|
d@0
|
1 <html lang="en">
|
|
d@0
|
2 <head>
|
|
d@0
|
3 <title>Basic and advanced distribution interfaces - FFTW 3.2alpha3</title>
|
|
d@0
|
4 <meta http-equiv="Content-Type" content="text/html">
|
|
d@0
|
5 <meta name="description" content="FFTW 3.2alpha3">
|
|
d@0
|
6 <meta name="generator" content="makeinfo 4.8">
|
|
d@0
|
7 <link title="Top" rel="start" href="index.html#Top">
|
|
d@0
|
8 <link rel="up" href="MPI-data-distribution.html#MPI-data-distribution" title="MPI data distribution">
|
|
d@0
|
9 <link rel="prev" href="MPI-data-distribution.html#MPI-data-distribution" title="MPI data distribution">
|
|
d@0
|
10 <link rel="next" href="Load-balancing.html#Load-balancing" title="Load balancing">
|
|
d@0
|
11 <link href="http://www.gnu.org/software/texinfo/" rel="generator-home" title="Texinfo Homepage">
|
|
d@0
|
12 <!--
|
|
d@0
|
13 This manual is for FFTW
|
|
d@0
|
14 (version 3.2alpha3, 14 August 2007).
|
|
d@0
|
15
|
|
d@0
|
16 Copyright (C) 2003 Matteo Frigo.
|
|
d@0
|
17
|
|
d@0
|
18 Copyright (C) 2003 Massachusetts Institute of Technology.
|
|
d@0
|
19
|
|
d@0
|
20 Permission is granted to make and distribute verbatim copies of
|
|
d@0
|
21 this manual provided the copyright notice and this permission
|
|
d@0
|
22 notice are preserved on all copies.
|
|
d@0
|
23
|
|
d@0
|
24 Permission is granted to copy and distribute modified versions of
|
|
d@0
|
25 this manual under the conditions for verbatim copying, provided
|
|
d@0
|
26 that the entire resulting derived work is distributed under the
|
|
d@0
|
27 terms of a permission notice identical to this one.
|
|
d@0
|
28
|
|
d@0
|
29 Permission is granted to copy and distribute translations of this
|
|
d@0
|
30 manual into another language, under the above conditions for
|
|
d@0
|
31 modified versions, except that this permission notice may be
|
|
d@0
|
32 stated in a translation approved by the Free Software Foundation.
|
|
d@0
|
33 -->
|
|
d@0
|
34 <meta http-equiv="Content-Style-Type" content="text/css">
|
|
d@0
|
35 <style type="text/css"><!--
|
|
d@0
|
36 pre.display { font-family:inherit }
|
|
d@0
|
37 pre.format { font-family:inherit }
|
|
d@0
|
38 pre.smalldisplay { font-family:inherit; font-size:smaller }
|
|
d@0
|
39 pre.smallformat { font-family:inherit; font-size:smaller }
|
|
d@0
|
40 pre.smallexample { font-size:smaller }
|
|
d@0
|
41 pre.smalllisp { font-size:smaller }
|
|
d@0
|
42 span.sc { font-variant:small-caps }
|
|
d@0
|
43 span.roman { font-family:serif; font-weight:normal; }
|
|
d@0
|
44 span.sansserif { font-family:sans-serif; font-weight:normal; }
|
|
d@0
|
45 --></style>
|
|
d@0
|
46 </head>
|
|
d@0
|
47 <body>
|
|
d@0
|
48 <div class="node">
|
|
d@0
|
49 <p>
|
|
d@0
|
50 <a name="Basic-and-advanced-distribution-interfaces"></a>
|
|
d@0
|
51 Next: <a rel="next" accesskey="n" href="Load-balancing.html#Load-balancing">Load balancing</a>,
|
|
d@0
|
52 Previous: <a rel="previous" accesskey="p" href="MPI-data-distribution.html#MPI-data-distribution">MPI data distribution</a>,
|
|
d@0
|
53 Up: <a rel="up" accesskey="u" href="MPI-data-distribution.html#MPI-data-distribution">MPI data distribution</a>
|
|
d@0
|
54 <hr>
|
|
d@0
|
55 </div>
|
|
d@0
|
56
|
|
d@0
|
57 <h4 class="subsection">6.4.1 Basic and advanced distribution interfaces</h4>
|
|
d@0
|
58
|
|
d@0
|
59 <p>As with the planner interface, the `<samp><span class="samp">fftw_mpi_local_size</span></samp>'
|
|
d@0
|
60 distribution interface is broken into basic and advanced
|
|
d@0
|
61 (`<samp><span class="samp">_many</span></samp>') interfaces, where the latter allows you to specify the
|
|
d@0
|
62 block size manually and also to request block sizes when computing
|
|
d@0
|
63 multiple transforms simultaneously. These functions are documented
|
|
d@0
|
64 more exhaustively by the FFTW MPI Reference, but we summarize the
|
|
d@0
|
65 basic ideas here using a couple of two-dimensional examples.
