Chris@10: @node Installation and Customization, Acknowledgments, Upgrading from FFTW version 2, Top Chris@10: @chapter Installation and Customization Chris@10: @cindex installation Chris@10: Chris@10: This chapter describes the installation and customization of FFTW, the Chris@10: latest version of which may be downloaded from Chris@10: @uref{http://www.fftw.org, the FFTW home page}. Chris@10: Chris@10: In principle, FFTW should work on any system with an ANSI C compiler Chris@10: (@code{gcc} is fine). However, planner time is drastically reduced if Chris@10: FFTW can exploit a hardware cycle counter; FFTW comes with cycle-counter Chris@10: support for all modern general-purpose CPUs, but you may need to add a Chris@10: couple of lines of code if your compiler is not yet supported Chris@10: (@pxref{Cycle Counters}). (On Unix, there will be a warning at the end Chris@10: of the @code{configure} output if no cycle counter is found.) Chris@10: @cindex cycle counter Chris@10: @cindex compiler Chris@10: @cindex portability Chris@10: Chris@10: Chris@10: Installation of FFTW is simplest if you have a Unix or a GNU system, Chris@10: such as GNU/Linux, and we describe this case in the first section below, Chris@10: including the use of special configuration options to e.g. install Chris@10: different precisions or exploit optimizations for particular Chris@10: architectures (e.g. SIMD). Compilation on non-Unix systems is a more Chris@10: manual process, but we outline the procedure in the second section. It Chris@10: is also likely that pre-compiled binaries will be available for popular Chris@10: systems. Chris@10: Chris@10: Finally, we describe how you can customize FFTW for particular needs by Chris@10: generating @emph{codelets} for fast transforms of sizes not supported Chris@10: efficiently by the standard FFTW distribution. Chris@10: @cindex codelet Chris@10: Chris@10: @menu Chris@10: * Installation on Unix:: Chris@10: * Installation on non-Unix systems:: Chris@10: * Cycle Counters:: Chris@10: * Generating your own code:: Chris@10: @end menu Chris@10: Chris@10: @c ------------------------------------------------------------ Chris@10: Chris@10: @node Installation on Unix, Installation on non-Unix systems, Installation and Customization, Installation and Customization Chris@10: @section Installation on Unix Chris@10: Chris@10: FFTW comes with a @code{configure} program in the GNU style. Chris@10: Installation can be as simple as: Chris@10: @fpindex configure Chris@10: Chris@10: @example Chris@10: ./configure Chris@10: make Chris@10: make install Chris@10: @end example Chris@10: Chris@10: This will build the uniprocessor complex and real transform libraries Chris@10: along with the test programs. (We recommend that you use GNU Chris@10: @code{make} if it is available; on some systems it is called Chris@10: @code{gmake}.) The ``@code{make install}'' command installs the fftw Chris@10: and rfftw libraries in standard places, and typically requires root Chris@10: privileges (unless you specify a different install directory with the Chris@10: @code{--prefix} flag to @code{configure}). You can also type Chris@10: ``@code{make check}'' to put the FFTW test programs through their paces. Chris@10: If you have problems during configuration or compilation, you may want Chris@10: to run ``@code{make distclean}'' before trying again; this ensures that Chris@10: you don't have any stale files left over from previous compilation Chris@10: attempts. Chris@10: Chris@10: The @code{configure} script chooses the @code{gcc} compiler by default, Chris@10: if it is available; you can select some other compiler with: Chris@10: @example Chris@10: ./configure CC="@r{@i{}}" Chris@10: @end example Chris@10: Chris@10: The @code{configure} script knows good @code{CFLAGS} (C compiler flags) Chris@10: @cindex compiler flags Chris@10: for a few systems. If your system is not known, the @code{configure} Chris@10: script will print out a warning. In this case, you should re-configure Chris@10: FFTW with the command Chris@10: @example Chris@10: ./configure