From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 9462 invoked by alias); 5 Feb 2004 06:25:20 -0000 Mailing-List: contact gcc-bugs-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Archive: List-Post: List-Help: Sender: gcc-bugs-owner@gcc.gnu.org Received: (qmail 9451 invoked by alias); 5 Feb 2004 06:25:19 -0000 Date: Thu, 05 Feb 2004 06:25:00 -0000 Message-ID: <20040205062519.9450.qmail@sources.redhat.com> From: "cgd at broadcom dot com" To: gcc-bugs@gcc.gnu.org In-Reply-To: <20020207120601.5625.emaste@sandvine.com> References: <20020207120601.5625.emaste@sandvine.com> Reply-To: gcc-bugzilla@gcc.gnu.org Subject: [Bug libstdc++/5625] [mips] exception unwinding creates invalid pointer on mips X-Bugzilla-Reason: CC X-SW-Source: 2004-02/txt/msg00593.txt.bz2 List-Id: ------- Additional Comments From cgd at broadcom dot com 2004-02-05 06:25 ------- Subject: Re: [mips] exception unwinding creates invalid pointer on mips At 5 Feb 2004 02:45:07 -0000, wilson at gcc dot gnu dot org wrote: > This patch changes the MIPS backend to use DImode for Pmode when generating > 64-bit ILP32 code, instead of SImode as it currently does. I'm a bit concerned about this comment, because SImode values in registers on MIPS should be in their properly sign-extended forms. If they aren't, patterns like addsi3 (-> addu) will not work properly under all conditions. (addu and other 32-bit arithmetic ops are UNPREDICTABLE if operands don't have bits 63:32 matching bit 31, i.e., if they're not sign extended. mipsisa64 sim should be reporting that, as I believe you've noticed.) Sorry if i manage to help get bug closed prematurely. That's what happens when a PHB tries to do real work. 8-) chris -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=5625