ffc team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Sat, Jan 30, 2010 at 08:32:50PM +0100, Kristian Oelgaard wrote:
>
>
> On 30 January 2010 00:24, Anders Logg <logg@xxxxxxxxx> wrote:
> >On Fri, Jan 29, 2010 at 10:25:52PM +0100, Kristian Oelgaard wrote:
> >>
> >>I don't know if it is a bug in the new quadrature code, it depends on how you look at it. :)
> >>The generated code in tabulate_tensor is exactly identical (apart from formatting of float values and some comments).
> >>However, this means that I no longer reset the values of A before going crazy with the += operator. This might explain the different results. I guess I should add back the reset A code since we decided to move away form allowing different representations on the same subdomain.
> >>
> >>Kristian
> >
> >I think that might be the problem. You need to set all entries to zero
> >in case you happen not to assign to all of them in the code. DOLFIN
> >does not reset the element tensor when it assembles so you need to do
> >that as part of the generated code.
>
> Did you check this again? I have pushed a changeset that should fix this issue.
>
> Kristian

Yes, it works perfect. The DOLFIN unit tests all pass again.

--
Anders

Attachment: signature.asc
Description: Digital signature