Hi Roy, I just checked in a change, I believe this is fixed now --- at least reduced_basis_ex6 works with --enable-parmesh for me now in r6280.
But I'd appreciate if you could have a look at the diff. In RBEIMConstruction::evaluate_mesh_function I use the mesh function on a ghosted vector now, and broadcast the result. Incidentally, is there a better way to get the root_id for the broadcast? (I used Parallel::sum with libMesh::processor_id()) David On 10/31/2012 05:33 PM, Roy Stogner wrote: > > And it doesn't look like it's ParallelMesh's fault. > > RBEIMConstruction::evaluate_mesh_function is querying a MeshFunction > object, and that query returns the _out_of_mesh_value (an empty > DenseVector by default) if it's asked for the value of a point that's > not within any active local or active ghosted elements. > > Even with SerialMesh, this would normally fail, except that > RBEIMConstruction built a serialized solution vector to hand to the > MeshFunction, so results should be correct but performance should > scale poorly. > > enrich_RB_space() and update_RB_system_matrices() are doing a localize > to that serialized vector, so they must already be parallel_only > methods - would it be possible to change them to only query local > elements, and then if necessary sync the resulting data afterwards? > > For now I'm going to disable reduced_basis_ex6 in the --enable-parmesh > case. > --- > Roy ------------------------------------------------------------------------------ Everyone hates slow websites. So do we. Make your web apps faster with AppDynamics Download AppDynamics Lite for free today: http://p.sf.net/sfu/appdyn_sfd2d_oct _______________________________________________ Libmesh-devel mailing list Libmesh-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libmesh-devel
