From owner-freebsd-current  Mon Jun  3 18:13: 5 2002
Delivered-To: freebsd-current@freebsd.org
Received: from mailman.zeta.org.au (mailman.zeta.org.au [203.26.10.16])
	by hub.freebsd.org (Postfix) with ESMTP
	id 1302937B409; Mon,  3 Jun 2002 18:13:01 -0700 (PDT)
Received: from bde.zeta.org.au (bde.zeta.org.au [203.2.228.102])
	by mailman.zeta.org.au (8.9.3/8.8.7) with ESMTP id LAA26854;
	Tue, 4 Jun 2002 11:12:58 +1000
Date: Tue, 4 Jun 2002 11:16:44 +1000 (EST)
From: Bruce Evans <bde@zeta.org.au>
X-X-Sender: bde@gamplex.bde.org
To: current@FreeBSD.ORG
Cc: obrien@FreeBSD.ORG
Subject: Re: memset() broken in gcc-3.1 on i386's
In-Reply-To: <20020604084202.Q939-100000@gamplex.bde.org>
Message-ID: <20020604110707.H1982-100000@gamplex.bde.org>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: owner-freebsd-current@FreeBSD.ORG
Precedence: bulk
List-ID: <freebsd-current.FreeBSD.ORG>
List-Archive: <http://docs.freebsd.org/mail/> (Web Archive)
List-Help: <mailto:majordomo@FreeBSD.ORG?subject=help> (List Instructions)
List-Subscribe: <mailto:majordomo@FreeBSD.ORG?subject=subscribe%20freebsd-current>
List-Unsubscribe: <mailto:majordomo@FreeBSD.ORG?subject=unsubscribe%20freebsd-current>
X-Loop: FreeBSD.ORG

On Tue, 4 Jun 2002, I wrote:

> gcc now generates inline code for memset in some cases.  Broken code.

Actually, it only generates inline code for memset in a few more cases,
and the case of a non-constant length is broken (and some cases of
constant lengths are pessimized (e.g., length 7)).

> ...
> This broke newfs (newfs left some garbage in a bitmap).

Actually, it broke fsck_ffs.

Workaround to avoid the known broken case:

%%%
Index: builtins.c
===================================================================
RCS file: /home/ncvs/src/contrib/gcc/builtins.c,v
retrieving revision 1.1.1.3
diff -u -2 -r1.1.1.3 builtins.c
--- builtins.c	13 May 2002 03:35:47 -0000	1.1.1.3
+++ builtins.c	4 Jun 2002 00:53:22 -0000
@@ -2195,4 +2195,9 @@
       len_rtx = expand_expr (len, NULL_RTX, VOIDmode, 0);

+      /* Give up for non-constant lengths.  They are broken on at least
+	 i386's.  */
+      if (GET_CODE (len_rtx) != CONST_INT)
+	return 0;
+
       dest_mem = get_memory_rtx (dest);
       set_mem_align (dest_mem, dest_align);
%%%

Bruce


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