This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: GCC trunk SPEC2000 performance
- From: law at redhat dot com
- To: Andreas Jaeger <aj at suse dot de>
- Cc: Toon Moene <toon at moene dot indiv dot nluug dot nl>, Richard Henderson <rth at redhat dot com>, Jan Hubicka <jh at suse dot cz>, Diego Novillo <dnovillo at redhat dot com>, David Edelsohn <dje at watson dot ibm dot com>, Mark Mitchell <mark at codesourcery dot com>, gcc at gcc dot gnu dot org
- Date: Sat, 22 Jun 2002 09:55:36 -0600
- Subject: Re: GCC trunk SPEC2000 performance
- Reply-to: law at redhat dot com
In message <u87kkr8hm6.fsf@gromit.moeb>, Andreas Jaeger writes:
> law@redhat.com writes:
>
> > In message <3D125DB3.F3B600B5@moene.indiv.nluug.nl>, Toon Moene writes:
> > > Richard Henderson wrote:
> > >
> > > > On Fri, Jun 21, 2002 at 12:02:00AM +0200, Toon Moene wrote:
> > > > > So, OK - that's it. Addressof makes a MEM out of this register, so
> we
> > > > > should accept it down the road.
> > >
> > > > Alternately, we may want to remove it from the reg_rtx array.
> > >
> > > Well, that would make this code a lot clearer for unsuspecting onlooker
> s
> > > like me :-)
> > Or even for suspecting folks ;-)
> >
> > I think all we'd need to do is clear the entry in the 4 places where we mo
> dify
> > REGs in-place and turn them into MEMs. 3 are in function.c plus one in re
> load.
> >
> > There's the possibility we have code which walks the entries in the reg ar
> ray
> > and isn't prepared for an entry that has been wiped away. So we should go
> > through the usual testing procedures for such a change.
>
> Who's going to do this? Honza is going on holiday and I fear he won't
> have time to fix it before.
>
> Shall I file a GNATS entry for this referring to the discussion?
I've been poking at it. Let's just say there's certainl code which doesn't
deal well with clearing those entries. Depending on how many more problems
we find, we may fall back to Jan's patch + comment updates.
jeff