dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Mon, Feb 01, 2010 at 08:03:32PM +0100, Marie Rognes wrote:
> Anders Logg wrote:
> >On Mon, Feb 01, 2010 at 08:00:18PM +0100, Marie Rognes wrote:
> >>Anders Logg wrote:
> >>>On Sun, Jan 31, 2010 at 03:26:30PM -0800, Johan Hake wrote:
> >>>>Hello.
> >>>>
> >>>>The mesh registered in the forms (a and L) is the original mesh from the first
> >>>>itteration, as it is cached in the form_data. ffc.jit does not recognize when
> >>>>a mesh is changed.
> >>>This shouldn't matter. The code should be the same when assembling on
> >>>different meshes. Something else must be going wrong. Don't know what
> >>>yet.
> >>>
> >>A quick fix might be to not cache the form_data in jit_form?
> >
> >Then it has to be recomputed which can take some time. What is it in
> >form_data that changes when we move to a new mesh?
> >
>
> original_arguments/coefficients

I think this should be fixed now. I changed the caching so it caches
on id(form) instead of form.

Then the new form will get new form data even if it has the same
__repr__.

--
Anders

Attachment: signature.asc
Description: Digital signature