Since its release, a few important corrections have been added to the current version ET_2013_11 "Noether". In fact, one error in particular can lead silently to wrong results, so we strongly suggest to update checkouts. Updating is easy: from within your Cactus checkout (cd into it), execute
./bin/GetComponents -u
New tags for this "point release" have been created: ET_2013_11_v1, but it is not recommended to use these for a checkout, to be able to easily receive possible future updates. The tar ball has been updated as well.
In the following we briefly describe the changes included in this update.
The ET_2013_11 "Noether" release of the Einstein Toolkit contained an error in Carpet which was introduced since the previous ET_2013_05 "Gauss" release. The problem was related to synchronising ghost zones after data is restricted (copied from fine grids to coarse grids). The problem has been observed to result in incorrect data on coarse grids for multi-process runs with more than 2 refinement levels, and data on coarse grids which depends on the number of processes used. This problem has now been fixed in the ET_2013_11 release branch.
The MoL integrator RK87 was not functional in the Noether release, triggering an error message instead. This message was meant to warn that RK87 is not implemented for grid arrays. However, this error message and abort was always triggered, even for evolutions of only grid functions (the usual case). This has been corrected, and now only triggers if evolutions of grid arrays are requested with RK87.
In IDLinearWaves the calculation of one coeffient for the m=2 mode of the Teukolsky waves was missing, resulting in usage of uninitialized memory for that mode. This is not a regression from a previous release, but was included in the point release because of its trivial and non-invasive fix.
Frank Loeffler for the Einstein Toolkit maintainers