[MITgcm-devel] bottom_ctrl_5x5 with latest changes in generic_advdiff

[Jean-Michel Campin]

Hi,

The latest changes I made in pkg/generic_advdiff (for OBCS)
are affecting the gradient (not the cost) of experiment
bottom_ctrl_5x5, which is strange since pkg/obcs is not
used (not even compiled) in this experiment.

the gradient check are not worse (with gfortran):
 new grad-check        previous grad-check
 -2.4301652872483E-05  2.1834576695579E-06
 -1.3699229403596E-04  7.9297453342886E-04
 -5.6692644379019E-07 -1.0471477467855E-03
  1.0175127298717E-04 -1.2977611548681E-04
  8.0446560912451E-04  8.8851931593537E-04
 -3.4441448658695E-04 -1.2336868908112E-03
  1.2587843741296E-04  9.4122897678772E-04
 -1.3934928245529E-05  4.4966888357534E-05
  2.7184821029813E-05  1.8678744130973E-05

This is simlar to the differences we were getting between g77
and gfortran on baudelaire.

I am going to update bottom_ctrl_5x5/results/output_adm.txt
with the new output file.

Cheers,
Jean-Michel