>>>>> "James" == James Amundson <amundson@fnal.gov> writes:
James> On Tue, 2002-04-16 at 11:52, Raymond Toy wrote:
>> >>>>> "James" == James Amundson <amundson@fnal.gov> writes:
>>
James> -- update bessel function code with slatec (?)
>>
>> Well, I'm finally back to maxima and have looked at some of the slatec
>> stuff. A quick check of derf shows the converted version works just
>> fine, so I don't expect any problems with Bessel functions.
James> Great.
>> However, I'd like some guidance here.
>>
>> Currently the code is spread across several files. I'm thinking of
>> removing (or commenting out) the numeric code in the files and moving
>> them to, say, slatec.lisp which will contain all of the converted
>> routines. I'd also probably want a separate file to hold the f2cl
>> macros and stuff we need.
James> (1) I don't want to make very large changes as we get closer to the
James> 5.9.0 release.
James> (2) I think it's a good idea to arrange the code as you suggest.
James> I think (2) outweighs (1), so go ahead.
How about if I create a directory src/slatec that holds all of the
slatec routines, Fortran and Lisp? Then I can put in a readme or
scripts that convert the Fortran to Lisp appropriately, without
cluttering everything up. Alternatively, defsystem can be told to
automatically convert the Fortran code to lisp and compile up the
result.
This is convenient for me because I've been pulling over just the
needed routines one at a time, but they sometimes contain duplicated
code. So I run fsplit on them, and save all of the individual files,
and duplicates are ignored. Plus, some of the routines I think needed
to be compiled slightly differently from the others due to the
assumptions f2cl makes.
This is all working for me right now with a suitably modified
maxima.system for erf, j0, j1. jn needs a bit more work because the
Fortran code uses an EQUIVALENCE statement that isn't really
necessary.
Ray