dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Wed, Apr 18, 2007 at 12:17:49PM +0200, Garth N. Wells wrote:
> Anders Logg wrote:
> > The PDE classes are now back on track again and LinearPDE is working.
> > Garth, could you look at NonlinearPDE?
> >
> 
> Sure, but I don't know how soon.
> 
> > It's somewhat different from before, but in essence the same. Both
> > LinearPDE and NonlinearPDE are sub classes of GenericPDE which have
> > protected members a (bilinear form), L (linear form), the mesh, a list
> > of boundary conditions to be applied and the solution vector x. For a
> > linear PDE, we have the variational problem
> > 
> >    a(v, u) = L(v) for all v
> > 
> > whereas for a nonlinear PDE, we have the variational problem
> > 
> >    L_u(v) = 0 for all v
> > 
> > and the bilinear form a is the derivative of L.
> > 
> > Note that the PDE class owns the solution vector and is responsible
> > for its creation and deletion. The Function class only holds a
> > reference to it.
> > 
> > The next step should be to get Stokes working and for that we need to
> > do some work on both the FFC and DOLFIN sides.
> > 
> 
> What about elasticity as the next step? This should require work on the 
> DOLFIN side only (?), mainly related applying boundary conditions to 
> selected components of a vector.
> 
> Garth

Yes why not, but I think the two are related. The same solution
applies to both since vector elements are now special cases of mixed
elements.

/Anders