Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.5/doc/html/FFTW-MPI-Wisdom.html @ 79:91c729825bca pa_catalina

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Update build for AUDIO_COMPONENT_FIX
author Chris Cannam
date Wed, 30 Oct 2019 12:40:34 +0000
parents 2cd0e3b3e1fd
children
rev   line source
Chris@42 1 <!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
Chris@42 2 <html>
Chris@42 3 <!-- This manual is for FFTW
Chris@42 4 (version 3.3.5, 30 July 2016).
Chris@42 5
Chris@42 6 Copyright (C) 2003 Matteo Frigo.
Chris@42 7
Chris@42 8 Copyright (C) 2003 Massachusetts Institute of Technology.
Chris@42 9
Chris@42 10 Permission is granted to make and distribute verbatim copies of this
Chris@42 11 manual provided the copyright notice and this permission notice are
Chris@42 12 preserved on all copies.
Chris@42 13
Chris@42 14 Permission is granted to copy and distribute modified versions of this
Chris@42 15 manual under the conditions for verbatim copying, provided that the
Chris@42 16 entire resulting derived work is distributed under the terms of a
Chris@42 17 permission notice identical to this one.
Chris@42 18
Chris@42 19 Permission is granted to copy and distribute translations of this manual
Chris@42 20 into another language, under the above conditions for modified versions,
Chris@42 21 except that this permission notice may be stated in a translation
Chris@42 22 approved by the Free Software Foundation. -->
Chris@42 23 <!-- Created by GNU Texinfo 5.2, http://www.gnu.org/software/texinfo/ -->
Chris@42 24 <head>
Chris@42 25 <title>FFTW 3.3.5: FFTW MPI Wisdom</title>
Chris@42 26
Chris@42 27 <meta name="description" content="FFTW 3.3.5: FFTW MPI Wisdom">
Chris@42 28 <meta name="keywords" content="FFTW 3.3.5: FFTW MPI Wisdom">
Chris@42 29 <meta name="resource-type" content="document">
Chris@42 30 <meta name="distribution" content="global">
Chris@42 31 <meta name="Generator" content="makeinfo">
Chris@42 32 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
Chris@42 33 <link href="index.html#Top" rel="start" title="Top">
Chris@42 34 <link href="Concept-Index.html#Concept-Index" rel="index" title="Concept Index">
Chris@42 35 <link href="index.html#SEC_Contents" rel="contents" title="Table of Contents">
Chris@42 36 <link href="Distributed_002dmemory-FFTW-with-MPI.html#Distributed_002dmemory-FFTW-with-MPI" rel="up" title="Distributed-memory FFTW with MPI">
Chris@42 37 <link href="Avoiding-MPI-Deadlocks.html#Avoiding-MPI-Deadlocks" rel="next" title="Avoiding MPI Deadlocks">
Chris@42 38 <link href="An-improved-replacement-for-MPI_005fAlltoall.html#An-improved-replacement-for-MPI_005fAlltoall" rel="prev" title="An improved replacement for MPI_Alltoall">
Chris@42 39 <style type="text/css">
Chris@42 40 <!--
Chris@42 41 a.summary-letter {text-decoration: none}
Chris@42 42 blockquote.smallquotation {font-size: smaller}
Chris@42 43 div.display {margin-left: 3.2em}
Chris@42 44 div.example {margin-left: 3.2em}
Chris@42 45 div.indentedblock {margin-left: 3.2em}
Chris@42 46 div.lisp {margin-left: 3.2em}
Chris@42 47 div.smalldisplay {margin-left: 3.2em}
Chris@42 48 div.smallexample {margin-left: 3.2em}
Chris@42 49 div.smallindentedblock {margin-left: 3.2em; font-size: smaller}
Chris@42 50 div.smalllisp {margin-left: 3.2em}
Chris@42 51 kbd {font-style:oblique}
Chris@42 52 pre.display {font-family: inherit}
Chris@42 53 pre.format {font-family: inherit}
Chris@42 54 pre.menu-comment {font-family: serif}
Chris@42 55 pre.menu-preformatted {font-family: serif}
Chris@42 56 pre.smalldisplay {font-family: inherit; font-size: smaller}
Chris@42 57 pre.smallexample {font-size: smaller}
Chris@42 58 pre.smallformat {font-family: inherit; font-size: smaller}
Chris@42 59 pre.smalllisp {font-size: smaller}
Chris@42 60 span.nocodebreak {white-space:nowrap}
Chris@42 61 span.nolinebreak {white-space:nowrap}
Chris@42 62 span.roman {font-family:serif; font-weight:normal}
