From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-patches-return-207621-listarch-gcc-patches=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 24488 invoked by alias); 13 Nov 2007 10:30:44 -0000
Received: (qmail 24471 invoked by uid 22791); 13 Nov 2007 10:30:42 -0000
X-Spam-Check-By: sourceware.org
Received: from mu-out-0910.google.com (HELO mu-out-0910.google.com) (209.85.134.189)     by sourceware.org (qpsmtpd/0.31) with ESMTP; Tue, 13 Nov 2007 10:30:17 +0000
Received: by mu-out-0910.google.com with SMTP id g7so1790889muf         for <gcc-patches@gcc.gnu.org>; Tue, 13 Nov 2007 02:30:15 -0800 (PST)
Received: by 10.64.28.3 with SMTP id b3mr7645544qbb.1194949814057;         Tue, 13 Nov 2007 02:30:14 -0800 (PST)
Received: by 10.65.232.20 with HTTP; Tue, 13 Nov 2007 02:30:14 -0800 (PST)
Message-ID: <84fc9c000711130230y5d56d07br33b83ce2c4e8cb94@mail.gmail.com>
Date: Tue, 13 Nov 2007 11:18:00 -0000
From: "Richard Guenther" <richard.guenther@gmail.com>
To: "Bernd Schmidt" <bernds_cb1@t-online.de>
Subject: Re: [PATCH] Fix PR34027, SCEV const-prop pessimizes -Os
Cc: gcc-patches <gcc-patches@gcc.gnu.org>
In-Reply-To: <4738EFF2.1010505@t-online.de>
MIME-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
References: <84fc9c000711110433x4c5fb5ebnfbb78cfac73f987a@mail.gmail.com> 	 <4738EFF2.1010505@t-online.de>
X-IsSubscribed: yes
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: 2007-11/txt/msg00696.txt.bz2

On Nov 13, 2007 1:29 AM, Bernd Schmidt <bernds_cb1@t-online.de> wrote:
> Richard Guenther wrote:
> > With the now famous enough testcase
> >
> > unsigned long long foobar(unsigned long long ns)
> > {
> >   while(ns >= 1000000000L)
> >     ns -= 1000000000L;
> >   return ns;
> > }
> >
> > SCEV const-prop creates the expression
> >
> >   ns + (ns /[fl] 1000000000L) / -1000000000L
> >
> > for the final value of ns
>
> Am I the only one here who thinks this is a spectacularly bad idea?
> Divisions are expensive, especially on targets without hardware divide.

No you are not ;)  In fact the kernel people complained about this
already.

Richard.