Re: [glob2-devel] gradient improvement

[Kai Antweiler]

> If we want to rely on the fact that listedAddr is ordered for further
> optimization, we have to fork the cases. One who can call the optimized one, 
> and
> one who can only call the less-optimized one. This would be smart actually.

I need this forking.

I have just implemented a new gradient algorithm and got a interesting
situation.  It's getting good results - unless I use AICastor, in this
case it uses so much cpu that I can't even close glob2.

It's in current Map.cpp, if you want to see it yourself.
Put in "#define KAI" to test it.

I don't know why this happens, it should not.
Actually it cannot.  In the worst, case there are only a few conditionals
more than in the simple version.
But it does happen.
-- 
Kai Antweiler