|
Chris@42
|
1 @node Installation and Customization, Acknowledgments, Upgrading from FFTW version 2, Top
|
|
Chris@42
|
2 @chapter Installation and Customization
|
|
Chris@42
|
3 @cindex installation
|
|
Chris@42
|
4
|
|
Chris@42
|
5 This chapter describes the installation and customization of FFTW, the
|
|
Chris@42
|
6 latest version of which may be downloaded from
|
|
Chris@42
|
7 @uref{http://www.fftw.org, the FFTW home page}.
|
|
Chris@42
|
8
|
|
Chris@42
|
9 In principle, FFTW should work on any system with an ANSI C compiler
|
|
Chris@42
|
10 (@code{gcc} is fine). However, planner time is drastically reduced if
|
|
Chris@42
|
11 FFTW can exploit a hardware cycle counter; FFTW comes with cycle-counter
|
|
Chris@42
|
12 support for all modern general-purpose CPUs, but you may need to add a
|
|
Chris@42
|
13 couple of lines of code if your compiler is not yet supported
|
|
Chris@42
|
14 (@pxref{Cycle Counters}). (On Unix, there will be a warning at the end
|
|
Chris@42
|
15 of the @code{configure} output if no cycle counter is found.)
|
|
Chris@42
|
16 @cindex cycle counter
|
|
Chris@42
|
17 @cindex compiler
|
|
Chris@42
|
18 @cindex portability
|
|
Chris@42
|
19
|
|
Chris@42
|
20
|
|
Chris@42
|
21 Installation of FFTW is simplest if you have a Unix or a GNU system,
|
|
Chris@42
|
22 such as GNU/Linux, and we describe this case in the first section below,
|
|
Chris@42
|
23 including the use of special configuration options to e.g. install
|
|
Chris@42
|
24 different precisions or exploit optimizations for particular
|
|
Chris@42
|
25 architectures (e.g. SIMD). Compilation on non-Unix systems is a more
|
|
Chris@42
|
26 manual process, but we outline the procedure in the second section. It
|
|
Chris@42
|
27 is also likely that pre-compiled binaries will be available for popular
|
|
Chris@42
|
28 systems.
|
|
Chris@42
|
29
|
|
Chris@42
|
30 Finally, we describe how you can customize FFTW for particular needs by
|
|
Chris@42
|
31 generating @emph{codelets} for fast transforms of sizes not supported
|
|
Chris@42
|
32 efficiently by the standard FFTW distribution.
|
|
Chris@42
|
33 @cindex codelet
|
|
Chris@42
|
34
|
|
Chris@42
|
35 @menu
|
|
Chris@42
|
36 * Installation on Unix::
|
|
Chris@42
|
37 * Installation on non-Unix systems::
|
|
Chris@42
|
38 * Cycle Counters::
|
|
Chris@42
|
39 * Generating your own code::
|
|
Chris@42
|
40 @end menu
|
|
Chris@42
|
41
|
|
Chris@42
|
42 @c ------------------------------------------------------------
|
|
Chris@42
|
43
|
|
Chris@42
|
44 @node Installation on Unix, Installation on non-Unix systems, Installation and Customization, Installation and Customization
|
|
Chris@42
|
45 @section Installation on Unix
|
|
Chris@42
|
46
|
|
Chris@42
|
47 FFTW comes with a @code{configure} program in the GNU style.
|
|
Chris@42
|
48 Installation can be as simple as:
|
|
Chris@42
|
49 @fpindex configure
|
|
Chris@42
|
50
|
|
Chris@42
|
51 @example
|
|
Chris@42
|
52 ./configure
|
|
Chris@42
|
53 make
|
|
Chris@42
|
54 make install
|
|
Chris@42
|
55 @end example
|
|
Chris@42
|
56
|
|
Chris@42
|
57 This will build the uniprocessor complex and real transform libraries
|
|
Chris@42
|
58 along with the test programs. (We recommend that you use GNU
|
|
Chris@42
|
59 @code{make} if it is available; on some systems it is called
|
|
Chris@42
|
60 @code{gmake}.) The ``@code{make install}'' command installs the fftw
|
|
Chris@42
|
61 and rfftw libraries in standard places, and typically requires root
|
|
Chris@42
|
62 privileges (unless you specify a different install directory with the
|
|
Chris@42
|
63 @code{--prefix} flag to @code{configure}). You can also type
|
|
Chris@42
|
64 ``@code{make check}'' to put the FFTW test programs through their paces.
