specfun update



>>>>> "Barton" == Barton Willis <willisb@unk.edu> writes:

    Barton> I don't think CL supports setting  the floating point
    Barton> rounding mode  (maybe CMUCL does as an extension); at

Definitely true.  Plus getting rounding right on x86 is pretty
difficult with its 80-bit registers.

    Barton> least for now, I think it would be okay to have error
    Barton> bounds that were almost certainly correct.  Really it

I think the difficulty will be how to handle intervals when you're
given just a single number x.  You could create an interval
[x*(1-eps), x*(1+eps)] and work with that.  The interval package for
cmucl was geared towards handling Lisp's interval types, not real
interval stuff.

    Barton> seems that the hard part of doing interval analysis is
    Barton> blending an interval type into Maxima. If you still have a
    Barton> copy of your code and if you could donate it, I'll look at
    Barton> it.

The core of the code was originally written to be completely portable,
but I don't know if I have that anymore.  I (or any one else) can pull
out some stuff from cmucl since it's truly public domain.

    Barton> The specfun name was inherited from Macsyma.  

Perhaps the name should be kept the same for compatibility?  Did
Macsyma's specfun include more than just orthogonal polynomials?  But
I think the choice is yours.

Ray