[scilab-Users] Sparse preprocessing
Stephane Mottelet
stephane.mottelet at utc.fr
Sat Oct 2 15:04:25 CEST 2010
Pedro Ledoux <p_ledoux at yahoo.com.br> a écrit :
> The method of finite diferences frequently give us systens of sparse
> matrix. Are you using this one? I had some recent experience with
> it. When the sparse matrix has a 3 diagonal behavior the most
> eficient thing I tested was thomas method. But We can't use it when
> the non 0 elements has no pattern :/
>
> I've used the finite difference method to solve a problem of the
> propagation of a wave in a wave guide. In this the sparse system to
> a "L" shaped wave guide gave me a sparse matrix with no pattern to
> non 0 elements. In the case of waveguide it lead me to a eigenvalue
> problem. Unfortunately I didn't found a function like eig() from
> Matlab, is there something similat in Scilab?
the functions you might be interested in are "spec" and "bdiag"
S.
>
>
> --- Em sáb, 2/10/10, Wojciech Artichowicz <w_artichowicz at o2.pl> escreveu:
>
> De: Wojciech Artichowicz <w_artichowicz at o2.pl>
> Assunto: [scilab-Users] Sparse preprocessing
> Para: users at lists.scilab.org
> Data: Sábado, 2 de Outubro de 2010, 7:30
>
>
>
>
>
>
>
>
>
>
>
>
> Hello,
>
> Fluid mechanics computations often lead to sparse
> matrix systems. Scilab has marvellous sparse functions, however
>
> I cannot find any function for sparse preprocessing. For
> example:
>
>
>
> I have built quite big system with sparse matrix with
> no explicit pattern of non-zero elements. The best way to solve this
> system is
> to use
>
> modified SOR method (my own implementation). My
> question is:
>
> Is there any function in Scilab that would remove
> zero entries from diagonal (which isn?t in this case very easy), and that
> would do some additional stuff like
>
> placing at diagonal the biggest possible element.
>
>
>
> If no, is there any well known not too complicated
> algorithm for such a purpose?
>
>
>
> Thank You in advance,
>
> Regards,
>
> W.A.
>
>
>
>
>
>
>
>
>
More information about the users
mailing list