[MITgcm-devel] [MITgcm-cvs] MITgcm/pkg/ctrl CVS Commit
Jean-Michel Campin
jmc at mit.edu
Wed Apr 5 11:51:03 EDT 2017
Hi Ou,
Just a summary of where we are now:
1) the addition of cost_gencost_glbmean.F in pkg/ecco was a little bit late
to be part of most of last night testreport run, but early enough for
Gael's test (better today).
So I am still unsure regarding experiments that use pkg/ecco
(adm test: 1D_ocean_ice_column, lab_sea, obcs_ctrl), will wait until tomorrow.
2) Regarding other ADM test, something is not right with experiment
global_ocean.90x40x15 (does not use pkg/ecco) which compiles but does not run:
Last night, on reference machine (baudelaire) with gfortran -devel,
( http://mitgcm.org/testing/results/2017_04/tr_baudelaire-b_20170405_0/summary.txt )
17,20c17,20
< Y Y Y N .. .. .. N/O global_ocean.90x40x15 (e=0, w=17)
< Y Y Y N .. .. .. N/O global_ocean.90x40x15.bottomdrag
< Y Y Y N .. .. .. N/O global_ocean.90x40x15.kapgm
< Y Y Y N .. .. .. N/O global_ocean.90x40x15.kapredi
---
Compare to last Monday (April 03):
> Y Y Y Y 16>16<16 pass global_ocean.90x40x15 (e=0, w=15)
> Y Y Y Y 16>16<16 pass global_ocean.90x40x15.bottomdrag
> Y Y Y Y 16>16<16 pass global_ocean.90x40x15.kapgm
> Y Y Y Y 16>16<16 pass global_ocean.90x40x15.kapredi
Note also that there is 2 more recomputation warnings from TAF.
I don't think this "fail" is specific to this platform/compiler version
so it should be easy to reproduce on most (if not any) linux box
with gfortran.
3) I will switch back on pkg/profile in global_with_exf (FWD) experiment
since I turned it off as pkg/profile was not compiling.
Cheers,
Jean-Michel
On Wed, Apr 05, 2017 at 04:59:03AM +0000, Wang, Ou (329B) wrote:
> Hi Jean-Michel,
>
> Thanks for pointing that out. I have checked in cost_gencost_glbmean.F.
>
> Cheers,
> Ou
>
> > On Apr 4, 2017, at 9:33 PM, Jean-Michel Campin <jmc at mit.edu> wrote:
> >
> > Hi Ou,
> >
> > I think there is (at least) one more problem left:
> > cost_gencost_glbmean.f
> > had been added to: pkg/ecco/ecco_ad_diff.list
> > but the file "cost_gencost_glbmean.F" does not exist (as far as I can see).
> >
> > Cheers,
> > Jean-Michel
> >
> >> On Wed, Apr 05, 2017 at 03:20:38AM +0000, Wang, Ou (329B) wrote:
> >> Thanks, Jean-Michel. I will check in my updated code.
> >>
> >> Ou
> >>
> >>
> >>> On Apr 4, 2017, at 8:14 PM, Jean-Michel Campin <jmc at mit.edu> wrote:
> >>>
> >>> Hi Ou,
> >>>
> >>> The ADM testreport output (tr_pfe21_20170404_17) looks better than the one
> >>> produced last night (available on testing page: http://mitgcm.org/public/testing.html)
> >>> So I would like to encourage you to check-in these modifications/fixes.
> >>>
> >>> Not sure why there are few "fail" but since I don't have a reference output
> >>> on the same machine from before, it's hard to compare.
> >>>
> >>> Cheers,
> >>> Jean-Michel
> >>>
> >>>> On Wed, Apr 05, 2017 at 02:04:49AM +0000, Wang, Ou (329B) wrote:
> >>>> Hi Jean-Michel and Gael,
> >>>>
> >>>> I have run both forward and adjoint testreports after correcting a couple of bugs. Attached please find the the reports. All but a few of the experiments passed the tests. Are those failed experiments a concern or not? I ran the tests on Pleiades.
> >>>>
> >>>> I have also run the adjoint testreport for Gael's experiments of global_oce_cs32 and global_oce_llc90. They both passed.
> >>>>
> >>>> If the report looks fine, I will check in my updated code.
> >>>>
> >>>> Thanks,
> >>>> Ou
> >>>> ________________________________________
> >>>> From: Jean-Michel Campin [jmc at mit.edu]
> >>>> Sent: Monday, April 03, 2017 10:15 PM
> >>>> To: mitgcm-devel at mitgcm.org
> >>>> Cc: Wang, Ou (329B)
> >>>> Subject: Re: [MITgcm-devel] [MITgcm-cvs] MITgcm/pkg/ctrl CVS Commit
> >>>>
> >>>> Hi Ou,
> >>>>
> >>>> Seems that even with forward testreport, one experiment (global_with_exf)
> >>>> fails to compile, because of changes in pkg/profiles
> >>>> For now, I've just turned this pkg off in this experiment.
> >>>>
> >>>> Regarding the TAF adjoint testreport, 4 experiments are no longer compiling:
> >>>> 1D_ocean_ice_column
> >>>> global_ocean.90x40x15
> >>>> global_ocean_ebm
> >>>> lab_sea
> >>>> And the interesting thing is that it's not due to a single error but
> >>>> the first one that shows up in the log file is different from one
> >>>> experiment to the other.
