From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Luis Claudio R. Goncalves" Subject: Re: 2.6.24-rc8-rt1: Strange latencies on mpc5200 powerpc Date: Thu, 24 Jan 2008 11:49:51 -0200 Message-ID: <20080124134951.GG4776@unix.sh> References: <1200544050.318.18.camel@localhost.localdomain> <478F2A46.5070506@grandegger.com> <20080123145340.GE6200@pengutronix.de> <47976D06.1010900@grandegger.com> <47986E44.3010101@grandegger.com> <20080124112847.GE4776@unix.sh> <47987D73.8090904@grandegger.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: linux-rt-users@vger.kernel.org To: Wolfgang Grandegger Return-path: Received: from marova.hst.terra.com.br ([200.176.10.215]:51753 "EHLO marova.hst.terra.com.br" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751675AbYAXNv7 (ORCPT ); Thu, 24 Jan 2008 08:51:59 -0500 Received: from sr06-04.mta.terra.com.br (sr06-04.mta.terra.com.br [200.154.152.73]) by marova.hst.terra.com.br (Postfix) with ESMTP id 502A83B6A36 for ; Thu, 24 Jan 2008 11:51:57 -0200 (BRST) Content-Disposition: inline In-Reply-To: <47987D73.8090904@grandegger.com> Sender: linux-rt-users-owner@vger.kernel.org List-ID: On Thu, Jan 24, 2008 at 12:58:43PM +0100, Wolfgang Grandegger wrote: | Hi Luis, | | I re-added CC to the mailing list. | | Luis Claudio R. Goncalves wrote: | > On Thu, Jan 24, 2008 at 11:53:56AM +0100, Wolfgang Grandegger wrote: | > | Wolfgang Grandegger wrote: | > ... | > I deleted lots of lines just to make it easy to see my question: | > ... | > | OK, here is the full report. | > ... | > | bash-3.00# ./cyclictest -n -p80 -t1 -i1000 | > | 2.91 4.81 13.72 1/50 23887 | > | | > | T: 0 ( 976) P:80 I:1000 C:1634520 Min: 15 Act: 45 Avg: 68 Max: 138 | > | > ... | > | > | bash-3.00# ./cyclictest -n -p80 -i1000 | > | 52.31 96.08 61.61 2/51 9129 | > | | > | T: 0 ( 976) P:80 I:1000 C: 795180 Min: 14 Act: 75 Avg: 69 Max: 134 | > | > Are the loadavg values right? If so, the results were obtained in very | > different circunstances. In the first case, load was around 2.91... in the | > second one, 52.31. This is also true for the other tests below. | > | > Please compare /proc/loadavg and /proc/loadavgrt (a simple cat do the | > trick). If loadavgrt is reporting bogus values, let us know. It was fixed a | > looong time ago, but I never tested that in a ppc. | | Here is the output: | | bash-3.00# cat /proc/loadavg; cat /proc/loadavgrt | 3.12 3.13 4.57 3/46 5541 | 75.35 73.84 72.94 0/46 5543 | | There is no constant load as I run: | | "while ./hackbench 10; do ./calibrator 400 32M cali; sleep 30; done" | | in the second terminal window. What you see is the load when I | interrupted the test manually. | | Wolfgang. The bogus readings in /proc/loadvgrt were fixed a while ago and the fix is in the 2.6.24-rcX-rtX series. Not that this could be leading you to wrong results, but that's really strange seeing it. OTOH, cyclictest reads the load from /proc/loadavg and system pressure (at least for the last 15 minutes) was really different on the tests, even though results were close. Luis -- [ Luis Claudio R. Goncalves Bass - Gospel - RT ] [ Fingerprint: 4FDD B8C4 3C59 34BD 8BE9 2696 7203 D980 A448 C8F8 ]