Ok - I know that we've all been complaining about METIS for a while now...
but is it possible that we're using it wrong?
I'm writing some code that is very partition sensitive right now (that's
actually the point of the code). In my particular case _bad_ partitionings
actually help me show that my new algorithm is awesome... so the fact that
METIS is returning terrible partitionings is helpful ;-)
BUT: It's also giving me a good look at what METIS is doing and just WTH.
Look at this one. This is a 100x100 perfect square running on _two_ procs:
[image: 2procs.png]
In what world would that be "optimal" in any way?
Surely METIS can do better than that! Has anyone tried playing with the
options we send to METIS? Is it possible that all of our terrible
partitionings we see with it are because we're giving it bad parameters?
Derek
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785351&iu=/4140
_______________________________________________
Libmesh-devel mailing list
Libmesh-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/libmesh-devel
