[Mitgcm-support] lab-sea high sensitivity

mitgcm-support at dev.mitgcm.org mitgcm-support at dev.mitgcm.org
Wed Jul 9 15:29:38 EDT 2003


Hi Dimitris,

Thanks for you answer.

About the "high sensitivity" of lab-sea exp,
I can see 2 different problems:

a) on Linux with g77, I still don't know why my PC gives
different "cg2d_ini_res" than the PC used by Patrick (twain: Linux, g77).
The differences remain if I reduce the optimisation (-O0)
but vanishe when I turn off sea-ice dynamics.

b) I have seen few lines in dynsolver.F where 
constants are use with no "_d 0", for instance :
          DAIRN(I,J,bi,bj)=RHOAIR*SEAICE_drag*
     &                     (2.70+0.142*AAA+0.0764*AAA*AAA)  
If I change it to :
          DAIRN(I,J,bi,bj)=RHOAIR*SEAICE_drag*
     &      (2.7 _d 0 +0.142 _d 0*AAA+0.0764 _d 0*AAA*AAA)
the testcript results change significantly.

Y Y Y Y  4  6 10 12 11 13 13 12 10  5  8  4  5  8  7  4  4 FAIL  lab_sea

I suspect that on 2 different machines (e.g: Linux & SGI),
the double precision part of single precision constant might 
change from one machine to the other (like when adding _d 0).

I checked also that (b) is not causing (a).

Cheers,

Jean-Michel



More information about the MITgcm-support mailing list