From owner-svn-src-head@FreeBSD.ORG Wed Nov 18 03:58:27 2009 Return-Path: Delivered-To: svn-src-head@FreeBSD.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 2885710656DA; Wed, 18 Nov 2009 03:58:27 +0000 (UTC) (envelope-from imp@bsdimp.com) Received: from harmony.bsdimp.com (bsdimp.com [199.45.160.85]) by mx1.freebsd.org (Postfix) with ESMTP id D99BF8FC1B; Wed, 18 Nov 2009 03:58:26 +0000 (UTC) Received: from localhost (localhost [127.0.0.1]) by harmony.bsdimp.com (8.14.3/8.14.1) with ESMTP id nAI3rSaX049243; Tue, 17 Nov 2009 20:53:28 -0700 (MST) (envelope-from imp@bsdimp.com) Date: Tue, 17 Nov 2009 20:53:50 -0700 (MST) Message-Id: <20091117.205350.1682637096.imp@bsdimp.com> To: jilles@stack.nl From: "M. Warner Losh" In-Reply-To: <20091117182501.GA70742@stack.nl> References: <4B01E548.7040708@gmail.com> <20091117182501.GA70742@stack.nl> X-Mailer: Mew version 5.2 on Emacs 21.3 / Mule 5.0 (SAKAKI) Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Cc: svn-src-head@FreeBSD.org, lynx.ripe@gmail.com, svn-src-all@FreeBSD.org, src-committers@FreeBSD.org, edwin@FreeBSD.org Subject: Re: svn commit: r194783 - head/lib/libc/stdtime X-BeenThere: svn-src-head@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: SVN commit messages for the src tree for head/-current List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 18 Nov 2009 03:58:27 -0000 In message: <20091117182501.GA70742@stack.nl> Jilles Tjoelker writes: : On Tue, Nov 17, 2009 at 01:50:32AM +0200, Dmitry Pryanishnikov wrote: : > > Author: edwin : > > Date: Tue Jun 23 22:28:44 2009 : > > New Revision: 194783 : > > URL: http://svn.freebsd.org/changeset/base/194783 : : > > Log: : > > Remove duplicate if-statement on gmt_is_set in gmtsub(). : : > > MFC after: 1 week : : > > Modified: : > > head/lib/libc/stdtime/localtime.c : : > This change looks like a (small?) pessimization to me: before it, : > _MUTEX_LOCK/_MUTEX_UNLOCK pair would be skipped for the case gmt_is_set : > == TRUE (all invocations except the first one), now it won't. I'm not : > sure whether this is critical here though... : : It is certainly less efficient, but the old code was (most likely) : wrong. It used an idiom known as "double checked locking", which is : incorrect in most memory models. The problem is that the store to : gmt_is_set may become visible without stores to other memory (gmtptr and : what it points to) becoming visible. Wouldn't a memory barrier solve that at a much lower cost than a full lock? Warner