From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S261423AbVGWSGm (ORCPT ); Sat, 23 Jul 2005 14:06:42 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S262385AbVGWSGl (ORCPT ); Sat, 23 Jul 2005 14:06:41 -0400 Received: from smtp.osdl.org ([65.172.181.4]:9365 "EHLO smtp.osdl.org") by vger.kernel.org with ESMTP id S261423AbVGWSFl (ORCPT ); Sat, 23 Jul 2005 14:05:41 -0400 Date: Sat, 23 Jul 2005 11:02:16 -0700 (PDT) From: Linus Torvalds To: Arjan van de Ven cc: Chuck Ebbert <76306.1226@compuserve.com>, Andrew Morton , linux-kernel@vger.kernel.org, Adrian Bunk Subject: Re: [patch 2.6.13-rc3a] i386: inline restore_fpu In-Reply-To: <1122140791.3582.25.camel@localhost.localdomain> Message-ID: References: <200507230313_MC3-1-A554-6927@compuserve.com> <1122140791.3582.25.camel@localhost.localdomain> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Sat, 23 Jul 2005, Arjan van de Ven wrote: > > For the kernel the runtime measurement is obviously tricky (kgprof > anyone?) Ehh, doesn't "opgprof" do that already? Anyway, judging by real life, people just don't _do_ profile runs. Any build automation that depends on the user profiling the result is a total wet dream by compiler developers - it just doesn't happen in real life. So: > but the static analysis method really shouldn't be too hard. I really think that the static analysis is the only one that actually would matter in real life. It's also the only one that is repeatable, which is a big big bonus, since at least that way different people are running basically the same code (heisenbugs, anyone?) and benchmarks are actually meaningful, since they don't depend on whatever load was picked to generate the layout. So dynamic analysis with dynamic profile data is just one big theoretical compiler-writer masturbation session, in my not-so-humble opinion. I bet static analysis (with possibly some hints from the programmer, the same way we use likely/unlikely) gets you 90% of the way, without all the downsides. Linus