From owner-freebsd-current@FreeBSD.ORG Sun Jun 5 08:49:56 2005 Return-Path: X-Original-To: freebsd-current@freebsd.org Delivered-To: freebsd-current@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 3719316A41C for ; Sun, 5 Jun 2005 08:49:56 +0000 (GMT) (envelope-from nakal@nurfuerspam.de) Received: from mail.gmx.net (pop.gmx.net [213.165.64.20]) by mx1.FreeBSD.org (Postfix) with SMTP id 6D5CF43D49 for ; Sun, 5 Jun 2005 08:49:55 +0000 (GMT) (envelope-from nakal@nurfuerspam.de) Received: (qmail invoked by alias); 05 Jun 2005 08:49:53 -0000 Received: from p5090ECBB.dip.t-dialin.net (EHLO klotz.local) [80.144.236.187] by mail.gmx.net (mp023) with SMTP; 05 Jun 2005 10:49:53 +0200 X-Authenticated: #989277 Received: from [127.0.0.1] (localhost [127.0.0.1]) by klotz.local (8.13.3/8.13.3) with ESMTP id j558nBPF001374; Sun, 5 Jun 2005 10:49:16 +0200 (CEST) (envelope-from nakal@nurfuerspam.de) Message-ID: <42A2BC87.5040103@nurfuerspam.de> Date: Sun, 05 Jun 2005 10:49:11 +0200 From: Martin User-Agent: Mozilla Thunderbird 1.0.2 (X11/20050403) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Matt Emmerton References: <42A242D1.4000002@nurfuerspam.de> <000701c5696f$0dee3b10$1200a8c0@gsicomp.on.ca> In-Reply-To: <000701c5696f$0dee3b10$1200a8c0@gsicomp.on.ca> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Y-GMX-Trusted: 0 Cc: freebsd-current@freebsd.org Subject: Re: Optimizing libc/string for amd64(?) X-BeenThere: freebsd-current@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Discussions about the use of FreeBSD-current List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sun, 05 Jun 2005 08:49:56 -0000 Matt Emmerton wrote: >>Wouldn't it be better to set it to "typedef long word"? > It would probably be better to use int64_t and int32_t instead of long/int, > in order to make the intention more obvious. The decision up to the developers. I personally would still use "long", because it is auto-adjusting and has always the size of the word on the concerning architecture, AFAIK. I usually try to avoid ifdefs, whereever it is possible. But there might be already some other places in the whole codebase where the type for word has been already computed. It would be even better to use that, because then it is done just in one place. But before thinking about a change, it would be better if someone confirms the degraded performance (just run nbench). I've only checked it on 1 machine and cannot access it very often (it does not belong to me). Martin