dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Fri, Oct 20, 2006 at 01:48:37PM +0200, Garth N. Wells wrote:
> 
> 
> Anders Logg wrote:
> > Garth, do we really need assembleResidualBC() in FEM?
> > 
> > I looks to me that
> > 
> >   assembleResidualBC(A, b, x, ...);
> > 
> > does the same thing as
> > 
> >   applyBC(A, b, ...);
> >   b -= x;
> > 
> > ?
> >
> 
> Not quite. The difference is at nodes where Dirichlet boundary
> conditions are applied. In this case, the "residual" is the difference
> between the computed solution x (not b) and the boundary condition value
> x_bc.
> 
> For linear problems it is easy because prior to the solve step b = x =
> 0.0. The complications arise when using Newton steps or solving
> time-dependent problems with non-zero Dirichlet bc's.
> 
> Garth

ok, I see that now.

/Anders