F2CL: Fortran to Lisp converter
Welcome to f2cl - a Fortran to Common Lisp Translator
Contained in this directory are source code files and some documentation. The translator is written in Common Lisp making installation simple.
If this has been extracted as a part of CLOCC, the way to build f2cl is to simply run "make system" from a shell, as is usual with CLOCC packages.
A second method is to use defsystem from CLOCC. Then load
f2cl.system, and then finally run
(mk:oos "f2cl" :compile).
A third method is to use asdf. You can load
f2cl.asd and run
(asdf:oos 'asdf:load-op :f2cl).
Finally, a fourth method, if you have none of the above, is to manually run everything as follows:
Start your favorite Common Lisp implementation use the function
compile-fileto compile each of the source files:
f2cl0.l f2cl1.l f2cl2.l f2cl3.l f2cl4.l f2cl5.l f2cl6.l f2cl7.l macros.l
Load up all of the files
(load "f2cl0.l") (load "f2cl1.l") (load "f2cl2.l") (load "f2cl3.l") (load "f2cl4.l") (load "f2cl5.l") (load "f2cl6.l") (load "f2cl7.l") (load "macros.l")
to load all the compiled files.
Converting Fortran Code
To use f2cl:
will convert the file
<path>/fortran.f to Lisp code and places the
result in the file
will also run
compile-file on the Lisp file so you can load it
directly into your lisp.
For those anxious to use the translator without studying the documentation here is a short list of restrictions that may result in obscure errors:
- input code is assumed to be valid Fortran 77
- no tabs are permitted in the source,
- $comments should only be turned on if the Fortran code has comments exclusively within subroutines,
- linebreaks must occur within whitespace,
- spaces are required to separate symbols.
Note also that an intermediate file called "prep.tmp" is produced by the preprocessing stage of the translation.
These are the options available to
File to contain Lisp code
Verbose output. Default = NIL. Mostly for debugging.
Prune unused labels. Default = NIL.
Include Fortran comments in the Lisp output. Default = NIL
Variables in DATA statements are automatically SAVE'd. Default = T.
Declarations of array sizes are relaxed in formal parameters to functions. That is, any array length declarations (except lower limits) are ignored if possible, like old Fortran used to. Default = T.
:ALWAYS, all assignment statements automatically coerce the RHS to the appropriate type for the assignment. If
:NEVER, coercion never happens. If
:AS-NEEDED, f2cl applies coercion if it thinks it is needed. Default =
The extension to use for the output file, if needed. Defaults to
T, the temporary file is not deleted. This is mostly for debugging f2cl. Default =
The type of array f2cl should use. Should be
:array. For some compilers, there can be significant speed up if the array can be declared as simple-arrays. But this is incompatible with array-slicing, so care must be used if you choose
:simple-array. Default =
NIL, f2cl assumes that, whenever we do an array reference in a call to a subroutine or function, we are really passing a subarray to the routine instead of just the single value, unless f2cl knows the function takes a scalar arg that is not modified. Default =
A string or symbol specifying what package the resulting code should be in. (Basically puts a
(in-package <p>)at the top.) Default is "COMMON-LISP-USER".
:DECLAIMDeclaim compilation options (Basically puts a
(declaim <declaim>)at the top.) Default is none.
NIL, any structures definitions for common blocks are defined when converting this file. Otherwise, the structures for the common blocks are expected to be defined elsewhere. This should be used only once for one subprogram that will be used to define the common block. See below for more information. Default is
Float format to use when printing the result. Default is
Instead of defining a common block as a structure with the same slot names as variables in the common block, the common block is defined as a set of arrays. The actual common block variables are defined as offsets into these arrays. For more information see below. This mimics the memory layout of how Fortran treats common blocks. Default =
Using Converted Code
Once you've converted the code, you do not need to load up all of f2cl to use the converted code. In fact, you only need f2cl0.l and macros.l to define the necessary packages and functions used by the converted code. Actually, you really only need the defpackage for f2cl-lib in f2cl0.l.
For a more detailed list of issues and notes, see src/NOTES.
We highlight just a few issues here.
Block data statements
In Fortran, block data statments are used to initialize common blocks. Since Fortran executables are loaded and run just once, this is not a problem. However, in Lisp, this might not be true, and you may want to run the main program many times. Thus, it is up to you to run the block data initializer at the right time, as needed. f2cl cannot know when and where to call the initializer.
F2cl converts common blocks to structures. However, common blocks may be referenced in several different files, so the user must tell f2cl when to define the structure. Use the :declare-common parameter to tell f2cl to define the structure. This should be done exactly once for each common block that is defined. This should also be done for the first file that is compiled and loaded, so that subsequent files know about the definition.
In addition, there is another option, :common-as-array. This changes how f2cl handles common blocks. A rather common use of common blocks has the same common block using different variable names. For example, one routine might have
COMMON /foo/ a(10), b, i(4)
and another might say
COMMON /foo/ b(9), c, d, j(2), k(2)
In Fortran, this is perfectly acceptable. Normally, f2cl expects all common blocks to use the same variable names, and then f2cl creates a structure for the common block using the variable names as the names of the slots. However, for a case like the above, f2cl gets confused. Hence, :common-as-array. We treat the common block as an array of memory. So this gets converted into a structure somewhat like
(defstruct foo (part-0 (make-array 11 :element-type 'real)) (part-1 (make-array 4 :element-type 'integer4)))
(In a more general case, we group all contiguous variables of the same type into one array. f2cl and Lisp cannot handle the case where a real and integer value are allocated to the same piece of memory.)
Then in the individual routines, symbol-macrolets are used to create accessors for the various definitions. Hence, for the second version, we would do something like
(symbol-macrolet (b (make-array 9 :displaced-to (foo-part-0 *foo*) :diplaced-offset 0)) (c (aref (foo-part-0 *foo*) 9)) (d (aref (foo-part-0 *foo*) 10)) (j (make-array 2 :displaced-to (foo-part-1 *foo*) :displaced-offset 0)) (k (make-array 2 :displaced-to (foo-part-1 *foo*) :displaced-offset 2)) ...)
Thus, we access the right parts of the common block, independent of the name. Note that this has a performance impact since we used displaced arrays.
While not necessary, f2cl can do a significantly better job in generating code if the functions are compiled in the correct order. This means any function, F, that is called by another function, G, should compiled first. In this way, f2cl can determine the calling conventions for F, and generate the appropriate call for F in G. This is important if F takes an array argument and G passes a slice of an array to F, or conversely if F takes a simple variable, and G calls F with an array reference.
If this is not done, the user may have to modify either the Fortran code or the resulting Lisp code to pass arguments correctly.
F2cl cannot always determine whether a slice of an array should be used or just the single element.
See also the file src/NOTES which contains a change log. But there are also various notes about about restrictions and enhancements on various features supported by f2cl.
The translator was written by Kevin Broughan and Diane Koorey Willcock at the University of Waikato. Reports should be sent to firstname.lastname@example.org and should include examples of Fortran code which fails to translate or which translates incorrectly.
Major changes have be written by Raymond Toy and the entire translator is now part of CLOCC, with permission from Kevin Broughan. Send bug reports and other comments to http://clocc.sourceforge.net.
The code is also placed under the GPL, by permission of Kevin Broughan. The exception is macros.l which is released under the LGPL so that it can be incorporated into other packages.