[MITgcm-devel] restarting cube sphere

[Jean-Michel Campin]

Hi Ichiro,

The restart needs to be exact, meaning that you should be able to
reproduce exactly the same trajectory, whether or not you do 1 long
run or several short runs with restart between.

This has always been the case with the linear free surface
but was not always exact (to the machine precision) with non-linear 
free-surface, until I changed this part (after checkpoint54).
The problem (before c54) was due to truncation errors only, especially
when using the realFreshWater option and time varying fresh water flux.
I checked with several set-up that 2.it + 2.it gives exactly
the same results as 4.iterations (diff of the final pickup file: zero 
difference).
 
Please let me know what version are you using (before or after 
checkpoint 54c_post); 
- if it is older than c54, I would recommend to test with the most 
  recent code.
- if not, I will try to reproduce the problem.

Thanks,

Jean-Michel