Chris@42 63 span.sansserif {font-family:sans-serif; font-weight:normal}
Chris@42 64 ul.no-bullet {list-style: none}
Chris@42 65 -->
Chris@42 66 </style>
Chris@42 67
Chris@42 68
Chris@42 69 </head>
Chris@42 70
Chris@42 71 <body lang="en" bgcolor="#FFFFFF" text="#000000" link="#0000FF" vlink="#800080" alink="#FF0000">
Chris@42 72 <a name="FFTW-MPI-Wisdom"></a>
Chris@42 73 <div class="header">
Chris@42 74 <p>
Chris@42 75 Next: <a href="Avoiding-MPI-Deadlocks.html#Avoiding-MPI-Deadlocks" accesskey="n" rel="next">Avoiding MPI Deadlocks</a>, Previous: <a href="FFTW-MPI-Transposes.html#FFTW-MPI-Transposes" accesskey="p" rel="prev">FFTW MPI Transposes</a>, Up: <a href="Distributed_002dmemory-FFTW-with-MPI.html#Distributed_002dmemory-FFTW-with-MPI" accesskey="u" rel="up">Distributed-memory FFTW with MPI</a> &nbsp; [<a href="index.html#SEC_Contents" title="Table of contents" rel="contents">Contents</a>][<a href="Concept-Index.html#Concept-Index" title="Index" rel="index">Index</a>]</p>
Chris@42 76 </div>
Chris@42 77 <hr>
Chris@42 78 <a name="FFTW-MPI-Wisdom-1"></a>
Chris@42 79 <h3 class="section">6.8 FFTW MPI Wisdom</h3>
Chris@42 80 <a name="index-wisdom-2"></a>
Chris@42 81 <a name="index-saving-plans-to-disk-2"></a>
Chris@42 82
Chris@42 83 <p>FFTW&rsquo;s &ldquo;wisdom&rdquo; facility (see <a href="Words-of-Wisdom_002dSaving-Plans.html#Words-of-Wisdom_002dSaving-Plans">Words of Wisdom-Saving Plans</a>) can
Chris@42 84 be used to save MPI plans as well as to save uniprocessor plans.
Chris@42 85 However, for MPI there are several unavoidable complications.
Chris@42 86 </p>
Chris@42 87 <a name="index-MPI-I_002fO-1"></a>
Chris@42 88 <p>First, the MPI standard does not guarantee that every process can
Chris@42 89 perform file I/O (at least, not using C stdio routines)&mdash;in general,
Chris@42 90 we may only assume that process 0 is capable of I/O.<a name="DOCF7" href="#FOOT7"><sup>7</sup></a> So, if we
Chris@42 91 want to export the wisdom from a single process to a file, we must
Chris@42 92 first export the wisdom to a string, then send it to process 0, then
Chris@42 93 write it to a file.
Chris@42 94 </p>
Chris@42 95 <p>Second, in principle we may want to have separate wisdom for every
Chris@42 96 process, since in general the processes may run on different hardware
Chris@42 97 even for a single MPI program. However, in practice FFTW&rsquo;s MPI code
Chris@42 98 is designed for the case of homogeneous hardware (see <a href="Load-balancing.html#Load-balancing">Load balancing</a>), and in this case it is convenient to use the same wisdom
Chris@42 99 for every process. Thus, we need a mechanism to synchronize the wisdom.
Chris@42 100 </p>
Chris@42 101 <p>To address both of these problems, FFTW provides the following two
Chris@42 102 functions:
Chris@42 103 </p>
Chris@42 104 <div class="example">
Chris@42 105 <pre class="example">void fftw_mpi_broadcast_wisdom(MPI_Comm comm);
Chris@42 106 void fftw_mpi_gather_wisdom(MPI_Comm comm);
Chris@42 107 </pre></div>
Chris@42 108 <a name="index-fftw_005fmpi_005fgather_005fwisdom"></a>
Chris@42 109 <a name="index-fftw_005fmpi_005fbroadcast_005fwisdom"></a>
Chris@42 110
Chris@42 111 <p>Given a communicator <code>comm</code>, <code>fftw_mpi_broadcast_wisdom</code>
Chris@42 112 will broadcast the wisdom from process 0 to all other processes.
Chris@42 113 Conversely, <code>fftw_mpi_gather_wisdom</code> will collect wisdom from all
Chris@42 114 processes onto process 0. (If the plans created for the same problem
Chris@42 115 by different processes are not the same, <code>fftw_mpi_gather_wisdom</code>
Chris@42 116 will arbitrarily choose one of the plans.) Both of these functions
Chris@42 117 may result in suboptimal plans for different processes if the
Chris@42 118 processes are running on non-identical hardware. Both of these
Chris@42 119 functions are <em>collective</em> calls, which means that they must be
Chris@42 120 executed by all processes in the communicator.
