Chris@42: @node Calling FFTW from Legacy Fortran, Upgrading from FFTW version 2, Calling FFTW from Modern Fortran, Top Chris@42: @chapter Calling FFTW from Legacy Fortran Chris@42: @cindex Fortran interface Chris@42: Chris@42: This chapter describes the interface to FFTW callable by Fortran code Chris@42: in older compilers not supporting the Fortran 2003 C interoperability Chris@42: features (@pxref{Calling FFTW from Modern Fortran}). This interface Chris@42: has the major disadvantage that it is not type-checked, so if you Chris@42: mistake the argument types or ordering then your program will not have Chris@42: any compiler errors, and will likely crash at runtime. So, greater Chris@42: care is needed. Also, technically interfacing older Fortran versions Chris@42: to C is nonstandard, but in practice we have found that the techniques Chris@42: used in this chapter have worked with all known Fortran compilers for Chris@42: many years. Chris@42: Chris@42: The legacy Fortran interface differs from the C interface only in the Chris@42: prefix (@samp{dfftw_} instead of @samp{fftw_} in double precision) and Chris@42: a few other minor details. This Fortran interface is included in the Chris@42: FFTW libraries by default, unless a Fortran compiler isn't found on Chris@42: your system or @code{--disable-fortran} is included in the Chris@42: @code{configure} flags. We assume here that the reader is already Chris@42: familiar with the usage of FFTW in C, as described elsewhere in this Chris@42: manual. Chris@42: Chris@42: The MPI parallel interface to FFTW is @emph{not} currently available Chris@42: to legacy Fortran. Chris@42: Chris@42: @menu Chris@42: * Fortran-interface routines:: Chris@42: * FFTW Constants in Fortran:: Chris@42: * FFTW Execution in Fortran:: Chris@42: * Fortran Examples:: Chris@42: * Wisdom of Fortran?:: Chris@42: @end menu Chris@42: Chris@42: @c ------------------------------------------------------- Chris@42: @node Fortran-interface routines, FFTW Constants in Fortran, Calling FFTW from Legacy Fortran, Calling FFTW from Legacy Fortran Chris@42: @section Fortran-interface routines Chris@42: Chris@42: Nearly all of the FFTW functions have Fortran-callable equivalents. Chris@42: The name of the legacy Fortran routine is the same as that of the Chris@42: corresponding C routine, but with the @samp{fftw_} prefix replaced by Chris@42: @samp{dfftw_}.@footnote{Technically, Fortran 77 identifiers are not Chris@42: allowed to have more than 6 characters, nor may they contain Chris@42: underscores. Any compiler that enforces this limitation doesn't Chris@42: deserve to link to FFTW.} The single and long-double precision Chris@42: versions use @samp{sfftw_} and @samp{lfftw_}, respectively, instead of Chris@42: @samp{fftwf_} and @samp{fftwl_}; quadruple precision (@code{real*16}) Chris@42: is available on some systems as @samp{fftwq_} (@pxref{Precision}). Chris@42: (Note that @code{long double} on x86 hardware is usually at most Chris@42: 80-bit extended precision, @emph{not} quadruple precision.) Chris@42: Chris@42: For the most part, all of the arguments to the functions are the same, Chris@42: with the following exceptions: Chris@42: Chris@42: @itemize @bullet Chris@42: Chris@42: @item Chris@42: @code{plan} variables (what would be of type @code{fftw_plan} in C), Chris@42: must be declared as a type that is at least as big as a pointer Chris@42: (address) on your machine. We recommend using @code{integer*8} everywhere, Chris@42: since this should always be big enough. Chris@42: @cindex portability Chris@42: Chris@42: @item Chris@42: Any function that returns a value (e.g. @code{fftw_plan_dft}) is Chris@42: converted into a @emph{subroutine}. The return value is converted into Chris@42: an additional @emph{first} parameter of this subroutine.