[dune-pdelab] Adaptive refinement of Taylor-Hood elements

Lars von Wolff lars.von-wolff at mathematik.uni-stuttgart.de
Mon Apr 15 11:01:07 CEST 2019


Hello Experts,

I am trying to solve coupled flow problems with pdelab. For this I 
construct Taylor-Hood elements in the following way:

     typedef Dune::PDELab::PkLocalFiniteElementMap<GV, DF, RF, 2> V_FEM;
     typedef Dune::PDELab::PkLocalFiniteElementMap<GV, DF, RF, 1> P_FEM;

     V_FEM vFem(gv);
     P_FEM pFem(gv);

     typedef Dune::PDELab::ConformingDirichletConstraints 
ConstraintsAssembler;
     typedef Dune::PDELab::ISTL::VectorBackend<> VectorBackend;
     using OrderingTagL = Dune::PDELab::LexicographicOrderingTag;

     typedef Dune::PDELab::VectorGridFunctionSpace<GV, V_FEM, dim, 
VectorBackend, VectorBackend, ConstraintsAssembler> V_GFS;
     V_GFS vGfs(gv, vFem);

     typedef Dune::PDELab::GridFunctionSpace<GV, P_FEM, 
ConstraintsAssembler, VectorBackend> P_GFS;
     P_GFS pGfs(gv, pFem);

     typedef Dune::PDELab::CompositeGridFunctionSpace<VectorBackend, 
OrderingTagL, V_GFS, P_GFS> NS_GFS;
     NS_GFS nsGfs(vGfs, pGfs);

I need to implement adaptive refinement and it seems that I have run 
into the following open bug:

https://gitlab.dune-project.org/pdelab/dune-pdelab/issues/96

Only the first sub-Gfs of the CompositeGridFunctionSpace is updated when 
calling "nsGfs.update(true)". Is there any workaround to this problem? 
(For example, is there a reasonable way to define Taylor-Hood elements 
without using the LexicographicOrderingTag?)

Thanks in advance,

Lars von Wolff





More information about the dune-pdelab mailing list