[MITgcm-devel] SEAICE_OLD_AND_BAD_DISCRETIZATION

[Martin Losch]

Hi there,

a year has passed, it's time for yet another push towards removing SEAICE_OLD_AND_BAD_DISCRETIZATION from the seaice package. None of the verification experiments is using it anymore, it's wrong, and potentially unstable. I suggest to remove the code and add a check that stops the code, if someone is still using this code, and then see what happens.

I have run this code with cs32-grids with long intergrations (order 350 years), and others have done even longer runs (also with ice around the corners), without ever noticing anything funny in the results.

Any objections?

Martin

On Oct 26, 2009, at 8:53 AM, Martin Losch wrote:

> OK, fine with me, I have enough other things on my agenda. But it would be good not to have this code around for longer than necessary.
> 
> Martin
> 
> On Oct 23, 2009, at 6:19 PM, Jean-Michel Campin wrote:
> 
>> Hi Martin,
>> 
>> Can we wait until Thanksgiving ?
>> I have not yet done long integration with "new code"
>> for test exp. global_ocean.cs32x15.icedyn
>> (and would be like to test what happens when seaice
>> extend further than face 3 & 6 and reach the corners)
>> Thanks,
>> Jean-Michel
>> 
>> On Fri, Oct 23, 2009 at 09:17:01AM +0200, Martin Losch wrote:
>>> Hi seaice modellers,
>>> 
>>> while I am struggling to convince TAF,  that it should write
>>> vectorizable code, I stumbled over this old flag:
>>> SEAICE_OLD_AND_BAD_DISCRETIZATION
>>> Defining this CPP-flag turns on a discretization that is not yet that
>>> old (6months or so), but really bad. We kept this in the code to make
>>> sure that we can recover old results (in particular for CS510 and ECCO
>>> GODAE). I personally never want to use this code again, so I am much in
>>> favor of removing this code (because it sometimes means that I have to
>>> code "innovations" (in German: Verschlimmbesserungen) twice). Is there
>>> anyone you still needs this code? Can we retire this flag? (It means
>>> changing the seaice part of global_ocean.cs32x15)
>>> 
>>> 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