I just traced down a "ParallelMesh bug" to an underlying bug in the Hermite finite elements: we use the element id to decide whether to use certain cached values or not, but it's possible for the id of the last element of one assembly pass to match the id of the first element in the next assembly pass due to renumbering rather than due to identity. It's a very improbable bug, but I think I could get this to trigger on SerialMesh too with just the right coarsening+refinement step.
For now I'm disabling caching there; if the resulting performance drop causes any screams, I'm open to suggestions. Cache based on element centroid rather than element id? --- Roy ------------------------------------------------------------------------------ LogMeIn Central: Instant, anywhere, Remote PC access and management. Stay in control, update software, and manage PCs from one command center Diagnose problems and improve visibility into emerging IT issues Automate, monitor and manage. Do more in less time with Central http://p.sf.net/sfu/logmein12331_d2d _______________________________________________ Libmesh-devel mailing list Libmesh-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libmesh-devel
