On 12/11/06, Raymond Toy <raymond.toy at ericsson.com> wrote:
> How am I supposed to run this? Esparse(A,27,30,1e-10)?
>
> If so, I have no problems running it on my copy of cmucl (on solaris).
> However, Esparse just says I should increase MAX or decrease ERR.
Esparse works OK for me too, as others have reported.
Esparse runs much faster if makelist(sum(A[i,j]*v_old[j],j,1,s),i,1,s)
is replaced by A . v_old -- the . indicates noncommutative
multiplication in Maxima.
Dunno if we've discussed this before, but there is a Fortran
package named ARPACK which computes a subset of the
eigenvectors and eigenvalues. IIRC it is much faster than
computing all eigenvectors & eigenvalues when just a few
need to be computed. ARPACK is suitable for sparse matrices
because only a matrix-vector product needs to be computed --
no need to store the entire matrix. If there is in interest in such
problems, ARPACK could be a candidate for translation to Lisp.
FWIW
Robert