@footnote{The Chris@42: reason for this is that some Fortran implementations seem to have Chris@42: trouble with C function return values, and vice versa.} Chris@42: Chris@42: @item Chris@42: @cindex column-major Chris@42: The Fortran routines expect multi-dimensional arrays to be in Chris@42: @emph{column-major} order, which is the ordinary format of Fortran Chris@42: arrays (@pxref{Multi-dimensional Array Format}). They do this Chris@42: transparently and costlessly simply by reversing the order of the Chris@42: dimensions passed to FFTW, but this has one important consequence for Chris@42: multi-dimensional real-complex transforms, discussed below. Chris@42: Chris@42: @item Chris@42: Wisdom import and export is somewhat more tricky because one cannot Chris@42: easily pass files or strings between C and Fortran; see @ref{Wisdom of Chris@42: Fortran?}. Chris@42: Chris@42: @item Chris@42: Legacy Fortran cannot use the @code{fftw_malloc} dynamic-allocation routine. Chris@42: If you want to exploit the SIMD FFTW (@pxref{SIMD alignment and fftw_malloc}), you'll Chris@42: need to figure out some other way to ensure that your arrays are at Chris@42: least 16-byte aligned. Chris@42: Chris@42: @item Chris@42: @tindex fftw_iodim Chris@42: @cindex guru interface Chris@42: Since Fortran 77 does not have data structures, the @code{fftw_iodim} Chris@42: structure from the guru interface (@pxref{Guru vector and transform Chris@42: sizes}) must be split into separate arguments. In particular, any Chris@42: @code{fftw_iodim} array arguments in the C guru interface become three Chris@42: integer array arguments (@code{n}, @code{is}, and @code{os}) in the Chris@42: Fortran guru interface, all of whose lengths should be equal to the Chris@42: corresponding @code{rank} argument. Chris@42: Chris@42: @item Chris@42: The guru planner interface in Fortran does @emph{not} do any automatic Chris@42: translation between column-major and row-major; you are responsible Chris@42: for setting the strides etcetera to correspond to your Fortran arrays. Chris@42: However, as a slight bug that we are preserving for backwards Chris@42: compatibility, the @samp{plan_guru_r2r} in Fortran @emph{does} reverse the Chris@42: order of its @code{kind} array parameter, so the @code{kind} array Chris@42: of that routine should be in the reverse of the order of the iodim Chris@42: arrays (see above). Chris@42: Chris@42: @end itemize Chris@42: Chris@42: In general, you should take care to use Fortran data types that Chris@42: correspond to (i.e. are the same size as) the C types used by FFTW. Chris@42: In practice, this correspondence is usually straightforward Chris@42: (i.e. @code{integer} corresponds to @code{int}, @code{real} Chris@42: corresponds to @code{float}, etcetera). The native Fortran Chris@42: double/single-precision complex type should be compatible with Chris@42: @code{fftw_complex}/@code{fftwf_complex}. Such simple correspondences Chris@42: are assumed in the examples below. Chris@42: @cindex portability Chris@42: Chris@42: @c ------------------------------------------------------- Chris@42: @node FFTW Constants in Fortran, FFTW Execution in Fortran, Fortran-interface routines, Calling FFTW from Legacy Fortran Chris@42: @section FFTW Constants in Fortran Chris@42: Chris@42: When creating plans in FFTW, a number of constants are used to specify Chris@42: options, such as @code{FFTW_MEASURE} or @code{FFTW_ESTIMATE}. The Chris@42: same constants must be used with the wrapper routines, but of course the Chris@42: C header files where the constants are