On Dec 6, 2013, at 11:06 , Ian Hinder ian.hinder@aei.mpg.de wrote:
Hi,
Erik and I have found and fixed a bug in the way Carpet handles restriction and synchronisation. Numerical results on coarse grids might change as a result of this. Previously, the results would have dependent on the number of processes.
I believe there is still an error when higher order restriction operators are used (which makes sense only for cell-centred simulations). In this case, ghost zones are synchronized, but boundary and symmetry conditions are not applied. This may lead to differences for simulations employing symmetries to refined regions.
See https://trac.einsteintoolkit.org/ticket/1499 for more details. Erik, should this be backported to the release branch at some point?
Yes, it probably should.
-erik