<?xml version="1.0" encoding="utf-8" ?> <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"> <html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en" lang="en"> <head> <meta http-equiv="Content-Type" content="text/html; charset=utf-8" /> <meta name="generator" content="Docutils 0.6: http://docutils.sourceforge.net/" /> <title>Parallel BGL Breadth-First Search</title> <link rel="stylesheet" href="../../../../rst.css" type="text/css" /> </head> <body> <div class="document" id="logo-breadth-first-search"> <h1 class="title"><a class="reference external" href="http://www.osl.iu.edu/research/pbgl"><img align="middle" alt="Parallel BGL" class="align-middle" src="pbgl-logo.png" /></a> Breadth-First Search</h1> <!-- Copyright (C) 2004-2008 The Trustees of Indiana University. Use, modification and distribution is subject to the Boost Software License, Version 1.0. (See accompanying file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt) --> <pre class="literal-block"> // named parameter version template <class Graph, class P, class T, class R> void breadth_first_search(Graph& G, typename graph_traits<Graph>::vertex_descriptor s, const bgl_named_params<P, T, R>& params); // non-named parameter version template <class Graph, class Buffer, class BFSVisitor, class ColorMap> void breadth_first_search(const Graph& g, typename graph_traits<Graph>::vertex_descriptor s, Buffer& Q, BFSVisitor vis, ColorMap color); </pre> <p>The <tt class="docutils literal"><span class="pre">breadth_first_search()</span></tt> function performs a distributed breadth-first traversal of a directed or undirected graph. The distributed BFS is syntactically equivalent to its <a class="reference external" href="http://www.boost.org/libs/graph/doc/breadth_first_search.html">sequential counterpart</a>, which provides additional background and discussion. Differences in semantics are highlighted here, and we refer the reader to the documentation of the <a class="reference external" href="http://www.boost.org/libs/graph/doc/breadth_first_search.html">sequential breadth-first search</a> for the remainder of the details.</p> <p>This distributed breadth-first search algorithm implements a <em>level-synchronized</em> breadth-first search, meaning that all vertices in a given level of the BFS tree will be processed (potentially in parallel) before any vertices from a successive level in the tree are processed. Distributed breadth-first search visitors should account for this behavior, a topic discussed further in <a class="reference internal" href="#visitor-event-points">Visitor Event Points</a>.</p> <div class="contents topic" id="contents"> <p class="topic-title first">Contents</p> <ul class="simple"> <li><a class="reference internal" href="#where-defined" id="id1">Where Defined</a></li> <li><a class="reference internal" href="#parameter-defaults" id="id2">Parameter Defaults</a></li> <li><a class="reference internal" href="#complexity" id="id3">Complexity</a></li> <li><a class="reference internal" href="#visitor-event-points" id="id4">Visitor Event Points</a><ul> <li><a class="reference internal" href="#making-visitors-safe-for-distributed-bfs" id="id5">Making Visitors Safe for Distributed BFS</a></li> <li><a class="reference internal" href="#distributed-bfs-visitor-example" id="id6">Distributed BFS Visitor Example</a></li> </ul> </li> <li><a class="reference internal" href="#performance" id="id7">Performance</a></li> </ul> </div> <div class="section" id="where-defined"> <h1><a class="toc-backref" href="#id1">Where Defined</a></h1> <p><<tt class="docutils literal"><span class="pre">boost/graph/breadth_first_search.hpp</span></tt>></p> </div> <div class="section" id="parameter-defaults"> <h1><a class="toc-backref" href="#id2">Parameter Defaults</a></h1> <p>All parameters of the <a class="reference external" href="http://www.boost.org/libs/graph/doc/breadth_first_search.html">sequential breadth-first search</a> are supported and have essentially the same meaning. Only differences are documented here.</p> <dl class="docutils"> <dt>IN: <tt class="docutils literal"><span class="pre">Graph&</span> <span class="pre">g</span></tt></dt> <dd>The graph type must be a model of <a class="reference external" href="DistributedGraph.html">Distributed Graph</a>.