defined can't be incorporated Chris@42: directly into Fortran code. Chris@42: Chris@42: Instead, we have placed Fortran equivalents of the FFTW constant Chris@42: definitions in the file @code{fftw3.f}, which can be found in the same Chris@42: directory as @code{fftw3.h}. If your Fortran compiler supports a Chris@42: preprocessor of some sort, you should be able to @code{include} or Chris@42: @code{#include} this file; otherwise, you can paste it directly into Chris@42: your code. Chris@42: Chris@42: @cindex flags Chris@42: In C, you combine different flags (like @code{FFTW_PRESERVE_INPUT} and Chris@42: @code{FFTW_MEASURE}) using the @samp{@code{|}} operator; in Fortran Chris@42: you should just use @samp{@code{+}}. (Take care not to add in the Chris@42: same flag more than once, though. Alternatively, you can use the Chris@42: @code{ior} intrinsic function standardized in Fortran 95.) Chris@42: Chris@42: @c ------------------------------------------------------- Chris@42: @node FFTW Execution in Fortran, Fortran Examples, FFTW Constants in Fortran, Calling FFTW from Legacy Fortran Chris@42: @section FFTW Execution in Fortran Chris@42: Chris@42: In C, in order to use a plan, one normally calls @code{fftw_execute}, Chris@42: which executes the plan to perform the transform on the input/output Chris@42: arrays passed when the plan was created (@pxref{Using Plans}). The Chris@42: corresponding subroutine call in legacy Fortran is: Chris@42: @example Chris@42: call dfftw_execute(plan) Chris@42: @end example Chris@42: @findex dfftw_execute Chris@42: Chris@42: However, we have had reports that this causes problems with some Chris@42: recent optimizing Fortran compilers. The problem is, because the Chris@42: input/output arrays are not passed as explicit arguments to Chris@42: @code{dfftw_execute}, the semantics of Fortran (unlike C) allow the Chris@42: compiler to assume that the input/output arrays are not changed by Chris@42: @code{dfftw_execute}. As a consequence, certain compilers end up Chris@42: optimizing out or repositioning the call to @code{dfftw_execute}, Chris@42: assuming incorrectly that it does nothing. Chris@42: Chris@42: There are various workarounds to this, but the safest and simplest Chris@42: thing is to not use @code{dfftw_execute} in Fortran. Instead, use the Chris@42: functions described in @ref{New-array Execute Functions}, which take Chris@42: the input/output arrays as explicit arguments. For example, if the Chris@42: plan is for a complex-data DFT and was created for the arrays Chris@42: @code{in} and @code{out}, you would do: Chris@42: @example Chris@42: call dfftw_execute_dft(plan, in, out) Chris@42: @end example Chris@42: @findex dfftw_execute_dft Chris@42: Chris@42: There are a few things to be careful of, however: Chris@42: Chris@42: @itemize @bullet Chris@42: Chris@42: @item Chris@42: You must use the correct type of execute function, matching the way Chris@42: the plan was created. Complex DFT plans should use Chris@42: @code{dfftw_execute_dft}, Real-input (r2c) DFT plans should use use Chris@42: @code{dfftw_execute_dft_r2c}, and real-output (c2r) DFT plans should Chris@42: use @code{dfftw_execute_dft_c2r}. The various r2r plans should use Chris@42: @code{dfftw_execute_r2r}. Chris@42: Chris@42: @item Chris@42: You should normally pass the same input/output arrays that were used when Chris@42: creating the plan. This is always safe. Chris@42: Chris@42: @item Chris@42: @emph{If} you pass @emph{different} input/output arrays compared to Chris@42: those used when creating the plan, you must abide by all the Chris@42: restrictions of the new-array execute functions (@pxref{New-array Chris@42: Execute Functions}). The