[MITgcm-support] how to run for exactly 1 month with calendarDumps

[Ryan Abernathey]

Dimitris,

Thanks for the suggestion. What you say is reasonable in some cases,
especially with high-frequency output (as with your llc runs). But in my
case, I am doing "online" averages over relatively long (i.e. 1 month)
intervals. If the run is killed before the averaging interval finishes, I
have wasted considerable computer time. So I am trying to be very
conservative.

I think an acceptable solution is to just run each month for exactly 31
days. I will waste exactly 7 days per year if I do this, which doesn't
sound too bad.

-Ryan

On Mon, Sep 18, 2017 at 1:22 PM, Dimitris Menemenlis <
menemenlis at jpl.nasa.gov> wrote:

> Hi Ryan, hopefully someone will have better answer.
> Typically what I do is run the model as long as possible,
> until end of job or machine crashes, then use a script like
> http://wwwcvs.mitgcm.org/viewvc/MITgcm/MITgcm_contrib/high_
> res_cube/input/modpickup
> to pick-up pieces from latest available rotating pickup.
>
> Dimitris Menemenlis
>
>
> On 09/18/2017 10:04 AM, Ryan Abernathey wrote:
>
>> Hi,
>>
>> I am new to doing "realistic" simulations with the cal package activated.
>> I would like to run for one calendar month and then dump diagnostics and
>> pickup files. I will then restart from pickup and run the next month, etc.,
>> for many years.
>>
>> I am using the following data.cal
>>
>>   &CAL_NML
>>   TheCalendar='gregorian',
>>   startDate_1=20100101,
>>   startDate_2=000000,
>>   calendarDumps = .TRUE.,
>>   &
>>
>> I understand that calendarDumps causes chkPtFreq, pChkPtFreq, taveFreq,
>> etc. to automatically conform their output to one calendar month if I set
>> chkptFreq  = 259200.0.
>>
>> My question is: how to I make the model stop running after exactly one
>> month, given that the length of a month is variable?
>>
>> The two options I am familiar with for setting the length of the run
>> are nTimeSteps or endTime. I don't want to manually specify the endTime
>> after each run, and anyway, that would be hard because I would need to
>> account for the variable month length. Alternatively, I could just always
>> run for 31 days, but this would waste some computer time for 30 and 28-day
>> months. (A small amount, but the model is quite expensive already.)
>>
>> What do people commonly do in this situation?
>>
>> Thanks for the help!
>>
>> -Ryan
>>
>>
>> _______________________________________________
>> MITgcm-support mailing list
>> MITgcm-support at mitgcm.org
>> http://mailman.mitgcm.org/mailman/listinfo/mitgcm-support
>>
>> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mailman.mitgcm.org/mailman/listinfo/mitgcm-support
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.mitgcm.org/pipermail/mitgcm-support/attachments/20170918/146adbf4/attachment.html>