cannam@167: .\" cannam@167: .\" Copyright (c) 2003, 2007-14 Matteo Frigo cannam@167: .\" Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology cannam@167: .\" cannam@167: .\" This program is free software; you can redistribute it and/or modify cannam@167: .\" it under the terms of the GNU General Public License as published by cannam@167: .\" the Free Software Foundation; either version 2 of the License, or cannam@167: .\" (at your option) any later version. cannam@167: .\" cannam@167: .\" This program is distributed in the hope that it will be useful, cannam@167: .\" but WITHOUT ANY WARRANTY; without even the implied warranty of cannam@167: .\" MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the cannam@167: .\" GNU General Public License for more details. cannam@167: .\" cannam@167: .\" You should have received a copy of the GNU General Public License cannam@167: .\" along with this program; if not, write to the Free Software cannam@167: .\" Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA cannam@167: .\" cannam@167: .TH FFTW-WISDOM 1 "February, 2003" "fftw" "fftw" cannam@167: .SH NAME cannam@167: fftwf\-wisdom \- create wisdom (pre-optimized FFTs) cannam@167: .SH SYNOPSIS cannam@167: .B fftwf\-wisdom cannam@167: [\fIOPTION\fR]... [\fISIZE\fR]... cannam@167: .SH DESCRIPTION cannam@167: .PP cannam@167: .\" Add any additional description here cannam@167: .I fftwf\-wisdom cannam@167: is a utility to generate FFTW cannam@167: .B wisdom cannam@167: files, which contain saved information about how to optimally compute cannam@167: (Fourier) transforms of various sizes. FFTW is a free library to cannam@167: compute discrete Fourier transforms in one or more dimensions, for cannam@167: arbitrary sizes, and of both real and complex data, among other cannam@167: related operations. More information on FFTW can be found at the FFTW cannam@167: home page: cannam@167: .I http://www.fftw.org cannam@167: cannam@167: Programs using FFTW can be written to load wisdom from an arbitrary file, cannam@167: string, or other source. Moreover, it is likely that many FFTW-using cannam@167: programs will load the \fBsystem wisdom\fR file, which is stored in cannam@167: .I /etc/fftw/wisdomf cannam@167: by default. cannam@167: .I fftwf\-wisdom cannam@167: can be used to create or add to such wisdom files. In its most cannam@167: typical usage, the wisdom file can be created to pre-plan a canonical cannam@167: set of sizes (see below) via: cannam@167: cannam@167: .ce cannam@167: fftwf\-wisdom \-v \-c \-o wisdomf cannam@167: cannam@167: (this will take many hours, which can be limited by the cannam@167: .B \-t cannam@167: option) and the output cannam@167: .I wisdomf cannam@167: file can then be copied (as root) to cannam@167: .I /etc/fftw/ cannam@167: or whatever. cannam@167: cannam@167: The cannam@167: .I fftwf\-wisdom cannam@167: program normally writes the wisdom directly to standard output, but this cannam@167: can be changed via the cannam@167: .B \-o cannam@167: option, as in the example above. cannam@167: cannam@167: If the system wisdom file cannam@167: .I /etc/fftw/wisdomf cannam@167: already exists, then cannam@167: .I fftwf\-wisdom cannam@167: reads this existing wisdom (unless the cannam@167: .B \-n cannam@167: option is specified) and outputs both the old wisdom and any cannam@167: newly created wisdom. In this way, it can be used to add new transform cannam@167: sizes to the existing system wisdom (or other wisdom file, with the cannam@167: .B \-w cannam@167: option). cannam@167: .SH SPECIFYING SIZES cannam@167: Although a canonical set of sizes to optimize is specified by the cannam@167: .B \-c cannam@167: option, the user can also specify zero or more non-canonical transform cannam@167: sizes and types to optimize, via the cannam@167: .I SIZE cannam@167: arguments following the option flags. Alternatively, the sizes to cannam@167: optimize can be read from standard input (whitespace-separated), if a cannam@167: .I SIZE cannam@167: argument of "\-" is supplied. cannam@167: cannam@167: Sizes are specified by the syntax: cannam@167: cannam@167: .ce cannam@167: <\fItype\fR><\fIinplace\fR><\fIdirection\fR><\fIgeometry\fR> cannam@167: cannam@167: <\fItype\fR> is either \'c\' (complex), \'r\' (real, r2c/c2r), or cannam@167: \'k\' (r2r, per-dimension kinds, specified in the geometry, below). cannam@167: cannam@167: <\fIinplace\fR> is either \'i\' (in place) or \'o\' (out of place). cannam@167: cannam@167: <\fIdirection\fR> is either \'f\' (forward) or \'b\' (backward). The cannam@167: <\fIdirection\fR> should be omitted for \'k\' transforms, where it is cannam@167: specified via the geometry instead. cannam@167: cannam@167: <\fIgeometry\fR> is the size and dimensionality of the transform, cannam@167: where different dimensions are separated by \'x\' (e.g. \'16x32\' for cannam@167: a two-dimensional 16 by 32 transform). In the case of \'k\' cannam@167: transforms, the size of each dimension is followed by a "type" string, cannam@167: which can be one of f/b/h/e00/e01/e10/e11/o00/o01/o10/o11 for cannam@167: R2HC/HC2R/DHT/REDFT00/.../RODFT11, respectively, as defined in the cannam@167: FFTW manual. cannam@167: cannam@167: For example, \'cif12x13x14\' is a three-dimensional 12 by 13 x 14 cannam@167: complex DFT operating in-place. \'rob65536\' is a one-dimensional cannam@167: size-65536 out-of-place complex-to-real (backwards) transform cannam@167: operating on Hermitian-symmetry input. \'ki10hx20e01\' is a cannam@167: two-dimensional 10 by 20 r2r transform where the first dimension is a cannam@167: DHT and the second dimension is an REDFT01 (DCT-III). cannam@167: cannam@167: .SH OPTIONS cannam@167: .TP cannam@167: \fB\-h\fR, \fB\-\-help\fR cannam@167: Display help on the command-line options and usage. cannam@167: .TP cannam@167: \fB\-V\fR, \fB\-\-version\fR cannam@167: Print the version number and copyright information. cannam@167: .TP cannam@167: \fB\-v\fR, \fB\-\-verbose\fR cannam@167: Verbose output. (You can specify this multiple times, or supply a numeric cannam@167: argument greater than 1, to increase the verbosity level.) Note that the cannam@167: verbose output will be mixed with the wisdom output (making it impossible cannam@167: to import), unless you write the wisdom to a file via the cannam@167: .B \-o cannam@167: option. cannam@167: .TP cannam@167: \fB\-c\fR, \fB\-\-canonical\fR cannam@167: Optimize/pre-plan a canonical set of sizes: all powers of two and ten cannam@167: up to 2^20 (1048576), including both real and complex, forward and cannam@167: backwards, in-place and out-of-place transforms. Also includes two- cannam@167: and three-dimensional transforms of equal-size dimensions cannam@167: (e.g. 16x16x16). cannam@167: .TP cannam@167: \fB\-t\fR \fIhours\fR, \fB\-\-time\-limit\fR=\fIhours\fR cannam@167: Stop after a time of cannam@167: .I hours cannam@167: (hours) has elapsed, outputting accumulated wisdom. (The problems are planned cannam@167: in increasing order of size.) Defaults to 0, indicating no time limit. cannam@167: .TP cannam@167: \fB\-o\fR \fIfile\fR, \fB\-\-output-file\fR=\fIfile\fR cannam@167: Send wisdom output to cannam@167: .I file cannam@167: rather than to standard output (the default). cannam@167: .TP cannam@167: \fB\-m\fR, \fB\-\-measure\fR; \fB\-e\fR, \fB\-\-estimate\fR; \fB\-x\fR, \fB\-\-exhaustive\fR cannam@167: Normally, cannam@167: .I fftwf\-wisdom cannam@167: creates plans in FFTW_PATIENT mode, but with these options you can instead cannam@167: use FFTW_MEASURE, FFTW_ESTIMATE, or FFTW_EXHAUSTIVE modes, respectively, cannam@167: as described in more detail by the FFTW manual. cannam@167: cannam@167: Note that wisdom is tagged with the planning patience level, and a cannam@167: single file can mix different levels of wisdom (e.g. you can mostly cannam@167: use the patient default, but plan a few sizes that you especially care cannam@167: about in cannam@167: .B \-\-exhaustive cannam@167: mode). cannam@167: .TP cannam@167: \fB\-n\fR, \fB\-\-no\-system\-wisdom\fR cannam@167: Do not import the system wisdom from cannam@167: .I /etc/fftw/wisdomf cannam@167: (which is normally read by default). cannam@167: .TP cannam@167: \fB\-w\fR \fIfile\fR, \fB\-\-wisdom\-file\fR=\fIfile\fR cannam@167: Import wisdom from cannam@167: .I file cannam@167: (in addition to the system wisdom, unless cannam@167: .B \-n cannam@167: is specified). Multiple wisdom files can be read via multiple cannam@167: .B \-w cannam@167: options. If cannam@167: .I file cannam@167: is "\-", then read wisdom from standard input. cannam@167: .TP cannam@167: \fB\-T\fR \fIN\fR, \fB\--threads\fR=\fIN\fR cannam@167: Plan with cannam@167: .I N cannam@167: threads. This option is only present if FFTW was configured with cannam@167: thread support. cannam@167: .SH BUGS cannam@167: Send bug reports to fftw@fftw.org. cannam@167: .SH AUTHORS cannam@167: Written by Steven G. Johnson and Matteo Frigo. cannam@167: cannam@167: Copyright (c) 2003, 2007-14 Matteo Frigo cannam@167: .br cannam@167: Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology cannam@167: .SH "SEE ALSO" cannam@167: fftw-wisdom-to-conf(1)