[Mitgcm-support] Re: testscript FAIL

[mitgcm-support at dev.mitgcm.org]

It's probably optimization that's breaking the results.

It would be helpful to add an IEEE compliant option to genmake
for the SGIs. Don't know what the option is but look under
case Linux*: for lines of the form
   if ($IEEE) set FFLAGS = ($FFLAGS option)
to see how to add it to genmake. The option should limit optimization
where it will change the numerical results. It shouldn't be the
default (i.e. with $IEEE unset) since it normally slows the code
down a lot.

A.


Dimitris Menemenlis wrote:
> >> Seems like lab_sea is extremely sensitive.
> >> For the identical code that I obtained 16 digits accuracy (also linux)
> >> Dimitris got as low as low as only 5 on an O2K.
> 
> Jean-Michel, just to reinforce Patrick's statement, these are the numbers I
> get for "testscript lab_sea" on an SGI Origin 2000:
> 
> Y Y Y Y  5  5  7  7  7  8 10  9  6  6  6  6  7  5  7  5  7 FAIL  lab_sea
> 
> and on Linux using g77:
> 
> Y Y Y Y 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 pass  lab_sea
> 
> for the same code (release1_p8).  I tried turning off sea-ice dynamics, but
> the high sensitivity remains.  During next couple of weeks, I plan to work on
> development of adjoint for sea-ice thermodynamics.  Hopefully I will be able
> to identify and reduce this high sensitivity.
> 
> D.
> 
> --
> Dimitris Menemenlis                      menemenlis at jpl.nasa.gov
> Jet Propulsion Lab, MS 300-323           tel: 818-354-1656
> 4800 Oak Grove Dr, Pasadena CA 91109     fax: 818-393-6720