</dd> <dt>IN: <tt class="docutils literal"><span class="pre">vertex_descriptor</span> <span class="pre">s</span></tt></dt> <dd>The start vertex must be the same in every process.</dd> <dt>IN: <tt class="docutils literal"><span class="pre">visitor(BFSVisitor</span> <span class="pre">vis)</span></tt></dt> <dd>The visitor must be a distributed BFS visitor. The suble differences between sequential and distributed BFS visitors are discussed in the section <a class="reference internal" href="#visitor-event-points">Visitor Event Points</a>.</dd> <dt>UTIL/OUT: <tt class="docutils literal"><span class="pre">color_map(ColorMap</span> <span class="pre">color)</span></tt></dt> <dd>The color map must be a <a class="reference external" href="distributed_property_map.html">Distributed Property Map</a> with the same process group as the graph <tt class="docutils literal"><span class="pre">g</span></tt> whose colors must monotonically darken (white -> gray -> black). The default value is a distributed <tt class="docutils literal"><span class="pre">iterator_property_map</span></tt> created from a <tt class="docutils literal"><span class="pre">std::vector</span></tt> of <tt class="docutils literal"><span class="pre">default_color_type</span></tt>.</dd> <dt>UTIL: <tt class="docutils literal"><span class="pre">buffer(Buffer&</span> <span class="pre">Q)</span></tt></dt> <dd><p class="first">The queue must be a distributed queue that passes vertices to their owning process. If already-visited vertices should not be visited again (as is typical for BFS), the queue must filter duplicates itself. The queue controls synchronization within the algorithm: its <tt class="docutils literal"><span class="pre">empty()</span></tt> method must not return <tt class="docutils literal"><span class="pre">true</span></tt> until all local queues are empty.</p> <dl class="last docutils"> <dt><strong>Default:</strong> A <tt class="docutils literal"><span class="pre">distributed_queue</span></tt> of a <tt class="docutils literal"><span class="pre">filtered_queue</span></tt> over a</dt> <dd>standard <tt class="docutils literal"><span class="pre">boost::queue</span></tt>. This queue filters out duplicate vertices and distributes vertices appropriately.</dd> </dl> </dd> </dl> </div> <div class="section" id="complexity"> <h1><a class="toc-backref" href="#id3">Complexity</a></h1> <p>This algorithm performs <em>O(V + E)</em> work in <em>d + 1</em> BSP supersteps, where <em>d</em> is the diameter of the connected component being searched. Over all supersteps, <em>O(E)</em> messages of constant size will be transmitted.</p> </div> <div class="section" id="visitor-event-points"> <h1><a class="toc-backref" href="#id4">Visitor Event Points</a></h1> <p>The <a class="reference external" href="http://www.boost.org/libs/graph/doc/BFSVisitor.html">BFS Visitor</a> concept defines 9 event points that will be triggered by the <a class="reference external" href="http://www.boost.org/libs/graph/doc/breadth_first_search.html">sequential breadth-first search</a>. The distributed BFS retains these nine event points, but the sequence of events triggered and the process in which each event occurs will change depending on the distribution of the graph.</p> <dl class="docutils"> <dt><tt class="docutils literal"><span class="pre">initialize_vertex(s,</span> <span class="pre">g)</span></tt></dt> <dd>This will be invoked by every process for each local vertex.</dd> <dt><tt class="docutils literal"><span class="pre">discover_vertex(u,</span> <span class="pre">g)</span></tt></dt> <dd>This will be invoked each time a process discovers a new vertex <tt class="docutils literal"><span class="pre">u</span></tt>. Due to incomplete information in distributed property maps, this event may be triggered many times for the same vertex <tt class="docutils literal"><span class="pre">u</span></tt>.</dd> <dt><tt class="docutils literal"><span class="pre">examine_vertex(u,</span> <span class="pre">g)</span></tt></dt> <dd>This will be invoked by the process owning the vertex <tt class="docutils literal"><span class="pre">u</span></tt>. If the distributed queue prevents duplicates, it will be invoked only once for a particular vertex <tt class="docutils literal"><span class="pre">u</span></tt>.