From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-return-70669-listarch-gcc=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 19513 invoked by alias); 19 Mar 2003 13:06:27 -0000
Mailing-List: contact gcc-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Archive: <http://gcc.gnu.org/ml/gcc/>
List-Post: <mailto:gcc@gcc.gnu.org>
List-Help: <http://gcc.gnu.org/ml/>
Sender: gcc-owner@gcc.gnu.org
Received: (qmail 19466 invoked from network); 19 Mar 2003 13:06:26 -0000
Received: from unknown (HELO mail.libertysurf.net) (213.36.80.91)
  by sources.redhat.com with SMTP; 19 Mar 2003 13:06:26 -0000
Received: from localhost.localdomain (212.83.190.180) by mail.libertysurf.net (6.5.026)
        id 3DE4935E010F8C98; Wed, 19 Mar 2003 14:06:23 +0100
Content-Type: text/plain;
  charset="iso-8859-1"
From: Eric Botcazou <ebotcazou@libertysurf.fr>
To: Joern Rennecke <joern.rennecke@superh.com>
Subject: Re: Reload question
Date: Wed, 19 Mar 2003 19:39:00 -0000
User-Agent: KMail/1.4.3
References: <3E7862AE.38774FFD@superh.com>
In-Reply-To: <3E7862AE.38774FFD@superh.com>
Cc: gcc@gcc.gnu.org
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Message-Id: <200303191405.56799.ebotcazou@libertysurf.fr>
X-SW-Source: 2003-03/txt/msg01252.txt.bz2

> This address looks suspicious to start with.  Where does it
> come from?

It comes from the combination during the reload pass of

(subreg:SI (reg:DF 770) 4))

and 

(set (reg:DF 770)
        (mem/s:DF (plus:SI (reg/f:SI 767)
                (reg:SI 769)) [16 dense.eden+0 S8 A64]))

> The only kind-of legitimate way I can think of
> is if you want to rematerialize reg+reg, and this has
> suceeded because both registers are function invariants.

I have to say that I don't fully understand this sentence.

-- 
Eric Botcazou