[MITgcm-devel] gfortran 4.4.5

[Jean-Michel Campin]

Hi,

I have extended the modification from exp4/input/data.ptracers
to other data.ptracers which have similar issue with 
gfortran 4.4.5,  so that all the tests
are passing again on baudelaire (but without diagnostics).

I will wait for the test run during the WE to see if
the data.ptracers modified syntax causes problem.
Regarding pgi, the switch to pgf90 seems to do the trick
(tested on aces).

Jeff, if you have time to do a short test on your cluster
(you are using pgi, right ?), such as to run the
tutorial_global_oce_biogeo, with the modified (checked-in)
data.ptracers, it would be useful.

Cheers,
Jean-Michel

On Thu, Dec 09, 2010 at 12:30:10AM -0500, Jean-Michel Campin wrote:
> Hi,
> 
> Just an update: 
> I am temporary running the gfortran testreport on baudelaire
> with useDiagnostics=F. Still have few exp that fails on
> baudelaire with gfortran, because of data.ptracers:
>     tutorial_global_oce_biogeo (FWD & AD) 
> and tutorial_dic_adjoffline (AD)
> 
> And regarding the try with modified exp4/input/data.ptracers,
> looks like (until now) that only pgf77 has problem,
> and going to see if switching to pgf90 can solve this.
> 
> Cheers,
> Jean-Michel
> 
> On Tue, Dec 07, 2010 at 06:15:01PM -0500, Jean-Michel Campin wrote:
> > Hi,
> > 
> > It looks like the recent version of gfortran (from 2010-11-12)
> >  gcc 4.4.5 (with FC.13), which is now on baudelaire 
> > is having problems with many namelist in several verification 
> > experiments (38 FWD + 3 AD experiments failed).
> > 
> > The problem is related to things like:
> >  fields(1,1) = 'THETA   ','SALT    ',
> > found in many data.diagnostics
> > or similarly, in exp4/input/data.ptracers:
> >  PTRACERS_ref(1,1) = 8*35.,
> > 
> > If I changed data.ptracers to:
> >  PTRACERS_ref(1:8,1) = 8*35.,
> > then it works with gfortran 4.4.5 but it does not work with
> > some version of pgf77.
> > 
> > As a test, I was going to change exp4/input/data.ptracers 
> > to see which compiler can read: PTRACERS_ref(1:8,1) = 8*35.,
> > 
> > But if someone has a better idea ...
> > 
> > Cheers,
> > Jean-Michel
> > 
> > _______________________________________________
> > MITgcm-devel mailing list
> > MITgcm-devel at mitgcm.org
> > http://mitgcm.org/mailman/listinfo/mitgcm-devel
> 
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel