From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 7414 invoked by alias); 30 Sep 2010 16:44:31 -0000 Received: (qmail 7390 invoked by uid 22791); 30 Sep 2010 16:44:29 -0000 X-SWARE-Spam-Status: No, hits=-2.3 required=5.0 tests=ALL_TRUSTED,AWL,BAYES_00,MISSING_MID X-Spam-Check-By: sourceware.org Received: from localhost (HELO gcc.gnu.org) (127.0.0.1) by sourceware.org (qpsmtpd/0.43rc1) with ESMTP; Thu, 30 Sep 2010 16:44:19 +0000 From: "hubicka at gcc dot gnu.org" To: gcc-bugs@gcc.gnu.org Subject: [Bug tree-optimization/45781] [4.6 Regression] GCC incorrectly puts function in .text.unlikely X-Bugzilla-Reason: CC X-Bugzilla-Type: changed X-Bugzilla-Watch-Reason: None X-Bugzilla-Product: gcc X-Bugzilla-Component: tree-optimization X-Bugzilla-Keywords: build X-Bugzilla-Severity: normal X-Bugzilla-Who: hubicka at gcc dot gnu.org X-Bugzilla-Status: UNCONFIRMED X-Bugzilla-Priority: P3 X-Bugzilla-Assigned-To: unassigned at gcc dot gnu.org X-Bugzilla-Target-Milestone: 4.6.0 X-Bugzilla-Changed-Fields: In-Reply-To: References: X-Bugzilla-URL: http://gcc.gnu.org/bugzilla/ Auto-Submitted: auto-generated Content-Type: text/plain; charset="UTF-8" MIME-Version: 1.0 Date: Thu, 30 Sep 2010 19:24:00 -0000 Mailing-List: contact gcc-bugs-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Id: List-Archive: List-Post: List-Help: Sender: gcc-bugs-owner@gcc.gnu.org X-SW-Source: 2010-09/txt/msg03280.txt.bz2 Message-ID: <20100930192400.6xH42Cgw81OvcCZcloMN5_Yh3yFpJ4Np1YAvvKECFuU@z> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45781 --- Comment #2 from Jan Hubicka 2010-09-30 16:44:17 UTC --- IA-64 seems to be fine with unlikely section at least at our periodic tester setup, otherwise SPEC2000 FDO testing would break. So it might be specific for ia64 HP-UX and in that case indeed correct fix is to simply define cold and hot sections to be .text sections (or fix it at binutils side) for HP-UX IA-64 (and possibly PA target too). What I am confused about is how partial inlining affect placement of init_target_chars. If it was because function got partially inlined, the wrong call would be named init_target_chars.part.XXX and it is not. It is possible that partial inlining affect profile in some of the callers and then it might be some bug in profile updating, so I would like to see a testcase. x.c in the PR is truncated and when i compile builtins.c from my GCC tree (x86-64) I do not get init_target_chars in unlikely function. For some funny reason I however get gimple_rewrite_call_expr. All uses of it are preceeded by very many exists from the function that makes them appear cold. Funny but probably not harmful. Honza