dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Tue, Sep 19, 2006 at 07:58:07PM +0200, Johan Jansson wrote:
> On Tue, Sep 19, 2006 at 07:35:28PM +0200, Anders Logg wrote:
> 
> ...
> 
> > ok. I guess there are cases when you need to operate on the vector
> > (like computing 1/u or |u| or things FFC can't do) but otherwise,
> > you should never need to look at the vector.
> > 
> > /Anders
> 
> In this case it's that we have components of a function which do not
> need space discretization, such as the wave equation:
> 
> dot(u_1) = u_2
> (dot(u_2) * v) * dx = -(grad(u_1) * grad(v)) * dx
> 
> I don't know if FFC should be involved with the first equation at all
> since it's not a form. Or what do you think?
> 
>   Johan

It depends on how you discretize it. You could multiply with v in the
first equation as well and lump the mass matrix.

/Anders