On LLNL ibm frost (pwr3 nighthawk 2 - very similar to seaborg), I ran both cam 3.1 and cam 3.0 (tag 3_0_5) for FV dycore, 2x2.5, 4x4 decomposition with 4 OMP threads, for 5 days. Cam 3.1 takes about 7% longer. I did each case twice. Increased time is more or less across the board. Physpkg takes a good 10% longer. Main FV dycore takes longer as well. Land model does not show time increase. I did not do any other comparison cases. Is this to be expected?
