Mercurial > hg > sv-dependency-builds

<html lang="en">
<head>
<title>2d MPI example - FFTW 3.3.3</title>
<meta http-equiv="Content-Type" content="text/html">
<meta name="description" content="FFTW 3.3.3">
<meta name="generator" content="makeinfo 4.13">
<link title="Top" rel="start" href="index.html#Top">
<link rel="up" href="Distributed_002dmemory-FFTW-with-MPI.html#Distributed_002dmemory-FFTW-with-MPI" title="Distributed-memory FFTW with MPI">
<link rel="prev" href="Linking-and-Initializing-MPI-FFTW.html#Linking-and-Initializing-MPI-FFTW" title="Linking and Initializing MPI FFTW">
<link rel="next" href="MPI-Data-Distribution.html#MPI-Data-Distribution" title="MPI Data Distribution">
<link href="http://www.gnu.org/software/texinfo/" rel="generator-home" title="Texinfo Homepage">
<!--
This manual is for FFTW
(version 3.3.3, 25 November 2012).

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.
   -->
<meta http-equiv="Content-Style-Type" content="text/css">
<style type="text/css"><!--
  pre.display { font-family:inherit }
  pre.format  { font-family:inherit }
  pre.smalldisplay { font-family:inherit; font-size:smaller }
  pre.smallformat  { font-family:inherit; font-size:smaller }
  pre.smallexample { font-size:smaller }
  pre.smalllisp    { font-size:smaller }
  span.sc    { font-variant:small-caps }
  span.roman { font-family:serif; font-weight:normal; }
  span.sansserif { font-family:sans-serif; font-weight:normal; }
--></style>
</head>
<body>
<div class="node">
<a name="g_t2d-MPI-example"></a>
<p>
Next:&nbsp;<a rel="next" accesskey="n" href="MPI-Data-Distribution.html#MPI-Data-Distribution">MPI Data Distribution</a>,
Previous:&nbsp;<a rel="previous" accesskey="p" href="Linking-and-Initializing-MPI-FFTW.html#Linking-and-Initializing-MPI-FFTW">Linking and Initializing MPI FFTW</a>,
Up:&nbsp;<a rel="up" accesskey="u" href="Distributed_002dmemory-FFTW-with-MPI.html#Distributed_002dmemory-FFTW-with-MPI">Distributed-memory FFTW with MPI</a>
<hr>
</div>

<h3 class="section">6.3 2d MPI example</h3>

<p>Before we document the FFTW MPI interface in detail, we begin with a
simple example outlining how one would perform a two-dimensional
<code>N0</code> by <code>N1</code> complex DFT.

<pre class="example">     #include &lt;fftw3-mpi.h&gt;

     int main(int argc, char **argv)
     {
         const ptrdiff_t N0 = ..., N1 = ...;
         fftw_plan plan;
         fftw_complex *data;
         ptrdiff_t alloc_local, local_n0, local_0_start, i, j;

         MPI_Init(&amp;argc, &amp;argv);
         fftw_mpi_init();

         /* <span class="roman">get local data size and allocate</span> */
         alloc_local = fftw_mpi_local_size_2d(N0, N1, MPI_COMM_WORLD,
                                              &amp;local_n0, &amp;local_0_start);
         data = fftw_alloc_complex(alloc_local);

         /* <span class="roman">create plan for in-place forward DFT</span> */
         plan = fftw_mpi_plan_dft_2d(N0, N1, data, data, MPI_COMM_WORLD,
                                     FFTW_FORWARD, FFTW_ESTIMATE);

         /* <span class="roman">initialize data to some function</span> my_function(x,y) */
         for (i = 0; i &lt; local_n0; ++i) for (j = 0; j &lt; N1; ++j)
            data[i*N1 + j] = my_function(local_0_start + i, j);

         /* <span class="roman">compute transforms, in-place, as many times as desired</span> */
         fftw_execute(plan);

         fftw_destroy_plan(plan);