> >>>>
> >>>> We always advise to run testreport (at least a forward one) before any check-in;
> >>>> But if you have access to TAF and specially when any of these 4 pkgs:
> >>>> autodiff/ctrl/cost/ecco is involved, a second "-adm" one seems necessary,
> >>>> at least to me.
> >>>>
> >>>> So, at this point, I don't know what to suggest.
> >>>> But let's be clear, witching off all the "-adm" daily testreport (easy)
> >>>> would save me significant amount of time.
> >>>>
> >>>> Cheers,
> >>>> Jean-Michel
> >>>>
> >>>>> On Mon, Apr 03, 2017 at 09:03:34PM -0400, Jean-Michel Campin wrote:
> >>>>> Hi Ou,
> >>>>>
> >>>>> Are you done with today check-in ? I am asking this because I tried to
> >>>>> run a tesrreport -adm and does not seem to work very well (but I
> >>>>> was also testing some of my changes).
> >>>>>
> >>>>> But in case you plan to to check-in more changes, just a comment
> >>>>> regarding the time in the day to check-in changes to the code
> >>>>> (since you are on Pacific time), it's generally better to avoid the
> >>>>> midnight - 1.am (EST) (or even better, 2.am) window since most of
> >>>>> the automatic testreport we run here are starting around this time.
> >>>>>
> >>>>> Cheers,
> >>>>> Jean-Michel
> >>>>>
> >>>>>> On Mon, Apr 03, 2017 at 07:16:38PM -0400, Wang, Ou wrote:
> >>>>>> Update of /u/gcmpack/MITgcm/pkg/ctrl
> >>>>>> In directory forge:/tmp/cvs-serv18128/pkg/ctrl
> >>>>>>
> >>>>>> Modified Files:
> >>>>>> ctrl_get_gen.F ctrl_map_ini_genarr.F
> >>>>>> Log Message:
> >>>>>> Add and update some files related to ECCO v4 release 3. The main changes are
> >>>>>> adding cost_gencost_glbmean.F to compute the costs for global mean time-series of OBP
> >>>>>> & SSH, splitting the profile cost to sample-mean and -anomaly parts,
> >>>>>> updating the cost formula for the proxy sea-ice costs, adding initial velocity
> >>>>>> as part of the control variables, and applying the correction of global
> >>>>>> mean steric height change to OBP.
> >>>>>> - cost_gencost_glbmean.F: compute cost for global mean OBP & SSH time-series
> >>>>>> - cost_profiles.F: add a CPP option (ALLOW_PROFILES_SAMPLESPLIT_COST)
> >>>>>> that if defined splits the profile cost to
> >>>>>> sample-mean and sample-anomaly costs.
> >>>>>> - cost_gencost_seaicev4.F: add a CPP option (SEAICECOST_JPL) that if defined
> >>>>>> reads in the weights for the proxy sea-ice costs.
> >>>>>> - cost_gencost_curstomize.F: if ALLOW_PSBAR_STERIC is defined, correct OBP with
> >>>>>> the global mean steric ssh change to be consistent
> >>>>>> with SSH.
> >>>>>> - ecco_phys.F: include sterGloH in a common block to be used in
> >>>>>> cost_gencost_customize.F to correct OBP when
> >>>>>> ALLOW_PSBAR_STERIC is defined.
> >>>>>> - ctrl_map_ini_genarr.F: add CPP options to include initial UV as the control
> >>>>>> variables.
> >>>>>> - cost_gencost_all.F: changes related to cost for the global mean time-series.
> >>>>>> - cost_gencost_bpv4.F: minor change related to cost for the global mean time-series.
> >>>>>> - ecco_readparms.F: changes related to cost for the global mean time-series.
> >>>>>> - ecco_ad_diff.list: include the new added cost_gencost_glbmean.F.
> >>>>>> - ctrl_get_gen.F: check if the first seven characters of xx_tauu_file &
> >>>>>> xx_tauv_filei matches "xx_tauu" or "xx_tauv".
> >>>>>> - ecco_cost_final.F: add modification related to the sample-mean profile cost.
> >>>>>> - ecco_cost_init_fixed.F: initialize the data and weights for the global mean costs
> >>>>>> for OBP & SSH, if ALLOW_GENCOST_1D is defined.
> >>>>>> - profiles_readvector.F: read in climatology field in the profile file, and mask
> >>>>>> out the weight where climatology data is flagged. Active
> >>>>>> if ALLOW_PROFILES_CLIMMASK is defined.
> >>>>>> - PROFILES_SIZE.h: changes related to the split of profile cost to sample-mean
> >>>>>> and sample-anomaly.
> >>>>>> - profiles.h: changes related to the split of profile cost to sample-mean
> >>>>>> and sample-anomaly.
> >>>>>> - ecco.h: changes related to the global mean costs
> >>>>>> - profiles_readparms.F: changes related to the profile cost splitting and masking
> >>>>>> by climatology data.
> >>>>>> - profiles_init_fixed.F: changes related to the profile cost splitting.
> >>>>>> - profiles_init_varia.F: changes related to the profile cost splitting.
> >>>>>>
> >>>>>>
> >>>>>>
> >>>>>>
> >>>>>> _______________________________________________
> >>>>>> 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
More information about the MITgcm-devel
mailing list