most difficult of these, in Fortran, is the Chris@42: requirement that the new arrays have the same alignment as the Chris@42: original arrays, because there seems to be no way in legacy Fortran to obtain Chris@42: guaranteed-aligned arrays (analogous to @code{fftw_malloc} in C). You Chris@42: can, of course, use the @code{FFTW_UNALIGNED} flag when creating the Chris@42: plan, in which case the plan does not depend on the alignment, but Chris@42: this may sacrifice substantial performance on architectures (like x86) Chris@42: with SIMD instructions (@pxref{SIMD alignment and fftw_malloc}). Chris@42: @ctindex FFTW_UNALIGNED Chris@42: Chris@42: @end itemize Chris@42: Chris@42: @c ------------------------------------------------------- Chris@42: @node Fortran Examples, Wisdom of Fortran?, FFTW Execution in Fortran, Calling FFTW from Legacy Fortran Chris@42: @section Fortran Examples Chris@42: Chris@42: In C, you might have something like the following to transform a Chris@42: one-dimensional complex array: Chris@42: Chris@42: @example Chris@42: fftw_complex in[N], out[N]; Chris@42: fftw_plan plan; Chris@42: Chris@42: plan = fftw_plan_dft_1d(N,in,out,FFTW_FORWARD,FFTW_ESTIMATE); Chris@42: fftw_execute(plan); Chris@42: fftw_destroy_plan(plan); Chris@42: @end example Chris@42: Chris@42: In Fortran, you would use the following to accomplish the same thing: Chris@42: Chris@42: @example Chris@42: double complex in, out Chris@42: dimension in(N), out(N) Chris@42: integer*8 plan Chris@42: Chris@42: call dfftw_plan_dft_1d(plan,N,in,out,FFTW_FORWARD,FFTW_ESTIMATE) Chris@42: call dfftw_execute_dft(plan, in, out) Chris@42: call dfftw_destroy_plan(plan) Chris@42: @end example Chris@42: @findex dfftw_plan_dft_1d Chris@42: @findex dfftw_execute_dft Chris@42: @findex dfftw_destroy_plan Chris@42: Chris@42: Notice how all routines are called as Fortran subroutines, and the Chris@42: plan is returned via the first argument to @code{dfftw_plan_dft_1d}. Chris@42: Notice also that we changed @code{fftw_execute} to Chris@42: @code{dfftw_execute_dft} (@pxref{FFTW Execution in Fortran}). To do Chris@42: the same thing, but using 8 threads in parallel (@pxref{Multi-threaded Chris@42: FFTW}), you would simply prefix these calls with: Chris@42: Chris@42: @example Chris@42: integer iret Chris@42: call dfftw_init_threads(iret) Chris@42: call dfftw_plan_with_nthreads(8) Chris@42: @end example Chris@42: @findex dfftw_init_threads Chris@42: @findex dfftw_plan_with_nthreads Chris@42: Chris@42: (You might want to check the value of @code{iret}: if it is zero, it Chris@42: indicates an unlikely error during thread initialization.) Chris@42: Chris@42: To transform a three-dimensional array in-place with C, you might do: Chris@42: Chris@42: @example Chris@42: fftw_complex arr[L][M][N]; Chris@42: fftw_plan plan; Chris@42: Chris@42: plan = fftw_plan_dft_3d(L,M,N, arr,arr, Chris@42: FFTW_FORWARD, FFTW_ESTIMATE); Chris@42: fftw_execute(plan); Chris@42: fftw_destroy_plan(plan); Chris@42: @end example Chris@42: Chris@42: In Fortran, you would use this instead: Chris@42: Chris@42: @example Chris@42: double complex arr Chris@42: dimension arr(L,M,N) Chris@42: integer*8 plan Chris@42: Chris@42: call dfftw_plan_dft_3d(plan, L,M,N, arr,arr, Chris@42: & FFTW_FORWARD, FFTW_ESTIMATE) Chris@42: call dfftw_execute_dft(plan, arr, arr) Chris@42: call dfftw_destroy_plan(plan) Chris@42: @end example Chris@42: @findex dfftw_plan_dft_3d Chris@42: Chris@42: Note that we pass the array dimensions in the ``natural'' order in both C Chris@42: and Fortran. Chris@42: Chris@42: To transform a one-dimensional real array in Fortran, you might do: Chris@42: Chris@42: @example Chris@42: double precision