         MPI_Finalize();
     }
</pre>
   <p>As can be seen above, the MPI interface follows the same basic style
of allocate/plan/execute/destroy as the serial FFTW routines.  All of
the MPI-specific routines are prefixed with &lsquo;<samp><span class="samp">fftw_mpi_</span></samp>&rsquo; instead
of &lsquo;<samp><span class="samp">fftw_</span></samp>&rsquo;.  There are a few important differences, however:

   <p>First, we must call <code>fftw_mpi_init()</code> after calling
<code>MPI_Init</code> (required in all MPI programs) and before calling any
other &lsquo;<samp><span class="samp">fftw_mpi_</span></samp>&rsquo; routine.
<a name="index-MPI_005fInit-357"></a><a name="index-fftw_005fmpi_005finit-358"></a>

   <p>Second, when we create the plan with <code>fftw_mpi_plan_dft_2d</code>,
analogous to <code>fftw_plan_dft_2d</code>, we pass an additional argument:
the communicator, indicating which processes will participate in the
transform (here <code>MPI_COMM_WORLD</code>, indicating all processes).
Whenever you create, execute, or destroy a plan for an MPI transform,
you must call the corresponding FFTW routine on <em>all</em> processes
in the communicator for that transform.  (That is, these are
<em>collective</em> calls.)  Note that the plan for the MPI transform
uses the standard <code>fftw_execute</code> and <code>fftw_destroy</code> routines
(on the other hand, there are MPI-specific new-array execute functions
documented below).
<a name="index-collective-function-359"></a><a name="index-fftw_005fmpi_005fplan_005fdft_005f2d-360"></a><a name="index-MPI_005fCOMM_005fWORLD-361"></a>

   <p>Third, all of the FFTW MPI routines take <code>ptrdiff_t</code> arguments
instead of <code>int</code> as for the serial FFTW.  <code>ptrdiff_t</code> is a
standard C integer type which is (at least) 32 bits wide on a 32-bit
machine and 64 bits wide on a 64-bit machine.  This is to make it easy
to specify very large parallel transforms on a 64-bit machine.  (You
can specify 64-bit transform sizes in the serial FFTW, too, but only
by using the &lsquo;<samp><span class="samp">guru64</span></samp>&rsquo; planner interface.  See <a href="64_002dbit-Guru-Interface.html#g_t64_002dbit-Guru-Interface">64-bit Guru Interface</a>.)
<a name="index-ptrdiff_005ft-362"></a><a name="index-g_t64_002dbit-architecture-363"></a>

   <p>Fourth, and most importantly, you don't allocate the entire
two-dimensional array on each process.  Instead, you call
<code>fftw_mpi_local_size_2d</code> to find out what <em>portion</em> of the
array resides on each processor, and how much space to allocate.
Here, the portion of the array on each process is a <code>local_n0</code> by
<code>N1</code> slice of the total array, starting at index
<code>local_0_start</code>.  The total number of <code>fftw_complex</code> numbers
to allocate is given by the <code>alloc_local</code> return value, which
<em>may</em> be greater than <code>local_n0 * N1</code> (in case some
intermediate calculations require additional storage).  The data
distribution in FFTW's MPI interface is described in more detail by
the next section.
<a name="index-fftw_005fmpi_005flocal_005fsize_005f2d-364"></a><a name="index-data-distribution-365"></a>

   <p>Given the portion of the array that resides on the local process, it
is straightforward to initialize the data (here to a function
<code>myfunction</code>) and otherwise manipulate it.  Of course, at the end
of the program you may want to output the data somehow, but
synchronizing this output is up to you and is beyond the scope of this
manual.  (One good way to output a large multi-dimensional distributed
array in MPI to a portable binary file is to use the free HDF5
library; see the <a href="http://www.hdfgroup.org/">HDF home page</a>.)
<a name="index-HDF5-366"></a><a name="index-MPI-I_002fO-367"></a>
<!--  -->

   </body></html>
author	Chris Cannam <cannam@all-day-breakfast.com>
date	Mon, 02 Mar 2020 14:03:47 +0000
parents	89f5e221ed7b
children