← Back to team overview

dolfin team mailing list archive

Re: Release

 



Anders Logg wrote:
On Tue, Jun 09, 2009 at 01:56:05PM +0200, Johan Hake wrote:
It would be nice to move towards a release in the not too distant
future. Can we try and knock a few points off the TODO list? The stand
out TODO items are:

- Parsing of command-line arguments
This one should be in place. But we need to sort out how we should combine
command line parsing with the PETSc options. Refering to what was
discussed in a previous thread.

Yes, this is done. And I think we agreed that we prefix PETSc options
with "petsc." so

 --petsc.foo=bar


Yes.

results in passing foo=bar to PETSc. Just needs to be implemented.

- Move to new parameter system

I will try to get to this soon but other things are more pressing at
the moment.


Is there much to do or is it just renaming files?

- Move to new XML format/implementation

Are these largely in place already?

I think so. Ola?

Then there's also the parallelization which is close to
finished. Depending on how it goes, it might just need a few more
hours of work to get an initial version going. Everything should be
ready up to actually initializing the (PETSc) matrix, which is also
mostly prepared.


I expect that this will require a lot of testing and quite some small and not so small tweaks, so I wouldn't hold off the next release for parallelisation.

Garth

If anyone's interested, look at the else {} clause in

void PETScMatrix::init(const GenericSparsityPattern& sparsity_pattern)

This needs to call num_nonzeros_diagonal() and num_nonzeros_off_diagonal()
for the sparsity pattern. We also need to finish the insertion
(commented out now) in SparsityPattern.cpp.



------------------------------------------------------------------------

_______________________________________________
DOLFIN-dev mailing list
DOLFIN-dev@xxxxxxxxxx
http://www.fenics.org/mailman/listinfo/dolfin-dev


Follow ups

References