in Chris@42: dimension in(N) Chris@42: double complex out Chris@42: dimension out(N/2 + 1) Chris@42: integer*8 plan Chris@42: Chris@42: call dfftw_plan_dft_r2c_1d(plan,N,in,out,FFTW_ESTIMATE) Chris@42: call dfftw_execute_dft_r2c(plan, in, out) Chris@42: call dfftw_destroy_plan(plan) Chris@42: @end example Chris@42: @findex dfftw_plan_dft_r2c_1d Chris@42: @findex dfftw_execute_dft_r2c Chris@42: Chris@42: To transform a two-dimensional real array, out of place, you might use Chris@42: the following: Chris@42: Chris@42: @example Chris@42: double precision in Chris@42: dimension in(M,N) Chris@42: double complex out Chris@42: dimension out(M/2 + 1, N) Chris@42: integer*8 plan Chris@42: Chris@42: call dfftw_plan_dft_r2c_2d(plan,M,N,in,out,FFTW_ESTIMATE) Chris@42: call dfftw_execute_dft_r2c(plan, in, out) Chris@42: call dfftw_destroy_plan(plan) Chris@42: @end example Chris@42: @findex dfftw_plan_dft_r2c_2d Chris@42: Chris@42: @strong{Important:} Notice that it is the @emph{first} dimension of the Chris@42: complex output array that is cut in half in Fortran, rather than the Chris@42: last dimension as in C. This is a consequence of the interface routines Chris@42: reversing the order of the array dimensions passed to FFTW so that the Chris@42: Fortran program can use its ordinary column-major order. Chris@42: @cindex column-major Chris@42: @cindex r2c/c2r multi-dimensional array format Chris@42: Chris@42: @c ------------------------------------------------------- Chris@42: @node Wisdom of Fortran?, , Fortran Examples, Calling FFTW from Legacy Fortran Chris@42: @section Wisdom of Fortran? Chris@42: Chris@42: In this section, we discuss how one can import/export FFTW wisdom Chris@42: (saved plans) to/from a Fortran program; we assume that the reader is Chris@42: already familiar with wisdom, as described in @ref{Words of Chris@42: Wisdom-Saving Plans}. Chris@42: Chris@42: @cindex portability Chris@42: The basic problem is that is difficult to (portably) pass files and Chris@42: strings between Fortran and C, so we cannot provide a direct Fortran Chris@42: equivalent to the @code{fftw_export_wisdom_to_file}, etcetera, Chris@42: functions. Fortran interfaces @emph{are} provided for the functions Chris@42: that do not take file/string arguments, however: Chris@42: @code{dfftw_import_system_wisdom}, @code{dfftw_import_wisdom}, Chris@42: @code{dfftw_export_wisdom}, and @code{dfftw_forget_wisdom}. Chris@42: @findex dfftw_import_system_wisdom Chris@42: @findex dfftw_import_wisdom Chris@42: @findex dfftw_export_wisdom Chris@42: @findex dfftw_forget_wisdom Chris@42: Chris@42: Chris@42: So, for example, to import the system-wide wisdom, you would do: Chris@42: Chris@42: @example Chris@42: integer isuccess Chris@42: call dfftw_import_system_wisdom(isuccess) Chris@42: @end example Chris@42: Chris@42: As usual, the C return value is turned into a first parameter; Chris@42: @code{isuccess} is non-zero on success and zero on failure (e.g. if Chris@42: there is no system wisdom installed). Chris@42: Chris@42: If you want to import/export wisdom from/to an arbitrary file or Chris@42: elsewhere, you can employ the generic @code{dfftw_import_wisdom} and Chris@42: @code{dfftw_export_wisdom} functions, for which you must supply a Chris@42: subroutine to read/write one character at a time. The FFTW package Chris@42: contains an example file @code{doc/f77_wisdom.f} demonstrating how to Chris@42: implement @code{import_wisdom_from_file} and Chris@42: @code{export_wisdom_to_file} subroutines in this way. (These routines Chris@42: cannot be compiled into the FFTW library itself, lest all FFTW-using Chris@42: programs be required to link with the Fortran I/O library.)