[MITgcm-devel] lab_sea AD test

[Jean-Michel Campin]

Hi Patrick,

I run the AD lab_sea test (on faulks) with my latest changes 
and got plenty of NANs.
Then I tweak the Makefile by adding -version 1.9.8
(in 1rst position, before -z -reverse) to the variable 
AD_TAF_FLAGS
and did again a make adall (got the same taf_ad.log,
appart from the TAF version), but this time it's OK, does 
not produce any NAN, and I checked that 
precision_derivative_grad & precision_derivative_cost are
the same as in results/output_adm.txt

So, something strange in the new TAF version (but only
for the lab_sea experiment, all the others pass with
no changes).

Jean-Michel