On Tue, 27 Sep 2016, Julian Andrej wrote: > Hi, thanks for the hint. Paul also pointed me to FEMSystem which you > use in GRINS. I don't really understand what FEMContexts purpose is > here though. I also saw this in fem_context.C > > // FIXME - we don't yet handle mixed finite elements from > // different families which require different quadrature rules > // libmesh_assert_equal_to (fe_type.family, hardest_fe_type.family); > > which led me to the thought that i can't use Q2Q1 for Stokes here?
You certainly can. We can handle different quadrature rules so long as they're from the same family of rules: e.g. if you use Q2Q1, then the Q2 rule will integrate the Q1 terms exactly, the FEMContext will figure that out and set up the proper rule automatically, and you're fine. What you can't currently do properly with FEMContext is mix completely different finite element families with different quadrature rule requirements. E.g. if you try to use a Clough-Tocher macroelement for one variable and a Lagrange triangle for another variable, then the Clough-Tocher quadrature rule will integrate both exactly, but FEMContext can't figure that out. --- Roy ------------------------------------------------------------------------------ _______________________________________________ Libmesh-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/libmesh-users
