[ODE] Non-deterministic behavior

[Gary R. Van Sickle]

> On Sun, 31 Aug 2003, Gary R. Van Sickle wrote:
>
> > I wonder if it wouldn't make sense to:
> >
> > 1.  Implement a portable PRNG in ODE used just for randomizing joint solving
> > order.
> > 2.  Change that to #define PSEUDORANDOM_JOINT_ORDER.
> > 3.  Reseed the PRNG apropriately on reset.
> >
> > ...so that even with random joint solving order, behavior is still
> > deterministic.
>
> The best of both worlds - pseudo-randomness and determinism.  That sounds
> to me like a good idea.
>

In fact it might be better to eliminate the #define entirely, since IIRC the
scrambled ordering is always a win, isn't it?

--
Gary R. Van Sickle
Brewer.  Patriot.