From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754460AbaFCNmA (ORCPT ); Tue, 3 Jun 2014 09:42:00 -0400 Received: from mx1.redhat.com ([209.132.183.28]:39509 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751796AbaFCNl7 (ORCPT ); Tue, 3 Jun 2014 09:41:59 -0400 Date: Tue, 3 Jun 2014 15:41:45 +0200 From: Jiri Olsa To: David Ahern Cc: Jiri Olsa , linux-kernel@vger.kernel.org, Arnaldo Carvalho de Melo , Corey Ashford , Frederic Weisbecker , Ingo Molnar , Jean Pihet , Namhyung Kim , Paul Mackerras , Peter Zijlstra Subject: Re: [PATCH 06/13] perf tools: Cache dso data file descriptor Message-ID: <20140603134145.GD2864@krava.brq.redhat.com> References: <1401743927-398-1-git-send-email-jolsa@kernel.org> <1401743927-398-7-git-send-email-jolsa@kernel.org> <538DCF59.6000808@gmail.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <538DCF59.6000808@gmail.com> User-Agent: Mutt/1.5.21 (2010-09-15) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Tue, Jun 03, 2014 at 07:36:25AM -0600, David Ahern wrote: > On 6/2/14, 3:18 PM, Jiri Olsa wrote: > >We keep dsos data file descriptors open until their count reaches > >the half of the current fd open limit (RLIMIT_NOFILE). In this case > >we close file descriptor of the first opened dso object. > > > >We've got overall speedup (24% for my workload) of report: > > 'perf report --stdio -i perf-test.data' (3 runs) > > (perf-test.data size was around 12GB) > > > > current code: > > 446,826,474,217 cycles ( +- 0.23% ) > > 669,435,100,846 instructions ( +- 0.03% ) > > > > 266.380674551 seconds time elapsed ( +- 0.37% ) > > > > change: > > 558,345,509,819 cycles ( +- 0.55% ) > > 817,283,647,637 instructions ( +- 0.03% ) > > > > 351.564209617 seconds time elapsed ( +- 0.89% ) > > message is confusing. Looks like time went up. aaargh.. right ;-) it should bethe other way around thanks, jirka