[MITgcm-devel] another bug in growth.F ?

[Dimitris Menemenlis]

Arghhh!

Patrick, would you be able to try your finite-difference test with 5 or 6 
intermediary growth.F routines (that I can generate and send you) in order to 
isolate which of the 5 or 6 changes that have taken place cause the forward 
sensitivity to blow up.

Dimitris


> Prelim. result:
> Your checkins make the finite-difference (not adjoint!)
> sensitivities increase by 4 orders of magnitude.
> That's worrying.
> (hard to see how the f.d. gradients could go wrong).
> -p.