Hello all,
Present were: Josh, Matt, Roland, Peter, Yosef, Steve, Christine, Erik
Merge rewrite of McLachlan: * Ian not present. Skipped.
The RIT group tried McLachlan starting from ET parfile. Find that McLachlan runs significantly slower than LazEv. Will move discussion to the mailing list.
Number of timelevels: * Roland raises the question of whether CCTK_MaxTimeLevels should be changed to return what CCTK_MaxActiveTimeLevels such that CCTK_MaxTimeLevels is an upper bound on CCTK_ActiveTimelevels at this point in time * consensus: rename CCTK_MaxTimeLevels to CCTK_MaxDeclaredTimeLevels * consensus: require that interface.ccl declared at least 2 timelevels to allow dynamic increase
* no clear agreement yet about whether to have the driver copy data forward in time * currently Carpet internally copies data during time level cycling * suggestion to add restriction that grid functions with multiple timelevels must exist globally * worry about situation where a third party thorn enables more timelevels for variables of a thorn * want to provide as many safety features as we can rather than allow all possible uses of the code
* static_assert: accept with suggestion to add CCTK_ATTRIBUTE_UNUSED
Yours, Roland