|
|
d@0
|
66
|
|
d@0
|
67 <p>For the 100 × 200 complex-DFT example, above, we would find
|
|
d@0
|
68 the distribution by calling the following function in the basic
|
|
d@0
|
69 interface:
|
|
d@0
|
70
|
|
d@0
|
71 <pre class="example"> ptrdiff_t fftw_mpi_local_size_2d(ptrdiff_t n0, ptrdiff_t n1, MPI_Comm comm,
|
|
d@0
|
72 ptrdiff_t *local_n0, ptrdiff_t *local_0_start);
|
|
d@0
|
73 </pre>
|
|
d@0
|
74 <p><a name="index-fftw_005fmpi_005flocal_005fsize_005f2d-352"></a>
|
|
d@0
|
75 Given the total size of the data to be transformed (here, <code>n0 =
|
|
d@0
|
76 100</code> and <code>n1 = 200</code>) and an MPI communicator (<code>comm</code>), this
|
|
d@0
|
77 function provides three numbers.
|
|
d@0
|
78
|
|
d@0
|
79 <p>First, it describes the shape of the local data: the current process
|
|
d@0
|
80 should store a <code>local_n0</code> by <code>n1</code> slice of the overall
|
|
d@0
|
81 dataset, in row-major order (<code>n1</code> dimension contiguous), starting
|
|
d@0
|
82 at index <code>local_0_start</code>. That is, if the total dataset is
|
|
d@0
|
83 viewed as a <code>n0</code> by <code>n1</code> matrix, the current process should
|
|
d@0
|
84 store the rows <code>local_0_start</code> to
|
|
d@0
|
85 <code>local_0_start+local_n0-1</code>. Obviously, if you are running with
|
|
d@0
|
86 only a single MPI process, that process will store the entire array:
|
|
d@0
|
87 <code>local_0_start</code> will be zero and <code>local_n0</code> will be
|
|
d@0
|
88 <code>n0</code>. See <a href="Row_002dmajor-Format.html#Row_002dmajor-Format">Row-major Format</a>.
|
|
d@0
|
89 <a name="index-row_002dmajor-353"></a>
|
|
d@0
|
90 Second, the return value is the total number of data elements (e.g.,
|
|
d@0
|
91 complex numbers for a complex DFT) that should be allocated for the
|
|
d@0
|
92 input and output arrays on the current process (ideally with
|
|
d@0
|
93 <code>fftw_malloc</code>, to ensure optimal alignment). It might seem that
|
|
d@0
|
94 this should always be equal to <code>local_n0 * n1</code>, but this is
|
|
d@0
|
95 <em>not</em> the case. FFTW's distributed FFT algorithms require data
|
|
d@0
|
96 redistributions at intermediate stages of the transform, and in some
|
|
d@0
|
97 circumstances this may require slightly larger local storage. This is
|
|
d@0
|
98 discussed in more detail below, under <a href="Load-balancing.html#Load-balancing">Load balancing</a>.
