[Mitgcm-support] Re: numbers

[mitgcm-support at dev.mitgcm.org]

A.,

Didn't manage to get the numbers with the plain checkout.
Did get comparable numbers (12 digitis) on soyinka and xolas and
identical to 14 digitis on xolas and on xolas + 6-threading.

But isn't the code actually wrong without
the UV exchange of hFacW/hFacS fix? Not sure
why since they are symmetric in this config. However, that fix 
changes the numbers and gets rid of blow up at 720 secs.
Think it should be an _RS exchange, but I am
running #define REAL4_IS_SLOW at the mom.

Just trying a run on soyinka with the old ini_depths
and _exch_r4 stuff to see if that blows too.
If it works I can change ini_depths in a way
that definitely has no impact. Bit confused though,
can't see how the new form differs from the old - 
do you? 

Chris
Alistair Adcroft wrote:
> 
> Did you reproduce the numbers in hs94.cs*/results/output.txt?
> 
> I tried using your fixed-up ini_depths.F and couldn't.
> 
> A.