dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Tue, Dec 06, 2005 at 09:51:35PM +0100, Johan Jansson wrote:
> On Tue, Dec 06, 2005 at 07:38:14PM +0100, Garth N. Wells wrote:
> > I've had a look at Preconditioner.h and Preconditioner.cpp to try to transfer
> > precondtioner stuff in KrylovSolver to Preconditioner. The code in
> > Preconditioner.cpp is pretty cryptic though. Is it possible to simply it?
> > 
> > Garth  
> > 
> 
> Good idea. It should be possibly to simply move everything from
> KrylovSolver and put it as a function/member of Preconditioner.
> 
> The only issue is as Anders describes, that there are two
> "preconditioner databases", one in PETSc and one consisting of
> preconditioners defined using the DOLFIN interface. Maybe the enum
> solution would unify this? Then there would be only one database
> (DOLFIN's), which could have either PETSc or DOLFIN preconditioners in
> it.

It also needs to be possible for a user to implement her own
preconditioner which is not in the databse.

But the enum solution is probably fine. At least it's the simplest and
it doesn't break anything.

/Anders

> I've cleaned up the Preconditioner implementation a bit, I hope it's
> easier to follow now. This is the reference I've used now (before I
> had some other reference/example code which was much more detailed):
> 
> http://www-unix.mcs.anl.gov/petsc/petsc-as/snapshots/petsc-current/src/ksp/pc/impls/jacobi/jacobi.c.html#PCJacobi
> 
> 
>   Johan
> 
> _______________________________________________
> 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/