Re: [ublas] Inversion of a sparse matrix

Subject: Re: [ublas] Inversion of a sparse matrix
From: James C. Sutherland (James.Sutherland_at_[hidden])
Date: 2009-02-09 11:59:14

To my knowledge there is no sparse iterative solver interface for  
ublas.  You may want to consider using a package designed for sparse  
matrix solves.  If your sparse matrix is very big, or if you may want  
to do a parallel solve, you may want to consider PETSc or Trilinos.   
Both are national lab packages that are suitable for solving large  
linear systems in serial or parallel.

James

On Feb 9, 2009, at 4:47 AM, Moritz Lenz wrote:

> Hi,
>
> I'm looking for an implementation that inverts a sparse matrix, and  
> returns a "full" (or dense, or whatever it's called...) matrix.
>
> I currently assign it to a dense matrix and use the LU-decomposition  
> which I found here:
> http://www.crystalclearsoftware.com/cgi-bin/boost_wiki/wiki.pl?LU_Matrix_Inversion
> but it's rather slow, and currently the bottleneck of my program -  
> any thoughts on how to improve it are very welcome.
>
> Cheers,
> Moritz
> _______________________________________________
> ublas mailing list
> ublas_at_[hidden]
> http://lists.boost.org/mailman/listinfo.cgi/ublas