#1795: PITTNullCode results depend on number of processors -----------------------------------+---------------------------------------- Reporter: rhaas | Owner: Type: defect | Status: new Priority: major | Milestone: Component: EinsteinToolkit thorn | Version: development version Keywords: PITTNullCode | -----------------------------------+---------------------------------------- As part of trying to add SphericalHarmonicReconGen to the ET I noticed that PITTNullCode produces different results depending on the number of cores used. This already happens with the %%regression_test%% test if SphericalHarmonicRecon that is already in the ET. The issue is hidden right now since the test uses very lose tolerances of {{{ ABSTOL 1.5e-5 RELTOL 1.5e-4 }}}
Changing them to the standard Cactus accuracies and regenerating the tests shows that the test that was generated using 2 processors fails when run with a single MPI process.
Test data is in the branch rhaas/procdependency .
This needs to be fixed or the arrangement must be marked as broken unless this is the intended behaviour in which case this needs to be documented prominently in the documenation.