From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-patches-return-339261-listarch-gcc-patches=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 24195 invoked by alias); 4 Apr 2013 15:15:29 -0000
Mailing-List: contact gcc-patches-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Id: <gcc-patches.gcc.gnu.org>
List-Archive: <http://gcc.gnu.org/ml/gcc-patches/>
List-Post: <mailto:gcc-patches@gcc.gnu.org>
List-Help: <mailto:gcc-patches-help@gcc.gnu.org>
Sender: gcc-patches-owner@gcc.gnu.org
Received: (qmail 24186 invoked by uid 89); 4 Apr 2013 15:15:28 -0000
X-Spam-SWARE-Status: No, score=-0.9 required=5.0 tests=BAYES_00,KHOP_DYNAMIC2,KHOP_THREADED,RCVD_IN_SEMBLACK,RCVD_IN_SORBS_DUL,RDNS_DYNAMIC autolearn=no version=3.3.1
Received: from c-24-4-193-8.hsd1.ca.comcast.net (HELO mrs.kithrup.com) (24.4.193.8)    by sourceware.org (qpsmtpd/0.84/v0.84-167-ge50287c) with ESMTP; Thu, 04 Apr 2013 15:15:26 +0000
Received: from up.mrs.kithrup.com (up.mrs.kithrup.com [IPv6:2001:470:81e3::1])	by mrs.kithrup.com (Postfix) with ESMTPSA id 4F42E574C630;	Thu,  4 Apr 2013 08:15:23 -0700 (PDT)
Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0 (Mac OS X Mail 6.3 \(1503\))
Subject: Re: [RTL] Canonicalize commutative operations more
From: Mike Stump <mrs@mrs.kithrup.com>
In-Reply-To: <1420303.Yr0JhII12W@polaris>
Date: Thu, 04 Apr 2013 18:57:00 -0000
Cc: Marc Glisse <marc.glisse@inria.fr>, gcc-patches@gcc.gnu.org
Content-Transfer-Encoding: quoted-printable
Message-Id: <85308AA0-C2BC-42DC-8D15-4C7D759F57C0@mrs.kithrup.com>
References: <alpine.DEB.2.02.1303191715280.4515@stedding.saclay.inria.fr> <2628063.UuGJrOXufS@polaris> <alpine.DEB.2.02.1303301322310.12904@laptop-mg.saclay.inria.fr> <1420303.Yr0JhII12W@polaris>
To: Eric Botcazou <ebotcazou@adacore.com>
X-SW-Source: 2013-04/txt/msg00262.txt.bz2

On Apr 2, 2013, at 1:55 AM, Eric Botcazou <ebotcazou@adacore.com> wrote:
> The question is: do we really need to canonicalize everything?

> The fear (at least mine) is that, by canonicalizing everything, you will =
make=20
> changes behind the back of back-ends that could disable some of their pat=
terns=20
> silently.

My take, yes, we really need to.  My take, during construction, the canonic=
alization code should kick in and build it canonically.  If you had a total=
 ordering predicate, and defined an ordering for any two rtls, then the bui=
lder could use it and then there would be no case in which things are not c=
anonicalized.  If one tries to do this after the fact, we'll, let's just sa=
y, no good come from that.