dolfin team mailing list archive

["Matthew Knepley" <knepley@xxxxxxxxx>]

On Tue, Sep 9, 2008 at 3:18 PM, Joachim B Haga <jobh@xxxxxxxxxxxx> wrote:
>>>>> FYI: SCOTCH can now do parallel graph partitioning (haven't tried it
>>>
>>> However, it uses a slow partitioning method and doesn't support weighted
>>> graphs (according to the announcement[1]). But it should only be a problem
>>> for fancy adaptive computations with load balancing/repartitioning.
>>>
>>
>> Ok, too bad since we like fancy stuff. /J
>
> Have you guys looked at zoltan at all? It looks quite versatile, including
> weighted graph and weighted hypergraph methods. I tested it briefly a while
> back, but only the geometric partitioner. Which worked ok but was obviously no
> match for parmetis.

Note that Zoltan uses ParMetis for graph partitioning. The guys that work on it
are mostly involved in developing new hypergraph partitioners and load
balancers.

  Matt

> http://www.cs.sandia.gov/Zoltan/
>
> List of algorithms (for graphs/hypergraphs, I believe only PHG is relevant to
> dolfin, licence-wise):
> http://www.cs.sandia.gov/Zoltan/ug_html/ug_alg.html
>
> -j.
>
> _______________________________________________
> DOLFIN-dev mailing list
> DOLFIN-dev@xxxxxxxxxx
> http://www.fenics.org/mailman/listinfo/dolfin-dev
>



-- 
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