From owner-freebsd-hackers  Sun Jul  2 17:35:43 2000
Delivered-To: freebsd-hackers@freebsd.org
Received: from overcee.netplex.com.au (peter1.corp.yahoo.com [208.48.107.4])
	by hub.freebsd.org (Postfix) with ESMTP id 10F2737C240
	for <hackers@FreeBSD.ORG>; Sun,  2 Jul 2000 17:35:31 -0700 (PDT)
	(envelope-from peter@netplex.com.au)
Received: from netplex.com.au (localhost [127.0.0.1])
	by overcee.netplex.com.au (Postfix) with ESMTP
	id 9B6AB1CD7; Sun,  2 Jul 2000 17:35:30 -0700 (PDT)
	(envelope-from peter@netplex.com.au)
X-Mailer: exmh version 2.1.1 10/15/1999
To: John Polstra <jdp@polstra.com>
Cc: hackers@FreeBSD.ORG
Subject: Re: GCC extended asm experts please look at this 
In-Reply-To: Message from John Polstra <jdp@polstra.com> 
   of "Sun, 02 Jul 2000 15:01:04 PDT." <XFMail.000702150104.jdp@polstra.com> 
Date: Sun, 02 Jul 2000 17:35:30 -0700
From: Peter Wemm <peter@netplex.com.au>
Message-Id: <20000703003530.9B6AB1CD7@overcee.netplex.com.au>
Sender: owner-freebsd-hackers@FreeBSD.ORG
Precedence: bulk
X-Loop: FreeBSD.ORG

John Polstra wrote:
> I _thought_ I was an expert in gcc's extended asm feature, but I
> can't figure out why this won't compile when optmization is
> disabled:
> 
> =============================================================================
    ==
> #define xchgl(v, m) ({                          \
>         int __result;                           \
>         __asm __volatile ("xchgl %0, %1"        \
>             : "=r"(__result), "=m"(m)           \
>             : "0"(v), "1"(m));                  \
>         (__result); })
> 
> void
> lock80386_acquire(volatile int *lock)
> {
>         while (xchgl(1, *lock) != 0)
>                 while (*lock != 0)
>                         ;
> }
> =============================================================================
    ==
> 
> It compiles and works fine with -O or higher; but without -O gcc
> says:
> 
>     locktest.c: In function `lock80386_acquire':
>     locktest.c:11: inconsistent operand constraints in an `asm'
> 
> This happens with both gcc-2.95.2 (the version in -current) and with
> the much older gcc-2.7.2.3.
> 
> I believe the code is correct according to the documentation in the
> gcc info pages.  I tried changing several things anyway to make it
> more conservative, but I haven't been able to make it compile
> without optimization.

If I change it to use a static inline function, it seems to work and
will generate identical code (with -O):

static inline int
xchgl(int v, volatile int *m)
{
        int __result;                           

        __asm __volatile ("xchgl %0, %1"        
            : "=r"(__result), "=m"(*m)           
            : "0"(v), "1"(*m));                  

        return __result;
}
void
lock80386_acquire(volatile int *lock)
{
        while (xchgl(1, lock) != 0)
                while (*lock != 0)
                        ;
}

It appears to generate valid code without -O, but I am not 100% sure.  It
is very inefficient without -O.  (Beware the different indirection of "m")

Cheers,
-Peter
--
Peter Wemm - peter@FreeBSD.org; peter@yahoo-inc.com; peter@netplex.com.au
"All of this is for nothing if we don't go to the stars" - JMS/B5



To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message