Thanks, it turned out to be the solver. Changing it to a direct solver by adding -ksp_type preonly -pc_type lu -pc_factor_mat_solver_package mumps solved this issue. I need to investigate what happened with the parallel iterative solver.
Miguel On Nov 12, 2015 13:59, "Roy Stogner" <[email protected]> wrote: > > On Thu, 12 Nov 2015, Miguel Angel Salazar de Troya wrote: > > Do these functions access the ghost elements? >> > > No. > > What I need is to have each processor iterate through its elements, >> calculate something and then add that value to a vector with size >> equal to the number of active elements. If two processors are >> accessing the same element, the result will be incorrect. What would >> be the best way to iterate for this goal? >> > > active local elements are what you would want to iterate through, yes. > --- > Roy > ------------------------------------------------------------------------------ _______________________________________________ Libmesh-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/libmesh-users
