Richard Fateman wrote:
> I haven't tried to run it, but I just looked at
> loading maxima-lapack and running your eigensys program.
> It would need to be bfloat-ed if the objective is to use it for
> arbitrary precision rootfinding, using the companion-matrix idea.
>
> A more interesting task might be to make a modified f2cl that generates
> bfloat code.
Actually, I don't have to modify f2cl. Since the code is Lisp, we can
just read in the translated code, modify the type declarations and such,
and we're done. Well, assuming we leave the basic arithmetic operators
as is, and shadow the CL versions and use macros and CLOS to implement
the operations on the appropriate types. I've know you've done
something like this with quad-doubles, and oct has done something similar.
Or we can just transform all of the operators to new ones that work on
the desired types. This would remove the cost of CLOS dispatch.
An interesting idea....
Ray