From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755275Ab1KUQFo (ORCPT ); Mon, 21 Nov 2011 11:05:44 -0500 Received: from mx1.redhat.com ([209.132.183.28]:33612 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751163Ab1KUQFo (ORCPT ); Mon, 21 Nov 2011 11:05:44 -0500 Message-ID: <4ECA76C9.2080209@redhat.com> Date: Mon, 21 Nov 2011 11:05:29 -0500 From: William Cohen User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.2.24) Gecko/20111104 Red Hat/3.1.16-2.el6_1 Thunderbird/3.1.16 MIME-Version: 1.0 To: Vince Weaver CC: Peter Zijlstra , mingo@elte.hu, Stephane Eranian , Arun Sharma , linux-kernel@vger.kernel.org Subject: Re: [RFC][PATCH 0/6] perf: x86 RDPMC and RDTSC support References: <20111121145114.049265181@chello.nl> In-Reply-To: Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 11/21/2011 10:02 AM, Vince Weaver wrote: > On Mon, 21 Nov 2011, Peter Zijlstra wrote: > >> These few patches implement x86 RDPMC support and add an extention to the self >> monitoring data to also allow additional time updates using userspace TSC reads. >> >> There's a few loose ends, but it mostly seems to work. > > I'll have to test these out. > > I have some low-level benchmarks I've been working on, you can see some > preliminary results here: > http://web.eecs.utk.edu/~vweaver1/projects/perf-events/benchmarks/ > > perf_events was lagging behind perfmon2 and perfctr in self-monitoring > overhead, but maybe these changes will help that. > > perf_event overhead got a lot worse between 2.6.32 and 3.0, I'm mid-bisect > on that trying to isolate the cause. > > Vince > Hi Vince, Any improvements in performance of self-monitoring would be welcomed. International Symposium on computer Architecture (ISCA 2011) has paper about a low overhead mechanism to read out the performance counters, LiMiT (http://castl.cs.columbia.edu/limit/index.php/LiMiT) that avoids using system calls to read the performance counter data. That paper shows a fair amount of overhead for PAPI. -Will