[MITgcm-devel] Problem with pkg/bbl addition in adjoint lab_sea exp.

Mon Aug 8 14:23:04 EDT 2011

Jean-Michel, thanks for isolating problem
and Patrick thank you for reinitialization fix.

What I could also do, if it is needed, is to revert to
an #ifdef construct that is closer to what was there before, i.e.,

#if defined ALLOW_DOWN_SLOPE || ALLOW_BBL
          IF ( useDOWN_SLOPE ) CALL DWNSLP_CALC_RHO
          ELSEIF (useBBL)  CALL BBL_CALC_RHO
          ELSE
#endif
          CALL FIND_RHO_2D
#if defined ALLOW_DOWN_SLOPE || ALLOW_BBL
          ENDIF
#endif

Let me know if above is necessary.

Dimitris Menemenlis

On Aug 8, 2011, at 5:47 AM, Jean-Michel Campin wrote:

> Hi Dimitris,
> 
>> But before doing this, let me figure out why the lab_sea AD test is 
>> taking so much time today ( > 160 longer than yesterday).
> 
> Looks like the pkg/bbl changes in do_oceanic_phys.F are likely
> responsible for many more recomputations in lab_sea AD test
> (because pkg/down_slope is in lab_sea/code_ad/packages.conf):
> yesterday:
> TAF reports 1 Errors and 25 Recomputation Warnings
> today:
> TAF reports 1 Errors and 91 Recomputation Warnings
> 
> And the additional ones are like:
> < TAF RECOMPUTATION WARNING IF_STMT ad_input_code.f:XXX in the_main_loop
> <  triggered by      : rhoinsitu
> <  still to be recomputed : rhoinsitu
> 
> I suggest that this should be fixed (+ get 1 night clean AD testing)
> before turning on useBBL in global_ocean.cs32x15/input/data.pkg
> 
> Cheers,
> Jean-Michel
> 
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel