> On 23 Feb 2017, at 17:08, Daniel Shapero <shapero.dan...@gmail.com> wrote: > > The vectors you have in this equation P(u+h*v) =..., which of those have > constrains distributed and which zeroed? > If you assembly matrices with ConstraintMatrix.distribute_local_to_global() > the diagonal elements corresponding to constrained DoFs > ... > Along the same lines: does you test work with non-homogeneous Dirichlet BC? > The discussion above applies there as well. > > Hi Denis! This sounds like the kind of thing I would have overlooked. I > assemble both the PDE operator f and its linearization using > ConstraintMatrix.distribute_local_to_global. The vector field u has > non-homogeneous Dirichlet boundary conditions. I took another vector field ub > that's equal to u along the boundary, and set v = u - ub. So v should be 0 > along the boundary, but I never explicitly enforced that using e.g. > MatrixTools::apply_boundary_values or something. > > Are you thinking that I should assemble without the constraints, and then > only add them later in, say, the Newton solver?
You should NOT need to, just be careful doing matrix-vector products. And if you enforce constraints, I would do ConstrantMatrix::distribute() as you would have extra constraints in h-FEM case. How you setup those constraints should be irrelevant, most likely you would call VectorTools::interpolate_boundary_values(). Denis. -- The deal.II project is located at http://www.dealii.org/ For mailing list/forum options, see https://groups.google.com/d/forum/dealii?hl=en --- You received this message because you are subscribed to the Google Groups "deal.II User Group" group. To unsubscribe from this group and stop receiving emails from it, send an email to dealii+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
