This is the mail archive of the
mailing list for the GCC project.
Re: Loop optimizer issues
- From: Andrew Pinski <pinskia at physics dot uc dot edu>
- To: wilson at tuliptree dot org (Jim Wilson)
- Cc: dje at watson dot ibm dot com (David Edelsohn), gcc at gcc dot gnu dot org
- Date: Mon, 25 Aug 2003 22:27:55 -0400 (EDT)
- Subject: Re: Loop optimizer issues
> David Edelsohn wrote:
> > What is the status of merging these optimizations into GCC 3.4?
> > GCC 3.4 currently exhibits a noticeable performance regression. I want to
> > make sure the regression and the proposed patches are noticed now so that
> > this is not a surprise later as we get closer to the release.
> If these are regressions resulting from my loop/alias bug fixes, then it
> might be a better strategy to improve the handling of alias info inside
> loop, rather than add new loop optimizations, or rewrite existing ones.
David was not taking about the aliasing bug fixes but rather the loop unroll
does not work for PPC with most loops as it really does not understand the
doloop optimizations which benfit the PPC greatly.