|
|
d@0
|
99 <a name="index-fftw_005fmalloc-354"></a>
|
|
d@0
|
100 The advanced-interface `<samp><span class="samp">local_size</span></samp>' function for multidimensional
|
|
d@0
|
101 transforms returns the same three things (<code>local_n0</code>,
|
|
d@0
|
102 <code>local_0_start</code>, and the total number of elements to allocate),
|
|
d@0
|
103 but takes more inputs:
|
|
d@0
|
104
|
|
d@0
|
105 <pre class="example"> ptrdiff_t fftw_mpi_local_size_many(int rnk, const ptrdiff_t *n,
|
|
d@0
|
106 ptrdiff_t howmany,
|
|
d@0
|
107 ptrdiff_t block0,
|
|
d@0
|
108 MPI_Comm comm,
|
|
d@0
|
109 ptrdiff_t *local_n0,
|
|
d@0
|
110 ptrdiff_t *local_0_start);
|
|
d@0
|
111 </pre>
|
|
d@0
|
112 <p><a name="index-fftw_005fmpi_005flocal_005fsize_005fmany-355"></a>
|
|
d@0
|
113 The two-dimensional case above corresponds to <code>rnk = 2</code> and an
|
|
d@0
|
114 array <code>n</code> of length 2 with <code>n[0] = n0</code> and <code>n[1] = n1</code>.
|
|
d@0
|
115 This routine is for any <code>rnk > 1</code>; one-dimensional transforms
|
|
d@0
|
116 have their own interface because they work slightly differently, as
|
|
d@0
|
117 discussed below.
|
|
d@0
|
118
|
|
d@0
|
119 <p>First, the advanced interface allows you to perform multiple
|
|
d@0
|
120 transforms at once, of interleaved data, as specified by the
|
|
d@0
|
121 <code>howmany</code> parameter. (<code>hoamany</code> is 1 for a single
|
|
d@0
|
122 transform.)
|
|
d@0
|
123
|
|
d@0
|
124 <p>Second, here you can specify your desired block size in the <code>n0</code>
|
|
d@0
|
125 dimension, <code>block0</code>. To use FFTW's default block size, pass
|
|
d@0
|
126 <code>FFTW_MPI_DEFAULT_BLOCK</code> (0) for <code>block0</code>. Otherwise, on
|
|
d@0
|
127 <code>P</code> processes, FFTW will return <code>local_n0</code> equal to
|
|
d@0
|
128 <code>block0</code> on the first <code>P / block0</code> processes (rounded down),
|
|
d@0
|
129 return <code>local_n0</code> equal to <code>n0 - block0 * (P / block0)</code> on
|
|
d@0
|
130 the next process, and <code>local_n0</code> equal to zero on any remaining
|
|
d@0
|
131 processes. In general, we recommend using the default block size
|
|
d@0
|
132 (which corresponds to <code>n0 / P</code>, rounded up).
|
|
d@0
|
133 <a name="index-FFTW_005fMPI_005fDEFAULT_005fBLOCK-356"></a><a name="index-block-distribution-357"></a>
|
|
d@0
|
134 For example, suppose you have <code>P = 4</code> processes and <code>n0 =
|
|
d@0
|
135 21</code>. The default will be a block size of <code>6</code>, which will give
|
|
d@0
|
136 <code>local_n0 = 6</code> on the first three processes and <code>local_n0 =
|
|
d@0
|
137 3</code> on the last process. Instead, however, you could specify
|
|
d@0
|
138 <code>block0 = 5</code> if you wanted, which would give <code>local_n0 = 5</code>
|
|
d@0
|
139 on processes 0 to 2, <code>local_n0 = 6</code> on process 3. (This choice,
|
|
d@0
|
140 while it may look superficially more “balanced,” has the same
|
|
d@0
|
141 critical path as FFTW's default but requires more communications.)
|
|
d@0
|
142
|
|
d@0
|
143 </body></html>
|
|
d@0
|
144
|
