[MITgcm-support] Simulation precision

Martin Losch Martin.Losch at awi.de
Sun Nov 2 05:38:19 EST 2008


David,
all "prognostic variables" such as (THETA, SALT, U/VVEL, ETAN) are  
always declared as REAL*8. Additionally, by default ALL reals are  
REAL*8.
Last week we fixed a broken flag in CPP_EEOPTIONS.h, that allows you  
to define some variables (mostly related to the grid, e.g. hFacC,  
etc.) as REAL*4. put #undef REAL4_IS_SLOW in CPP_EEOPTIONS.h to turn  
that on. However, I would not recommend that, unless you run into  
memory limitation (because it is not well tested and the memory  
effect is not so large).

For some opt files we still have the -r8 flag (or equivalent), but I  
believe that that is not really necessary, although I may be wrong  
about that.

Martin

On 2 Nov 2008, at 01:50, David Hebert wrote:

> Hi all,
>
> Does MITgcm simulation always use REAL*8 precision for the  
> timestepping, or does that need to be specified explicitly via -r8  
> compiler flag?
>
>
> Thanks,
>
> David
>
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support




More information about the MITgcm-support mailing list