On 8 Dec 2015, at 17:43, Ian Hinder ian.hinder@aei.mpg.de wrote:
Hi,
I just created a ticket (https://trac.einsteintoolkit.org/ticket/1846) to describe a problem that Barry Wardell and I found with the current (2015_11) release. It has probably been present for a while, but is not present in the 2014_05 release. This would show up as large grids and memory usage, and low speeds, in simulations with multiple components per refinement level, as Carpet seems to be replacing these components with their containing bbox. There are two simple workarounds described in the ticket.
The problem has now been fixed (thanks for the quick response Erik!) on both master and ET_2015_11. To update, you should be able to run
cd Cactus/repos/carpet git pull
You can check that you have the fix with "git log", which should show
Author: Erik Schnetter schnetter@gmail.com
CarpetLib: Correct bboxset2::size
Return the number of points in the set, not the volume. Correct integer overflow on pre-C++11 compilers.
We also now have a test (https://build.barrywardell.net/job/EinsteinToolkit/lastCompletedBuild/testRe...) on master which checks the grid structure when using multiple components on a refinement level, so hopefully this sort of problem would be caught earlier in future.