On 4/28/10 1:38 PM, Roy Stogner wrote: > > On Wed, 28 Apr 2010, Boyce Griffith wrote: > >> On 4/28/10 1:26 PM, Roy Stogner wrote: >>> >>> On Wed, 28 Apr 2010, Boyce Griffith wrote: >>> >>>> For AMR meshes, should these sorts of constraints be applied only for >>>> the nodes of level 0 elements? Or is this something that I don't have >>>> to worry about explicitly? >>> >>> You need to apply your constraints for all dofs that aren't already >>> constrained - everything except hanging nodes in adaptive h refinement >>> and their equivalent high-p dofs in adaptive p refinement. In your >>> case it shouldn't matter whether or not you constrain hanging nodes as >>> well, but if you do then turn off forbid_constraint_overwrite when >>> adding them; by default we assume adding two different constraints to >>> the same node is a bug. >> >> Looking ahead to inhomogenous boundary conditions --- > > A patch for that would be very appreciated, even if it only worked for > LAGRANGE variables at first, as long as the API was general enough to > be extensible later. This has *forever* been on our list of ideas > that are important/interesting enough to design but not quite > important/urgent enough to implement.
It seems like it should suffice to allow a non-zero right-hand-side for the constraint equation. Am I missing anything? >> is there a mechanism for determining if a particular node is a hanging >> node, so that I can avoid re-constraining those DOFs? > > DofMap::is_constrained_dof() Each time that user_constrain() is called, have the constraints already been reset? I.e., in user_constrain(), is it safe to assume that the only constraints are hanging nodes (or their p-refinement equivalent) and periodic boundaries? -- Boyce ------------------------------------------------------------------------------ _______________________________________________ Libmesh-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/libmesh-users
