dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

Hey!

The plan is that we will use the Sieve library, developed resently by
Matt Knepley and Dmitry Karpeev, to represent meshes in DOLFIN. Sieve
is very powerful and can represent both simplicial meshes and
hexahedral/quadrilateral meshes (and other mesh topologies). It's also
completely parallel.

You can read about Sieve in the document called "PETSc and unstructured
finite elements" available at http://www.fenics.org/docs/.

Sieve is currently available in the development version of PETSc (in
the subdirectory src/dm/mesh/sieve) which can be downloaded from

    http://www-unix.mcs.anl.gov/petsc/petsc-2/developers/index.html

I have just started to look at Sieve and it takes some effort to get
used to the new abstractions (chain, cone, closure, support, star etc)
which are different from the usual vertex, edge, face, cell.

As with the linear algebra, we will most probably create some wrapper
for Sieve in DOLFIN that presents a simplified and specialized
interface suitable for DOLFIN. I don't know the exact time frame, but
I hope we will be well on our way before the end of the year.

As always, any comments, suggestions or code is always welcome. Check
out Appendix C in the manual for some simple guidelines on how to
contribute code to DOLFIN.

/Anders

On Fri, Oct 14, 2005 at 04:41:22PM -0400, Harish Narayanan wrote:
> Greetings everyone.
> 
> I was recently introduced to dolfin and have eagerly been playing with 
> it. Most of the problems I work with are on hexahedral meshes, so I got 
> to poking around seeing what it would take to add 
> hexahedral/quadrilateral cells (additionally aiming to familiarize 
> myself with the code). I've been working closely following what has been 
> done for the existing Triangle and Tetrahedron derivatives of GenericCell.
> 
> I am mailing this list to first make sure this doesn't exist elsewhere 
> and I am not duplicating effort (I didn't find anything similar in the 
> archives of this list over the past few months), and also to find out 
> what the general procedure is--do I just try it, run test problems and 
> e-mail the new source files/diffs to the mailing list if it works? Is 
> this of interest to other people?
> 
> I was also slightly concerned about the constructor for the class Cell 
> which currently generates new Triangles/Tetrahedra based only on the 
> number of nodes with which the constructor is called. I was wondering if 
> this needed to be tweaked to something that knows about the cell type 
> (it came up when I was going through it and Tetrahedra and 
> Quadrilaterals both have four nodes).
> 
> Thanks,
> Harish
> 
> _______________________________________________
> DOLFIN-dev mailing list
> DOLFIN-dev@xxxxxxxxxx
> http://www.fenics.org/cgi-bin/mailman/listinfo/dolfin-dev
> 

-- 
Anders Logg
Research Assistant Professor
Toyota Technological Institute at Chicago
http://www.tti-c.org/logg/