From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1759297Ab1JFVkh (ORCPT ); Thu, 6 Oct 2011 17:40:37 -0400 Received: from claw.goop.org ([74.207.240.146]:34491 "EHLO claw.goop.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757215Ab1JFVkg (ORCPT ); Thu, 6 Oct 2011 17:40:36 -0400 Message-ID: <4E8E2025.30004@goop.org> Date: Thu, 06 Oct 2011 14:39:49 -0700 From: Jeremy Fitzhardinge User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:7.0.1) Gecko/20110930 Thunderbird/7.0.1 MIME-Version: 1.0 To: "H. Peter Anvin" CC: Jason Baron , Steven Rostedt , "David S. Miller" , David Daney , Michael Ellerman , Jan Glauber , the arch/x86 maintainers , Xen Devel , Linux Kernel Mailing List , Jeremy Fitzhardinge , peterz@infradead.org, rth@redhat.com Subject: Re: [PATCH RFC V2 3/5] jump_label: if a key has already been initialized, don't nop it out References: <477dead9647029012f93c651f2892ed0e86b89e7.1317506051.git.jeremy.fitzhardinge@citrix.com> <20111003150205.GB2462@redhat.com> <4E89E28C.7010700@goop.org> <20111004141011.GA2520@redhat.com> <4E8B3489.60902@zytor.com> <4E8CF348.4080405@goop.org> <4E8CF385.2080804@zytor.com> <4E8DEB19.1050509@goop.org> <20111006181055.GA2505@redhat.com> <4E8DEFDE.4030706@zytor.com> In-Reply-To: <4E8DEFDE.4030706@zytor.com> X-Enigmail-Version: 1.3.2 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 10/06/2011 11:13 AM, H. Peter Anvin wrote: > On 10/06/2011 11:10 AM, Jason Baron wrote: >> It would be cool if we could make the total width 2-bytes, when >> possible. It might be possible by making the initial 'JUMP_LABEL_INITIAL_NOP' >> as a 'jmp' to the 'l_yes' label. And then patching that with a no-op at boot >> time or link time - letting the compiler pick the width. In that way we could >> get the optimal width... >> > Yes, that would be a win just based on icache footprint alone. I'm not sure it would be a win, necessarily. My test with back-to-back jmp2 was definitely slower than with the nop padding it out to 5 bytes; I suspect that's a result of having too many jmps within one cacheline. Of course, there's no reason why the CPU would optimise for jumps to jumps, so perhaps its just hitting a "stupid programmer" path. J