[Mitgcm-support] Re: GMREDI test code

[mitgcm-support at dev.mitgcm.org]

Ok Patrick, I won't redo it then, since it might affect the adjoint. 
And thanks for looking into the GMredi problem. 
I was going to suggest that small denominator values in the GM code
might be 
the cause of the crashing, but probably you guys know better.

- Benny


Patrick Heimbach wrote:
> 
> No. It's not as simple as that, but it's now simpler
> than ever; I've tried to bring both codes as close
> together as possible to facilitate the plug-in
> of the MITgcm into the ECCO environment.
> Among others, you have to replace the_main_loop.F
> (there is an ecco_the_main_loop in pkg/ecco/).
> That is related to the much more complex cost function
> for ECCO, as well as problems with the adjoint when separating
> the_main_loop and forward_step in the ECCO environment
> (very weired).
> The packages ecco, cost, ctrl, which are in Sealion
> have to be replaced by those in the ECCO environment.
> That's more or less it (I may have forgotten something,
> I assume you don't want to redo it, do you?).
> 
> Linking the verif's of Sealion to ECCO is probably
> trivial, but I don't have the time to do it right now.
> Maybe I'll find some time to at least make the front_relax
> work in ECCO.
> 
> We'll look into the GMRedi problem.
> Ultimately, we want to have GMRedi switched-on in the
> natl_box test case as well, which is available as test case
> for both Sealion and ECCO, thus providing a testable link
> between the two.
> 
> Patrick
> 
> Benny Cheng wrote:
> >
> > Patrick, is it true that to adapt the Sealion release GMREDI test code
> > to the ecco_env version,
> > all you have to do is change the links ??
> >
> > The GMredi test case for Sealion release works as is (with deltaT=1800).
> > But, with our configuration ( slope-limiter =  ldd97 , telescoping
> > grid, deltaT= 3600, etc.) it crashes.
> >
> > - Benny
> >
> > Patrick Heimbach wrote:
> > >
> > > Hi Benny,
> > >
> > > hardly any of the verification experiments of the MITgcm
> > > are properly linked to the ECCO environment.
> > > That is partly because the top-level routine relevant for the
> > > adjoint code generation, the_main_loop.F, differs between
> > > MITgcm and ECCO.
> > > Given the many other work on the code I haven't bothered
> > > adopting those to the ECCO environment, since they
> > > can be easily tested on the MITgcm side.
> > > Sorry, if I wasn't clear on this aspect.
> > >
> > > I am forwarding your bug report to the list support at mitgcm.org
> > > and urge you to do likewise if you find further problems
> > > so we know what they are and can (hopefully) solve them.
> > > Which slope-limiter are you using in your config?
> > >
> > > Cheers
> > > Patrick
> > >
> > > Benny Cheng wrote:
> > > >
> > > > For the record,
> > > > the GMREDI test code is not with Patrick's ecco_env code, as
> > > > inaccurately reported earlier, but is actually in the MITgcmUV
> > > > release_1beta_1
> > > > version. This test code bombs after 2 days with a deltaT = 3600 secs.
> > > > The results are the same with the ldd97 GM scheme. The only time the
> > > > test code works is when deltaT = 1800 sec., so it will not work in
> > > > our configuration.
> > > >
> > > > - Benny
> > >
> > > --
> > > _/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/
> > >
> > > Patrick Heimbach ........................ FON: +1/617/253-5259
> > > Massachusetts Institute of Technology ... FAX: +1/617/253-4464
> > > EAPS, Room 54-1518 ...........................................
> > > 77 Massachusetts Avenue .............. mailto:heimbach at mit.edu
> > > Cambridge MA 02139, U.S.A. ..... http://www.mit.edu/~heimbach/
> 
> --
> _/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/
> 
> Patrick Heimbach ........................ FON: +1/617/253-5259
> Massachusetts Institute of Technology ... FAX: +1/617/253-4464
> EAPS, Room 54-1518 ...........................................
> 77 Massachusetts Avenue .............. mailto:heimbach at mit.edu
> Cambridge MA 02139, U.S.A. ..... http://www.mit.edu/~heimbach/