Chris@42 121 <a name="index-collective-function-1"></a>
Chris@42 122 </p>
Chris@42 123
Chris@42 124 <p>So, for example, a typical code snippet to import wisdom from a file
Chris@42 125 and use it on all processes would be:
Chris@42 126 </p>
Chris@42 127 <div class="example">
Chris@42 128 <pre class="example">{
Chris@42 129 int rank;
Chris@42 130
Chris@42 131 fftw_mpi_init();
Chris@42 132 MPI_Comm_rank(MPI_COMM_WORLD, &amp;rank);
Chris@42 133 if (rank == 0) fftw_import_wisdom_from_filename(&quot;mywisdom&quot;);
Chris@42 134 fftw_mpi_broadcast_wisdom(MPI_COMM_WORLD);
Chris@42 135 }
Chris@42 136 </pre></div>
Chris@42 137
Chris@42 138 <p>(Note that we must call <code>fftw_mpi_init</code> before importing any
Chris@42 139 wisdom that might contain MPI plans.) Similarly, a typical code
Chris@42 140 snippet to export wisdom from all processes to a file is:
Chris@42 141 <a name="index-fftw_005fmpi_005finit-2"></a>
Chris@42 142 </p>
Chris@42 143 <div class="example">
Chris@42 144 <pre class="example">{
Chris@42 145 int rank;
Chris@42 146
Chris@42 147 fftw_mpi_gather_wisdom(MPI_COMM_WORLD);
Chris@42 148 MPI_Comm_rank(MPI_COMM_WORLD, &amp;rank);
Chris@42 149 if (rank == 0) fftw_export_wisdom_to_filename(&quot;mywisdom&quot;);
Chris@42 150 }
Chris@42 151 </pre></div>
Chris@42 152
Chris@42 153 <div class="footnote">
Chris@42 154 <hr>
Chris@42 155 <h4 class="footnotes-heading">Footnotes</h4>
Chris@42 156
Chris@42 157 <h3><a name="FOOT7" href="#DOCF7">(7)</a></h3>
Chris@42 158 <p>In fact,
Chris@42 159 even this assumption is not technically guaranteed by the standard,
Chris@42 160 although it seems to be universal in actual MPI implementations and is
Chris@42 161 widely assumed by MPI-using software. Technically, you need to query
Chris@42 162 the <code>MPI_IO</code> attribute of <code>MPI_COMM_WORLD</code> with
Chris@42 163 <code>MPI_Attr_get</code>. If this attribute is <code>MPI_PROC_NULL</code>, no
Chris@42 164 I/O is possible. If it is <code>MPI_ANY_SOURCE</code>, any process can
Chris@42 165 perform I/O. Otherwise, it is the rank of a process that can perform
Chris@42 166 I/O ... but since it is not guaranteed to yield the <em>same</em> rank
Chris@42 167 on all processes, you have to do an <code>MPI_Allreduce</code> of some kind
Chris@42 168 if you want all processes to agree about which is going to do I/O.
Chris@42 169 And even then, the standard only guarantees that this process can
Chris@42 170 perform output, but not input. See e.g. <cite>Parallel Programming
Chris@42 171 with MPI</cite> by P. S. Pacheco, section 8.1.3. Needless to say, in our
Chris@42 172 experience virtually no MPI programmers worry about this.</p>
Chris@42 173 </div>
Chris@42 174 <hr>
Chris@42 175 <div class="header">
Chris@42 176 <p>
Chris@42 177 Next: <a href="Avoiding-MPI-Deadlocks.html#Avoiding-MPI-Deadlocks" accesskey="n" rel="next">Avoiding MPI Deadlocks</a>, Previous: <a href="FFTW-MPI-Transposes.html#FFTW-MPI-Transposes" accesskey="p" rel="prev">FFTW MPI Transposes</a>, Up: <a href="Distributed_002dmemory-FFTW-with-MPI.html#Distributed_002dmemory-FFTW-with-MPI" accesskey="u" rel="up">Distributed-memory FFTW with MPI</a> &nbsp; [<a href="index.html#SEC_Contents" title="Table of contents" rel="contents">Contents</a>][<a href="Concept-Index.html#Concept-Index" title="Index" rel="index">Index</a>]</p>
Chris@42 178 </div>
Chris@42 179
Chris@42 180
Chris@42 181
Chris@42 182 </body>
Chris@42 183 </html>