|
|
Chris@42
|
65 If you have problems during configuration or compilation, you may want
|
|
Chris@42
|
66 to run ``@code{make distclean}'' before trying again; this ensures that
|
|
Chris@42
|
67 you don't have any stale files left over from previous compilation
|
|
Chris@42
|
68 attempts.
|
|
Chris@42
|
69
|
|
Chris@42
|
70 The @code{configure} script chooses the @code{gcc} compiler by default,
|
|
Chris@42
|
71 if it is available; you can select some other compiler with:
|
|
Chris@42
|
72 @example
|
|
Chris@42
|
73 ./configure CC="@r{@i{<the name of your C compiler>}}"
|
|
Chris@42
|
74 @end example
|
|
Chris@42
|
75
|
|
Chris@42
|
76 The @code{configure} script knows good @code{CFLAGS} (C compiler flags)
|
|
Chris@42
|
77 @cindex compiler flags
|
|
Chris@42
|
78 for a few systems. If your system is not known, the @code{configure}
|
|
Chris@42
|
79 script will print out a warning. In this case, you should re-configure
|
|
Chris@42
|
80 FFTW with the command
|
|
Chris@42
|
81 @example
|
|
Chris@42
|
82 ./configure CFLAGS="@r{@i{<write your CFLAGS here>}}"
|
|
Chris@42
|
83 @end example
|
|
Chris@42
|
84 and then compile as usual. If you do find an optimal set of
|
|
Chris@42
|
85 @code{CFLAGS} for your system, please let us know what they are (along
|
|
Chris@42
|
86 with the output of @code{config.guess}) so that we can include them in
|
|
Chris@42
|
87 future releases.
|
|
Chris@42
|
88
|
|
Chris@42
|
89 @code{configure} supports all the standard flags defined by the GNU
|
|
Chris@42
|
90 Coding Standards; see the @code{INSTALL} file in FFTW or
|
|
Chris@42
|
91 @uref{http://www.gnu.org/prep/standards/html_node/index.html, the GNU web page}.
|
|
Chris@42
|
92 Note especially @code{--help} to list all flags and
|
|
Chris@42
|
93 @code{--enable-shared} to create shared, rather than static, libraries.
|
|
Chris@42
|
94 @code{configure} also accepts a few FFTW-specific flags, particularly:
|
|
Chris@42
|
95
|
|
Chris@42
|
96 @itemize @bullet
|
|
Chris@42
|
97
|
|
Chris@42
|
98 @item
|
|
Chris@42
|
99 @cindex precision
|
|
Chris@42
|
100 @code{--enable-float}: Produces a single-precision version of FFTW
|
|
Chris@42
|
101 (@code{float}) instead of the default double-precision (@code{double}).
|
|
Chris@42
|
102 @xref{Precision}.
|
|
Chris@42
|
103
|
|
Chris@42
|
104 @item
|
|
Chris@42
|
105 @cindex precision
|
|
Chris@42
|
106 @code{--enable-long-double}: Produces a long-double precision version of
|
|
Chris@42
|
107 FFTW (@code{long double}) instead of the default double-precision
|
|
Chris@42
|
108 (@code{double}). The @code{configure} script will halt with an error
|
|
Chris@42
|
109 message if @code{long double} is the same size as @code{double} on your
|
|
Chris@42
|
110 machine/compiler. @xref{Precision}.
