[MITgcm-devel] Problem with lab_sea adjoint test

[Patrick Heimbach]

The lab_sea adjoint is broken due move of
IF (DIFFERENT_MULTIPLE...
into SEAICE_DYNSOLVER
(if an IF/ENDIF block is added somewhere extending over
multiple pages without indentation, it might be good practice
to comment where the ENDIF statements belongs to).
The recomputation/storing structure is thoroughly messed up now.

Second thing is I can't find the
#ifdef SEAICE_BREAKS_VERIFICATION
that used to be in seaice_growth anymore
(it was bracketing code that broke the Arctic state estimation).
There are recomputation issues also in seaice_growth.

Third thing is that it seems code as of 1 or 2 weeks ago
no longer work for more than 1 iteration of Arcitc 1-year,
which I don't understand yet.

-Patrick



On Jan 19, 2008, at 1:20 PM, Jean-Michel Campin wrote:

> Hi Patrick and others,
>
> The AD testing of the 3 lab_sea experiments is taking much more time
> than it use to (more than 100 times more):
> On Jan.15 early morning (before David's modifs which prevent
>  to get a test on Jan 16 & 17):
>> lab_sea.noseaicedyn User: 29.0885785 System: 0.889865002 Wall  
>> clock: 30.82884
>> lab_sea.noseaice    User: 28.0057424 System: 0.799878975 Wall  
>> clock: 28.845278
>> lab_sea             User: 35.6285841 System: 0.846872005 Wall  
>> clock: 37.1708851
>
> and since Jan.18:
>> lab_sea.noseaicedyn  User: 4384.07557 System: 32.1821097 Wall  
>> clock: 4509.27592
>> lab_sea.noseaice     User: 4258.07266 System: 27.9497508 Wall  
>> clock: 4400.82334
>> lab_sea              User: 4714.99844 System: 30.8383126 Wall  
>> clock: 4829.53805
>
> Would be good to fix this.
>
> Jean-Michel
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel

---
Patrick Heimbach | heimbach at mit.edu | http://www.mit.edu/~heimbach
MIT | EAPS 54-1518 | 77 Massachusetts Ave | Cambridge MA 02139 USA
FON +1-617-253-5259 | FAX +1-617-253-4464 | SKYPE patrick.heimbach