[MITgcm-devel] changes in pkg/ptracers

[Jean-Michel Campin]

Hi,

I am going to check-in changes in pkg/ptracers, to allow to
start each tracer at a different time.
I had to disable this feature for the adjoint (was getting
too many recomputation), but still have an issue with
tutorial_global_oce_biogeo/input_ad : the printed cost function
is zero but the forward gradient did not change at all.
(I tried to use an older version of TAF, 2.1.15, but the
problem is still there, i.e., cost-function=0).
And the reason is that with my modification, PACKAGES_INIT_VARIABLES
is called at the end of the backward run (was not the case before
although I have the impression that INITIALISE_VARIA was already
called) which reset the cost-function.
I don't know how this can be fixed.

Cheers,
Jean-Michel