|
|
Chris@42
|
111
|
|
Chris@42
|
112 @item
|
|
Chris@42
|
113 @cindex precision
|
|
Chris@42
|
114 @code{--enable-quad-precision}: Produces a quadruple-precision version
|
|
Chris@42
|
115 of FFTW using the nonstandard @code{__float128} type provided by
|
|
Chris@42
|
116 @code{gcc} 4.6 or later on x86, x86-64, and Itanium architectures,
|
|
Chris@42
|
117 instead of the default double-precision (@code{double}). The
|
|
Chris@42
|
118 @code{configure} script will halt with an error message if the
|
|
Chris@42
|
119 compiler is not @code{gcc} version 4.6 or later or if @code{gcc}'s
|
|
Chris@42
|
120 @code{libquadmath} library is not installed. @xref{Precision}.
|
|
Chris@42
|
121
|
|
Chris@42
|
122 @item
|
|
Chris@42
|
123 @cindex threads
|
|
Chris@42
|
124 @code{--enable-threads}: Enables compilation and installation of the
|
|
Chris@42
|
125 FFTW threads library (@pxref{Multi-threaded FFTW}), which provides a
|
|
Chris@42
|
126 simple interface to parallel transforms for SMP systems. By default,
|
|
Chris@42
|
127 the threads routines are not compiled.
|
|
Chris@42
|
128
|
|
Chris@42
|
129 @item
|
|
Chris@42
|
130 @code{--enable-openmp}: Like @code{--enable-threads}, but using OpenMP
|
|
Chris@42
|
131 compiler directives in order to induce parallelism rather than
|
|
Chris@42
|
132 spawning its own threads directly, and installing an @samp{fftw3_omp} library
|
|
Chris@42
|
133 rather than an @samp{fftw3_threads} library (@pxref{Multi-threaded
|
|
Chris@42
|
134 FFTW}). You can use both @code{--enable-openmp} and @code{--enable-threads}
|
|
Chris@42
|
135 since they compile/install libraries with different names. By default,
|
|
Chris@42
|
136 the OpenMP routines are not compiled.
|
|
Chris@42
|
137
|
|
Chris@42
|
138 @item
|
|
Chris@42
|
139 @code{--with-combined-threads}: By default, if @code{--enable-threads}
|
|
Chris@42
|
140 is used, the threads support is compiled into a separate library that
|
|
Chris@42
|
141 must be linked in addition to the main FFTW library. This is so that
|
|
Chris@42
|
142 users of the serial library do not need to link the system threads
|
|
Chris@42
|
143 libraries. If @code{--with-combined-threads} is specified, however,
|
|
Chris@42
|
144 then no separate threads library is created, and threads are included
|
|
Chris@42
|
145 in the main FFTW library. This is mainly useful under Windows, where
|
|
Chris@42
|
146 no system threads library is required and inter-library dependencies
|
|
Chris@42
|
147 are problematic.
|
|
Chris@42
|
148
|
|
Chris@42
|
149 @item
|
|
Chris@42
|
150 @cindex MPI
|
|
Chris@42
|
151 @code{--enable-mpi}: Enables compilation and installation of the FFTW
|
|
Chris@42
|
152 MPI library (@pxref{Distributed-memory FFTW with MPI}), which provides
|
|
Chris@42
|
153 parallel transforms for distributed-memory systems with MPI. (By
|
|
Chris@42
|
154 default, the MPI routines are not compiled.) @xref{FFTW MPI
|
|
Chris@42
|
155 Installation}.
|
|
Chris@42
|
156
|
|
Chris@42
|
157 @item
|
|
Chris@42
|
158 @cindex Fortran-callable wrappers
|
|
Chris@42
|
159 @code{--disable-fortran}: Disables inclusion of legacy-Fortran
|
|
Chris@42
|
160 wrapper routines (@pxref{Calling FFTW from Legacy Fortran}) in the standard
|
|
Chris@42
|
161 FFTW libraries. These wrapper routines increase the library size by
|
|
Chris@42
|
162 only a negligible amount, so they are included by default as long as
|
|
Chris@42
|
163 the @code{configure} script finds a Fortran compiler on your system.
