Re: [PING][PATCH][REVISED] Fix PR middle-end/PR28690, modify swap_commutative_operands_p

["H. J. Lu" <hjl at lucon dot org>]

On Thu, Jun 21, 2007 at 07:16:03AM -0700, H. J. Lu wrote:
> On Thu, Jun 21, 2007 at 03:16:38PM +0200, Richard Guenther wrote:
> > On 6/21/07, H. J. Lu <hjl@lucon.org> wrote:
> > >On Wed, Jun 20, 2007 at 12:17:38PM -0500, Peter Bergner wrote:
> > >> On Tue, Jun 19, 2007 at 05:38:14PM +0200, Paolo Bonzini wrote:
> > >> > 2) This has nothing to do with this patch which -- in my opinion -- is
> > >> > now ready to be committed (as soon as the regression period finishes).
> > >>
> > >> Now that we seem to have resolved the x86_64 performance regressions,
> > >> I'd like to ask for another review of this patch.
> > >>
> > >> I had to refresh the patch (attached below) since the last time I posted 
> > >it
> > >> due to the dataflow merge.  I bootstrapped and regtested it on 
> > >powerpc64-linux
> > >> and everything still looks good.  Is this patch ok for mainline once the
> > >> freeze is over?
> > >>
> > >
> > >Gcc 4.3 revision 125901 miscompiles 447.dealII in SPEC CPU 2006
> > >and revision 125849 is OK. I have to rebuild my gcc with revision
> > >125849 and restart my SPEC 2000/2006 runs. It will take a while.
> > 
> > For me it's revision 125761 that is bad and revision 125748 that is ok.
> > Which would be the pointer-plus merge.
> > 
> 
> Oops. Wrong info. Gcc 4.3 revision 125740 is the last good one for me.
> 

Peter, do you have a patch against revision 125740?


H.J.