dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Sun, Nov 06, 2005 at 03:23:21PM +0100, Johan Hoffman wrote:

> >> Ok, so the functionality of writing out the time discretization would
> >> rather be in the ODE solver then.
> >
> > Yes, but you would have to pick one of the time discretizations that
> > are implemented: cG(q), dG(q), mcG(q) or mdG(q) for general q.
> >
> > /Anders
> >
> 
> Very nice. Is the Newton jacobian generated automatically in the ODE
> solver/Newton solver? Or is that a job for FFC?
> 
> /Johan

If not provided, it will be generated by numerical differentiation but
the ODE solver also lets the user provide the Jacobian (or rather its
action). The user in this case would be the new PDE class, so FFC
should generate the Jacobian for the PDE class so that the PDE class
can pass it on to the ODE solver.

I have some ideas for how to differentiate forms in FFC. It should be
pretty straightforward.

/Anders