Comment (by Roland Haas):

tcmalloc is indead the better solution to this and may well actually take care of your out-of-memory situation altogether. It is very easy to install and can be used as well on clusters. In the past we have also found that running with tcmalloc can give a noticeable speedup for some codes compared to running just with glibc (a speedup measurable in the physical_time_per_hour measurement, though the code was not actually Cactus based).

Having said this, that is also the reason why I did not use it for this test since I wanted to see if I can still get the increase in rss (which does not come from mallinfo but from reading /proc so is "safe") before/after the patch. In this case I think mallinfo() is still trustworthy since I am running with less than 2GB per rank anyway.

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