On Thu, 14 Jan 2016, Roy Stogner wrote: > You know what? It doesn't for me either, my previous reports to the > contrary. But that use case *should* work, and does work on lots of > other meshes I've tried. I'll keep looking into it.
And now I think I understand the problem. Our solution restart file node ids are made canonical by using libHilbert ordering... but our solution restart file *partitioning* is made canonical by nothing more than faith in a guarantee which ParMETIS doesn't make. In the long term the fix is to finally implement parallel xda/xdr output of the mesh files, too, then don't repartition when reading such a mesh. In the short term the workaround ought to be to use WRITE_SERIAL_FILES; I'm not sure that's working but if it's not I'll try to fix it ASAP. --- Roy ------------------------------------------------------------------------------ Site24x7 APM Insight: Get Deep Visibility into Application Performance APM + Mobile APM + RUM: Monitor 3 App instances at just $35/Month Monitor end-to-end web transactions and take corrective actions now Troubleshoot faster and improve end-user experience. Signup Now! http://pubads.g.doubleclick.net/gampad/clk?id=267308311&iu=/4140 _______________________________________________ Libmesh-users mailing list Libmesh-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libmesh-users