[Mitgcm-support] Re: [Fwd: Code import]

[mitgcm-support at dev.mitgcm.org]

John,

I've also just been through the process of testing the suite of expts
on a Sun-Blade-1000 (Workshop 6 update 1)

John Stark wrote:
> Currently the model passes exp0 to 4, but crashes during exp5.  I'm still 
> looking into that one.

exp5 failed due to stacksize limits - it gave a seg fault.
"unlimit" at the command prompt was all I had to do to fix that.

The numerical accuracy of the hs* experiments has been resolved. The type
casting
of constants behaves differently on the Sun, Dec and Intel box. I've reconciled
all those platforms and the new accurate output.txt files will be in c40.

Currently, pkg/aim and the associated expts in verification/aim.* are very
unportable and is being worked on currently. I saw an email yesterday
claiming that fixes were about to be posted...  I suggest you avoid
running the aim.* experiments and use genmake -disable=aim to avoid the
source code until those fixes are available.

The conflict with the "-u" option: was that in pkg/exf of routines from
model/src?
I didn't follow what you were referring to.

Thanks for input,

Alistair.