[MITgcm-devel] [MITgcm-cvs] MITgcm/pkg/obcs CVS Commit
Jean-Michel Campin
jmc at ocean.mit.edu
Tue Apr 19 17:06:43 EDT 2011
Hi Matt,
Regarding this:
> I hope JMC or yourself can automate it to work with the testers...
I can finish/polish the new verification experiment (e.g., .cvsignore
files, or even results/output_ad.txt if you prefer me to do it)
As for the input_ad and code_ad, it's better if you add them
(rather than me)i, but otherwise, let me know.
Thanks,
Jean-Michel
On Tue, Apr 19, 2011 at 11:20:17AM -0700, Matthew Mazloff wrote:
> Hi Martin,
>
> I cleaned up
> ctrl_init_obcs_variables.F
> and removed a lot of the comments, instead referring to
> c MITgcm/verification/obcs_ctrl/input/gendata.m
>
> When I switch to
> >CALL MDSFINDUNIT( dUnit, myThid )
>
> I get a (I think benign) recomputation error so I will work on
> fixing that. Once I straighten that out I can check everything in.
>
> And yeah, I switched xx_obcs?period=0.0, in data.ctrl
>
>
> I hope JMC or yourself can automate it to work with the testers...
>
> Thanks!
> Matt
>
>
>
> On Apr 19, 2011, at 8:54 AM, Martin Losch wrote:
>
> >Hi Matt,
> >
> >I am fine with the code, except for maybe the reading of the mode-
> >matrix ctrl_init_obcs_variables.F:
> >1. the unit 117 is hard wired, that may not be portable, can we
> >use "CALL MDSFINDUNIT( dUnit, myThid )" instead (stolen from
> >mdsio_read_field.F), or is that too pkg-specific?
> >2. the unit 117 is not closed
> >
> >3. in data.ctrl, why is xx_obcs?period=9999999999.0? Do you mean
> >"0" but that didn't work? Please try now with xx_obcs?period=0.0,
> >it should work now.
> >
> >Martin
> >
> >On Apr 19, 2011, at 3:13 PM, Jean-Michel Campin wrote:
> >
> >>Hi Matt,
> >>
> >>I think we can start with this (but would wait for Martin's
> >>feedback).
> >>
> >>2 little things:
> >>1) the name of the experiment to add in verification
> >>(I prefer dir with lower case letters, so that it's ordered
> >>in the same way on all platforms).
> >>could be something with obcs_ctrl in it (similar to bottom_ctrl),
> >>but you need to decide.
> >>2) I would be tempted to reduce even further the number of
> >>time-steps (presently 72, 24.h).
> >>Could we try to run for 6.h (18.iter) ?
> >>(does not need to be changed in MITgcm_contrib/SOSE/BoxAdj, can
> >>be done directly in verification/*/input_ad)
> >>
> >>Cheers,
> >>Jean-Michel
> >>
> >>On Mon, Apr 18, 2011 at 05:08:48PM -0700, Matthew Mazloff wrote:
> >>>Hello Martin and JMC,
> >>>
> >>>OK, so I updated
> >>>MITgcm_contrib/SOSE/BoxAdj/
> >>>and added in gradient check -- it runs for 1 day and checks gradient
> >>>on western obcs. It passes with 1-FDGRD/ADGRD < 1e-5.
> >>>
> >>>For now I believe it is just checking the temperature gradient,
> >>>however, and it should probably check the normal velocity gradient
> >>>too. It's been awhile since I used the gradient check package and
> >>>I'm having trouble remembering how to do that for obcs...I can work
> >>>on that later, or maybe one of you can fix that in
> >>>data.grdchk....its not an immediate issue though...
> >>>
> >>>Ideally I would like to see
> >>>ctrl_getobcs(n,s,e,w).F ctrl_init_obcs_variables.F and OBCS.h
> >>>merged into the main code and then the BoxAdj code, minus those 6
> >>>files of course, made into a verification experiment. Let me know
> >>>what you think
> >>>
> >>>Thanks
> >>>Matt
> >>>
> >>>
> >>>
> >>>
> >>>On Apr 18, 2011, at 3:50 PM, Matthew Mazloff wrote:
> >>>
> >>>>Hi Jean-Michel -- sorry about calling you Martin last email :o)
> >>>>
> >>>>so -- anyway, the BoxAdjoint is very simple -- but does use
> >>>>modes to
> >>>>control OBCS. To be most useful this code (modifications to
> >>>>ctrl_getobcs(n,s,e,w).F ctrl_init_obcs_variables.F and
> >>>>OBCS.h) should
> >>>>be checked into the main branch, thus minimizing code in the
> >>>>verification folder. Should I go ahead and check in this
> >>>>stuff (all
> >>>>mods are within a preprocessor block), or do you want me to just
> >>>>update SOSE/BoxAdj and you can incorporate it all into /
> >>>>verification/
> >>>>BoxAdj? How's best to proceed?
> >>>>
> >>>>Thanks
> >>>>Matt
> >>>>
> >>>>
> >>>>
> >>>>
> >>>>On Apr 18, 2011, at 2:31 PM, Jean-Michel Campin wrote:
> >>>>
> >>>>>Hi Matt,
> >>>>>
> >>>>>Regarding lab_sea, it has already many things turned on,
> >>>>>so probably better not to add obcs.
