From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-patches-return-270038-listarch-gcc-patches=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 3698 invoked by alias); 8 Aug 2010 08:44:23 -0000
Received: (qmail 3687 invoked by uid 22791); 8 Aug 2010 08:44:22 -0000
X-SWARE-Spam-Status: No, hits=-3.4 required=5.0	tests=AWL,BAYES_00,T_RP_MATCHES_RCVD
X-Spam-Check-By: sourceware.org
Received: from cantor2.suse.de (HELO mx2.suse.de) (195.135.220.15)    by sourceware.org (qpsmtpd/0.43rc1) with ESMTP; Sun, 08 Aug 2010 08:44:17 +0000
Received: from relay1.suse.de (charybdis-ext.suse.de [195.135.221.2])	by mx2.suse.de (Postfix) with ESMTP id 9AA438655F;	Sun,  8 Aug 2010 10:44:15 +0200 (CEST)
Date: Sun, 08 Aug 2010 08:44:00 -0000
From: Richard Guenther <rguenther@suse.de>
To: Eric Botcazou <ebotcazou@adacore.com>
Cc: gcc-patches@gcc.gnu.org, Arnaud Charlet <charlet@adacore.com>
Subject: Re: [PATCH] Keep lattice abstract in the SSA propagator, substitue lattice values at defs
In-Reply-To: <201008081000.41310.ebotcazou@adacore.com>
Message-ID: <alpine.LNX.2.00.1008081039350.25856@zhemvz.fhfr.qr>
References: <alpine.LNX.2.00.1008041417350.25856@zhemvz.fhfr.qr> <20100804123142.GA83420@adacore.com> <alpine.LNX.2.00.1008041437130.25856@zhemvz.fhfr.qr> <201008081000.41310.ebotcazou@adacore.com>
User-Agent: Alpine 2.00 (LNX 1167 2008-08-23)
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
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
X-SW-Source: 2010-08/txt/msg00584.txt.bz2

On Sun, 8 Aug 2010, Eric Botcazou wrote:

> >   D.2544_253 = prephitmp.53_302 + 2147483647;
> >   D.2553_254 = (size_type) D.2544_253;
> 
> The signed integer overflow is introduced during the reassoc1 pass.  For the 
> attached reduced testcase, we start from:
> 
> <bb 45>:
>   D.2615_217 = L43b_216 + -1;
>   iftmp.30_218 = D.2615_217 + -2147483648;
>   D.2624_225 = (size_type) iftmp.30_218;
> 
> and we go to:
> 
> <bb 45>:
>   iftmp.30_218 = L43b_216 + 2147483647;
>   D.2624_225 = (size_type) iftmp.30_218;

Well, I remember a similar kind of problem fixed in fold - basically
you can't re-associate (a + b) + c to a + (b + c) with signed
arithmetic.  But - we can do so if reassociating constants only
as in the example above.  Because if (L43b_216 - 1) - 2147483648
doesn't overflow then L43b_216 + 2147483647 doesn't either.

So I believe this specific problem pre-exists before re-assoc.

Richard.