[MITgcm-devel] Re: lab_sea

[Dimitris Menemenlis]

Jean-Michel, thanks for feedback on lab_sea.  As far as I know, the 
tile-ness of seaice dynamics has been fixed.  But there may be new problems?

The edge differences can be made arbitrarily small by setting the 
run-time variable LSR_ERROR.  By default LSR_ERROR=0.0001, which is a 
copromise between allowing a little bit of the tile edge to show, 
especially during initialization, when the sea-ice is not yet in 
balance, and computational cost of the iterative solver.

If it's easy for you to repeat your 2-CPU vs 1-CPU test, can you set 
"LSR_ERROR=1e-40" in data.seaice, to see if the tile-edge difference 
goes away?

There is a runtime flag called SEAICEuseDYNAMICS, which is specified
in data.seaice, which can be used to turn off dynamics.

D.

-- 
Dimitris Menemenlis <menemenlis at jpl.nasa.gov>
Jet Propulsion Lab, California Institute of Technology
MS 300-323, 4800 Oak Grove Dr, Pasadena CA 91109-8099
tel: 818-354-1656;  fax: 818-393-6720