[MITgcm-devel] global_ocean.cs32x15/input.icedyn

[Jean-Michel Campin]

Hi Martin,

Regarding this new test-experiment, I did not have time to look 
at this "high sensitivity" issue (pgf & ifort test fail @ level 2,
which is not really good on a 10 iterations test).
Given that, I am not surprised by (2); regarding the metric terms,
tanPhiAtU should be zero in this experiment, and this is a little
bit strange. Did you also rearrange some of the computation ?

And finally, I generally prefer to have separated fixes (it makes
easier to track in case of a problem), but since all those changes
are in the same S/R (right ?), it's less a problem, and you can 
check-in all at once.

Jean-Michel

On Mon, Jun 18, 2007 at 11:16:03AM +0200, Martin Losch wrote:
> Good morning Jean-Michel,
> 
> I am a little puzzled by your new experiment
> 1. I made a small change (bug fix, not yet checked in), that only  
> affecteds metric terms in seaice_lsr.F (all involve tanPhiAtU), but  
> still global_ocean.cs32x15/input.icedyn.
> How is that possible, I thought that with usingCurvilinearGrid  
> = .true., all tanPhiAtU/V = 0?
> 
> 2. This experiment is extremely sensitive, right? This is what I get  
> on my Apple with g77
> Y Y Y Y 11 13 13 14 16 16 16 16 16 13 13 14 16 16 13 14 16 FAIL   
> global_ocean.cs32x15
> Y Y Y Y  2  2 13  7  7 13 14  9  6 10 10  4  6 10 10  5  6 FAIL   
> global_ocean.cs32x15.icedyn
> Y Y Y Y 11 16 16 16 16 16 16 16 16 13 13 14 16 13 13 14 16 FAIL   
> global_ocean.cs32x15.thsice
> Y Y Y Y 10 16 16 16 16 16 16 16 16 13 13 13 16 13 13 13 16 FAIL   
> global_ocean.cs32x15.viscA4
> 
> 3. Have made changes that affect lab_sea at the truncation level  
> (rearrange a few terms), that is cg2d is accurated to 13 digits, but  
> global_ocean.cs32x15.icedyn fails with 2!
> 
> I also found a serious bug in the LSR solver that will affect all LSR  
> results dramatically. Now,
> shall I just check in all of the above (small bug only affecting  
> metric terms, big bug affecting everything, and rearrangement of  
> terms affecting truncation, which really changes  
> global_ocean.cs32x15.icedyn) at the same time, glossing over all  
> subtelties?
> 
> Martin
> 
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel