#1795: PITTNullCode results depend on number of processors ------------------------------------+--------------------------------------- Reporter: rhaas | Owner: Type: defect | Status: confirmed Priority: major | Milestone: Component: EinsteinToolkit thorn | Version: development version Resolution: | Keywords: PITTNullCode ------------------------------------+---------------------------------------
Comment (by yosef@…):
The remaining 'issue' is that the interpolator is sensitive to the domain decomposition. This is due to the fact (I think) that grid spacing is calculated locally via something like dspace = (upper - lower)/nzones. I'll attach a parfile where dspace is a power of 2, which should give near identical answers (to within the default tolerance) on any number of processors