This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: [tree-profiling] Fix return predictors
- From: Steven Bosscher <stevenb at suse dot de>
- To: kenner at vlsi1 dot ultra dot nyu dot edu (Richard Kenner), jh at suse dot cz
- Cc: gcc at gcc dot gnu dot org
- Date: Sat, 14 Aug 2004 11:54:01 +0200
- Subject: Re: [tree-profiling] Fix return predictors
- Organization: SUSE Labs
- References: <10408140121.AA18215@vlsi1.ultra.nyu.edu>
On Saturday 14 August 2004 03:21, Richard Kenner wrote:
> [Moved from gcc-patches to gcc.]
>
> > IMHO it is an absolute _must_ that the old loop optimizer can
> > be removed before GCC 3.5 is released.
>
> Yep, that would be nice tought ;)
>
> I very strongly disagree. We have no hope of catching all loop invariants
> at tree level without putting a tremendous amount of knowlege into the tree
> formats and, even if we did it, that knowlege would interfere with other
> optimizations.
And to catch loop invariants at the RTL level, we need the old loop
optimizer? I think you're missing the point...
The old loop optimizer is the one that needs loop notes and destroys
the CFG so we can't run the RTL loop optimizers earlier and we can't
have profile information earlier. That's loop.c and unroll.c. Those
should go away. The replacements are already there, except strength
reduction, which is on the LNO branch waiting to be merged.
Gr.
Steven