From: Roy Stogner <roystgnr@ic...>  20060828 23:27:39

On Mon, 28 Aug 2006, Manav Bhatia wrote: > I would appreciate if someone could help me understand the > conceptual difference between the elem_soln and nodal_soln (as is > being used in the FE method nodal_soln). The elem_soln is a vector of coefficients corresponding to the particular degrees of freedom on your finite element space; the nodal_soln is a vector of coefficients corresponding to the function values at your elements' geometric nodes. Except for isoparametric Lagrange elements, they're different vectors. > If I want to implement a new FE, and not use AMR, Why wouldn't you want to use AMR? If your FE space has a simple (discontinuous, C0, C1) level of interelement continuity, the DofMap and the FE::compute_proj_constraints function will do all of the hard work for you. > should I be concerned about the implementation of this? Yes. FE::nodal_soln() isn't for AMR, it's actually used by the MeshOutput classes for plotting. It's not a tough implementation, though; if I were less lazy I'd factor it out of all the separate fe_clough.C, fe_hermite.C, etc. instantiations, because it can be implemented with essentially the same function for all of them. There are minor optimizations possible for many FE types, but except for LAGRANGE I don't think we do any of them.  Roy 