furnace/extern/fftw/doc/html/An-improved-replacement-for...

128 lines
5.3 KiB
HTML

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
<html>
<!-- This manual is for FFTW
(version 3.3.10, 10 December 2020).
Copyright (C) 2003 Matteo Frigo.
Copyright (C) 2003 Massachusetts Institute of Technology.
Permission is granted to make and distribute verbatim copies of this
manual provided the copyright notice and this permission notice are
preserved on all copies.
Permission is granted to copy and distribute modified versions of this
manual under the conditions for verbatim copying, provided that the
entire resulting derived work is distributed under the terms of a
permission notice identical to this one.
Permission is granted to copy and distribute translations of this manual
into another language, under the above conditions for modified versions,
except that this permission notice may be stated in a translation
approved by the Free Software Foundation. -->
<!-- Created by GNU Texinfo 6.7, http://www.gnu.org/software/texinfo/ -->
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
<title>An improved replacement for MPI_Alltoall (FFTW 3.3.10)</title>
<meta name="description" content="An improved replacement for MPI_Alltoall (FFTW 3.3.10)">
<meta name="keywords" content="An improved replacement for MPI_Alltoall (FFTW 3.3.10)">
<meta name="resource-type" content="document">
<meta name="distribution" content="global">
<meta name="Generator" content="makeinfo">
<link href="index.html" rel="start" title="Top">
<link href="Concept-Index.html" rel="index" title="Concept Index">
<link href="index.html#SEC_Contents" rel="contents" title="Table of Contents">
<link href="FFTW-MPI-Transposes.html" rel="up" title="FFTW MPI Transposes">
<link href="FFTW-MPI-Wisdom.html" rel="next" title="FFTW MPI Wisdom">
<link href="Advanced-distributed_002dtranspose-interface.html" rel="prev" title="Advanced distributed-transpose interface">
<style type="text/css">
<!--
a.summary-letter {text-decoration: none}
blockquote.indentedblock {margin-right: 0em}
div.display {margin-left: 3.2em}
div.example {margin-left: 3.2em}
div.lisp {margin-left: 3.2em}
kbd {font-style: oblique}
pre.display {font-family: inherit}
pre.format {font-family: inherit}
pre.menu-comment {font-family: serif}
pre.menu-preformatted {font-family: serif}
span.nolinebreak {white-space: nowrap}
span.roman {font-family: initial; font-weight: normal}
span.sansserif {font-family: sans-serif; font-weight: normal}
ul.no-bullet {list-style: none}
-->
</style>
</head>
<body lang="en">
<span id="An-improved-replacement-for-MPI_005fAlltoall"></span><div class="header">
<p>
Previous: <a href="Advanced-distributed_002dtranspose-interface.html" accesskey="p" rel="prev">Advanced distributed-transpose interface</a>, Up: <a href="FFTW-MPI-Transposes.html" accesskey="u" rel="up">FFTW MPI Transposes</a> &nbsp; [<a href="index.html#SEC_Contents" title="Table of contents" rel="contents">Contents</a>][<a href="Concept-Index.html" title="Index" rel="index">Index</a>]</p>
</div>
<hr>
<span id="An-improved-replacement-for-MPI_005fAlltoall-1"></span><h4 class="subsection">6.7.3 An improved replacement for MPI_Alltoall</h4>
<p>We close this section by noting that FFTW&rsquo;s MPI transpose routines can
be thought of as a generalization for the <code>MPI_Alltoall</code> function
(albeit only for floating-point types), and in some circumstances can
function as an improved replacement.
<span id="index-MPI_005fAlltoall"></span>
</p>
<p><code>MPI_Alltoall</code> is defined by the MPI standard as:
</p>
<div class="example">
<pre class="example">int MPI_Alltoall(void *sendbuf, int sendcount, MPI_Datatype sendtype,
void *recvbuf, int recvcnt, MPI_Datatype recvtype,
MPI_Comm comm);
</pre></div>
<p>In particular, for <code>double*</code> arrays <code>in</code> and <code>out</code>,
consider the call:
</p>
<div class="example">
<pre class="example">MPI_Alltoall(in, howmany, MPI_DOUBLE, out, howmany MPI_DOUBLE, comm);
</pre></div>
<p>This is completely equivalent to:
</p>
<div class="example">
<pre class="example">MPI_Comm_size(comm, &amp;P);
plan = fftw_mpi_plan_many_transpose(P, P, howmany, 1, 1, in, out, comm, FFTW_ESTIMATE);
fftw_execute(plan);
fftw_destroy_plan(plan);
</pre></div>
<p>That is, computing a P&nbsp;&times;&nbsp;P
transpose on <code>P</code> processes,
with a block size of 1, is just a standard all-to-all communication.
</p>
<p>However, using the FFTW routine instead of <code>MPI_Alltoall</code> may
have certain advantages. First of all, FFTW&rsquo;s routine can operate
in-place (<code>in == out</code>) whereas <code>MPI_Alltoall</code> can only
operate out-of-place.
<span id="index-in_002dplace-8"></span>
</p>
<p>Second, even for out-of-place plans, FFTW&rsquo;s routine may be faster,
especially if you need to perform the all-to-all communication many
times and can afford to use <code>FFTW_MEASURE</code> or
<code>FFTW_PATIENT</code>. It should certainly be no slower, not including
the time to create the plan, since one of the possible algorithms that
FFTW uses for an out-of-place transpose <em>is</em> simply to call
<code>MPI_Alltoall</code>. However, FFTW also considers several other
possible algorithms that, depending on your MPI implementation and
your hardware, may be faster.
<span id="index-FFTW_005fMEASURE-3"></span>
<span id="index-FFTW_005fPATIENT-4"></span>
</p>
</body>
</html>