CFLAGS="@r{@i{}}" Chris@10: @end example Chris@10: and then compile as usual. If you do find an optimal set of Chris@10: @code{CFLAGS} for your system, please let us know what they are (along Chris@10: with the output of @code{config.guess}) so that we can include them in Chris@10: future releases. Chris@10: Chris@10: @code{configure} supports all the standard flags defined by the GNU Chris@10: Coding Standards; see the @code{INSTALL} file in FFTW or Chris@10: @uref{http://www.gnu.org/prep/standards/html_node/index.html, the GNU web page}. Chris@10: Note especially @code{--help} to list all flags and Chris@10: @code{--enable-shared} to create shared, rather than static, libraries. Chris@10: @code{configure} also accepts a few FFTW-specific flags, particularly: Chris@10: Chris@10: @itemize @bullet Chris@10: Chris@10: @item Chris@10: @cindex precision Chris@10: @code{--enable-float}: Produces a single-precision version of FFTW Chris@10: (@code{float}) instead of the default double-precision (@code{double}). Chris@10: @xref{Precision}. Chris@10: Chris@10: @item Chris@10: @cindex precision Chris@10: @code{--enable-long-double}: Produces a long-double precision version of Chris@10: FFTW (@code{long double}) instead of the default double-precision Chris@10: (@code{double}). The @code{configure} script will halt with an error Chris@10: message if @code{long double} is the same size as @code{double} on your Chris@10: machine/compiler. @xref{Precision}. Chris@10: Chris@10: @item Chris@10: @cindex precision Chris@10: @code{--enable-quad-precision}: Produces a quadruple-precision version Chris@10: of FFTW using the nonstandard @code{__float128} type provided by Chris@10: @code{gcc} 4.6 or later on x86, x86-64, and Itanium architectures, Chris@10: instead of the default double-precision (@code{double}). The Chris@10: @code{configure} script will halt with an error message if the Chris@10: compiler is not @code{gcc} version 4.6 or later or if @code{gcc}'s Chris@10: @code{libquadmath} library is not installed. @xref{Precision}. Chris@10: Chris@10: @item Chris@10: @cindex threads Chris@10: @code{--enable-threads}: Enables compilation and installation of the Chris@10: FFTW threads library (@pxref{Multi-threaded FFTW}), which provides a Chris@10: simple interface to parallel transforms for SMP systems. By default, Chris@10: the threads routines are not compiled. Chris@10: Chris@10: @item Chris@10: @code{--enable-openmp}: Like @code{--enable-threads}, but using OpenMP Chris@10: compiler directives in order to induce parallelism rather than Chris@10: spawning its own threads directly, and installing an @samp{fftw3_omp} library Chris@10: rather than an @samp{fftw3_threads} library (@pxref{Multi-threaded Chris@10: FFTW}). You can use both @code{--enable-openmp} and @code{--enable-threads} Chris@10: since they compile/install libraries with different names. By default, Chris@10: the OpenMP routines are not compiled. Chris@10: Chris@10: @item Chris@10: @code{--with-combined-threads}: By default, if @code{--enable-threads} Chris@10: is used, the threads support is compiled into a separate library that Chris@10: must be linked in addition to the main FFTW library. This is so that Chris@10: users of the serial library do not need to link the system threads Chris@10: libraries. If @code{--with-combined-threads} is specified, however, Chris@10: then no separate threads library is created, and threads are included Chris@10: in the main FFTW library. This is mainly useful under Windows, where Chris@10: no system threads library is required and inter-library dependencies Chris@10: are problematic. Chris@10: Chris@10: @item Chris@10: @cindex MPI Chris@10: @code{--enable-mpi}: Enables compilation and installation of the FFTW Chris@10: MPI library (@pxref{Distributed-memory FFTW with MPI}), which provides Chris@10: parallel transforms for distributed-memory systems with MPI. (By Chris@10: default, the MPI routines are not compiled.) @xref{FFTW MPI Chris@10: Installation}. Chris@10: Chris@10: @item Chris@10: @cindex Fortran-callable wrappers Chris@10: @code{--disable-fortran}: Disables inclusion of legacy-Fortran Chris@10: wrapper routines (@pxref{Calling FFTW from Legacy Fortran}) in the standard Chris@10: FFTW libraries. These wrapper routines increase the library size by Chris@10: only a negligible amount, so they are included by default as long as Chris@10: the @code{configure} script finds a Fortran compiler on your system. Chris@10: (To specify a particular Fortran compiler @i{foo}, pass Chris@10: @code{F77=}@i{foo} to @code{configure}.) Chris@10: Chris@10: @item Chris@10: @code{--with-g77-wrappers}: By default, when Fortran wrappers are Chris@10: included, the wrappers employ the linking conventions of the Fortran Chris@10: compiler detected by the @code{configure} script. If this compiler is Chris@10: GNU @code{g77}, however, then @emph{two} versions of the wrappers are Chris@10: included: one with @code{g77}'s idiosyncratic convention of appending Chris@10: two underscores to identifiers, and one with the more common Chris@10: convention of appending only a single underscore. This way, the same Chris@10: FFTW library will work with both @code{g77} and other Fortran Chris@10: compilers, such as GNU @code{gfortran}. However, the converse is not Chris@10: true: if you configure with a different compiler, then the Chris@10: @code{g77}-compatible wrappers are not included. By specifying Chris@10: @code{--with-g77-wrappers}, the @code{g77}-compatible wrappers are Chris@10: included in addition to wrappers for whatever Fortran compiler Chris@10: @code{configure} finds. Chris@10: @fpindex g77 Chris@10: Chris@10: @item Chris@10: @code{--with-slow-timer}: Disables the use of hardware cycle counters, Chris@10: and falls back on @code{gettimeofday} or @code{clock}. This greatly Chris@10: worsens performance, and should generally not be used (unless you don't Chris@10: have a cycle counter but still really want an optimized plan regardless Chris@10: of the time). @xref{Cycle Counters}. Chris@10: Chris@10: @item Chris@10: @code{--enable-sse}, @code{--enable-sse2}, @code{--enable-avx}, Chris@10: @code{--enable-altivec}, @code{--enable-neon}: Enable the compilation of Chris@10: SIMD code for SSE (Pentium III+), SSE2 (Pentium IV+), AVX (Sandy Bridge, Chris@10: Interlagos), AltiVec (PowerPC G4+), NEON (some ARM processors). SSE, Chris@10: AltiVec, and NEON only work with @code{--enable-float} (above). SSE2 Chris@10: works in both single and double precision (and is simply SSE in single Chris@10: precision). The resulting code will @emph{still work} on earlier CPUs Chris@10: lacking the SIMD extensions (SIMD is automatically disabled, although Chris@10: the FFTW library is still larger). Chris@10: @itemize @minus Chris@10: @item Chris@10: These options require a compiler supporting SIMD extensions, and Chris@10: compiler support is always a bit flaky: see the FFTW FAQ for a list of Chris@10: compiler versions that have problems compiling FFTW. Chris@10: @item Chris@10: With AltiVec and @code{gcc}, you may have to use the Chris@10: @code{-mabi=altivec} option when compiling any code that links to FFTW, Chris@10: in order to properly align the stack; otherwise, FFTW could crash when Chris@10: it tries to use an AltiVec feature. (This is not necessary on MacOS X.) Chris@10: @item Chris@10: With SSE/SSE2 and @code{gcc}, you should use a version of gcc that Chris@10: properly aligns the stack when compiling any code that links to FFTW. Chris@10: By default, @code{gcc} 2.95 and later versions align the stack as Chris@10: needed, but you should not compile FFTW with the @code{-Os} option or the Chris@10: @code{-mpreferred-stack-boundary} option with an argument less than 4. Chris@10: @item Chris@10: Because of the large variety of ARM processors and ABIs, FFTW Chris@10: does not attempt to guess the correct @code{gcc} flags for generating Chris@10: NEON code. In general, you will have to provide them on the command line. Chris@10: This command line is known to have worked at least once: Chris@10: @example Chris@10: ./configure --with-slow-timer --host=arm-linux-gnueabi \ Chris@10: --enable-single --enable-neon \ Chris@10: "CC=arm-linux-gnueabi-gcc -march=armv7-a -mfloat-abi=softfp" Chris@10: @end example Chris@10: @end itemize Chris@10: Chris@10: @end itemize Chris@10: Chris@10: @cindex compiler Chris@10: To force @code{configure} to use a particular C compiler @i{foo} Chris@10: (instead of the default, usually @code{gcc}), pass @code{CC=}@i{foo} to the Chris@10: @code{configure} script; you may also need to set the flags via the variable Chris@10: @code{CFLAGS} as described above. Chris@10: @cindex compiler flags Chris@10: Chris@10: @c ------------------------------------------------------------ Chris@10: @node Installation on non-Unix systems, Cycle Counters, Installation on Unix, Installation and Customization Chris@10: @section Installation on non-Unix systems Chris@10: Chris@10: It should be relatively straightforward to compile FFTW even on non-Unix Chris@10: systems lacking the niceties of a @code{configure} script. Basically, Chris@10: you need to edit the @code{config.h} header (copy it from Chris@10: @code{config.h.in}) to @code{#define} the various options and compiler Chris@10: characteristics, and then compile all the @samp{.c} files in the Chris@10: relevant directories. Chris@10: Chris@10: The @code{config.h} header contains about 100 options to set, each one Chris@10: initially an @code{#undef}, each documented with a comment, and most of Chris@10: them fairly obvious. For most of the options, you should simply Chris@10: @code{#define} them to @code{1} if they are applicable, although a few Chris@10: options require a particular value (e.g. @code{SIZEOF_LONG_LONG} should Chris@10: be defined to the size of the @code{long long} type, in bytes, or zero Chris@10: if it is not supported). We will likely post some sample Chris@10: @code{config.h} files for various operating systems and compilers for Chris@10: you to use (at least as a starting point). Please let us know if you Chris@10: have to hand-create a configuration file (and/or a pre-compiled binary) Chris@10: that you want to share. Chris@10: Chris@10: To create the FFTW library, you will then need to compile all of the Chris@10: @samp{.c} files in the @code{kernel}, @code{dft}, @code{dft/scalar}, Chris@10: @code{dft/scalar/codelets}, @code{rdft}, @code{rdft/scalar}, Chris@10: @code{rdft/scalar/r2cf}, @code{rdft/scalar/r2cb}, Chris@10: @code{rdft/scalar/r2r}, @code{reodft}, and @code{api} directories. Chris@10: If you are compiling with SIMD support (e.g. you defined Chris@10: @code{HAVE_SSE2} in @code{config.h}), then you also need to compile Chris@10: the @code{.c} files in the @code{simd-support}, Chris@10: @code{@{dft,rdft@}/simd}, @code{@{dft,rdft@}/simd/*} directories. Chris@10: Chris@10: Once these files are all compiled, link them into a library, or a shared Chris@10: library, or directly into your program. Chris@10: Chris@10: To compile the FFTW test program, additionally compile the code in the Chris@10: @code{libbench2/} directory, and link it into a library. Then compile Chris@10: the code in the @code{tests/} directory and link it to the Chris@10: @code{libbench2} and FFTW libraries. To compile the @code{fftw-wisdom} Chris@10: (command-line) tool (@pxref{Wisdom Utilities}), compile Chris@10: @code{tools/fftw-wisdom.c} and link it to the @code{libbench2} and FFTW Chris@10: libraries Chris@10: Chris@10: @c ------------------------------------------------------------ Chris@10: @node Cycle Counters, Generating your own code, Installation on non-Unix systems, Installation and Customization Chris@10: @section Cycle Counters Chris@10: @cindex cycle counter Chris@10: Chris@10: FFTW's planner actually executes and times different possible FFT Chris@10: algorithms in order to pick the fastest plan for a given @math{n}. In Chris@10: order to do this in as short a time as possible, however, the timer must Chris@10: have a very high resolution, and to accomplish this we employ the Chris@10: hardware @dfn{cycle counters} that are available on most CPUs. Chris@10: Currently, FFTW supports the cycle counters on x86, PowerPC/POWER, Alpha, Chris@10: