I'm running into what appears to be either a bug or a misunderstanding on my part. I'm trying to run a 3D adaptive simulation for the first time on a phase-field problem and running into problems accessing elemental DOFs that I thought that I owned but apparently whose values are not in the PETSc vector.
My understanding has been that this assumption should hold for all libMesh simulations: local dofs + ghosted dofs = semi-local dofs However, I'm not sure that this is even remotely correct, at least for elemental dofs. My code is rather complex and doesn't limit itself to any of the normal element iterators in libMesh so to keep myself from accessing DOFs that the current processor doesn't own, I've always relied on making sure the element reports back that it is semi-local before attempting to access the elemental (or nodal DOFs) and this has worked for me until yesterday when I pushed beyond the 2D (with and without adaptivity) and 3D (without adaptivity) simulations I've been running. I only hit this bug when trying to run a 3D simulation with adaptivity. I need some clarity on accessing ghosted DOFs. I believe John discovered that the existing ghosted element iterator may be incorrect (because it uses the semi-local predicate). I'm sure I can always ask the PetscVector if it contains a particular DOF before accessing but I shouldn't have to do this. BTW - I built a small MOOSE example using generated mesh (3D adaptivity) that can easily replicate the issue. I know I can make it even smaller by dropping my fake physics and just using a "Marker" to adapt several elements in the mesh and allowing things to fall apart. Help! Cody ------------------------------------------------------------------------------ What NetFlow Analyzer can do for you? Monitors network bandwidth and traffic patterns at an interface-level. Reveals which users, apps, and protocols are consuming the most bandwidth. Provides multi-vendor support for NetFlow, J-Flow, sFlow and other flows. Make informed decisions using capacity planning reports. http://pubads.g.doubleclick.net/gampad/clk?id=1444514421&iu=/41014381 _______________________________________________ Libmesh-users mailing list Libmesh-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libmesh-users
