I think the Morton ordering is not amenable to just linear cutting, which is essentially what we're doing. http://en.wikipedia.org/wiki/Z-order_curve
Maybe we should be splicing it up some other way? I'm not sure I see anything *wrong* with the hilbert result, but some are definitely worse than others The metis stuff: wow! There are options that can get passed into the partitioner, and we are leaving them as defaults. Probably worth playing with some of those… -Ben On Nov 7, 2013, at 3:34 PM, Cody Permann <codyperm...@gmail.com> wrote: > I'm seeing some less than optimal partitioning for several processor counts > using both Metis and Hilbert SFC. Also from what I can tell, Morton SFC is > completely broken. I realize that I'm splitting a small mesh among a lot of > processors, but the behavior is still pretty bad. > > The following pictures are using mesh generation (16x16), no refinement: > https://drive.google.com/folderview?id=0B8csupg5nQaady14UFZjVUMzSm8&usp=sharing > ------------------------------------------------------------------------------ > November Webinars for C, C++, Fortran Developers > Accelerate application performance with scalable programming models. Explore > techniques for threading, error checking, porting, and tuning. Get the most > from the latest Intel processors and coprocessors. See abstracts and register > http://pubads.g.doubleclick.net/gampad/clk?id=60136231&iu=/4140/ostg.clktrk_______________________________________________ > Libmesh-devel mailing list > Libmesh-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/libmesh-devel ------------------------------------------------------------------------------ November Webinars for C, C++, Fortran Developers Accelerate application performance with scalable programming models. Explore techniques for threading, error checking, porting, and tuning. Get the most from the latest Intel processors and coprocessors. See abstracts and register http://pubads.g.doubleclick.net/gampad/clk?id=60136231&iu=/4140/ostg.clktrk _______________________________________________ Libmesh-devel mailing list Libmesh-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libmesh-devel