UltraSPARC (SPARC v9), IA64, PA-RISC, and MIPS processors. Chris@10: Chris@10: @cindex compiler Chris@10: Access to the cycle counters, unfortunately, is a compiler and/or Chris@10: operating-system dependent task, often requiring inline assembly Chris@10: language, and it may be that your compiler is not supported. If you are Chris@10: @emph{not} supported, FFTW will by default fall back on its estimator Chris@10: (effectively using @code{FFTW_ESTIMATE} for all plans). Chris@10: @ctindex FFTW_ESTIMATE Chris@10: Chris@10: You can add support by editing the file @code{kernel/cycle.h}; normally, Chris@10: this will involve adapting one of the examples already present in order Chris@10: to use the inline-assembler syntax for your C compiler, and will only Chris@10: require a couple of lines of code. Anyone adding support for a new Chris@10: system to @code{cycle.h} is encouraged to email us at @email{fftw@@fftw.org}. Chris@10: Chris@10: If a cycle counter is not available on your system (e.g. some embedded Chris@10: processor), and you don't want to use estimated plans, as a last resort Chris@10: you can use the @code{--with-slow-timer} option to @code{configure} (on Chris@10: Unix) or @code{#define WITH_SLOW_TIMER} in @code{config.h} (elsewhere). Chris@10: This will use the much lower-resolution @code{gettimeofday} function, or even Chris@10: @code{clock} if the former is unavailable, and planning will be Chris@10: extremely slow. Chris@10: Chris@10: @c ------------------------------------------------------------ Chris@10: @node Generating your own code, , Cycle Counters, Installation and Customization Chris@10: @section Generating your own code Chris@10: @cindex code generator Chris@10: Chris@10: The directory @code{genfft} contains the programs that were used to Chris@10: generate FFTW's ``codelets,'' which are hard-coded transforms of small Chris@10: sizes. Chris@10: @cindex codelet Chris@10: We do not expect casual users to employ the generator, which is a rather Chris@10: sophisticated program that generates directed acyclic graphs of FFT Chris@10: algorithms and performs algebraic simplifications on them. It was Chris@10: written in Objective Caml, a dialect of ML, which is available at Chris@10: @uref{http://caml.inria.fr/ocaml/index.en.html}. Chris@10: @cindex Caml Chris@10: Chris@10: Chris@10: If you have Objective Caml installed (along with recent versions of Chris@10: GNU @code{autoconf}, @code{automake}, and @code{libtool}), then you Chris@10: can change the set of codelets that are generated or play with the Chris@10: generation options. The set of generated codelets is specified by the Chris@10: @code{@{dft,rdft@}/@{codelets,simd@}/*/Makefile.am} files. For example, you can add Chris@10: efficient REDFT codelets of small sizes by modifying Chris@10: @code{rdft/codelets/r2r/Makefile.am}. Chris@10: @cindex REDFT Chris@10: After you modify any @code{Makefile.am} files, you can type @code{sh Chris@10: bootstrap.sh} in the top-level directory followed by @code{make} to Chris@10: re-generate the files. Chris@10: Chris@10: We do not provide more details about the code-generation process, since Chris@10: we do not expect that most users will need to generate their own code. Chris@10: However, feel free to contact us at @email{fftw@@fftw.org} if Chris@10: you are interested in the subject. Chris@10: Chris@10: @cindex monadic programming Chris@10: You might find it interesting to learn Caml and/or some modern Chris@10: programming techniques that we used in the generator (including monadic Chris@10: programming), especially if you heard the rumor that Java and Chris@10: object-oriented programming are the latest advancement in the field. Chris@10: The internal operation of the codelet generator is described in the Chris@10: paper, ``A Fast Fourier Transform Compiler,'' by M. Frigo, which is Chris@10: available from the @uref{http://www.fftw.org,FFTW home page} and also Chris@10: appeared in the @cite{Proceedings of the 1999 ACM SIGPLAN Conference on Chris@10: Programming Language Design and Implementation (PLDI)}. Chris@10: