[ceph.git] / ceph / src / boost / libs / graph_parallel / doc / html / mpi_bsp_process_group.html

<?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 MPI BSP Process Group</title>
<link rel="stylesheet" href="../../../../rst.css" type="text/css" />
</head>
<body>
<div class="document" id="logo-mpi-bsp-process-group">
<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> MPI BSP Process Group</h1>

<!-- Copyright (C) 2004-2009 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) -->
<div class="contents topic" id="contents">
<p class="topic-title first">Contents</p>
<ul class="simple">
<li><a class="reference internal" href="#introduction" id="id1">Introduction</a></li>
<li><a class="reference internal" href="#where-defined" id="id2">Where Defined</a></li>
<li><a class="reference internal" href="#reference" id="id3">Reference</a></li>
</ul>
</div>
<div class="section" id="introduction">
<h1><a class="toc-backref" href="#id1">Introduction</a></h1>
<p>The MPI <tt class="docutils literal"><span class="pre">mpi_process_group</span></tt> is an implementation of the <a class="reference external" href="process_group.html">process
group</a> interface using the Message Passing Interface (MPI). It is the
primary process group used in the Parallel BGL at this time.</p>
</div>
<div class="section" id="where-defined">
<h1><a class="toc-backref" href="#id2">Where Defined</a></h1>
<p>Header <tt class="docutils literal"><span class="pre">&lt;boost/graph/distributed/mpi_process_group.hpp&gt;</span></tt></p>
</div>
<div class="section" id="reference">
<h1><a class="toc-backref" href="#id3">Reference</a></h1>
<pre class="literal-block">
namespace boost { namespace graph { namespace distributed {

class mpi_process_group
{
public:
  typedef boost::mpi::communicator communicator_type;

  // Process group constructors
  mpi_process_group(communicator_type comm = communicator_type());
  mpi_process_group(std::size_t num_headers, std::size_t buffer_size,
                    communicator_type comm = communicator_type());

  mpi_process_group();
  mpi_process_group(const mpi_process_group&amp;, boost::parallel::attach_distributed_object);

  // Triggers
  template&lt;typename Type, typename Handler&gt;
    void trigger(int tag, const Handler&amp; handler);

  template&lt;typename Type, typename Handler&gt;
    void trigger_with_reply(int tag, const Handler&amp; handler);

  trigger_receive_context trigger_context() const;

  // Helper operations
  void poll();
  mpi_process_group base() const;
};

// Process query
int process_id(const mpi_process_group&amp;);
int num_processes(const mpi_process_group&amp;);

// Message transmission
template&lt;typename T&gt;
  void send(const mpi_process_group&amp; pg, int dest, int tag, const T&amp; value);

template&lt;typename T&gt;
  void receive(const mpi_process_group&amp; pg, int source, int tag, T&amp; value);

optional&lt;std::pair&lt;int, int&gt; &gt; probe(const mpi_process_group&amp; pg);

// Synchronization
void synchronize(const mpi_process_group&amp; pg);

// Out-of-band communication
template&lt;typename T&gt;
  void send_oob(const mpi_process_group&amp; pg, int dest, int tag, const T&amp; value);

template&lt;typename T, typename U&gt;
  void
  send_oob_with_reply(const mpi_process_group&amp; pg, int dest, int
                      tag, const T&amp; send_value, U&amp; receive_value);

template&lt;typename T&gt;
  void receive_oob(const mpi_process_group&amp; pg, int source, int tag, T&amp; value);