|
|
Chris@42
|
164 (To specify a particular Fortran compiler @i{foo}, pass
|
|
Chris@42
|
165 @code{F77=}@i{foo} to @code{configure}.)
|
|
Chris@42
|
166
|
|
Chris@42
|
167 @item
|
|
Chris@42
|
168 @code{--with-g77-wrappers}: By default, when Fortran wrappers are
|
|
Chris@42
|
169 included, the wrappers employ the linking conventions of the Fortran
|
|
Chris@42
|
170 compiler detected by the @code{configure} script. If this compiler is
|
|
Chris@42
|
171 GNU @code{g77}, however, then @emph{two} versions of the wrappers are
|
|
Chris@42
|
172 included: one with @code{g77}'s idiosyncratic convention of appending
|
|
Chris@42
|
173 two underscores to identifiers, and one with the more common
|
|
Chris@42
|
174 convention of appending only a single underscore. This way, the same
|
|
Chris@42
|
175 FFTW library will work with both @code{g77} and other Fortran
|
|
Chris@42
|
176 compilers, such as GNU @code{gfortran}. However, the converse is not
|
|
Chris@42
|
177 true: if you configure with a different compiler, then the
|
|
Chris@42
|
178 @code{g77}-compatible wrappers are not included. By specifying
|
|
Chris@42
|
179 @code{--with-g77-wrappers}, the @code{g77}-compatible wrappers are
|
|
Chris@42
|
180 included in addition to wrappers for whatever Fortran compiler
|
|
Chris@42
|
181 @code{configure} finds.
|
|
Chris@42
|
182 @fpindex g77
|
|
Chris@42
|
183
|
|
Chris@42
|
184 @item
|
|
Chris@42
|
185 @code{--with-slow-timer}: Disables the use of hardware cycle counters,
|
|
Chris@42
|
186 and falls back on @code{gettimeofday} or @code{clock}. This greatly
|
|
Chris@42
|
187 worsens performance, and should generally not be used (unless you don't
|
|
Chris@42
|
188 have a cycle counter but still really want an optimized plan regardless
|
|
Chris@42
|
189 of the time). @xref{Cycle Counters}.
|
|
Chris@42
|
190
|
|
Chris@42
|
191 @item
|
|
Chris@42
|
192 @code{--enable-sse} (single precision),
|
|
Chris@42
|
193 @code{--enable-sse2} (single, double),
|
|
Chris@42
|
194 @code{--enable-avx} (single, double),
|
|
Chris@42
|
195 @code{--enable-avx2} (single, double),
|
|
Chris@42
|
196 @code{--enable-avx512} (single, double),
|
|
Chris@42
|
197 @code{--enable-avx-128-fma},
|
|
Chris@42
|
198 @code{--enable-kcvi} (single),
|
|
Chris@42
|
199 @code{--enable-altivec} (single),
|
|
Chris@42
|
200 @code{--enable-vsx} (single, double),
|
|
Chris@42
|
201 @code{--enable-neon} (single, double on aarch64),
|
|
Chris@42
|
202 @code{--enable-generic-simd128},
|
|
Chris@42
|
203 and
|
|
Chris@42
|
204 @code{--enable-generic-simd256}:
|
|
Chris@42
|
205
|
|
Chris@42
|
206 Enable various SIMD instruction sets. You need compiler that supports
|
|
Chris@42
|
207 the given SIMD extensions, but FFTW will try to detect at runtime
|
|
Chris@42
|
208 whether the CPU supports these extensions. That is, you can compile
|
|
Chris@42
|
209 with@code{--enable-avx} and the code will still run on a CPU without AVX
|
|
Chris@42
|
210 support.
|
|
Chris@42
|
211
|
|
Chris@42
|
212 @itemize @minus
|
|
Chris@42
|
213 @item
|
|
Chris@42
|
214 These options require a compiler supporting SIMD extensions, and
|
|
Chris@42
|
215 compiler support is always a bit flaky: see the FFTW FAQ for a list of
|
|
Chris@42
|
216 compiler versions that have problems compiling FFTW.
