dolfin team mailing list archive

["Garth N. Wells" <g.n.wells@xxxxxxxxxx>]

Anders Logg wrote:
> Sounds good. The problem with Metis is that it's not open-source. The
> license restricts both usage and distribution.

True. I had a quick look at some others, and only Scotch 
(http://www.labri.fr/perso/pelegrin/scotch/) appears to be both actively 
maintained and LGPL. The partitioning step is pretty simple at the 
moment (one function call to METIS), so I can change to another library 
later without much effort. I won't touch configure.ac until this is 
sorted out (flags are added manually at the moment for the test file).

Garth



> /Anders
>
>
> On Fri, Dec 01, 2006 at 12:13:55PM +0100, Garth N. Wells wrote:
> > I just started this morning. I have the mesh partitioned (using METIS), 
> > and a MeshFunction contains the partition number for each cell. I plan 
> > initially to let each processor have a copy of the entire mesh and just 
> > assemble its partition. Distributing the mesh will come later.
> >
> > Garth
> >
> >
> > Anders Logg wrote:
> > > Excellent! I have an MSc student starting after Christmas to work on
> > > parallelization of the mesh data structures, including partitioning.
> > > We should try to coordinate the effort.
> > >
> > > /Anders
> > >
> > >
> > > On Fri, Dec 01, 2006 at 12:02:25PM +0100, Garth N. Wells wrote:
> > > > OK. I'm working on parallel assembly now, so I'll take a look at the 
> > > > degree of freedom numbering when I get to assembling vector-valued problems.
> > > >
> > > > I'll check in the beginnings of parallel assembly after the new release 
> > > > is made.
> > > >
> > > > Garth
> > > >
> > > >
> > > > Johan Hoffman wrote:
> > > > > Sounds good. This is the natural way I think.
> > > > >
> > > > > /Johan
> > > > >
> > > > > > Should we consider changing the degree of freedom numbering such that
> > > > > > degrees of freedom at a node are located next to each other in the
> > > > > > matrix/vector? This will be needed for parallel assembly. It may also
> > > > > > speed up the assembly of sparse matrices by allowing blocks to be
> > > > > > assembled.
> > > > > >
> > > > > > Garth
> > > > > > _______________________________________________
> > > > > > DOLFIN-dev mailing list
> > > > > > DOLFIN-dev@xxxxxxxxxx
> > > > > > http://www.fenics.org/mailman/listinfo/dolfin-dev
> > _______________________________________________
> > DOLFIN-dev mailing list
> > DOLFIN-dev@xxxxxxxxxx
> > http://www.fenics.org/mailman/listinfo/dolfin-dev
> _______________________________________________
> DOLFIN-dev mailing list
> DOLFIN-dev@xxxxxxxxxx
> http://www.fenics.org/mailman/listinfo/dolfin-dev