fenics team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Fri, Jun 09, 2006 at 11:24:00AM -0500, Matthew Knepley wrote:
> On 6/7/06, Anders Logg <logg@xxxxxxxxx> wrote:
> 
>     On Thu, Jun 08, 2006 at 02:13:39PM -0400, tomtzigt wrote:
>     > Any word on the Sieve and its integration into FEniCS yet?
>     >
>     >
>     >
>     > Theo
> 
>     Not that I have heard, but I'm working on a much improved rewrite of
>     the DOLFIN mesh library that will most likely be merged into the DOLFIN
>     main tree within a week or two.
> 
> 
> Cool, you have to tell me what's in it. So for the Sieve stuff, it now appears
> to be working in a visco-elastic-plastic post-rupture easrthquake code
> called PyLith (formerly Tecton). Dmitry is cleaning up the Field
> representation,
> but basically the alpha now has everything working.

It's not as fancy as Sieve. It's a light-weight minimal and very
simple implementation of a dimension-independent storage for mesh
entities and connectivity. Our current mesh library is used here at
Simula and they need something simple and fast. I haven't benchmarked
the new mesh yet but I suspect orders of magnitude speedup (for
computation of connectivity and iteration over entities).

> Petsc just switched to Mercurial, and you can pull the latest version from
> 
>     hg clone old-http://www-unix.mcs.anl.gov/~petsc/BuildSystem
>     hg clone old-http://www-unix.mcs.anl.gov/~petsc/petsc-dev
>
> and look at the examples in src/dm/mesh/examples/tutorials first, and then
> there are some preliminary KSP examples. Peter Brune (student of Rob Kirby)
> will be here this summer helping make unstructured multigrid work in this
> framework.
> 
>    Matt

Cool. How's Mercurial working out so far? Have you had any problems
with it?

/Anders