dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

It should be enough to just run the existing demo and compare with the
old solver, both using the current size of the velocity field (which I
think is v = (0, -10)) and also for v = (0, -100). In the latter case,
there should be some oscillations near the boundary layer on the
dolphin using the old solver with a significant improvement when using
the new. I have limited computer/internet access for another couple of
weeks so I can't try it myself.

/Anders

On Tue, Aug 02, 2005 at 01:21:10PM +0200, jhoffman@xxxxxxxxxxx wrote:
> Garth,
> 
> I have added the SUP-stabilized module. It compiles, but I have not yet
> tried it in a solver.
> 
> /Johan
> 
> > On Tue, Jul 12, 2005 at 06:49:40PM +0200, Garth N. Wells wrote:
> >> Attached are files for SUPG stabilised convection-diffusion. They
> >> simply replace existing files in src/modules/convdiff and
> >> src/modules/convdiff/dolfin.
> >
> > Great! I've been waiting for someone to come along and stabilize the
> > convection-diffusion module.
> >
> > I'll add this, but probably not today (and perhaps not until after my
> > vacation).
> >
> >> The stabilisation terms are computed in a very simple
> >> fashion. Things could be made simpler if it were possible to perform
> >> more algebraic operations on functions with ffc such as taking
> >> square roots, inverses, etc. Would this be possible Anders?
> >
> > Yes, I think it's possible to extend the capabilities of FFC to handle
> > almost any kind of operations applied to Function (but not to
> > BasisFunction). This can be done by replacing each operation
> >
> >     op
> >
> > with
> >
> >     pi op pi
> >
> > where pi is a projection or interpolant onto the finite element space.
> >
> > In the simplest case, pi is just nodal interpolation and op is defined
> > point-wise, so if w is a given function, then
> >
> >     op w --> pi op pi w = pi op sum_i w_i phi_i
> >                         = sum_i op(w_i) phi_i
> >
> > In the case of the square-root, we would just take the square-roots of
> > the expansion coefficients and this might as well be done by FFC as by
> > DOLFIN. (FFC would generate a couple of extra lines of code for doing
> > the operations on the expansion coefficients.)
> >
> > I'll add this to my TODO list.
> >
> > /Anders
> >
> > _______________________________________________
> > DOLFIN-dev mailing list
> > DOLFIN-dev@xxxxxxxxxx
> > http://www.fenics.org/cgi-bin/mailman/listinfo/dolfin-dev
> >
> 
> 
> 
> _______________________________________________
> 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/