S. Newhouse wrote:
> It is interesting that there don't seem to be lisp (I did not see any
> that I recognized) based projects on their list.
I believe LispNYC has been accepted to the Google Summer of Code. There
are quite a few Lisp projects there, and I think they're asking for
people to submit additional projects. Perhaps some Lisp changes for
maxima can be submitted and accepted?
> One direction that seems worth developing is the use of high precision
> floating point calculations. Rational arithmetic takes forever in
> even the simplest types of iterative calculations. But bigfloat
> calculations occur fast enough with even precision of 500 or 1000 digits
> that I think they would be usable in many situations. In this
> connection it would be significant to have tools such as the lapack and
> optimization routines available in extended precision.
I think this was mentioned here, but perhaps not. I did some work with
Richard Fateman on a quad-double implementation. I hacked up the
blas/lapack implementation in maxima to use quad-double arithmetic. It
successfully computed some eigenvalues to quad-double (212-bit) precision.
A similar thing could be done using maxima bigfloat implementation, but
I haven't done it. What made quad-double work with lapack was that
quad-double shadowed all of the basic Lisp arithmetic operations with
its own. If there were something similar in maxima (just needs a
volunteer), then arbitrary precision lapack would be possible.
There are, of course, other possible ways of getting arbitrary precision
lapack into maxima.
Ray