[MITgcm-devel] compiler optimization on SunOS

[Dimitris Menemenlis]

Martin, et al., slightly off topic for this discussion, but I just received 
following e-mail from NASA Ames, which indicates that -O3 may not always be 
faster than -O2 optimization.  D.

> Hi Dimitris:
> 
> 
> Best times at 1:20 (PBS wall clock time) are -O2 with both v10 beta and
> 9.1.039.
> 
> Second best time is 1:24 (PBS wall clock time) using -O3 and v10 beta
> 
> And in third and last place: 1:30 using -O3 and v9.1.039.
> 
> I'm going to ask Ken to do a -O2 timing with his version of the code.
> 
> I suppose we all assumed -O3 would be better and sometimes it just isn't.
> 
> I'll pluck out the exact times from the runs a bit later on.
> 
> Art