[MITgcm-devel] adjoint file size changed dramatically (for the better)

[Martin Losch]

Hi there,

Olaf and I are running the adjoint MITgcm with seaice, obcs, kpp, and gmredi. This week Olaf did an update of the code that was as "old" as early December. Suddenly the executable requires much less core memory: The same domain size, and even slightly larger nchklev_1 (21 instead of 14), suddenly requires only 6GB instead of 28GB of RAM. Not that I want to complain about that, but I suspect that we made some other mistake somewhere, as most of the changes since December 2009 I found involved additional store directives, so I expected a larger memory footprint with the update. Before we start digging into our configuration, is there any plausible reason for our observation?

Martin