On 6/9/12 2:40 PM, Richard Hennessy wrote:
> I was not doing a two symbolic matrix multiply. I was doing two, 400 x
> 400 numerical matrices created by genmatrix(). This is a pure
This is just a simple matter of programming. :-) There already exists
multi-core matrix multiply routines. All you (the generic you, not you,
Richard) need to do is hook them up to maxima using the FFI of your
favorite lisp. There will be some cost of converting maxima's matrix
format to the format wanted by the matrix routines, but these are
already solved pretty much for the LAPACK interface.
But if I were doing this all the time, I think I would not use maxima,
even if it were fast enough. It would seem more natural to me to use
octave or R or your favorite matrix language.
Ray