[MITgcm-devel] 3 time levels in seaice

[Jean-Michel Campin]

Martin,

OK, I will check-in also seaice_output.F then (right now, everything is
broken).

And for the 3-levels -> 1-level changes, we should hear from Patrick
first (from what I know, it's unlikely that it will have zero impact
on the adjoint; but I agree it should not be delayed too much).

Cheers,
Jean-Michel

On Wed, Jun 17, 2009 at 04:08:41PM +0200, Martin Losch wrote:
> Jean-Michel,
>
> I don't know what got into me, but this morning I wanted to do something 
> useful but not immediately important, so I started doing all of these 
> things. Go ahead and check in your stuff. I will wait and merge your new 
> code with mine (shouldn't be hard, since I did not touch the I/O yet).
>
> Then, if you agree, I'll make a tag tomorrow or even later, check-in my 
> crap, then hope that Patrick can fix the adjoint if I really break it, 
> and then we can worry about the pickups? (I guess it's enough to switch 
> from rec_lev to rec_3d routines in reading and writing pickup files).
>
> Martin
>
> On Jun 17, 2009, at 3:53 PM, Jean-Michel Campin wrote:
>
>> Hi Martin,
>>
>> I though we were going to keep those 3 levels field for some time
>> so I changed the seaice_do_diags.F (and renamed it seaice_output.F)
>> and am about to check-in this stuff (+ few other renaming in other  
>> pkg).
>> Would it make sense to wait a little bit for your seaice check-in
>> until we are sure of what the adjoint and Patrick pathological
>> set-up (seems silent for now) are going ?
>>
>> Apart from that, I think we only store level1 in pickup-files.
>> + pickup_seaice has been switch to "clever pickup" so that even
>> if you need to change the number of fields/levels, no need for an  
>> other
>> flag.
>>
>> Jean-Michel
>>
>> On Wed, Jun 17, 2009 at 02:42:51PM +0200, Martin Losch wrote:
>>> Jean-Michel, Patrick
>>>
>>> I have started to get rid of the 3 time levels in pkg/seaice.
>>> So far I am still using the old 3d fields in the code, but only  
>>> level1.
>>> This works fine; the next step will be to replace uice(:,:,3,:,:) etc
>>> with uice(:,:,:,:), but that will require changing the pickup files. 
>>> I am
>>> not sure how to do that in order to be backward compatible. Is there 
>>> a
>>> way to tell from the meta files? Should I introduce a flag
>>> useSeaicePickupPreCkpt?? and write code for these different cases?
>>>
>>> Further, the biggest problem was dealing with the old advect.F and
>>> diffus.F routines. The most important application for these is the
>>> adjoint (because the flux-limited schemes are not yet  
>>> differentiated). I
>>> had to change them quite a bit (among that removing the leap-frog  
>>> code
>>> that hopefully nobody uses), so i am afraid that I might have broken 
>>> the
>>> adjoint.
>>>
>>> Should I make a tag, check in the stuff that I have now, and let  
>>> Patrick
>>> fix the adjoint before we worry about the pickup files?
>>>
>>> Martin
>>>
>>>
>>> _______________________________________________
>>> 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