This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: g77 performance on ALPHA
- To: John Wehle <john at feith dot com>
- Subject: Re: g77 performance on ALPHA
- From: Jeffrey A Law <law at cygnus dot com>
- Date: Thu, 02 Sep 1999 00:34:11 -0600
- cc: torvalds at transmeta dot com, gcc at gcc dot gnu dot org, toon at moene dot indiv dot nluug dot nl, rth at cygnus dot com, martin dot kahlert at provi dot de
- Reply-To: law at cygnus dot com
In message <199909011920.PAA06128@jwlab.FEITH.COM>you write:
> > Linus Torvalds <torvalds@transmeta.com> wrote:
> > >
> > >It looks like the Compaq compiler knows the stores cannot alias the
> > >loads, while g77 thinks (or tells the instruction scheduler) that the
> > >stores could alias and thus they never get moved down.
> >
> > Naah, I didn't notice that the gcc output actually has all the
> > compare-and-branch instructions still in the unrolled body of the loop,
> > and those will obviously be serialization points regardless of any
> > aliases: the reason the loads haven't migrated up and the stores down
> > like they should is all those control dependencies.
>
> Not sure if this is relevant, however attached is a patch I'm playing with
> which eliminates redundant conditional jumps resulting from loop unrolling.
> I'm not aware of any problems with it, though it's still a work in progress
> I would be interested in hearing if it helps.
This looks like something "thread_jumps" is supposed to be handling.
jeff