[petsc-dev] Possible New Ksp Tests For PETSc

[Jed Brown]

On Wed, Oct 6, 2010 at 16:27, Travis Austin <austin at txcorp.com> wrote:
> Hi,
> I have a DOE SBIR where we are looking at solving high-order finite element
> problems with a sparse preconditioner that is
> inverted using BoomerAMG.  As a result of this work, we have developed
> several 2D and 3D examples that use high-order
> finite elements, i.e., the order of the approximation is an input.  We have
> built these examples to closely conform with example
> src/ksp/ksp/examples/tests/ex4.c
> which solves Poisson's equation on a square using bilinear finite elements.
>   Is there any interest from the PETSc team to
> include this example in the set of ksp examples.  Our hope was that by
> making this available to users through the test suite
> that new solvers could be tested for these types of problems.

Does it "stand alone" in any sense, or does it rely on large external
dependencies like your p-FEM library?  From the above, It sounds like
the tests can stand alone, in which case they would certainly be
welcome.

BTW, I've done some work on this with unstructured grids.

http://dx.doi.org/10.1007/s10915-010-9396-8

I'd be interested to compare experiences.

Jed