This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: comments on loop changes
- To: Michael Hayes <m dot hayes at elec dot canterbury dot ac dot nz>
- Subject: Re: comments on loop changes
- From: Jeffrey A Law <law at hurl dot cygnus dot com>
- Date: Thu, 25 Feb 1999 02:59:04 -0700
- cc: Richard Henderson <rth at cygnus dot com>, egcs-patches at egcs dot cygnus dot com
- Reply-To: law at cygnus dot com
In message <14036.34683.62275.477061@ongaonga.elec.canterbury.ac.nz>you write
:
> Jeffrey A Law writes:
> > In message <14022.16263.433300.727000@ongaonga.elec.canterbury.ac.nz>y
> ou writ
> > e:
> > > One other thing, is the second loop optimisation pass still necessar
> y?
> > We've already discussed this several times -- we need to benchmark befor
> e
> > it goes away.
>
> With J"oern and Richards mods to strength reduction, I am noticing
> minimal benefit of rerunning the loop optimiser.
>
> However, the experimentation I've done with the C4x indicates that it
> is desirable to rerun the loop optimiser after loop unrolling.
>
> I would be interested if other folks notice an improvement with this
> patch when unrolling loops, especially on machines with autoincrement
> addressing.
>
> Thu Feb 25 12:22:56 1999 Michael Hayes <m.hayes@elec.canterbury.ac.nz>
>
> * toplev.c (rest_of_compilation): Unroll loops on the first
> call to loop_optimize.
I don't think this is necessarily a safe thing to do. We had a fix a slew
of obscure bugs already due to weird interactions between unrolling and
the multiple loop optimization passes.
If you're going to suggest doing this, I'm going to ask that you grub through
the archives for messages related to this to see what issues need to be
resolved if one wants to run loop after unrolling.
jeff