From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-return-154449-listarch-gcc=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 9904 invoked by alias); 5 Jun 2009 15:36:54 -0000
Received: (qmail 9896 invoked by uid 22791); 5 Jun 2009 15:36:53 -0000
X-SWARE-Spam-Status: No, hits=-2.4 required=5.0 	tests=AWL,BAYES_00,SPF_PASS
X-Spam-Check-By: sourceware.org
Received: from mail-bw0-f224.google.com (HELO mail-bw0-f224.google.com) (209.85.218.224)     by sourceware.org (qpsmtpd/0.43rc1) with ESMTP; Fri, 05 Jun 2009 15:36:41 +0000
Received: by bwz24 with SMTP id 24so1798029bwz.8         for <gcc@gcc.gnu.org>; Fri, 05 Jun 2009 08:36:38 -0700 (PDT)
Received: by 10.103.226.20 with SMTP id d20mr2298039mur.8.1244216198507;         Fri, 05 Jun 2009 08:36:38 -0700 (PDT)
Received: from ?192.168.2.99? (cpc2-cmbg8-0-0-cust61.cmbg.cable.ntl.com [82.6.108.62])         by mx.google.com with ESMTPS id e9sm386636muf.2.2009.06.05.08.36.37         (version=SSLv3 cipher=RC4-MD5);         Fri, 05 Jun 2009 08:36:37 -0700 (PDT)
Message-ID: <4A293E50.4010301@gmail.com>
Date: Fri, 05 Jun 2009 15:36:00 -0000
From: Dave Korn <dave.korn.cygwin@googlemail.com>
User-Agent: Thunderbird 2.0.0.17 (Windows/20080914)
MIME-Version: 1.0
To: fearyourself <fearyourself@gmail.com>
CC: "gcc@gcc.gnu.org" <gcc@gcc.gnu.org>
Subject: Re: Expanding a load instruction
References: <c568a2600906050832y1d344449l5aef478548ad3bca@mail.gmail.com>
In-Reply-To: <c568a2600906050832y1d344449l5aef478548ad3bca@mail.gmail.com>
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Mailing-List: contact gcc-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Id: <gcc.gcc.gnu.org>
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
X-SW-Source: 2009-06/txt/msg00103.txt.bz2

fearyourself wrote:

> In the instruction set of my architecture, the offsets of a half-load
> (HImode) have to be multiples of 2. However, if I set up a structure
> in a certain way, the compiler will generate:
> 
> (mem/s/j:HI (plus:DI (reg:DI 134 [ ivtmp.23 ])
>         (const_int 1 [0x1])) [0 <variable>.geno+0 S2 A16])
> 
> As the memory operand for the load.
> 
> Now, one solution I am going to try to fix this is to use
> define_expand and add a move into another register before this load
> and then load from that register (thus removing the offset of 1).
> 
> My question is: Is that how it should be done or is there another solution?

  This looks like what you need:

 -- Macro: STRICT_ALIGNMENT
     Define this macro to be the value 1 if instructions will fail to
     work if given data not on the nominal alignment.  If instructions
     will merely go slower in that case, define this macro as 0.

    cheers,
      DaveK