Serious performance regression -- some tree optimizer questions

[Zdenek Dvorak]

Hello,

> > *************** add_to_evolution_1 (unsigned loop_nb, 
> > *** 704,709 ****
> > --- 716,722 ----
> >   	      var = loop_nb;
> >   	      left = chrec_before;
> >   	      right = build_int_cst (type, 0);
> > + 	      chrec_before = NULL;
> 
> I would prefer
>   chrec_before = chrec_not_analyzed_yet;
> 
> but this is not very important.

I would prefer not to; chrec_not_analyzed_yet seems like that it
should be used when the chrec is not analyzed yet, which obviously
is not the case here.

Zdenek