} } }
</pre>
<p>Since the <tt class="docutils literal"><span class="pre">mpi_process_group</span></tt> is an implementation of the <a class="reference external" href="process_group.html">process
group</a> interface, we omit the description of most of the functions in
the prototype. Two constructors need special mentioning:</p>
<pre class="literal-block">
mpi_process_group(communicator_type comm = communicator_type());
</pre>
<p>The constructor can take an optional MPI communicator. As default a communicator
constructed from MPI_COMM_WORLD is used.</p>
<pre class="literal-block">
mpi_process_group(std::size_t num_headers, std::size_t buffer_size,
                  communicator_type comm = communicator_type());
</pre>
<p>For performance fine tuning the maximum number of headers in a message batch
(num_headers) and the maximum combined size of batched messages (buffer_size)
can be specified. The maximum message size of a batch is
16*num_headers+buffer_size. Sensible default values have been found by optimizing
a typical application on a cluster with Ethernet network, and are num_header=64k
and buffer_size=1MB, for a total maximum batches message size of 2MB.</p>
<hr class="docutils" />
<p>Copyright (C) 2007 Douglas Gregor</p>
<p>Copyright (C) 2007 Matthias Troyer</p>
</div>
</div>
<div class="footer">
<hr class="footer" />
Generated on: 2009-05-31 00:22 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>
Commit	Line	Data
7c673cae FG	1	<?xml version="1.0" encoding="utf-8" ?>
	2	<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
	3	<html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en" lang="en">
	4	<head>
	5	<meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
	6	<meta name="generator" content="Docutils 0.6: http://docutils.sourceforge.net/" />
	7	<title>Parallel BGL MPI BSP Process Group</title>
	8	<link rel="stylesheet" href="../../../../rst.css" type="text/css" />
	9	</head>
	10	<body>
	11	<div class="document" id="logo-mpi-bsp-process-group">
	12	<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> MPI BSP Process Group</h1>
	13
	14	<!-- Copyright (C) 2004-2009 The Trustees of Indiana University.
	15	Use, modification and distribution is subject to the Boost Software
	16	License, Version 1.0. (See accompanying file LICENSE_1_0.txt or copy at
	17	http://www.boost.org/LICENSE_1_0.txt) -->
	18	<div class="contents topic" id="contents">
	19	<p class="topic-title first">Contents</p>
	20	<ul class="simple">
	21	<li><a class="reference internal" href="#introduction" id="id1">Introduction</a></li>
	22	<li><a class="reference internal" href="#where-defined" id="id2">Where Defined</a></li>
	23	<li><a class="reference internal" href="#reference" id="id3">Reference</a></li>
	24	</ul>
	25	</div>
	26	<div class="section" id="introduction">
	27	<h1><a class="toc-backref" href="#id1">Introduction</a></h1>
	28	<p>The MPI <tt class="docutils literal"><span class="pre">mpi_process_group</span></tt> is an implementation of the <a class="reference external" href="process_group.html">process
	29	group</a> interface using the Message Passing Interface (MPI). It is the
	30	primary process group used in the Parallel BGL at this time.</p>
	31	</div>
	32	<div class="section" id="where-defined">
	33	<h1><a class="toc-backref" href="#id2">Where Defined</a></h1>
	34	<p>Header <tt class="docutils literal"><span class="pre"><boost/graph/distributed/mpi_process_group.hpp></span></tt></p>
	35	</div>
	36	<div class="section" id="reference">
	37	<h1><a class="toc-backref" href="#id3">Reference</a></h1>
	38	<pre class="literal-block">
	39	namespace boost { namespace graph { namespace distributed {
	40
	41	class mpi_process_group
	42	{
	43	public:
	44	typedef boost::mpi::communicator communicator_type;
	45
	46	// Process group constructors
	47	mpi_process_group(communicator_type comm = communicator_type());
	48	mpi_process_group(std::size_t num_headers, std::size_t buffer_size,
	49	communicator_type comm = communicator_type());
	50
	51	mpi_process_group();
	52	mpi_process_group(const mpi_process_group&, boost::parallel::attach_distributed_object);
	53
	54	// Triggers
	55	template<typename Type, typename Handler>
	56	void trigger(int tag, const Handler& handler);
	57
	58	template<typename Type, typename Handler>
	59	void trigger_with_reply(int tag, const Handler& handler);
	60
	61	trigger_receive_context trigger_context() const;
	62
	63	// Helper operations
	64	void poll();
65	mpi_process_group base() const;
66	};
67
68	// Process query
69	int process_id(const mpi_process_group&);
70	int num_processes(const mpi_process_group&);
71
72	// Message transmission
73	template<typename T>
74	void send(const mpi_process_group& pg, int dest, int tag, const T& value);
75
76	template<typename T>
77	void receive(const mpi_process_group& pg, int source, int tag, T& value);
78
79	optional<std::pair<int, int> > probe(const mpi_process_group& pg);
80
81	// Synchronization
82	void synchronize(const mpi_process_group& pg);
83
84	// Out-of-band communication
85	template<typename T>
86	void send_oob(const mpi_process_group& pg, int dest, int tag, const T& value);
87
88	template<typename T, typename U>
89	void
90	send_oob_with_reply(const mpi_process_group& pg, int dest, int
91	tag, const T& send_value, U& receive_value);
92
93	template<typename T>
94	void receive_oob(const mpi_process_group& pg, int source, int tag, T& value);
95
96	} } }
97	</pre>
98	<p>Since the <tt class="docutils literal"><span class="pre">mpi_process_group</span></tt> is an implementation of the <a class="reference external" href="process_group.html">process
99	group</a> interface, we omit the description of most of the functions in
100	the prototype. Two constructors need special mentioning:</p>
101	<pre class="literal-block">
102	mpi_process_group(communicator_type comm = communicator_type());
103	</pre>
104	<p>The constructor can take an optional MPI communicator. As default a communicator
105	constructed from MPI_COMM_WORLD is used.</p>
106	<pre class="literal-block">
107	mpi_process_group(std::size_t num_headers, std::size_t buffer_size,
108	communicator_type comm = communicator_type());
109	</pre>
110	<p>For performance fine tuning the maximum number of headers in a message batch
111	(num_headers) and the maximum combined size of batched messages (buffer_size)
112	can be specified. The maximum message size of a batch is
113	16*num_headers+buffer_size. Sensible default values have been found by optimizing
114	a typical application on a cluster with Ethernet network, and are num_header=64k
115	and buffer_size=1MB, for a total maximum batches message size of 2MB.</p>
116	<hr class="docutils" />
117	<p>Copyright (C) 2007 Douglas Gregor</p>
118	<p>Copyright (C) 2007 Matthias Troyer</p>
119	</div>
120	</div>
121	<div class="footer">
122	<hr class="footer" />
123	Generated on: 2009-05-31 00:22 UTC.
124	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.
125
126	</div>
127	</body>
128	</html>