Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/doc/html/Combining-MPI-and-Threads.html @ 82:d0c2a83c1364

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW 3.3.8 source, and a Linux build
author Chris Cannam
date Tue, 19 Nov 2019 14:52:55 +0000
parents
children
comparison
equal deleted inserted replaced
81:7029a4916348 82:d0c2a83c1364
1 <!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
2 <html>
3 <!-- This manual is for FFTW
4 (version 3.3.8, 24 May 2018).
5
6 Copyright (C) 2003 Matteo Frigo.
7
8 Copyright (C) 2003 Massachusetts Institute of Technology.
9
10 Permission is granted to make and distribute verbatim copies of this
11 manual provided the copyright notice and this permission notice are
12 preserved on all copies.
13
14 Permission is granted to copy and distribute modified versions of this
15 manual under the conditions for verbatim copying, provided that the
16 entire resulting derived work is distributed under the terms of a
17 permission notice identical to this one.
18
19 Permission is granted to copy and distribute translations of this manual
20 into another language, under the above conditions for modified versions,
21 except that this permission notice may be stated in a translation
22 approved by the Free Software Foundation. -->
23 <!-- Created by GNU Texinfo 6.3, http://www.gnu.org/software/texinfo/ -->
24 <head>
25 <title>FFTW 3.3.8: Combining MPI and Threads</title>
26
27 <meta name="description" content="FFTW 3.3.8: Combining MPI and Threads">
28 <meta name="keywords" content="FFTW 3.3.8: Combining MPI and Threads">
29 <meta name="resource-type" content="document">
30 <meta name="distribution" content="global">
31 <meta name="Generator" content="makeinfo">
32 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
33 <link href="index.html#Top" rel="start" title="Top">
34 <link href="Concept-Index.html#Concept-Index" rel="index" title="Concept Index">
35 <link href="index.html#SEC_Contents" rel="contents" title="Table of Contents">
36 <link href="Distributed_002dmemory-FFTW-with-MPI.html#Distributed_002dmemory-FFTW-with-MPI" rel="up" title="Distributed-memory FFTW with MPI">
37 <link href="FFTW-MPI-Reference.html#FFTW-MPI-Reference" rel="next" title="FFTW MPI Reference">
38 <link href="FFTW-MPI-Performance-Tips.html#FFTW-MPI-Performance-Tips" rel="prev" title="FFTW MPI Performance Tips">
39 <style type="text/css">
40 <!--
41 a.summary-letter {text-decoration: none}
42 blockquote.indentedblock {margin-right: 0em}
43 blockquote.smallindentedblock {margin-right: 0em; font-size: smaller}
44 blockquote.smallquotation {font-size: smaller}
45 div.display {margin-left: 3.2em}
46 div.example {margin-left: 3.2em}
47 div.lisp {margin-left: 3.2em}
48 div.smalldisplay {margin-left: 3.2em}
49 div.smallexample {margin-left: 3.2em}
50 div.smalllisp {margin-left: 3.2em}
51 kbd {font-style: oblique}
52 pre.display {font-family: inherit}
53 pre.format {font-family: inherit}
54 pre.menu-comment {font-family: serif}
55 pre.menu-preformatted {font-family: serif}
56 pre.smalldisplay {font-family: inherit; font-size: smaller}
57 pre.smallexample {font-size: smaller}
58 pre.smallformat {font-family: inherit; font-size: smaller}
59 pre.smalllisp {font-size: smaller}
60 span.nolinebreak {white-space: nowrap}
61 span.roman {font-family: initial; font-weight: normal}
62 span.sansserif {font-family: sans-serif; font-weight: normal}
63 ul.no-bullet {list-style: none}
64 -->
65 </style>
66
67
68 </head>
69
70 <body lang="en">
71 <a name="Combining-MPI-and-Threads"></a>
72 <div class="header">
73 <p>
74 Next: <a href="FFTW-MPI-Reference.html#FFTW-MPI-Reference" accesskey="n" rel="next">FFTW MPI Reference</a>, Previous: <a href="FFTW-MPI-Performance-Tips.html#FFTW-MPI-Performance-Tips" accesskey="p" rel="prev">FFTW MPI Performance Tips</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>
75 </div>
76 <hr>
77 <a name="Combining-MPI-and-Threads-1"></a>
78 <h3 class="section">6.11 Combining MPI and Threads</h3>
79 <a name="index-threads-2"></a>
80
81 <p>In certain cases, it may be advantageous to combine MPI
82 (distributed-memory) and threads (shared-memory) parallelization.
