Thanks Ben, That worked great! The partitions are much more stable using
parmetis. I didn't even know if that was an option with serial mesh, so I
learned something new.
Cody
On Tue, Jul 9, 2013 at 1:09 PM, Kirk, Benjamin (JSC-EG311) <
[email protected]> wrote:
> On Jul 9, 2013, at 12:40 PM, Cody Permann <[email protected]> wrote:
>
> > While I'm on the subject of partitioners, does anyone know of a way to
> make the Metis partitioner less sporadic in moving the partitions around?
> With each time step of this PF simulation, the set of elements owned by a
> specific process is almost completely disjoint. The partitions look good
> but they move around in a seemingly random fashion which is why I was using
> the centroid partitioner instead. My intention was to make a movie of the
> partitions overlaid with a "grain-tracking" capability that I've developed.
>
> For parmetis, and metis as well I think, there is an "adaptive
> repartitioning" scheme which seeks to improve a portioning with a diffusion
> algorithm (again, IIRC) thereby minimizing movement.
>
> I believe we tap into that in the parmetis interface because the
> expectation is it is important for ParallelMesh, but we've always opted for
> the highest quality partitioning for the SerialMesh and not worrying about
> the redistribution aspects. So I'd say (1) try Parmetis if the erratic
> behavior bothers you, and (2) we could consider implementing the same
> scheme inside metis.
>
> -Ben
>
>
>
------------------------------------------------------------------------------
See everything from the browser to the database with AppDynamics
Get end-to-end visibility with application monitoring from AppDynamics
Isolate bottlenecks and diagnose root cause in seconds.
Start your free trial of AppDynamics Pro today!
http://pubads.g.doubleclick.net/gampad/clk?id=48808831&iu=/4140/ostg.clktrk
_______________________________________________
Libmesh-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-devel