> >>>>>
> >>>>>But the SOSE/BoxAdj does not look too big (64x64x8) to be turned
> >>>>>into a verification test. Are all the input files there ?
> >>>>>We would need to turn on grdchk. Anything else ?
> >>>>>
> >>>>>Cheers,
> >>>>>Jean-Michel
> >>>>>
> >>>>>On Mon, Apr 18, 2011 at 01:36:24PM -0700, Matthew Mazloff wrote:
> >>>>>>Hi Martin,
> >>>>>>
> >>>>>>I would love an obcs adjoint verification experiment.
> >>>>>>We can make
> >>>>>>a new one -- or perhaps just modify a present one...e.g. lab sea
> >>>>>>would be a good candidate. What do people prefer.
> >>>>>>
> >>>>>>As for the obcs control modes code, I have not checked it in the
> >>>>>>main branch yet... It is still in
> >>>>>>http://mitgcm.org/viewvc/MITgcm/MITgcm_contrib/SOSE/BoxAdj/code_ad/
> >>>>>>
> >>>>>>-Matt
> >>>>>>
> >>>>>>
> >>>>>>
> >>>>>>
> >>>>>>On Apr 18, 2011, at 1:38 AM, Martin Losch wrote:
> >>>>>>
> >>>>>>>Hi Matt,
> >>>>>>>
> >>>>>>>what's the status of the verification experiment for
> >>>>>>>adjoint obcs?
> >>>>>>>I cannot remember ... is the ball in my half of the
> >>>>>>>playing field?
> >>>>>>>Having this experiment would help to avoid these
> >>>>>>>recurring issues!
> >>>>>>>
> >>>>>>>Martin
> >>>>>>>
> >>>>>>>On Apr 18, 2011, at 7:38 AM, Matthew Mazloff wrote:
> >>>>>>>
> >>>>>>>>Hi JMC,
> >>>>>>>>
> >>>>>>>>I just updated code and am getting some odd TAF
> >>>>>>>>errors from your
> >>>>>>>>latest changes to obcs....not sure whats going on -- I'll let
> >>>>>>>>you know when I figure it out....or if you have any
> >>>>>>>>ideas please
> >>>>>>>>let me know
> >>>>>>>>
> >>>>>>>>thanks
> >>>>>>>>matt
> >>>>>>>>
> >>>>>>>>
> >>>>>>>>On Apr 16, 2011, at 8:15 PM, Jean-Michel Campin wrote:
> >>>>>>>>
> >>>>>>>>>Update of /u/gcmpack/MITgcm/pkg/obcs
> >>>>>>>>>In directory baudelaire:/srv/scratch/jmc/gcm_current/pkg/obcs
> >>>>>>>>>
> >>>>>>>>>Added Files:
> >>>>>>>>> obcs_exchanges.F
> >>>>>>>>>Removed Files:
> >>>>>>>>> obcs_prescribe_exchanges.F
> >>>>>>>>>Log Message:
> >>>>>>>>>rename obcs_prescribe_exchanges -> obcs_exchanges
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>_______________________________________________
> >>>>>>>>>MITgcm-cvs mailing list
> >>>>>>>>>MITgcm-cvs at mitgcm.org
> >>>>>>>>>http://mitgcm.org/mailman/listinfo/mitgcm-cvs
> >>>>>>>>
> >>>>>>>>
> >>>>>>>>_______________________________________________
> >>>>>>>>MITgcm-devel mailing list
> >>>>>>>>MITgcm-devel at mitgcm.org
> >>>>>>>>http://mitgcm.org/mailman/listinfo/mitgcm-devel
> >>>>>>>
> >>>>>>>
> >>>>>>>_______________________________________________
> >>>>>>>MITgcm-devel mailing list
> >>>>>>>MITgcm-devel at mitgcm.org
> >>>>>>>http://mitgcm.org/mailman/listinfo/mitgcm-devel
> >>>>>>
> >>>>>
> >>>>>>_______________________________________________
> >>>>>>MITgcm-devel mailing list
> >>>>>>MITgcm-devel at mitgcm.org
> >>>>>>http://mitgcm.org/mailman/listinfo/mitgcm-devel
> >>>>>
> >>>>>
> >>>>>_______________________________________________
> >>>>>MITgcm-devel mailing list
> >>>>>MITgcm-devel at mitgcm.org
> >>>>>http://mitgcm.org/mailman/listinfo/mitgcm-devel
> >>>>
> >>>>
> >>>>_______________________________________________
> >>>>MITgcm-devel mailing list
> >>>>MITgcm-devel at mitgcm.org
> >>>>http://mitgcm.org/mailman/listinfo/mitgcm-devel
> >>>
> >>>
> >>>_______________________________________________
> >>>MITgcm-devel mailing list
> >>>MITgcm-devel at mitgcm.org
> >>>http://mitgcm.org/mailman/listinfo/mitgcm-devel
> >>
> >>_______________________________________________
> >>MITgcm-devel mailing list
> >>MITgcm-devel at mitgcm.org
> >>http://mitgcm.org/mailman/listinfo/mitgcm-devel
> >
> >
> >_______________________________________________
> >MITgcm-devel mailing list
> >MITgcm-devel at mitgcm.org
> >http://mitgcm.org/mailman/listinfo/mitgcm-devel
>
>
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel
More information about the MITgcm-devel
mailing list