|
|
Chris@42
|
217 @item
|
|
Chris@42
|
218 Because of the large variety of ARM processors and ABIs, FFTW
|
|
Chris@42
|
219 does not attempt to guess the correct @code{gcc} flags for generating
|
|
Chris@42
|
220 NEON code. In general, you will have to provide them on the command line.
|
|
Chris@42
|
221 This command line is known to have worked at least once:
|
|
Chris@42
|
222 @example
|
|
Chris@42
|
223 ./configure --with-slow-timer --host=arm-linux-gnueabi \
|
|
Chris@42
|
224 --enable-single --enable-neon \
|
|
Chris@42
|
225 "CC=arm-linux-gnueabi-gcc -march=armv7-a -mfloat-abi=softfp"
|
|
Chris@42
|
226 @end example
|
|
Chris@42
|
227 @end itemize
|
|
Chris@42
|
228
|
|
Chris@42
|
229 @end itemize
|
|
Chris@42
|
230
|
|
Chris@42
|
231 @cindex compiler
|
|
Chris@42
|
232 To force @code{configure} to use a particular C compiler @i{foo}
|
|
Chris@42
|
233 (instead of the default, usually @code{gcc}), pass @code{CC=}@i{foo} to the
|
|
Chris@42
|
234 @code{configure} script; you may also need to set the flags via the variable
|
|
Chris@42
|
235 @code{CFLAGS} as described above.
|
|
Chris@42
|
236 @cindex compiler flags
|
|
Chris@42
|
237
|
|
Chris@42
|
238 @c ------------------------------------------------------------
|
|
Chris@42
|
239 @node Installation on non-Unix systems, Cycle Counters, Installation on Unix, Installation and Customization
|
|
Chris@42
|
240 @section Installation on non-Unix systems
|
|
Chris@42
|
241
|
|
Chris@42
|
242 It should be relatively straightforward to compile FFTW even on non-Unix
|
|
Chris@42
|
243 systems lacking the niceties of a @code{configure} script. Basically,
|
|
Chris@42
|
244 you need to edit the @code{config.h} header (copy it from
|
|
Chris@42
|
245 @code{config.h.in}) to @code{#define} the various options and compiler
|
|
Chris@42
|
246 characteristics, and then compile all the @samp{.c} files in the
|
|
Chris@42
|
247 relevant directories.
|
|
Chris@42
|
248
|
|
Chris@42
|
249 The @code{config.h} header contains about 100 options to set, each one
|
|
Chris@42
|
250 initially an @code{#undef}, each documented with a comment, and most of
|
|
Chris@42
|
251 them fairly obvious. For most of the options, you should simply
|
|
Chris@42
|
252 @code{#define} them to @code{1} if they are applicable, although a few
|
|
Chris@42
|
253 options require a particular value (e.g. @code{SIZEOF_LONG_LONG} should
|
|
Chris@42
|
254 be defined to the size of the @code{long long} type, in bytes, or zero
|
|
Chris@42
|
255 if it is not supported). We will likely post some sample
|
|
Chris@42
|
256 @code{config.h} files for various operating systems and compilers for
|
|
Chris@42
|
257 you to use (at least as a starting point). Please let us know if you
|
|
Chris@42
|
258 have to hand-create a configuration file (and/or a pre-compiled binary)
|
|
Chris@42
|
259 that you want to share.
|
|
Chris@42
|
260
|
|
Chris@42
|
261 To create the FFTW library, you will then need to compile all of the
|
|
Chris@42
|
262 @samp{.c} files in the @code{kernel}, @code{dft}, @code{dft/scalar},
|
|
Chris@42
|
263 @code{dft/scalar/codelets}, @code{rdft}, @code{rdft/scalar},
|
|
Chris@42
|
264 @code{rdft/scalar/r2cf}, @code{rdft/scalar/r2cb},
|
|
Chris@42
|
265 @code{rdft/scalar/r2r}, @code{reodft}, and @code{api} directories.
