From: Michele Milano <milano@ifd.mavt.ethz.ch>
Newsgroups: comp.parallel.mpi
Subject: Re: large matrix inversion
Date: Tue, 25 Aug 1998 14:47:31 -0700
Organization: NAS - NASA Ames Research Center, Moffett Field, CA
Message-Id: <35E330F2.DCC5F728@ifd.mavt.ethz.ch>
References: <35E1D24D.B1B78994@ifd.mavt.ethz.ch>
    <01bd75d5$8a5768b0$e8b7060a@eraui187> <omaf4spykx.fsf@prancer.cs.utk.edu>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit


Victor Eijkhout wrote:

> "R. Manus" <renaud.manus@hol.fr> writes:
>
> > > I need an MPI subroutine to fastly perform inversion of large, non
> > > sparse, symmetric positive definite matrices.
>
> Really? You sure you don't just want to solve linear systems?

Actually this is the reason why I need the matrix inversion.

>
>
> > Get PETSC package (www.ornl.gov)
> > It's a large collection of routines for scientific calculation based on
> > MPI.
> >
> > Great :-)
>
> Oh, it's great all right, but it doesn't do distributed dense systems.
> For that you can use Scalapack. Look on http://www.netlib.org/.

I would have preferred a smaller, ad hoc subroutine, but I think that
finally I will install one of these libraries. Thanks to you both for your
help!

Michele