Jack Scheible wrote:
> I respectfully submit that, if your models are so sensitive to "change
> in number format from Cray to ieee, and the calculation order changes
> required for MPP," then your algorithms are decidedly unstable.
It is fairly widely known that atmospheric and ocean models
of all kinds are thoroughly chaotic and ill-conditioned. I
have heard this referred to as "the Butterfly Effect"; The
rate at which a butterfly moves its wings in Tokyo today
will have as significant effect on the weather in Chicago
ten days from now as will most other measurable quantities.
So, I would agree that the algorithms are unstable, but then
again, the theory says they should be. So there is no
negative connotation to instability, just a mountain of
painstaking work to partially compensate for it.
-------- Cray Research --------- Roger Glover
-- A Silicon Graphics Company -- http://home.cray.com/~glover
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|