Re: [Libmesh-devel] problems with the p_level

[Roy S. <roy...@ic...>]

On Wed, 26 Apr 2006, Steffen Petersen wrote:

> I just tried to clean up some shape function files
> and code the Bernstein shapes in a more generic way
> (including arbitrary orders for quadrilaterals).

Oh, and out of curiosity - can you make the high order Bernstein bases
hierarchic?  (i.e. such that the basis functions at each node for
order p+1 begin with the basis functions at that node for order p)

Currently we only support p-adaptive constraints on elements with
hierarchic bases on side nodes, System::project_vector will only
handle p refinement correctly on elements with hierarchic bases on
every node, and I have no plans to do the work necessary to handle
more general cases anytime soon.  ;-)

Of course, non-hierarchic high-p elements are still useful, and just
getting the correct function numbering on a hierarchic basis can be
tricky (check out my shameful hack at src/utils/number_lookups.h
sometime), but I'd hate for you to start getting bad adaptive p
results without warning you about why.

In fact, should I add a "hierarchic_bases" FEInterface function to let
the library know which elements do and don't have that characteristic?
It would allow the code to switch to the more expensive constraint and
projection calculations that non-hierarchic elements require in the
future, and would allow us to barf with a warning when someone tries
to do p adaptivity and p projections on elements which don't support
them today.
---
Roy