[MITgcm-devel] lab_sea.hb87 restart problem

[Martin Losch]

All I tested was the hb87 experiment, but I started to modify the  
data files, so that I turned of hb87 and also turned of evp and then  
even turned off seaice (in data.pkg), and only then the pickup  
worked, but I will repeat the tests asap on a linux box (rather than  
on my apple laptop on the train).

Martin

On 9 Nov 2007, at 16:30, Jean-Michel Campin wrote:

> Hi Martin,
>
> I recently updated the tst1+1 script so that it does not use anymore
> this "add0upto10c" stupid script.
> Might not be more portable (I only use it on Linux box), but
> it might be easier to check what the script is doing.
> Could you run it on a linux box 1rst ?
>
> My impression (from your e-mail) is that the script is not working,
> does not move the pickup correctly, and the results is that
> it just repeat 3 times the same run (4 it long).
>
> Regarding the seaice pickup,
> I tested recently (few days ago) and both the standard
> lab_sea (with evp) and labsea.lsr were passing the test 2+2=4
> with zero diff in the pickup files. I also checked with the
> seaice B-grid dynamics and seems to work also.
> When I do the same test with hb87, get large differences.
> I will repeat the test and copy to you the cg2d output
> later today.
>
> Cheers,
> Jean-Michel
>
> On Fri, Nov 09, 2007 at 03:47:51PM +0100, Martin Losch wrote:
>> Jean-Michel,
>> I seem to have problems with you script an cannot reproduce your
>> observation that the pickup does not work: This is what I get after
>> running "tst1+1 2" (error messages are the same as for  "tst1+1 1",
>> but there are no differences in the cg2d output, probably a
>> consequence of the script returning errors?
>>> csysm3::tr_run.hb87> ./tst1+1 2
>>> expr: syntax error
>>> use script: P10c='../../../../MITgcm_contrib/jmc_script/add0upto10c'
>>>             RNPs='../../../../MITgcm_contrib/jmc_script/rnp -s'
>>>        and  DIFd='../../../../MITgcm_contrib/jmc_script/dif_data'
>>> expr: syntax error
>>> expr: syntax error
>>> expr: syntax error
>>> ../../../../MITgcm_contrib/jmc_script/add0upto10c: line 9: test: =:
>>> unary operat
>>> or expected
>>> ../../../../MITgcm_contrib/jmc_script/add0upto10c: line 6: test: !
>>> =: unary opera
>>> tor expected
>>> ../../../../MITgcm_contrib/jmc_script/add0upto10c: line 9: test: =:
>>> unary operat
>>> or expected
>>> ../../../../MITgcm_contrib/jmc_script/add0upto10c: line 6: test: !
>>> =: unary opera
>>> tor expected
>>> ../../../../MITgcm_contrib/jmc_script/add0upto10c: line 9: test: =:
>>> unary operat
>>> or expected
>>> ../../../../MITgcm_contrib/jmc_script/add0upto10c: line 6: test: !
>>> =: unary opera
>>> tor expected
>>>
>>>
>>> gcmExc=mitgcmuv
>>> -- compare cg2d_init_res :
>>> run 1iA:
>>>    6.08551932893086E-02
>>>    6.55586521348751E-02
>>>    7.01960854205512E-02
>>>    7.29481919712837E-02
>>> run 1iB:
>>>    6.08551932893086E-02
>>>    6.55586521348751E-02
>>>    7.01960854205512E-02
>>>    7.29481919712837E-02
>>> run 2it:
>>>    6.08551932893086E-02
>>>    6.55586521348751E-02
>>>    7.01960854205512E-02
>>>    7.29481919712837E-02
>>
>> With my own inconvenient and error-prone scheme I can repoduce that
>> the pickup does not work properly. However, it's not related to hb87.
>> It does not work, when useHB87stressCoupling = .false., either.
>> Rather the problem seems related to the seaice pickup in general
>> (both evp and lsr). When did that happen?
>>
>> I don't have a fast internet connection right now, so I cannot test
>> this with older checkpoints. The last change to the pickups is
>> related to HSALT.
>>
>> Martin
>>
>>
>> On 4 Nov 2007, at 21:19, Jean-Michel Campin wrote:
>>
>>> Hi Martin,
>>>
>>> On Sun, Nov 04, 2007 at 09:14:48AM +0100, Martin Losch wrote:
>>>> Hi Jean-Michel,
>>>>
>>>> no, I did not check that, because the Hibler+Bryan stress  
>>>> formulation
>>>> should not require any restarts ... I can't think of a reason  
>>>> why it
>>>> should:
>>>> for useHB87stressCoupling=.true., the stress on the ocean is  
>>>> computed
>>>> differently at the end of seaice_dynamics (in the case of the
>>>> verification experiment it's based on the internal stress  
>>>> divergence
>>>> computed earlier in seaice_evp.F). So if the restart does not work,
>>>> it should be related to the evp solver rather than the hb87
>>>> stress ...
>>>>
>>>> Strange. Where do I find you restart test script, so I can start
>>>> playing with this, too.
>>>> Martin
>>>>
>>>
>>> I updated the little script that I use:
>>> MITgcm_contrib/jmc_script/tst1+1
>>>
>>> and the sequence of command I typed to check the restart was:
>>> % cd verification
>>> % ./testreport -dd -t lab_sea
>>> % cd lab_sea/tr_run.hb87
>>> % mkdir temp
>>> % mv pickup* temp
>>> % cp data data.tst
>>> % rm data
>>>
>>> edit data.tst:
>>> (useSingleCpuIO is commented because of limitation of the tst1+1
>>> script)
>>> % diff data data.tst
>>> 53c53
>>> <  useSingleCpuIO=.TRUE.,
>>> ---
>>>> #useSingleCpuIO=.TRUE.,
>>> 86a87,88
>>>> nIter0=10,
>>>> nTimeSteps=4,
>>> 88,89c90,91
>>> <  startTime=0.0,
>>> <  endTime=36000.,
>>> ---
>>>> #startTime=0.0,
>>>> #endTime=36000.,
>>>
>>> % cp ../../../../MITgcm_contrib/jmc_script/tst1+1 .
>>> % ./tst1+1 1
>>> % ./tst1+1 2
>>> And then you should see some differences in cg2d_ini output.
>>> All the results are put in separated dir, so you can make some
>>> plots and see where the differences are located.
>>>
>>> the last step is to make a diff of pickup files:
>>> % ./tst1+1 3
>>>
>>> Jean-Michel
>>> _______________________________________________
>>> 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