Loop optimizer issues

[Zdenek Dvorak]

Hello,

> >>>>> Steven Bosscher writes:
> 
> Steven> Also, look at the deficiencies list.  Can all those missing pieces be
> Steven> provided in time for 3.4?  (Do they make a difference?)
> 
> 	The current "new" loop unroller optimizer already has a number of
> deficiencies on the GCC 3.4 trunk.  The question is whether this rewrite
> addresses some of those regressions as good or better than reverting to
> the old unroller on the trunk.  In other words, this is new functionality,
> but is it also a regression fix early in Stage 2 of development?

some of the reported deficiences would be fixed. Many new would be
introduced.

Zdenek