[MITgcm-devel] [MITgcm-cvs] MITgcm/pkg/streamice CVS Commit
Jean-Michel Campin
jmc at mit.edu
Tue Apr 12 12:00:20 EDT 2016
Dan, sorry for double posting,
but wanted to reply to mitgcm-devel
(since we do not have much documentation in tag-index ...)
Jean-Michel
--------------------------------------------
Hi Dan,
I don't know which code (and parameters) you used to generate this new
halfpipe_streamice/results/output_oadm.txt
but agreement degrade compared to the old one:
Y Y Y Y 8> 7< 6 FAIL halfpipe_streamice
and from last night:
Y Y Y Y 16>10<16 FAIL halfpipe_streamice
The new output seems to have been generated with an old code:
> (PID.TID 0000.0001) // MITgcmUV version: checkpoint65u
> (PID.TID 0000.0001) // Build user: dgoldberg
and with the wrong "data.streamice" parameter file:
197,198c197,198
< (PID.TID 0000.0001) > streamice_nonlin_tol = 1.0e-20,
< (PID.TID 0000.0001) > streamice_nonlin_tol_fp = 1.e-7,
---
> (PID.TID 0000.0001) > streamice_nonlin_tol = 1.0e-8,
> (PID.TID 0000.0001) > streamice_nonlin_tol_fp = 1.e-20,
200c200
< (PID.TID 0000.0001) > streamice_max_nl_iter = 100,
---
> (PID.TID 0000.0001) > streamice_max_nl_iter = 200,
I will update the TAF adm output soon.
Cheers,
Jean-Michel
On Tue, Apr 12, 2016 at 03:17:52PM +0100, Daniel Goldberg wrote:
> Hi J-M
>
> Apologies -- I knew this change would lead to a change in results and I
> thought I had update the results file, but apparently I had not. I just
> did, however, hopefully it will pass from now on...?
>
> I refer to halfpipe_streamice/OAD, not TAF...
>
> Dan
>
> On Tue, Apr 12, 2016 at 2:50 PM, Daniel Goldberg <dngoldberg at gmail.com>
> wrote:
>
> > Hi J-M --
> >
> > Many apologies, I must not have seen this. I will check output_oadm.txt on
> > baudelaire now, and re-generate if necessary. As I mentioned, I cannot
> > re-generate output_adm.txt but if it is easy enough for you to do (no build
> > errors, recomputation loops) then that would be good.
> >
> > Dan
> >
> > On Tue, Apr 12, 2016 at 2:23 PM, Jean-Michel Campin <jmc at mit.edu> wrote:
> >
> >> Hi Dan,
> >>
> >> I didn't see any answer yet.
> >> Are you receiving msg posted to mitgcm-devel ?
> >>
> >> Cheers,
> >> Jean-Michel
> >>
> >> On Sat, Apr 02, 2016 at 08:48:32PM -0400, Jean-Michel Campin wrote:
> >> > Hi Dan,
> >> >
> >> > Regarding the confusion, it's my fault, I quoted the g77 output instead
> >> of
> >> > the gfortran output !
> >> >
> >> > However, I did not check last time but the OpenAD output has changes
> >> too:
> >> > from March 19:
> >> > Y Y Y Y 16>10<16 FAIL halfpipe_streamice
> >> > and before your changes (e.g., March 18):
> >> > Y Y Y Y 16>16<16 pass halfpipe_streamice
> >> > Could you check this one and update output_oadm.txt if you think these
> >> > changes are expected ?
> >> >
> >> > And coming back to TAF adm, it's not clear if results did change or not
> >> > since the output_adm.txt had not been updated (after changes made at
> >> the end
> >> > of June 2015 or in July 2015) with only 10 digits agreement since that
> >> time:
> >> > Y Y Y Y 12>10< 5 FAIL halfpipe_streamice (e=0, w=8)
> >> > Should I update this output_adm.txt (agreement between fwd grad and AD
> >> grad
> >> > are not significantly different) ?
> >> >
> >> > Cheers,
> >> > Jean-Michel
> >> >
> >> > On Mon, Mar 21, 2016 at 07:47:46AM +0000, Daniel Goldberg wrote:
> >> > > Hi J-M --
> >> > >
> >> > > Sorry I misunderstood, I thought new changes had caused the
> >> > > halfpipe_streamice taf expt to not compile any more. If the inaccuracy
> >> > > reflects earlier changes this makes sense. That is fine then if you
> >> are
> >> > > able to compile with TAF now; I do not have any license I can use.
