Hi Praveen,
then you might as well run integration_loop twice. during the first pass you
can
assemble all the face terms (say, using MeshWorker::Assembler::CellsAndFaces
with separate_faces = true).
then, the second pass takes care of the cell integration and finishes up the
matrix/function assembly (use the previously computed face values).
it will be more expensive but it's a "quick" solution...
does that make sense?
-- Mihai
________________________________
Von: Praveen C <[email protected]>
An: mihai alexe <[email protected]>
CC: deal.ii <[email protected]>
Gesendet: Mittwoch, den 13. April 2011, 19:17:20 Uhr
Betreff: Re: [deal.II] MeshWorker integration_loop: Do faces first
Hello Mehai,
I have a term like this on each cell K
mu(K) * int(K)( grad(W) * grad(phi) )
where
mu(K) = int(dK)( some jump terms )
which is integral over boundary of K. Hence I need all face integrations to be
done first.
Thanks
praveen
On Wed, Apr 13, 2011 at 10:08 PM, mihai alexe <[email protected]> wrote:
Praveen,
>
>
>the assembly/integration loop is executed in parallel by several threads
>(regardless of debug/optimized mode). You shouldn't use the sequence of
>printed
>outputs as an indication of the true order of execution.
>
>
>From the documentation, it looks like cells_first is not what you want.
>
>
>Could you elaborate on your problem more? Why would you need the jump terms in
>the cell integration? Is this required by your integration scheme? This
>requirement sounds a bit odd to me. What equation are you trying to solve?
>
>
>Perhaps there is a way to get around this problem by re-engineering the
>implementation a bit.
>
>
>Regards,
>Mihai
_______________________________________________
dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
