[ODE] ANN : Update ODE distribution!!!

[Adam D. Moss]

Nguyen Binh wrote:
> ADM> I'm trying to pick apart the precise differences between this
> ADM> and CVS-HEAD so that the juicy bits can be tested individually
> ADM> and submitted back to Russ for inclusion in CVS.  (It would be
> ADM> good if you could try to do this yourself Nguyen, but...)
> 
>      I have an messy notes about what I've done but it's too
>      embarassed to show. :) . Moreove, I don't have enough time...

That's okay, time is always a problem.  I can separate out
some parts myself, but others will have to be extracted by
people who can test them (like the VC.NET stuff or anything
new that's not Makefile'd like your terrain demo).  The risk
of course is that some changes will get lost in the noise
and never get applied to CVS.

> ADM> The thing I'm most interested in isolating is the removal of
> ADM> many dSetZero() calls, which seems like a universally useful
> ADM> optimization.  I'll try to get to those first.
>      Yes. I expect some performance gain after removals.I did it quite
>      a while ago when I SIMDize ODE.Recently, some one post on the
>      list about this so I publish my patch.
> 
>      To anyone use VTune, please do experiments and publish the results.

I'll do some testing myself when I manage to un-break my
test harness.  :)

BTW, I don't see that dSetZeroSSE() is being called from
anywhere... is that intentional?

Thanks,
--Adam
-- 
Adam D. Moss   . ,,^^   adam at gimp.org   http://www.foxbox.org/   co:3
"i18n ought to be abbreviated i2n to make it quicker to write!"-snout