[MITgcm-devel] switching seaice_obcs from legacy to devel

[Jean-Michel Campin]

Hi Dimitris,

Thanks for fixing this for the 2+2=4 test.
I've just checked, and the 3 seaice_obcs tests pass.

I've also re-run a quick AD testreport on baudelaire (with gfortran):
http://mitgcm.org/testing/results/2012_12/tr_baudelaire_20121219_9/summary.txt
and lab_sea is back to normal.

Cheers,
Jean-Michel

On Wed, Dec 19, 2012 at 04:34:22PM +0000, Menemenlis, Dimitris (3248) wrote:
> Jean-Michel, I confirm that there is only sufficient OBCS records for 10 time steps.
> What I will do is reduce the test duration to 6 time steps.
> This is sufficient for what is being tested and it avoids making input files bigger.
> This is also what we did in 2008, when the 2+2 test was introduced.
> 
> The reason for not changing time-interval is that the test is supposed to
> reproduce as exactly as possible a subdomain of lab_sea/input.salt_plume
> as a way to test and improve seaice obcs.
> There is a matlab script seaice_obcs/input/compare_lab_sea.m
> which does this comparison.
> Also for longer test integrations, longer open boundary input files
> can be generated using seaice_obcs/input/mk_input.m
> 
> Sorry for breaking so many things with changes yesterday.
> I notice that I also broke the lab_sea ad experiments and that
> you fixed them this morning.  Thanks!
> 
> Dimitris Menemenlis
> 
> On Dec 19, 2012, at 7:45 AM, Jean-Michel Campin wrote:
> 
> > Hi Dimitris,
> > 
> > There is a problem with (I think) the new set of input files:
> > my impression is that there are not enough records in some 
> > input files to run for 4 more iterations (what the 2+2=4 test does), 
> > up to iter=14. Can you confirm ?
> > 
> > And if this the case, it would be nice to allow to run this 2+2=4 test, 
> > either by adding more records or by increasing the time-interval
> > between 2 read.
> > 
> > Cheers,
> > Jean-Michel
> 
> 
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel