[MITgcm-support] Deviated results after restart

[Martin Losch]

Hi Weiyang,

this can happen, if your machine/compiler combination uses higher precision for internal storage that the nominal 64bit in the restart file, which can lead to different truncation errors (although the differences are very large in your plots).

Can you try to tweak the options of your compiler, so that I uses exact IEEE arithmetics and repeat the test. I would expect that then the problem goes away, or at least get’s much smaller.

Martin


> On 12. Oct 2021, at 00:06, Weiyang Bao <wbao at udel.edu> wrote:
> 
> Dear all,
> 
> I’m running a simulation for a freshwater-discharged coastal channel. The model is run with multi processors and it could restart from the pickup files and end normally without reporting any error. However, the output results (T, S, U) become different from those at the same timestep in a non-restart run, after initial conditions being loaded correctly from the pickup. That is to say, the 1st-iteration output in the restart is identical to the same-realtime output in a non-restart run but consequent results begin to differ. A plot showing the divergence at one grid point is attached here:
> 
> 
> <restart_deviation.png>
> 
> The data file I used is as follow:
> 
>  &PARM01
>  tRef=90*0.,
>  sRef=90*34.,
>  viscAz=0.00001,
>  viscC2smag=2.2,
>  diffKzT=0.00001,
>  diffKhT=20,
>  diffKzS=0.00001,
>  diffKhS=20,
>  bottomDragQuadratic=2.5E-3,
>  implicitViscosity=.TRUE.,
>  implicitDiffusion=.TRUE.,
>  no_slip_sides=.FALSE.,
>  no_slip_bottom=.TRUE.,
>  rigidLid=.FALSE.,
>  implicitFreeSurface=.TRUE.,
>  hFacMin=0.05,
>  readBinaryPrec=64,
>  staggerTimeStep=.TRUE.,
>  tempAdvScheme=33,
>  saltAdvScheme=33,
>  eosType='JMD95Z',
>  f0=1.2E-4,
>  beta=0.E-11,
>  selectAddFluid=1.,
>  useSingleCpuIO=.TRUE.,
>  &
> 
> # Elliptic solver parameters
>  &PARM02
>  cg2dMaxIters=300,
>  cg2dTargetResidual=1.E-13,
>  cg3dMaxIters=20,
>  cg3dTargetResidual=1.E-8,
>  &
> 
> # Time stepping parameters
>  &PARM03
>  nIter0=0,
>  endTime=20000.,
>  deltaT=2.,
>  abEps=0.01,
>  pChkptFreq=1000.,
>  chkptFreq=0.,
>  pickupSuff='',
>  dumpFreq=1000.,
>  monitorSelect=1,
>  monitorFreq=1000.,
>  periodicExternalForcing=.FALSE.,
>  &
> 
> # Gridding parameters
>  &PARM04
>  usingCartesianGrid=.TRUE.,
>  delXfile='delx.bin',
>  delYfile='dely.bin',
>  delZ= 20*2,30*4,40*6.,
>  &
> 
> # Input datasets
>  &PARM05
>  bathyFile='bathymetry.bin',
>  hydrogThetaFile='tempini.bin',
>  hydrogSaltFile ='saltini.bin',
>  &
> 
> I only modified nIter0 every time before restart, is there anything else needs attention? Thanks a lot for your time!
> 
> Best, Weiyang
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mailman.mitgcm.org/mailman/listinfo/mitgcm-support

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 1665 bytes
Desc: not available
URL: <http://mailman.mitgcm.org/pipermail/mitgcm-support/attachments/20211019/0f64b69b/attachment.p7s>