</dd> <dt><tt class="docutils literal"><span class="pre">examine_edge(e,</span> <span class="pre">g)</span></tt></dt> <dd>This will be invoked by the process owning the source vertex of <tt class="docutils literal"><span class="pre">e</span></tt>. If the distributed queue prevents duplicates, it will be invoked only once for a particular edge <tt class="docutils literal"><span class="pre">e</span></tt>.</dd> <dt><tt class="docutils literal"><span class="pre">tree_edge(e,</span> <span class="pre">g)</span></tt></dt> <dd>Similar to <tt class="docutils literal"><span class="pre">examine_edge</span></tt>, this will be invoked by the process owning the source vertex and may be invoked only once. Unlike the sequential BFS, this event may be triggered even when the target has already been discovered (but by a different process). Thus, some <tt class="docutils literal"><span class="pre">non_tree_edge</span></tt> events in a sequential BFS may become <tt class="docutils literal"><span class="pre">tree_edge</span></tt> in a distributed BFS.</dd> <dt><tt class="docutils literal"><span class="pre">non_tree_edge(e,</span> <span class="pre">g)</span></tt></dt> <dd>Some <tt class="docutils literal"><span class="pre">non_tree_edge</span></tt> events in a sequential BFS may become <tt class="docutils literal"><span class="pre">tree_edge</span></tt> events in a distributed BFS. See the description of <tt class="docutils literal"><span class="pre">tree_edge</span></tt> for additional details.</dd> <dt><tt class="docutils literal"><span class="pre">gray_target(e,</span> <span class="pre">g)</span></tt></dt> <dd>As with <tt class="docutils literal"><span class="pre">tree_edge</span></tt> not knowing when another process has already discovered a vertex, <tt class="docutils literal"><span class="pre">gray_target</span></tt> events may occur in a distributed BFS when <tt class="docutils literal"><span class="pre">black_target</span></tt> events may occur in a sequential BFS, due to a lack of information on a given processor. The source of edge <tt class="docutils literal"><span class="pre">e</span></tt> will be local to the process executing this event.</dd> <dt><tt class="docutils literal"><span class="pre">black_target(e,</span> <span class="pre">g)</span></tt></dt> <dd>See documentation for <tt class="docutils literal"><span class="pre">gray_target</span></tt></dd> <dt><tt class="docutils literal"><span class="pre">finish_vertex(e,</span> <span class="pre">g)</span></tt></dt> <dd>See documentation for <tt class="docutils literal"><span class="pre">examine_vertex</span></tt>.</dd> </dl> <div class="section" id="making-visitors-safe-for-distributed-bfs"> <h2><a class="toc-backref" href="#id5">Making Visitors Safe for Distributed BFS</a></h2> <p>The three most important things to remember when updating an existing BFS visitor for distributed BFS or writing a new distributed BFS visitor are:</p> <ol class="arabic simple"> <li>Be sure that all state is either entirely local or in a distributed data structure (most likely a property map!) using the same process group as the graph.</li> <li>Be sure that the visitor doesn't require precise event sequences that cannot be guaranteed by distributed BFS, e.g., requiring <tt class="docutils literal"><span class="pre">tree_edge</span></tt> and <tt class="docutils literal"><span class="pre">non_tree_edge</span></tt> events to be completely distinct.</li> <li>Be sure that the visitor can operate on incomplete information. This often includes using an appropriate reduction operation in a <a class="reference external" href="distributed_property_map.html">distributed property map</a> and verifying that results written are "better" that what was previously written.</li> </ol> </div> <div class="section" id="distributed-bfs-visitor-example"> <h2><a class="toc-backref" href="#id6">Distributed BFS Visitor Example</a></h2> <p>To illustrate the differences between sequential and distributed BFS visitors, we consider a BFS visitor that places the distance from the source vertex to every other vertex in a property map. The sequential visitor is very simple:</p> <pre class="literal-block"> template<typename DistanceMap> struct bfs_discovery_visitor : bfs_visitor<> { bfs_discovery_visitor(DistanceMap distance) : distance(distance) {} template<typename Edge, typename Graph> void tree_edge(Edge e, const Graph& g) { std::size_t new_distance = get(distance, source(e, g)) + 1; put(distance, target(e, g), new_distance); } private: DistanceMap distance; }; </pre> <p>To revisit this code for distributed BFS, we consider the three points in the section <a class="reference internal" href="#making-visitors-safe-for-distributed-bfs">Making Visitors Safe for Distributed BFS</a>:</p> <ol class="arabic"> <li><p class="first">The distance map will need to become distributed, because the distance to each vertex should be stored in the process owning the vertex. This is actually a requirement on the user to provide such a distributed property map, although in many cases the property map will automatically be distributed and no syntactic changes will be required.</p> </li> <li><p class="first">This visitor <em>does</em> require a precise sequence of events that may change with a distributed BFS. The extraneous <tt class="docutils literal"><span class="pre">tree_edge</span></tt> events that may occur could result in attempts to put distances into the distance map multiple times for a single vertex. We therefore need to consider bullet #3.</p> </li> <li><p class="first">Since multiple distance values may be written for each vertex, we must always choose the best value we can find to update the distance map. The distributed property map <tt class="docutils literal"><span class="pre">distance</span></tt> needs to resolve distances to the smallest distance it has seen. For instance, process 0 may find vertex <tt class="docutils literal"><span class="pre">u</span></tt> at level 3 but process 1 finds it at level 5: the distance must remain at 3. To do this, we state that the property map's <em>role</em> is as a distance map, which introduces an appropriate reduction operation:</p> <pre class="literal-block"> set_property_map_role(vertex_distance, distance); </pre> </li> </ol> <p>The resulting distributed BFS visitor (which also applies, with no changes, in the sequential BFS) is very similar to our original sequential BFS visitor. Note the single-line difference in the constructor:</p> <pre class="literal-block"> template<typename DistanceMap> struct bfs_discovery_visitor : bfs_visitor<> { bfs_discovery_visitor(DistanceMap distance) : distance(distance) { set_property_map_role(vertex_distance, distance); } template<typename Edge, typename Graph> void tree_edge(Edge e, const Graph& g) { std::size_t new_distance = get(distance, source(e, g)) + 1; put(distance, target(e, g), new_distance); } private: DistanceMap distance; }; </pre> </div> </div> <div class="section" id="performance"> <h1><a class="toc-backref" href="#id7">Performance</a></h1> <p>The performance of Breadth-First Search is illustrated by the following charts. Scaling and performance is reasonable for all of the graphs we have tried.</p> <img align="left" alt="chart_php_generator_ER_SF_SW_dataset_TimeSparse_columns_4.png" class="align-left" src="chart_php_generator_ER_SF_SW_dataset_TimeSparse_columns_4.png" /> <img alt="chart_php_generator_ER_SF_SW_dataset_TimeSparse_columns_4_speedup_1.png" src="chart_php_generator_ER_SF_SW_dataset_TimeSparse_columns_4_speedup_1.png" /> <img align="left" alt="chart_php_generator_ER_SF_SW_dataset_TimeDense_columns_4.png" class="align-left" src="chart_php_generator_ER_SF_SW_dataset_TimeDense_columns_4.png" /> <img alt="chart_php_generator_ER_SF_SW_dataset_TimeDense_columns_4_speedup_1.png" src="chart_php_generator_ER_SF_SW_dataset_TimeDense_columns_4_speedup_1.png" /> <hr class="docutils" /> <p>Copyright (C) 2004 The Trustees of Indiana University.</p> <p>Authors: Douglas Gregor and Andrew Lumsdaine</p> </div> </div> <div class="footer"> <hr class="footer" /> Generated on: 2009-05-31 00:21 UTC. Generated by <a class="reference external" href="http://docutils.sourceforge.net/">Docutils</a> from <a class="reference external" href="http://docutils.sourceforge.net/rst.html">reStructuredText</a> source. </div> </body> </html>