
            return_slot(&return_slot_),
            parent_sum(NULL),
            is_final(true),
            is_right_child(false),
            range(range_),
            partition(partitioner_)
        {
            __TBB_ASSERT( !*return_slot, NULL );
        }

        static void run(  const Range& range, Body& body, const Partitioner& partitioner ) {
            if( !range.empty() ) {
                typedef internal::start_scan<Range,Body,Partitioner> start_pass1_type;
                internal::sum_node<Range,Body>* root = NULL;
                typedef internal::final_sum<Range,Body> final_sum_type;
                final_sum_type* temp_body = new(task::allocate_root()) final_sum_type( body );
                start_pass1_type& pass1 = *new(task::allocate_root()) start_pass1_type(
                    /*return_slot=*/root,
                    range,
                    *temp_body,
                    partitioner );
                task::spawn_root_and_wait( pass1 );
                if( root ) {
                    root->body = temp_body;
                    root->incoming = NULL;
                    root->stuff_last = &body;
                    task::spawn_root_and_wait( *root );
                } else {
                    body.assign(temp_body->body);
                    temp_body->finish_construction( range, NULL );
                    temp_body->destroy(*temp_body);
                }
            }
        }
    };

    template<typename Range, typename Body, typename Partitioner>
    task* start_scan<Range,Body,Partitioner>::execute() {
        typedef internal::finish_scan<Range,Body> finish_pass1_type;
        finish_pass1_type* p = parent_sum ? static_cast<finish_pass1_type*>( parent() ) : NULL;
        // Inspecting p->result.left_sum would ordinarily be a race condition.
        // But we inspect it only if we are not a stolen task, in which case we
        // know that task assigning to p->result.left_sum has completed.
        bool treat_as_stolen = is_right_child && (is_stolen_task() || body!=p->result.left_sum);
        if( treat_as_stolen ) {
            // Invocation is for right child that has been really stolen or needs to be virtually stolen
            p->right_zombie = body = new( allocate_root() ) final_sum_type(body->body);
            is_final = false;
        }
        task* next_task = NULL;
        if( (is_right_child && !treat_as_stolen) || !range.is_divisible() || partition.should_execute_range(*this) ) {
            if( is_final )
                (body->body)( range, final_scan_tag() );
            else if( sum )
                (body->body)( range, pre_scan_tag() );
            if( sum ) 
                *sum = body;
            __TBB_ASSERT( !*return_slot, NULL );
        } else {
            sum_node_type* result;
            if( parent_sum ) 
                result = new(allocate_additional_child_of(*parent_sum)) sum_node_type(range,/*left_is_final=*/is_final);
            else
                result = new(task::allocate_root()) sum_node_type(range,/*left_is_final=*/is_final);
            finish_pass1_type& c = *new( allocate_continuation()) finish_pass1_type(*return_slot,sum,*result);
            // Split off right child
            start_scan& b = *new( c.allocate_child() ) start_scan( /*return_slot=*/result->right, *this, result );
            b.is_right_child = true;    
            // Left child is recycling of *this.  Must recycle this before spawning b, 
            // otherwise b might complete and decrement c.ref_count() to zero, which
            // would cause c.execute() to run prematurely.
            recycle_as_child_of(c);
            c.set_ref_count(2);
            c.spawn(b);
            sum = &result->left_sum;
            return_slot = &result->left;
            is_right_child = false;
            next_task = this;
            parent_sum = result; 
            __TBB_ASSERT( !*return_slot, NULL );
        }
        return next_task;
    } 
} // namespace internal

// Requirements on Range concept are documented in blocked_range.h



template<typename Range, typename Body>
void parallel_scan( const Range& range, Body& body ) {
    internal::start_scan<Range,Body,__TBB_DEFAULT_PARTITIONER>::run(range,body,__TBB_DEFAULT_PARTITIONER());
}


template<typename Range, typename Body>
void parallel_scan( const Range& range, Body& body, const simple_partitioner& partitioner ) {
    internal::start_scan<Range,Body,simple_partitioner>::run(range,body,partitioner);
}


template<typename Range, typename Body>
void parallel_scan( const Range& range, Body& body, const auto_partitioner& partitioner ) {
    internal::start_scan<Range,Body,auto_partitioner>::run(range,body,partitioner);
}

} // namespace tbb

#endif /* __TBB_parallel_scan_H */