From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752057AbbBJH0q (ORCPT ); Tue, 10 Feb 2015 02:26:46 -0500 Received: from metis.ext.pengutronix.de ([92.198.50.35]:58482 "EHLO metis.ext.pengutronix.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751211AbbBJH0o (ORCPT ); Tue, 10 Feb 2015 02:26:44 -0500 Date: Tue, 10 Feb 2015 08:26:34 +0100 From: Markus Pargmann To: Peter Zijlstra Cc: Huang Ying , Ingo Molnar , LKML , LKP ML Subject: Re: [LKP] [sched/core] 9edfbfed3f5: +88.2% hackbench.time.involuntary_context_switches Message-ID: <20150210072634.GZ16879@pengutronix.de> References: <1423461513.5968.56.camel@intel.com> <20150209083120.GL5029@twins.programming.kicks-ass.net> MIME-Version: 1.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="HjGz34yYE1BNnn/x" Content-Disposition: inline In-Reply-To: <20150209083120.GL5029@twins.programming.kicks-ass.net> X-Sent-From: Pengutronix Hildesheim X-URL: http://www.pengutronix.de/ X-IRC: #ptxdist @freenode X-Accept-Language: de,en X-Accept-Content-Type: text/plain X-Uptime: 08:14:03 up 19 days, 21:21, 56 users, load average: 0.07, 0.31, 0.37 User-Agent: Mutt/1.5.23 (2014-03-12) X-SA-Exim-Connect-IP: 2001:67c:670:100:1d::7 X-SA-Exim-Mail-From: mpa@pengutronix.de X-SA-Exim-Scanned: No (on metis.ext.pengutronix.de); SAEximRunCond expanded to false X-PTX-Original-Recipient: linux-kernel@vger.kernel.org Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org --HjGz34yYE1BNnn/x Content-Type: text/plain; charset=utf-8 Content-Disposition: inline Content-Transfer-Encoding: quoted-printable Hi, On Mon, Feb 09, 2015 at 09:31:20AM +0100, Peter Zijlstra wrote: > On Mon, Feb 09, 2015 at 01:58:33PM +0800, Huang Ying wrote: > > FYI, we noticed the below changes on > >=20 > > git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git sched/core > > commit 9edfbfed3f544a7830d99b341f0c175995a02950 ("sched/core: Rework rq= ->clock update skips") > >=20 > >=20 > > testbox/testcase/testparams: xps2/hackbench/performance-1600%-process-s= ocket > >=20 > > cebde6d681aa45f9 9edfbfed3f544a7830d99b341f =20 > > ---------------- -------------------------- =20 > > %stddev %change %stddev > > \ | \ =20 > > 1839273 =C2=B1 6% +88.2% 3462337 =C2=B1 4% hackbench.time.= involuntary_context_switches > > 41965851 =C2=B1 5% +5.6% 44307403 =C2=B1 1% hackbench.time.= voluntary_context_switches > > 388 =C2=B1 39% -58.6% 160 =C2=B1 10% sched_debug.cfs= _rq[1]:/.tg_load_contrib > > 12957 =C2=B1 14% -60.5% 5117 =C2=B1 11% sched_debug.cfs= _rq[2]:/.tg_load_avg > > 30505 =C2=B1 14% -57.7% 12905 =C2=B1 6% sched_debug.cfs= _rq[3]:/.tg_load_avg > > 2790 =C2=B1 24% -65.4% 964 =C2=B1 32% sched_debug.cfs= _rq[3]:/.blocked_load_avg > > 2915 =C2=B1 23% -62.2% 1101 =C2=B1 29% sched_debug.cfs= _rq[3]:/.tg_load_contrib > > 1839273 =C2=B1 6% +88.2% 3462337 =C2=B1 4% time.involuntar= y_context_switches > > 1474 =C2=B1 28% -61.7% 565 =C2=B1 43% sched_debug.cfs= _rq[2]:/.tg_load_contrib > > 11830 =C2=B1 15% +63.0% 19285 =C2=B1 11% sched_debug.cpu= #4.sched_goidle > > 19319 =C2=B1 29% +91.1% 36913 =C2=B1 7% sched_debug.cpu= #3.sched_goidle > > 5899 =C2=B1 31% -35.6% 3801 =C2=B1 11% sched_debug.cfs= _rq[4]:/.blocked_load_avg > > 5999 =C2=B1 30% -34.5% 3929 =C2=B1 11% sched_debug.cfs= _rq[4]:/.tg_load_contrib > > 37884 =C2=B1 13% -33.5% 25207 =C2=B1 7% sched_debug.cfs= _rq[4]:/.tg_load_avg > > 229547 =C2=B1 5% +47.9% 339519 =C2=B1 5% cpuidle.C1-NHM.= usage > > 35712 =C2=B1 3% +31.7% 47036 =C2=B1 9% cpuidle.C3-NHM.= usage > > 5010 =C2=B1 9% -29.0% 3556 =C2=B1 20% sched_debug.cfs= _rq[6]:/.blocked_load_avg > > 5139 =C2=B1 9% -28.2% 3690 =C2=B1 19% sched_debug.cfs= _rq[6]:/.tg_load_contrib > > 49568 =C2=B1 6% +24.8% 61867 =C2=B1 7% sched_debug.cpu= #1.sched_goidle > > 26369 =C2=B1 35% -42.0% 15289 =C2=B1 29% cpuidle.C6-NHM.= usage > > 18 =C2=B1 16% +36.5% 25 =C2=B1 7% sched_debug.cpu= #4.nr_running > > 1.41 =C2=B1 12% -19.3% 1.14 =C2=B1 13% perf-profile.cp= u-cycles.sock_wfree.unix_destruct_scm.skb_release_head_state.skb_release_al= l.consume_skb > > 25 =C2=B1 15% +28.7% 32 =C2=B1 9% sched_debug.cpu= #3.nr_running > > 1.63 =C2=B1 11% -18.0% 1.34 =C2=B1 12% perf-profile.cp= u-cycles.unix_destruct_scm.skb_release_head_state.skb_release_all.consume_s= kb.unix_stream_recvmsg > > 0.57 =C2=B1 8% +9.6% 0.62 =C2=B1 5% turbostat.CPU%c1 > > 148 =C2=B1 11% -16.7% 123 =C2=B1 7% sched_debug.cfs= _rq[1]:/.load > > 109 =C2=B1 6% +17.1% 128 =C2=B1 6% sched_debug.cpu= #6.cpu_load[0] > > 2.41 =C2=B1 8% -13.3% 2.09 =C2=B1 11% perf-profile.cp= u-cycles.skb_release_head_state.skb_release_all.consume_skb.unix_stream_rec= vmsg.sock_aio_read > > 147 =C2=B1 12% -16.4% 123 =C2=B1 7% sched_debug.cpu= #1.load > > 111 =C2=B1 5% +15.4% 129 =C2=B1 5% sched_debug.cpu= #6.cpu_load[2] > > 110 =C2=B1 5% +14.9% 127 =C2=B1 5% sched_debug.cfs= _rq[6]:/.runnable_load_avg > > 112 =C2=B1 5% +14.5% 128 =C2=B1 4% sched_debug.cpu= #6.cpu_load[3] > > 113 =C2=B1 5% +13.2% 128 =C2=B1 3% sched_debug.cpu= #6.cpu_load[4] > > 789953 =C2=B1 2% -10.8% 704528 =C2=B1 4% sched_debug.cpu= #3.avg_idle > > 15471 =C2=B1 5% -7.7% 14278 =C2=B1 2% sched_debug.cpu= #5.curr->pid > > 2675106 =C2=B1 10% +16.2% 3109411 =C2=B1 1% sched_debug.cpu= #4.nr_switches > > 2675140 =C2=B1 10% +16.2% 3109440 =C2=B1 1% sched_debug.cpu= #4.sched_count > > 155201 =C2=B1 5% +14.6% 177901 =C2=B1 3% softirqs.RCU > > 8.64 =C2=B1 6% -9.6% 7.82 =C2=B1 5% perf-profile.cp= u-cycles.skb_release_all.consume_skb.unix_stream_recvmsg.sock_aio_read.sock= _aio_read > > 2658351 =C2=B1 11% +13.7% 3021564 =C2=B1 2% sched_debug.cpu= #5.sched_count > > 2658326 =C2=B1 11% +13.7% 3021539 =C2=B1 2% sched_debug.cpu= #5.nr_switches > > 71443 =C2=B1 5% +9.9% 78486 =C2=B1 0% vmstat.system.cs > > 8209 =C2=B1 5% +7.3% 8805 =C2=B1 0% vmstat.system.in > >=20 >=20 > OK, so the interesting number is total runtime; I cannot find it. > Therefore I cannot say what if anything changed. This is just a bunch of > random numbers afaict. The total runtime of hackbench in v3.19 compared to v3.15 to v3.18 is shown here [1] (Group 5 -> linux_perf.hackbench). It is not the same project so the profiling data is not related and not recorded. The results should be reproducable by simply running hackbench with the same options as cbenchsuite [2] did. The options are always listed in the result browser [1] below the plots. Best Regards, Markus [1] http://results.cbenchsuite.org/plots/2015-02-09__v3.15-v3.19-quad/detai= led/ [2] http://cbenchsuite.org --HjGz34yYE1BNnn/x Content-Type: application/pgp-signature; name="signature.asc" Content-Description: Digital signature -----BEGIN PGP SIGNATURE----- Version: GnuPG v1 iQIcBAEBAgAGBQJU2bKqAAoJEEpcgKtcEGQQMmIP/1FPNcevPTb6m2V7OVkcRIAA hHsOTa00UWxJAE3S+lrqBYJKo5xjY/7YLCPZDx2qbFotfmujoMbgIxFZx/pTVeoe ekKYFp2+SuQJyf23XiHnO8naYpf1AtR184kDdOwgYcRG1fVcgfBSSZE/z+LhBUKZ 8FgVF4sE4aZxs7II8INevJbZ65EMAB7pz1Tb0Vnx+oxWsjZFF58sX0uNVEYN7+A/ 5r5zzqVALwx4tuqECxFcA1OdDFAs7K6gcKwzQjMAfh8sLTyhIym8qr81T8cLh4RO Mh0sH76TnBEDvU8Ij2tmMl+Bsz6fnQ+CFxbPpJA3xmrAuXQibZpAZwDWG3FsJylu 8TcQyYoZIJI0G6TdFQvYmaHZlM4r1tlu2Ld37WM8uvHsL+zqyK6ty51ZGGINLpvM VE2IUE43RH+FmjwRxUQQSF9opNBxhEP7D739eq0fNvvs21ApoL6gHn0BP4F6DntW lWFLOcHpYhmNZiMmGS7oLeag0u0nBNmzXRtMhvQYDtCpg85oXLPckLmdrWPzQ8yi PlwdACulnMaqar+8KSLbssbrVXTBbHVv31Nojt8UNTLRpiZusXxjhx4+fyeURWKn y7p6Qu8Sibos/irHOXxnYwqLlO2N9Gd66kTQbP0rQnop4o18Eh9Oj6SM0TzvHRb9 YhsQ635sm0zNnFWMRhw1 =AKYk -----END PGP SIGNATURE----- --HjGz34yYE1BNnn/x--