On Wed, 13 Oct 2010, Ondrej Certik wrote:
On Wed, Oct 13, 2010 at 12:01 AM, Robert Cimrman cimr...@ntc.zcu.cz wrote:
On Tue, 12 Oct 2010, Ondrej Certik wrote:
Hi,
what is the status of the higher order elements? Do you think it will be working (in 3D) till next February?
Ondrej
The chance here it is...
We will need, apart other things, hierarchical base functions for that - are you aware of a BSD-licensed Python/Cython code that could be reused?
Yes, hermes1d:
Yeah, I have thought about that.
it's BSD licensed, and here are the functions:
http://github.com/hpfem/hermes/blob/master/hermes1d/src/lobatto.cpp
It looks like the file could be easily converted to plain old C, right? Or even Cython.
you can also reuse this:
http://github.com/hpfem/hermes/blob/master/hermes1d/src/gen/lobatto.py
but the problem is that the sympy generated C++ code is not precise for high order (~40) because of some numeric issues. The better algorithm is to use recursion, see lobatto.cpp above.
The 1D functions can be used in 2D, 3D to make tensor-product spaces, that's fine. What about simplex elements?
r.