Comment (by Miguel Zilhão):

i've finally found some time to give this a try.

from my very brief test, though, i think i'm seeing the same sort of behaviour with and without this commit (i've checked the current branch ET_2019_03 vs the branch rhaas/commstate_free_buffers). memory usage as reported by systemstatistics for the simple run ML_QC3_memcheck.par with and without the commit looks the same in both, and it resembles what i had observed before (though now i’ve performed a much shorter run) in that the memory increases a lot after regridding. am i not looking at the right things, or should i be looking at longer evolutions?

--
Ticket URL: https://bitbucket.org/einsteintoolkit/tickets/issues/2194/memory-increase-during-regridding