|
|
Chris@42
|
266 If you are compiling with SIMD support (e.g. you defined
|
|
Chris@42
|
267 @code{HAVE_SSE2} in @code{config.h}), then you also need to compile
|
|
Chris@42
|
268 the @code{.c} files in the @code{simd-support},
|
|
Chris@42
|
269 @code{@{dft,rdft@}/simd}, @code{@{dft,rdft@}/simd/*} directories.
|
|
Chris@42
|
270
|
|
Chris@42
|
271 Once these files are all compiled, link them into a library, or a shared
|
|
Chris@42
|
272 library, or directly into your program.
|
|
Chris@42
|
273
|
|
Chris@42
|
274 To compile the FFTW test program, additionally compile the code in the
|
|
Chris@42
|
275 @code{libbench2/} directory, and link it into a library. Then compile
|
|
Chris@42
|
276 the code in the @code{tests/} directory and link it to the
|
|
Chris@42
|
277 @code{libbench2} and FFTW libraries. To compile the @code{fftw-wisdom}
|
|
Chris@42
|
278 (command-line) tool (@pxref{Wisdom Utilities}), compile
|
|
Chris@42
|
279 @code{tools/fftw-wisdom.c} and link it to the @code{libbench2} and FFTW
|
|
Chris@42
|
280 libraries
|
|
Chris@42
|
281
|
|
Chris@42
|
282 @c ------------------------------------------------------------
|
|
Chris@42
|
283 @node Cycle Counters, Generating your own code, Installation on non-Unix systems, Installation and Customization
|
|
Chris@42
|
284 @section Cycle Counters
|
|
Chris@42
|
285 @cindex cycle counter
|
|
Chris@42
|
286
|
|
Chris@42
|
287 FFTW's planner actually executes and times different possible FFT
|
|
Chris@42
|
288 algorithms in order to pick the fastest plan for a given @math{n}. In
|
|
Chris@42
|
289 order to do this in as short a time as possible, however, the timer must
|
|
Chris@42
|
290 have a very high resolution, and to accomplish this we employ the
|
|
Chris@42
|
291 hardware @dfn{cycle counters} that are available on most CPUs.
|
|
Chris@42
|
292 Currently, FFTW supports the cycle counters on x86, PowerPC/POWER, Alpha,
|
|
Chris@42
|
293 UltraSPARC (SPARC v9), IA64, PA-RISC, and MIPS processors.
|
|
Chris@42
|
294
|
|
Chris@42
|
295 @cindex compiler
|
|
Chris@42
|
296 Access to the cycle counters, unfortunately, is a compiler and/or
|
|
Chris@42
|
297 operating-system dependent task, often requiring inline assembly
|
|
Chris@42
|
298 language, and it may be that your compiler is not supported. If you are
|
|
Chris@42
|
299 @emph{not} supported, FFTW will by default fall back on its estimator
|
|
Chris@42
|
300 (effectively using @code{FFTW_ESTIMATE} for all plans).
|
|
Chris@42
|
301 @ctindex FFTW_ESTIMATE
|
|
Chris@42
|
302
|
|
Chris@42
|
303 You can add support by editing the file @code{kernel/cycle.h}; normally,
|
|
Chris@42
|
304 this will involve adapting one of the examples already present in order
|
|
Chris@42
|
305 to use the inline-assembler syntax for your C compiler, and will only
|
|
Chris@42
|
306 require a couple of lines of code. Anyone adding support for a new
|
|
Chris@42
|
307 system to @code{cycle.h} is encouraged to email us at @email{fftw@@fftw.org}.
