From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1030417AbXDNNbJ (ORCPT ); Sat, 14 Apr 2007 09:31:09 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1161196AbXDNNbJ (ORCPT ); Sat, 14 Apr 2007 09:31:09 -0400 Received: from 65-117-135-114.dia.static.qwest.net ([65.117.135.114]:46218 "EHLO postfix.timesys.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1030403AbXDNNbI (ORCPT ); Sat, 14 Apr 2007 09:31:08 -0400 X-Greylist: delayed 1471 seconds by postgrey-1.27 at vger.kernel.org; Sat, 14 Apr 2007 09:31:07 EDT Subject: Re: [RFC] [Patch 0/4] timer_stats slimmed down From: Thomas Gleixner Reply-To: tglx@timesys.com To: Martin Peschke Cc: linux-kernel@vger.kernel.org, Andrew Morton , linux-s390@vger.kernel.org, d.pensator@gmail.com, mingo@redhat.com In-Reply-To: <1176545387.5462.108.camel@dix> References: <1176545387.5462.108.camel@dix> Content-Type: text/plain Date: Sat, 14 Apr 2007 15:07:27 +0200 Message-Id: <1176556047.3387.64.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Evolution 2.6.1 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Martin, On Sat, 2007-04-14 at 12:09 +0200, Martin Peschke wrote: > 2.6.21-rc* kernels come with a feature that adds timer statistics > to /proc/timer_stats (CONFIG_TIMER_STATS). This patch set makes it use > the proposed statistics infrastructure (-mm tree), and adds prequisite > functions to the statistics code. > > The result are noticeable lines of code savings. > kernel/time/timer_stats.c which comes with several functions which can > be made redundant by using corresponding functions of lib/statistic.c, > that is: > > - accounting > - printing output in sequential files > - user interface for switching accounting on/off, reset etc. > - measuring the accounting time > - per-cpu optimisations > > Function that were missing in lib/statistics, and which are added by > this patch set, are: > > - allow arbitrary keys for statistics entries (e.g. structures) > - allow client to label output of statistics entries (e.g. using keys) > > A secondary effect of the patch set is that we won't add another entry > to good old /proc. The statistics infrastructure uses debugfs. > > This patch set is a prototype that works for me, but which might allow > for optimisations. Suggestions are welcome. I'm a bit short of time to take a closer look, but on the first glance it looks good to me. I give it a whirl, when I'm back from travelling. tglx