[Carpet] regridding?
Bela Szilagyi
szilagyi at aei.mpg.de
Thu May 4 17:35:53 CEST 2006
A few more tests I have run since indicate that, indeed, the problem is not
neccessarily within CarpetRegrid. If, at any point, I have reasons, again,
to be suspicous of the regridding, I'll let you know.
Thanks.
Bela.
On Thursday 04 May 2006 17:31, Erik Schnetter wrote:
> On May 4, 2006, at 09:30:17, Bela Szilagyi wrote:
> > Can one (or -- is one supposed to) adjust carpet grid structure by
> > steering
> > dynamically the parameter CarpetRegrid::coordinates ?
>
> Yes, this is supposed to work.
>
> [schnipp]
>
> > A short time after this parameter change AHFinderDirect quit with the
> > complaint
> >
> > WARNING level -1 in thorn AHFinderDirect processor 2 host ic0141
> > (line 78
> > of /home/szilagyi/Cactus/configs/harm/build/AHFinderDirect/jtutil/
> > error_exit.cc):
> > -> ***** modulo_reduce(): no modulo value is fuzzily within
> > specified range!
> > x = 0 xmod = 6.28319
> > [xmin,xmax] = [0.610865,1.74533]
> > ==> xx = 0
> >
> > which suggest incorrect coordinate range. AHFinderDirect had no
> > problem
> > working until the parameter change. Erik, just to make your
> > argument against
> > Jonathan's code more difficult, this is YOUR branch.
>
> This tries to find a value between 0 and 2pi; this looks like it is
> finding a value on a sphere, not in 3D. AHFinderDirect does not care
> at all about the shapes of the 3D grids, it only calls the
> interpolator. Your problem may be that the spacetime variables have
> gone bad, and the AH finder now gets confused internally because
> certain algorithms do not converge.
>
> If you have a stack backtrace and variable values, Jonathan may be
> able to help pointing at the real source of the problem. Of course,
> ultimately still the change of the grid structure could be to blame.
>
> -erik
--
Bela Szilagyi
----------------------------------------------------
Max-Planck-Institut für Gravitationsphysik
Albert-Einstein-Institut
Tel: +49 331 567 7632
Fax: +49 331 567 7649
----------------------------------------------------
More information about the developers
mailing list