[MITgcm-devel] taf dropping cost_averagesfields

[Matthew Mazloff]

Hello,

Well adding in the cost_averages call worked for the forward model --  
the cost came out as expected.  But the cost was never realized by  
the  adjoint model -- so the adjoint variables were all zero.  As far  
as I can tell adcost_obcsn is being called -- I have no idea why the  
adjoint model is all zeros....

I will work on debugging this, but please let me know if you have any  
ideas.

-Matt





On Nov 28, 2010, at 11:18 AM, Matthew Mazloff wrote:

> Hello,
>
> I am running the divided adjoint.  When my only cost term is the open
> boundary cost:
> #define ALLOW_OBCSN_COST_CONTRIBUTION
> taf is dropping the call to
>                   call cost_averagesfields( mytime,mythid )
> in the main loop
>
> It leaves the final call, the one at the end of the_main_loop -- but
> the actual one inside the loop is omitted.
>
> What's interesting is the adjoint model has both calls -- so there is
> the call to
>                   call adcost_averagesfields( mytime,mythid )
> in the adjoint loops.
>
> So I could put this call back in to the forward model....I think it
> will work...and I'll try doing this now
>
> I wanted to know if anyone has come across this type of behavior  
> before?
>
> Thanks
> Matt
>
>
>
>
>
>
>
>
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel