Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754855AbcDKMiV (ORCPT ); Mon, 11 Apr 2016 08:38:21 -0400 Received: from mx1.redhat.com ([209.132.183.28]:41884 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753233AbcDKMiT (ORCPT ); Mon, 11 Apr 2016 08:38:19 -0400 Message-ID: <1460378295.25336.5.camel@redhat.com> Subject: Re: [regression] cross core scheduling frequency drop bisected to 0c313cb20732 From: Rik van Riel To: Mike Galbraith , "Rafael J. Wysocki" , Peter Zijlstra Cc: "Rafael J. Wysocki" , "Rafael J. Wysocki" , LKML , Linux PM list , Doug Smythies Date: Mon, 11 Apr 2016 08:38:15 -0400 In-Reply-To: <1460343894.3682.11.camel@gmail.com> References: <1460092854.4051.1.camel@gmail.com> <20160408064510.GK3448@twins.programming.kicks-ass.net> <1460098254.5582.17.camel@gmail.com> <2428384.mEkP3EOpsR@vostro.rjw.lan> <20160409110729.GS3448@twins.programming.kicks-ass.net> <1460302797.4383.44.camel@gmail.com> <1460319890.25336.2.camel@redhat.com> <1460343894.3682.11.camel@gmail.com> Content-Type: multipart/signed; micalg="pgp-sha256"; protocol="application/pgp-signature"; boundary="=-2yhfBHSLyspRvLXXLps0" Mime-Version: 1.0 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 2329 Lines: 65 --=-2yhfBHSLyspRvLXXLps0 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable On Mon, 2016-04-11 at 05:04 +0200, Mike Galbraith wrote: > On Sun, 2016-04-10 at 16:24 -0400, Rik van Riel wrote: > >=20 > > On Sun, 2016-04-10 at 17:39 +0200, Mike Galbraith wrote: > >=20 > > >=20 > > > Should the default idle state not then be governor > > > dependent?=C2=A0=C2=A0When I > > > set gov=3Dperformance, I'm expecting box to go just as fast as it > > > can > > > go > > > without melting.=C2=A0=C2=A0Does polling risk CPU -> lava conversion? > > Current CPUs can only have some cores run at full speed > > (turbo mode) if other cores are idling and/or running at > > lower speeds. > The real world is very unlikely to miss the prettier numbers I'm > grieving over one tiny bit.=C2=A0=C2=A0Knowing that doesn't make giving t= hem up > any easier though.. byebye cycles (sniff) ;-) I suspect your pipe benchmark could be very relevant to network performance numbers, too. I would like to go into polling a little bit more aggressively in a future kernel, and I think we can get away with it if we teach the polling loop to exit after we have spent enough time there that the menu governor will pick HLT after a few timed out poll loops. That way while we run a workload that actually benefits from polling, we will get polling, but once we run a workload that actually sleeps longer than the HLT threshold, we will quickly fall back to HLT. With 10Gbps network traffic, it could make a real difference whether or not the CPU can wake up immediately, or takes a microsecond to wake up... --=20 All Rights Reversed. --=-2yhfBHSLyspRvLXXLps0 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: This is a digitally signed message part Content-Transfer-Encoding: 7bit -----BEGIN PGP SIGNATURE----- Version: GnuPG v2 iQEcBAABCAAGBQJXC5q3AAoJEM553pKExN6DMdUIAJv8ddCouEs6FDFNHR3cxjbK bLmJHupjLFylhSmLIKbIWQkyobgpUje0048kW9LlCQ4S8Mp31x0Z0NJHJ+mvm+go 4p//nFPlgTIjDhG7/u6/Rl0XZ6b+nNEs/U5eUfT2l3CD1/+1nm7kkq8emIpB+rfh A2c+zrlvJDMzdXsr2rpP0Hm46qck/YRZ0h4UxQtNkHbG2u0Vs0Vp1oyxnmgqCnc1 gVQjthfFUCxiAezs5JLPMDsC2pYu9muBWyMYJErLa/Sm33HlA+pQcyeE/se8bqV2 rLq+ligPk7WTSI/Ki43g6X4JY8tv/RO1cPPGDU1S0RlQdF5r8lXKVZdHB3b+GX4= =wuqi -----END PGP SIGNATURE----- --=-2yhfBHSLyspRvLXXLps0--