[patch] Speed up # of iterations analysis

[Richard Guenther]

On Apr 4, 2005 1:17 PM, Sebastian Pop <sebastian.pop@cri.ensmp.fr> wrote:
> Zdenek Dvorak wrote:
> > actually ivopts perform this optimization, but obviously this is too
> > late for the rest of the loop optimization passes.
> >
> 
> Ah, I didn't know.  Probably we could split this optimization and
> schedule it a little earlier before loop optimizations?

Actually, as we're now trying to preserve loop information in more
passes and possibly
doing loop_init very early after going into SSA we might do an early
(extra) run of cunroll and
ivopts without much cost.  See PR18754 and related.

Richard.