[MITgcm-support] ifort/g77 issue

Thomas Haine Thomas.Haine at jhu.edu
Wed Jan 23 13:54:49 EST 2008


Hi Guys,

On Tue, 2008-01-22 at 11:40 -0500, Jean-Michel Campin wrote: 
> Hi Thomas,
> 
> Did you already try to lower the optimization to the lowest level -O0 ?
> We have several examples of compiler version where 
> 1 subroutine (and only few set-up which use it in a particular
> way) does't work properly with relatively standard optimization
> level.
> But otherwise, I don't have any other suggestion right now.
> 
> Jean-Michel

Advection is set to 2. There are no packages selected in data.pkg. When
I make the model levels thicker by 7% the ifort problem still occurs
after about 10 steps (compared to 3 before).

ifort options:

-O0 makes the code run OK apparently, but very very slow.

-check all -fpe0 -traceback (and -O2) also makes the code run OK
apparently (at least to step 28), but the build and run are very slow.
This confuses me. Aren't -check all -fpe0 -traceback just diagnostic
options?

This is starting to look like a compiler bug?  Anything else I should
check?

Cheers, Tom.


        



More information about the MITgcm-support mailing list