Modeling Maxima's bigfloats as a separate package would make it more
plausible to
drop in an alternative, especially MPFR, at some future stage, into Maxima.
I would prefer that MPFR be callable from within Maxima, rather than
looking to
outside control. E.g. Sage would presumably look at a problem
and try to decide between MPFR and Maxima, making Maxima appear to be,
and in fact, less capable than Sage in this respect [assuming MPFR is
"better" than what the 2 or 3 people
here have done with bfloats, and that MPFR will be pushed further by
others.]
I thought we had better root-finding in the FORTRAN routines. There is a
Laguerre
method that is supposedly better than Jenkins-Traub. Converting that to
lisp bfloats
might be good. (In fact, I think I have a Laguerre method written in
Lisp from first
principles, but it was done by a student who didn't totally inspire
confidence.)
Good luck.
RJF
Robert Dodier wrote:
> On 12/12/08, Raymond Toy <raymond.toy at ericsson.com> wrote:
>
>
>> Perhaps this can be placed in cpoly.lisp or maybe as a contrib in
>> share/numeric.
>>
>
> Since it 's a bigfloat version of an existing algorithm,
> I would put it in the same place as the existing one.
>
>
>> Also, I should probably also convert rpoly-sl to support bfloats, and
>> then allroots and bfallroots will be functionally identical.
>>
>
> OK by me.
>
> best
>
> Robert
> _______________________________________________
> Maxima mailing list
> Maxima at math.utexas.edu
> http://www.math.utexas.edu/mailman/listinfo/maxima
>