|
|
Chris@42
|
308
|
|
Chris@42
|
309 If a cycle counter is not available on your system (e.g. some embedded
|
|
Chris@42
|
310 processor), and you don't want to use estimated plans, as a last resort
|
|
Chris@42
|
311 you can use the @code{--with-slow-timer} option to @code{configure} (on
|
|
Chris@42
|
312 Unix) or @code{#define WITH_SLOW_TIMER} in @code{config.h} (elsewhere).
|
|
Chris@42
|
313 This will use the much lower-resolution @code{gettimeofday} function, or even
|
|
Chris@42
|
314 @code{clock} if the former is unavailable, and planning will be
|
|
Chris@42
|
315 extremely slow.
|
|
Chris@42
|
316
|
|
Chris@42
|
317 @c ------------------------------------------------------------
|
|
Chris@42
|
318 @node Generating your own code, , Cycle Counters, Installation and Customization
|
|
Chris@42
|
319 @section Generating your own code
|
|
Chris@42
|
320 @cindex code generator
|
|
Chris@42
|
321
|
|
Chris@42
|
322 The directory @code{genfft} contains the programs that were used to
|
|
Chris@42
|
323 generate FFTW's ``codelets,'' which are hard-coded transforms of small
|
|
Chris@42
|
324 sizes.
|
|
Chris@42
|
325 @cindex codelet
|
|
Chris@42
|
326 We do not expect casual users to employ the generator, which is a rather
|
|
Chris@42
|
327 sophisticated program that generates directed acyclic graphs of FFT
|
|
Chris@42
|
328 algorithms and performs algebraic simplifications on them. It was
|
|
Chris@42
|
329 written in Objective Caml, a dialect of ML, which is available at
|
|
Chris@42
|
330 @uref{http://caml.inria.fr/ocaml/index.en.html}.
|
|
Chris@42
|
331 @cindex Caml
|
|
Chris@42
|
332
|
|
Chris@42
|
333
|
|
Chris@42
|
334 If you have Objective Caml installed (along with recent versions of
|
|
Chris@42
|
335 GNU @code{autoconf}, @code{automake}, and @code{libtool}), then you
|
|
Chris@42
|
336 can change the set of codelets that are generated or play with the
|
|
Chris@42
|
337 generation options. The set of generated codelets is specified by the
|
|
Chris@42
|
338 @code{@{dft,rdft@}/@{codelets,simd@}/*/Makefile.am} files. For example, you can add
|
|
Chris@42
|
339 efficient REDFT codelets of small sizes by modifying
|
|
Chris@42
|
340 @code{rdft/codelets/r2r/Makefile.am}.
|
|
Chris@42
|
341 @cindex REDFT
|
|
Chris@42
|
342 After you modify any @code{Makefile.am} files, you can type @code{sh
|
|
Chris@42
|
343 bootstrap.sh} in the top-level directory followed by @code{make} to
|
|
Chris@42
|
344 re-generate the files.
|
|
Chris@42
|
345
|
|
Chris@42
|
346 We do not provide more details about the code-generation process, since
|
|
Chris@42
|
347 we do not expect that most users will need to generate their own code.
|
|
Chris@42
|
348 However, feel free to contact us at @email{fftw@@fftw.org} if
|
|
Chris@42
|
349 you are interested in the subject.
|
|
Chris@42
|
350
|
|
Chris@42
|
351 @cindex monadic programming
|
|
Chris@42
|
352 You might find it interesting to learn Caml and/or some modern
|
|
Chris@42
|
353 programming techniques that we used in the generator (including monadic
|
|
Chris@42
|
354 programming), especially if you heard the rumor that Java and
|
|
Chris@42
|
355 object-oriented programming are the latest advancement in the field.
|
|
Chris@42
|
356 The internal operation of the codelet generator is described in the
|
|
Chris@42
|
357 paper, ``A Fast Fourier Transform Compiler,'' by M. Frigo, which is
|
|
Chris@42
|
358 available from the @uref{http://www.fftw.org,FFTW home page} and also
|
|
Chris@42
|
359 appeared in the @cite{Proceedings of the 1999 ACM SIGPLAN Conference on
|
|
Chris@42
|
360 Programming Language Design and Implementation (PLDI)}.
|
|
Chris@42
|
361
|
