[Mitgcm-support] status

mitgcm-support at dev.mitgcm.org mitgcm-support at dev.mitgcm.org
Wed Jul 9 15:31:14 EDT 2003


Hi guys,

here's a prelim. conclusion of the remaining problems with GM:

1.
I think (almost certain, but I'm still cautious) that the
remaining problem are entirely to be found in the
Visbeck-part of the code.
If I switch this one off and leave all other GMREDI_OPTIONS.h on,
I get reasonable gradient checks for all tapering schemes.

2.
It's very likely that this is again a numerical problem.
I made tangent linear gradient checks and the yield very
similar results to the adjoint gradient checks, so no bug
in the adjoint code this time.
I will play around again in the Visbeck code.
Hope, I won't screw up the forward-GM this time ;o)

Talk of the day from Rehovot:
They have a real(!) espresso machine (not our kind of
monster with only an "espresso" button) in the lab!

A bientot
-Patrick



_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/
Patrick Heimbach
from January 14th to 31st:
.... Weizmann Institute of Science
.... Rehovot, 76100, Israel
.... FON: +972/8/934-4238
.... FAX: +972/8/934-4124
.... mailto:heimbach at mit.edu




More information about the MITgcm-support mailing list