[ET Trac] [Einstein Toolkit] #941: CCTK_ActiveTimeLevels returns "wrong" number

Einstein Toolkit trac-noreply at einsteintoolkit.org
Sat Jun 2 04:34:39 CDT 2012


#941: CCTK_ActiveTimeLevels returns "wrong" number
----------------------------------------+-----------------------------------
 Reporter:  reisswig@…                  |        Type:  enhancement          
   Status:  new                         |    Priority:  optional             
Milestone:                              |   Component:  Cactus               
  Version:                              |    Keywords:  CCTK_ActiveTimeLevels
----------------------------------------+-----------------------------------
 CCTK_ActiveTimeLevels is supposed to return the current number of active
 timelevels for a given grid function.
 Currently, however, with multiple reflevels (and also maps),
 CCTK_ActiveTimeLevels returns "number of active time levels"*"number of
 reflevels"*"number of maps".
 Unfortunately, e.g. CarpetInterp can get confused when asked to time
 interpolate. A grid function, which technically only has one level per
 reflevel active, will have 3 levels active according to
 CCTK_ActiveTimeLevels when there are 3 reflevels. CarpetInterp may then
 think that indeed there are enough active time levels to time interpolate
 even though there are not. This previously led to segfaults in
 CarpetInterp in certain circumstances (now fixed by avoiding to call
 CCTK_ActiveTimeLevels).

 CCTK_ActiveTimeLevels really calls GroupStorageCrease in
 Carpet/src/Storage.cc. At line, 207, the total number of timelevels is
 computed. Instead of making this a product between number of reflevels,
 timelevels, and maps, would it be possible to instead compute the minimum
 over reflevels, maps? According to Erik, different reflevels/maps can have
 different number of active timelevels. If we return the minimum, then we
 are on the safe side and CCTK_ActiveTimeLevels would return a correct
 number.

-- 
Ticket URL: <https://trac.einsteintoolkit.org/ticket/941>
Einstein Toolkit <http://einsteintoolkit.org>
The Einstein Toolkit


More information about the Trac mailing list