83 FFTW supports this, with certain caveats. For example, if you have a
84 cluster of 4-processor shared-memory nodes, you may want to use
85 threads within the nodes and MPI between the nodes, instead of MPI for
86 all parallelization.
87 </p>
88 <p>In particular, it is possible to seamlessly combine the MPI FFTW
89 routines with the multi-threaded FFTW routines (see <a href="Multi_002dthreaded-FFTW.html#Multi_002dthreaded-FFTW">Multi-threaded FFTW</a>). However, some care must be taken in the initialization code,
90 which should look something like this:
91 </p>
92 <div class="example">
93 <pre class="example">int threads_ok;
94
95 int main(int argc, char **argv)
96 {
97 int provided;
98 MPI_Init_thread(&amp;argc, &amp;argv, MPI_THREAD_FUNNELED, &amp;provided);
99 threads_ok = provided &gt;= MPI_THREAD_FUNNELED;
100
101 if (threads_ok) threads_ok = fftw_init_threads();
102 fftw_mpi_init();
103
104 ...
105 if (threads_ok) fftw_plan_with_nthreads(...);
106 ...
107
108 MPI_Finalize();
109 }
110 </pre></div>
111 <a name="index-fftw_005fmpi_005finit-3"></a>
112 <a name="index-fftw_005finit_005fthreads-2"></a>
113 <a name="index-fftw_005fplan_005fwith_005fnthreads-1"></a>
114
115 <p>First, note that instead of calling <code>MPI_Init</code>, you should call
116 <code>MPI_Init_threads</code>, which is the initialization routine defined
117 by the MPI-2 standard to indicate to MPI that your program will be
118 multithreaded. We pass <code>MPI_THREAD_FUNNELED</code>, which indicates
119 that we will only call MPI routines from the main thread. (FFTW will
120 launch additional threads internally, but the extra threads will not
121 call MPI code.) (You may also pass <code>MPI_THREAD_SERIALIZED</code> or
122 <code>MPI_THREAD_MULTIPLE</code>, which requests additional multithreading
123 support from the MPI implementation, but this is not required by
124 FFTW.) The <code>provided</code> parameter returns what level of threads
125 support is actually supported by your MPI implementation; this
126 <em>must</em> be at least <code>MPI_THREAD_FUNNELED</code> if you want to call
127 the FFTW threads routines, so we define a global variable
128 <code>threads_ok</code> to record this. You should only call
129 <code>fftw_init_threads</code> or <code>fftw_plan_with_nthreads</code> if
130 <code>threads_ok</code> is true. For more information on thread safety in
131 MPI, see the
132 <a href="http://www.mpi-forum.org/docs/mpi-20-html/node162.htm">MPI and
133 Threads</a> section of the MPI-2 standard.
134 <a name="index-thread-safety-2"></a>
135 </p>
136
137 <p>Second, we must call <code>fftw_init_threads</code> <em>before</em>
138 <code>fftw_mpi_init</code>. This is critical for technical reasons having
139 to do with how FFTW initializes its list of algorithms.
140 </p>
141 <p>Then, if you call <code>fftw_plan_with_nthreads(N)</code>, <em>every</em> MPI
142 process will launch (up to) <code>N</code> threads to parallelize its transforms.
143 </p>
144 <p>For example, in the hypothetical cluster of 4-processor nodes, you
145 might wish to launch only a single MPI process per node, and then call
146 <code>fftw_plan_with_nthreads(4)</code> on each process to use all
147 processors in the nodes.
148 </p>
149 <p>This may or may not be faster than simply using as many MPI processes
150 as you have processors, however. On the one hand, using threads
151 within a node eliminates the need for explicit message passing within
152 the node. On the other hand, FFTW&rsquo;s transpose routines are not
153 multi-threaded, and this means that the communications that do take
154 place will not benefit from parallelization within the node.
155 Moreover, many MPI implementations already have optimizations to
156 exploit shared memory when it is available, so adding the
157 multithreaded FFTW on top of this may be superfluous.
158 <a name="index-transpose-4"></a>
159 </p>
160 <hr>
161 <div class="header">
162 <p>
163 Next: <a href="FFTW-MPI-Reference.html#FFTW-MPI-Reference" accesskey="n" rel="next">FFTW MPI Reference</a>, Previous: <a href="FFTW-MPI-Performance-Tips.html#FFTW-MPI-Performance-Tips" accesskey="p" rel="prev">FFTW MPI Performance Tips</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>
164 </div>
165
166
167
168 </body>
169 </html>