[patch] Fix behavior of TER on unrolled loops

[Zdenek Dvorak]

Hello,

> >Yes, it is a workaround in the sense that it does not really fix
> >all the possible problems.  If you try hard enough, you are still able
> >to persuade TER to cause this behavior.  This patch is just good 
> >enough to
> >catch the common case coming from unrolling.
> 
> And you most likely could reproduce it even without the tree level
> optimizations pre 4.0.0 so this is not the correct place for this.
> This is what Daniel is trying to say, you don't need TER to reproduce
> it.

yes, I do -- if I did not, I would not be changing TER, would I?
Of course I can also produce the code created by TER by hand.  But then,
I would not consider it a bug in compiler if it is not able to do
anything with it; minimizing register pressure is NP-complete, thus I
cannot reasonably expect compiler to do it in all cases.

Zdenek