[MITgcm-devel] problems with pickup and CD code

[Jean-Michel Campin]

Hi Samar,

I did take a look at this "pickupSuff" recently, and it seems
to me that this special case (using pickupSuff with nIter0 = 0 or
startTime=0) that Alistair added is only implemented for the main 
pickup files (which is not very logical).
I aggree that this should be either extended to all pickups or
removed, and you just provide a good argument to use it.

Cheers,
Jean-Michel

On Sun, Aug 16, 2009 at 12:26:23PM -0400, Samar Khatiwala wrote:
> Hi,
>
> I encountered a problem while attempting to restart a run from a pickup 
> file with nIter0=0 and pickupSuff='0000000000'.
> The problem is that in  S/R CD_CODE_INI_VARS the following lines:
>
>       IF ( nIter0.NE.0 ) THEN
>         CALL CD_CODE_READ_PICKUP( nIter0, myThid )
>       ENDIF
>
> prevent the pickup_cd file from being read. Is this the intended  
> behavior?
>
> I thought the point of pickupSuff was that one could rename the  
> iteration numbers of pickup files (e.g., when going from
> asynchronous to synchronous time stepping which requires changing the  
> time step) so as to be able to do a clean restart.
> However, because of the above, I get different results when starting  
> from the exact same pickup files, but with one of them
> renamed. (I could of course set nIter0 to something other than 0, but  
> that would partially defeat the purpose of the flexibility
> that pickupSuff provides.)
>
> S/R INI_FIELDS correctly treats this situation by having the following:
>
> C     Set model variables to initial/restart conditions
>       IF ( startTime .EQ. baseTime .AND.  nIter0 .EQ. 0
>      &     .AND. pickupSuff .EQ. ' ' ) THEN
>         CALL INI_VEL( myThid )
>
>          etc
>
>       ELSEIF ( .NOT.useOffLine ) THEN
>         CALL READ_PICKUP( nIter0, myThid )
>
>       ENDIF
>
> Is there a reason why S/R CD_CODE_INI_VARS does not have a similar IF  
> (... pickupSuff .NE. ' ' ...)
>
> Thanks,
>
> Samar
>
>
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel