[MITgcm-devel] linux_ia64_ifort+mpi_altix_gfdl

[Alistair Adcroft]

GFDL uses -fltconsistency for FP reproducability.

Patrick Heimbach wrote:
> On Wed, 2006-12-06 at 08:27 -0800, Dimitris Menemenlis wrote:
>   
>> Patrick, let me know what you find.
>>
>> On the _nas option file we have replaced -mp with -fp-model precise, which I 
>> think is a slightly less stringent requirement for floating point arithmetic 
>> accuracy.  The other difference is the -ip flag.
>>     
>
> Yep, I'm aware.
>
>   
>> More than likely your code fails at GFDL because of wrong modules/paths for the 
>> libraries, not because of above options.
>>     
>
> Same optfile/module/paths works for older code...
>
> -p.
>
>   
>> D.
>> _______________________________________________
>> MITgcm-devel mailing list
>> MITgcm-devel at mitgcm.org
>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>>