[MITgcm-support] Uncanny output frequency after some time of simulation

Martin Losch martin.losch at awi.de
Tue Jul 5 11:30:14 EDT 2016


Hi Wlademir,

this is probably a round-off error problem in the function “different_multiple".F, see comments in eesupp/src/different_multiple.F

-recompile this routine without optimization (i.e. put into the NOOPTFILES list in your build options file and set NOOPTFLAGS=‘-O0’ or similar, see examples of this in the build options files in tools/build_options).

Martin

> On 05 Jul 2016, at 00:28, Wlademir Santis <wsantisj at gmail.com> wrote:
> 
> Hi
> 
> After some time of simulation I'm getting an output outside the frequency I set in the data file.
> 
> I'm using an experiment based on the cpl_aim+ocn, with the following parameters for the atmosphere "data" file:
>  ...
>  &PARM03
>  nIter0=0,
>  nTimeSteps=38880000,
>  deltaT=400.0,
>  abEps=0.1,
>  forcing_In_AB=.FALSE.,
>  cAdjFreq=0.,
>  pChkptFreq=2592000.,
>  taveFreq=2592000.,
>  dumpFreq=25920000.,
>  monitorFreq=2592000.,
> ...
> 
> 
> Which leads to an output frequency like:
> ...
> U.0016588800.001.001.data
> U.0016653600.001.001.data
> U.0016718400.001.001.data
> U.0016783199.001.001.data
> U.0016783200.001.001.data
> U.0016783201.001.001.data
> U.0016847999.001.001.data
> U.0016848000.001.001.data
> U.0016848001.001.001.data
> U.0016912799.001.001.data
> U.0016912800.001.001.data
> U.0016912801.001.001.data
> ...
> 
> Note that at some point it starts to produce outputs with +1 and -1 timestep around the correct one (at every 6480 timesteps). After more time it also produces +2 and -2 outputs and so forward.
> 
> Doe's anyone knows what I'm doing wrong?
> 
> I've noticed that the time averaged variables (e.g. vveltave...)  have problems after this weird output starts to happen. 
> 
> Thanks in advance!
> Wlademir
> 
> 
>  
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support




More information about the MITgcm-support mailing list