From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 8869 invoked by alias); 23 Sep 2010 14:34:43 -0000 Received: (qmail 8650 invoked by uid 22791); 23 Sep 2010 14:34:41 -0000 X-SWARE-Spam-Status: No, hits=-1.8 required=5.0 tests=AWL,BAYES_00,DKIM_SIGNED,DKIM_VALID,DKIM_VALID_AU,FREEMAIL_FROM,RCVD_IN_DNSWL_NONE,TW_CF X-Spam-Check-By: sourceware.org Received: from mail-gx0-f175.google.com (HELO mail-gx0-f175.google.com) (209.85.161.175) by sourceware.org (qpsmtpd/0.43rc1) with ESMTP; Thu, 23 Sep 2010 14:34:31 +0000 Received: by gxk10 with SMTP id 10so692041gxk.20 for ; Thu, 23 Sep 2010 07:34:29 -0700 (PDT) MIME-Version: 1.0 Received: by 10.150.57.9 with SMTP id f9mr2871747yba.341.1285252469556; Thu, 23 Sep 2010 07:34:29 -0700 (PDT) Received: by 10.231.17.130 with HTTP; Thu, 23 Sep 2010 07:34:28 -0700 (PDT) In-Reply-To: <4C9B26F7.3080000@codesourcery.com> References: <4BB3CCCA.7000600@codesourcery.com> <201004101235.54302.ebotcazou@adacore.com> <4BC62EB2.40609@codesourcery.com> <201004200005.53901.ebotcazou@adacore.com> <4C460A1F.4000509@codesourcery.com> <4C56EABB.4030401@redhat.com> <4C56EB55.7010408@codesourcery.com> <4C56ECB1.9060302@redhat.com> <4C56EF21.1060801@codesourcery.com> <4C58232D.8090805@codesourcery.com> <4C58329F.6010902@redhat.com> <4C583639.9070704@codesourcery.com> <4C584E0E.2040408@redhat.com> <4C596CA5.2060207@codesourcery.com> <4C7BCCFA.90409@codesourcery.com> <4C97316B.6090902@codesourcery.com> <4C97843D.7070104@redhat.com> <4C9B26F7.3080000@codesourcery.com> Date: Thu, 23 Sep 2010 22:00:00 -0000 Message-ID: Subject: Re: ifcvt/crossjump patch: Fix PR 42496, 21803 From: Richard Guenther To: Bernd Schmidt Cc: Jeff Law , Eric Botcazou , gcc-patches@gcc.gnu.org, Steven Bosscher , Jim Wilson Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable X-IsSubscribed: yes Mailing-List: contact gcc-patches-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Id: List-Archive: List-Post: List-Help: Sender: gcc-patches-owner@gcc.gnu.org X-SW-Source: 2010-09/txt/msg01902.txt.bz2 On Thu, Sep 23, 2010 at 12:07 PM, Bernd Schmidt w= rote: > On 09/20/2010 05:56 PM, Jeff Law wrote: >> =A0On 09/20/10 04:03, Bernd Schmidt wrote: >>> On 08/30/2010 05:23 PM, Bernd Schmidt wrote: >>>> On 08/04/2010 03:35 PM, Bernd Schmidt wrote: >>>>> On 08/03/2010 07:12 PM, Jeff Law wrote: >>>>>> OK. =A0WRT the comment, we might want to just say that BB_MODIFIED is >>>>>> set >>>>>> at the same time as a block is marked dirty, but is not cleared >>>>>> during a >>>>>> df_analyze allowing a pass to update the DF information and still kn= ow >>>>>> what blocks were modified. >>>>> >>>>> New patch below. >>> >>> Ping^2. >>> http://gcc.gnu.org/ml/gcc-patches/2010-08/msg00280.html >> Sorry, I thought this had been approved long ago. =A0Approved :-) > > Thanks. =A0Committed with a small fix found while testing on x86_64: In > try_head_merge_bb, in the case where we try to move across multiple > blocks to optimize for a switch statement, the final destination block > must be the only predecessor of the block we're looking at. I now get an bootstrap fail building Ada on x86_64-linux in Program received signal SIGSEGV, Segmentation fault. 0x0000000000c9ebf6 in df_simulate_one_insn_backwards (bb=3D0x7ffff3f0ad68, insn=3D0x0, live=3D0x50bab30) at /space/rguenther/src/svn/trunk/gcc/df-problems.c:3640 3640 if (!NONDEBUG_INSN_P (insn)) (gdb) up #1 0x0000000000c9f112 in simulate_backwards_to_point (bb=3D0x7ffff3f0ad68, live=3D0x50bab30, point=3D0x7ffff3a461b0) at /space/rguenther/src/svn/trunk/gcc/df-problems.c:3813 3813 df_simulate_one_insn_backwards (bb, insn, live); #2 0x0000000002d2fe2f in try_head_merge_bb (bb=3D0x7ffff3f0ad00) at /space/rguenther/src/svn/trunk/gcc/cfgcleanup.c:2094 #3 0x0000000002d30c2e in try_optimize_cfg (mode=3D2) at /space/rguenther/src/svn/trunk/gcc/cfgcleanup.c:2437 #4 0x0000000002d311a1 in cleanup_cfg (mode=3D2) at /space/rguenther/src/svn/trunk/gcc/cfgcleanup.c:2627 #5 0x0000000002d842ad in rest_of_handle_stack_adjustments () at /space/rguenther/src/svn/trunk/gcc/combine-stack-adj.c:555 and your patch is in range and looks related. Yes I do have another patch applied, but that affects profiling only which is disabled. Checking w/o now. Richard.