[MITgcm-devel] [MITgcm-cvs] MITgcm/pkg/shelfice CVS Commit

[Jean-Michel Campin]

Hi Dan,

Regarding the 2 isomip (TAF) AD results, the gradient check give a 
similar agreement (~1.e-9) than before:
primary test: new and old:
new: grdchk  summary  :  RMS of    5 ratios =  3.4900607714524E-09
old: grdchk  summary  :  RMS of    5 ratios =  3.8886209914663E-09
secondary test (htd):
new: grdchk  summary  :  RMS of    5 ratios =  7.2226584095097E-09
old: grdchk  summary  :  RMS of    5 ratios =  6.7287773250031E-09
So I will check-in the 2 updated output.

One thing that would be nice would be to add few lines of description
(and mention the updated reference outputs) in MITgcm/doc/tag-index
regarding the changes you made to pkg/shelfice

Thanks,
Jean-Michel.

On Thu, Apr 23, 2015 at 11:06:28PM +0100, Daniel Goldberg wrote:
> Hi J-M --
> 
> i just checked in the oad results, accuracies were at same level as before.
> TAF accuracies should hopefully remain close to 1e-9. (Please let me know
> if not..)
> 
> Many thanks
> Dan
> 
> On Thu, Apr 23, 2015 at 10:26 PM, Jean-Michel Campin <jmc at ocean.mit.edu>
> wrote:
> 
> > Hi Dan,
> >
> > OK, I will look at the TAF results ;
> > I will not update the TAF output before you check-in the updated OpenAD
> > results.
> >
> > Cheers,
> > Jean-Michel
> >
> > On Thu, Apr 23, 2015 at 09:09:49PM +0100, Daniel Goldberg wrote:
> > > Hi Jean Michel
> > >
> > > I have looked into the OAD experiment and seems that drLoc from
> > > shelfic_forcing.F (where ShelficeBoundaryLayer = .TRUE.) is negative (and
> > > small) in some places, but somehow my change to set it to zero does not
> > > affect the forward results. The adjoint results are affected -- but the
> > > gradient checks are still accurate (accuracies 1e-8 to 1e-9 range).
> > >
> > > Thus I think the changes are ok, as Martin was fine with the change,
> > unless
> > > you think otherwise. I will check in the new results_oadm.txt on
> > > baudelaire. I wonder if you could possibly run TAF and check in the
> > results
> > > -- as long as the gradient check accuracies are still small? (if they are
> > > not still small let me know, i will figure out a way to fix this.)
> > >
> > > Thanks
> > > Dan
> > >
> > > On Thu, Apr 23, 2015 at 3:54 PM, Jean-Michel Campin <jmc at ocean.mit.edu>
> > > wrote:
> > >
> > > > Hi Dan,
> > > >
> > > > The three AD isomip test (2 TAF, 1 OpenAD) have changed, and it's very
> > > > likely due to the changes you made yesterday:
> > > >
> > > > On baudelaire with gfortran (=reference platform):
> > > > OpenAD
> > > >  Y Y Y Y 11> 9< 8 FAIL  isomip
> > > > TAF:
> > > >  Y Y Y Y 11> 9< 8 FAIL  isomip  (e=0, w=4)
> > > >  Y Y Y Y 12>11< 9 FAIL  isomip.htd
> > > > (previously, was all 16)
> > > >
> > > > Can you try to identify which modification is responsible for these
> > > > changes and check if this was expected (in which cae we will need to
> > update
> > > > the reference output) or if there is something to fix in the modified
> > code.
> > > >
> > > > Cheers,
> > > > Jean-Michel
> > > >
> > > > On Wed, Apr 22, 2015 at 09:12:19AM -0400, Daniel N. Goldberg, 54-1423,
> > > > 6172532977,  wrote:
> > > > > Update of /u/gcmpack/MITgcm/pkg/shelfice
> > > > > In directory forge:/tmp/cvs-serv5262
> > > > >
> > > > > Modified Files:
> > > > >       shelfice_forcing.F shelfice_thermodynamics.F
> > > > > Log Message:
> > > > > limit effects of shelficeboundarylayer to where hFacC(kTopC)<1
> > > > >
> > > > >
> > > > > _______________________________________________
> > > > > MITgcm-cvs mailing list
> > > > > MITgcm-cvs at mitgcm.org
> > > > > http://mitgcm.org/mailman/listinfo/mitgcm-cvs
> > > >
> > >
> > >
> > >
> > > --
> > >
> > > Daniel Goldberg, PhD
> > > Lecturer in Glaciology
> > > School of Geosciences, University of Edinburgh
> > > Geography Building, Drummond Street, Edinburgh EH8 9XP
> > >
> > >
> > > em: D <dgoldber at mit.edu>an.Goldberg at ed.ac.uk
> > > web: http://ocean.mit.edu/~dgoldberg
> >
> 
> 
> 
> -- 
> 
> Daniel Goldberg, PhD
> Lecturer in Glaciology
> School of Geosciences, University of Edinburgh
> Geography Building, Drummond Street, Edinburgh EH8 9XP
> 
> 
> em: D <dgoldber at mit.edu>an.Goldberg at ed.ac.uk
> web: http://ocean.mit.edu/~dgoldberg