> >> (At some
> >> > > point I might make a change that will lead to recomputation loops and
> >> > > too-long run times and will not be able to debug -- this is why i
> >> suggested
> >> > > not to run it anymore...)
> >> > >
> >> > > On Sun, Mar 20, 2016 at 7:00 AM, Jean-Michel Campin <jmc at mit.edu>
> >> wrote:
> >> > >
> >> > > > Hi Dan,
> >> > > >
> >> > > > I've just checked and the time it take to run the AD-TAF experiment
> >> has
> >> > > > not increased significantly. So we ccould continue to run it
> >> everyday.
> >> > > > But it would be good to update the output_adm.txt if we think that
> >> > > > it reflect past changes in the code (I don't remember exactly when
> >> > > > the results did change and there was some back and forth uodate at
> >> the
> >> > > > time).
> >> > > >
> >> > > > Cheers,
> >> > > > Jean-Michel
> >> > > >
> >> > > > On Sat, Mar 19, 2016 at 07:06:12PM +0000, Daniel Goldberg wrote:
> >> > > > > hi j-m
> >> > > > >
> >> > > > > i was afraid of this, not having any access to taf anymore makes
> >> it very
> >> > > > > difficult to maintain the output_results or check whether a change
> >> > > > "breaks"
> >> > > > > the experiment. i was confident that it would not because my
> >> changes do
> >> > > > not
> >> > > > > break OpenAD -- but I don't know what the issue is with TAF. do
> >> we want
> >> > > > to
> >> > > > > maybe just stop running the TAF experiment?
> >> > > > >
> >> > > > > I'll look into the other issue with jim.
> >> > > > >
> >> > > > > thanks
> >> > > > > dan
> >> > > > >
> >> > > > > On Sat, Mar 19, 2016 at 6:20 PM, Jean-Michel Campin <jmc at mit.edu>
> >> wrote:
> >> > > > >
> >> > > > > > Hi Dan and Jim,
> >> > > > > >
> >> > > > > > It seems that these changes have increased the number of TAF
> >> > > > recomputation
> >> > > > > > warning:
> >> > > > > > previously:
> >> > > > > > > Y Y N N .. .. .. N/O halfpipe_streamice (e=0, w=8)
> >> > > > > > and now:
> >> > > > > > < Y Y N N .. .. .. N/O halfpipe_streamice (e=0, w=10)
> >> > > > > > Note that this experiment has been failing (only 10 matching
> >> digits)
> >> > > > > > for a long time (I remember sending an email at that time).
> >> > > > > >
> >> > > > > > Also, experiment shelfice_remeshing is now broken (not compile
> >> > > > anymore).
> >> > > > > >
> >> > > > > > Cheers,
> >> > > > > > Jean-Michel
> >> > > > > >
> >> > > > > > On Fri, Mar 18, 2016 at 03:18:29PM -0400, Daniel N. Goldberg,
> >> 54-1423,
> >> > > > > > 6172532977, wrote:
> >> > > > > > > Update of /u/gcmpack/MITgcm/pkg/streamice
> >> > > > > > > In directory forge:/tmp/cvs-serv5616
> >> > > > > > >
> >> > > > > > > Modified Files:
> >> > > > > > > STREAMICE.h streamice_get_fp_err_oad.F
> >> > > > > > > streamice_get_vel_fp_err.F streamice_get_vel_resid_err.F
> >> > > > > > > streamice_get_vel_resid_err_oad.F streamice_readparms.F
> >> > > > > > > Log Message:
> >> > > > > > > allow for differing error norms in nonlinear iteration
> >> > > > > > >
> >> > > > > > >
> >> > > > > > > _______________________________________________
> >> > > > > > > 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
> >> > > > > >
> >> > > > >
> >> > > > >
> >> > > > >
> >> > > > > --
> >> > > > >
> >> > > > > 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
> >> > > >
> >> > > > > _______________________________________________
> >> > > > > 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
> >> > > >
> >> > >
> >> > >
> >> > >
> >> > > --
> >> > >
> >> > > 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
> >> >
> >> > > _______________________________________________
> >> > > 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
> >>
> >
> >
> >
> > --
> >
> > 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
More information about the MITgcm-devel
mailing list