From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-return-45025-listarch-gcc=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 8131 invoked by alias); 24 Jan 2002 01:21:06 -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 8047 invoked from network); 24 Jan 2002 01:21:04 -0000
Received: from unknown (HELO dberlin.org) (64.246.6.106)
  by sources.redhat.com with SMTP; 24 Jan 2002 01:21:04 -0000
Received: from localhost (localhost.localdomain [127.0.0.1])
	by dberlin.org (8.11.6/8.11.6) with ESMTP id g0O1Kwn20889;
	Wed, 23 Jan 2002 20:20:58 -0500
Date: Wed, 23 Jan 2002 18:27:00 -0000
From: Daniel Berlin <dan@dberlin.org>
To: Dan Nicolaescu <dann@godzilla.ICS.UCI.EDU>
cc: gcc@gcc.gnu.org
Subject: Re: g++ and aliasing bools
In-Reply-To: <200201231638.aa22100@gremlin-relay.ics.uci.edu>
Message-ID: <Pine.LNX.4.44.0201232018510.20726-100000@dberlin.org>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
X-SW-Source: 2002-01/txt/msg01489.txt.bz2

On Wed, 23 Jan 2002, Dan Nicolaescu 
wrote:

> 
> It looks like g++ has some problems with aliasing bools, and gcc
> doesn't.

It's a side effect of C++ not making alias sets for classes, i think.
Try commenting out the AGGREGATE_TYPE_P check in cxx_get_alias_set and see 
if it goes away.
> 
> 
> The following program: 
> 
> #ifdef __cplusplus
> 
> struct bool_struct         
> {
>   bool f0;
>   bool f1;
>   bool f2; 
>   bool f3;
> };
> #else
> struct bool_struct         
> {
>   _Bool f0;
>   _Bool f1;
>   _Bool f2; 
>   _Bool f3;
> };
> 
> #endif
> 
> 
> struct bool_struct *str; 
> 
> 
> void 
> g (void)
> {
>   str->f0 = 1;
>   str->f1 = 0;
>   str->f2 = 1;
>   str->f3 = 0; 
> }
> 
> compiles to: 
> 
> _Z1gv:
> .LLFB1:
>         !#PROLOGUE# 0
>         !#PROLOGUE# 1
>         sethi   %hi(str), %o2
>         ld      [%o2+%lo(str)], %o1
>         mov     1, %o3
>         stb     %o3, [%o1]
>         ld      [%o2+%lo(str)], %o0   <- this load is not needed
>         stb     %g0, [%o0+1]
>         ld      [%o2+%lo(str)], %o1   <- this load is not needed
>         stb     %o3, [%o1+2]
>         ld      [%o2+%lo(str)], %o0   <- this load is not needed
>         retl
>         stb     %g0, [%o0+3]
> 
> when compiled with g++ -O2  on sparc-sun-solaris2.8 using a GCC from
> CVS as of this morning. 
> 
> but the code is much better when compiling with gcc -O2 
> 
> g:
>         !#PROLOGUE# 0
>         !#PROLOGUE# 1
>         sethi   %hi(str), %o0
>         ld      [%o0+%lo(str)], %o1
>         mov     1, %o2
>         stb     %o2, [%o1+2]
>         stb     %g0, [%o1+3]
>         stb     %o2, [%o1]
>         retl
>         stb     %g0, [%o1+1]
> 
> 
> so it looks like g++ has some issues with aliasing bools. 
> 
> Any idea what is wrong? 
> 
>         
>