From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S934351AbZLFWbF (ORCPT ); Sun, 6 Dec 2009 17:31:05 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1757683AbZLFWbE (ORCPT ); Sun, 6 Dec 2009 17:31:04 -0500 Received: from one.firstfloor.org ([213.235.205.2]:47735 "EHLO one.firstfloor.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753683AbZLFWbD (ORCPT ); Sun, 6 Dec 2009 17:31:03 -0500 To: Jiri Kosina Cc: Peter Zijlstra , Justin Piszcz , linux-kernel@vger.kernel.org Subject: Re: What is the performance when using frame pointers in the kernel? From: Andi Kleen References: <1259868529.3977.1300.camel@laptop> Date: Sun, 06 Dec 2009 23:31:03 +0100 In-Reply-To: (Jiri Kosina's message of "Fri, 4 Dec 2009 14:23:55 +0100 (CET)") Message-ID: <874oo391lk.fsf@basil.nowhere.org> User-Agent: Gnus/5.1008 (Gnus v5.10.8) Emacs/22.3 (gnu/linux) MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Jiri Kosina writes: > On Thu, 3 Dec 2009, Peter Zijlstra wrote: > >> > What is the performance hit when the kernel is compiled with frame >> > pointers? >> >> build both kernels and run your favourite workload to find out. > > But generally speaking, frame pointers impose quite some performance > penalty indeed. lmbench syscall microbenchmark can give you some hint. I > expect you'll see approx. 10% performance increase for various syscalls if > you disable them, as that's what we have measured lately. It'll depend on the CPU. On many common cores frame pointer cause a a few cycles stall on each function entry, but not on all. -Andi -- ak@linux.intel.com -- Speaking for myself only.