dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Thu, Apr 09, 2009 at 11:26:49AM +0200, Anders Logg wrote:
> On Tue, Apr 07, 2009 at 12:04:33PM -0500, Robert Kirby wrote:
> > Hi all, 
> > In addition to my matrix-free technique I'm implementing behind the
> > KrylovMatrix interface, I'm also
> > trying to do a sanity check and build the matrix for comparison.
> > 
> > However, there are "set" methods for uBLASSparseMatrix and PETScSparseMatrix
> > (and uBLASDenseMatrix and just Matrix),
> > but not "add" methods in the pydolfin interface.  Is this straightforward to
> > add, or is there some issue here?
> > 
> > I can always do manual indexing since this is just for comparison (if I'm after
> > performance, I won't use Python anyway),
> > but for research in new methods that aren't ready to automate yet, it can be
> > helpful to have as much of DOLFIN exposed
> > to Python as possible.
> > 
> > Thanks,
> > Rob
> 
> I've made an attempt to add add. See if it helps.

This didn't work. I tried to see what's been done for set() in
dolfin_la_{pre,post}.i and did the same for add().

Does anyone know why set() is exposed and not add()? Johan?

-- 
Anders

Attachment: signature.asc
Description: Digital signature