[MITgcm-devel] order in initialised_fixed

[Martin Losch]

I haven't tried yet (other problems that I want to fix first), but  
write_grid was the place that I was thinking about, too.

More generally, the grid-information (2D fields) could be part of all  
netcdf files including the files that are written by the diagnostics  
package. I have heard complaints about the grid information being in a  
different netcdf-file. Where would one do that, ie. call a "call  
mnc_write_coord_vars" ?

M.
On Oct 8, 2009, at 2:50 PM, Jean-Michel Campin wrote:

> Hi Martin,
>
> I kind of agree with you. And what about calling this new S/R
> from S/R write_grid.F ? Is it too late ? I have not checked
> but may be you know.
> Cheers,
> Jean-Michel
>
> On Thu, Oct 08, 2009 at 09:27:08AM +0200, Martin Losch wrote:
>> Hi Jean-Michel,
>>
>> the simplest fix would probably just to call autodiff_ini_model_io  
>> just
>> (or sometimes) after ini_grid ini initialised_fixed. I'll try that.
>>
>> On the other hand the name of this routine is misleading, since it's
>> deals only with mnc-related stuff. I might want to invent another
>> routine, like mnc_write_coord_vars.F that does nothing but this:
>>      CALL MNC_CW_SET_UDIM('adstate', 0, myThid)
>>      CALL MNC_CW_RS_W('R','adstate',0,0,'XC',xC, myThid)
>>      CALL MNC_CW_RS_W('R','adstate',0,0,'YC',yC, myThid)
>>      CALL MNC_CW_RS_W('R','adstate',0,0,'XU',xG, myThid)
>>      CALL MNC_CW_RS_W('R','adstate',0,0,'YU',yC, myThid)
>>      CALL MNC_CW_RS_W('R','adstate',0,0,'XV',xC, myThid)
>>      CALL MNC_CW_RS_W('R','adstate',0,0,'YV',yG, myThid)
>>      CALL MNC_CW_RS_W('R','adstate',0,0,'XG',xG, myThid)
>>      CALL MNC_CW_RS_W('R','adstate',0,0,'YG',yG, myThid)
>>      CALL MNC_CW_RS_W('R','adstate',0,0,'RC',rC, myThid)
>>      CALL MNC_CW_RS_W('R','adstate',0,0,'RF',rF, myThid)
>>
>> and call it somewhere (for all relevant files) after ini_grid. Does  
>> that
>> make sense?
>>
>> M.
>>
>>
>>
>> On Oct 6, 2009, at 5:48 PM, Jean-Michel Campin wrote:
>>
>>> Hi Martin,
>>>
>>> I think INI_MODEL_IO needs to be called first, before ini_grid.
>>> The reason is that some of the IO stuff need to be set before
>>> reading grid arrays (mdsioLocalDir, buffer initialisation)
>>> and monitor initialisation needs to happen before ini_grid.
>>> So, if there is a problem with the content of AUTODIFF_INI_MODEL_IO
>>> and/or where it is called, this S/R or the call should be fixed.
>>> Can it be splitted in 2 ? or called after ini_grid ?
>>>
>>> Cheers,
>>> Jean-Michel
>>>
>>> On Tue, Oct 06, 2009 at 11:14:03AM +0200, Martin Losch wrote:
>>>> Hi,
>>>>
>>>> is there any reason why
>>>> INI_MODEL_IO( myThid )
>>>> cannot be called after
>>>> INI_GRID( myThid )
>>>> in initialised fixed? It's just a cosmetic issue, but the grid
>>>> parameters (e.g. XC, RC, etc) are already filled in the netcdf file
>>>> for
>>>> the adjoint variables (ad*.nc) in ini_model_io, but they are all  
>>>> zero
>>>> because they are computed in ini_grid. Can I change the order?
>>>>
>>>> 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
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel