>>>>> "James" == James Amundson <amundson@fnal.gov> writes:
James> Right. We might consider them a bug fix. I'm really not sure. I have to
James> confess that at the time you were discussing your more accurate
James> numerical routines I wasn't paying as much attention as I am now. I
James> think we should re-open the topic for discussion. I should also point
James> out that I think Richard has the final say on whether the new routines
James> go in.
I was playing around with my elliptic function package and noticed
that the values my code computed differed a bit from the analytical
expressions involving the gamma function.
A peek at the gamma routine showed that most of the constants were
given to only 5-6 digits. So I replaced it with a more accurate
version and got the expected agreement. That routine was sent to this
list.
Barton Willis pointed out that the the Bessel functions also didn't
seem very accurate, according to the tables in A&S. I just replaced
these with routines converted from Fortran.
I don't think there's any downside since I'm not changing anything
other than the implementation of the numerical part of the functions.
>> The things that I did a while ago had some
>> simple dependencies on f2cl. Should those be removed before being
>> integrated? It would be much easier if we just added the f2cl
>> macros/functions that we need so that we can just take the converted
>> code without hand modifications.
James> I don't know. I'd have to take a look at the code you have and the f2cl
James> routines.
I didn't send these to the list, so if you want to look at them, I can
send them here or to you directly, as you wish. The Fortran routines
are pretty clean so the Lisp is too.
>> For updates and things, should we send patches here for possible
>> inclusion in to the sources? Or is there some other way?
James> If the patches are small, please send them to the list. If they are not
James> small, say so and we will work something out.
Ok.
Ray