Order (N) Sparse Banded Linear Solver

Matthew Knepley knepley at gmail.com
Thu Oct 4 09:01:54 CDT 2007


On 10/4/07, Dr. Timothy Stitt <timothy.stitt at ichec.ie> wrote:
> Matthew,
>
> Out of curiosity what would you term as very very large? The matrices I
> am using will probably be of order 10^3 x 10^3 elements.

I believe this is fairly small. SuperLU and MUMPS routinely report results for
matrices of order 10^5x10^5 (on large parallel systems).

  Matt

> Thanks,
>
> Tim.
>
> Matthew Knepley wrote:
> > On 10/4/07, Dr. Timothy Stitt <timothy.stitt at ichec.ie> wrote:
> >
> >> Dear PETSc Community,
> >>
> >> I was wondering if anyone knew of any sparse/dense banded linear solvers
> >> that run in order (N)? Does PETSc provide such a solver or link to an
> >> appropriate external solver? There seems to be a lot of links online to
> >> theoretical approaches but we are having difficulties in sourcing an
> >> implementation? I am hoping PETSc might be our saviour ???
> >>
> >
> > Have you tried MUMPS or SuperLU, both available through PETSc's configure?
> > They are not truly O(N), but its hard to tell the difference until you get very
> > very large, or have a nice elliptic PDE on a regular grid. In the
> > latter case, you
> > should use the PETSc DMMG.
> >
> >     Matt
> >
> >
> >> Thanks in advance,
> >>
> >> Tim.
> >>
> >> --
> >> Dr. Timothy Stitt <timothy_dot_stitt_at_ichec.ie>
> >> HPC Application Consultant - ICHEC (www.ichec.ie)
> >>
> >> Dublin Institute for Advanced Studies
> >> 5 Merrion Square - Dublin 2 - Ireland
> >>
> >> +353-1-6621333 (tel) / +353-1-6621477 (fax) / +353-874195427 (mobile)
> >>
> >>
> >>
> >
> >
> >
>
>
> --
> Dr. Timothy Stitt <timothy_dot_stitt_at_ichec.ie>
> HPC Application Consultant - ICHEC (www.ichec.ie)
>
> Dublin Institute for Advanced Studies
> 5 Merrion Square - Dublin 2 - Ireland
>
> +353-1-6621333 (tel) / +353-1-6621477 (fax) / +353-874195427 (mobile)
>
>


-- 
What most experimenters take for granted before they begin their
experiments is infinitely more interesting than any results to which
their experiments lead.
-- Norbert Wiener




More information about the petsc-users mailing list