#1106: Possible issue in TmunuBase with support_old_CalcTmunu_mechanism=no ------------------------------------------+--------------------------------- Reporter: wolfgang.kastaun@… | Owner: Type: defect | Status: review Priority: minor | Milestone: Component: EinsteinToolkit thorn | Version: development version Resolution: | Keywords: ------------------------------------------+---------------------------------
Comment (by rhaas):
Nono. I had to check this recently, remember. The scheduling is such that the flesh processes the schedule.ccl files for each thorn in alphabetical order (since this is the order in which thorns are initialized) of the thorn names (hence the alphabetical) but within each schedule.ccl file (and hence within each thorn), the order is (initially, before BEFORE/AFTER are enforced) given by the order in which the statements appear (see discussion in http://cactuscode.org/pipermail/users/2012-July/003163.html and eg the schedule that results from TmunuBase where TmunuBase_Zero is not the last item appearing in Cactus' schedule output even though it is alphabetically after TmunuBase_AddTmunu).
schedule.ccl are translated almost verbatim (or at least one could treat SCHEDULE and STORAGE as some funny macros) into C code in Cactus/configs/configname/bindings/ScheduleThornName.c and the schedule sorter itself never looks at routine names.
If you want alphabetical ordering, try Cactus::schedule_sort_mode = "ascending" (or "descending" for reverse alphabetical ordering).