[MITgcm-support] ifort/g77 issue

[Constantinos Evangelinos]

On Wed 23 Jan 2008, Thomas Haine wrote:

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

Unfortunately this can be a sign of either a compiler bug or of a hidden bug 
in the code that is not exposed with -O0 because of things like zero-ing out 
of uninitialized vars that may be happening in that case.

> -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?

-check all does runtime array bounds checking and many more things and thus 
does slow things down quite a lot.

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

To prove a compiler bug without knowing exactly what it is and where it messes 
things up you have to show that the code works fine with most/all other 
compilers - then probability weighs in and points the finger at Intel. I'd 
try gfortran, g95, open64 and pgi.

Constantinos
-- 
Dr. Constantinos Evangelinos
Department of Earth, Atmospheric and Planetary Sciences
Massachusetts Institute of Technology