[Mitgcm-support] ECCO on SunOS and optimization levels

[mitgcm-support at dev.mitgcm.org]

What I first thought to be problem of MITgcm c44
w.r.t. TAMC (used instead of TAF)
turns out to be yet another compiler problem.
Essentially, a NaN suddenly appears out of nowhere.
If I reduce to optimization level 0 the TAMC-generated
code runs fine on the Sun's.

The question is whether we should, by default,
exclude the optimization for the Sun's since they've
caused so many troubles in the past.